Introduction

Fenton and Fenton-like reactions, as advanced oxidation processes (AOPs), are widely recognized for their effectiveness in removing environmental pollutants and degrading recalcitrant organic matter in water and soil1,2,3,4. The activity of Fenton or Fenton-like reactions arises from the activation of peroxides such as H2O2, peroxymonosulfate (PMS) and peroxydisulfate (PDS), which generate reactive oxygen species (ROS), including •OH, O2•−, SO4•−, SO5•−, and 1O25,6,7,8,9. Most ROS generated are radicals released from the surface of Fenton catalysts, initiating homogeneous oxidative reactions with the pollutants. In contrast, the high-valent metal-oxo species (e.g., FeIV = O10,11,12, CoIV = O13) display distinct behaviors in oxidation processes compared to radicals, as they typically engage in two-electron oxygen atom transfer (OAT) processes14, whereas radicals favor single-electron oxidation pathways. This pathway is particularly efficient in processes such as water oxidation15, epoxidation16, alkane hydroxylation17, and the oxidation of sulfides/phosphides18,19, decreasing energy barriers and avoiding side reactions. Additionally, since metal-oxo species are surface-bound, their reaction dynamics significantly differ from the homogeneous reactions of other ROS. However, compared to radicals or 1O2 with state-of-the-art techniques, such as electron paramagnetic resonance (EPR) spectroscopy, to characterize their appearance in the reactions, metal-oxo species remain somewhat elusive, possibly due to their surface-bound nature; these typical techniques applicable for homogeneous radicals are ineffective for the detection of metal-oxo. As a result, the existence of metal-oxo species and their roles in pollutant degradation, particularly in heterogeneous Fenton or Fenton-like systems, is still a subject of debate20.

Currently, approaches for identifying metal-oxo species can be broadly categorized into two main strategies: the use of molecular probes21,22,23,24 and direct detection via vibrational spectroscopy techniques, such as ATR-FTIR and resonance Raman Spectroscopy24,25,26,27. Each method presents inherent challenges in terms of sensitivity, selectivity, and the transient nature of reactive intermediates. In the first category, methyl phenyl sulfoxide (PMSO) is commonly employed, with metal-oxo species believed to oxidize PMSO directly to methyl phenyl sulfone (PMSO2) via oxygen-atom transfer. While the formation of PMSO2 is often cited as evidence for the presence of metal-oxo, this method is not entirely conclusive, as other oxidizing agents can also convert PMSO to PMSO2 under certain conditions28. Vibrational spectroscopy is another approach used to detect metal-oxo species, though it has limitations. For example, Mössbauer spectroscopy requires a very low temperature to detect Fe-oxo, which may not reflect actual reaction conditions. Techniques like infrared or Raman spectroscopy often exhibit low sensitivity to metal-oxo species. Furthermore, to date, the spectroscopic identification of metal-oxo in various systems—whether enzyme-like, photoelectrochemical, or Fenton—has been largely qualitative, aiming only to confirm the presence of the intermediate4,10,11,12,13,23. These spectroscopic techniques are typically steady-state, meaning they cannot provide real-time kinetics information about the formation and depletion of metal-oxo, and are incapable of tracking its life cycle. The long intervals between steady-state spectral collections also introduce baseline variations, obscuring the weak signals of metal-oxo.

To achieve time-dependent spectroscopy monitoring, it is crucial to precisely synchronize the reaction timeline with the spectra collection process. In other words, the starting points of the reaction and spectral data collection must align accurately. In the case of the Fenton reaction, which is initiated by mixing peroxide with a Fenton catalyst, employing the stopped-flow technique in conjunction with rapid-scan spectroscopy offers a potential solution. However, the classical stopped-flow technique is designed for homogeneous systems, facilitating the rapid mixing of two liquid-phase reactants and the synchronized monitoring of real-time changes in reagents, intermediates, and products via UV-visible spectroscopy29. To date, the application of the stopped-flow technique in heterogeneous systems alongside vibrational spectroscopy has not been reported.

Herein, we employ an adapted Stopped-Flow technique in conjunction with rapid-scan attenuated total reflection (ATR) Infrared spectroscopy, enabling in-situ monitoring of the heterogeneous Fenton reactions with a maximum time resolution of 60 ms. Supported by this technique, our investigation of the PMS-based Fenton-like reaction on a Fe-N4 single-atom catalyst (Fe-N-C) not only confirms the presence of surface FeIV = O but also provides the experimental identification of its lifetime on the timescale of seconds (3.96 s) through spectral kinetics analysis. Additionally, we uncover an oxygen exchange between FeIV = O and water molecules, further substantiating the notion of FeIV = O as a long-lived reactive species in aqueous systems. This innovative in-situ technique offers an approach to elucidate reaction pathways in Fenton reactions and can be applied to other reactions initiated by mixing reactants. Furthermore, based on the spectroscopic results, leveraging the specific reactivity of FeIV = O towards aqueous AsIII along with its extended lifetime, we achieve effective preferential removal of AsIII even in complex wastewater environments using the Fe-N-C/PMS system.

