The optimized Fenton-like activity of Fe single-atom sites by Fe atomic clusters–mediated electronic configuration modulation

Significance

The single-atom catalysts (SACs) have been investigated and recognized as promising alternatives for initializing peroxymonosulfate (PMS)-based advanced oxidation processes (AOPs). Of note, the reasonable property modulation may further drive its practical application. This work suggests that tuning the electronic structure of single-atom sites is of great importance to achieve superior PMS activation kinetics. The Fe-based single-atom catalyst with Fe atomic cluster (ACs) modulation exhibits superior performance than the pure atomically dispersed Fe catalysts. In addition, the visible light can contribute to the formation of electron-deficient Fe species, therefore further improving the reaction activity. This work provides insights into the electronic structure regulation of metal centers at the atomic level.

Abstract

The performance optimization of isolated atomically dispersed metal active sites is critical but challenging. Here, TiO₂@Fe species-N-C catalysts with Fe atomic clusters (ACs) and satellite Fe-N₄ active sites were fabricated to initiate peroxymonosulfate (PMS) oxidation reaction. The AC-induced charge redistribution of single atoms (SAs) was verified, thus strengthening the interaction between SAs and PMS. In detail, the incorporation of ACs optimized the HSO₅^- oxidation and SO₅^·− desorption steps, accelerating the reaction progress. As a result, the Vis/TiFeAS/PMS system rapidly eliminated 90.81% of 45 mg/L tetracycline (TC) in 10 min. The reaction process characterization suggested that PMS as an electron donor would transfer electron to Fe species in TiFeAS, generating ¹O₂. Subsequently, the h_VB⁺ can induce the generation of electron-deficient Fe species, promoting the reaction circulation. This work provides a strategy to construct catalysts with multiple atom assembly–enabled composite active sites for high-efficiency PMS-based advanced oxidation processes (AOPs).

Single-atom catalysts (SACs), featuring the maximum atom utilization efficiency and the tunable electronic structure, are highly efficient for peroxymonosulfate (PMS)-based advanced oxidation processes (AOPs), and therefore have recognized as promising alternatives to homogeneous catalysts (e.g., FeSO₄ or FeCl₂) in water purification (1, 2). For example, the Co-N₂₊₂ directed the PMS activation by a nonradical pathway, the constant of Co single atoms (SAs) was 10.55 and 6.27 times higher than Co nanoparticles (NPs) and Co atomic clusters (ACs) (3). Moreover, Fe SACs supported on N-doped C showed excellent catalytic activity toward bisphenol A (i.e., 97% degradation in 5 min) and the pyrrolic N-coordinated Fe(III) dominated the reaction (4). Besides, a high selectivity of 99.71% toward ¹O₂ generation was obtained using Cu-SA/MXene, showing outstanding catalytic activities and strong resistances to environmental interferences (5). On this basis, various strategies have been developed to further accelerate the PMS reaction kinetics, for example, coordination regulation (e.g., CoN₄ → CoN₂₊₂) (3), elevating active site density (e.g., high Fe loading of 11.2 wt%) (6), and heteroatom doping (e.g., Cu-N₄/C-B) (1). However, limited studies have focused on the modulation effect of ACs on SA sites in PMS-based AOP fields.

Fe-N-C is the most active component among all M-N-C catalysts for PMS activation field (7, 8). Fe-N-C catalysts include multiscale metal phase from SAs and ACs to NPs, depending on the fabrication method and metal content (9). Viewed from the thermodynamic theories, the single-metal atoms tend to form aggregates (i.e., ACs or NPs) due to the decreased surface free energy (10), and its stable preparation has been an academic foreland (11). However, the coupling between ACs/NPs and SAs, benefiting from the intrinsic accessible characteristics (i.e., the inclined aggregation of partial SAs to ACs or NPs), can effectively tune the electronic structure of active metal species, eventually increasing the reaction activity. For example, the synergistic function of Au-Cu alloy NPs and Cu-SAs in CO₂ photoreduction enhanced the adsorption activation of CO₂ and H₂O, simultaneously lowering the activation energy barrier (12). The synergistic function of Pd₁ and Pd_NPs in ketone and aldehyde hydrogenation was verified, which activated C=O group and promoted H₂ dissociation, respectively (13). The electron donating effect from the embedded NPs and nearby Cu-N_x to Co-N_x was observed, increasing the oxygen reduction reaction (ORR) catalytic performance (14). Particularly, the Cu SAs would evolve into Cu ACs during the electrochemical reduction of nitrate to ammonia, highlighting the important role of ACs in catalytic reaction (15). Therefore, the incorporation of ACs offers a substantial potential method to regulate electronic structure of active metal species. Theoretically, the metal ACs, potentially existed as electron-withdrawing metal⁰@metal^x entity (16), can possibly extract electrons from the nearby metal SAs. With this strategy, incorporating Fe ACs into the substrate is a potential method to deplete the electronic density of nearby Fe SAs, tuning the electronic structure of SA sites to promote PMS oxidation kinetics (1). From another perspective, the visible light, as a readily available and cost-effective energy source, can be used to promote the formation of electron-deficient Fe active sites as well (17–19).

Several advantages can be expected in the incorporation of Fe ACs: i) The formation of Fe ACs during high-temperature annealing process is spontaneous, enabling a favorable operability, ii) higher atomic utilization efficiency, and thereby lower cost of Fe ACs compared to Fe NPs, due to the potential similar regulatory effects of small particle size ACs and larger counterparts (9); iii) the composition and morphology (i.e., electronic configuration and atomic coordination) of Fe ACs can be easily manipulated upon varying the Fe integration ratio or annealing procedure compared to the multicomponent alloy ACs/NPs (20), endowing the catalysts with promising application potential. Herein, the Fe AC-functionalized dispersed Fe SAs on a TiO₂@NC substrate (TiNC) were fabricated for highly efficient visible light-assisted PMS oxidation reaction. The formation of Fe ACs can deplete the electronic density around Fe SA sites, endowing the Fe SA with higher oxidation state to drive the oxidation of PMS to ¹O₂. As a result, the Vis/(Fe ACs + Fe SAs)@TiNC/PMS system exhibited a rapid tetracycline (TC) elimination efficiency (i.e., 90.81 and 98.84% in 10 min with 40 and 20 mg/L TC) and PMS utilization efficiency (51.04% in 10 min). Experimental and theoretical analyses demonstrated that the Fe ACs induced charge redistribution of active metal centers (i.e., SAs and ACs) and TiNS substrate. The optimized charge configuration effectively regulated the reaction energy, enabling a favorable PMS adsorption, activation, and desorption. The photogenerated h_VB⁺ can promote the formation of electron-deficient Fe species through electron transfer, further intensifying the interaction between PMS and Fe sites. This study provides a strategy to construct catalysts with multiple atom assemblies-enabled composite active sites while unveiling the related mechanism for the high-efficiency PMS-based AOPs.

Results and Discussion

Key Model Selection and Morphological Characteristic Identification.

