Manipulating the Second Coordination Shell of Single-Atom Fe for Enhanced Fenton Reaction

Abstract

While current methods use oxidizable metals as electron donors to effectively reduce Fe³⁺, they suffer from the irreversible oxidation of these metals, ultimately compromising the catalyst’s longevity. To address this challenge, we engineered the second coordination shell of a single-atom Fe center by doping boron (B) onto a graphene-based support (Fe₁/B-graphene) and utilized H₂O₂ as the electron source for efficient Fe²⁺ regeneration. Experimental results, supported by theoretical calculations, revealed that the Fe–O–B motif functions like a micro galvanic cell, with intermediary O atoms facilitating electron transfer between electrodes. Specifically, electrons consumed during H₂O₂ activation at Fe₁ sites (positive electrode) are replenished by electrons extracted from H₂O₂ at B atoms (negative electrode), where the activation energy for H₂O₂ oxidation is significantly lower than that at Fe₁ sites. This study offers inspirational insights into the design of Fenton catalysts through precise regulation of the second coordination shell, demonstrating the potential of tailoring the outer coordination environment of single-atom catalysts to enhance catalytic performance across various reactions.

Among various water remediation strategies, the Fenton reaction, which features H₂O₂ activation by Fe²⁺ (eq ), has been extensively studied and widely employed to generate ^•OH for the oxidative degradation of a variety of organic contaminants. However, the conventional homogeneous Fenton reaction involving dissolved Fe is constrained by a narrow pH range (typically 2.8–3.2). To address this, heterogeneous Fenton reactions have been explored to widen the working pH range and avoid the formation of Fe-containing sludge. Nevertheless, another fundamental challenge remains: the sluggish Fe²⁺ regeneration (eq ), which is the rate-limiting step in the overall catalysis kinetics. Current Fenton research endeavors are therefore focused on developing efficient approaches to expedite the Fe²⁺ regeneration.

$${\mathrm{H}}_2{\mathrm{O}}_2+{\mathrm{Fe}}^{2+}\to {\mathrm{OH}}^{-}+{}^{•}\mathrm{OH}+{\mathrm{Fe}}^{3+},k=40\sim 80{\mathrm{M}}^{-1}{\mathrm{s}}^{-1}$$ 1

$${\mathrm{Fe}}^{3+}+{\mathrm{H}}_2{\mathrm{O}}_2\to {\mathrm{Fe}}^{2+}+{{\mathrm{HO}}_2}^{•}+{\mathrm{H}}^{+},k=0.001\sim 0.1{\mathrm{M}}^{-1}{\mathrm{s}}^{-1}$$ 2

Introducing oxidizable metals that serve as electron donors, such as Mo⁴⁺, Cu⁺, and W⁴⁺, can address this issue by facilitating electron transfer (ET) from low-valent metals to Fe³⁺. Unfortunately, this strategy entails a significant trade-off of Fe²⁺ regenerating at the expense of oxidizing these metals (i.e., into Mo⁶⁺, Cu²⁺ and W⁶⁺). Without a means to ″recharge″ these metals, the catalyst’s longevity is inevitably compromised. Given the dual role of H₂O₂ as both electron donor and acceptor due to the intermediate valence state of O (−1) in the peroxide bond, we postulate that locating an electron-deficient oxidation site can induce the electron-donating behavior of H₂O₂, thus facilitating the electron extraction for Fe²⁺ regeneration via eq . Since the ET rate between two redox centers decreases exponentially with increasing distance through bridging atoms (k _ET ∝ exp­(-βR _DA ), where k _ET is the ET rate, R _DA is the donors-acceptor distance, and β is the exponential decay constant), , it is essential to position oxidation sites in close proximity to the Fe atoms. But engineering such precisely controlled architecture surrounding catalytic centers poses a challenge in material development.

Single-atom catalysts, such as Fe loaded on graphene, present a compelling design option. , Since Fe has a strong tendency to lose its outermost electrons, it is impractical to stably coordinate an electron-deficient atom in its first shell. Therefore, we hypothesize that an ideal alternative is to locate the oxidation sites, such as heteroatoms, in its second shell, achieving the theoretically shortest ET distance. Among potential heteroatom candidates, we consider B to be the most promising for the Fenton reaction, because its low electronegativity (2.04) promotes the shift of electron density to first shell atoms, increasing the electron density of the Fe center via the remote electronic induction effect. Moreover, while other nonmetallic heteroatoms (e.g., P, and S) have been employed to regulate the second shell, they are also capable of directly coordinating with Fe in the first shell (e.g., Fe–P, and Fe–S), leading to the complex interactions between Fe and heteroatoms across both the first and second shells. In contrast, the semimetallic nature of B can inhibit its direct bonding to the Fe center, confining it exclusively to the second shell and therefore providing an appropriate model to exploit second shell effects. ,

We test this hypothesis by anchoring a single-atom Fe catalyst onto B-doped graphene (Fe₁/B-graphene) to construct the Fe–O–B motif. We evaluate the Fenton activity of Fe₁/B-graphene by catalyzing the oxidative removal of 1,4-dioxane, a representative contaminant of significant environmental concern due to its widespread occurrence and resistance to common adsorption and membrane filtration techniques. We further construct a catalytic membrane consisting of Fe₁/B-graphene and test its performance for flow-through treatment in practical water treatment scenarios. Comprehensive material characterization, accompanied by in-depth theoretical simulations, is presented to uncover the distinctive role of B atoms, offering inspirational insights into the design of efficient and stable Fenton catalysts.