Results

Methodology of rapid-scan ATR-IR spectroscopy in conjunction with Stopped-Flow technique (SF-ATR-IR)

To enable real-time monitoring of the heterogeneous Fenton reaction, an approach coupling in situ IR technique with an adapted stopped-flow method is developed. The conventional stopped-flow technique is a rapid mixing method developed for homogeneous reactions. In this method, two or more reactants are quickly mixed in an observation chamber, and the flow is then abruptly stopped to allow the detector to monitor the reaction kinetics starting from the moment of mixing. However, this typical setup is not suitable for heterogeneous reactions. Here, we adopted the concept of “stopped-flow” but modified it to accommodate the nature of the heterogeneous Fenton-like reaction, where the reaction is initiated by the contact between the Fenton catalyst and PMS (or PDS).

To enable surface-specific monitoring, we first deposited the Fenton catalyst onto an ATR diamond crystal by evacuating the catalyst suspension (Fig. 1 and Supplementary Fig. 1), which serves as a role analogous to the observation chamber in the conventional setup. Given that the detection depth of ATR-IR is limited to just a few micrometers above the crystal surface, only reactions occurring on the deposited catalyst layer are detected. Therefore, the catalyst film thickness is controlled within 1–2 μm to ensure effective signal capture (Supplementary Fig. 2a, b). On the other hand, a digitally controlled stepping syringe pump was utilized to precisely inject PMS or PDS solutions. We then used a syringe pump to inject the PMS solution, abruptly stopping it at the catalyst-coated surface, ensuring that only the surface reaction is “observed”. This modified stopped-flow configuration allows us to selectively track the interfacial reaction between PMS and the catalyst in real-time.

Fig. 1: Schematic diagram of the instrument.
Fig. 1: Schematic diagram of the instrument.The alternative text for this image may have been generated using AI.
Full size image

The instrumental setup for rapid-scan SF-ATR-FTIR spectroscopy.

Prior to PMS/PDS injection, 10 μL of distilled water was added to the catalyst membrane. This serves a dual purpose: it acts as a buffer layer for the PMS/PDS drops and mitigates baseline variations in the IR spectrum, which could arise from the transition between a gas-solid interface and a water-solid interface when PMS/PDS is applied directly to the dry catalyst membrane. Upon injection of a drop of PMS/PDS (approximately 10 μL), the PMS/PDS molecules diffuse through the buffer layer, abruptly “stopping” at the catalyst membrane’s surface to initiate the Fenton-like reaction and ensure that only the surface reaction is “observed”.

The operation of the syringe pump is synchronized with the IR spectrometer, triggering the injection of PMS/PDS in response to the forward movement of the spectrometer’s mirror. Consequently, IR data acquisition begins simultaneously with the PMS/PDS injection from the syringe. The data collection occurs in rapid-scan mode, capturing IR spectra with a maximum time resolution of 60 ms per spectrum. Importantly, the ATR-IR technique only detects signals within 1–2 μm above the ATR crystal surface30; therefore, only when PMS/PDS molecules diffuse close to the catalyst membrane can they be observed in the IR spectra. This SF-ATR-FTIR technique thus provides valuable insights specifically focused on the surface reactions between PMS/PDS and Fenton catalysts.

Noteworthy, to observe the weak IR signals of Fe-oxo and peroxide intermediates, the operating conditions of ATR-IR are carefully optimized. The use of an oxidation-resistant diamond crystal prevents Fe-oxo species from reacting with the ATR crystal. Additionally, the use of a mid-band MCT detector, with a cutoff at 625 cm−1, enhances sensitivity for detecting the low-frequency Fe-oxo band. The entire IR spectroscopy sample chamber is continuously purged with dry argon to minimize atmospheric water vapor interference and improve the signal-to-noise ratio (Supplementary Fig. 2c). All of these optimizations, combined with the flexibility to introduce various catalysts and reagents, make this technique applicable to a wide range of reaction types—whether heterogeneous or homogeneous—including mixing-triggered reactions such as the Fenton reaction, enzyme catalysis, and catalytic organic synthesis.

Preparation and Characterization of Fe-N-C Catalyst

To validate the capability of our spectroscopy technique for dynamically identifying heterogeneous metal-oxo species, we conducted a study using Fe-N-C, a single-atom catalyst with Fe-N4 coordination. Single-atom Fe-N-C catalysts have attracted significant attention in energy- and environment-related applications and have recently emerged as promising Fenton catalysts31. These catalysts are typically fabricated by stabilizing monodispersed metal atoms on nitrogen-doped carbon substrates (M-N-C) through nitrogen-coordination (MNx)32. This approach allows for precise tuning of the MNx structure to control the types of reactive species generated during Fenton or Fenton-like processes. Of particular interest is the possible formation of metal-oxo species on M-N-C catalysts, as suggested by prior studies employing PMSO-probe experiments23,33,34. The formation of surface metal-oxo on single-atom catalysts would offer additional advantages: on conventional catalysts with adjacent metal sites, the synergistic interaction between two adjacent metal-oxo species makes them prone to rapid consumption through reaction with water, limiting their effective utilization in pollutant removal. In contrast, the isolated nature of metal sites in single-atom catalysts prevents such undesirable interactions, effectively shielding the metal-oxo species from premature decay. This extends the lifetime of metal-oxo species, allowing them to serve as more efficient oxidants for pollutant removal rather than being wasted in reactions with the aqueous matrix. Therefore, exploring the formation and dynamic kinetics of metal-oxo species on single-atom catalysts holds profound significance.

