Introduction

Developing highly efficient and durable heterogeneous catalysts is crucial for remediating growing water contamination1,2,3. In recent decades, single-atom catalysts (SACs) have been extensively explored for basic research and practical applications in peroxymonosulfate (PMS)-based advanced oxidation processes (AOPs) owing to numerous unique properties4,5,6. SACs based on diverse metal centers and supports have been designed and employed for PMS activation to generate reactive oxygen species (ROS)7,8. SACs for industrial catalytic applications are expected to possess excellent catalytic performance coupled with long-term operation lifetime. The intense interaction between reactive metal centers and reaction intermediate products renders metal sites of SACs easily occupied and deactivated, leading to declines in catalytic stability9,10. Moreover, although SACs can maximize the utilization of metal atoms to reduce metal consumption, the synthesis involves cumbersome and complex processes. The high production costs associated with expensive precursors or supports (e.g., metal-organic frameworks (MOFs) and MXene) and high energy consumption, as well as the potential environmental impact of toxic chemicals, contravene the environmental concept of sustainable development11,12. Thus, developing SACs with high activity in a green and economical manner persists in pivotal challenges to extend to real-world applications.

Carbon-based Fe SACs equipped with Fe−N−C moieties are favored by researchers due to the merits of high reactivity, low toxicity, and environmental friendliness13,14. Further upgrading the intrinsic activity of SACs and understanding molecular-scale reaction mechanisms are current spotlights, and most efforts are devoted to unfolding the catalytic potential of single-atom sites by tuning metal center and local structures in the first coordination shell (FCS)15,16,17. Typically, regulating configurations, coordination numbers, and coordinated-atom types in FCS are effective ways to manipulate the properties of SACs. However, although FCS directly bonded to metal atoms intensely governs SACs, studies have rarely focused on long-range structural components beyond FCS. Owing to multiple underlying non-metallic functionalized species (e.g., functional groups, heteroatom doping sites, and intrinsic carbon defects) that generally exist in carbon supports of carbon-based SACs, it is urgent to maneuver these long-range species precisely and interpret modulating effects on electronic structures and activity.

Intrinsic topological defects contribute significantly to the properties of carbon nanomaterials, enhancing the performance of critical catalytic reactions18,19,20. Concerning carbon-based SACs, defects probably alter charge and energy distributions of neighboring metal−N−C (M−N−C) sites and create abundant reactant-accessible active sites to potentiate the transfer of substrates and electrons, thus accelerating reaction kinetics21,22,23,24,25. In contrast, current reported SACs-based systems ignore the underlying contributions of intrinsic carbon defects, which are prevalent in carbon-based catalysts prepared by pyrolysis. Unfortunately, the high vacancy-formation energy in pristine graphene suggests enormous difficulties involving removing lattice carbon atoms26. Alternatively, eliminating external N atoms by thermal treatment permits readier access to vacancies in graphene, inducing the formation of topological defects27. However, before elucidating modulation mechanisms of intrinsic defects on M−N−C sites, a significant challenge is preserving the original FCS configurations of M−N−C sites intact following the construction of intrinsic defects. Regulating pyrolysis temperatures to produce intrinsic carbon defects may be accompanied by alterations in FCS microenvironments in M−N−C sites, such as transitions from M−N4 to M−N3 and M−N2 or introducing M−C coordination28,29. Hence, the controllable construction of microscopic defects in SACs without disturbing local FCS configurations of M−N−C sites is crucial for unraveling the synergistic optimization of overlooked defect engineering in SACs.

Herein, we design a superior Fenton-like catalyst by constructing abundant intrinsic defects within carbon planes anchored with Fe−N4 sites (D−FeN4−C). Low-cost sources and eco-friendly assembly strategy induce N elimination to create intrinsic defects without shifting the FCS coordination environments of Fe−N4 sites. Coupling intrinsic defects and Fe−N4 sites performed robust synergistic effects and significantly strengthened PMS activation. Density functional theory (DFT) calculations demonstrate that long-range interactions from intrinsic defects on neighboring Fe−N4 sites elevate the intrinsic reactivity of Fe−N4 sites and promote reactive FeN4=O generation. Furthermore, intrinsic defects can function as cleaner for adjacent Fe−N4 sites to realize long-term stability and environmental sustainability. This study provided promising perspectives for optimizing the intrinsic reactivity and revealing the structure-activity relationship of SACs by synergistic topological carbon defects.

Results

Fabrication of intrinsic carbon defects in Fe SAC

A facile thermal treating strategy based on temperature-dependent N elimination was developed for controlled constructions of intrinsic defects without altering FCS coordination structures of Fe−N−C sites, as illustrated in Fig. 1a. C-riched Fe-doped g-C3N4 was selected as precursors derived from thermal condensations of dicyandiamide (DCD) and 1-(2-cyanoethyl)-2-phenylimidazole (CEPI) molecules incorporated with trace amounts of Fe salts. CEPI contributed to the formation of C-riched Fe-g-C3N4, providing additional carbon sources during carbonation and improving the defect levels and the yields of products (Supplementary Figs. 1, 2). Further carbonized at higher pyrolysis temperatures to eliminate partial N atoms, and C atoms are involved in the rearrangement, creating substantial topological defects in carbon products anchored with isolated Fe−N4 sites to yield D−FeN4−C30. In contrast, relatively low-temperature thermal treatments obtained FeN4−C with low-defect levels, and corresponding metal-free samples were fabricated without adding Fe salts (defect-rich sample denoted as D−NC, low-defect sample denoted as NC) (Supplementary Fig. 3).

