Atomistic Simulations for Nuclear Materials Science and Engineering: Recent Advances Enabled by Machine Learning

with Takuji Oda, Professor
Department of Nuclear Engineering at Seoul National University, South Korea

Friday, September 11, 10:10 am
440 Goodwin Hall, Blacksburg

Atomistic simulations, which describe the interactions and motions of individual atoms, provide a fundamental basis for multiscale and multiphysics modeling in nuclear materials science and engineering. Such modeling is indispensable because atomic-scale processes can influence microstructural evolution, material properties, and ultimately the performance and safety of nuclear-energy systems. Although experiments are essential, their scope is often constrained by the availability and capability of facilities and equipment. In particular, for advanced reactor concepts that have not yet been deployed, reliable prediction of materials behavior under actual operating conditions cannot rely on experiments alone and requires modeling grounded in a fundamental understanding of the relevant phenomena.
Traditionally, atomistic simulations have often been used to interpret experimental observations and identify their underlying atomic-scale mechanisms. Recent applications of machine-learning methodologies, however, have substantially improved their accuracy and practical applicability. These advances are increasingly enabling computational studies to generate quantitative data for materials, conditions, and phenomena that are difficult to investigate experimentally, and to provide insights that can guide materials development. In this talk, I will present representative examples of machine-learning-based atomistic simulations of liquid metals relevant to nuclear engineering. The examples will address impurity behavior, chemical reactivity, and corrosion phenomena, illustrating how such simulations can enable “virtual experiments” and support the development and assessment of advanced nuclear-energy systems.

Takuji Oda is a Professor in the Department of Nuclear Engineering at Seoul National University, South Korea. He received his Ph.D. in Nuclear Engineering from the University of Tokyo in 2007. Before joining Seoul National University in 2013, he worked at the University of Tokyo and the University of Tennessee, Knoxville. His research focuses on computational materials science and engineering for nuclear-energy systems, particularly atomistic simulations of radiation damage and hydrogen-isotope behavior in solid metals and ceramics, as well as impurity behavior, corrosion, and materials chemistry in liquid metals and molten salts. In recent years, his group has extensively applied machine-learning methodologies, including machine-learning interatomic potentials and data-driven models, to improve the accuracy and applicability of nuclear-materials simulations. In 2026, he is on sabbatical leave, visiting Tohoku University in Japan and Virginia Tech.