Introduction

The burgeoning urbanization and industrialization have significantly exacerbated organic wastewater discharge, necessitating the utilization of specific reactive species for efficient water purification1,2. High-valent metal-oxo species (HVMOs), a ubiquitous intermediate in synthetic and enzymatic reactions, are currently emerging as pivotal reactive species in a wide range of biological and chemical oxidation processes such as alkane hydroxylation, epoxidation, water oxidation, and heterogeneous Fenton-like oxidation3,4,5,6. Peroxymonosulfate (PMS)-based Fenton-like reactions offer a feasible solution for organic pollutant treatment, benefiting from the generation of multiple radical and nonradical species7. Radicals generally suffer from ultrashort lifetimes (10−9–10−6 s), low steady-state concentrations (10−15–10−12 M), and rapid scavenging by background water constituents, while nonradical singlet oxygen (1O2) is also limited by mild oxidation capacity (0.81 VNHE) and high decay rate in water (2.5 × 105 s−1)8. In contrast with the mentioned reactive species, HVMOs possess a longer lifetime (7–10−1 s), higher steady-state concentration (~10–8 M), and robust resistance against complex water matrices while maintaining appreciable redox potential (>1.95 VNHE)9,10. These typical features perfectly retain the merits of radical and nonradical species, enabling HVMOs as the prevailing candidate for the decontamination of refractory organic pollutants in complex wastewater scenarios. With the advent of single-atom catalysis, atomically dispersed transition metal–nitrogen–carbon (M–N–C) catalysts are renowned for their unparalleled activity and selectivity to enable HVMOs generation via two-electron transfer or proton-coupled electron transfer process11,12,13,14,15,16. Nevertheless, radicals and 1O2 frequently perform as the primary or secondary reactive species, despite sharing identical local coordination like the conventional Fe–N4 motif17,18,19. To achieve meticulous control over HVMOs formation, regulating the electronic structure of metal centers is of great necessity yet challenging.

Precisely modulating the electron density and configuration of active metal centers has become the fundamental strategy for the targeted production of HVMOs. Theoretically, reducing the electron density of metal center enables PMS adsorption and HVMOs generation with nearly 100% selectivity, which has been exemplified by O/B/S doped single-atom Fe/Co catalysts13,15,20,21. In reality, radicals and 1O2 may also be preferentially generated via single-electron transfer or PMS oxidation21,22,23,24. Additionally, the high-spin electronic configuration with singly occupied π antibonding orbitals (π*) favors the HVMOs generation through a low energy barrier12,14,25. However, the existing findings indicate that the surface-activated PMS complexes (catalyst-PMS*) or radicals are exclusively formed at high-spin metal sites26,27,28. The key difficulty in the selective generation of HVMOs lies in the complex interfacial redox nature between catalysts and PMS, leading to both continuous electron loss and dynamic configuration changes at the individual metal site29,30. Accordingly, there is an ongoing debate as to whether such a mononuclear metal center can fulfill the requirements of manipulating electron density and configuration concurrently for the highly selective formation of HVMOs.

Recently, the metastable catalyst-PMS* complexes have been proposed to induce the generation of HVMOs31,32. Some researchers reported that the homonuclear Fe–N–C catalysts pairing neighboring Fe sites are prerequisites for generating FeIV = O from Fe–N–C–PMS* while requiring rigorous inter-site distance and adequate PMS dosage33. The incorporation of heteronuclear metal sites offers a versatile approach to improving the selectivity of HVMOs generation. For instance, the interplay of Fe−Mn diatomic sites promotes electronic delocalization and band gap narrowing for sustainable generation of Fe(V)–O–O–Mn(IV)34. Besides, Fe−Co diatomic pairs are also unraveled to rearrange the electronic configuration via spin-state reconstruction to produce stable FeIV−O−CoIV species35. Nevertheless, these diatomic metal categories commonly used in PMS activation are primarily limited to d-block transition metal atoms, which possess unfilled 3d orbitals and can readily trigger electron competition with metal counterparts. This behavior inevitably leads to inadequate electron transfer or reverse electron flow to generate radicals/catalyst-PMS* complexes and 1O2 species, thereby diminishing the selectivity of HVMOs generation36,37,38,39. Sharply distinguished from d-block transition metals, p-block metals with fully occupied d10 configurations restrict electron mobility and typically exhibit inferior catalytic reactivity, which might prevent electron competition with d-block transition metals during PMS activation40. Moreover, p-block metals can serve as Lewis acid sites to chemically bond with electronegative oxygen atoms, including terminal oxygen and ether oxygen, highly similar to the adsorption of peroxanion (HSO5−) by metal centers41,42. This feature facilitates the construction of electron-rich regions surrounding d-block transition metals and thereby potentially provides the electron sources required for the HVMOs generation43,44. To date, several pioneering studies have found that the electronic structure of the d-block Fe active center can be precisely modulated by introducing p-block metals such as indium (In), antimony (Sb), and tin (Sn), which demonstrate remarkable electrocatalytic activity and selectivity for oxygen reduction reaction and electrochemical CO2 reduction45,46,47. Therefore, it can be reasonably postulated that the introduction of p-block metals can effectively modulate the electron configuration of Fe active center and boost electron transfer to PMS, ultimately achieving the selective generation of FeIV = O.

Inspired by the motivations, we propose a p-block metal coordination strategy to dynamically modulate the electronic structure of single-atom Fe sites for selective FeIV = O generation in PMS activation. Bismuth (Bi) has been known as an environmentally benign and cost-effective p-block metal element and gained widespread applications in catalytic chemistry48,49,50,51. Here, Bi is selected as a representative p-block metal to coordinate with the Fe active center in the distinct configuration of two nitrogen-bridged Fe−Bi diatomic pairs, and the obtained FeBi–N–C is adopted as a typical model catalyst to initiate the Fenton-like reaction by PMS activation, with single-atom Fe–N–C and prevalent diatomic catalysts as counterparts. The effectiveness and applicability of FeBi–N–C for PMS activation are systematically evaluated by the degradation of diverse organic pollutants. The selective generation of FeIV = O and dynamic electronic modulation mechanism are further confirmed by electron paramagnetic resonance, scavenging tests, and in-situ experiments, coupled with theoretical calculations. The universality of the p-block metal coordination strategy is also validated using other catalytic models such as FeIn–N–C and FeSb–N–C. Overall, our study provides a deeper atomic-level understanding of the selective generation of FeIV = O in Fenton-like chemistry through a electronic engineering strategy.

