Dresden 2026 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 13: Data-driven Materials Science: Big Data and Workflows I

MM 13.7: Vortrag

Dienstag, 10. März 2026, 12:00–12:15, SCH/A251

Where Are Large Language Models Actually Useful for Materials Design? — •Hedda Oschinski, Maximilian L. Ach, David Greten, Konstantin S. Jakob, Christian Carbogno, and Karsten Reuter — Fritz-Haber-Institut der MPG, Berlin

The rapid development of large language models (LLMs) and LLM-based agents has opened new possibilities for accelerating materials discovery and design. In this work, we explore their potential in the context of solar cell materials, a class of systems requiring complex, multi-property optimization across chemistry and materials science. By systematically evaluating a range of tasks for a well-known test set of Elpasolites - from context preparation and descriptor prioritization to design hypothesis generation and autonomous validation within an agent framework - we identify where current LLMs provide genuine utility and where critical limitations remain. Our findings offer a grounded perspective on how these tools can be integrated into materials discovery workflows, and what developments are needed to expand their impact in the future.

Keywords: LLM; Agent; Materials Design; Materials Discovery; Solar Cells