Results and Discussion

Synthesis and Characterization of Fe₁/B-graphene

We synthesized Fe₁/B-graphene catalyst via a two-step procedure (Figure S1), which consisted of oxidation of the graphene substrate followed by thermal treatment to pyrolyze the Fe precursors and incorporate B doping. , The first step involved the acid treatment of a graphene substrate to introduce O-containing groups; these functionalities either served as anchoring sites for the Fe₁ center in Fe₁/O-graphene or were reductively removed to form substitutional doping sites for B in Fe₁/B-graphene. The uniform distribution of O and B (Figures a, b and S2) was crucial for well-dispersed Fe₁ sites. X-ray photoelectron spectroscopy (XPS) analysis of Fe₁/B-graphene revealed that the deconvoluted B 1s peak corresponded to BC₂O (191.2 eV, 64.4%) and BCO₂ (192.3 eV, 35.6%) moieties, indicating the substitution of C atoms with B atoms at the defective and periphery sites in the graphene lattice, respectively (Figure S3). , The loading amount of Fe was determined by inductively coupled plasma-mass spectrometry (ICP-MS) to be 5.3 wt % for Fe₁/O-graphene and 4.1 wt % for Fe₁/B-graphene, in agreement with the XPS results presented in Table S1. Aberration-corrected high-angle annular-dark-field scanning transmission electron microscopy (AC-HAADF-STEM) images in Figures c and d visually demonstrated the single-atom configuration of Fe in both Fe₁/O-graphene and Fe₁/B-graphene, with no Fe nanoparticles present.

1.

TEM images and the corresponding element mapping of (a) Fe₁/O-graphene, and (b) Fe₁/B-graphene; AC-HAADF-STEM images of (c) Fe₁/O-graphene and (d) Fe₁/B-graphene; (e) normalized X-ray absorption near edge structure (XANES) measurements at the Fe K-edge of Fe foil, Fe₂O₃, Fe₁/O-graphene and Fe₁/B-graphene; (f) Fourier-transformed extended XAFS (FT-EXAFS) of Fe references and Fe₁ catalysts; (g) fitting results for Fe₁/O-graphene and Fe₁/B-graphene (the parameters extracted from the fit are provided in Table S2). The data ranges used for data fitting in k-space and R-space were 3.0–10 and 1.2–2.5 Å, respectively.

The white line intensities in the Fe K-edge XANES spectra (Figure e) provided qualitative insight into the oxidation state of Fe, showing that Fe in both Fe₁/B-graphene and Fe₁/O-graphene existed in an oxidation state that is close to +3 (Fe₂O₃). XPS was further employed to quantify the oxidation state (Figure S4), whereas Fe in Fe₁/B-graphene was slightly more electron-rich (+2.4) than in Fe₁/O-graphene (+2.7). Comparisons of Fe₁ with the Fe foil reference using FT-EXAFS (Figure f) confirmed the absence of an Fe–Fe peak, excluding the existence of metallic Fe. This is consistent with the X-ray diffraction results (Figure S5), which showed no detectable crystalline Fe peaks. The primary peak for Fe₁/O-graphene was attributed to Fe–O coordination based on the Fe₂O₃ reference. Notably, the Fe–O peak shifted slightly left for Fe₁/B-graphene, compared to Fe₁/O-graphene, indicating a possible change in the local coordination environment around the Fe atoms. Similar peak shifts have previously been attributed to modifications in the second coordination shell surrounding the metal center. Based on these results, along with the O–B–C motifs (i.e., BCO₂ and BC₂O) observed from XPS analysis (Figure S3 ), we propose that B doping altered the Fe–O–C coordination in Fe₁/O-graphene to Fe–O–B in Fe₁/B-graphene. Best-fit parameters (Figures g, S6, S7 and Table S2) further revealed that the Fe–O coordination number for both Fe₁/O-graphene and Fe₁/B-graphene was approximately 4 (in contrast to 6 for Fe₂O₃). Such unsaturated coordination environment of Fe₁ could potentially benefit the H₂O₂ adsorption and subsequent activation. ,

Catalyst Performance

As illustrated in Figures a and b, Fe₁/B-graphene exhibited significantly higher activity toward 1,4-dioxane degradation compared to Fe₁/O-graphene over a broad pH range. Although increasing pH led to a decrease in TOF (defined as the number of contaminant molecules degraded per active site per minute) for both Fe₁ catalysts, Fe₁/B-graphene demonstrated much better pH adaptability (Figures S8 and S9). Specifically, the TOF for Fe₁/B-graphene under neutral conditions retained approximately 35% of that at pH 3, whereas the TOF for Fe₁/O-graphene only retained 5% (Table S3). This suggests that Fe₁/B-graphene is more adaptable to a wider range of pH conditions, making it particularly promising for practical applications.