Results

Synthesis and characterization of catalysts

The heteronuclear FeBi–N–C catalyst was synthesized through a facile two-stage pyrolysis strategy as illustrated in Fig. 1a. Initially, melamine underwent an in-situ polymerization process to form graphitic carbon nitride, while iron acetate and bismuth acetate coordinated with the highly reactive –NH2 group on L-alanine during the first pyrolysis process. Due to the electron delocalization over the ligand π-system, the electrophilic metal centers accepted the lone pair electron and were reduced to low oxidation state metal atoms. Subsequently, the higher pyrolysis temperature resulted in the carbonization of graphitic carbon nitride to catalyze the growth of graphene-like nanosheets, which served as a soft template to prevent the aggregation of low-valent metal species and thus enabled the homogeneous dispersion of Fe and Bi atoms on the carbon substrate. The Fe and Bi contents of FeBi–N–C, measured by inductively coupled plasma mass spectrometry, are detected to be 1.43 and 0.90 wt%, respectively (Table S1). Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images of FeBi–N–C show a two-dimensional porous nanosheet morphology without visible metallic agglomerates or nanoparticles (Figs. 1b and S1). Neither crystalline metal nanoparticles nor clusters are observed by powder X-ray diffraction patterns (Fig. S2). Dark-field scanning transmission electron microscopy (STEM) and corresponding energy-dispersive X-ray spectroscopy elemental mapping demonstrate the uniform distribution of C, N, Fe, and Bi elements (Fig. 1c, d). Atomic force microscopy analysis depicts a sheet thickness of 0.8–1.2 nm (Fig. 1c inset), consistent with the layered graphene-like structure. Aberration-corrected high-angle annular dark field-scanning transmission electron microscopy (AC-HAADF-STEM) images were used to reveal the existing forms of Fe and Bi. Numerous bright spots in Fig. 1e confirm the atomic-level dispersion of Fe and Bi on the porous carbon substrate. Z-contrast analysis can directly distinguish the existence of heteronuclear diatomic pairs from all spot pairs, due to the Z-contrast difference between heavier Bi (83) and lighter Fe (26) atoms on nonmetallic species52. Analysis of representative regions marked as 1 and 2 reveals the adjacent bright spots with distinct intensities, confirming the prevalence of heteronuclear Fe–Bi diatomic pairs with an average interatomic distance of 0.25 nm (Fig. 1f, g). Statistical analyses of >100 spot pairs indicate that almost 90% can be identified as Fe–Bi diatomic sites, despite the existence of individual Fe and Bi atoms (Fig. S3). These findings suggest that the heteronuclear Fe–Bi diatomic pairs likely serve as the major active sites for PMS activation. Other catalysts, including FeIn–N–C, FeSb–N–C, control samples, and corresponding single-atom counterparts, which follow similar synthetic protocols, also present similar morphological characteristics to FeBi–N–C (Figs. S4–S22). Notably, FeIn–N–C and FeSb–N–C primarily consist of the heteronuclear Fe–In and Fe–Sb diatomic pairs, respectively, without detectable agglomerates of metallic nanoparticles. The higher ID/IG ratios of FeBi–N–C, FeIn–N–C, and FeSb–N–C than Fe–N–C suggest that the p-block metal coordinated diatomic pairs contribute to the formation of defective structures (Fig. S23), which were verified by the sp2/sp3 ratios in the high-resolution C 1s XPS spectra (Fig. S24). The structural defects of catalysts facilitate the creation of a large specific surface area and abundant mesoporous architecture (Fig. S25 and Table S2). Specifically, the type-IV N2 adsorption-desorption isotherm and the pore size distribution curve of FeBi−N−C indicate the predominantly mesoporous structure with an average pore diameter of 8.11 nm, in alignment with the AC-HAADF-STEM images. The highly exposed surface area and mesoporous feature are conducive to the availability of active sites and efficient mass transport between PMS and catalysts53.

Fig. 1: Synthesis and morphological structure of FeBi–N–C.
Fig. 1: Synthesis and morphological structure of FeBi–N–C.
Full size image

a Schematic illustration of the synthetic strategy for FeBi–N–C. b Transmission electron microscopy (TEM) image. c High-angle annular dark field-scanning transmission electron microscopy (HAADF-STEM) image and atomic force microscopy (AFM) image (inset). d Energy-dispersive X-ray spectroscopy (EDS) elemental mapping images. e Aberration-corrected HAADF-STEM image. f Enlarged spot pairs from Figs. 1e and 3D atom-overlapping Gaussian-function fitting map along with intensity profiles derived from region 1 and g region 2.

Using FeBi–N–C as a representative model, X-ray absorption spectroscopy was conducted to investigate the electronic state and local coordination of Fe and Bi atoms. The Fe K-edge X-ray absorption near-edge structure (XANES) spectra demonstrate that the absorption edge positions of FeBi–N–C and Fe–N–C are located between Fe foil and Fe2O3, and the edge position of FeBi–N–C aligns closer to FeO but deviates from Fe foil (Fig. 2a). Deriving from the first-derivative XANES curves and fitted results, the average oxidation state of Fe in FeBi–N–C is 2.17 (Fig. S26a, b). In the Fe K-edge Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS) spectra, FeBi–N–C and Fe–N–C showcase the major peak and broad peak at ~1.48 Å and ~2.45 Å, corresponding to the first-shell Fe–N and Fe–N–C scattering paths, respectively (Fig. 2b)54,55. The Fe–Fe bonds in the Fe foil (~2.21 Å) are not present for FeBi–N–C and Fe–N–C, indicating the absence of Fe-containing nanoparticles. The EXAFS fitting curves for Fe–N and Fe–N–C paths coincide well with the experimental data (Fig. S27a–d), and the relative intensity (χ(R)) for Fe–N–C path is significantly weaker than that for Fe–Fe, revealing that Fe tends to be atomically dispersed on carbon substrate instead of forming metal-metal coordination (Fig. S28). The quantitative results suggest that Fe atoms in FeBi–N–C are coordinated with two types of N atoms, and the average coordination numbers of Fe–N scattering paths are determined to be 2.1 and 1.8 (Table S3). Wavelet-transform (WT) contour plot of FeBi–N–C in the Fe K-edge EXAFS spectra exhibits the intensity maxima at ~4 Å−1 assigned to the Fe–N coordination, which is almost identical to that of Fe–N–C but distinctly different from Fe foil, FeO, and Fe2O3 references (Figs. 2c, d and S29)56. Likewise, the Bi L3-edge XANES spectra along with fitted first-derivative curves demonstrate that the Bi atom in FeBi–N–C has an average oxidation state of about 2.6 (Figs. 2e and S26c, d). The FT-EXAFS spectra of Bi L3-edge for FeBi–N–C and Bi–N–C suggest the existence of Bi–N and Bi–N–C scattering paths without observable Bi–Bi bonds (Fig. 2f)57. As compared with the Bi–Bi bonds, the relatively lower χ(R) intensity for Bi–N–C path reveals that Bi is atomically dispersed on the carbon substrate (Fig. S28). The Bi atoms in FeBi–N–C are coordinated with two types of N atoms, and the average coordination numbers of Bi–N scattering paths are determined to be 2.0 and 1.9 (Fig. S27e–h and Table S4). The intensity maxima of FeBi–N–C in the WT contour plot of Bi L3-edge EXAFS spectra are located at ~5 Å−1, coinciding with the Bi–N coordination (Fig. 2g, h)58. In summary, the above analyses further confirm that the heteronuclear Fe–Bi diatomic pairs coordinate with nitrogen to construct a unique FeBi–N6 configuration.