The typical single-atom Fe catalyst was synthesized via an adapted method from literature22,35: the Fe(Phen)3 precursors were first templated by nano-MgO, followed by calcination under argon gas and removal of templates by nitric acid leaching to obtain Fe-N-C (Fig. 2a). Scanning electron microscopy (SEM) images revealed that the resulting Fe-N-C featured a pleated and porous structure (Supplementary Fig. 3), while energy-dispersive X-ray spectroscopy (EDS) mapping confirmed the uniform distribution of Fe, N, and C elements (Fig. 2b). X-ray diffraction (XRD) analysis showed no diffraction peaks corresponding to crystalline or aggregated Fe (Supplementary Fig. 4). High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images further verified that Fe atoms were atomically dispersed across the carbon matrix without signs of clustering (Fig. 2c, d). This evidence underscored the effective synthesis of single-atom Fe catalysts with uniform dispersion.

Fig. 2: Structural and spectroscopic characterization of single-atom catalyst Fe-N-C.
Fig. 2: Structural and spectroscopic characterization of single-atom catalyst Fe-N-C.The alternative text for this image may have been generated using AI.
Full size image

a Schematic illustration of the chemical structure of Fe-N-C. b Energy-dispersive X-ray spectroscopy (EDS) mapping images. c Transmission electron microscopy (TEM) image and d High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of Fe-N-C. (Blue circles highlight a portion of Fe single atoms) e Normalized Fe K-edge X-ray absorption near-edge structure (XANES) spectra and f Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS) spectra of Fe-N-C, Fe2O3, FePc (iron phthalocyanine), FeO and Fe foil. Source data for Fig. 2e, f are provided as a Source Data file.

The chemical state of Fe was analyzed using X-ray photoelectron spectroscopy (XPS). The high-resolution Fe 2p spectrum revealed that Fe primarily exists in the FeII state, with a peak-area ratio of 61.6% (Supplementary Fig. 5). Additionally, the N 1 s spectrum displayed an Fe-N peak at 399.13 eV, confirming the presence of Fe in the N-coordinated Fe-Nx form36. The K-edge X-ray absorption near-edge structure (XANES) spectrum of Fe showed that the absorption edge of Fe-N-C closely aligns with that of standard iron phthalocyanine (FePc), indicating a predominant Fe2+ oxidation state, consistent with the XPS findings (Fig. 2e). Extended X-ray absorption fine structure (EXAFS) spectroscopy was utilized to probe the coordination environment of Fe in Fe-N-C. The spectrum exhibited a prominent peak at ~1.5 Å, corresponding to the Fe-N first coordination shell. Notably, the absence of Fe-Fe peaks at ~2.2 Å (metallic Fe) or ~2.7 Å (Fe-Fe in FeOx) further validated the atomic dispersion of Fe within the Fe-N-C matrix37 (Fig. 2f). The EXAFS wavelet transformation of Fe-N-C displayed a maximum intensity at ~4.0 Å−1, associated with the Fe-N interaction (Supplementary Fig. 6), confirming the atomically dispersed nature of Fe without detectable crystalline metallic phases. Finally, curve-fitting analysis of the EXAFS data indicated an Fe-N4 coordination number of approximately 3.8 (Supplementary Fig. 7, Supplementary Table 1), affirming the presence of the Fe-N4 structure.

The dynamic spectroscopic identification of FeIV = O in the Fenton-like reaction between Fe-N-C and PMS

The active species generated during the Fenton-like reaction between Fe-N-C and PMS was monitored in situ using SF-ATR-FTIR. As discussed above and shown in Fig. 3a, due to the detecting scope of ATR-IR that is limited to the near-surface region of the ATR crystal (or deposited catalyst membrane), there is an inherent delay between the injection of PMS and the observation of PMS or its reaction intermediates in the IR spectra, arising from the time required for PMS to diffuse from the syringe pump to the ATR crystal. To estimate this diffusion duration, we first conducted a controlled experiment on a bare diamond ATR crystal.

Fig. 3: PMS diffusion and reaction dynamics on bare and Fe-N-C-loaded ATR crystals.
Fig. 3: PMS diffusion and reaction dynamics on bare and Fe-N-C-loaded ATR crystals.The alternative text for this image may have been generated using AI.
Full size image

a Schematic illustration of PMS diffusion and reaction in the ATR-IR chamber. b Time-dependent IR spectra collected on a bare diamond ATR crystal, showing PMS diffusion without reaction. c Time-dependent IR spectra collected on an Fe-N-C loaded ATR crystal, capturing both PMS diffusion and reaction. The right panel highlights the FeIV = O band in an enlarged IR region. d Kinetics of HSO5− (1250 cm−1), HSO4− (1193 cm−1) and SO42− (1100 cm−1) growth derived from time-dependent spectra. Solid symbols represent the bare diamond case, while empty symbols correspond to the Fe-N-C-loaded diamond. The inset zooms in on the time region from 0 to 2 s. e Kinetics for the growth and decay of HSO5− (1250 cm−1) and FeIV = O (833 cm−1) on the Fe-N-C loaded diamond. f Enlarged view of the time region from 228 to 236 s for more specific details of the degradation. The error bars in Fig. 3d are based on data from at least three parallel experiments. Source data for Fig. 3b–f are provided as a Source Data file.

