Data-Driven Deciphering Structure-Mössbauer Spectroscopy Relationships in Iron-Based Compounds.

Accurate identification of material structures is crucial for establishing reliable structure-property relationships, yet this task is often hindered by the coexistence of multiple metastable phases and their facile transformations under reaction conditions. Although spectroscopic techniques are powerful tools for probing structural motifs, the lack of comprehensive reference data remains a critical bottleneck for experimental analysis. Here, we propose a unified strategy that integrates structure prediction, thermodynamic stability calculations, machine learning models, and Mössbauer spectroscopy simulations, tailored for iron-based intermetallic compounds. By constructing quantitative relationships between local structural features and Mössbauer parameters with machine learning, and by leveraging a theoretical spectral database to interpret experimental spectra, we successfully identified candidate iron sulfide metastable phases that were previously unresolved. This data-driven framework provides a robust complement to experimental approaches, enabling more accurate and efficient identification of metastable phases in dynamically evolving systems.