Fig. 2: X-ray absorption spectroscopy (XAS) spectra of Fe–N–C, Bi–N–C, and FeBi–N–C.
Fig. 2: X-ray absorption spectroscopy (XAS) spectra of Fe–N–C, Bi–N–C, and FeBi–N–C.
Full size image

a Normalized Fe K-edge X-ray absorption near-edge structure (XANES) spectra and b Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS) spectra for Fe K-edge of Fe–N–C, FeBi–N–C, and reference samples. c Wavelet-transform (WT) contour plot in the Fe K-edge EXAFS spectra of Fe–N–C and d FeBi–N–C. e Normalized Bi L3-edge XANES spectra and f FT-EXAFS spectra for Bi L3-edge of Bi–N–C, FeBi–N–C, and reference samples. g WT contour plot in the Bi L3-edge EXAFS spectra of Bi–N–C and h FeBi–N–C.

Enhanced Fenton-like performance enabled by p-block metal coordination

The catalytic performance of FeBi–N–C, along with controlled samples in PMS-based Fenton-like reactions was initially assessed by the degradation of phenol (PhOH), a bio-refractory organic pollutant commonly detected in wastewater. In optimization experiments, FeBi–N–C with 0.9 wt% Bi loading and 1.4 wt% Fe loading showed superior activity and was thus determined as the typical catalyst with an optimal dosage of 0.1 g L–1 (Figs. S30 and S31). As shown in Fig. S32 and Table S5, individual PMS hardly degraded PhOH, and less than 10% of PhOH could be removed by the adsorption of catalysts. Notably, Bi–N–C showed inferior catalytic activity for PMS activation due to the closed-shell d10 electronic configuration of p-block Bi metal (Fig. 3a)59,60. Strikingly, the single-atom Fe site with Bi coordination achieved complete PhOH removal within 30 min, substantially outperforming Fe–N–C/PMS, Bi–N–C/PMS, and the physically mixed Fe–N–C+Bi–N–C/PMS systems. The observed reaction rate constant (kobs) of PhOH degradation by FeBi–N–C was 3.5, 9.3, and 4.7 folds that of Fe–N–C, Bi–N–C, and Fe–N–C+Bi–N–C, respectively (Fig. S33). In addition, the FeBi–N–C/PMS system also demonstrated exceptional mineralization efficiency, as evidenced by higher removal of total organic carbon (Fig. S34). These results confirm that the incorporated p-block Bi atom in FeBi–N–C drastically enhances PMS activation and catalytic degradation of PhOH.

Fig. 3: Fenton-like performance evaluation of different PMS-based reaction systems.
Fig. 3: Fenton-like performance evaluation of different PMS-based reaction systems.
Full size image

a Kinetic curves of phenol degradation by PMS activated with different catalysts. b Comparison of normalized kobs values with previously documented catalysts. c PMS utilization efficiency (PUE) values of different catalyst-activated PMS systems. d Phenol degradation curves with and without the addition of KSCN in different reaction systems. e Comparison of kobs and kper-site values in Fe–N–C/PMS, FeMd–N–C/PMS, and FeMp–N–C/PMS systems. f Comparison of kper-site values in different catalyst-activated systems. g Phenol degradation performance of FeBi–N–C/PMS system in the presence of different initial pH, inorganic anions, humic acid, various water matrices, and consecutive cycles by measuring the removal efficiency and kper-site values. h Comparison of kper-site values for the degradation of diverse organic pollutants in Fe–N–C/PMS and FeBi–N–C/PMS systems. Reaction condition: [pollutants] = 10 mg L−1, catalyst dosage = 0.1 g L−1, [PMS] = 0.2 g L−1, [KSCN] = 2 mM, [Cl−] = [NO3−] = [HCO3–] = [H2PO4−] = [HA] = 10 mM, T = 298 K, initial pH = 7. Error bars in the figure represent the standard deviation from three replicate experiments.

Motivated by the unusually excellent reactivity of FeBi–N–C, the single-atom Fe site was coordinated with other p-block and d-block metals to prepare the FeMp/d–N–C catalysts (Mp: In and Sb, Md: Cu, Mn, and Ni) for systematic evaluation of Fenton-like performance. Notably, the FeIn–N–C and FeSb–N–C catalysts also delivered higher catalytic activity compared to their single-atom counterparts, even outperforming the generally recognized FeMd–N–C catalysts (Fig. S35). By normalizing kobs values to PMS concentration and catalyst dosage, the performance of FeMp–N–C stood out among the state-of-the-art single-atom M–N–C (M = Fe, Co, etc.) and diatomic catalysts reported to date (Fig. 3b and Table S6). These results profoundly underscore the universality of the p-block metal coordination strategy in boosting PMS activation and Fenton-like performance. In addition, the FeMp–N–C catalysts also yielded ultrahigh PMS utilization efficiency (PUE), which matched closely with the trend of PhOH degradation performance (Figs. 3c and S36). To elucidate the origin of enhanced reactivity, we investigated the structure-activity relationship between the degradation performance and specific properties of catalysts (Fig. S37). Taking into account the insignificant correlation of the kobs values in different systems with nonmetallic sites of catalysts, including specific surface area, defect, and N content, single-atom Fe sites serve as the primary active centers for PMS activation. To verify this hypothesis, potassium thiocyanate (KSCN) was employed to mask the isolated Fe sites9. The PhOH degradation was completely suspended after KSCN addition, validating the pivotal role of Fe sites during PMS activation (Fig. 3d). It should be noted that these leached metal ions exerted no influence on PhOH degradation, indicating the negligible contributions from homogeneous catalysis (Fig. S38). Furthermore, the kobs values did not correlate with the Fe content of catalysts (Fig. S37d). Intriguingly, the Fe content normalized reaction rate constants (kmass) of FeMp–N–C were found to be significantly higher than those of other catalysts, implying that enhanced reactivity should not be attributed to the loading of Fe sites (Fig. S39). Further normalizing reaction rate constants to the molar ratio of Fe content (kper-site) endowed FeBi–N–C with the highest value, followed closely by FeIn–N–C and FeSb–N–C, which were 3.6–4.4 folds that of Fe–N–C and far superior to ever-reported values when compared with most of the advanced Fe-based single-atom and diatomic catalysts (Fig. 3e, f). The above results strongly support the enhanced intrinsic activity of single-atom Fe sites with p-block metal coordination.

