Presentation Information
[8p-P11-4]Analysis of Gas Molecule Diffusion Behavior in Titanium-Silica Membranes Using Machine-Learning Molecular Dynamics Simulations
〇Meguru Yamazaki1, Yuta Yoshimoto1, Isshin Gosha2, Taku Fujisawa2, Tomoya Matsuda1,2, Naoki Matsumura1, Yuto Iwasaki1, Atsuki Inoue2, Tomohisa Yoshioka2, Hiroshi Kawaguchi2, Yasufumi Sakai1 (1.Fujitsu Limited, 2.Kobe University)
Keywords:
Titanium-Silica Membrane,Machine Learning Potential
Neural Network Potentials (NNPs) enable highly accurate molecular dynamics (MD) simulations for large systems. In this study, we used an active learning-based NNP generation framework, GeNNIP4MD, to develop NNPs for a titania-silica membrane and gas molecules. The generated NNP model successfully realized stable 4-nanosecond MD simulations, accurately reproducing the diffusion behavior of hydrogen molecules and calculating the self-diffusion coefficient.