Fig. 1: Synthetic scheme and structural properties of as-fabricated SACs.
figure 1

a Schematic illustration of the synthetic procedure of catalysts. b–d HAADF and AC-HAADF-STEM images of FeN4−C. e Atom-overlapping Gaussian-function-fitting mapping of the square and intensity profile along X-Y in d. f–h HAADF and AC-HAADF-STEM images of D−FeN4−C. i Atom-overlapping Gaussian-function-fitting mapping of the square and intensity profile along X-Y in h. j Distribution of N species and high-resolution N 1s XPS spectra, k C 1s XPS spectra, l Raman spectra, and m EPR spectra of FeN4−C and D−FeN4−C. Source data are provided as a Source Data file.

Full size image

High-angle annular dark-field (HAADF) and aberration-corrected HAADF scanning transmission electron microscopy (AC-HAADF-STEM) images reveal Fe species evenly doped in carbon matrixes in both FeN4−C and D−FeN4−C, evidencing atomic dispersion of Fe species (Fig. 1b–d,f–h). The relevant elemental mapping diagrams (Supplementary Fig. 4) show that C, N, and Fe elements are uniformly distributed. Besides, the atom-overlapping Gaussian-function-fitting results from AC-HAADF-STEM displayed additional spatial information concerning the atomically dispersed Fe centers (Fig. 1e, i). X-ray diffraction (XRD) patterns of samples present only two broad peaks at 25.2° and 43.6°, correlating to the graphitized (002) and (101) planes, respectively, lacking Fe nanostructures or clusters (Supplementary Fig. 5). According to inductively coupled plasma optical emission spectrometry (ICP-OES) measurements, the mass loading of Fe was 0.72 wt% in FeN4−C and 0.69 wt% in D−FeN4−C (Supplementary Fig. 6).

No significant Fe 2p peak can be detected in X-ray photoelectron spectroscopy (XPS) due to the ultralow contents (Supplementary Fig. 7). High-resolution N 1s XPS spectra in Fig. 1j show that the dopant N atoms were eliminated more effectively as pyrolysis temperatures elevated (from FeN4−C to D−FeN4−C). The N signal intensity of D−FeN4−C is much weaker than that of FeN4−C, and the total N content reduces from 17.7% to 7.9%, suggesting substantial intrinsic carbon defects may be created by N atoms removal in D−FeN4−C. In C 1s spectra (Fig. 1k), the proportion of sp3 energy levels assigned to defective sp3 carbon atoms increased in D−FeN4−C, and the sp2/sp3 ratio dropped from 3.7 to 2.8 relative to FeN4−C, signifying more carbon defects in D−FeN4−C24,31. Structural variations are verified by Raman spectroscopy with the intensity of the D band sensitive to in-plane structural imperfection increased significantly as treatment temperatures increased (Fig. 1l), and the D band to G band intensity ratio (ID/IG) also elevated in D−FeN4−C31. Electron paramagnetic resonance (EPR) (Fig. 1m and Supplementary Fig. 8) further evaluated the defect levels and the substantially higher EPR signal intensities at g = 2.004 of defective samples (D−FeN4−C and D−NC) than those of FeN4−C and NC, demonstrating the high-concentrations topological defects23. Combined with the micro/mesoporous pore size distribution in Brunauer-Emmett-Teller (BET) analysis and the intact carbon planes of D−FeN4−C showed in atomic force microscope (AFM) images, indicated the limited holes and edge defects (Supplementary Figs. 9, 10). AC-TEM images showed direct evidence for the dominance of topological carbon defects created by eliminating several N atoms in the D−FeN4−C (Supplementary Fig. 11). These results collectively confirmed that removing N atoms by thermal treatment for fabricating intrinsic topological defects is practicable.

Atomic structure and chemical state analysis of Fe SACs

To reveal the local electronic structures and chemical coordination environments of FeN4−C and D−FeN4−C at atomic scales, X-ray absorption spectroscopy (XAS) was performed. The Fe K-edge X-ray absorption near-edge structure (XANES) spectra show that the energies of FeN4−C and D−FeN4−C are situated between Fe foil and Fe2O3 and almost overlap with Fe phthalocyanine (FePc) (Fig. 2a), implying partial oxidation states of Fe species. As for the Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS) spectra of FeN4−C and D−FeN4−C, only a prominent peak located at 1.45 Å assigned to Fe−N scattering paths is observed for FeN4−C and D−FeN4−C (Fig. 2b), and no Fe−Fe interaction peak emerges at 2.15 Å, confirms the atomically dispersed Fe species23. Further EXAFS fitting analysis on k-space and R-space yielded quantitative structural parameters surrounding central Fe atoms, Fe−N4 configuration is verified in both FeN4−C and D−FeN4−C (Fig. 2c, d; Supplementary Figs. 12, 13 and Supplementary Table 1). EXAFS wavelet transforms (WT) show that a clear WT intensity maximum is observed only at 4.1 Å−1 for FeN4−C and D−FeN4−C, ascribed to the Fe−N coordination and perfectly coincides with FePc (Fig. 2e), in line with the above results13.