As a proof of concept, FeBi–N–C was thereafter selected to evaluate the application prospects in water purification. FeBi–N–C/PMS system maintained appreciable degradation efficiency and stability, as evidenced by the wide pH adaptability from 3 to 11, negligible interference by environmentally widespread inorganic anions (Cl−, NO3−, HCO3−, and H2PO4−) and humic acid (HA), robust resistance in various water matrics (tap water, lake water, groundwater, and wastewater), as well as satisfactory reusability after five consecutive cycles (Figs. 3g and S40). The excellent stability was further revealed by the unaffected morphological and chemical structures of FeBi–N–C after the reaction (Fig. S41). FeBi–N–C/PMS system was also utilized for the degradation of diverse organic pollutants, including phenolics, antibiotics, and organic dyes. The broad-spectrum efficacy of FeBi–N–C was demonstrated by superior intrinsic activity toward 12 typical pollutants, showcasing the versatility of the FeBi–N–C/PMS system (Fig. 3h). Nevertheless, the electron-deficient pollutants such as benzoic acid, p-nitrobenzene, and atrazine exhibited significantly lower degradation efficiencies (less than 20%) (Fig. S42). The substrate-dependent reactivity of the FeBi–N–C–C/PMS system confirms a highly catalytic selectivity toward specific pollutants.

Reactive species transformation with p-block metal coordination

Electron paramagnetic resonance (EPR) tests were performed to discern the generated reactive species during PMS activation using 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) and 2,2,6,6-tetramethyl-4-piperidone (TEMP) as spin-trapping agents35. As shown in Fig. S43a, the characteristic signals of DMPO-•OH, DMPO-SO4•−, and DMPO-O2•− were not observed, ruling out the generation of free radicals. In contrast, the EPR spectra of all catalysts/PMS systems showed the typical seven-line signals of 5,5-dimethyl-2-oxopyrroline-1-oxyl adduct, potentially derived from the DMPO oxidation by 1O2, high-valent metal-oxo, and other oxidative species61. Furthermore, the attenuated signal intensity of TEMP-1O2 adduct in all catalysts/PMS systems relative to PMS alone indicated that 1O2 mainly originated from PMS self-decay rather than heterogeneous PMS activation (Fig. S43b). The generation of 1O2 by uncatalyzed PMS self-decay is relatively inefficient with a low rate constant of only 0.2 M−1 s−1, and may not be responsible for pollutant degradation, which aligns with the unaffected PhOH removal using β-carotene as a 1O2-specific probe (Fig. S44). The above results were further corroborated by scavenging experiments, where tert-butanol (TBA), methanol (MeOH), p-benzoquinone (p-BQ), TEMP, and dimethyl sulfoxide (DMSO) served as specific scavengers for •OH, •OH/SO4•−, O2•−, 1O2, and FeIV = O, respectively62. These selected scavengers did not significantly consume PMS in the absence of catalysts, which ensures the reliability of scavenging results (Fig. S45). As shown in Figs. 4a and S46, the negligibly inhibitory effect of TBA, MeOH, p-BQ, and TEMP on PhOH degradation suggested that neither radicals nor 1O2 were the primary reactive species in all catalysts/PMS systems. However, the addition of DMSO, which moderately suppressed the removal efficiency of PhOH in Fe–N–C/PMS and FeMd–N–C/PMS systems, exerted considerable inhibition on PhOH degradation in FeMp–N–C/PMS systems, thus implying the involvement of FeIV = O species. Consistent with the scavenging results, trans-stilbene and methyl phenyl sulfoxide (PMSO), as the representative chemical probes of FeIV = O, could be selectively oxidized into trans-stilbene oxide and methyl phenyl sulfone (PMSO2) via an oxygen atom transfer reaction, respectively, consolidating the predominance of FeIV = O in FeMp–N–C/PMS systems (Figs. 4b–d and S47–51)16. According to the 18O isotope labeling experiments, the mass spectra in FeMp–N–C/PMS systems exhibited two fragments of PMS16O18O at m/z ~ 159.036 and m/z ~ 80.989, compared to the PMS16O16O at m/z 157.0323 and m/z 78.9850 in the PMS alone system, further providing conclusive evidence for the FeIV = O generation (Figs. S52–55). The calculated steady-state concentrations of FeIV = O in FeMp–N–C/PMS systems (1.89–2.24 × 10–8 M), based on the transformation kinetics from PMSO oxidation into PMSO2 generation, were found to be substantially improved by almost one order of magnitude relative to Fe–N–C/PMS and FeMd–N–C/PMS systems, which even exceed those in recently reported HVMOs-dominated oxidation systems (Figs. 4e and S56–59 and Table S7). The above results provide convincing evidence to support that FeMp–N–C catalysts with p-block metal coordination contribute to the selective generation of FeIV = O and excellent degradation of PhOH.

Fig. 4: Reactive species identification.
Fig. 4: Reactive species identification.
Full size image

a Effect of different scavengers including methanol (MeOH), tert-butanol (TBA), p-benzoquinone (p-BQ), 2,2,6,6-tetramethyl-4-piperidone (TEMP), and dimethyl sulfoxide (DMSO) on phenol degradation in different catalysts-activated PMS systems. b The conversion rate of trans-stilbene to trans-stilbene in different catalysts-activated PMS systems. c Kinetic curves of methyl phenyl sulfoxide (PMSO) consumption and methyl phenyl sulfone (PMSO2) generation, and d the calculated transformation rate (η) in FeBi–N–C/PMS system. e Steady-state concentrations of FeIV = O in different catalysts-activated PMS systems. f In-situ Raman spectra of different systems. g Contributions of radicals, 1O2, FeIV = O, and electron transfer pathway to phenol degradation in different reaction systems. Reaction condition: [pollutants] = 10 mg L−1, catalyst dosage = 0.1 g L−1, [PMS] = 0.2 g L−1 or 50 g L−1 (if needed), [MeOH] = [TBA] = 100 mM, [TEMP] = 10 mM, [DMSO] = 20 mM, [PMSO] = 50 μM, [trans-stilbene] = 1.0 mM, T = 298 K, initial pH = 7. Error bars in the figure represent the standard deviation from three replicate experiments.