The preparation of the catalysts was referred to the two-step annealing method proposed by Hai et al. (11) with modification (SI Appendix, Scheme S1), namely, i) the transformation of NH₂-MIL-125 to TiNC (calcination), ii) Fe anchoring (mixing and adsorption), iii) low-temperature annealing (controllable anchoring), and iv) high-temperature annealing (ligands removal). Initially, the correlation between degradation performance and Fe doping amounts was determined (Fig. 1 and SI Appendix, Fig. S1). The calculated degradation rates were 53.08, 61.44, 72.78, 90.81, 63.83, and 77.13% in 10 min, corresponding to the Fe doping amounts of 0.1, 0.25, 0.5, 1, 1.5, and 2 wt%, respectively. Interestingly, the TiNC with 1% Fe doping (saturated adsorption) showed the optimal degradation rate (90.81%) compared to the TiNC with higher doping amounts (63.83 and 77.13% for 1.5 and 2%, respectively). In addition, negligible TC was removed via the TiNC adsorption, TiNC photocatalysis, and direct PMS oxidation. Therefore, the nonlinear correlation had been explored with selection of three models, namely, TiO₂@Fe SAs-N-C (TiFeSA, 0.5 wt%), TiO₂@Fe ACs/SAs-N-C (TiFeAS, 1 wt%), and TiO₂@Fe NPs-N-C (TiFeNP, 1.5 wt%).

Fig. 1.

The determination of correlation between degradation efficiency/rate and Fe doping amounts. (A) Plots of TC concentration versus time. (B) The plot of degradation rate and k value versus different catalytic systems. Reaction system: volume = 100 mL, [PMS]₀ = 0.5 g/L, [TC]₀ = 45 mg/L, [catalysts]₀ = 0.2 g/L, [pH]₀ = 3.5, illumination condition: 300 W Xe lamp and 10-cm distance.

The physicochemical characteristics of TiNC, TiFeSA, TiFeAS, and TiFeNP were first investigated using scanning electron microscopy (SEM), transmission electron microscopy (TEM), and atomic-resolution high-angle annular dark-field scanning TEM (HAADF-STEM) observation. The high-resolution TEM (HRTEM) analysis for TiNC identified the spherical structure with interlayer distance of 0.352, 0.3248, and 0.2487 nm, which matched the [1 0 1] plane of anatase and [1 1 0]/[1 0 1] plane of rutile, respectively, suggesting the transformation of Ti-O clusters to TiO₂ under calcination (SI Appendix, Fig. S2 A and B). As shown in SI Appendix, Fig. S4A, the SEM image of TiFeAS showed a nanometer-sized 3D polyhedron structure with average diameter of ~380.66 nm (SI Appendix, Fig. S5). The TEM image exhibited dispersive polyhedron structure as well, without visible Fe agglomerates (Fig. 2A). HAADF-STEM was performed to investigate the distribution of Fe species at the atomic scale (Fig. 2B). The coexistence of Fe SAs and Fe ACs was observed in the TiNC support. The magnified image in Fig. 2C clearly displayed that several Fe atoms (red circle) closely distributed around a Fe cluster (yellow circle), suggesting the coexistence of Fe ACs and satellite Fe SAs in TiFeAS. The short intersite distance was conducive to the rapid electron transfer (9). The element mapping further verified the uniform distribution of the hybrid sites on TiNC (Fig. 2 D–F and SI Appendix, Fig. S4 D–H). The size distribution of the clusters in TiFeAS was in the range of 0.62 to 3.12 nm, with average diameter of 1.69 nm (SI Appendix, Fig. S9). Notably, the monodispersed SAs, which was far away from the ACs, would initiate the catalytic process like regular SACs. The SEM observation of TiFeNP showed a similar structure (SI Appendix, Fig. S6A). However, the TEM observation revealed the presence of Fe NPs (SI Appendix, Fig. S10A), further evidenced by the element mapping (SI Appendix, Figs. S6 B and H and and S7) and HRTEM results (SI Appendix, Fig. S10B), showing agglomeration of Fe species and Fe [1 1 0] lattice plane, respectively. Of note, the oxidized outsize surface (Fe₂O₃) was observed, verified by the existing Fe₂O₃ lattice plane [1 0 4] (SI Appendix, Fig. S10B). For TiFeSA, uniformly dispersed single-atom Fe sites were observed, indicating the formation of Fe single-atom catalyst (SI Appendix, Figs. S3 and S8). It was preliminarily concluded that the catalytic activity was in the order of ACs + SAs > SAs > NPs. The structure characteristics were further verified by X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), and synchrotron X-ray absorption spectroscopy (XAS) analysis.

Fig. 2.

Morphology characterization of TiFeAS. (A) The TEM image of TiFeAS. (B) The HAADF-STEM image of TiFeAS and (C) its magnification. (D– F) The HAADF and element mapping images of TiFeAS.

Intrinsic Structural Characterization of TiFeSA, TiFeAS, and TiFeNP.

The morphological characterization results were further verified by XRD analysis. As shown in Fig. 3A, the TiNC was successfully fabricated after a calcination of the NH₂-MIL-125(Ti) for 5 h at 900 °C, evidenced by the diffraction mode of rutile and anatase phase TiO₂, accounting for 54.1 and 45.9%, respectively (SI Appendix, Table S2). The incorporation of Fe species with high-temperature pyrolysis further promoted the transformation of anatase to rutile phase TiO₂ (SI Appendix, Table S2). The graphitic C peaks (i.e., (002) and (101) at 23° and 44°) were not detected (21), due possibly to their overlaps with anatase phase at 25° and 44°. Similar to the previously reported Fe SA catalysts, no diffraction peaks of crystalline Fe species were identified in the XRD pattern of TiFeSA (21). Notably, the crystalline components were still absent in the XRD pattern of TiFeAS, manifesting high dispersion and small size of Fe ACs. However, a diffraction peak at 44.72° indexed to Fe [1 1 0] lattice plane was observed, verifying the formation of Fe NPs in the TiFeNP.

Fig. 3.

The (A) XRD, (B) Raman, (C) Fe 2p XPS, (D) N 1s XPS, (E) XANES, and (F) EXAFS analysis of TiFeSA, TiFeAS, and TiFeNP.

It is previously reported that the structural defects would be introduced into the C substrate upon doping N or other metal species (22, 23). Two prominent peaks positioned at 1,350 (D band) and 1,580 (G band) cm⁻¹ were related to the defective or disordered C system and in-plane vibrations of graphitic structures (Fig. 3B). Higher proportioned intensity between D and G bands (I_D/I_G) can reflect more abundant defects and disorders in C-based materials. The calculated I_D/I_G values were 0.93, 0.96, and 0.94 for TiFeSA, TiFeAS, and TiFeNP, respectively, thus demonstrating the formation of more defective C structures in TiFeAS, contributing to modulate surface and electronic structures (i.e., increasing electron exchange), thereby optimizing reactivity. In addition, it was demonstrated that the Fe doping amount can directly influence the structure of C support. The other two peaks at 142 and 606 cm⁻¹ were indexed to anatase TiO₂ and the peaks at 252 and 432 should be related to rutile TiO₂.

The XPS N 1s spectrum indicated the presence of pyridinic N, Fe–N coordination, graphitic N, and oxidized N, indicating the existence of Fe–N moieties in three catalysts (Fig. 3D). The Fe 2p spectrum of TiFeSA and TiFeAS showed the positively charged Fe species (709.48 and 712.72 eV for Fe²⁺ and Fe³⁺ in TiFeSA/709.39 and 712.83 eV for Fe²⁺ and Fe³⁺ in TiFeAS) without obvious zero-valent Fe (706.50 to 708.70 eV) (National Institute of Standards and Technology (NIST) XPS Database, Version 4.1), suggesting that the Fe atoms in ACs were coordinated by the C/N/O atoms located in the matrix (Fig. 3C). On the contrary, the peak located at 707.19 eV verified the generation of Fe NPs in TiFeNP. The accompanied Fe²⁺ and Fe³⁺ peaks suggested the oxidation of generated Fe⁰. The increased Fe³⁺ content in TiFeAS (37.22%) compared to that in TiFeSA (19.78%) in XPS analysis (Fig. 3C) and elevated Fe absorption peak of TiFeAS compared to that of TiFeSA in X-ray absorption near edge structure (XANES) analysis (Fig. 3E) supported that the ACs induced electron depletion of SAs. In addition, the binding energy of pyridinic N in TiFeAS (398.34 eV) was positively shifted compared to that in TiFeSA (398.65 eV) (Fig. 3D) and the positive shift of C-N binding energy in TiFeAS (i.e., 285.76 eV in TiFeSA to 285.48 eV in TiFeAS) indicated that the depleted electron around Fe species would transfer to N-doped C layer (SI Appendix, Fig. S16A). These results were further confirmed by density functional theory (DFT) calculation.