Fig. 2: Atomic structural analysis of as-fabricated SACs.
figure 2

a XANES (inset: magnification of local areas) and b FT-EXAFS spectra in R-space of FeN4−C, D−FeN4−C and reference samples at Fe K-edge. c, d Corresponding FT-EXAFS fitting of Fe K-edge in R-space of FeN4−C and D−FeN4−C (inset: theoretical model structures). e WT spectra of Fe K-edge EXAFS for samples. f, g 57Fe Mössbauer spectra of FeN4−C and D−FeN4−C. h Different Fe species distribution in FeN4−C and D−FeN4−C. Source data are provided as a Source Data file.

Full size image

57Fe Mössbauer spectra measurements were conducted to analyze the components of Fe species in FeN4−C and D−FeN4−C. The Mössbauer spectrum for both FeN4−C and D−FeN4−C could be deconvoluted into doublet and no singlet and sextet, excluding the existence of zero-valence Fe phases, in accordance with XAS and AC-HAADF-STEM results (Fig. 2f, g). All the curves can be fitted with two components, assigned to doublet 1 (D1, high-spin FeIII−N4 species) and doublet 2 (D2, high-spin FeII−N4 species), respectively (fitting parameters are listed in Supplementary Table 2)32,33. Quantitative analyses reveal that the high-spin FeIII−N4 species was the dominant component in D−FeN4−C with a relative content of 72.7%, much higher than that of FeN4−C (57.6%). Simultaneously, the specific ratio of the two contents (D1/D2) in D−FeN4−C was elevated to 2.66 than FeN4−C of 1.36 (Fig. 2h). The above results indicated that the construction of neighboring intrinsic carbon defects modified the distribution of Fe species and resulted in more conversion of Fe−N4 sites to high-spin FeIII−N4 species in D−FeN4−C. Coupling with the XAS results and Mössbauer spectra analysis, atomically dispersed Fe species with different components extend throughout the carbon planes of FeN4−C and D−FeN4−C. Essentially, the homogeneity of the carbon substrate properties (except for intrinsic defects) and the FCS coordination structures for FeN4−C and D−FeN4−C are confirmed. All the above findings permit reasonable attempts to attribute the catalytic behavioral differences to the primary parametric variable of intrinsic defects.

Evaluation of Fenton-like reactivity

The Fenton-like reaction activity of the catalysts was evaluated through PMS activation to eliminate sulfisoxazole (SIZ), a typical antibiotic contaminant. Although adsorption alone could remove about 20% of SIZ, it has negligible effects on the high-speed catalytic degradation with incorporated PMS (Supplementary Fig. 14). Considerable oxygen-containing groups (C=O and C−O−C) and N-doped sites in NC and high-concentration intrinsic defects in D−NC as potential reactive sites failed to efficiently activate PMS suffered from the lack of reactive Fe−N4 sites (Fig. 3a). The catalytic activity of FeN4−C significantly enhanced after introducing isolated Fe−N4 sites and the observed rate constants (kobs) increased by 4.2-fold compared with referential NC (Fig. 3b). The poisoning tests conducted by potassium thiocyanate (KSCN) further support that Fe−N4 sites serve as authentic active sites for FeN4−C and D−FeN4−C (Supplementary Fig. 15). Concurrently, the kobs for D−FeN4−C demonstrated a tremendous 9.3-fold improvement over corresponding non-metallic defective D−NC. Notably, D−FeN4−C carrying both intrinsic defects and Fe−N4 sites exhibited remarkable efficiency with the kobs as high as 0.548 min−1, which is 3.1-fold superior to that of FeN4−C (0.178 min−1). These results suggest intense associations between the abundant intrinsic defects and atomic Fe−N4 sites, and modifying intrinsic defects enhanced the reactivity of Fe−N4 sites.

Fig. 3: Fenton-like catalytic performance, reactive intermediates identification and environmental applicability analysis.
figure 3

a SIZ degradation in different systems. b Comparison of kinetics of catalytic systems. c Comparison of TOF values with state-of-the-art Fenton-like catalysts. d Time-dependent fluorescence intensity obtained from 1O2 captured by SOSG. e PMSO consumption and PMSO2 generation in catalytic systems. f Influence of SIZ on PMSO conversion. g Extracted ion chromatography of PMS16O16O and PMS16O18O generated from PMS16O in the D−FeN4−C/PMS system in H218O. h MS2 spectrum of PMS16O16O and PMS16O18O. i In situ Raman spectra. j Decline ratio of kobs in reusability tests for SIZ removal. k The kobs of reaction systems under different water substrates. l Removal ratio of multiple model pollutants in the D−FeN4−C/PMS system. Reaction condition: [pollutant]0 = 20 μM, [PMS]0 = 0.2 mM, [catalyst] = 0.05 g L−1, [TBA] = [MeOH] = [FFA] = [Carotene] = 0.1–500 mM (if any), initial pH = 6.5. Error bars are standard error values of three tests (n = 3). Source data are provided as a Source Data file.

Full size image

It is worth noting that although higher-temperature pyrolysis contributed to removing more N species and producing more intrinsic defects, the agglomeration of Fe species renders the inferior reactivity to atomic FeN4−C and D−FeN4−C (Supplementary Fig. 16). D−FeN4−C possesses the lowest N content and also contains identical oxygen-containing groups as FeN4−C but exhibits the best reactivity, suggesting that the reactivity is primarily governed by defects level rather than other non-metallic functional species in carbon substrates. The turnover frequency (TOF) was employed to quantitatively evaluate the intrinsic reactivity versus previously reported Fenton-like catalysts. Figure 3c and Supplementary Table 3 show that D−FeN4−C exhibits excellent normalized intrinsic reactivity compared to the reported Fe-based catalysts and state-of-the-art SACs, as evidenced by a high TOF efficiency (0.62 min−1).

