Single-iron catalyst with heterogenized TEMPO for selective toluene electrooxidation to benzaldehyde at low potentials

Abstract

Toluene electrooxidation offers a sustainable route for the synthesis of valuable oxygen-containing chemicals, but existing methods often rely on strongly acidic or basic electrolytes, high cell voltages, or soluble mediators that complicate catalyst separation and increase waste. In this study, we show that combining grafted 2,2,6,6-tetramethylpiperidine N-oxyl (TEMPO) units with atomically dispersed iron sites in nitrogen-doped carbon enables efficient and selective electrooxidation of toluene to benzaldehyde under mild conditions. The catalyst operates at low potentials of 1.5 ~ 1.8 V vs. Ag/Ag⁺ without the need for added molecular hydrogen-atom-transfer reagents, soluble metal redox couples, or strong acids/bases. Mechanistic studies indicate that the grafted TEMPO not only participates in oxidation chemistry but also reshapes the carbon framework by introducing carbonyl groups, which modulate the electronic structure and lower the spin state of the iron centers. These coupled electronic and spin effects promote oxygen activation, thereby promoting the formation and release of reactive superoxide species that drive selective toluene oxidation. This strategy provides a cleaner platform for aerobic electrooxidation and offers a design principle for selective oxidation catalysis powered by electricity.

Selective electrooxidation of toluene under mild conditions is challenging. Here, the authors report grafted TEMPO on Fe–N–C catalysts modulates local electronic structure, lowers iron spin state, enhances O₂ activation, and promotes superoxide formation for toluene oxidation.

Introduction

The selective oxidation of C–H bonds in hydrocarbons to value-added oxygenates is an important transformation in the chemical industry for producing fine chemicals, pharmaceuticals, fragrances, etc^1–5. Among various hydrocarbons, the oxidation of toluene is particularly important in this scenario because of the wide range of industrial applications of various derived oxygen-containing compounds, such as benzyl alcohol, benzaldehyde and benzoic acid^6–9. Despite advances in catalytic performance, current methodologies rely predominantly on chlorination or thermocatalytic oxidation processes that involve stoichiometric oxidants (Fig. 1a)^10–14. These strategies require harsh reaction conditions and generate hazardous byproducts, thereby resulting in high energy consumption and significant environmental burdens. Moreover, achieving high activity without compromising selectivity remains intrinsically challenging. Therefore, the development of sustainable catalytic systems for selective toluene oxidation under mild conditions remains an important yet unresolved challenge.

Fig. 1. Pathways for toluene oxidation.

a Toluene chlorination hydrolysis approach and thermally catalytic oxidation. b Direct electrochemical processes. c Indirect electrooxidation processes with the use of molecular catalysts in the presence of additives. d Electrocatalytic toluene oxidation by electrocatalysts with heterogeneous metal redox pairs and molecular TEMPO as electron/spin modulators.

In comparison, electrochemical organic synthesis involves the use of electrons as effective redox reagents, which serve as green alternatives to toxic and strong oxidants, and offers distinct advantages in terms of sustainability, atom economy, and controllability, particularly under aerobic conditions at room temperature and atmospheric pressure¹⁵. Recently, the electrocatalytic activation of C‒H bonds has garnered extensive attention for the selective synthesis of high-value compounds^16–18. Notably, the methyl moiety in toluene, which is a ubiquitous structural motif in organic chemistry, requires a high redox potential for the effective activation of C–H bonds. This chemical inertness necessitates the development of advanced electrocatalysts that can lower activation barriers while maintaining precise selectivity.

Current methodologies for toluene electrooxidation mainly rely on two primary routes: direct and indirect pathways. The direct electrooxidation route typically requires a high voltage (>2.0 V vs. SCE)^19,20, which often induces the transient dissolution of most metal catalysts and concomitant oxidative degradation of electrolytes, solvents, and additives (Fig. 1b)^21–24. To overcome these challenges, indirect electrooxidation strategies have been developed using various molecular hydrogen-atom transfer (HAT) mediators (N-hydroxyphthalimide (PINO)^25,26, 2,2,6,6-tetramethylpiperidine N-oxyl (TEMPO)^23,27, etc.) and/or soluble metal redox pairs^28,29 (Co²⁺/Co^{3+ 25,30–33}, Mn²⁺/Mn^3+ 34–37, Fe²⁺/Fe^3+ 38,39, Cu⁺/Cu^2+ 40, etc.), which are defined as complementary systems that comprise metal species with different oxidation states as redox couples through reversible electron transfer. These mediators and catalysts generate reactive species that facilitate C–H oxygenation under strongly acidic/basic conditions (Fig. 1c). However, this indirect approach introduces significant challenges for catalyst separation, the purification of target products and the treatment of massive solid waste^27,41,42. In comparison, heterogeneous catalysts for toluene electrooxidation via the indirect pathway have been infrequently explored, despite their promising potential for practical applications by enabling catalyst recycling. Recently, a microfiltration membrane electrode loaded with Co₃O₄-based nanoarrays was reported to convert toluene (at a rate of 47.6%) with a benzaldehyde selectivity of 90% at catalytic current densities below 3 mA/cm² under ambient conditions⁴³. Nonetheless, achieving high toluene conversion rates inevitably reduces benzaldehyde selectivity because of overoxidation to benzoic acid. Therefore, developing heterogeneous electrocatalysts that can simultaneously improve activity and selectivity for toluene electrooxidation at low potentials remains important.