Noteworthy, the commercially available PMS is not solely a purified sample of potassium hydrogen persulfate (KHSO5); it also contains redox-inert KHSO4 and K2SO4. Therefore, we need to identify characteristic IR bands for each component to separately estimate their near-surface concentrations. These characteristic bands should be distinct and not overlap with those of other components. As shown in Supplementary Fig. 8a, the standard IR spectrum of the PMS mixture reveals several distinct bands. Notably, the characteristic IR bands for persulfates appear at 1250 and 1060 cm−1, corresponding to the S-O (or S = O) vibrations of the SO3 moiety, as well as the peroxyl (O-O) bond at 880 cm−1. For the other two components, KHSO4 exhibits IR bands at 1193 and 1050 cm−1, while K2SO4 shows a band at 1100 cm−1. To avoid interference among the components, we used the intensity of the 1250, 1193 and 1100 cm−1 bands to represent HSO5−, HSO4− and SO42−, respectively.

On the bare diamond ATR crystal, as illustrated in Fig. 3b, the IR bands corresponding to HSO5−, HSO4− and SO42−, became detectable 0.6 s after syringe injection. The kinetics of all PMS components showed a similar pattern (Fig. 3d and Supplementary Fig. 8b, solid symbols): their intensities sharply increased until stabilizing around 18.0 s, indicating that PMS diffusion had reached equilibrium. However, when Fe-N-C was deposited on the diamond crystal, as shown in Fig. 3c and d (empty symbols), the observation delay for the HSO5− band at 1250 cm−1 extended by an additional 0.5 s. Following this delay, the increase in intensity was slower. Notably, after peaking at 12.0 s, the intensity began to decline and was completely depleted at approximately 100.0 s. This behavior suggested that HSO5− was actively reacting with the Fe-N-C catalyst, with the apparent intensity of the 1250 cm−1 band reflecting the difference between HSO5− that diffused to the surface and that consumed in the reaction. In contrast, the intensities of HSO4− and SO42− exhibited a monotonic increase, reaching higher equilibrium values than on the bare diamond. This accelerated increase was attributed to the generation of HSO4− and SO42− from the reaction between HSO5⁻ and the Fe-N-C catalyst.

Significantly, almost concurrently with the appearance of the HSO5− band, a new band at 833 cm−1 emerged (Fig. 3c), closely matching the frequency of surface FeIV = O reported in the literature38. As shown in the kinetics (Fig. 3e, f), the IR bands of HSO5− (1250 cm−1) began to decay at approximately 12 s, marking the point where the consumption of near-surface HSO5− outpaced its supplement from the bulk solution. Complete depletion of the 1250 cm−1 band was observed at 230.16 s, suggesting the exhaustion of HSO5− in both the near-surface region and the bulk system. In contrast, the 833 cm−1 band exhibited a delayed decay, initiating at approximately 66 s, with complete disappearance occurring at 234.12 s (Fig. 3e, f). The extended existence of FeIV = O, lasting an additional 3.96 s beyond the depletion of HSO5−, highlights its long lifetime on the Fe-N-C on the scale of seconds. This result provided the experimental identification of the lifetime of surface metal-oxo generated during the Fenton reaction.

The effect of pH on the lifetime of FeIV = O was also examined, given that FeIV = O can degrade via reaction with H+ to form FeIII. As shown in Supplementary Fig. 8c–e, adjusting the pH of the Fenton-like system from approximately 2 (the original value used in the above experiments) to 7 using NaOH led to a slight increase in the FeIV = O lifetime from 3.96 s to 4.08 s, an increase of only 0.12 s, about 3% of the original lifetime. This indicates that although FeIV = O is slightly more stable under neutral conditions, the overall extension is minimal, suggesting that pH has a limited impact on its lifetime (Supplementary Fig. 8c–e).

To further consolidate the assignment of the IR band corresponding to surface FeIV = O on Fe-N-C, we conducted an 18O isotope labeling experiment. Due to the lack of commercially available 18O-labeled persulfate, we opted to use 18O-labeled water (H218O) instead. Since surface FeIV = O is produced from the oxidation of persulfate, it implies that the oxo O originates from persulfate rather than water. Consequently, the 18O labeling of water should have a minimal impact on the source of the oxo groups. However, previous studies have indicated that an oxygen exchange can occur between FeIV = O and water. When using PMSO as a molecular probe to detect FeIV = O, Jiang et al.39 found that, when 16O-PMS and H218O are employed, the oxidation of PMS16O by FeIV = O results in the formation of mixed-labeled PMS16O18O. This observation raises the possibility that following the formation of FeIV = 16O from 16O-PMS, an oxygen exchange with H218O may lead to the formation of FeIV = 18O and the subsequent oxidation products of PMS16O18O (Fig. 4a).