Identification of reactive species in Fenton-like reactions

The introduction of intrinsic defects can potentially impact catalytic reaction behaviors of Fe−N4 sites. Chemical quenching experiments were conducted to identify the prevailing ROS generated within these systems. Weak inhibition on SIZ degradation by tert-butanol (TBA), methanol (MeOH), and p-benzoquinone (p-BQ) as quenchers of radicals and signal-free EPR trapping imply negligible contributions of radicals (Supplementary Fig. 17)34. In contrast, furfuryl alcohol (FFA) and carotene significantly suppressed SIZ removal in the FeN4−C/PMS system, confirming the dominant role of 1O2, yet the contributions of 1O2 in the D−FeN4−C/PMS system evidently weakened. For further probe 1O2, EPR tests show that introducing Fe−N4 sites considerably boosts the 1O2 generation and signal intensities recorded in FeN4−C/PMS and D−FeN4−C/PMS systems proved equivalent (Supplementary Figs. 18, 19). Highly supported by Singlet Oxygen Sensor Green (SOSG) probe tests, with strengthened 1O2 generation induced by Fe−N4 sites and consistent 1O2 intensities in both systems (Fig. 3d)17. The lifetime of 1O2 in D2O (~ 68 μs) far exceeds that in H2O (~ 3.7 μs)35, and the kinetically accelerated oxidation in FeN4−C/PMS is higher than D−FeN4−C/PMS after transferring the 1O2-based systems into D2O (Supplementary Fig. 20). Collectively, it is speculated that creating substantial intrinsic defects in D−FeN4−C strengthens other nonradical mechanisms beyond 1O2 (electron transfer mechanisms or high-valent metal-oxo species).

Electron transfer pathways may be involved in multi-phase catalytic oxidation, especially for carbon materials36. Involving the pre-mixing of catalyst and PMS, SIZ degradation in catalytic systems is inhibited as the extended pre-mixing time (Supplementary Fig. 21). Combined with the depletion of PMS (Supplementary Fig. 22), PMS interacts with Fe−N4 sites, thus decomposed and continuously consumed in catalytic processes, and this behavior deviates from the characteristics of electron transfer mechanisms. Open circuit voltage tests further recorded the voltage variation in catalytic systems (Supplementary Fig. 23), and PMS injection resulted in an elevated potential at the surface of the catalyst-coated electrode. However, the subsequent addition of SIZ triggered no discernible voltage jump. The above results demonstrate the absence of electron transfer mechanisms in principal mechanisms for SIZ elimination in FeN4−C/PMS and D−FeN4−C/PMS systems.

Upon the above exclusion, it is reasonable to assume that the enhanced catalytic performance of D−FeN4−C stems from the contribution of high-valent FeN4=O species. Chemical probing of methyl phenyl sulfoxide (PMSO) was performed to identify FeN4=O species by selectively oxidizing PMSO to generate methyl phenyl sulfone (PMSO2) via oxygen atoms transfer (OAT)35,37,38. PMS alone cannot oxidize PMSO to produce PMSO2, and FeN4−C/PMS converted limited PMSO with a yield (ηΔPMSO2/ΔPMSO) of only 25.8% (Fig. 3e and Supplementary Fig. 24). As anticipated, considerable conversion of PMSO and high concentrations of PMSO2 were detected in the D−FeN4−C/PMS system. The ηΔPMSO2/ΔPMSO representing the selectivity of OAT reaction reached 84.3%, suggesting the crucial role of high-valent FeN4=O. Consequently, PMSO oxidization was significantly suppressed by adding competing SIZ to the PMSO conversion system (Fig. 3f and Supplementary Fig. 25), indicating that the FeN4=O intermediates are highly reactive to SIZ, highlighting the essential contribution of high-valent FeN4=O in SIZ removal. The Fe−O bond in FeN4=O can exchange O atoms with solvent water, so 18O isotopically labeled PMSO2 (C7H816O18O) was also determined. When the reaction system transferred to the H218O matrix, two peaks with m/z = 157.0327 and m/z = 159.0361 were detected (Fig. 3g), ascribed to PMS16O16O and PMS16O18O, suggesting that 18O isotope occurs OAT reaction among H218O, FeN4=O and PMS16O16O. Furthermore, a deviation of two mass units was observed in characteristic segments of PMSO2, demonstrating the presence of FeN4=O species (Fig. 3h).

In situ Raman was performed to observe the surface interactions during PMS activation (Fig. 3i). The PMS solution showed characteristic peaks of HSO5− at 882/1060 cm−1 and SO42− at 978 cm−1. After adding FeN4−C, the characteristic peaks attributed to HSO5− almost disappeared, indicating the rapid decomposition of PMS upon contact with the catalyst. A blueshift of the residual peak was also observed, suggesting that the chemical adsorption process transferred electrons from the catalyst to PMS and formed a few adsorption complexes. However, beyond the fast decomposition of PMS observed during the coupling of D−FeN4−C to PMS, a new peak at 838 cm−1 representing the surface-active Fe species (high-valent FeN4=O) was also identified37,39. Meanwhile, high-valent FeN4=O oxidized amino and aromatic rings and induced SIZ degradation, as abundant OAT products were found during SIZ degradation, consistent with the analysis of reaction mechanisms (Supplementary Fig. 26 and Supplementary Table 4). These results demonstrate that intrinsic defects strengthened the formation of FeN4=O species as complementary ROS with 1O2 catalyzed by D−FeN4−C.