As for the Fe–N–C and FeMd–N–C without p-block metal coordination, other nonradical processes, e.g., the electron-transfer pathway (ETP), might be responsible for PMS activation and PhOH degradation. In-situ Raman measurement was initially performed to monitor the ETP process. As shown in Fig. 4f, the new characteristic peaks at ~834 cm−1 corresponding to surface-activated PMS complexes (PMS*) appeared after the interaction with catalysts, but completely disappeared upon the addition of PhOH63. This result indicates that Fe–N–C and FeMd–N–C enable the PMS adsorption to form the reactive PMS* intermediates, which subsequently withdraw electrons from PhOH to initiate the ETP oxidation. In addition, the open-circuit potentials on Fe–N–C and FeMd–N–C first increased and reached the equilibrium potential upon the addition of PMS, and then declined immediately after adding electron-donating PhOH as reductant, suggesting the formation and decomposition of surface PMS* complexes (Fig. S60). The ETP-dominated oxidation process with PMS* intermediates was also verified by the accelerated consumption of PMS after PhOH addition, indicative of a typical ETP-based catalytic system (Fig. S61). Moreover, the ETP process mainly occurred on the surface of catalysts because K2Cr2O7, as the solution-phase electron scavenger, hardly inhibited the PhOH removal (Fig. S62)64. Combining the above results from EPR tests and scavenging experiments, the contribution of reactive species to PhOH degradation in different catalytic systems was further quantified. In the Fe–N–C/PMS and FeMd–N–C/PMS systems, the generated PMS* complexes were more inclined to directly oxidize PhOH via the ETP process, accompanied by the minor contribution of FeIV = O to PhOH degradation (<25%) (Figs. 4g and S63). With p-block metal coordination, however, FeIV = O species accounted for 82.7–90.1% of all reactive species. These results collectively confirm that p-block metal coordination in FeMp–N–C markedly promotes the selective generation of FeIV = O from PMS* intermediates, ultimately facilitating the substantially enhanced catalytic activity and ultrahigh PUE.

Mechanisms of selective FeIV = O generation

The generation of FeIV = O species has been realized by the transformation of PMS* intermediates formed on traditional Fe–N–C with specific electronic structures, yet remains elusive in the FeMp–N–C/PMS systems. Therefore, disclosing how the electronic structure of single-atom Fe sites affects the FeIV = O generation with p-block metal coordination is of great necessity. To advance the understanding of interfacial electron transfer between the PMS and catalysts, time-dependent in-situ surface-enhanced Raman spectroscopy was initially used to uncover the dynamic evolution of the electronic structure of FeMp–N–C, with Fe–N–C and FeMd–N–C as comparative references. As shown in Fig. 5a, the PMS alone displayed three characteristic peaks at around 882, 980, and 1060 cm−1, which could be attributed to the stretching vibrations of O–O, SO42−, and SO3− in PMS molecule35. As the reaction time prolonged, the distinct peak of PMS* was first observed at 2 min upon the activation of FeBi–N–C, followed by the appearance of a new peak at ~843 cm−1 corresponding to the FeIV = O species, which suggested the transformation from PMS* to FeIV = O33. Subsequently, the characteristic peak of FeIV = O gradually disappeared due to the rapid self-decay of FeIV = O species in water. The new peak of FeIV = O could also be discerned in another FeIn–N–C/PMS and FeSb–N–C/PMS systems, but was undetectable in Fe–N–C/PMS and FeMd–N–C systems (Figs. S64–S68 and 5b), in alignment with scavenging results. The emerging peaks of FeIV = O exhibited a pronounced redshift with 18O-substituted water (H218O) as the solvent via the oxygen atom exchange, further highlighting the selective FeIV = O generation with p-block metal coordination (Fig. S69)13.

Fig. 5: Mechanistic insights into the selective FeIV = O generation.
Fig. 5: Mechanistic insights into the selective FeIV = O generation.
Full size image

a Time-dependent in-situ surface-enhanced Raman spectra of FeBi–N–C/PMS and b Fe–N–C/PMS system. The zoomed regions in this figure (a) denote the peak of PMS* and FeIV = O at 4 and 6 min (left) and vibrational peak of Bi–OH within 10 min (right). The dashed arrow indicates the shift of Bi–OH peak. The zoomed regions in this figure (b) denote the peak of PMS* at 4 and 6 min. c Operando Bi 4f, In 3d, and Sb 3d X-ray photoelectron spectroscopy (XPS) spectra of FeBi–N–C, FeIn–N–C, and FeSb–N–C before and after PMS activation. d Fe 2p XPS spectra of Fe–N–C, FeBi–N–C, FeIn–N–C, and FeSb–N–C after PMS activation. e Calculated Koutecky–Levich (K–L) plots of different systems from linear sweep voltammetry (LSV) curves. f Relationship between the electron transfer number of different reaction systems with steady-state concentrations of FeIV = O and g kper-site values. h In-situ solid electron paramagnetic resonance (EPR) spectra of different catalysts after PMS activation at 110 K. i Temperature-dependent magnetic susceptibilities of FeBi–N–C, FeIn–N–C, and FeSb–N–C after PMS activation. j Schematic illustration of electronic configuration of high-spin state FeIII.

Intriguingly, the FeBi–N–C/PMS system exhibited a noticeable peak at ~1080 cm−1 associated with metal–OH motifs55. Such a similar peak only appeared in the Bi–N–C/PMS system relative to the Fe–N–C/PMS system (Fig. S70), implying that the Bi atom serving as a Lewis acid site prefers to coordinate with the terminal hydroxyl oxygen to construct an electron-rich Bi–OH microenvironment. During the PMS activation, the vibrational peak of Bi–OH in the FeBi–N–C/PMS system was red-shifted to a lower wavenumber with the prolonged time but remained in the Bi–N–C/PMS system (Fig. 5a), which obviously indicated the decreased electron density and dynamic electron donation from Bi–OH to the adjacent Fe site. The same phenomenon also occurred in FeIn–N–C/PMS and FeSb–N–C/PMS systems due to the electron-donor properties of In–OH and Sb–OH analogs (Figs. S64 and S65). However, such metal–OH peaks were not observed in FeMd–N–C/PMS systems, which correspondingly produced no FeIV = O species due to the limited electron transfer (Figs. S66–S68). Therefore, it can be postulated that the dynamic electron supply surrounding single-atom Fe sites is the key to generating FeIV = O. The Mp–OH motifs as indispensable intermediates, derived from p-block metal coordination in FeMp–N–C, provide a new platform for the selective formation of FeIV = O by a sustained electron transfer process.

