LLM‐Assisted Design of Ultralow‐κ Chalcogenides

ABSTRACT Accelerating the discovery of materials with extreme lattice thermal conductivity () is limited by the high cost of first‐principles screening. Here, we develop an LLM‐assisted low‐ discovery workflow, where Bayesian optimization (BO) proposes candidates, an LLM performs evidence‐conditioned re‐ranking, and physics‐based structure generation, stability validation, and calculations serve as the acceptance criteria. This integration enables efficient candidate prioritization while maintaining rigorous physical validation. The workflow identifies dynamically stable Ag‐based chalcogenides, including Ag 3 SbSe 4 , Ag 4 TiSe 4 , and Ag 4 TiTe 4 . Mechanistic analysis reveals that their ultralow originates from strong phonon renormalization and coherence‐dominated heat transport. Ag 3 SbSe 4 is predicted to reach a maximum of 2.94 at 500 K due to its ultralow of 0.11 W . Furthermore, automated high‐throughput screening of ternary compounds containing 14 elements identifies 2742 dynamically stable materials with W . These results demonstrate that evidence‐conditioned LLM re‐ranking effectively enhances BO‐driven, physics‐validated materials discovery.