Environmental applicability analysis

Subsequently, the catalytic stability of Fe SACs was studied. D−FeN4−C exhibits superior reusability with a slight decline in reactivity after five cycles and maintained complete SIZ removal within 10 min (Supplementary Fig. 27). In contrast, the recycling performance of FeN4−C is relatively inferior, with kobs declining by 74.7% in the fifth cycle, whereas D−FeN4−C dropped by only 39.6% (Fig. 3j). Pyridinic-N species distribution before and after reactions of D−FeN4−C varied less than that of FeN4−C, moderating the poisoning effects induced by the occupation of the pyridinic-N-coordinated Fe−N4 sites by intermediates (Supplementary Fig. 28). Moreover, the reduction of intrinsic defect levels during cycling is responsible for the progressive inactivation of D−FeN4−C (Supplementary Fig. 29). Interestingly, the catalytic reactivity of D−FeN4−C could recover entirely after regeneration associated with the reconstructed pyridinic-N species and defect levels by thermal treatment (Supplementary Figs. 30, 31). Accordingly, the proportion of Fe leaching from reaction systems in cycling is minimal and consistent under 25 μg L−1, far below the limit of 300 μg L−1 in standards for drinking water quality of China (GB 5749-2022) (Supplementary Fig. 32)39. Simultaneously, the catalytic systems were highly resistant to coexisting anions and organic compounds, and D−FeN4−C sustained high reactivity in tap and river water matrices with only 20.4% loss in kobs compared with 43.3% decreases in FeN4−C (Fig. 3k, see Supplementary Figs. 33, 34 with accompanying detailed discussion, parameters are listed in Supplementary Table 5), implying that intrinsic defects endow D−FeN4−C with superior stability and practicability. The D−FeN4−C/PMS system can also selectively oxidize multiple electron-donating model pollutants efficiently with various molecular structures and chemical properties, including phenols, antibiotics and dye-based substances, demonstrating the broad environmental applicability (Fig. 3l and Supplementary Fig. 35).

Mechanisms of defects-induced long-range interactions

DFT calculations were carried out to understand the critical role of intrinsic defects and the formation mechanism of high-valent FeN4=O species. Considering the intrinsic topological defects dominate in D−FeN4−C from characterizations, the 585-type topological defect structure (two five-membered rings sandwiching an eight-membered ring) rearranged from optimizing V2 configuration (vacancies missing two carbon atoms) was constructed neighboring the Fe−N4 site (Supplementary Fig. 36)26,27,40. Creating adjacent intrinsic defects at Fe−N4 sites significantly promotes PMS adsorption according to adsorption energy (Eads) calculations (Supplementary Fig. 37). Simultaneously, modifying topological defects converted from V1 and V3 vacancies via C-atom rearrangements shows consistent behavioral patterns of activity enhancement on adjacent Fe−N4 sites (Supplementary Fig. 38). Furthermore, the role of intrinsic defects and oxygen-containing groups as potential active sites deserves to be explored41,42. As evidenced by Eads calculations, PMS adsorption capacities of the 585 defects and oxygen-containing groups (C−O−C and C=O) (−1.56 eV to −0.95 eV) are much inferior to Fe−N4 sites (−2.83 eV to −2.19 eV), indicating the weak coupling to PMS and matching the experimental results (Fig. 4a, Supplementary Figs. 39, 40). Compared with FeN4−C, the doped-585-defect site in D−FeN4−C modifies the electronic structure of graphene, releasing more free electrons and accumulating into the Fe−N4 site to form electron-rich regions (Supplementary Fig. 41). Neighboring defects induced alterations in Fe−N bond lengths and charge redistribution of Fe−N4 sites, disrupting the original symmetric charge state (Supplementary Figs. 42, 43). High electron density facilitates catalyzing the adsorbed PMS and modulating the reactivity of reaction intermediates.

Fig. 4: Electronic structures and reaction pathways of FeN4−C and D−FeN4−C by inducing intrinsic carbon defects.
figure 4

a PMS adsorption energy on different sites. b Projected DOS of Fe 3d in FeN4−C and D−FeN4−C, Ef is at 0 eV. c Fe 3d projected DOS of FeN4−C and D−FeN4−C after adsorbing PMS, Ef is at 0 eV. d Correlation among the oxidation capacity, coupling strength to PMS, d-band structure and the effective magnetic moment (μeff) of FeN4−C and D−FeN4−C. e Charge density (rouge and green represent the dissipation and aggregation of electrons, respectively) and the S−O bond length (lS−O) in PMS adsorption configurations on FeN4−C and D−FeN4−C. f Reaction pathways and the energy profile diagram of PMS dissociation and FeN4=O generation on FeN4−C and D−FeN4−C (inset: corresponding intermediate structures). Source data are provided as a Source Data file.

