About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Late News Poster Session
|
| Presentation Title |
H-33: CALPHAD-Type Thermodynamic Database from Machine Learning Potential |
| Author(s) |
Arkapol Saengdeejing, Hiori Kino, Yoshiyuki Kawazoe, Kazuyuki Higashino, Ryoji Sahara |
| On-Site Speaker (Planned) |
Arkapol Saengdeejing |
| Abstract Scope |
We demonstrated that using universal interatomic potential developed by machine learning can achieve the similar accuracy in constructing CALPHAD-type thermodynamic database comparing to what first-principles calculations data. The thermodynamic database of the Al-Ni binary system is constructed using only the ground state and finite-temperature data obtained from machine learning potential. The mixing energies between Al and Ni elements are obtained from the special quasi-random structures (SQS) which represented the random mixing configurations between both elements in the specific lattice. The calculated free energies are used for the assessment of the thermodynamic database based on the CALPHAD approach. Both calculated phase diagrams achieve the similar results, but the computational resources required from the machine learning potential are several orders of magnitude lower than the first-principles calculations. |
| Proceedings Inclusion? |
Undecided |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, Copper / Nickel / Cobalt |