Fig. 4: Oxygen exchange mechanism between FeIV = O and water.
Fig. 4: Oxygen exchange mechanism between FeIV = O and water.The alternative text for this image may have been generated using AI.
Full size image

a Schematic illustration for the oxygen exchange between FeIV = O and water. b Time-dependent IR spectra conducted with 16O-PMS and H218O, the dashed curve exhibits a comparison with 16O-PMS and H216O (copied from Fig. 3c). The right panel highlights the FeIV = 16O and FeIV = 18O bands in an enlarged IR region. Source data for Fig. 4b is provided as a Source Data file.

Interestingly, as shown in Fig. 4b, in the IR spectra, replacing regular H216O with H218O revealed a new band at 795 cm−1. The original 833 cm−1 band persisted in the early stages of the reaction, but with decreased intensity compared to the scenario with only 16O. As the reaction progressed, the 833 cm−1 band gradually diminished, leaving the 795 cm−1 band as the sole observed feature. The theoretical value for the 18O/16O isotope shift can be predicted using vibration frequency calculation equations (“Isotope shift calculation” section in Supplementary Information). If the 833 cm−1 band is assigned to Fe=16O, its 18O counterpart should shift down to 796 cm−1, closely aligning with the observed band at 795 cm−1. Literature also supports a theoretical shift of approximately 36 cm−1 for the Fe-O harmonic oscillator17,40. We also performed a control experiment mixing regular 16O-PMS with H218O to check for the presence of 18O-persulfate through oxygen exchange (Supplementary Fig. 9a). However, no isotope-shifted band for persulfate was observed in the IR spectra, indicating that direct oxygen exchange does not occur between PMS and H2O. In addition, we dissolved 16O-labeled PMS in H218O and allowed the solution to stand for 24 h. High-resolution mass spectrometry (HRMS, Supplementary Fig. 9b) was then performed, and no peak corresponding to 18O-labeled PMS (m/z = 114.96) was detected, thereby ruling out the possibility of oxygen exchange under these conditions. Based on both infrared spectroscopy and high-resolution mass spectrometry analyses, we can confidently exclude the possibility of oxygen exchange between H218O and PMS.

Therefore, the observed 795 cm−1 band can be assigned to 18O-labelled FeIV = 18O, resulting from an immediate O-exchange between FeIV = 16O (produced from the oxidation by 16O-persulfate) and H218O. Notably, the growth of the FeIV = 18O signal exhibited a delayed kinetic profile compared to FeIV = 16O. Since the direct exchange of the oxo-O in a metal-oxo double bond is generally challenging, this kinetic delay would suggest a gradual oxygen-exchange pathway: it resembles the O-exchange between water and the carbonyl group of aldehyde or ketone, which occurs via the formation of a geminal diol intermediate41,42,43. The addition of H218O on FeIV = 16O generates a geminal-hydroxyl structure of FeIV(16OH)(18OH), subsequent dehydration of this intermediate yields either FeIV = 18O or reverts to FeIV = 16O, due to the relatively small energy difference between the Fe-18O-H and Fe-16O-H bonds. As a result, multiple exchange cycles are needed to accumulate a significant amount of the FeIV = 18O species. This requirement for repeated exchange events explains why the FeIV = 18O signal does not emerge with the same kinetics as FeIV = 16O, and instead appears with a noticeable delay. Nevertheless, this isotope experiment further supports our assignment of the 833 cm−1 band to surface FeIV = O. More importantly, the observed oxygen exchange indicates that this surface FeIV = O species possesses a relatively long lifetime and stability in the aqueous matrix, making it a potentially high-activity species for degrading aqueous pollutants.

The dynamic spectroscopic identification of FeIV = O in the Fenton-like reaction between Fe-N-C and PDS

For comparison, we performed similar experiments with the replacement of PMS with peroxydisulfate (PDS), using the same SF-ATR-FTIR characterization method (Fig. 5a). Unlike PMS, commercial-available PDS is a pure sample that exhibits characteristic IR peaks at 1274 cm−1 and 1050 cm−1, corresponding to the SO4 moiety in the S2O82− group (Supplementary Fig. 10a). During the in-situ monitoring the Fenton reaction between PDS and Fe-N-C, as shown in Fig. 5a, the appearance of PDS peaks exhibited a similar delay owing to PDS diffusion. The kinetic behavior mirrored that of PMS, initially increasing, reaching a maximum, and then declining to extinction. We also observed the growth of SO42− and HSO4− peaks, demonstrating a reaction between PDS and Fe-N-C. However, no new characteristic peaks were observed in the 800 ~ 900 cm−1 range. Replacing H216O with H218O did not change the characteristic peaks of PDS, and no new characteristic peaks emerged (Supplementary Fig. 10b). Therefore, no FeIV = O species is produced in the PDS and Fe-N-C system.