2.

Kinetic plots of 1,4-dioxane degradation using (a) Fe₁/O-graphene and (b) Fe₁/B-graphene under various pH conditions. Conditions: catalysts = 0.5 g/L, 1,4-dioxane concentration = 50 mg/L, H₂O₂ = 0.5 g/L, temperature = 25 ± 1 °C. (c) Electron paramagnetic resonance (EPR) spectroscopy for Fe₁/O-graphene and Fe₁/B-graphene, using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) and 2,2,6,6-tetramethylpiperidine (TEMP) as the probes. The signals for ^•OH and ¹O₂ were collected in water at pH 4, while the signal for ^•O₂ ^– were detected in methanol at pH 7. (d) Turnover frequency (TOF) comparisons for Fe₁/B-graphene with the addition of different quenchers under pH 4 (tert-butyl alcohol (TBA), 100 mM; nitro blue tetrazolium chloride (NBT), 10 mM; potassium thiocyanate (KSCN), 20 mM); (e) 10 h longevity tests using Fe₁/B-graphene in a flow-through mode under neutral pH conditions. The flux was 24 L/(m²·h) driven by the peristaltic pump. The pore size and residence time (RT) were determined to be ∼6.7 nm and 1.32 S, respectively. The test conducted in three separate trials using the same membrane, with flushing the membrane for 1 h between trials to remove residual H₂O₂ and pollutants.

Reaction Mechanisms

We attribute 1,4-dioxane degradation to reactive oxygen species that are generated from the activation of H₂O₂ by Fe₁ catalysts. We employed EPR analysis to distinguish the roles of different reactive oxygen species during the reaction. As shown in Figure c, both ^•OH and ^•O₂ ^– signals were much more intense when using Fe₁/B-graphene, while the ¹O₂ signal was comparable between the two Fe₁ catalysts, thus excluding the potential contribution from ¹O₂. Additionally, both ^•OH and ^•O₂ ^– signals diminished with increasing pH values (Figure S10), consistent with the trend observed in TOF.

To evaluate whether ^•OH and ^•O₂ ^– cocontribute to 1,4-dioxane degradation, we added TBA (^•OH quencher) and NBT (^•O₂ ^– quencher) to the reaction suspension of Fe₁/B-graphene (Figures d, S11, and Table S4). The addition of TBA completely inhibited the reaction, suggesting the dominant role of ^•OH. Interestingly, adding NBT increased TOF by 13%. At pH 4, only 15.8% of HO₂ ^• (pK _a = 4.8) would be deprotonated into ^•O₂ ^–. However, the excessive addition of NBT, which continuously quenched ^•O₂ ^– (NBT + ^•O₂ ^– → formazan + O₂ ), drove the equilibrium toward further dissociation (HO₂ ^• ⇌ H⁺ + ^•O₂ ^–), resulting in a reduced proportion of HO₂ ^• in the system. The decreased HO₂ ^• would benefit the Fenton reaction through two pathways: (i) promoting Fe²⁺ regeneration (eq ) and (ii) suppressing side reactions caused by HO₂ ^•, such as Fe²⁺ reoxidation (eq ) and ^•OH consumption (eq ).

We observed that 1,4-dioxane degradation was almost completely inhibited when we blocked Fe₁ by adding KSCN (Figure S11), forming a stable Fe-SCN complex. This confirmed the critical role of Fe₁ sites in catalyzing H₂O₂ activation. Furthermore, the absence of methyl phenyl sulfone (PMSO₂) throughout the reaction (Figure S12) excluded the involvement of high-valent Fe species (i.e., Fe^IV=O and Fe^V=O), widely reported as oxidants for the selective oxidization of methyl phenyl sulfoxide (PMSO) into PMSO₂. These results are consistent with the in situ Raman spectra (Figure S13), which exhibit no characteristic peaks for high-valent Fe species.

$${{\mathrm{HO}}_2}^{•}+{\mathrm{Fe}}^{2+}\to {{\mathrm{HO}}_2}^{-}+{\mathrm{Fe}}^{3+},k=0.72\sim 1.5\times {10}^6{\mathrm{M}}^{-1}{\mathrm{s}}^{-1}$$ 3

$${{\mathrm{HO}}_2}^{•}+{}^{•}\mathrm{OH}\to {\mathrm{H}}_2\mathrm{O}+{\mathrm{O}}_2,k=1.4\times {10}^{10}{\mathrm{M}}^{-1}{\mathrm{s}}^{-1}$$ 4

Catalyst Stability

We examined the stability of Fe₁ for potential Fe leaching by measuring the concentration of Fe ions in suspension after the 8-h Fenton reaction. The Fe concentration was below the detection limit of ICP-MS (approximately 100 ng/L) above pH 4 (Figure S14). The high stability is attributed to the strong metal–support interaction of interfacial Fe–O bonds. When the pH was lowered below 4, the high-concentration H₂SO₄ caused acidic etching of the Fe₁ centers. A comprehensive comparison of reaction kinetics across various reaction systems (Figure S15 and Table S5) further demonstrated the impressive reactivity of Fe₁/B-graphene within the pH range of 4 to 7.