The fine structure of the Fe species was examined by XAS. Fig. 3E exhibits the Fe K-edge XANES spectra of TiFeSA, TiFeNS, and reference samples. The absorption threshold positions TiFeSA and TiFeAS were identical to that of phthalocyanine (FePc) and located between FePc and Fe₃O₄, suggesting that the average chemical valence of Fe species was in the range of 2 to 3, consistent with the XPS results (Fig. 3C). The Fourier-transformed extended X-ray absorption fine structure (EXAFS) of TiFeSA and TiFeAS both showed predominant peaks at 1.5 Å in R space, similarly to the Fe-N peak of FePc (Fig. 3F). However, a small peak indexed to Fe-Fe path at 2.3 Å in EXAFS spectra of TiFeAS was observed, but not TiFeSA, indicating the coexistence of ACs and SAs in TiFeAS and dominant SAs in TiFeSA. The FT-EXAFS spectra of TiFeAS and TiFeSA were well fitted using backscattering paths of Fe-N/O and Fe-Fe (R factor of 0.0157) and Fe-N/O (R factor of 0.0018), respectively (SI Appendix, Fig. S11 and Table S3). The coordination numbers of Fe-N/O and Fe-Fe in TiFeAS were approximately 5.37 and 0.57, respectively. For Fe-N/O coordination in TiFeSA, this value was 4.09 (SI Appendix, Table S3). EXAFS wavelet transforms (WT) plot, an effective method to distinguish backscattering atoms, showed a maximum intensity at 5.8 Å⁻¹ in k space that can be ascribed to the Fe-N/O in TiFeSA and TiFeAS (SI Appendix, Fig. S12 D and E), and a second maximum intensity at 6.7 Å⁻¹ that was assigned to Fe-Fe in TiFeAS (SI Appendix, Fig. S12E). The negative shift of k value in TiFeAS (6.7 Å⁻¹) compared to that in Fe foil (8 Å⁻¹) may be related to the different coordination numbers between bulk Fe (8) and Fe ACs (0.57) (SI Appendix, Fig. S12 A and E and Table S3). Of note, abundant Fe ACs were observed in the HAADF-STEM images (Fig. 2 B and C), however, accompanied by the less significant Fe-Fe peak intensity compared to Fe-N/O (Fig. 3F). These results may result from the partial oxidation of Fe ACs into FeO_x ACs by air. It had been reported that PtO_x ACs or small PtO_x NPs (<5 nm) can exhibit similar EXAFS spectra as Pt SAs on metal oxide supports due to the high degree of disorder of the small PtO_x particles (24). Similar results were observed in N-anchored Fe ACs and satellite Fe–N₄ sites on two-dimensional porous carbon (FeSA/FeAC−2DNPC) as well, showing the existence of Fe-O coordination on N-doped C support (9).

Overall, the existing forms of Fe species can be gradually tuned from SAs, ACs/SAs, and NPs by varying the ratio of FeCl₂·4H₂O in the precursor. Subsequently, the physicochemical characteristics of TiFeSA, TiFeAS, and TiFeNP were further characterized by Fourier transform infrared spectroscopy (FTIR), Brunauer–Emmet–Teller (BET), XPS, electron paramagnetic resonance (EPR), electrochemical impedance spectroscopy (EIS), UV–Vis diffuse reflectance spectroscopy (UV–Vis DRS), etc (viewed at SI Appendix).

The Identification of Catalytic Efficiency Differences Between TiFeSA, TiFeAS, and TiFeNP: Focusing on Lewis Acid Site Density and Reaction Energy.

The PMS molecule, as Lewis base, can interact with the Lewis acid sites in TiFeSA, TiFeAS, and TiFeNP, mainly Fe species (25). Therefore, determining the quantity of Lewis acid sites (i.e., authentic interaction site density) would be more meaningful than simply measuring the total metal amount or specific surface area. The pyridine-infrared spectroscopy (Py-IR) spectra were then collected for TiFeSA, TiFeAS, and TiFeNP (Fig. 4A and SI Appendix, Fig. S21A) The normalized spectra of three catalysts showed vibrational bands near 1450 and 1540 cm⁻¹, ascribed to Lewis acid sites and Brønsted acid sites, respectively, overall showing the dominated Lewis acid sites compared to Brønsted acid sites on the catalysts surface. Not surprisingly, the L-acid amount was in the order of TiFeSA (91.22), TiFeAS (75), and TiFeNP (71.63). The single-atom Fe catalyst (TiFeSA, 0.5 wt% Fe incorporation) with high atom efficiency and dispersibility showed maximum L-acid amount, however, displayed inferior catalytic activity compared to TiFeAS (1 wt% Fe incorporation) (Fig. 1). Therefore, the increment in catalytic activity should be not simply attributed to the increased metal loading amount. For clarification, the catalytic activity was normalized using the L-acid amount (Fig. 4B and SI Appendix, Fig. S21B), showing the intrinsic advantage of Fe ACs + SAs (TiFeAS) compared to Fe SAs (TiFeSA) and Fe NPs (TiFeNP). The in-depth enhancement mechanisms were explored using theoretical calculation, focusing on the geometric and electronic structures and reaction energy.

Fig. 4.

The activity analysis of TiFeSA, TiFeAS, and TiFeNP. (A) The Py-IR spectra of TiFeSA, TiFeAS, and TiFeNP recorded at 200 °C. (B) The normalized k average values using L-acid amounts. (C) The proposed models for TiFeSA (C1), TiFeAS (C2), and TiFeNP (C3) and Fe Barder charge analysis (− and + denoted e⁻ loss and gain, respectively). (D) The proposed reaction pathway. (E) The Fe d orbit DOS of FeN₄ in TiFeSA and TiFeAS. (F) Reaction energy diagrams for PMS oxidation on Fe-N₄, Fe₄N₆/*FeN₄, Fe[111], and *Fe₄N₆/FeN₄. (G) The correlational analysis between valence electron number and average relative energy.

