Understanding High-Pressure Material Synthesis Through Machine-Learning Accelerated Simulation

Rebecca K. Lindsey

Department of Chemical Engineering, University of Michigan, Ann Arbor

Abstract: Extreme events including cometary impact, volcanic eruptions, and explosions have captivated scientists for millennia, largely due to the exotic conditions they impart. However, information on atomic-scale phenomena underlying these events remains elusive. Beyond implications for understanding natural systems, these missing insights have direct relevance to synthesis science and could enable finer control in e.g., laser, electron- and ion-beam, sonication, and shock methods toward bespoke material structure and property modulation. This knowledge gap stems from an inability to experimentally derive detailed structural and chemical information on the short (sub us) characteristic timescales. Atomistic simulations can, in principle provide these missing insights, but efforts have been largely confined to prohibitively expensive quantum-based (QM) simulation approaches due to complicated underlying chemical physics that can entail conformationally complex species undergoing complicated condensed-phase covalent chemistry, phase separation/transformation, and band-gap closure.

Machine learning provides a disruptive means of overcoming this capability gap, enabling atomistically-resolved “quantum-accurate” simulations on unprecedented scales, but many grand challenges exist surrounding model training set generation ,reproducibility, uncertainty quantification, interpretability, and the overall computational expense associated with model development. To meet this challenge, we developed ChIMES, a unique, physics-informed machine-learned interatomic model and semi-autonomous fitting framework. In this seminar, an overview of the ChIMES approach and recent methodological developments will be provided. Sample applications toward elucidating chemistry under extreme conditions, including toward nanocarbon synthesis via shock-compression will also be presented.

Bio: Dr. Lindsey is an Assistant Professor of Chemical Engineering and by courtesy, of Applied Physics, Materials Science& Engineering, and Nuclear Engineering & Radiological Sciences at the University of Michigan. She is a member of the KIM Review Editorial Board and the Institute of Computational Science & Engineering Advisory Board, and is an Executive Committee Member of the American Physical Society Topical Group on Shock Compression of Condensed Matter. Dr. Lindsey received her B.S. in Chemical Engineering from Wayne State University and her M.S. and Ph.D. in Chemical Physics from the University of Minnesota, Twin Cities. She went on to become a Postdoctoral Research Fellow at Lawrence Livermore National Laboratory (LLNL), where she worked as a Staff Research Scientist for several years, leading research teams within the Energetic Materials Center. Dr. Lindsey has worked in computational chemistry throughout her career, with applications spanning, e.g., sorption in soft materials, possible mechanisms for the origins of life, and detonation synthesis of unusual carbon nanoparticles. Throughout, her work has been underpinned by a strong interest in developing tools enabling work in previously inaccessible problem spaces. Her group also leverages data science, machine learning, and computer vision to aid in interpretation of large experimental datasets and develop material and device performance models from them. Her efforts were recently recognized through the LLNL Physical and Life Sciences Directorate Research Award and the Young Investigator Award from the American Institute of Chemical Engineering’s Computational Molecular Science and Engineering Forum.