Full size image

The long-range interactions of the 585 defects on the neighboring Fe−N4 sites remarkably narrow the d-band of Fe atoms in calculated density of electronic states (DOS), and the optimized electron distribution of the Fe atom elevates the d-band center from −1.11 eV for FeN4−C to −0.32 eV for D−FeN4−C (Fig. 4b). Following the Newns-Anderson-Grimley theory, the more centralized d orbitals tend to split to create bonding and antibonding orbitals when interacting with the adsorbate under energy level matching to firmer adsorption43. Furthermore, the disparity in Fe 3d DOS for FeN4−C and D−FeN4−C after PMS adsorption was also investigated. The adsorption strength of the intermediates can be inferred from the position of the highest electronic peak states (Ep). As illustrated in Fig. 4c, the energy level position of Ep after PMS adsorption by D−FeN4−C is closer to the Fermi level (Ef), which implies that the lower filling of the antibonding state and responds to stronger bonding to adsorbates44. To further elucidate the regulation of the spin states at Fe centers by intrinsic defects, we calculated the effective magnetic moments (μeff) based on the χ−1−T curves (Supplementary Fig. 44)45. Larger magnetic moments of D−FeN4−C (μeff = 2.31 μB, and μB is Bohr magneton) than FeN4−C (μeff = 1.94 μB) suggest that the intrinsic defect-modified D−FeN4−C possesses higher spin states. Meanwhile, consistent with experimental μeff values, the possible theoretical μeff values equally found the higher spin state of the Fe species in D−FeN4−C under the identical Fe spin numbers (Supplementary Table 6). The introduction of intrinsic defects still increased unpaired electron numbers, accompanied by enhancement of the electron density of the Fe−N4 sites, even though the Fe in both FeN4−C and D−FeN4−C retained high spin states (from Mössbauer spectroscopy analysis)46. Combined with the positive correlation among spin state, Eads and the oxidation capacities (Fig. 4d), incorporation defects induced the near-Fermi electronic states and high-spin states of Fe−N4 sites and decreases in orbital occupancy, thus boosting the bonding strength and reaction kinetics.

Electrochemical impedance spectroscopy (EIS) tests show that D−FeN4−C carrying intrinsic defects possesses less internal charge transfer resistance than FeN4−C (Supplementary Fig. 45)25. Meanwhile, stronger coupling and accelerated charge transfer between the Fe−N4 sites and PMS extended the S−O bond length, contributing to the generation of FeN4=O (Fig. 4e and Supplementary Fig. 46). The modification of neighboring intrinsic carbon defects on the free energy of the FeN4=O generation was further calculated (Fig. 4f). Firstly, the Fe−N4 site adsorbs the PMS molecule and releases energy. Then, the adsorbed PMS is activated, and the S−O bond undergoes stretching and breakage. Subsequently, HSO4− desorption forms FeN4=O intermediates. Inspiringly, upon modification with intrinsic defects, the transition state (TS) involved in the critical step of S−O bond cleavage to form FeN4=O is thermodynamically more favorable, and the energy barrier to be overcome is lowered by 0.39 eV, indicating the much higher conversion potential to form high-valent FeN4=O in D−FeN4−C. The total energy released by D−FeN4−C/PMS during reactions is 4.22 eV, which is higher than that of FeN4−C/PMS (3.01 eV). Previous literature has reported that high-spin FeIII−N4 species were thermodynamically and kinetically favorable for generating high-valent FeN4=O species upon binding to PMS47. Combined with Mössbauer spectroscopy analysis, the high spin state and high concentrations of high-spin FeIII−N4 species in D−FeN4−C resulted in relatively low formation energy of high-valent FeN4=O, thus accelerating the electron transfer from the organic pollutant molecules to the Fe−O π* orbitals47. However, the coupling of the defects to the Fe−N4 sites cannot thermodynamically enhance 1O2 generation, which is consistent with the mechanism analysis results (Supplementary Fig. 47). All these calculations essentially signify that intrinsic defect could regulate electronic structures of neighboring Fe−N4 sites and enhance intrinsic reactivity by strengthening charge transfer and reducing the reaction barrier to realizing highly efficient pollutant degradation.

Sustainability assessment and environmental applications

Life cycle assessment (LCA) was initiated to quantify the sustainability and environmental impacts of the D−FeN4−C/PMS process48. Concurrently, we adopted the Fe-based PMS catalytic systems constructed from the extensively reported Fe SACs derived from zeolitic imidazolate frameworks (ZIF-derived FeSA) and the conventional Fe3O4 catalysts as comparisons49,50. Specifically, we considered the entire water treatment process, including both catalyst preparation and pollutant degradation, with 18 relevant descriptors, and all data are at the lab scale (Supplementary Fig. 48 and Supplementary Tables 7, 8). As shown in Fig. 5a, it is apparent that the preparation process of Fe3O4 presents much lower environmental impacts than SACs for most descriptors, which is attributed to the more complicated synthesis conditions of SACs. As expected, the preparation of D−FeN4−C possesses evident superiority over ZIF-derived FeSA, indicating that the synthesis of D−FeN4−C in this work is more cost-effective and eco-friendly. Considering that the equipment can be operated under full load conditions and resources can be maximized when industrial production scale is achieved in the future, the environmental impact of SACs at the preparation stage will be highly reduced. It is noteworthy that the high-performance SACs provide distinct advantages versus Fe3O4 in the pollutant degradation stage. The D−FeN4−C system also has a lower environmental impact than the ZIF-derived FeSA in treating SIZ wastewater, owing to the less catalyst/PMS requirement and electricity consumption stemming from the higher intrinsic activity of D−FeN4−C. Lower raw material consumption also implies reduced transport costs from the place of production to the wastewater treatment plant in real-world applications, further reducing the environmental impact.