Herein, we present a rational design of a composite electrocatalyst (FeSA-T) composed of atomically dispersed Fe₁ sites anchored on N-doped carbon with surface-grafted TEMPO for the highly efficient and selective electrooxidation of toluene to benzaldehyde at low potentials (1.5 ~ 1.8 V vs. Ag/Ag⁺) with a current density of up to 36 mA/cm² (Fig. 1d). Notably, this catalytic system operated without molecular HAT reagents, soluble metal redox pairs, or strong acids/bases. FeSA-T achieved 96.1% benzaldehyde selectivity and 40.6% toluene conversion at 1.5 V vs. Ag/Ag⁺ over 4 h. The catalysts delivered a benzaldehyde yield of 91.2 mmol g_cat⁻¹ h⁻¹ with 74.6% toluene conversion and 81.4% benzaldehyde selectivity at 1.8 V vs. Ag/Ag⁺, thus outperforming state-of-the-art electrocatalysts and alleviating the trade-off between activity and selectivity. Structural characterization and mechanistic investigations revealed that the grafted TEMPO plays multiple critical roles: it acts as an oxidant to introduce more C=O moieties into the N-doped carbon, serves as an electronic regulator to induce significant surface charge redistribution of the electrocatalyst, and functions as a spin modulator to downregulate the spin state of Fe₁ sites, thereby increasing the capability for O₂ activation. Advanced spectroscopic studies (in situ Raman, in situ attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS), and electron paramagnetic resonance (EPR)) demonstrated that the spin-lowered monatomic Fe sites in FeSA-T effectively activated molecular O₂ to generate surface-bound *O₂⁻ species on the catalyst surface and readily released reactive O₂^•⁻ for C–H oxidation. On the basis of these findings, we elucidate the mechanisms by which TEMPO grafting modulates the electronic and spin states of atomically dispersed Fe, thereby facilitating selective toluene electrooxidation at low potentials. This study provides general design guidelines for sustainable C–H electrocatalysis under mild conditions.

Results

Synthesis and characterization of FeSA-T

The synthetic procedure for FeSA-T electrocatalysts is illustrated in Fig. 2a. First, atomically dispersed Fe anchored on N-doped carbon (FeSA) was prepared through a two-step approach, which involved the incorporation of Fe³⁺ ions into zeolitic imidazolate framework-8 (ZIF-8) to produce the iron-incorporated ZIF-8 (Fe@ZIF-8; Fig. S1) precursor and subsequent thermal treatments at 800 °C. Afterward, a typical solvothermal postmodification procedure was adopted to prepare FeSA-T by covalently grafting TEMPO onto the surface of FeSA.

Fig. 2. Synthesis and characterization of FeSA-T.

a Schematic illustration of the synthesis of FeSA-T. b HAADF-STEM image of FeSA-T. Atomically dispersed Fe single-atom sites are shown as bright dots. c Fe K-edge XANES spectra and d Fourier transform (FT)-EXAFS of FeSA, FeSA-T, and reference samples (Fe foil, FePc, and Fe₂O₃). e O K-edge XANES spectra and f C K-edge XANES spectra of FeSA and FeSA-T. The inset shows a schematic diagram of the coordination structure of TEMPO on FeSA-T. Source data for Fig. 2c–f are provided as a Source Data file.

Afterward, the ZIF-8 and Fe@ZIF-8 precursors underwent pyrolysis to volatilize the majority of the zinc species, which yielded nitrogen-doped carbon (NC) and FeSA electrocatalysts, respectively, with coordination-unsaturated N atoms serving as anchoring sites for Fe atoms^39,44. Synthetic details are provided in the Supporting Information. Transmission electron microscopy (TEM) and high-angle annular dark field-scanning TEM (HAADF-STEM) confirm that both the NC and FeSA preserve the rhombic dodecahedral morphology of the parent ZIF-8 structure (Fig. S2). The X-ray diffraction (XRD) patterns exhibit two broad reflections at ~24° and ~44°, which indicate graphitic carbon formation upon pyrolysis (Fig. S3). No peaks of metallic iron and/or iron oxides were observed for FeSA, thus indicating the high dispersion and low crystallinity of the iron species. Energy-dispersive X-ray spectroscopy (EDS) mapping revealed a uniform distribution of Fe, C, N, and O within the catalyst, as shown in Fig. S4. Using aberration-corrected HAADF-STEM with angstrom resolution, the observation of many isolated bright spots confirmed the presence of atomically dispersed Fe atoms in FeSA instead of Fe-based clusters and nanoparticles (Fig. S2c).

The FeSA catalysts were subsequently dispersed in dimethylformamide (DMF) with triethylamine (Et₃N) and TEMPO for chemical grafting under N₂. Compared with the pristine FeSA, the obtained FeSA-T electrocatalysts maintained structural integrity, as evidenced by their XRD patterns and TEM images, which revealed preserved morphology (Figs. S3, S5a). EDS mapping provided evidence that Fe was evenly distributed throughout FeSA-T (Fig. S5b). The atomically dispersed Fe atoms in FeSA-T were visualized by double aberration-corrected HAADF-STEM (Fig. 2b, S5c). The Fe loadings in FeSA and FeSA-T were 1.89 wt.% and 1.77 wt.%, respectively, as determined by inductively coupled plasma-atomic emission spectrometry (ICP‒AES). X-ray photoelectron spectroscopy (XPS) was used for the surface chemical analysis of the electrocatalysts. The N 1 s XPS spectra of FeSA and FeSA-T reveal five distinct nitrogen species in both catalysts: pyridinic-N (398.1 eV), pyrrole-N (400.6 eV), graphitic-N (401.9 eV), oxidized-N (403.6 eV), and Fe-N moieties (399.5 eV; Fig. S6). Notably, the oxidized-N fraction increased from 13.9% of FeSA to 18.7% of FeSA-T, which was attributable to successful TEMPO grafting by the introduction of additional N–O bonds in FeSA-T. Brunauer–Emmett–Teller (BET) N₂ adsorption/desorption isotherm measurements revealed hierarchical porosity in FeSA, which had a large surface area of 927.9 m²/g with mesopores of 3 ~ 4 nm and 10 ~ 50 nm. This porous structure enabled TEMPO to access the internal pores of FeSA for surface modification (Fig. S7). In comparison with that of FeSA, the reduced surface area of FeSA-T (744.9 m²/g) could be attributable to the occupation of grafted TEMPO inside the mesopores of FeSA. All the characterization results confirmed the successful surface functionalization of TEMPO on FeSA, with the atomic dispersion of Fe₁ sites preserved and the hierarchical porous architecture maintained.

