About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Local Chemical Ordering and Its Impact on Mechanical Behaviors, Radiation Damage, and Corrosion
|
| Presentation Title |
a-MCDFT: A Data-Driven Framework for Accelerated Discovery of Local Chemical Ordering in Complex Alloys |
| Author(s) |
David Cereceda, Md Rajib Khan Musa, Aasig Majeed |
| On-Site Speaker (Planned) |
David Cereceda |
| Abstract Scope |
Local chemical ordering (LCO) plays a critical role in determining the phase stability and microstructural evolution of compositionally complex materials. However, identifying minimum energy configurations (MECs) through their LCO remains computationally demanding. We present a-MCDFT, a novel framework that integrates Monte Carlo sampling, density functional theory (DFT), and machine learning to accelerate MEC discovery in multi-principal element alloys. Built on the Cluster Expansion formalism and enhanced with a Local Outlier Factor model, a-MCDFT enables rapid and accurate energy predictions for atomic swaps. Applied to a tungsten-based quaternary high-entropy alloy, the method achieves MEC predictions with a relative error of ~0.022% compared to full DFT, while significantly reducing computational cost. This efficiency enables broader configurational sampling and larger supercell exploration, providing a scalable approach to investigate LCO in chemically complex systems. Our results highlight the potential of data-driven, physics-informed methods to overcome current limitations in modeling LCO and guiding materials design.
|
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, High-Entropy Alloys, Machine Learning |