Fig. 5: Comparative spectroscopic identification of reactive species in PMS and PDS systems.
Fig. 5: Comparative spectroscopic identification of reactive species in PMS and PDS systems.The alternative text for this image may have been generated using AI.
Full size image

a Time-dependent IR spectra conducted with 16O-PDS and H216O, the dashed curve exhibited comparison with 16O-PMS and H216O (copied from Fig. 3c). The right panel highlights the FeIV = O band in an enlarged IR region. To improve the clarity of the spectral features and facilitate better observation of FeIV = O related signals, we have replaced the dashed lines with solid lines in the relevant figures. b The detection of reactive radicals in EPR spectra using 5,5-dimethyl-1-pyrrolidine-N-oxide (DMPO) as a trapping agent. c The EPR spectra using 2,2,6,6-tetramethyl-4-piperidinyloxyl (TEMP) as a trapping agent. d The high performance liquid chromatography (HPLC) absorption spectra for PMSO oxidation and PMSO2 generation. Source data for Fig. 5a–d are provided as a Source Data file.

To bolster the credibility of our IR findings, we conducted additional analyses using electron paramagnetic resonance (EPR) and PMSO molecular probe experiments. In EPR analysis (Fig. 5b), using 5,5-dimethyl-1-pyrrolidine-N-oxide (DMPO) as a trapping agent, in the PDS system, the intense signal with the typical feature of trapped SO4•− and •OH can be clearly observed, confirming the generation of SO4•− and •OH in PDS system. In contrast, these trapped radical signals were absent in the PMS system; instead, we detected the 5,5-dimethyl-2-pyrrolidinone-yloxy (DMPOX; aN = 7.2 G and aH = 4.1 G) signal, an oxidation product of DMPO44. The absence of significant signals for trapped SO4•− and •OH suggests the single-electron oxidation of FeII to FeIII is negligible in the PMS system, where this pathway is accompanied by the stoichiometric generation of SO4•− and •OH. Also, for both systems, the absence of 1O2 was also confirmed (Fig. 5c). These results confirm that the PMS and PDS systems generate distinct oxidative species, with the PMS system predominantly producing FeIV = O directly.

Furthermore, the PMSO molecular probe experiment was performed. As shown in Fig. 5d, the results revealed that PMSO was selectively oxidized to PMSO2 in the Fe-N-C/PMS system, whereas no PMSO2 signal was detected in the Fe-N-C/PDS system. Given that FeIV = O is known to selectively oxidize PMSO to PMSO2, whereas •OH and SO4•− cannot, these findings are consistent with the IR and EPR results, confirming the formation of FeIV = O species in the PMS system and suggesting their absence in the PDS system

Combining the results from in-situ FT-IR and EPR, we conclude that PMS effectively activates Fe-N-C, resulting in the formation of surface FeIV = O species as the dominant active species. In contrast, the presence of PDS primarily generates free radicals such as SO4•− and •OH, with no detectable FeIV = O formation. To elucidate the mechanistic differences between the PMS and PDS systems, we performed density functional theory (DFT) calculations with a focus on the detailed reaction intermediates and transition states, aiming to clarify the disparity in energy barriers between the two pathways. Our calculations revealed that the PMS system facilitates the rapid formation of the high-valent FeIV = O species via a low-barrier oxygen atom transfer (OAT) process, with an overall activation free energy of 13.5 kcal/mol (Supplementary Fig. 11, Supplementary Data 1). In contrast, the PDS system follows a distinct single electron transfer (SET) mechanism. Although the initial step is highly exergonic, then owing to the more complex dimeric structure of PDS compared to PMS, the formation of FeIV = O by PDS then requires involving multiple intermediates and transition states. The cumulative effect of these steps results in a significantly higher overall energy barrier along the reaction coordinate, making the formation of FeIV = O in the PDS system both not only thermodynamically and kinetically unfavorable (Supplementary Fig. 12, Supplementary Data 1). As a result, the FeIV = O species is effectively inaccessible under PDS conditions.

To further demonstrate the universality and reliability of our SF-ATR-FTIR technique in identifying FeIV = O species, we synthesized two additional Fe-Nx-based single-atom catalysts (Supplementary Fig. 13; Supplementary Table 4). During in situ monitoring of their Fenton-like reactions, a characteristic IR band around 830 cm−1 was consistently observed for both catalysts (Supplementary Fig. 14a, b). Complementary PMSO molecular probe experiments and EPR analyses also supported the presence of FeIV = O. These consistent findings confirmed that the ~830 cm−1 IR band detected by our technique is a reliable indicator of FeIV = O formation (Supplementary Fig. 14c–e).