Inspired by this result, we performed a long-term experiment (Figure S16) using Fe₁/B-graphene in a membrane configuration. We fabricated a membrane consisting of layers of Fe₁/B-graphene following the procedures described in Methods. The membrane exhibited a molecular weight cutoff of 18 kDa (Figure S17), similar to commercial ultrafiltration membranes used for dissolved organic removal. During the steady-state operation under a permeate flux of 24 L/(m²·h), we observed ∼ 90% of 1,4-dioxane removal under neutral pH with concurrent consumption of 29% of H₂O₂ (145 mg/L, 7.5 times the molar concentration of 1,4-dioxane in the feed solution) (Figure e). The absence of Fe leaching, as confirmed by ICP-MS after the membrane test, further validates the stability of Fe₁/B-graphene, consistent with the leaching results under neutral conditions (Figure S14). The RT of water within the membrane under this flux, and hence the contact time, was estimated to be approximately 1.3 s (Text S1), highlighting extremely fast kinetics (1.74 S^–1, ∼ 4600 times of reaction kinetics in batch experiment) achieved in the membrane configuration due to efficient mass transfer within the confined membrane pores. 1,4-dioxane removal through adsorption was ruled out based on control experiments conducted without H₂O₂ (Figure S18). Notably, the negligible adsorption additionally excludes the possible contribution from the surface-bound ^•OH (^•OH_ads), which has been broadly reported to participate in the ^•OH-driven reaction. To further validate this finding, F^– was introduced into the reaction system (Figure S19), which is known to efficiently convert ^•OH_ads into free ^•OH. The absence of significant inhibition demonstrated the dominant role of free ^•OH toward 1,4-dioxane degradation. To evaluate the practical applicability of Fe₁/B-graphene, we additionally conducted experiments with the presence of common background constituents (e.g., CO₃ ^2–, Cl^–, SO₄ ^2–, and humic acid (HA)) under pH 4, as well as a real water sample (Figure S20) with pH unadjusted. Fe₁/B-graphene retained high catalytic activity in the presence of CO₃ ^2– (predominantly exist as H₂CO₃ due to a pK _a of 6.3), SO₄ ^2–, and natural water matrix; the observed inhibitory effects of Cl^– and HA are ascribed to their well-documented scavenging of ^•OH radicals. ,

XPS analyses of the Fe 2p region after the Fe₁/B-graphene longevity test demonstrated that Fe₁ maintained an oxidation state of +2.4 (Figure S21) without undergoing oxidation, which is a critical challenge in Fenton process. Additionally, the constant electronic structure of B (validated by XPS results in Figure S22) and the marginal variation in elemental composition (Table S6) confirmed the high stability of Fe₁/B-graphene during the reaction. An AC-HAADF-STEM image further demonstrated that Fe₁ remained atomically dispersed without aggregation into Fe nanoparticles (Figure S23), underscoring the structural stability of this catalyst over extended use.

Role of B Doping

Given that different neighboring nonmetallic species have been demonstrated to influence the Fenton-like activity in previous studies, , we further investigated the relationship between B species (i.e., BC₂O at 191.2 eV and BCO₂ at 192.3 eV) and Fenton activity. The ratio of different B species was controlled by adjusting the temperature for thermal treatment from 400 to 700 °C (note that the subscript denotes the temperature, for example, Fe₁/B-graphene₇₀₀ refers to the sample treated at 700 °C). As presented in Figures a-d, increasing treatment temperature led to a notable increase in the proportion of BC₂O species, while no detectable variation in Fe 2p region was observed, excluding any potential influence from the valence state of Fe (Figure S24). AC-HAADF-STEM results (Figure S25) additionally exclude the formation of Fe nanoparticles under the various conditions. We observed that the degradation kinetics of 1,4-dioxane (Figures e, S26 and Table S7) were significantly impacted by the variation in treatment temperature; from 0.0385 min^–1 for Fe₁/B-graphense₄₀₀ to 0.0705 min^–1 for Fe₁/B-graphene₇₀₀. Combining the above results (Figure f), we observed a consistent increase in TOF with the increasing proportion of BC₂O. This result suggests a critical role of B species in the second coordination shell of Fe₁.

3.