To confirm the hypothesis, operando X-ray photoelectron spectroscopy (XPS) was used to analyze the electron transfer between PMS and catalysts. As shown in Fig. 5c, the high-resolution Bi 4f, In 3d, and Sb 3d XPS spectra of FeBi–N–C, FeIn–N–C, and FeSb–N–C positively shifted toward higher binding energy upon the PMS addition, indicating the decreased electron density after PMS activation. In contrast, Mn 2p, Cu 2p, and Ni 2p XPS spectra of reacted FeMn–N–C, FeCu–N–C, and FeNi–N–C still retained the original oxidation state (Fig. S71). This further reveals the electron-rich nature of Mp–OH motifs and continuously dynamic electron transfer to single-atom Fe sites. In the Fe 2p XPS spectra, the new peaks at 711.3 and 712.9 eV suggested the formation of FeIII and FeIV = O species in FeMp–N–C/PMS systems, in sharp contrast to the abundance of FeII in Fe–N–C/PMS system (Fig. 5d)65. All the results strongly confirm that Mp–OH motifs serve as endogenous electron sources of single-atom Fe sites to trigger the FeIV = O generation via a two-electron transfer process (FeII→FeIII → FeIV). To determine the process, linear sweep voltammetry (LSV) measurements of all catalysts were conducted using a rotating disk electrode. By plotting the Koutecky–Levich (K–L) curves, the electron transfer number (n) of FeMp–N–C catalysts during PMS activation was calculated to be 1.91–2.02, following the two-electron transfer pathway, which was twice as many as that of Fe–N–C and FeMd–N–C references (Figs. 5e and S72). Furthermore, the steady-state concentration of FeIV = O and kper-site were demonstrated to positively correlate with calculated n values (Fig. 5f, g). These findings provide compelling evidence that p-block metal coordination in FeMp–N–C catalysts facilitates the electron transfer process during PMS activation, consequently achieving selective generation of FeIV = O. In contrast, the inferior electron transfer efficiency in Fe–N–C/PMS and FeMd–N–C/PMS systems may suppress the conversion from PMS* to FeIV = O, thus disfavoring the selective FeIV = O formation and PhOH degradation.

Apart from the enhanced electron transfer, the electronic configuration of single-atom metal sites (e.g., spin state) also plays a pivotal role in the generation of HVMO species such as FeIV = O, CoIV = O, and MnIV = O12,13,14. Driven by the fact, in-situ solid-state EPR spectra of different systems at 110 K were recorded to determine the spin state of single-atom Fe sites. Low-temperature EPR technology is a powerful approach to distinguish the spin state by the g-factor66. Before the PMS activation, all the samples except N–C showcased an isotropic signal at the g value of ~2.003, attributed to the unpaired electrons in the 3d orbitals of single-atom Fe sites (Fig. S73)67. Notably, the new characteristic signal at the g value of ~4.25 was detected after PMS activation with FeMp–N–C catalysts relative to others, indicating the formation of high-spin FeIII species (Fig. 5h)12. The formed high-spin intermediates were further verified by temperature-dependent magnetic susceptibility measurements. The effective magnetic moments (μeff) are calculated by μeff = \(\sqrt{8{\mbox{C}}}\,\)μB that obeys the Langevin theory, which is used to obtain the unpaired d electron number (nd) of Fe sites following the equation of μeff = nd(nd + 2). The calculated μeff values for PMS-activated FeBi–N–C, FeIn–N–C, and FeSb–N–C were 6.23 μB, 6.13 μB, and 5.69 μB, which corresponded to the nd values of about 5.31, 5.21, and 4.78 (Fig. 5i). In contrast, the nd values of Fe–N–C and other FeMd–N–C were only 0.82–1.17 after PMS activation (Fig. S74). Based on the calculated nd values, the electronic configuration of high-spin FeIII sites was illustrated in Fig. 5j. The single-electron occupancy in antibonding orbitals with the highest energy level is more inclined to penetrate the π* orbitals of oxygen and ensures the sustainable electron transfer that drives the generation of FeIV = O12,13,68. However, in the Fe–N–C/PMS and FeMd–N–C/PMS systems, only a single-electron in the low-spin Fe dz2 orbital limits the availability of unpaired electrons required for hybridization with the O 2p orbital of PMS and subsequent electron transfer.

To further decipher the selective generation mechanism of FeIV = O species, density functional theory (DFT) calculations were conducted to distinguish the PMS activation by Fe–N–C, FeMn–N–C, and FeBi–N–C (Fe–N–C and FeMn–N–C represent the M–N–C and FeMd–N–C models, respectively). The different adsorption configurations of PMS on FeBi–N–C were constructed, and the optimized structures were categorized into single-site and dual-site modes (Fig. S75). In single-site modes, the Fe atom is coordinated with either hydroxyl O or terminal O bonded to the S atom by the end-on Pauling-type model. In dual-site modes, Fe and Bi atoms are simultaneously coordinated with both hydroxyl O and terminal O bonded to the S atom by the side-on Yeager-type model. The calculated adsorption energy (Eads) demonstrates that Bi and Fe atoms are more spontaneously and exothermally coordinated with hydroxyl O and terminal O bonded to the S atom, respectively. By comparison, the two types of O serve as the preferred adsorption sites for Fe and Mn atoms in FeMn–N–C (Fig. S76). Notably, FeBi–N–C exhibits a more negative Eads value (−4.24 eV) than FeMn–N–C (−2.56 eV) and Fe–N–C (−2.40 eV), suggesting that Bi coordination promotes the PMS adsorption on single-atom Fe sites and triggers the sequential elementary reactions (Fig. 6a). Subsequently, the adsorbed PMS complexes are activated and undergo significant elongation of O–O bond from 1.32 Å to 1.56 Å, consequently resulting in the spontaneous cleavage (Fig. S77).

Fig. 6: Theoretical calculations of PMS activation.
Fig. 6: Theoretical calculations of PMS activation.
Full size image

a Adsorption energies of PMS on Fe–N–C, FeMn–N–C, and FeBi–N–C. b d-band center of Fe–N–C, FeMn–N–C, and FeBi–N–C before and after PMS adsorption. c Charge density difference plots of PMS adsorption on Fe–N–C, FeMn–N–C, and FeBi–N–C. d Projected density of state (PDOS) plots of FeBi–N–C before and after PMS adsorption and corresponding theoretical magnetic moments. e Free energy profiles of PMS activation on FeMn–N–C and FeBi–N–C for FeIV = O generation.

The projected density of states (PDOS) was analyzed to reveal the electronic configuration of Fe sites during PMS activation. The d-band center of FeBi–N–C is nearer to the Fermi level compared with Fe–N–C and FeMn–N–C, implying the centralized electron distribution and enhanced reactivity of FeBi–N–C with PMS (Fig. 6b)69. After PMS adsorption, the d-band center of FeBi–N–C shifts downward and is farthest from the Fermi level, indicating that Bi coordination dynamically modulates the electronic configuration of the Fe sites, facilitating electron transfer from FeBi–N–C to PMS. The electronic structure evolution of Fe sites was further confirmed by charge density difference and Bader charge calculation (Fig. 6c). After PMS adsorption on FeBi–N–C, the Bi atom preferentially bonds to the hydroxyl O to create an electron-rich microenvironment (4.15 e). The unprecedented coordination structure serves as an endogenous electron source for Fe sites, enabling more electron transfer to PMS (1.52 e) for potential formation of FeIV = O. In contrast, only 1.13 e and 0.8 e are transferred from the Fe–N–C and FeMn–N–C to PMS, disfavoring the FeIV = O generation. These results are in good agreement with prior experimental findings that Bi coordination in FeBi–N–C is conducive to the continuous and enhanced electron transfer to PMS. Additionally, theoretical magnetic moments of Fe–N–C, FeMn–N–C, and FeBi–N–C were also calculated during PMS activation. In the FeBi–N–C/PMS system, the magnetic moment of Fe sites is about 1.92 μB after PMS adsorption, which is far higher than that of the original FeBi–N–C (Fig. 6d). In comparison with FeBi–N–C, Fe–N–C and FeMn–N–C almost retain the low magnetic moment after PMS adsorption (Fig. S78). The above results confirm the electronic configuration transition of Fe sites from low-spin to high-spin state with Bi coordination, aligning with the experimental μeff values. The high-spin Fe sites coupled with sufficient electron transfer lay the foundation for the formation of reactive FeIV = O species.