Next, UV‒vis and EPR measurements were systematically performed to investigate the interactions between TEMPO and FeSA. The UV‒vis spectrum of FeSA-T revealed the absence of a characteristic peak at 430 nm (assigned to the N-O· nitroxide radical of free TEMPO; Fig. S8a), which provided evidence for covalent bonding between TEMPO and the N-doped carbon. The EPR spectrum of FeSA-T was essentially silent, in contrast to the paramagnetic signature of free TEMPO (Fig. S8b, S8c). These complementary spectroscopic observations further confirm that TEMPO was covalently grafted onto FeSA rather than being physically adsorbed or confined within pores.

To further elucidate the chemical states and coordination environments of iron in FeSA and FeSA-T, the Fe K-edge X-ray absorption spectroscopy (XAS) results for the two electrocatalysts were analyzed with Fe foil, Fe₂O₃, and iron phthalocyanine (FePc) as reference standards. X-ray absorption near-edge structure (XANES) spectra revealed that the absorption edge positions of both FeSA and FeSA-T were located between those of FePc and Fe₂O₃ and closer to that of FePc (Fig. 2c). This finding indicates that the average valence state of Fe in the two electrocatalysts is between that of Fe²⁺ and Fe³⁺ and is predominantly close to that of Fe²⁺. In the Fourier transform k²-weighted extended X-ray absorption fine structure (EXAFS) spectra of various samples (Fig. 2d), the main peaks of both the FeSA and FeSA-T electrocatalysts were at ~1.5 Å, which corresponded to the first Fe‒N coordination shell, similar to that of FePc. Notably, the absence of an Fe‒Fe signal at 2 ~ 3 Å was observed for FeSA and FeSA-T in comparison with the Fe foil, which confirmed the atomic dispersion of Fe in the two electrocatalysts. Quantitative EXAFS fitting yielded an Fe‒N distance of 1.98 Å and an Fe‒N coordination number of 4.3 for FeSA-T, which closely matched those of FeSA (Fig. S9 and Table S1). The wavelet transform (WT) plots further validated the negligible intensity ascribable to the Fe‒Fe path (2 ~ 3 Å in R space and ~9 Å in k space), which confirmed that the grafting of TEMPO did not alter the monoatomic dispersion of Fe₁ centers (Fig. S10). Synchrotron-based spectroscopic analyses confirmed the Fe–N₄ coordination structure in FeSA and the preserved local environment of the atomically dispersed Fe centers in FeSA-T after TEMPO grafting.

Chemical binding of TEMPO onto FeSA

Compared with those of FeSA, the preserved coordination environments of monoatomic Fe in FeSA-T confirm the absence of direct chemical bonding between TEMPO and the Fe₁ center. Therefore, elucidating the covalent interaction between TEMPO and FeSA is essential for deciphering their synergistic effects on the electronic structure of the electrocatalyst and subsequent catalytic performance. Typically, TEMPO, which has been used extensively as a two-electron oxidizing agent in organic catalysis^42,45, might oxidize FeSA. As revealed by the C 1 s XPS spectra, three types of C binding environments (C–C/C=C, C–N/C–O and O=C–C) were observed for both FeSA and FeSA-T (Fig. S11). Notably, compared with FeSA, FeSA-T contained a greater fraction of C=O species, which could be attributable to the oxidative ability of TEMPO toward graphitic carbon. This TEMPO-induced carbon oxidation was quantitatively confirmed by the O 1 s XPS spectra, where the C=O proportion increased from 32% in FeSA to 39% in FeSA-T (Fig. S12). O K-edge XANES spectra of FeSA and FeSA-T further validated this phenomenon, with FeSA-T demonstrating significantly increased π* (∼534.7 eV) and σ*(∼542.2 eV) transitions of C=O moieties in comparison with those of FeSA (Fig. 2e)^46–48. These spectroscopic findings confirm that TEMPO is an oxidizing agent that is enriched in C=O species within the N-doped carbon framework of FeSA-T, which is consistent with the results of the XPS analysis.

In addition, previous studies have reported that the hydroxyl groups in reduced graphene oxide can undergo nucleophilic attack on the nitrogen of TEMPO to form covalent C‒O‒N bonds via Meisenheimer rearrangement to achieve surface grafting of TEMPO onto graphitic carbon⁴⁹. The C K-edge XANES spectra of the two catalysts were analyzed to experimentally corroborate this covalent bonding between TEMPO and FeSA (Fig. 2f). The C K-edge XANES spectra of FeSA and FeSA-T exhibited two characteristic features of sp²-hybridized carbon: including a 1 s → π* transition peak at 286.8 eV and a σ* resonance peak at 293.8 eV⁵⁰. Among these two transitions, two additional features were observed: a peak at ~289.6 eV, which was assigned to oxygen-functionalized N-doped carbon species, and a peak at ~290.3 eV, which was ascribed to Fe-N-C coordination environments⁵¹. Notably, compared with FeSA, FeSA-T showed noticeable attenuation of the 289.6 eV (1 s → π*) transition intensity. The observed spectral transformation can be attributed to TEMPO grafting, which perturbs the conjugated carbon framework in FeSA-T through two synergistic pathways: (1) interfacial C–O–N bond formation, which disrupts sp²-hybridized carbon, and (2) the introduction of more methyl groups from the TEMPO moiety⁵². Herein, the lone pair of electrons on the nitrogen atom in TEMPO can improve the conductivity and electron transfer efficiency of the support, thereby increasing the overall electron supply capacity of the catalyst and promoting the redox cycle. In line with these findings, both the infrared spectrum and the I_D/I_G values derived from the Raman spectra confirmed that TEMPO grafting reduced the sp² content in FeSA-T (Fig. S13). Importantly, the significantly increased intensity of the C–O peak in FeSA-T provided direct evidence for the formation of interfacial C–O–N bonds (Fig. 2f). Concurrently, the increased C‒N peak intensity was attributable to the successful grafting of TEMPO, which introduced additional C‒N bonds. Thus, a collective analysis of Fe K-edge, O K-edge, and C K-edge XANES spectra revealed substantial chemical interactions between TEMPO and FeSA.

