Machine Learning-Driven FT-ICR MS Analysis of Leachate DOM Ozonation and Membrane Fouling

Pre-ozonation mitigates forward osmosis membrane fouling by transforming dissolved organic matter (DOM); however, the dynamic interplay between ozonation-induced precursor-product evolution and fouling behavior remains unclear. We demonstrated that pre-ozonation preferentially oxidizes fulvic acids, followed by soluble proteins (S-PN), in landfill leachates, whereas excessive ozone increases S-PN in aged leachates. Based on interpretable machine learning and linkage analysis, we identified key molecular properties (O/C, molecular weight [MW], oxygen count, and double bond equivalents minus oxygen) governing ozone reactivity and unveiled the following transformation pathways: oxygen addition, dealkylation, and desulfonation, that collectively convert unsaturated low-oxygen compounds into saturated, oxygen-rich mid/small molecules. In particular, sulfur-containing compounds (CHOS and CHONS) undergo conversion into highly oxidized and saturated compounds (CHO and CHON). In addition, pre-ozonation reduced fouling by oxidizing lignin/carboxyl-rich alicyclic (CRAM)-like and aliphatic/protein structures, notably those containing sulfur, while lowering DOM hydrophobicity and zeta potential. Over-ozonation in aged leachates converts CHONS-lignin/CRAM into low-MW CHON-aliphatic/proteins enriched with carboxylic acids, aggravating irreversible fouling. This study elucidates the novel mechanisms underlying the impact of ozone-driven DOM transformations on membrane fouling and offers critical insights for optimizing quantitative treatment strategies for recalcitrant organic wastewater.