(a–d) B 1s XPS results for Fe₁/B-graphene_TEMP catalysts synthesized under different thermal treatment temperatures. (e) Linear fitting of reaction kinetics for 1,4-dioxane degradation using various Fe₁/B-graphene_TEMP catalysts. (f) Correlation between TOF and BCO₂ ratio. Notably, Fe₁/B-graphene was synthesized at 800 °C, the same pyrolysis temperature used for the synthesis of Fe₁/B-graphene₈₀₀.

To uncover the atomic-level mechanisms underlying the superior catalytic performance of Fe₁/B-graphene, we employed DFT calculations to simulate the relevant reaction pathways (Figure a). Although both catalysts exhibit negative adsorption energies for initial H₂O₂ adsorption, indicating an exothermic process, the lower adsorption energy on the Fe₁/B-graphene (−0.61 eV) compared to Fe₁/O-graphene (−0.35 eV) suggests that H₂O₂ adsorption is more thermodynamically favorable on the boron-doped catalyst. Notably, O atoms in H₂O₂ preferentially adsorbed onto the Fe₁ active sites rather than the terminal H atoms (Figures S27–29). This adsorption configuration facilitates the subsequent dissociation of O–O bond in the H₂O₂ molecule (IM1 to IM2), resulting in the formation of two OH* fragments, each carrying an unpaired electron. The reaction energy for this step is negative for both catalysts, signifying a barrierless and spontaneous process. The subsequent step, transitioning from IM2 to IM3, involves two critical events: (i) the ET from Fe²⁺ to OH*, leading to the formation of OH^–, and (ii) desorption of OH* to generate free ·OH radicals, which are essential for decontamination. The calculated reaction energy for this step is 2.93 and 2.39 eV for Fe₁/B-graphene and Fe₁/O-graphene, respectively, indicating that ·OH generation was thermodynamically more favorable on Fe₁/O-graphene.

4.

(a) Density functional theory (DFT) calculated minimum energy pathways for the H₂O₂ activation (eq ) and Fe²⁺ regeneration (eq ) catalyzed by Fe₁/O-graphene and Fe₁/B-graphene. Abbreviations: IM, intermediate; TS, transition state. All theoretical models involved in the reaction process are presented in Table S8 and Figures S27–S30. (b) Minimal basis iterative Stockholder (MBIS) charge analysis of ET during Fe²⁺ regeneration. Atoms displayed in green with positive numbers are in an electron-deficient state, wherein darker shades of green indicate a higher degree of electron deficiency. Conversely, atoms displayed in blue with negative numbers are in an electron-rich state, with darker blue suggesting a higher electron density. (c) Reaction energy for the reoxidation of Fe₁ by HO₂ ^•. Inset: electron distribution analyses upon HO₂ ^• adsorption onto two Fe₁ catalysts and by the charge density difference (CDD), yellow for the charge accumulation, and blue for electron depletion. (d) Schematic representation of the inhibited ·OH quenching by constructing the Fe–O–B moiety, conceptually resembling the micro galvanic cell configuration.

However, this observation appeared to contradict the observed degradation performance, where Fe₁/B-graphene exhibits much improved Fenton activity. To resolve this discrepancy, we further investigated the Fe²⁺ regeneration process through H₂O₂ oxidation (eq ), identified as the rate-limiting step governing the overall Fenton activity. Unlike Fe₁/O-graphene, where Fe₁ serves as the sole reactive site for H₂O₂ oxidation, the presence of B atoms in the second shell of Fe₁/B-graphene may offer an additional reactive site. To investigate this possibility, we first examined the adsorption energy of H₂O₂ on these sites. The results reveal that H₂O₂ exhibits a lower adsorption energy on B sites (−1.21 eV) compared to that for Fe₁ sites (−0.41 eV in Fe₁/B-graphene, and −0.92 eV in Fe₁/O-graphene), demonstrating a preferential adsorption of H₂O₂ on B sites in Fe₁/B-graphene. Further analysis of the reaction barrier for H₂O₂ oxidation highlights the critical role of B sites in facilitating this step. Specifically, the reaction barrier on B sites in Fe₁/B-graphene was calculated to be 2.63 eV (Figure S28), which is 0.77 eV lower than the barrier on Fe₁ sites within the same material (Figure S29). For Fe₁/O-graphene, the reaction barrier for H₂O₂ oxidation was substantially higher, at 3.99 eV (Figure S27), even exceeding the barrier for Fe₁ sites in Fe₁/B-graphene. These findings unequivocally underscore the pivotal role of the B atoms in accelerating H₂O₂ oxidation, thereby enhancing the overall Fenton efficiency. This mechanistic insight aligns well with the higher TOF observed experimentally (Table S3), providing a coherent explanation for the superior catalytic performance of Fe₁/B-graphene.