To investigate the role of Fe ACs in promoting the activity of satellite Fe-SAs, DFT calculations were performed. Previous reports have demonstrated that the size effect of ACs was not obviously significant [e.g., (ORR, 1 and 13 Fe atoms) and electrochemical reduction of nitrate to ammonia (<10 nm)] (9, 15). Therefore, this study used Fe₄N₆ to simulate Fe ACs, due to the main emphasis on the regulatory effects of Fe ACs. As shown in Fig. 4C  C2, a model of Fe₄N₆ with a closely adjacent Fe-N₄ was built on graphene/rutile TiO₂ to mimic the hybrid active sites of TiFeAS. The sole existence of Fe-N₄ and Fe[111] lattice plane was established for TiFeSA and TiFeNP, respectively (Fig. 4C C1 and C3). The PMS oxidation pathway was hypothesized to procced through HSO₅ adsorption, HSO₅ oxidation to SO₅^· (Fig. 4D and SI Appendix, Fig. S22), and SO₅^· desorption. Specifically, the Fe species in TiFeSA, TiFeAS, and TiFeNP would adsorb O site in -SO₄ size of PMS, promoting the PMS oxidation to SO₅^·, simultaneously losing the H atom (6). The SO₅^· can rapidly self-reaction to generate S₂O₈², SO₄, and ¹O₂ (Eqs. 1–3), due to the high reaction rate (≈ 2 × 10⁸ M⁻¹ s⁻¹) and low activation energy (7.4 ± 2.4 kcal mol⁻¹) (6). This is further discussed in The Investigation on Degradation Mechanisms.

$${\mathit{HSO}}_5^{-}\to {\mathit{SO}}_5^{·-}+H^{+}+e^{-},$$ [1]

$${\mathit{SO}}_5^{·-}+{\mathit{SO}}_5^{·-}\to S_2{O}_8^{2-}+{{}^{1}O}_2^{},$$ [2]

$${\mathit{SO}}_5^{·-}+{\mathit{SO}}_5^{·-}\to 2S{O}_4^{2-}+{{}^{1}O}_2^{}.$$ [3]

When the AC was introduced, the Fe-N₄ showed higher oxidation state (i.e., 1.28 e⁻ loss) with lower d-band center (−1.1221 eV) compared to the sole Fe-N₄ site (i.e., 1.07 e⁻ loss and −0.9601 eV of d-band center) (Fig. 4 C and E and SI Appendix, Fig. S23 and Table S8). It was thus demonstrated that the incorporation of Fe AC can induce the charge redistribution around Fe SA site (i.e., ACs induced electron depletion of SAs). Of note, the Barder charge analysis showed that the Fe₄ would exist as electron-rich sites (average value: 0.43 e⁻ loss, SI Appendix, Fig. S23A) compared to FeN₄ sites (1.28 e⁻ loss, SI Appendix, Fig. S23A) after the model optimization (setting the initial Fe valence electron number as 8). Further, the electron around the Fe species would transfer to the adjacent N-doped C substrate, making the Fe-N₄ center more positive and intensifying the interaction with HSO₅. These results were in accordance with the XPS and XANES analysis (Fig. 3 C–E and SI Appendix, Fig. S16A). Therefore, the Fe₄N₆-modified Fe-N₄ (TiFeAS) exhibited maximum interaction intensity with PMS (SI Appendix, Fig. S28), possibly originating from the flexible reactant adsorption and products desorption, due to the lower d-band center position (farther away from Fermi energy level) and valence electron number. The previous reports had demonstrated that the adjacent Fe NPs on the NC matrix can provide electrons to the NC matrix, which can increase the Fermi level of the whole system and simultaneously downshift the corresponding d-band center, endowing the Fe₄/FeN₃ catalysts with optimized free energy (20). Similar results were reported for the Fe₄/FeN₄/C as well (26).

The reaction energy differences between TiFeSA, TiFeAS, and TiFeNP were calculated (Fig. 4F and SI Appendix, Table S9). The minimum energy change for SO₅^· desorption was observed for Fe₄N₆/*Fe-N₄ (0.12 eV) compared to Fe-N₄ (0.44 eV) and Fe[111] (0.83 eV) (Fig. 4F), indicating that the incorporation of ACs optimized the binding strength of SO₅^· products, thus accelerating the HSO₅ re-adsorption. Similarly, the Fe₄N₆/*Fe-N₄ exhibited decreased energy difference (0.92 eV) for key step HSO₅ → SO₅·⁻ + H⁺ compared to Fe-N₄ (1.13 eV), suggesting that the TiFeAS was more favorable for PMS oxidation in terms of the thermodynamics. However, the Fe[111] can more rapidly initialize this transformation reaction (0.84 eV), but limited by its highest desorption energy requirement (0.83 eV). Ideally, the catalysts with optimized free energy close to average value needed in the reaction steps would show optimum performance, as it will allow the lowest free energy in rate-determining step (7). Therefore, the TiFeAS exhibited minimum average relative energy (−0.13 eV) toward HSO₅ transformation compared to the TiFeSA (−0.70 eV) and TiFeNP (−1.03 eV), thus showing optimum catalytic efficiency. Although the position of d-band center of TiFeNP (−0.99679 eV) was between TiFeSA (0.9601 eV) and TiFeAS (−1.1221 eV) (SI Appendix, Table S8), maximum HSO₅ adsorption energy and SO₅^· desorption barrier were verified, which can be ascribed to the formation of dual coordination bonds. The O atoms in -SO₄ side and -OH side can be simultaneously captured by metal sites due to the closely connected metal-metal bond, enabling an intensified interaction compared to metal SAs (27). The stronger adsorption energy of PMS on Fe₂O₃ (110), FeO (110), and Fe (100) compared to Fe-N₄ was verified as well (28). Of note, the *Fe₄N₆/Fe-N₄ was predicted with inferior activity (i.e., significantly increased HSO₅ adsorption and SO₅^• desorption energy compared to FeN₄, Fig. 4F and SI Appendix, Table S9), suggesting that the Fe ACs would participate in the reaction as an activity booster. Similar results were observed in Fe₄N₆/FeN₄ catalyzed ORR reaction (9).

Furthermore, the average relative energy was found to be linearly related to the valence electron number (Fig. 4G). TiFeSA and TiFeAS with lower valence electron number (6.93 and 6.72, respectively) exhibited lower average relative energy (−0.26 and −0.05 eV, respectively). On the contrary, TiFeNP with the highest valence electron number (8.15) showed the highest average relative energy (−0.39 eV), and thus low reaction intensity with PMS. The *Fe₄N₆/Fe-N₄ in TiFeAS also displayed unsatisfactory average relative energy (−0.94 eV) due to the high average valence number (7.57) (SI Appendix, Fig. S23A).

These results highlighted the pivotal role of optimized electronic configuration in PMS oxidation reaction. The DFT calculation supported a fact that profiting from the positive effect of AC incorporation on optimizing PMS binding, transformation (PMS→SO₅^·), and desorption. TiFeAS showed outstanding photocatalytic activity toward ¹O₂ production, which was in good agreement with the experimental results.

The Investigation on Degradation Mechanisms.

Reactive species in the Vis/TiFeAS/PMS system.

Electron spin resonance (ESR) experiments were conducted to identify the reactive species in the Vis/TiFeAS/PMS system. A typical seven-line EPR signal of 5,5-dimethyl-1-pyrrolidone-N-oxyl (DMPOX) was observed (Fig. 5A). The formation of DMPOX is related to the 5, 5-dimethyl-1-pyrroline N-oxide (DMPO) oxidation by immense amount of ·OH suddenly generated, high-valent metals, or ¹O₂ (2). Therefore, contrast experiments were further performed to identify the generated active species responsible for the DMPOX generation. The MeOH addition did not significantly change the peak shape and intensity (SI Appendix, Fig. S24B), indicating that the ·OH or SO₄^· was not generated in the system. Correspondingly, the radical quenching experiments showed that methanol (MeOH) and tert-butyl alcohol (TBA) all had negligible effects on TC degradation, suggesting that ·OH and SO₄^· were not massively generated in the system (Fig. 5 D and E), which were further verified by the corresponding active species probing experiments. The SO₄^· and ·OH yields were 20 and 0.5 μM in 10 min, respectively (Fig. 5E and SI Appendix, Figs. S25 A and B and S26B). The participation of O₂^·- was easily excluded due to its limited generation (4 μM in 10 min) (Fig. 5E and SI Appendix, Fig. S25C).

