| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Alloys and Compounds for Thermoelectric and Solar Cell Applications XIII
|
| Presentation Title |
Leveraging Machine Learning to Enhance the Performance of Filled Skutterudites Through Composition Optimization |
| Author(s) |
Yifan Sun, Sora-at Tanusilp, Masaya Kumagai, Hirofumi Tsuruta, Yuji Ohishi, Hiroaki Muta, Ken Kurosaki |
| On-Site Speaker (Planned) |
Yifan Sun |
| Abstract Scope |
Filled skutterudites, with a base formula of MxCoSb3, are renowned for their thermoelectric properties. The main method for optimizing their thermoelectric performance involves incorporating filler atoms such as Yb, La, Ce, and Ba to reduce lattice thermal conductivity through enhanced phonon scattering. Given the extensive variety of potential atom combinations, traditional experimental optimization of these materials is not feasible. In this study, we leverage the Starrydata2 database on thermoelectric materials and machine learning to systematically explore skutterudite compositions that minimize lattice thermal conductivity. Initially, we trained a machine learning model to identify skutterudites within the database and distinguish between the filler and host atoms. The composition and element properties of the filler and host systems, alongside temperature, are used as features to predict lattice thermal conductivity. Finally, the thermophysical properties of several selected candidates are experimentally verified to evaluate the reliability of our machine learning approach. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Energy Conversion and Storage, |