To gain deeper insight into the ET process from the H₂O₂ molecule to the oxidized Fe₁ sites in Fe₁/B-graphene, we employed the MBIS charge analysis. This approach partitions the electron density of a reaction system into its constituent atoms, enabling a detailed examination of ET during the reaction. The MBIS analysis assigns numerical values to each atom, representing their partial charges. Although these values do not directly denote the valence state, they exhibit a strong correlation with the oxidation state, offering a robust quantitative framework to understand the electron redistribution throughout the catalytic process. As revealed in Figure b, the transition from phase I to phase II reveals the notable ET within the Fe₁/B-graphene system. Specifically, the adsorption of O atoms from H₂O₂ onto the B sites results in an increase in the partial charge of the O atoms (from −0.86 to −0.80) in the adsorbed *OOH component, indicating a loss of electron density. Conversely, the partial charges of the B atom (from 0.90 to 0.84), the intermediary O atom (from −0.55 to −0.67), and the Fe₁ atom (from 1.45 to 1.33) decreased, signifying a gain of electron density on these atoms (i.e., Fe–O–B motif).

During the subsequent transition from phase II to phase III, ET occurred between the intermediary O atom and Fe₁ center. The intermediary O in an electron-rich state transferred electrons to the Fe₁, leading to a reduction in the partial charge of Fe₁ from 1.33 to 1.30, accomplishing the Fe²⁺ regeneration. Furthermore, the subsequent oxidation of Fe₁ sites from phase III to phase I (due to the H₂O₂ activation, eq ) increased the partial charge of the B atom from 0.84 to 0.90, rendering B more electron-deficient. This electronic state is advantageous for the electron extraction from H₂O₂ through eq . Thus, the Fe–O–B moiety functions analogously to a micro galvanic cell, where the Fe₁ atom serves as the positive electrode for H₂O₂ activation, the B acts as the negative electrode for H₂O₂ oxidation, and the intermediary O facilitates ET from B to Fe. The overall ET process within the Fe–O–B moiety underpins the catalytic efficiency of Fe₁/B-graphene, offering a mechanistic explanation for its superior performance in the Fenton reaction.

Apart from the core reactions related to the H₂O₂ activation and Fe²⁺ regeneration in Figure a, multiple competitive reactions also play the critical role in influencing the overall Fenton efficiency. Due to the unpaired electron in a π* antibonding orbital associated with the O–O bond, HO₂ ^• exhibits both oxidative and reductive properties. This dual reactivity enables HO₂ ^• to participate in rapid Fe²⁺ reoxidation (eq ) and quenching of ·OH (eq ) in a conventional Fenton process. It is noteworthy that although HO₂ ^• can lead to Fe³⁺ reduction as well (eq ), the kinetics were magnitudes lower than that of the reaction between HO₂ ^• and Fe²⁺. Consequently, the Fe³⁺ reduction pathway was excluded from the DFT simulations due to its negligible contribution under typical reaction conditions. The Fe²⁺ reoxidation reaction (eq ) is a one-electron transfer process that does not involve bond cleavage or formation, posing inherent challenge in accurately modeling transient states and therefore leading to inaccuracies in describing charge localization and delocalization. While kinetic analysis of this step is limited due to difficulties in defining activation energies, its thermodynamic feasibility can still be assessed. As shown in Figure c, the reaction energy for Fe²⁺ reoxidation was calculated to be 0.23 eV for Fe₁/O-graphene, significantly lower than that for Fe₁/B-graphene (2.06 eV for B sites, and 2.71 eV for Fe₁ sites), underscoring the intrinsic resistance of Fe₁/B-graphene to HO₂ ^• oxidation.

$${\mathrm{Fe}}^{3+}+{{\mathrm{HO}}_2}^{•}\to {\mathrm{Fe}}^{2+}+{\mathrm{O}}_2+{\mathrm{H}}^{+},k=2\times {10}^3{\mathrm{M}}^{-1}{\mathrm{s}}^{-1}$$ 5

To gain a more comprehensive understanding of the interaction between Fe₁ and HO₂ ^•, CDD analyses were performed to explore the electron distribution on the two Fe₁ catalysts. As shown in Figure c, the blue region surrounding Fe₁ in Fe₁/O-graphene represents electron depletion, indicating substantial electron transfer from Fe₁ to the adsorbed HO₂ ^• species. In contrast, the yellow region surrounding Fe₁ in Fe₁/B-graphene denotes a relatively higher electron density at the Fe₁ site, reflecting reduced electron transfer to HO₂ ^•. These results collectively suggest that the Fe–O–B coordination motif exhibits enhanced resistance to HO₂ ^•oxidation, contributing to improved catalyst stability. Moreover, the dual role of Fe₁ in Fe₁/O-graphene as the reactive site for H₂O₂ activation and oxidation leads to the inevitable quenching of ·OH by in situ generated HO₂ ^•, forming H₂O and O₂ via eq . In contrast, the Fe–O–B configuration in Fe₁/B-graphene enabled the spatial separation of Fe and B atoms, which served as the H₂O₂ activation (eq ) and oxidation (eq ) sites, respectively (Figure d), thus effectively inhibiting the rapid quenching of ·OH to improve the Fenton performance. Notably, the Fe–O–B differs from a conventional electrochemical cell, since H₂O₂ activation, not ion migration, transfers charges between positive and negative electrodes. We additionally analyze the nonradical decay of H₂O₂ (i.e., H₂O₂ →H₂O + O₂), a process that reduces the utilization efficiency of H₂O₂ and even accelerate under evaluated pH conditions. , The reaction barrier for Fe₁/B-graphene was 0.49 eV higher than that for Fe₁/O-graphene (Figure S30), indicating suppressed H₂O₂ self-decomposition. This suppression benefits H₂O₂ utilization and thereby prevents the impairment of Fenton activity, further contributing to the remarkable catalytic performance of Fe₁/B-graphene in the Fenton reaction.