Electrooxidation of toluene

Given that TEMPO is covalently grafted onto FeSA via interfacial C‒O‒N bonds, the electrooxidation of toluene to benzaldehyde was used as a model reaction to assess the catalytic performance. Electrochemical measurements were performed in an O₂-saturated electrolyte that contained 0.1 M tetrabutylammonium hexafluorophosphate (But₄NPF₆) in a three-electrode system. This setup involved the use of a catalyst-coated carbon paper as the working electrode, a Pt foil counter electrode, and an Ag/Ag⁺ reference electrode. Cyclic voltammetry (CV) curves revealed that both the FeSA and FeSA-T electrocatalysts delivered significantly reduced overpotentials in the presence of toluene, which demonstrated their intrinsic activity for toluene oxidation (Fig. 3a). In contrast, NC resulted in negligible changes in the overpotentials upon toluene addition, thus emphasizing the critical role of atomically dispersed Fe₁ species in the oxidation process. Importantly, compared with FeSA (with a catalytic current density of 20.7 mA cm⁻²), FeSA-T achieved a significantly higher catalytic current density (36.4 mA cm⁻²) at 1.8 V vs. Ag/Ag⁺, which highlighted the essential contribution of TEMPO to the enhanced catalytic activity of FeSA-T.

Fig. 3. Electrocatalytic performance.

a Cyclic voltammetry curves (the current density normalized by the geometric areas of the electroactive materials) of the FeSA-T, FeSA and NC in 0.1 M But₄NPF₆ with or without toluene (40 mM) at a scan rate of 50 mV s⁻¹. b Catalytic performance of FeSA-T, FeSA and NC for toluene oxidation within a wide potential range from 1.5 V to 1.8 V vs. Ag/Ag⁺. Reaction conditions: 40 mM toluene, 4 h, and 25 °C. c Productivities of benzaldehyde normalized by the ECSA and those of various electrocatalysts. d Stability of the FeSA-T electrocatalysts over four testing cycles under the following conditions: an external potential of 1.8 V vs. Ag/Ag⁺, toluene (40 mM), 4 h, and 25 °C. Data are presented as mean ± SD (n = 3). The potentials shown are not iR corrected. Source data for Fig. 3 are provided as a Source Data file.

The catalytic performance of FeSA, FeSA-T, and NC for toluene electrooxidation was further evaluated at various external potentials over 4 h. Product analysis was performed using high-resolution mass spectrometry (HRMS) and high-performance liquid chromatography (HPLC) (Fig. S14). Typically, toluene conversion increased with increasing applied potential (Fig. 3b). Crucially, FeSA-T exhibited the highest catalytic performance over a wide potential range from 1.5 V to 1.8 V vs. Ag/Ag⁺. At a low external potential of 1.5 V vs. Ag/Ag⁺, FeSA-T delivered 40.6% toluene conversion and 96.1% benzaldehyde selectivity, which. significantly exceeded those of FeSA (32.9% conversion and 86.4% selectivity). This trend was also observed at all the tested potentials (Fig. 3b). The results of the filtration test support the origin of the catalytic activity from the heterogeneous FeSA-T catalyst rather than from potentially leached Fe species and/or molecular mediators in solution (Fig. S15a). To gain insights into the catalytic pathway, the time-dependent performance profiles of FeSA-T were recorded throughout a 4-hour toluene electrooxidation process (Fig. S15b–c). Toluene conversion increased with both higher potentials and longer operation times. However, the benzaldehyde selectivity decreased, which was accompanied by increased formation of benzoic acid. Despite this, FeSA-T maintained 74.6% conversion and 81.4% selectivity at 1.8 V vs. Ag/Ag⁺ and substantially outperformed FeSA (59.5% conversion and 71.1% selectivity) and NC (16.1% conversion and 45.2% selectivity). This comparison suggests that incorporating Fe₁ sites and grafting TEMPO facilitates toluene electrooxidation and alters the reaction pathway, thereby ultimately improving selectivity toward aldehydes as the primary products.

To further assess the intrinsic activity of the atomically dispersed Fe in the two catalysts, the electrochemical surface areas (ECSAs) of FeSA-T, FeSA, and NC were determined using the double-layer capacitance (C_dl) via cyclic voltammetry (Fig. S16). As shown in Fig. 3c, the ECSA values of FeSA (2.9 mF cm^-2) and FeSA-T (2.8 mF cm⁻²) were greater than that of NC (2.4 mF cm⁻²). After normalization to the ECSA, FeSA-T exhibited a benzaldehyde productivity of 255.9 μmol cm⁻²_ECSA h⁻¹, which was 1.5 and 7.1 times higher than those of FeSA (172.1 μmol cm⁻²_ECSA h^-1) and NC (35.9 μmol cm^-2_ECSA h⁻¹), respectively.

