About this Abstract |
| Meeting |
2021 TMS Annual Meeting & Exhibition
|
| Symposium
|
Phase Stability, Phase Transformations, and Reactive Phase Formation in Electronic Materials XX
|
| Presentation Title |
Machine Learning for Perovskite Phase Stability |
| Author(s) |
Dane Morgan, Wei Li, Ryan Jacobs |
| On-Site Speaker (Planned) |
Dane Morgan |
| Abstract Scope |
Machine learning methods are a powerful tool to rapidly predict phase stability, particularly when large amounts of calculated stability data are available. In this talk I will discuss recent work on predicting formation energies of perovskite structures[1]. We use combinations of elemental features and find that an extra trees method yields a good 5-fold cross-validation accuracy on energy above the convex hull. We then demonstrate that the cross-validation is a very optimistic estimate appropriate only for those chemistries that are well-represented in the data set. This work illustrates some of the capabilities for machine learning to predict stability but also the challenges of extrapolation to new chemistries.[1] 1. Li, W., Jacobs, R. & Morgan, D. Predicting the thermodynamic stability of perovskite oxides using machine learning models. Computational Materials Science 150, 454–463 (2018). DOI: 10.1016/j.commatsci.2018.04.033 |
| Proceedings Inclusion? |
Planned: |