The formation pathway of FeIV = O species was further unraveled by calculating the free energies of different reaction systems (Fig. 6e). In the Fe–N–C/PMS system, the O–O bond is moderately elongated after PMS adsorption, probably requiring more external energy to break the O–O bond (Fig. S79). As a result, it is thermodynamically unfavorable for the generation of FeIV = O in the Fe–N–C/PMS system by the traditional two-electron transfer process. Then, the formation energies of FeIV = O were compared in the FeMn–N–C/PMS and FeBi–N–C/PMS systems. Upon the PMS activation, FeMn–N–C-PMS* adsorbs the energy of 1.99 eV to produce the transition state (TS), which involves the successive O–O and O–H cleavage for FeIV = O generation. However, the energy barrier to be overcome is lowered by FeBi–N–C-PMS* (1.54 eV) to produce TS, which is the rate-determining step for the formation of FeIV = O via the S–O cleavage. The total energy released by the FeBi–N–C/PMS system (3.02 eV) is remarkably improved relative to that by the FeMn–N–C/PMS system (0.12 eV). All calculation results profoundly affirm that Bi coordination enhances thermodynamic feasibility for the formation of FeIV = O species by the enhanced electron transfer and lower reaction barrier.

Practical wastewater purification and toxicity assessment

The scalable and sustained wastewater decontamination was conducted through a continuous-flow device. The practical wastewater was collected from secondary biochemical effluent in a municipal sewage treatment plant (Fig. 7a). The column reactor was filled with polyurethane sponges that were loaded with different catalyst powders and immersed in aqueous solutions (Fig. 7b). The homemade laboratory-scale reactor was continuously fed with a mixed solution containing wastewater and PMS solution to trigger the oxidation reaction. In the long-term and continuous operation, the FeBi–N–C/PMS system achieved stable and efficient degradation performance for PhOH and chemical oxygen demand (COD) during the running period of 120 h. Notably, the removal efficiency of PhOH could be maintained above 90%, and the COD concentration in the effluent was reduced to 50 mg L−1 (Fig. 7c, d), which reached the first-level A criteria in “Discharge Standard of Pollutants for Municipal Wastewater Treatment Plant of China” (GB 18918–2002). The superior catalytic activity of FeBi–N–C/PMS was universally applicable to FeIn–N–C/PMS and FeSb–N–C/PMS systems (Fig. S80). Moreover, the Fe contents of FeMp–N–C catalysts showed only a slight decline after the long-term and continuous operation (Fig. S81). In terms of exceptional durability and stability, the FeMp–N–C catalysts with p-block metal coordination hold promising application prospects in complex wastewater scenarios.

Fig. 7: Practical wastewater purification.
Fig. 7: Practical wastewater purification.
Full size image

a Practical wastewater collected from a municipal sewage treatment plant and chemical properties. b Schematic diagram of the continuous-flow device. c The long-term removal of phenol and d chemical oxygen demand (COD) in practical wastewater by FeBi–N–C/PMS system under the continuous-flow operation. Reaction condition: [phenol] = 10 mg L−1, catalyst dosage = 0.1 g L−1, [PMS] = 0.2 g L−1, flow rate = 0.42 L h−1, T = 298 K, initial pH = 7.

The ecological toxicity of phenol and its degradation intermediate products was initially predicted using the Toxicity Estimation Software Tool (T.E.S.T) software by Quantitative Structure-Activity Relationship (QSAR) modeling (Figs. S82 and S83 and Table S8). By combining the four toxicity indicators, including acute toxicity, bioconcentration factor, developmental toxicity, and mutagenicity, most intermediate products exhibited significantly lower toxicity than the parent PhOH molecule, indicating the great potential for practical applicability. Subsequently, the actual toxicity of different systems was further assessed through the seed culture tests by calculating the germination rate and measuring the root and shoot lengths of wheat seeds (Fig. S84). In comparison with the untreated PhOH solution that inhibited seed growth, the treated PhOH solution by FeMp–N–C/PMS systems resulted in a superior germination rate and developed root and shoot, even comparable to those grown in distilled water (Fig. S85). The combination of minimal ecological risk and excellent catalytic activity renders the FeMp–N–C as state-of-the-art catalysts for future practical applications in water purification.

Discussion

In this study, an p-block metal coordination strategy was applied to dynamically modulate the electronic structure of single-atom Fe sites, thereby contributing to the selective FeIV = O generation in PMS activation. We demonstrate that the constructed FeMp–N–C catalysts (Mp: Bi, In, and Sb) with p-block metal coordination trigger the activation pathway transition from electron transfer process mediated by Fe–N–C-PMS complexes to FeIV = O species-dominated oxidation process, thus delivering significantly enhanced catalytic activity for PMS activation and subsequent phenol degradation. Combining the in-situ characterizations and theoretical calculations, we confirm that the key Mp–OH intermediates as endogenous electron sources of single-atom Fe sites enable the selective FeIV = O generation via enhanced electron transfer and high-spin-state electronic configuration of Fe sites. Additionally, the efficacy of FeMp–N–C catalysts for PMS activation is further demonstrated by continuous-flow degradation of phenol in practical wastewater. We believe that the strategy will bring a significant breakthrough in the selective FeIV = O formation for water purification and beyond.

Methods

Materials and chemicals

Unless otherwise specified, all reagents were of analytical grade and used as received without further treatment. The details are provided in Text S1.