When normalized to the total catalyst weight, FeSA-T exhibited a benzaldehyde yield of 91.2 mmol g_cat⁻¹ h⁻¹, which was 1.3 and 8.3 times higher than those of FeSA (63.5 mmol g_cat⁻¹ h⁻¹) and NC (10.9 mmol g_cat^-1 h⁻¹), respectively, at 1.8 V vs. Ag/Ag⁺ (Fig. S17). These comparisons illustrate the significantly higher intrinsic activity of FeSA-T than that of FeSA. FeSA-T outperformed most previously reported catalysts for toluene electrooxidation in terms of both activity and selectivity (Table S2). Notably, the external potentials required for FeSA-T were significantly lower than those that were reported in previous studies to achieve similar catalytic current densities. FeSA-T also retained its activity over ten cycles, with a Faraday efficiency of ~50.8% at 1.8 V vs. Ag/Ag⁺ (Fig. 3d). Characterization of the spent catalyst confirmed the structural stability of FeSA-T during the electrooxidation process (Fig. S18).

Identification of the reactive intermediates in toluene electrooxidation

To elucidate the mechanism of toluene oxidation on the catalyst surface, the open-circuit potential (OCP) responses were analyzed to reveal the adsorption and activation characteristics of toluene molecules within the inner Helmholtz layer (Fig. S19)^53–55. After the introduction of toluene, the decrease in the OCP was most pronounced for NC (ΔU = 9.5 mV), whereas the decreases in the values for FeSA-T (ΔU = 7.4 mV) and FeSA (ΔU = 7.1 mV) were relatively small. Notably, all the values significantly exceeded that of the Fe foil (ΔU: 2.5 mV), which indicated substantial toluene adsorption and effective activation within the inner Helmholtz layers of FeSA and FeSA-T (Fig. 4a). This interpretation aligned with density functional theory (DFT) calculations, which revealed more thermodynamically favorable adsorption of toluene on FeSA-T (−0.64 eV) than on FeSA (−0.53 eV; Fig. S25). The reaction atmosphere was subsequently investigated to identify the active oxidant (Fig. 4b). Compared with its significantly lower activity under an Ar atmosphere, FeSA-T greatly increased toluene conversion under air and O₂; thus, O₂ was unambiguously identified as the essential oxidant. The even higher activity of the electrocatalysts under pure O₂ than under air was attributable to the increased concentration of dissolved O₂ in the electrolyte, which confirmed the oxidative role of O₂. However, excess O₂ promoted benzaldehyde overoxidation to benzoic acid, which indicated an optimal O₂ concentration for balancing benzaldehyde yield and selectivity. Previous studies have suggested that O₂ can be effectively activated on Fe₁ sites to generate active oxygen species in situ for oxidation^56–58. Hence, in situ Raman measurements were performed to provide intuitive insights into the reactive intermediates during toluene electrooxidation⁵⁹. As shown in Fig. 4c, the peak at 592 cm⁻¹ was attributable to the Fe‒O bond of FeSA-T. The doublet peaks at 1046 cm⁻¹ and 1095 cm⁻¹ were attributable to the adsorbed *O₂⁻ species, which strongly indicated its role as a reactive intermediate for toluene electrooxidation. After ten minutes of reaction, a peak at 1476 cm⁻¹, which was ascribed to the benzyl group, suggested the successful adsorption and activation of toluene on the catalyst surface. Additionally, the appearance of the C=O bond peak at 1665 cm⁻¹ validated the electrooxidation of toluene to benzaldehyde.

Fig. 4. Identification of the reactive intermediates.

a OCP changes of the FeSA-T, FeSA and NC electrocatalysts in 0.1 M But₄NPF₆ before and after the addition of toluene (40 mM). b Catalytic performance of FeSA-T for toluene oxidation under different atmospheres (Ar, air and O₂). Reaction conditions: toluene (40 mM), 1.8 V vs. Ag/Ag⁺, 4 h, and room temperature. Data are presented as mean ± SD (n = 3). c In situ Raman spectra of FeSA-T in 0.1 M But₄NPF₆ with toluene (40 mM) at 1.8 V vs. Ag/Ag⁺. Source data for Fig. 4 are provided as a Source Data file.

Identification of the role of TEMPO in toluene electrooxidation

The comparative catalytic performance of FeSA and FeSA-T highlights the critical role of grafted TEMPO in increasing both activity and selectivity for toluene oxidation. To elucidate the underlying mechanism, the Fe L-edge XAS profiles of both catalysts were analyzed to provide insights into how TEMPO grafting modulated the electronic structures of Fe₁ sites and influenced O₂ adsorption/activation (Fig. 5a). The increased peak intensity of FeSA-T indicated a spin-state transition of Fe sites upon TEMPO grafting, which induced a shift from the high-spin (HS) state characterized by high e_g filling to the intermediate-spin (IS) state with reduced e_g filling. Curie–Weiss fitting of the χ⁻¹-T curves yielded Curie constants of 3.8 and 2.6 emu·K·mol⁻¹ Fe for FeSA and FeSA-T (Fig. S20), which corresponded to effective magnetic moments of 5.5 and 4.6 μB per Fe atom, respectively. On the basis of the relation ${\mathrm{\mu }}_{\mathrm{eff}}=\sqrt{\mathrm{n}\left(\mathrm{n}+2\right)}$, the number of unpaired d-electrons (n) was estimated to be approximately 4.6 for FeSA and 3.7 for FeSA-T. The reduced μ_eff, weaker Curie tail and larger χ⁻¹ values observed over the entire temperature range for FeSA-T also indicated a suppressed spin state and weakened magnetic interactions compared with those of FeSA. Notably, this IS state of Fe₁ generates more vacant orbitals, thereby facilitating electron transfer and enabling a stronger interaction between d electrons of the Fe₁ site and π* antibonding orbitals of oxygen^60,61. Consequently, this electronic reconfiguration increases oxygen activation, thus demonstrating the pivotal role of TEMPO grafting in improving catalytic performance.

Fig. 5. Identification of the roles of TEMPO.

a Fe L-edge XANES spectra of FeSA and FeSA-T. In situ ATR-SEIRAS spectra of b FeSA-T and c FeSA systems. EPR spectra d over FeSA and FeSA-T in 0.1 M But₄NPF₆ without toluene and e over FeSA-T with toluene at 1.8 V vs. Ag/Ag⁺ using DMPO as a spin-trapping agent (O₂^•⁻ radicals (•) and carbon radicals (*)). Source data for Fig. 5 are provided as a Source Data file.