Fig. 5: Life cycle assessment (LCA) and potential in actual sewage treatment.
figure 5

a Relative environmental impact of environmentally relevant descriptors for treating 1 kg SIZ in Fenton-like processes with Fe3O4/PMS, ZIF-derived FeSA/PMS and D−FeN4−C/PMS systems. b Schematic diagram of the continuous-flow catalytic unit coupled with microreactor or enlarged microreactor. c, d Treatment of SIZ and actual hospital wastewater by the microreactor and the Fe leaching concentration. e, f Treatment of SIZ and actual hospital wastewater by the enlarged microreactor and the Fe leaching concentration. g Schematic illustration for the possible industrial hospital wastewater treatment with D−FeN4−C as the heterogeneous Fenton-like catalyst. Reaction condition for d and f: [TOC of hospital wastewater] = 49.70 mg L−1. Error bars are standard error values of three tests (n = 3). Source data are provided as a Source Data file.

Full size image

The excellent properties of D−FeN4−C inspire us to further extend the potential for practical applications at the device level. Two continuous-flow microreactors with different sizes were employed to evaluate the long-term stability by continuously treating the simulated and actual wastewater (Fig. 5b and Supplementary Figs. 49, 50). Catalysts were homogeneously fixed by cotton in the microreactor, while the enlarged microreactor was filled with catalysts-loaded polyurethane (PU) sponges. The microreactor system loaded with D−FeN4−C could totally remediate 6 L SIZ wastewater without cleaning at the flow rate of 2 mL min−1, but degradation rates gradually decreased at the terminal under the high flow rate of 5 mL min−1, indicating the overloading of pollutants (Fig. 5c). Conversely, FeN4−C showed inferior stability, with a significant decline in SIZ degradation under high flow rate conditions. In addition, D−FeN4−C also exhibited improved performance over FeN4−C in TOC removal for actual hospital wastewater (Fig. 5d). However, even at the low flow rate of 1 mL min−1, the D−FeN4−C system only achieved nearly 50% TOC removal after recollecting and re-pumping primary-treated effluent, indicating that the limited volume and catalyst loading of the microreactor cannot respond to higher-intensity pollutant loadings.

The enlarged microreactor was designed to broaden the practical application potential (see Supplementary Fig. 51 with accompanying detailed discussion for the characterization of catalysts-loaded PU sponges). SIZ removal in the device equipped with D−FeN4−C-loaded PU sponges was consistently maintained at 100% at high pumping speeds of 10 mL min−1 and even 20 mL min−1 in treating 60 L SIZ wastewater (Fig. 5e), demonstrating excellent catalytic capacity and stability of the scale-up reaction device. The D−FeN4−C-loaded PU sponges were also resistant to shock loading and remained uniformly distributed without significant loss in reactions attributed to the strong bonding strength. (Supplementary Fig. 52). Notably, upon application of the enlarged microreactor for actual hospital wastewater treatment, the D−FeN4−C/PMS system achieved high TOC removal after repeated operations. As shown in Fig. 5f and Supplementary Table 9, 72.2%–83.5% TOC removal and 67.1%–76.9% COD removal were achieved after continuous treatment, and the final effluent met the discharge standard for medical organizations in China (GB18466-2005) (Supplementary Fig. 53). Accordingly, the Fe leaching concentration during reaction systems was under 100 μg L−1 without severe secondary contamination, far below the limit of 300 μg L−1 in standards for drinking water quality in China (GB 5749-2022). Thus, the processing capacity of the catalytic systems can be scaled up simultaneously by expanding reactor capacity and catalyst loading, demonstrating the enormous potential for practical hospital wastewater treatment (Fig. 5g). Overall, the wastewater treatment process constructed by intrinsic carbon defects-modified D−FeN4−C and PMS offers a green and sustainable approach to sustainable water purification.

Discussion

In this study, intrinsic topological defects were successfully created in two-dimensional carbon planes loaded with Fe−N4 sites via an elaborate strategy of N species elimination. Integration of neighboring intrinsic carbon defects with Fe−N4 moieties significantly elevated the intrinsic reactivity and stability of Fe−N4 sites and boosted the PMS activation performance of D−FeN4−C for SIZ degradation. Mechanistic investigations revealed that introducing intrinsic carbon defects enhanced FeN4=O generation without influencing 1O2 production. DFT calculations confirmed that the constructed 585 defects strengthened the coupling of neighboring Fe−N4 sites with PMS and reduced reaction barriers for producing FeN4=O species by optimizing the Fe 3d orbitals through long-range interactions. Further life cycle assessment and device-level operation suggested that the D−FeN4−C-based reactor system achieved long-term stability with strong environmental viability during continuous simulated and actual wastewater treatment. These findings open a new avenue for developing highly efficient and durable SAC for sustainable water decontamination.

