| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Hume-Rothery Symposium on Thermodynamics of Microstructure Stability and Evolution
|
| Presentation Title |
Predicting Domain Structure and Switching in Ferroelectrics: Physics-Informed Machine Learning and Phase-field Modeling |
| Author(s) |
Samrat Choudhury, Benjamin Rhoads, Abigail Hogue, Joseph Hafen |
| On-Site Speaker (Planned) |
Samrat Choudhury |
| Abstract Scope |
Phase-field approach has been widely adapted to investigate the formation of domain structure and domain switching behavior in ferroelectric and multiferroic materials. However, these simulations require significant computational resources, especially for 3-D microstructures. Deep learning-based machine learning (ML) tools have recently shown significant potential for accelerating the prediction of materials microstructure and its evolution. However, training these machines also requires significant amount of data. In this research, we will demonstrate that physics-informed machine learning requires comparatively less data during the training process and predicts domain structure in ferroelectrics with a higher accuracy in less time each time the network is informed about a new energy component or physics-informed function. Finally, we will present a graph neural network based interpretable machine learning framework to extract the underlying physics governing microstructure evolution, along with predicting future evolution of microstructure and material properties beyond the time domain in which the ML-model is trained. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Modeling and Simulation, Electronic Materials |