Conclusions

In this study, we manipulated the coordination environment of Fe₁ through loading onto different graphene-based supports and evaluated the resulting Fenton activity on the degradation of 1, 4-dioxane as a representative pollutant. Although the first coordination shell, consisting of four O atoms, was identical, the presence of B atom in the second shell of Fe in Fe₁/B-graphene significantly improved Fenton performance. Experimental findings, supported by computational analyses, revealed that the engineered Fe–O–B moiety operates like a micro galvanic cell, with Fe₁ sites serving as the positive electrode for H₂O₂ activation, B atoms functioning as the negative electrode for H₂O₂ oxidation, and intermediary O atoms acting as the conduit for ET between the electrodes. This unique configuration enhanced overall Fenton efficiency by promoting Fe²⁺ regeneration and suppressing the competitive reactions of Fe²⁺ reoxidation and ^•OH quenching. This work underscores the opportunity to fully exploit the potential of the Fe catalyst used for Fenton reaction by tailoring the outer coordination environment of single-atom Fe.

Methods

Catalyst Synthesis

Graphene was first oxidized using an acid treatment as previously reported. Afterward, 0.5 g graphene oxide was dispersed in 50 mL deionized water, along with 5 mL FeCl₃ (10 mM). Boric acid (5 g) was added as the boron precursor. The resulting solution underwent probe sonication for 2.0 h, followed by drying in a freeze-dryer. The mixture was finely grounded and transferred to a tube furnace for the final thermal treatment. This process was conducted at 800 °C for 4 h, with a heating rate of 5 °C/min under a reductive atmosphere (H₂:Ar = 5:95). The obtained sample was designated as Fe₁/O-graphene or Fe₁/B-graphene, depending on the specific processing conditions.

Catalyst Characterization

AC-HAADF-STEM was performed using a Titan Themis Z STEM (Thermo Fisher Scientific, USA) operating at 200 kV and equipped with a probe aberration corrector, achieving an imaging spatial resolution of less than 1 Å. Transmission electron microscopy images were acquired using a Titan Cubed Themis G2 300 operating at 100 kV. XAFS spectroscopy at the Fe K-edge was conducted at the Inner Shell Spectroscopy beamline of the National Synchrotron Light Source II at Brookhaven National Laboratory. XPS analysis was performed using an ESCALAB 250Xi XPS system with monochromatic Al Kα radiation (1486.6 eV). EPR analysis was carried out using an ESR-300E spectrometer (Bruker Instruments) with DMPO as the spin-trapping agent for ^•O₂ ^– (existing as HO₂ ^• in methanol due to a pK _a of 4.8), ^•OH, and TEMP for ¹O₂. Samples for EPR analysis were collected after 10 min. In this work, all EPR spectra were collected under strictly identical experimental conditions (matching sample concentrations, solvent, instrument parameters, and measurement settings), with peak intensities serving exclusively for qualitative comparison of radical generation trends among different catalysts, not for absolute quantification. ICP-MS was performed using a PerkinElmer SCIEX Elan DRC-e to determine the Fe loading. A 10 mg sample was acid-digested using a mixture of 7.5 mL of 37% HCl, 2.5 mL of 70% HNO₃, and 2.0 mL of 47.5% HF in a Mars5 CEM microwave system. The digestion process involved two stages: (i) heating to 220 °C over 15 min, followed by (ii) holding at 220 °C for 40 min. The samples were subsequently cooled to room temperature for at least 3 h. To neutralize the HF, 1.2 g of ACS-grade boric acid was added to each solution, and the mixture was digested a second time under the same microwave conditions. The digested solutions were transferred to polypropylene centrifuge tubes and diluted to 50 mL with ultrapure water. From each solution, a 0.15 mL aliquot was extracted and further diluted to 12 mL for ICP-MS analysis.

Quantification of 1,4-Dioxane

The concentration of 1,4-dioxane was quantified using ultraperformance liquid chromatography (Waters, AcQuity UPLC H-class) equipped with an HSS T3 column (2.1 mm × 100 mm, 1.8 μm). Detection was performed at an absorption wavelength of 191 nm. The mobile phase consisted of a mixture of acetonitrile (A) and water (B) with the following gradient program: 0–5 min, 0% A and 100% B; 6–7 min, 100% A and 0% B. The flow rate was maintained at 0.3 mL/min.