Fig. 5.

EPR spectra of the different reaction systems with (A) DMPO in water, (B) 4-amion-2,2,6,6-tetramentylniperidine (TEMP) in water, and (C) TEMPO in acetonitrile (detecting h_VB⁺)/in water (detecting e_CB). (D) The quenching experiments with existence of MeOH, TBA, DMSO, FFA, ethylenediaminetetraacetic acid disodium salt (EDTA-2Na), and AgNO₃ in the Vis/TiFeAS/PMS system. (E) The active species yields in 10 min. (F) The comparison of the degradation performance between Vis/TiFeAS/TC, TiFeAS/PMS/TC, and Vis/TiFeAS/PMS/TC systems. (G) O1s XPS spectra of TiFeAS before and after the reaction. (H) The determination of the degradation efficiency in the atmosphere of N₂, O₂, or air. (I) Fe 2p XPS spectra of TiFeAS before and after the reaction. Reaction system: volume = 50 mL, [catalysts]₀ = 0.2 g/L, [PMS]₀ = 0.5 g/L, [TC]₀ = 20 mg/L, [pH]₀ = 3.5, illumination conditions: 300 W Xe lamp and 10-cm distance.

To verify the possible participation of high-valent Fe, the methyl phenyl sulfoxide (PMSO) was used as the probe, as it is generally accepted that PMSO can be oxidized to methyl phenyl sulfone (PMSO₂) through the Fe(IV)-mediated oxygen transfer pathway (29). However, contradictory results were obtained, showing maximum PMSO₂ generation in pure TiNC without Fe incorporation (SI Appendix, Fig. S27). Similarly, a very recent study reported that the PMSO was easily converted to PMSO₂ without metal ions, thus querying the rationality using PMSO as a high-valence metal probe (30). Therefore, the dimethyl sulfoxide (DMSO) was used as the effective scavenger for the high-valent Fe, showing nearly no effect on TC degradation, excluding the participation of high-valent Fe in the TC degradation. The limited generation of ·OH, potentially deriving from the self-decay of Fe(IV) = O (Fe(IV) = O → H₂O₂ → ·OH), also verified the limited generation of high-valent Fe (29). These results were comprehensible as the PMS would transfer electrons to Fe sites, thus alleviating the high-valent Fe generation (as discussed in in situ FTIR and Raman spectra). From this perspective, when only focusing and comparing the PMSO transformation rate between TiFeSA, TiFeAS, and TiFeNP (SI Appendix, Fig. S27), minimum high-valent Fe generation amount can be expected in the Vis/TiFeAS/PMS system compared to other two counterparts when comparing the PMSO₂ generation. The minimum generation amount of high-valent Fe in the Vis/TiFeAS/PMS system reasonably verified the maximum interaction between TiFeAS and PMS, leading to the more rapid PMS activation (SI Appendix, Fig. S28). Therefore, the interaction intensity between catalysts and PMS was in the order of TiFeAS/PMS > TiFeSA/PMS > TiFeNP/PMS. From another perspective, the high transformation rate for the TiFeAS/PMS system potentially suggested the critical role of high-valence Fe species as previously reported in Fe-based catalysts/PMS system (31). For Vis/TiFeAS/PMS, the high-valence Fe may be a transient species (i.e., rapidly interacting with PMS molecule), thus showing limited PMSO transformation efficiency, whereas, for the Vis/TiNC/PMS system, the role of h_VB⁺ in PMSO transformation may need further evaluation.

Of note, the furfuryl alcohol (FFA), as a specific probe detecting ¹O₂ (32), was used to qualitatively and quantificationally analyze the ¹O₂ generation. On the one hand, the FFA addition (1 mM) significantly inhibited the TC degradation, decreasing from 98 to 56% (Fig. 5D). On the other hand, the Vis/TiFeAS/PMS system produced 730 μM ¹O₂ in 10 min, much higher than the SO₄^·, ·OH, and O₂^· (SI Appendix, Figs. S25D and S26B), suggesting the dominant role of ¹O₂ in TC degradation. Correspondingly, typical triplet EPR peaks with intensity ratio of 1:1:1 were observed in the Vis/TiFeAS/PMS system (Fig. 5B), indicating ¹O₂ generation again. These results demonstrated that the ¹O₂ was dominant in the catalytic system and responsible for the DMPOX generation and TC degradation. The Vis/TiFeSA/PMS, Vis/TiFeAS/PMS, and Vis/TiFeNP/PMS systems all showed similar active species identification results (i.e., quenching experiments: SI Appendix, Fig. S29; probing experiments: SI Appendix, Figs. S25–S27; ESR: Fig. 5 A and B and SI Appendix, Fig. S24), however, with maximum production amount of ¹O₂ in the Vis/TiFeAS/PMS system (SI Appendix, Fig. S25D), indicating its superior PMS activation capacity. Particularly, it was previously reported that the interaction between Fe NPs and PMS would produce more radical species compared to ¹O₂ (6). Several reasons may account for the high ¹O₂ yield in the Vis/TiFeNP/PMS system: i) The existing h_VB⁺ induced electron transfer from Fe NPs to h_VB⁺, enabling the Fe NPs to be electron-deficient sites (i.e., Fe⁰@electron-deficient Fe) (SI Appendix, Fig. S10B) (17, 33), further initiating PMS oxidation and ii) the high-temperature pyrolysis in N₂ may transform partial Fe NPs to Fe SAs (34). Under this circumstance, the Fe NPs may function as reaction booster, rather than active sites (14). Similarly, dominant ¹O₂ generation was observed in the B,N-decorated carbocatalyst/PMS system, originating from the synergism between Fe NPs and B/N co-doped C matrix (35).

Generally, the ¹O₂ may be derived from i) stepwise reaction mediated by OVs (Eqs. 4 and 5), ii) O₂^·-dominated reaction pathway (Eqs. 6–8), and iii) h_VB⁺ or single-atom Fe-mediated reaction pathway (Eqs. 9–11) (33). In order to explore the origin of the ¹O₂, three models Vis/TiNC/PMS (h_VB⁺-PMS interaction model, approximating Vis/TiFeAS/PMS due to similar energy band structure, SI Appendix, Fig. S30), TiFeAS/PMS (Fe site-PMS interaction model), and Vis/TiFeAS/PMS (h_VB⁺ coupling with Fe site-PMS interaction model) were used. For pathway I, the OV concentration in the Vis/TiFeAS/PMS system was determined by XPS analysis before and after the reaction (Fig. 5G). Three peaks can be attributed to lattice oxygen (O²), OVs, and physical adsorbed oxygen (H₂O), respectively. However, the peak area of OVs increased from 21.01 to 37.31% and 36.28% after the reaction upon illumination (Vis/TiFeAS/PMS) and at dark (TiFeAS/PMS), respectively, thus excluding the involvement of pathway I. Of note, the OV concentration in the Vis/TiNC system without PMS addition decreased after the reaction (SI Appendix, Fig. S31), due possibly to the electron trapping (36). Therefore, the significantly increased OV concentration may associate closely with the reducing environment with PMS as an electron donor (as discussed in in situ FTIR and Raman spectra), which can function as adsorption and activation sites, further strengthening the contaminant elimination. In addition, the dominated role of pathway II can be easily excluded due to the limited O₂^· generation (<6 μM, SI Appendix, Fig. S25C). The N₂ and O₂ purging experiments further verified that the ¹O₂ was mainly derived from PMS decomposition (Fig. 5H). For pathway III, the ¹O₂ yield in the Vis/TiNC/PMS, Vis/TiFeAS/PMS, and TiFeAS/PMS systems was determined. The ¹O₂ generation in the Vis/TiNC/PMS system mainly derived from h_VB⁺–PMS interaction, due to its low reduction potential (i.e., low conduction band position 0.81 eV, SI Appendix, Fig. S30D), whereas ¹O₂ generation in the TiFeAS/PMS system should mainly come from the interaction between Fe SAs/ACs and PMS (37). The results showed that the ¹O₂ yield in the Vis/TiFeAS/PMS system increased by 53.08 and 34.31 %, respectively, compared to Vis/TiNC/PMS and TiFeAS/PMS, suggesting the dominant role of Fe SA/ACs compared to h_VB⁺ (SI Appendix, Fig. S34B). In fact, the h_VB⁺ was assumed to not only strengthen the interaction between Fe sites and PMS, but also interact with PMS to produce ¹O₂ directly, as discussed in the next section. Therefore, the ¹O₂ should mainly originate from pathway III.