To probe the influence of TEMPO on O₂ activation, in situ ATR-SEIRAS experiments were conducted on both catalysts in 0.1 M But₄NPF₆ at 1.8 V vs. Ag/Ag⁺. As shown in Fig. 5b, FeSA-T exhibited a distinct peak at ~847 cm⁻¹ during the initial toluene-free stage, which was assigned to the Fe–O₂* intermediate^62–64. Importantly, the intensity of this peak increased markedly over one hour without toluene. In contrast, FeSA exhibited a significantly weaker peak intensity for the Fe–O₂* intermediate during the same interval, which suggested substantially improved O₂ adsorption/activation on TEMPO-modified electrocatalysts (Fig. 5b, c). Upon subsequent toluene addition, the Fe–O₂* signals of both FeSA-T and FeSA at ~847 cm⁻¹ in the spectroscopic profile were gradually consumed and stabilized afterward, which implied the consumption of Fe-O₂* intermediates by toluene. To further investigate the role of grafted TEMPO in O₂ activation, EPR measurements were conducted with 5,5-dimethyl-1-pyrrolidine-N-oxide (DMPO) as the O₂^•⁻ capture agent. During electrolysis in 0.1 M But₄NPF₆, the generation of active O₂^•⁻ intermediates was evidenced by the characteristic six-line signal of the DMPO-O₂^•⁻ adduct (Figs. 5d, S21)^65,66. Notably, compared with FeSA, FeSA-T produced a markedly stronger O₂^•⁻ signal, which suggested that TEMPO-induced low-spined Fe₁ sites facilitated the generation of more reactive O₂^•⁻ species. This observation aligned with the catalytic performance of the two electrocatalysts. The quasi in situ EPR spectra of FeSA-T revealed the continuous generation of O₂^•⁻ in the absence of toluene for a period of 1 h, which suggested the continuous accumulation of the adsorbed *O₂⁻ species on FeSA-T and the efficient release of active O₂^•⁻ intermediates for oxidation. These findings imply the ability of TEMPO to stabilize active O₂^•⁻ species⁶⁷. After the addition of toluene and the use of DMPO as a capture agent, compared with FeSA, FeSA-T resulted in stronger carbon radical signals (Figs. 5e, S22)⁵³, which were attributable to both the increased generation of O₂^•⁻ intermediates and the increased activation of the C(sp³)–H bond of toluene upon TEMPO grafting.

Next, to probe C(sp³)–H bond activation in toluene electrooxidation, kinetic isotope effect (KIE) experiments were performed on FeSA-T and FeSA. The KIE results revealed decreased current densities for both electrocatalysts when toluene was substituted with C₆D₅CD₃ (Fig. S23). If TEMPO acted as a redox shuttle, it would transfer electrons through its own redox process, which might involve hydrogen transfer between TEMPO and toluene. In this case, the cleavage of H atoms in toluene that interact with either TEMPO or the FeSA catalyst would be a key step. For the hydrogen atom transfer of TEMPO, C–H activation is significantly enhanced, which leads to a very large KIE value (>10)^68,69. Herein, both electrocatalysts exhibited lower KIE values of k_H/k_D for FeSA-T (1.31) and FeSA (2.10), which ruled out the possibility of TEMPO directly participating in hydrogen transfer. The slightly lower KIE value of FeSA-T was also consistent with the stronger adsorption energy of toluene on FeSA-T, which indicated the enhanced activation of toluene on the TEMPO-modified FeSA (Fig. S25). These findings were in accordance with the EPR results, thus illustrating the pivotal role of grafted TEMPO in toluene activation. Together, these results demonstrate that chemical grafting of TEMPO modulates the electronic structure of FeSA-T and promotes the activation of both O₂ and toluene.

Catalytic mechanism

To gain deeper insights into the catalytic mechanism of FeSA-T, differential pulse voltammetry (DPV) measurements and operando XAFS analyses were performed (Fig. 6a). The monitoring of iron valence state transitions under electrochemical conditions confirmed that Fe₁ was the active site for toluene electrooxidation^70,71. Prior to toluene addition, a single Fe²⁺/Fe³⁺ redox peak was observed at -0.15 V. When toluene was introduced, two distinct oxidation peaks emerged at ~-0.15 V and ~0.21 V, which corresponded to Fe²⁺/Fe³⁺ and Fe³⁺/Fe⁴⁺ redox couples, respectively. Compared with FeSA, FeSA-T resulted in a lower Fe²⁺/Fe³⁺ peak potential and a significantly larger peak area. Crucially, the Fe²⁺/Fe³⁺ potential for FeSA-T resided substantially below the onset potential for toluene oxidation, which suggested that Fe²⁺ was preoxidized into Fe³⁺ prior to toluene oxidation and that Fe^2⁺ was thereafter established as the active site for O₂ adsorption/oxidation. Conversely, the proportion of Fe³⁺ transition peaks in FeSA was significantly greater than that in FeSA-T during toluene oxidation, which implied a poor ability of Fe single atoms in FeSA to release active oxygen. Operando Fe K-edge XANES spectra of FeSA-T were further collected under a series of applied potentials that started from the open-circuit potential (OCP) in the presence of toluene (Fig. S26). In a blank electrolyte, the average oxidation state of Fe in FeSA-T was between those of Fe²⁺ and Fe³⁺, and both the pre-edge and main-edge positions gradually shifted to higher energies with increasing potential, which was consistent with mild electrochemical oxidation of the Fe sites. In contrast, in the presence of toluene, the XANES features displayed the opposite trend, with the absorption edge shifting toward lower energies under anodic polarization, which indicated a redox process coupled to substrate oxidation. These results demonstrate that toluene oxidation on FeSA-T proceeds via two parallel pathways, namely, the metal redox pair mechanism and the O₂^•⁻ radical approach. This synergistic mechanism originates from the improved O₂ activation of FeSA-T with grafted TEMPO.

