About this Abstract |
Meeting |
2025 TMS Annual Meeting & Exhibition
|
Symposium
|
Computational Thermodynamics and Kinetics
|
Presentation Title |
Predicting Elemental Segregation Tendency via Ab Initio and Machine Learning Methods |
Author(s) |
Ho Lee, Sangtae Kim |
On-Site Speaker (Planned) |
Ho Lee |
Abstract Scope |
Understanding the grain boundary segregation energies of metastable metallic phases such as γ-Fe remains a challenge with classical molecular dynamics methods due to the potential transformation into ground state phases. Yet, many metallic materials employ metastable phases such as γ-Fe in stainless steel and understanding the segregation behavior among these phases is crucial in alloy designs. Here, we employ density functional theory calculations to first compute the segregation energies of γ-Fe phases for 11 transition elements in 4 symmetric tilt GBs. The 543 computed segregation energies are then employed to train a machine-learning model based on a gradient boosting algorithm to predict segregation tendencies for the total of 4499 segregation energy data among 23 undecorated GBs of γ-Fe. The trained model reveals that the size effects (Voronoi volume) and electronic effects (Hartigan dip) play synergistically for elements with fewer d-electrons than Fe, providing strong segregation tendencies for these elements. |
Proceedings Inclusion? |
Planned: |
Keywords |
Computational Materials Science & Engineering, Machine Learning, Iron and Steel |