Degradation Experiments

The degradation of 50 mg/L 1,4-dioxane was investigated in a suspension system containing 0.5 g/L catalyst. The pH was adjusted using H₂SO₄. The suspension was sonicated for 10 min to achieve dynamic adsorption/desorption equilibrium. Subsequently, H₂O₂ was introduced to initiate the reaction at a concentration of 0.5 g/L (14.7 mM). Aliquots (250 μL) were collected at specific time intervals and immediately filtered using a poly­(ether sulfone) filter (0.22 μm).

Membrane Parameter Determination

For the longevity test, 50 mg of Fe₁/B-graphene was loaded onto a polyvinylidene fluoride membrane (pore size: 0.45 μm) via vacuum filtration. The average pore size of the Fe₁/B-graphene membrane was determined using the molecular weight cutoff method with a 1 wt % polyethylene glycol solution, based on the following equation: (d, the hydrodynamic diameter of a molecule or particle; MW, the molecular weight; β, a proportionality constant)

$$d=\beta {(\mathrm{mw})}^n=0.09{(\mathrm{mw})}^{0.44}$$

The flow rate of the peristaltic pump was set to 5 mL/min, while the actual water outlet flow rate was approximately 0.5 mL/min. The flux was calculated using the following equation: (J, flux; V, volume of permeate; A, membrane area; t, time of filtration)

$$J=\frac{V}{A^{*}t}$$

$$J=\frac{0.5mL}{{3.14}^{*}{20}^2{mm}^{2^{*}}1\min }=24\mathrm{L}/\mathrm{m}2·\mathrm{h}$$

The RT was calculated based on the below equation: (V _pore , volume of membrane; W _wet , wet weight of the membrane; W _dry , dry weight of the membrane; ρ _liquid , density of the solution; Q, flow rate)

$$RT=\frac{V_{pore}}{Q}=\frac{W_{wet}-W_{dry}}{{{\rho }_{liquid}}^{*}Q}$$

$$RT=\frac{V_{pore}}{Q}=\frac{(276-217)-48\mathrm{mg}}{1000{\frac{mg}{mL}}^{*}(0.5/60)\frac{mL}{s}}=1.32\mathrm{S}$$

XAFS Data Fitting

XAFS data processing was performed using Athena software, which facilitated the conversion of raw data into μ­(E) spectra, background subtraction, normalization, Fourier transformation, and plotting. EXAFS data analysis was conducted using Artemis software, employing theoretical standards. This included defining the range of the Fourier transform in k-space and setting the fitting range parameters in R-space. The interatomic distance refers to the bond length between the central atom and the surrounding coordination atoms. The Debye–Waller factor (σ²) represents the thermal and static disorder in the absorber-scatterer distance.

Theoretical Analyses

All computations were carried out using the Vienna Ab Initio Simulation Package, with electron–ion interactions modeled through the projector-augmented wave method. The wave functions were expanded using a plane wave basis set, with a cutoff energy of 400 eV. The exchange-correlation energies were calculated using the generalized gradient approximation of Perdew–Burke–Ernzerhof, and dispersion interactions were accounted for by employing Grimme’s DFT-D3 correction method. Lattice atoms, not constrained to the bulk structure, were permitted to relax until the force on each atom was reduced to below 0.03 eV/Å. The electronic convergence criterion was set to a total energy tolerance of 10^–5 eV. To balance computational accuracy with efficiency, the gamma point was selected for the k-point sampling in the Brillouin zone. The climbing image nudged elastic band method was initially used to locate the transition state, with the maximum force limited to 0.1 eV/Å for initial convergence. The dimer method was then employed to refine the transition state location, with convergence achieved when the interatomic forces were below 0.03 eV/Å at the saddle point. Atomic charge distributions were obtained using the MBIS method, with the Multiwfn software. ,

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsnano.5c13343.

• MBIS analysis; quantification of ^•OH radical; schematic illustration of synthesis; TEM images and element mapping; XPS results and element ratio; XRD results; XAFS fitting; pseudo-first-order linear fitting; kinetics comparisons; EPR results; detection of PMSO₂ during the reaction; in situ Raman spectra; pH effect on Fe₁ leaching; schematic and real setup of membrane system; MWCO method for the pore size determination; kinetic plots and pseudo-first-order linear fitting of 1,4-dioxane degradation with background constituents and real natural water; AC-HAADF-STEM after the longevity experiment; proposed configurations; DFT simulations (PDF)

†.

D.H. and W.W. contributed equally.

D.H., W.W., and J.-H.K. designed the research; D.H., K.R., and X.W. synthesized the catalysts and conducted performance tests; W.W. and J.C. performed DFT calculations. D.H., K.R., and E.S. conducted XAFS measurements; D.H. and J.N. conducted STEM measurements. D.H., J.L., and J.-H.K. analyzed data; D.H., W.W., J.C., and J.-H.K. wrote the paper. All authors discussed the results and commented on the manuscript.

The authors declare no competing financial interest.