Methods

Synthetic strategy of Fe SACs

Under room temperature, 16 g DCD and 2 g CEPI were dissolved in 400 mL methanol. After stirring for 10 h, elevate the temperature to 80 °C to remove methanol by evaporation. The obtained solid was thoroughly ground and put into a crucible with a lid, then heated in a muffle furnace at 550 °C for 4 h with a heating rate of 2.5 °C min−1 to obtain CEPI-modified g-C3N4. The synthesis procedure of Fe-g-C3N4 was identical to that of g-C3N4 except for adding 24 mg or 15 mg Fe(NO3)3∙9H2O in the raw materials. Fe-g-C3N4 was carbonized at 800 °C (with 24 mg Fe(NO3)3∙9H2O added) and 900 °C (with 15 mg Fe(NO3)3∙9H2O added) for 2 h, with a heating rate of 5 °C min−1 under N2 atmosphere, to obtain FeN4−C and D−FeN4−C, respectively. Under the same conditions, NC and DNC were prepared as reference samples by pyrolyzing CEPI-modified g-C3N4 at 800 °C or 900 °C. FeNP−NC was obtained by calcining the precursors of D−FeN4−C at 1000 °C. D−FeN4−C was regenerated by pyrolysis at 900 °C for 1 h.

Catalytic activity evaluation for Fenton-like reaction

The degradation tests were conducted in a 250 mL reactor, and the reaction volume was 100 mL. Typically, 5 mg catalyst was dispersed into the SIZ solution. The reaction was triggered after adding PMS. 1 mL of liquid was withdrawn and quenched with 30 μL Na2S2O3 solution at each time interval, and the residual SIZ was measured using HPLC. All experiments are conducted at 25 °C. In the recycling test, the catalyst was collected after each run, fully washed with adequate amounts of ethanol and Milli-Q water, and dried before the next recycling. All experiments were conducted in duplicate or triplicate to ensure reproducibility.

Characterizations

Fe K-edge XAS analyzes were performed with Si(111) crystal monochromators at the BL14W Beamline at the Shanghai Synchrotron Radiation Facility (SSRF) (Shanghai, China). Samples were compressed into pellets, and the XAFS spectra were collected at the transmission mode. EXAFS data analysis was performed using the ATHENA and ARTEMIS programs in the Demeter computer package51. Concentrations of contaminants were analyzed by HPLC (Agilent, USA) equipped with a UV detector and an Eclipse XDB C18 (5 μm, 4.6 × 250 mm) (Supplementary Table 10). The Raman measurements were carried out on a Raman Spectrometer (HORIBA-800) with a 514 nm laser as the excitation source. Element content analysis was performed on an inductively coupled plasma mass spectrometer (ICP-OES730, Agilent). The atomic-resolution HAADF-STEM imaging and EDS and EELS analysis were performed on an FEI Titan Themis apparatus with an X-FEG electron gun and a DCOR aberration corrector operating at 300 kV. Atomic force microscopy (AFM) images were performed by an AFM microscope XE-100E in the air condition. Total organic carbon (TOC) analyzes were realized by TOC analyzer (Shimadzu TOC-L CPH). Chemical oxygen demand (COD) was measured by a COD rapid tester (5B-3C(V8), Lianhua Technology). The ammonia concentration (NH3-N) was measured by the HI-96715 ammonia analyzer (Hanna, Italy). Molybdenum antimony anti-spectrophotometry was used to determine total phosphorus (TP). BET-specific surface areas and pore size distribution of the catalysts were ascertained using a Micromeritics ASAP2020 Plus 2.00 system for N2 adsorption-desorption measurements. Electrochemical measurements were performed on a PARSTAT 4000 A potentiostat workstation (Princeton Applied Research) with a conventional three-electrode cell system. EPR spectra were obtained on Bruker EMX plus X-band CW EPR spectrometer (microwave frequency, 9.83 GHz; microwave power, 2.00 MW). Three-dimensional fluorescence spectroscopy (3DEEMs) was carried out using a fluorescence spectrophotometer (F-7000 FL Spectrophotometer, Hitachi, Japan). Zeta potentials of samples were measured by the Malvern Zetasizer Nano ZS90. Zero field cooled (ZFC) and field cooled (FC) measurements were executed by magnetic property measurement system (MPMS-3, Quantum Design) under H = 500 Oe for temperature-dependent (2–400 K) magnetization measurement. The 57Fe Mössbauer spectra of the samples were measured on a Topologic 500 A spectrometer using a 57Co:Rh source at 298 K and were calibrated with the α-iron foil at 298 K. The recorded Fe Mössbauer spectra were analyzed with the least-square methods.

Theoretical calculation

All DFT calculations were performed by the Vienna ab initio simulation package (VASP v5.4.1)52,53. The exchange-correlation interactions were described by using the Perdew-Burke-Ernzerhof in the generalized gradient approximation (GGA-PBE)54. The cut-off energy of plane wave was set at 400 eV for valence electrons55. The maximum Hellmann-Feynman force for each ionic optimization step is 0.01 eV/Å, and the convergence energy threshold of self-consistent calculations is 10−5 eV. The accuracy of geometry optimization converged to 0.02 eVÅ−1 and 10−5 eV. To ensure the accuracy of the calculation results on the simulated crystal surface, the vacuum layer was set as 20 Å. The Gaussian smearing width was set to 0.05 eV. Van der Waals interactions were accounted for via the zero damping DFT-D3 method. Due to the magnetism of transition metal elements, spin polarization was considered in all calculations56. We established initial magnetic moments for Fe atom based on the possible valence states and conducted single-point energy calculations.