Synthesis of catalysts

The synthesis of all diatomic catalysts was conducted by pyrolyzing the mixture of melamine, L-alanine, and different metal acetates with a molar ratio of 1:1 between these two metals. In a typical synthesis of FeBi–N–C, 15.5 mg of iron(II) acetate, 34.5 mg of bismuth(III) acetate, 2 × g of melamine, and 2 g of L-alanine were homogeneously ground by a ZrO2 mortar ball for 1 h. Subsequently, 10 mL mixed solution of anhydrous ethanol and hydrochloric acid was dropwise added, and the slurry was continuously ground until the ethanol completely evaporated. The resulting mixture was dried in an oven for 12 h and then subjected to ball milling for another 30 min. The obtained fine powder was transferred to a quartz crucible and calcined in a tubular furnace under an argon atmosphere. The temperature was first heated from 25 °C to 600 °C at a ramping rate of 2 °C/min and maintained for 2 h, then heated to 900 °C and held for 1 h. After naturally cooling to room temperature, the solid product was immersed in 2 M HCl at 80 °C for 12 h, followed by repeated washing with deionized water. Afterward, the precipitates were dried and then again calcined at 900 °C for 1 h in the tubular furnace. Finally, the black solid powder was collected and ground for further use. Other FeBi–N–C catalysts with different Bi and Fe loading were synthesized by changing the molar ratio of Fe to Bi as 1:0.5, 1:1.5, 1:2, 0.5:1, 1.5:1, and 2:1. The typical method was also used for the synthesis of other samples (Text S2).

Characterization methods

The morphologies of samples were observed by SEM (ZEISS Gemini 300, German) and high-resolution transmission electron microscopy (Talos F200X, Thermo Fisher Scientific) with an accelerating voltage of 200 kV. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were conducted on an FEI Themis Z equipped with double spherical aberration correctors operating at 300 kV. X-ray absorption fine structure (XAFS) spectra were recorded at the beamline 4B9A station of Beijing Synchrotron Radiation Facility (2.5 GeV, 250 mA). Temperature-dependent (2 K–300 K) magnetization was performed by a magnetic property measurement system (Quantum Design) at a magnetic field strength (H) of 1 kOe under zero-field-cooled and field-cooled procedures. LSV measurement was carried out in a standard three-electrode system (CHI 760E, Chenhua Instrument Co., China) using a rotating disk electrode at different rotation rates from 400 to 2500 rpm with the scan rate of 10 mV s−1. A glassy carbon working electrode, platinum wire counter electrode, and saturated Ag/AgCl reference electrode were inserted in an electrolytic cell containing 50 mM Na2SO4 electrolyte and 0.2 g L−1 PMS solution. To prepare the working electrode, 5 mg of the catalysts were added to 1 mL of water-ethanol solution with 5% Nafion solution and ultrasonically dispersed for 60 min to form a homogenous suspension. Then, 10 μL of mixed aliquots were dropped onto a polished glassy carbon disk electrode and dried at room temperature to obtain the catalyst-coated glassy carbon electrode. After the LSV scanning, all the current values were recorded when the potential ranged from 0 to 2.0 V in a positive sweep. Until the current steadily increased to fulfill the electron transfer reaction, the scanning potentials were selected for plotting the Koutecky–Levich (K–L) curves. Time-dependent in-situ surface-enhanced Raman spectroscopy was performed using a Raman spectroscopy instrument (LabRAM HR Evolution, Horiba Co., France) equipped with a sapphire window at a green laser emitting of 532 nm. Before measurement, a silicon wafer at the band of 520.7 cm−1 was used in the spectrometer calibration. The acquisition time for each Raman spectrum was set as 20 s, and multiple scans were accumulated for a favorable signal-to-noise ratio. In each measurement, the laser beam was refocused on the upper surface of the catalysts. The signal was collected after a sufficient time interval of 2 min and further enhanced by Au nanoparticles. Additional characterizations of catalysts and theoretical calculations are described in Text S3.

Catalytic activity evaluation and analyses

The degradation experiments were performed in a 100 mL batch reactor under continuous magnetic stirring at 25 ± 2 °C. 5 mg of catalyst was added into 50 mL of phenol solution (10 mg L−1) and stirred for 30 min to reach the adsorption-desorption equilibrium. The initial pH of the reaction solution was adjusted to 3–11 with diluted H2SO4 and NaOH. The degradation reaction was then initiated by adding a certain amount of PMS stock solution to obtain the desired concentration (0.2 g L−1). At fixed time intervals, 1 mL of aliquots was withdrawn with the syringe, filtered through a 0.22 μm polytetrafluoroethylene membrane, and immediately transferred to a vial containing 0.5 mL of sodium sulfite (0.1 M) to stop the reaction for further analysis. To evaluate the practical applicability of catalysts, the degradation of phenol and diverse organic pollutants (if needed) was also performed in specific aqueous solutions containing different inorganic anions and humic acid, or real waters (tap water, lake water, groundwater, and wastewater). In a cycling test, the used catalyst was gathered by centrifugation, washed with ethanol and deionized water several times, and dried at 80 °C for the next cycle. The long-term wastewater decontamination was conducted by a laboratory-scale continuous-flow device, and the influent of a homemade reactor was collected from the secondary biochemical effluent of a municipal sewage treatment plant in Hangzhou, China, mixed with a certain concentration of phenol. Unless otherwise specified, all the experiments were conducted in duplicates or triplicates, and the average values with standard deviations were presented. Detailed analyses, including the determination of diverse organic pollutants, the contribution of reactive oxygen species, the quantification of the steady-state concentration of FeIV = O, the calculation of the electron transfer number, and the toxicity assessment, are provided in Text S4.

Reactive species identification

Electron paramagnetic resonance (EPR) tests were performed to detect the generation of SO4•−/•OH and 1O2 with DMPO and TEMP as spin-trapping agents, respectively. Briefly, 1 mL of the filtered sample was added to a 160 mM DMPO or 160 mM TEMP solution. The mixed solution was transferred into a capillary tube and then quickly inserted into the cavity of the EPR spectrometer to collect the EPR signal. Scavenging experiments were further conducted to determine the reactive species generation and contribution to phenol degradation with TBA, MeOH, p-BQ, TEMP, and DMSO as specific scavengers for •OH, •OH/SO4•−, O2•−, 1O2, and FeIV = O, respectively. To avoid interference from scavengers that could lead to the misidentification of reactive species, the direct consumption of PMS by various scavengers was also estimated.

Toxicity assessment

The toxicity of phenol and its degradation intermediate products was initially predicted using the Toxicity Estimation Software Tool (T.E.S.T) software by QSAR modeling. The assessment mainly focused on indicators such as acute toxicity, bioconcentration factor, developmental toxicity, and mutagenicity. The acute toxicity was determined by calculating the LC50 (48 h) for Daphnia magna. Then, the actual toxicity was systematically evaluated through seed culture experiments. A hydroponic method was conducted to investigate the effects of different post-treatment solutions on the germination and growth of wheat. In detail, the uniformly plump wheat seeds were sterilized in a 3% hydrogen peroxide (H2O2) solution for 30 min and thoroughly washed with distilled water to remove residual H2O2. Then, 30 seeds were spread evenly on filter paper in a petri dish moistened with deionized water, untreated phenol solution, or treated phenol solution through the PMS-based catalytic oxidation system. After a 5-day cultivation at 25 °C, the germination rate was calculated by observing the germination number of wheat seeds. The root and shoot lengths of wheat seeds were measured by the ImageJ software.