Fig. 6. Catalytic mechanism.

a DPV on FeSA and FeSA-T electrodes in 0.1 M But₄NPF₆. DFT calculations. (b) ESP distribution of Fe^IIN₄C₁₀ (left) and TEMPO-Fe^IIN₄C₁₀ (right). The gray, blue, red, white, and brown spheres represent C, N, O, H, and Fe, respectively. PDOS of Fe d orbitals before and after O₂ adsorption: c Fe^IIN₄C₁₀ (left) and Fe^IIIN₄C₁₀-O₂ (right); d TEMPO-Fe^IIN₄C₁₀ (left) and TEMPO-Fe^IIIN₄C₁₀-O₂ (right). (e) Schematic plot of the O₂ adsorption mechanism. f DSPs of Fe^IIIN₄C₁₀-O₂ and TEMPO-Fe^IIIN₄C₁₀-O₂. Yellow and blue contours represent electron accumulation and depletion, respectively, between O₂ and Fe. The isosurface values are 0.003 eÅ^-3. Source data for Fig. 6a, c, d are provided as a Source Data file.

The above findings highlight the critical role of TEMPO in FeSA-T for electrocatalytic toluene oxidation. DFT calculations were subsequently conducted to investigate the catalytic contributions of TEMPO. Two optimized models were constructed: Fe^IIN₄C₁₀ (Fe₁–N₄ configuration) and TEMPO–Fe^IIN₄C₁₀ with TEMPO and a C=O moiety (Fig. S24). The optimized atomic coordinates derived from DFT calculations are provided in Supplementary Data 1. The electrostatic potential (ESP) distributions of the two electrocatalysts suggest that the C=O induced by TEMPO grafting, which acts as an electron acceptor, triggers noticeable charge redistribution (Figs. 6b, S27). The strongly electron-withdrawing C = O moiety is expected to decrease the electron density at the Fe₁ site through π*-antibonding interactions, thereby leading to an increased effective oxidation state of the Fe₁ sites in TEMPO-Fe^IIN₄C₁₀. Concurrently, TEMPO with oxidizing ability may further oxidize the Fe center via electron transfer, thereby synergistically increasing the valence state of Fe₁. Bader charge analysis quantitatively demonstrated that the Fe₁ center in TEMPO-Fe^IIN₄C₁₀ underwent electron depletion, with the net charge increasing from +1.10 |e | (Fe^IIN₄C₁₀) to +1.13 |e | (TEMPO-Fe^IIN₄C₁₀).

On the basis of experimental evidence of a TEMPO-induced spin-state transition from an HS configuration to an IS configuration (Fig. 5a), the O₂ adsorption on the Fe₁ centers in Fe^IIN₄C₁₀ and TEMPO-Fe^IIN₄C₁₀ to form Fe^IIIN₄C₁₀-O₂ and TEMPO-Fe^IIIN₄C₁₀-O₂ was analyzed by considering the spin states and the spin orientations of Fe₁ and O₂⁷². To model the spin transition of iron, the HS state of Fe₁ in Fe^IIN₄C₁₀ with a d-electron configuration of (d_xy)¹(d_xz)²(d_yz)¹(d_z²)¹(d_x²_-y²)¹ and the IS state of Fe₁ in TEMPO-Fe^IIN₄C₁₀ with a d-electron configuration of (d_xy)²(d_xz)²(d_yz)¹(d_z²)¹(d_x²_-y²)⁰ were explicitly used to calculate the adsorption geometries and energies of O₂, together with a comprehensive analysis of the d-orbital occupation (Table S3).

Projected density of states (PDOS) analysis revealed pronounced alterations in Fe d-orbital occupation upon O₂ adsorption (Fig. 6c, d, Tables S4–6). With respect to the HS Fe^IIN₄C₁₀, a d_xz-to-d_z² d–d transition and d_z²-to-2π* electron transfer were identified as O₂ approached the Fe₁ site (Fig. S28)⁷². Consequently, O₂ is predicted to preferentially binds to the Fe₁ site of Fe^IIN₄C₁₀ in an end-on configuration to form Fe^IIIN₄C₁₀-O₂ with a strong σ-bond between the Fe-d_z² orbital and the O₂-σ* orbital. This relatively strong interaction, however, may pose a challenge for the effective release of adsorbed *O₂⁻ species into active O₂^•⁻ intermediates. Conversely, for the IS TEMPO-Fe^IIIN₄C₁₀-O₂, some electrons are transferred from the fully occupied d_xz orbital of Fe₁ to the 2π* antibonding orbital of O₂ (Figs. 6d, e). Accordingly, O₂ adsorption is predicted to adopt a side-on configuration on the Fe₁ site in TEMPO-Fe^IIIN₄C₁₀-O₂, which results in a weaker d_xz–2π* antibonding interaction between the d_xz orbital of Fe₁ and the π* orbital of O₂. Spin-density isosurface analysis further suggests that the TEMPO-induced HS → IS transition suppresses excessive electron transfer to adsorbed O₂ and preserves its radical character, thereby facilitating the release of reactive O₂^•⁻ intermediates for oxidation (Fig. S29). Despite their distinct adsorption geometries and binding strengths, both Fe^IIN₄C₁₀ and TEMPO-Fe^IIN₄C₁₀ are capable of activating O₂ in the models. Notably, the IS TEMPO–Fe^IIIN₄C₁₀–O₂ is expected to facilitate favorable O₂^•⁻ release, which may contribute to enhanced toluene electrooxidation.

