Human-Machine Collaborative Design of SnTe-Based Thermoelectric Materials via a Multiagent Framework Leveraging Large Language Models

Herein, we present an innovative large language model (LLM)-driven multiagent collaborative framework for the human-machine collaborative design of SnTe-based thermoelectric materials. Composed of strategy planning and reasoning modules, this framework efficiently extracts implicit domain knowledge from extensive literature and integrates historical experimental data to deduce optimization strategies and compositional ratio ranges, representing a novel paradigm that transcends traditional empirical and computational approaches in material design. Under the guidance of LLM, experimental results validate its efficacy: the incorporation of Sb, Ge, and Cu elements not only optimizes carrier concentration but also induces multiscale defects, synergistically modulating electrical and thermal transport properties. Notably, the (Sn0.58Sb0.12Ge0.3Te)0.95(Cu2Te)0.05 sample achieves a zT value of ∼1.2, marking a 40% increase compared to Sb-doped SnTe and a 267% rise relative to SnTe prepared by the same method, directly demonstrating that LLM guidance can significantly enhance material performance. This research not only showcases the potential of LLMs to revolutionize thermoelectric material development but also provides a valuable reference for high-performance material design in broader energy-related fields with implications for improving waste heat recovery and solid-state cooling technologies.