Practical application of FeIV = O as an oxygen transfer reagent

FeIV = O has been recognized for its ability to facilitate an oxygen-atom transfer (OAT) process during the oxidation of organic and inorganic substrates18. This characteristic makes it especially effective in oxidizing substrates that possess lone-pair electrons, for instance, the aqueous arsenic (III) (AsIII). The aqueous AsIII, classified as a Group I carcinogen by the International Agency for Research on Cancer (IARC), has garnered significant attention due to its health risks45, even low-level exposure to AsIII can lead to various health issues. The preferred approach to remove AsIII is to oxidize it to arsenate (AsV), which has lower toxicity and is more readily removed by physical-chemical methods46. However, a significant challenge is that, in a real environmental system, aqueous AsIII typically exists at low concentrations and coexists with large amounts of low-toxicity organic and inorganic substrates. This presence complicates the use of broad-spectrum oxidizing agents, like radicals, which cannot selectively oxidize AsIII in the presence of interfering substrates.

We first assessed the effectiveness of the Fe-N-C/PMS system for removing 20 μM AsIII in a deionized water matrix. By quantifying the produced AsV (Supplementary Fig. 15), we found a significant generation of AsV within just 5 min, with approximately 85.5% of AsIII rapidly converted to AsV during this time (Fig. 6a). The final degradation rate of AsIII reached 99.55% within 15 min, demonstrating the high efficiency of the Fe-N-C/PMS system in treating AsIII. In contrast, when using the Fe-N-C/PDS system, where the ROS shifted from FeIV = O to free radicals, only trace amounts of AsV were detected in the initial 5 min, after which the generation of AsV became more pronounced. A scaled-up AsIII removal experiment (Supplementary Fig. 16) was also conducted to demonstrate the feasibility of Fe-N-C/PMS even in the enlarged system. In addition, we re-collected the Fe-N-C catalyst (Fe-N-C-r) after the AsIII degradation experiment and performed a series of post-reaction characterizations, including EXAFS, XPS, and HAADF-STEM (Supplementary Fig. 17), to verify the stability of the catalyst. The superior performance of Fe-N-C/PMS in the oxidative removal of AsIII can be attributed to the differing mechanisms. As shown in Fig. 6b, in the Fe-N-C/PDS system, the predominant radicals, with •OH and SO4•−, initiate a one-electron oxidation process that first converts AsIII to AsIV, this one-electron oxidation step faces a significant energy barrier of 2.40 eV, which slows the reaction in the initial stage. As a result, only a limited amount of AsV is generated until most AsIII is oxidized to AsIV. Once this occurs, the oxidation of AsIV to AsV becomes energetically favorable, leading to an increase in AsV production as observed in Fig. 6a. In contrast, in the Fe-N-C/PMS system, the dominant ROS, FeIV = O, facilitates a more efficient two-electron oxygen-atom transfer that directly transforms AsIII to AsV with a much lower energy barrier of only 0.56 eV. This mechanism results in more efficient oxidative removal of AsIII. The specific reactivity of FeIV = O in AsIII oxidation was also confirmed by our ST-ATR-FTIR measurements. First, the direct reaction between free PMS and AsIII was excluded by the control experiment in the absence of the Fe-N-C catalyst, as the introduction of AsIII did not cause any decay in the characteristic peak of PMS at 1250 cm−1 within 10 min’ reaction (Fig. 6c). Subsequently, when the Fe-N-C catalyst was immobilized on the ATR surface, co-injection of AsIII and PMS led to the disappearance of the FeIV = O characteristic band at 833 cm−1, indicating its rapid consumption by AsIII (Fig. 6d).

Fig. 6: Evaluation of AsIII oxidation performance and mechanism in different water matrices using Fe-N-C-based catalytic systems.
Fig. 6: Evaluation of AsIII oxidation performance and mechanism in different water matrices using Fe-N-C-based catalytic systems.The alternative text for this image may have been generated using AI.
Full size image

a Variation in AsV concentration during the oxidation of AsIII in a deionized water matrix using PMS/Fe-N-C and PDS/Fe-N-C (the complete data in Supplementary Table 2). b Schematic diagram illustrating the energy barrier for the one-electron oxidation of AsIII to AsIV and the direct oxygen atom transfer process. c Time-dependent IR spectra collected on a bare diamond ATR crystal in AsIII buffer solution, showing PMS diffusion without reaction. d The time-dependent IR spectra of AsIII oxidation by Fe-N-C/PMS, simultaneously capturing the diffusion and reaction of PMS. The right panel presents a magnified infrared region highlighting the FeIV = O band. e Variation of AsV concentration (solid line) and chemical oxygen demand (COD) (dashed line) during the removal of AsIII in a real lake water matrix (the complete data in Supplementary Table 3). The error bars in Fig. 6a and e are based on data from at least three parallel experiments. Experimental conditions for AsIII removal in panels (a) and (e): [AsIII] = 20 μM; [PMS] or [PDS] = 50 μM; [Fe-N-C] = 50 mg·L−1. Source data for Fig. 6a, c-e are provided as a Source Data file.

