Applications of Artificial Intelligence in Interrogating Complex Materials Structure-Property Relationships
Abstract: In most advanced energy applications such as next-generation fission, fusion, advanced batteries, or thermal storage systems, performance is limited by the ability for materials to perform under extreme environments (corrosive, high temperature, under irradiation). Monitoring, testing, and designing new materials in these environments poses significant practical challenges, which are exacerbated by the large possible space of compositions, thermodynamic conditions, and non-equilibrium phenomena that are possible. Modern techniques for understanding materials structure-property such as predictive simulation or in situ structural spectroscopies are limited by efficiency or information that can retrieved under such challenging conditions. This presentation covers latest advances in applying artificial intelligence to overcome these challenges, particularly in three areas: 1) accelerating predictive simulations using deep learning surrogate models, 2) learning implicit property relationships that would otherwise be difficult to uncover with human intuition alone, and 3) enhancing spectroscopic techniques with computer vision-based learning. Examples are given showing how these methods can be applied to understanding local structure, thermophysical properties, and thermodynamic properties in different contexts including high temperature ionic liquids, battery electrolytes, reactive surfaces, and more.
Bio: Dr. Lam is an Assistant Professor of Nuclear and Chemical Engineering at UMass Lowell. Prior to UMass Lowell, Dr. Lam received his MS (2017) and PhD (2020) in Nuclear Science and Engineering at MIT, and was a Computational Material Scientist at Robert Bosch LLC. His research interests include artificial intelligence-driven development of materials for advanced energy systems, machine learning-assisted simulation and spectroscopy, and non-conventional applications of nuclear.
