About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Algorithms Development in Materials Science and Engineering
|
| Presentation Title |
Temporal Acceleration of Molecular Dynamics Simulations for Iron using a Parareal Framework |
| Author(s) |
Aoi Watanabe, Ryuhei Sato, Bilal Mahmoud Hawchar, Frédéric Legoll, Laurent Brochard, Yasushi Shibuta |
| On-Site Speaker (Planned) |
Aoi Watanabe |
| Abstract Scope |
Simulating iron systems requires accurately capturing complex atomic interactions that govern various phase transformations. While highly accurate Machine Learning Potentials (MLPs) are actively developed to address this, their prohibitive computational costs hinder large-scale microstructural simulations. To overcome this limitation, we applied a hybrid Parareal framework to an equilibrium iron system, introducing time-parallelism as an orthogonal direction to spatial decomposition. This framework utilizes EAM-fs as a fast Coarse Solver to predict initial trajectories and a recently developed BNNP as an accurate Fine Solver to iteratively correct them in parallel. Enhanced by a dynamic window-shifting strategy that concentrates resources on these converging trajectories, the framework successfully achieved temporal acceleration. This demonstration proves the practical utility of Parareal for iron, establishing a viable pathway toward high-fidelity, large-scale simulations. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Modeling and Simulation, Iron and Steel, |