LLM-assisted meta-analysis reveals a global shift toward non-radical dominance in peroxymonosulfate-based advanced...

Despite rapid growth in peroxymonosulfate (PMS)-based advanced oxidation processes (AOPs) research, systematic quantification of field-wide mechanistic trends has been lacking. Here we develop and validate an LLM-assisted data extraction framework (Core F1 = 0.935 against a 49-paper Gold Standard) to analyze 531 peer-reviewed publications spanning 2019-2025. Our meta-analysis reveals a statistically significant trend toward non-radical dominance: non-radical mechanisms increased from 35.0% (2019-2022) to 46.0% (2023-2025), reaching 51.0% in 2025 (z = -2.13, p = 0.017, Cohen's h = 0.22). Singlet oxygen (¹O₂) has emerged as the dominant reactive species (67.2% prevalence in 2025), while hydroxyl radical declined to 2.5%. Catalyst architecture serves as a significant predictor of mechanistic outcomes (χ² = 75.7, p < 0.001): single-atom catalysts achieve 71.1% non-radical selectivity compared to 31.1% for metal oxides, with M-N₄ coordination environments optimal for selective pathways. These findings provide data-driven design principles for rational non-radical catalyst engineering in water treatment applications. The complete dataset and extraction framework are openly available at Zenodo (DOI: 10.5281/zenodo.18017674).