| About this Abstract |
| Meeting |
TMS Specialty Congress 2025
|
| Symposium
|
3rd World Congress on Artificial Intelligence in Materials & Manufacturing (AIM 2025)
|
| Presentation Title |
Active Learning for Rapid Targeted Manufacturing of Thermoelectric Thin Film Alloys |
| Author(s) |
Nathan S. Johnson, Aashwin Mishra, Apurva Mehta, Dylan Kirsch |
| On-Site Speaker (Planned) |
Nathan S. Johnson |
| Abstract Scope |
As the integration of machine learning into materials science becomes increasingly vital for the synthesis of advanced materials, this study presents an active learning (AL) strategy to hasten the creation of thin-film compositionally complex alloys (CCAs) targeting thermoelectric applications. Traditional methods for thin film synthesis are often labor-intensive and time-consuming, underscoring the necessity for more efficient yet accurate approaches. We utilized a high-throughput combinatorial sputtering technique to produce an extensive array of alloy compositions. Active machine learning guided a human researcher in efficiently determining the optimal synthesis conditions for a series of five-element alloys. Our findings reveal that the AL approach significantly surpasses conventional human-driven sampling in optimizing synthesis parameters. Additionally, we employed transfer learning to pre-train models, which further enhanced efficiency. Models initially trained on lower-dimensional alloy systems were applied to higher-dimensional systems, resulting in immediate improvements in predictive accuracy. |
| Proceedings Inclusion? |
Undecided |