Interpretable causal machine learning elucidates performance regulation via atomic Fe‑S synergistic pathways in Sulfidated zero‑valent Iron.

Sulfidated zero-valent iron (S-ZVI) is a promising material for the reductive remediation of persistent contaminants, yet its performance is constrained by an inherent trade-off between reactivity and electron selectivity. To overcome this limitation, this study introduced an interpretable causal machine learning (CML) framework to construct a quantitative "preparation-structure-performance" causal network for the precision design of S-ZVI. The model identified the S/Fe molar ratio and Fe °content as two core structural determinants and delineated their optimal parameter window (S/Fe = 0.08-0.23; Fe0 = 0.75-0.92). Experimental validation for removal of halogenated organic compounds and high-valent heavy metals confirmed the maximum reaction rate, minimal hydrogen evolution, excellent performance stability, and low levels of sulfur leaching within this window. Integrated multiscale characterization and density functional theory calculations jointly revealed the atomic Fe‑S synergistic pathway for regulating material performance under this window. An optimal S/Fe ratio tuned the coordination environment of sulfur, which weakened hydrogen adsorption and induced interfacial charge redistribution. Sufficient Fe° content ensured a robust source of electrons. This synergy collectively optimized the entire electron transfer chain from bulk supply (Fe °core) through interfacial transport to surface utilization for efficient contaminant remediation. This interpretable CML-driven, mechanism-guided strategy proposes a theoretical basis for the precision design of S-ZVI and offers a perspective for the intelligent regulation of remediation materials in aquatic environment.