We then assessed the removal of AsIII in a real lake water matrix, sourced from the artificial lake at Beijing Olympic Park, which had a chemical oxygen demand (COD) value of 118 mg·L−1. Using the Fe-N-C/PMS system, the aqueous contaminants were effectively degraded, resulting in a COD reduction to 3 mg·L−1 within 30 min (Fig. 6e). Notably, despite the presence of various interfering contaminants, the removal of AsIII remained highly effective and a removal ratio of 96.8% was achieved. In contrast, while the Fe-N-C/PDS system also led to a significant decrease in COD, the conversion of AsIII to AsV was substantially hindered compared to the distilled water experiments, with only a small amount of AsV detected after 30 min and an AsIII removal ratio of only 12.5%. Similarly, the presence of humic acid and pH adjustment had minimal impact on AsIII removal by the Fe-N-C/PMS system, highlighting its excellent resistance to environmental interferences (Supplementary Figs. 18, 19). These results underscore the superior performance of Fe-N-C/PMS in selectively removing AsIII even in complex wastewater environments. This effectiveness stems from the specific reactivity of FeIV = O towards AsIII and the long lifetime of the FeIV = O species generated in the Fe-N4 isolated structure. The extended stability of FeIV = O ensures that the oxidative potential of PMS is utilized efficiently to remove AsIII and other contaminants, rather than being lost to meaningless decay.

This Fe-N-C/PMS system would demonstrate strong potential for practical environmental remediation owing to its excellent recoverability, cost-efficiency, and compatibility with existing water treatment processes. The catalyst can be recovered through simple physical methods, and future development of floating-type configurations may further improve separation efficiency and operational feasibility. Its unique ability to selectively oxidize AsIII via the formation of FeIV = O, combined with the high catalytic efficiency and low metal loading requirements of a single-atom catalyst, significantly reduces material costs. Moreover, PMS is already widely used in advanced oxidation processes, and the Fe-N-C/PMS system enhances both degradation efficiency and selectivity, ensuring seamless integration with current treatment infrastructures. These features collectively position the system as a promising candidate for scalable and sustainable water treatment applications.

Discussion

In summary, with the utilization of a spectrometer-controlled syringe pump, we synchronized the initiation of the Fenton reaction with IR data acquisition, allowing for dynamic monitoring of the heterogeneous Fenton reaction, including the variations and kinetics of reactants and intermediates with millisecond time resolution. This operando technique is not only applicable to Fenton reactions but can also be utilized for other reactions initiated by mixing reactants, offering valuable insights into reaction mechanisms.

With the support of this operando approach, we have advanced our understanding of heterogeneous Fe-oxo species, ROS that exhibit distinct reaction pathways compared to other radical ROS generated in Fenton systems. Our findings not only confirmed the presence of Fe-oxo species in the Fenton reaction but also elucidated their kinetic behaviors. Notably, we provided the experimental evidence of their lifetime on the timescale of seconds in aqueous systems. In contrast to the short-lived •OH or SO4•−, which last only nanoseconds, the long-lived Fe-oxo species minimize the loss of oxidative equivalents during ROS decay or recombination. This maximizes the oxidative potential of persulfate and other Fenton reagents for pollutant removal. Furthermore, the specific oxygen-atom transfer capabilities of Fe-oxo species make them particularly effective for the preferential removal of aqueous arsenic pollutants, even in complex wastewater environments.

Methods

Catalyst synthesis

Fe-N-C was synthesized following the procedure of Zhang et al.23,35. First, 0.5 mmol of Fe(OAc)2 and 1.5 mmol of phen were dissolved in 50 mL of ethanol and sonicated for 10 min. Next, 3.16 g of nano-MgO was added and the mixture was sonicated for another 10 min. The mixture was then stirred and refluxed at 60 °C for 12 h. After refluxing, the ethanol solvent was removed by rotary evaporation, and the remaining solid was dried at 60 °C overnight. The dried solid was then transferred to a tube furnace and calcined in an N2 atmosphere at 700 °C for 2 h, with a ramp rate of 2 °C per minute. The resulting black solid was dispersed in 100 mL of 1.0 M HNO3 solution and stirred at room temperature for 2 h. This acid treatment was repeated three times to remove the MgO carrier. The recovered solid was washed with water until the filtrate reached neutral pH, then dried at 60 °C for 12 h to obtain Fe-N-C.

Characterization

The micro-morphology and nanostructure of the catalysts were characterized by field emission scanning electron microscopy (SEM, ZEISS GeminiSEM 300) and transmission electron microscopy (TEM, JEOL JEM-F200), along with energy dispersive X-ray spectroscopy (EDS). X-ray diffraction (XRD) patterns were recorded with a Rigaku D/max 2500 diffractometer using Cu Kα radiation. X-ray photoelectron spectroscopy (XPS) was performed on an ESCALAB250XI instrument. High-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images were captured using a JEM-ARM300F microscope. Electron paramagnetic resonance (EPR) spectra were obtained with a Bruker Elexsys E500-T spectrometer. In-situ attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectra were collected using a Bruker Vertex70V. UV-visible absorption spectra were measured with a Hitachi UV-visible spectrophotometer (UV-3900).

Sample preparation for SF-ATR-FTIR: Specifically, the Fe-N-C was dispersed in distilled water at a concentration of 2 mg·mL−1 to prepare a deposition ink. Then, 10 μL of the ink was dropped onto the ATR crystal and dried in situ using an evacuation line to form a uniform membrane. As a result, approximately 20 μg of Fe-N-C was deposited as a thin film on the ATR crystal for subsequent IR measurements.