Literature-Scale Mössbauer Data Enable Prediction, Electronic-State Interpretation, and Chemical Insight in Molecular Iron Chemistry

Experimental spectroscopy is central to molecular characterization, yet conventional quantitative analysis often compresses rich structural and electronic information into a few calculated observables, leaving much of the experimental context and expert interpretation reported in the literature underused. Here, using molecular 57Fe Mössbauer spectroscopy as a representative case, we reconstructed more than 300 publications into FeMössDB, a site-resolved dataset containing 502 Fe complexes and 624 Fe sites with linked structural, experimental, and interpretive information. Using FeMössDB, we developed FeMössML for three complementary aspects of Mössbauer analysis: isomer-shift/quadrupole-splitting (IS/QS) prediction, electronic-state interpretation, and analysis of structure–state–spectroscopy relationships. FeMössML reduces IS/QS prediction errors by about 30% relative to a single global density functional theory (DFT) calibration and achieves cross-validated electronic-state accuracies of up to 92%, while correctly assigning several challenging low-valent and electronically delocalized Fe complexes. Feature analysis further recovers established relationships between Mössbauer parameters and oxidation state, spin state, and Fe–L geometry from the literature-scale dataset. FeMössML can also use geometries rapidly generated by Universal Models for Atoms (UMA) or experimental X-ray structures without target-specific DFT hyperfine calculations, reducing wall-clock time by more than three orders of magnitude. More broadly, this work points to a new paradigm for spectroscopic analysis, moving from compound-by-compound quantum-chemical calculation and expert interpretation toward literature-scale reuse of experimental and interpretive data for prediction, interpretation, and systematic relationship analysis.