About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Computational Thermodynamics and Kinetics
|
| Presentation Title |
Machine Learning Assisted Compositional Exploration for Sluggish Vacancy and Interstitial Dumbbell Diffusion in High Entropy Alloys |
| Author(s) |
Wenjiang Huang, Xian-Ming Bai |
| On-Site Speaker (Planned) |
Xian-Ming Bai |
| Abstract Scope |
Sluggish diffusion kinetics is often used to interpret the unique properties of high entropy alloys (HEAs). However, its existence is still controversial, in part due to the limited compositions explored. To explore HEA’s vast compositional space, we have developed machine learning models to predict vacancy and <100> interstitial dumbbell migration barriers for arbitrary local atomic environment in an FeNiCrCuCo model HEA. These models are used as on-the-fly barrier calculators for kinetic Monte Carlo (ML-KMC) modeling to calculate vacancy- and interstitial-mediated diffusivities in various non-equiatomic compositions. For the more complex interstitial dumbbell diffusion, Bayesian optimization is coupled with ML-KMC to efficiently explore the compositional space. For both diffusion mechanisms, our results show that low-concentration of fast diffusers can slow down diffusion kinetics through enhanced local trapping effect, which may provide valuable insight for HEA compositional design. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, High-Entropy Alloys, Machine Learning |