About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Computational Thermodynamics and Kinetics
|
| Presentation Title |
Uncertainty Propagation for Ab Initio Thermodynamic Phase Diagrams |
| Author(s) |
Jan Janssen, Prabhath Chilakalapudi, Haitham Wael Ali Awad Gaafer, Joerg Neugebauer |
| On-Site Speaker (Planned) |
Jan Janssen |
| Abstract Scope |
Temperature-concentration phase diagrams form the basis of alloy design, and the CALPHAD community has provided extensive experimental benchmarks over several decades. These benchmarks allow ab initio thermodynamic methods, such as Density Functional Theory (DFT), to be validated and demonstrate their quantitative accuracy. To improve computational efficiency, machine-learned interatomic potentials (MLIPs) are used as an approximation in DFT calculations. However, uncertainties from various sources, such as exchange-correlation functionals, basis set limitations, MLIP hyperparameters and thermodynamic approximations, must be carefully considered. This presentation introduces an extended workflow for calculating phase diagrams that incorporates uncertainty propagation from DFT calculations to the phase diagram using the pyiron framework. The intention is to quantify the uncertainty in phase boundaries across the entire temperature-concentration range and provide error bars for phase diagrams. This approach improves the reliability of ab initio based alloy design. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, ICME |