About this Abstract |
Meeting |
2024 TMS Annual Meeting & Exhibition
|
Symposium
|
Computational Thermodynamics and Kinetics
|
Presentation Title |
Machine Learning Magnetic Ordering Prediction |
Author(s) |
Mingda Li |
On-Site Speaker (Planned) |
Mingda Li |
Abstract Scope |
Despite the central importance of magnetism in modern technology, the prediction of magnetic ordering, such as spin states and magnetic vectors, has been facing tremendous challenges. Classical methods like density functional theory are largely limited to collinear orders, while quantum methods like Quantum Monte Carlo or Exact Diagonalization experience exceedingly high computational complexity. As a result, reliable determination of magnetic orderings is largely limited to experimental approaches like magnetic diffraction, which has very limited resources. In this presentation, we introduce the Non-Abelian Integrative Neural Architecture for non-collinear magnetism prediction close to experimental accuracy. By representing the spins with special unitary group representations with free product with time-reversal and spatial inversion, the spin structures can be encoded while preserving crystallographic and magnetic symmetry. Our approach represents a unified machine-learning framework considering symmetry in composite spaces. |
Proceedings Inclusion? |
Planned: |
Keywords |
Machine Learning, Magnesium, Magnetic Materials |