Pathway I

$$OVs+{\mathit{HSO}}_5^{-}\to O^{\ast }+{\mathit{HSO}}_4^{-}\mathit{,}$$ [4]

$$O^{\ast }+{\mathit{HSO}}_5^{-}\to {}^1{O}_2^{}+{\mathit{HSO}}_4^{-}\mathit{,}$$ [5]

Pathway II

$$e^{-}+O_2\to O_2^{·-},$$ [6]

$$O_2^{·-}+h^{+}\to {}^1{O}_2,$$ [7]

$$O_2^{·-}+·OH\to {}^1{O}_2+O{H}^{-},$$ [8]

Pathway III

$${\mathit{HSO}}_5^{-}+{h_{\mathit{VB}}}^{+}\to {\mathit{SO}}_5^{·-}+H^{+},$$ [9]

$$\begin{array}{c}{\mathit{HSO}}_5^{-}+{Fe (high valence)-N}_4\to \hfill \\ {{Fe (low valence)-N}_4+SO}_5^{·-}+H^{+},\hfill \end{array}$$ [10]

$${\mathit{SO}}_5^{·-}+H_{2}O\to 1.5{}^1{O}_2+{HS{O}_4}^{-}.$$ [11]

Visible light-strengthened active species generation.

The superior photocatalytic/electrocatalytic characteristics were verified as previously mentioned (SI Appendix, Figs. S18–S20). Under these circumstances, one may argue that the optimized catalytic activity of TiFeAS may be originated from the optimized energy band structure; however, the catalytic activity was in the order of TiFeAS > TiFeSA > TiFeNP even at dark (shown as higher current) (SI Appendix, Fig. S32B), suggesting the intrinsic superiority of TiFeAS. This was further verified by the similar 2,2,6,6-tetramethylpiperidine 1-oxyl (TEMPO) intensity both in acetonitrile and water solution between TiFeSA, TiFeAS, and TiFeNP (indicating similar h_VB⁺ and e_CB contribution, SI Appendix, Fig. S33), thus excluding the possibility of visible light-induced degradation efficiency difference. Of note, the low conduction band positions of TiFeSA (0.08 eV), TiFeAS (0.77 eV), and TiFeNP (-0.06 eV) potentially explained the limited generation of SO₄^· and ·OH through PMS reduction, due to their weak electron donation capacity (SI Appendix, Fig. S20). Subsequently, the TiFeAS model was used to further reveal the strengthening role of light illumination. An enhancement in TC degradation was observed in Vis/TiFeAS/PMS compared to TiFeAS/PMS at dark (Fig. 5F), suggesting that the TC degradation can be enhanced upon light illumination, possibly originating from the accelerated ¹O₂ generation (SI Appendix, Fig. S34B). ESR tests (focusing on h_VB⁺ and e_CB) were performed to identify the visible-light contribution. In detail, the TEMPO, typical spin label molecule, was used to trap the h_VB⁺ and e_CB on the catalysts surface. The characteristic triplet TEMPO signals with intensity of 1:1:1 were observed in the ESR spectra. The signal intensity significantly decreased upon irradiation for 5 min in acetonitrile solution, due to the interaction between h_VB⁺ and TEMPO, transforming the latter into oxoammonium cations (TEMPO⁺) (Fig. 5C). In quenching experiment, the EDTA-Na₂ addition (interacting with h_VB⁺) significantly inhibited the TC degradation, decreasing from 98.84 to 40.44 % (Fig. 5D). These results verified the important contribution of h_VB⁺. However, the Vis/TiFeAS under irradiation showed unsatisfied degradation efficiency (Fig. 5F), thus indicating that the visible light may contribute to TC degradation both in direct (h_VB⁺ oxidation) and indirect (electron transfer) manners. This speculation was proved by XPS analysis, the Vis/TiFeAS system without PMS addition exhibited a negative shift of Fe binding energy (electron loss), demonstrating the existence of h_VB⁺-induced electron-deficient Fe (Figs. 5I  and 6C). Similar results were observed for single-atom Fe-modified macroporous/mesoporous TiO₂-SiO₂, which showed increased charge at Fe site upon introducing one positive charge approximating h_VB⁺ (17). After PMS addition, the Fe-binding energy in the Vis/TiFeAS/PMS system positively shifted compared to the TiFeAS/PMS system (Fig. 5I), further verifying the strengthened interaction between Fe sites and PMS under irradiation (higher ¹O₂ yield). Besides, limited e_CB^- contribution was verified (Fig. 5D and SI Appendix, Fig. S33B), evidenced by the negligible effect on TC degradation upon Ag⁺ addition and almost same intensity in TEMPO peaks in aqueous solution (the photogenerated electron would transform TEMPO to TEMPOH in aqueous solution) compared to the TEMPO control. To sum up, illumination (especially visible light), a wide and available energy form, can act as an electron transfer accelerator, significantly strengthening the catalytic performance.

Fig. 6.

(A) The ATR-FTIR and (B) Raman spectra of the Vis/TiFeAS/PMS system during the reaction collected at 2 min. (C) The proposed reaction mechanisms.

The reaction process characterization of the Vis/TiFeAS/PMS system.

To further gain insights into the electron-transfer process between TiFeAS and PMS, the Raman spectra were performed (Fig. 6A). Three distinct peaks in the PMS ([O₃S-O_I-O_II-H]⁻¹) solution were observed, located at 880, 980, and 1,059 cm⁻¹, corresponding to the stretching vibrations of O-O, SO₄², and SO₃, respectively. The O-O and SO₃ peaks (880 cm⁻¹) drastically decreased upon addition of TiFeAS within 2 min, indicating the rapid PMS activation. Of note, the characteristic vibration of SO₃ appeared significant blue shift (1,050 cm⁻¹). The blue shift typically derives from the decrease in electron density, suggesting the donation of electrons from PMS to catalysts (Eq. 10) (38). Subsequently, PMS is disposed to be SO₅^·-, which reacts with H₂O, generating ¹O₂ (Eq. 11).

