2025 AIChE Annual Meeting

(389aw) Li Ion Diffusion in Solid Electrolyte Analyzed Using Machine Learning Potential and Deep Generative Models

Authors

Hiroya Nitta - Presenter, JSOL Corporation
Teppei Fukuya, JSOL Corporation
Takayuki Nishio, JSOL Corporation
Taku Ozawa, JSOL Corporation
Kenji Yasuoka, Keio University
Molecular Dynamics (MD) is effective in providing insights at the atomic level. However, the computational costs are notably high, particularly for ab-initio MD (AIMD). Even though parallelization techniques have expanded the spatial scale of MD, extending the temporal scale remains a persistent challenge. There are several types of materials that require AIMD simulation to investigate their properties. Solid electrolytes are one of these kinds, since versatile force field parameters are not always available. Recently, new computational approaches, particularly those incorporating machine learning technology like machine learning (ML) potentials or generative models, are gaining traction for accelerating computation. ML potentials [1-4] provide universal potential of materials that are as accurate as ab-initio theory simulations. MD-GAN [5-7], based on Generative Adversarial Networks (GAN), has the potential to generate long-term data from short-term data, thus enabling faster computation through temporal parallelization. We have conducted AIMD simulation of LZP electrolyte system and have confirmed that MD-GAN can produce trajectories of Li ions in the system [8,9]. In this study, the MACE potential [4] is utilized to generate data for the learning process of MD-GAN, and it is applied to evaluate the diffusivity of Li ions at lower temperatures than those previously examined.
  1. Artrith and A. Urban, Comput. Mater. Sci. 114 (2016) 135-150
  2. Bartok, et al., PRB 87, 184115, 2013
  3. Drautz, PRB 99, 014104, 2019.
  4. I. Batatia, D. P. Kovacs, G. N. C. Simm, C. Ortner, G. Csanyi, NIPS'22, 830, 11423-11436, (2022).
  5. Endo, K. Tomobe, and K. Yasuoka, Proceedings of the AAAI Conference on Artificial Intelligence, 32, 2192 (2018).
  6. Kawada., K. Endo, K. Yasuoka, H. Kojima and N. Matubayasi, Journal of Chemical Information and Modeling, 63, 76-86(2022).
  7. Kawada, K. Endo, D. Yuhara and K. Yasuoka, Soft Matter, 18, 8446-8455(2022).
  8. Fukuya, T. Nishio, H. Nitta T. Ozawa, and K. Yasuoka , AIChE 2023 Annual Meeting
  9. Fukuya, T. Nishio, H. Nitta T. Ozawa, and K. Yasuoka , AIChE 2024 Annual Meeting