| About this Abstract |
| Meeting |
2024 TMS Annual Meeting & Exhibition
|
| Symposium
|
Phase Transformations and Microstructural Evolution
|
| Presentation Title |
C-14: Discovery of High-Pressure Phases – Integrating High-throughput DFT Simulations, Graphic Neural Networks, and Active Learning |
| Author(s) |
Ching-Chien Chen, Robert J Appleton, Saswat Mishra, Kat Nykiel, Alejandro Strachan |
| On-Site Speaker (Planned) |
Ching-Chien Chen |
| Abstract Scope |
Pressure-induced phase transformations in materials are of interest in a range of fields including geophysics, planetary sciences, and even shock physics. Importantly, some metastable phases can only be achieved via high pressures. Despite the significant interest, only a small subset of the known materials have been explored at high pressure either experimentally or theoretically. In this work, we combine high-throughput density functional theory (DFT) simulations and graphic neural networks machine learning models to predict the relative stability of different material structures at high pressure and identify potential phase transitions. The model was trained by an active learning loop, where new DFT simulations were appended after each cycle. As a result, the equation of state of more than 100,000 materials in the Materials Project were obtained through the model. Our investigation also led to the discovery of several novel high-pressure stable structures, expanding our understanding of materials under extreme conditions. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Phase Transformations, |