The attenuated total reflection Fourier transformed infrared spectroscopy (ATR-FTIR) test was performed to identify the reaction mechanism (Fig. 6B). The jagged peaks existed in the spectra can be assigned to the H₂O adsorption bands. The spectra of TiFeAS, PMS, and TiFeAS+PMS in solution showed broad peaks with maximum intensity at 3,263, 3,212, and 3,225 cm⁻¹. The upward peaks originated from the correction of background peaks of ·OH in H₂O molecule (SI Appendix, Fig. S35). Compared to PMS alone, the addition of TiFeAS enabled a blue shift, potentially indicating that the PMS (HSO₅^-) would replace the ·OH bonded to the catalyst surface during the reaction (Eq. 12) (39).

$$Fe-O{H}^{-}+{\mathit{HSO}}_5^{-}\to Fe-\left(OH\right)OS{O}_3^{-}+{\mathit{OH}}^{-}\mathit{.}$$ [12]

The bands observed at 2,320 to 2,357 cm⁻¹ were assigned to CO₂ in air (40). The bands at 759/885/1,245, 1,060, and 1,100 cm⁻¹ in PMS spectra can be indexed to stretching vibration of S-O bond, stretching vibration of S=O bond, and symmetrical stretching vibration of sulfonate, respectively. The intensities of these peaks significantly decreased within 2 min due to the rapid decomposition of PMS. In addition, a significant blue shift was observed, showing 1,245 cm⁻¹ in PMS solution and 1,191 cm⁻¹ in the mixed suspension of TiFeAS and PMS, indicative of the electron transfer from PMS to TiFeAS, as evidenced by the XPS and Raman analysis (Figs. 5I  and 6A) (39).

Besides, the process of electron transfer was further studied by linear sweep voltammetry (LSV) analysis (SI Appendix, Fig. S36). A noticeable increase in current for the TiFeAS electrode with PMS addition was observed, indicating that the electron would transfer from PMS to TiFeAS (3). In addition, the current would further increase upon light illumination and TC addition, suggesting the favorable reaction between PMS, TC, and TiFeAS upon light illumination.

To sum up, the incorporation of ACs influenced the local electronic configuration of SAs, making the Fe-N₄ site more positively charged (i.e., intrinsic property). Simultaneously, the visible light-induced h_VB⁺ generation further contributed to fabricate electron-deficient Fe species (i.e., reaction process, Eq. 13). Under this circumstance, the PMS molecule would function as an electron donor, transferring electron to h_VB⁺ and electron-deficient Fe species, simultaneously promoting ¹O₂ generation and reducing high-valence Fe species to low-valence Fe species (Eqs. 2, 3, 9–11). As for e_CB, it would be trapped by OVs, not interacting with PMS, O₂, or contaminants, due to the low CB position (0.77 eV) (Fig. 6C). Of note, the investigation on the stability and practical application potential is provided in SI Appendix, Fig. S37.

$${Fe \left(low valence\right) -N}_4+h_{\mathit{VB}}^{+}\to Fe \left(high valence\right)-N_{4.}$$ [13]

Conclusions

In summary, the TiO₂@Fe species-N-C was successfully fabricated with Fe ACs and satellite Fe-N₄ composite active sites. The incorporation of Fe ACs induced charge redistribution, making the Fe-N₄ sites more positively charged. Accordingly, the increased electron density around C-N coordination (i.e., lower binding energy) was observed, suggesting the electron transfer to C substrate. The superior charge configuration optimized the reaction energy. Specifically, the incorporation of ACs decreased the needed energy for PMS oxidation and SO₅^· desorption, thus promoting product desorption and reactant readsorption. It was thus demonstrated that the superior catalytic performance of TiFeAS compared to TiFeSA and TiFeNP originated from the synergism of ACs and SAs, rather than the increased metal species content. This was further verified by the Py-IR spectra, showing maximum L-acid amount for TiFeSA, not TiFeAS. As a result, the Vis/TiFeAS/PMS system showed highest degradation efficiency (i.e., 90.81% of 45 mg/L TC in 10 min) compared to Vis/TiFeSA/PMS (72.78 %) and Vis/TiFeNP/PMS systems (63.83 %). Subsequently, the electron transfer was investigated using reaction process characterizations, manifesting as h_VB⁺-induced electron-deficient Fe species-mediated PMS oxidation, thus selectively generating ¹O₂. In addition, the h_VB⁺ can also interact with contaminant and PMS directly. The proposed strategy in this study can further strengthen the catalytic capacity of SACs, especially in PMS-based AOP fields.

Materials and Methods

Materials.

All chemicals, purchased from Shanghai Macklin Biochemical Co., Ltd, are analytically pure and directly used as received without further purification.

Preparation of Catalysts.

The whole preparation strategy was referred to ref. 11 with modification. The NH₂-MIL-125 was synthesized as the TiNC precursor. Typically, 2-aminoterephthalic acid (1,136.2 mg) was dissolved in mixed solvents (40 mL N,N-dimethylformamide + 10 mL MeOH), stirred at room temperature till complete dissolution. Subsequently, titanium tetraisopropanolate (1.5 mL) was added, followed by stirring for another 5 min. The resulting solution was subjected to hydrothermal reaction in a 100-mL autoclave at 150 °C for 24 h (41). The resulting yellow precipitate (NH₂-MIL-125) was collected by centrifugation, then washed with MeOH for three times, and dried 60 °C overnight. To prepare TiNC, the NH₂-MIL-125 (5 g) was mixed with potassium chloride (KCl, 100 g) and ball-milling for 20 min, subsequently heated to 900 °C (5 °C/min) for 5 h in nitrogen flow. The product was washed with deionized water and MeOH, then dried at 60 °C overnight.

To incorporate Fe species, TiNC (400 mg) was dispersed in a series of Fe gradient solutions (50 mL deionized water) with 1.42 (0.1 wt% Fe), 3.55 (0.25 wt% Fe), 7.1 (0.5 wt% Fe), 14.2 (1 wt% Fe), 21.3 (1.5 wt% Fe with 40 mg NaBH₄), and 28.4 (2 wt% Fe with 54 mg NaBH₄) mg FeCl₂·4H₂O addition, followed by 4-h ultrasound for homogeneous dispersion and 12-h shaking for Fe adsorption, which was further collected by centrifugation and dried at 60 °C under vacuum overnight.

Subsequently, the resulting Fe-incorporated TiNC powder was subjected to two-step annealing. For low-temperature annealing, the powder was heated to 300 °C (5 °C/min) for 5 h in nitrogen flow. After thorough washing using a water–MeOH mixture (1:1), the dried powders (60 °C) were subjected to high-temperature annealing at 550 °C (5 °C/min) for 5 h in nitrogen flow. Of note, the catalysts can be categorized into TiFeSA (0.1, 0.25, and 0.5%), TiFeAS (1%), and TiFeNP (1.5 and 2%). Unless otherwise specified, the TiFeSA, TiFeAS, and TiFeNP were referred to Fe doping amounts of 0.5, 1, and 1.5%.

The reasonability for selecting NH₂-MIL-125 as a precursor was discussed as follows: Photocatalysis mediated by visible-light represents promising potential in various chemical reactions (e.g., hydrogen evolution reaction, CO₂ reduction reaction (CO₂RR), elimination of contaminants, etc.) (23, 42–44). It was previously demonstrated that the NH₂-MIL-125, assembled from Ti(OC₄H₉)₄ and H₂BDC-NH₂, can be transformed to TiO₂@NC upon high-temperature calcination (23). Therefore, the NH₂-MIL-125 was chosen as the precursor rather than the frequently used ZIF-8 to realize the synergistic catalysis between photocatalysis and ACs/SAs-mediated catalysis (11).

Characterizations.

XAS measurements were carried out at the Australian Synchrotron in Melbourne using a set of liquid nitrogen cooled Si(111) monochromator crystals. The electron beam energy is 3.0 GeV. With the associated beamline optics (Si-coated collimating mirror and Rh-coated focusing mirror), the harmonic content of the incident X-ray beam was negligible. Data were collected using the fluorescence mode, and energy was calibrated using a Fe foil. The beam size is about 1 × 1mm. Note that a single XAS scan takes approximately 1 h.

Field emission SEM (FESEM, TESCAN MIRA LMS) and TEM (Talos F200X G2) were used to characterize catalyst surface morphology. HRTEM, Talos F200X G2 was used to characterize the lattice fringe. HAADF-STEM (JEM-ARM200F) was used to characterize the atomic structure of catalysts. XRD (Ulitma IV) was used to characterize the crystalline structure of catalysts. XPS (Thermo Scientific K-Alpha) was used to characterize the characteristics of surface elements of catalysts. FTIR (Thermo Scientific Nicolet iS20) and Raman spectroscopy (Horiba JY, LabRAM ARAMIS) were used to characterize the functional groups and molecular skeleton of catalysts, respectively. BET (ASAP 2460) was used to characterize the surface area and pore characteristics of catalysts. EPR (Bruker EMXplus-6/1) was used to characterize the surface defects of catalysts. UV–Vis DRS (Shimadzu UV-3600 iplus) was used to characterize the energy band structures of catalysts. The photoluminescence spectroscopy (PL, Edinburgh FLS1000) was used to characterize the separation performance of photogenerated carriers. Py-IR (Tensor 27) was used to characterize the Lewis acid amount.

The electrochemical workstation (CHI660e, Instrument, USA) was used to characterize the electrochemical properties of catalysts. A typical three-electrode configuration (catalyst-loaded carbon paper, Pt sheet, and Ag/AgCl (filled with saturated KCl) were used as working electrode, counter electrode, and reference electrode, respectively) was used. The distance between the working electrode and counter electrode was 1 cm. LSV was collected at a scan rate of 10 mV/s. Cyclic voltammetry (CV) was collected at a scan rate of 10 mV/s in an electrolyte containing 100 mM NaClO₄ and 5 mM K₃[Fe(CN)₆]. EIS was collected at open circuit voltage with the frequency range of 0.01 Hz–1 MHz in the same electrolyte as CV.

Experimental and Analytical Methods.

The photocatalytic activity of the catalysts was evaluated by the TC degradation upon irradiation. The photodegradation reaction was carried out in a 50-mL water-jacketed quartz photochemical reactor with attached water condenser (LX-300, Beijing Coolium Scientific Instruments Co., Ltd). The visible light source was obtained by a 300 W xenon lamp (CEL-PUV300-T8E, Beijing China Education Au-Light Technology Co., Ltd) with a cutoff filter (>420 nm). The distance between the lamp and reactor was 10 cm. During the reaction, a water-cooling system cooled the water-jacketed photochemical reactor to maintain the reaction temperature. Predetermined volumes of organic contaminants and PMS stock solutions were first added to the reactor. Subsequently, the reactions were initiated by adding the desired dosage of catalysts with simultaneous illumination. Stirring was maintained to keep the mixture in suspension during the reaction. The 1 mL solution was taken out every 1 min, filtered through a 0.22-μm poly tetra fluoroethylene (PTFE) filter (Jinlong syringe filter, Tianjin Keyilong Lab Equipment Co., Ltd) to remove the photocatalyst, and quenched with 0.5 mL NaSO₃ (0.5 M) before analysis.

The TC concentration was analyzed by high-performance liquid chromatography (HPLC, 1260 Infinity II, Agilent Technologies, Inc., USA) equipped with a reverse-phase Poroshell 120 EC-C18 (4.6 × 150 mm, 2.7 μm particle size) at 40 °C. The mobile phase was the mixture of acetonitrile, water, and formic acid (25:75:1) with a flowing rate of 1 mL/min. The detection wavelength was set at 280 nm for TC. The rhodamine b, bisphenol A, carbamazepine, ciprofloxacin, and norfloxacin were determined at 554, 280, 285, 278, and 272.5 nm, respectively, using a UV–vis spectrophotometer (T600, PERSEE, Beijing, China). For the cyclic experiment, the catalyst was recycled after each run of the experiment by centrifugation, ultrasound, and washed thoroughly with water. The active species was examined by ESR using DMPO (50 mM), TEMP (50 mM), and TEMPO (50 mM) as trapping agents of ·OH/SO₄^· (in deionized water)/O₂^· (in MeOH), ¹O₂ (in deionized water), and e⁻ (in deionized water)/h⁺ (in acetonitrile), respectively (45, 46). ATR-FTIR, Raman spectra, and LSV were used to monitor reaction process. An iodometric method was used to determine the PMS concentration in solution. In detail, 0.2 mL reaction solution was added to a mixture of 0.3 g KI, 0.06 g NaHCO₃, and 2.8 mL H₂O, and the mixture was then examined at 319 nm using a UV-Vis spectrophotometer (T600, PERSEE, Beijing, China) (47).

For the active species scavenging experiment, MeOH (0.1 M) (48), TBA (0.1 M) (48), FFA (1 mM) (47), DMSO (10 mM) (2), AgNO₃ (1 mM), and EDTA-2Na (1 mM) (49) were used to quench SO₄^·-/·OH, ·OH, O₂^·-, ¹O₂, high-valence Fe, e^-, and h⁺, respectively.

For the active species probing experiments, FFA (1 mM) (32), Ce(III) (0.1 mM) (50), terephthalic acid (TA, 1 mM) (2), nitrotetrazolium blue chloride (NBT, 0.05 mM) (2), and PMSO (0.2 mM) (2) were used to quantitatively determine ¹O₂, SO₄^·-, ·OH, O₂^·-, and high-valence Fe, respectively.

The cyclic experiments, Fe leaching experiments, coexisting ion (i.e., Cl^-, HCO₃^-, and SO₃^2-) experiments, pH (i.e., 3.3, 7.2, and 8.6) experiments, water matrix (i.e., tap water, sea water, lake water, and river water) experiments, and contaminants (i.e., TC (20 mg/L), rhodamine b (20 mg/L), bisphenol A (10 mg/L), carbamazepine (10 mg/L), ciprofloxacin (10 mg/L), and norfloxacin (10 mg/L)) experiments were performed to explore the practical application potential.

Theoretical Calculation.

All the calculations were performed in the framework of the density functional theory with the projector augmented plane-wave method, as implemented in the Vienna ab initio simulation package. The generalized gradient approximation proposed by Perdew, Burke, and Ernzerhof is selected for the exchange-correlation potential. The long range van der Waals interaction is described by the DFT-D3 approach. The cutoff energy for plane wave is set to 520 eV. The energy criterion is set to 10⁻⁶ eV in the iterative solution of the Kohn–Sham equation. A vacuum layer of 15 Å is added perpendicular to the sheet to avoid artificial interaction between periodic images. The Brillouin zone integration is performed using a 2 × 2 × 1 k-mesh. All the structures are relaxed until the residual forces on the atoms have declined to less than 0.03 eV/Å.

The pathway for the PMS transformation to ¹O₂ has not reached an agreement. Gao et al. (37) proposed that the PMS would transform to ¹O₂ through PMS → OH^* → O^* → ¹O₂, accompanied by the release of H₂SO₄. However, we failed to detect the massive presence of intermediate OH^* (SI Appendix, Fig. S26). In addition, the ATR-FTIR showed that the PMS (HSO₅) would replace the ·OH bonded to the catalyst surface during the reaction (Eq. 12) (39). Thus, the possibility of OH^* as intermediate was excluded in this study. Of note, the SO₅^· was considered as an important intermediate in PMS oxidation reaction, which was verified and accepted by several studies (i.e., Co and Fe SACs, verified by the DFT calculations) (6, 27, 28). Therefore, the pathway HSO₅ → SO₅^· → ¹O₂ was used in the DFT calculation.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.