O₂ adsorption on the Fe₁ site induces a significant elongation of the O–O bond from 1.21 Å (free O₂) to 1.33 Å in both Fe^IIN₄C₁₀ and TEMPO–Fe^IIN₄C₁₀ (Fig. 6f), which provides evidence of the effective activation of O₂ in both systems. However, the charge difference and Bader charge analysis revealed that compared with electrons transferred in Fe_IIN₄C₁₀ (0.65 e), the Fe₁ center in TEMPO–Fe^IIN₄C₁₀ transfers fewer electrons (0.53 e) to the adsorbed O₂. These differences lead to a lower adsorption energy of activated oxygen species (E_ads, 0.21 eV) on TEMPO–Fe^IIIN₄C₁₀-O₂ with the IS configuration and a larger E_ads of -0.51 eV on Fe^IIIN₄C₁₀–O₂ with the HS state, which is consistent with previous reports⁷². This trend was also reflected by the Fe−O bond lengths: 2.13/2.14 Å for IS vs. 1.86 Å for HS. Consequently, TEMPO grafting modulates the spin state of the Fe₁ site from the HS to the IS, which may help balance O₂ activation and desorption, and potentially promotes the generation/release of reactive O₂^•⁻ for toluene oxidation. This optimized O₂ adsorption‒desorption equilibrium is consistent with improved performance of FeSA-T in toluene electrooxidation.

On the basis of the experimental and DFT results, two plausible indirect catalytic pathways for toluene electrooxidation catalyzed by FeSA-T are proposed, as illustrated in Fig. 7. Pathway I proceeds via a Fe²⁺/Fe³⁺ redox-mediated inner-sphere mechanism, where the Fe(III) site generated from the oxidation of Fe(II) in FeSA-T catalyzes the oxidation of toluene while it is being reduced to Fe(II). An inner-sphere single-electron transfer (SET) from toluene to Fe(III) generates a toluene radical cation, which subsequently undergoes deprotonation to form a •C₆H₅CH₂ radical; the latter can further react with oxygen to generate a peroxide and produce *C₆H₅CH₂OOH⁷³; *C₆H₅CH₂OOH decomposes to form benzaldehyde⁷⁴; and benzaldehyde is desorbed. The Fe(II) sites are reoxidized to Fe(III) at the anode, which completes the catalytic cycle. Pathway II follows a superoxide-mediated outer-sphere mechanism. Oxygen is activated on Fe(II) sites, which yields Fe(II)–*O₂ (Int-1). Afterward, charge transfer from Fe(II) to adsorbed *O₂ leads to the formation of adsorbed Fe(III)-*O₂⁻ species (Int-2). Subsequently, the released active O₂^•⁻ intermediates enable the oxidation of toluene, which is accompanied by the formation of an Fe(III) site (Int−3), as shown in Pathway II. Moreover, toluene electrooxidation could also take place on Fe(III) sites, accompanied by the regeneration of Fe(II) sites. DFT-calculated free-energy profiles (Fig. S30) reveal the coexistence of these two indirect pathways. Pathway I proceeds via a redox-mediated mechanism, wherein toluene activation initiates deprotonation to form a benzyl radical (•C₆H₅CH₂), accompanied by the generation of •OOH from O₂. In contrast, Pathway II involves a superoxide-mediated route in which O₂^•⁻ species oxidize adjacent toluene molecules, thereby yielding the same key radical intermediates. Both pathways converge after the initial toluene activation step. Interestingly, the energy difference in toluene adsorption is minimal, which suggests the coexistence of these two indirect pathways. Notably, compared with the redox-mediated pathway, Pathway II has consistently lower energy barriers for both the FeSA and FeSA-T catalysts. In particular, the rate-determining step associated with benzaldehyde desorption is significantly reduced on FeSA-T (0.46 eV) compared with that on FeSA (0.67 eV), which explains the experimentally observed increase in benzaldehyde selectivity and the suppression of overoxidation. These two catalytic pathways facilitate the highly efficient toluene electrooxidation of FeSA-T at low potentials.

Fig. 7. Proposed catalytic pathways for toluene electrooxidation.

The schematic summarizes the proposed reaction pathways for toluene electrooxidation over FeSA-T, including O₂ activation at the Fe site, electronic modulation by grafted TEMPO, and the selective formation of benzaldehyde.

Discussion

We have successfully designed a heterogeneous system that incorporates TEMPO as an electron and spin modulator and atomically dispersed Fe sites as a metal redox pair into N-doped carbon to achieve efficient electrooxidation of toluene at low potentials. TEMPO plays multiple roles: (1) as an oxidant to introduce additional -C=O moieties into the N-doped carbon, (2) as an electron modulator to regulate the electronic structure of the Fe₁ site, and (3) as a spin modulator to promote the activation of the C(sp³)–H bond in toluene. The resultant −C=O moieties and the electronic redistribution induced by TEMPO induce a spin-state transition of the atomically dispersed Fe in FeSA, thus shifting from the HS state to the IS state. This transition promotes the activation of O₂, generates more active *O₂⁻ species on the catalyst surface, and facilitates the release of O₂^•⁻ for oxidation. The activity and selectivity of FeSA-T at low potentials indicate the potential of this strategy, which involves the heterogenization of electron and spin modulators and metal redox pairs into N-doped carbon for the catalytic conversion of inert C–H bonds in toluene. This synthetic approach, which is characterized by sustainability under mild conditions with minimal use of molecular additives, also shows potential for the green transformation of other hydrocarbons.

Methods

Materials and Characterizations

Materials

Zinc nitrate hexahydrate (Zn(NO₃)₂•6H₂O, 99%), 2-methylimidazole, 2,2,6,6-Tetramethylpiperidinooxy (TEMPO, 98.5%), N, N-dimethylformamide (DMF, 99.5%), and sulfuric acid (H₂SO₄) were obtained from Energy Chemicals. Ferritin from Horse Spleen (Type I, saline solution) was purchased from Aladdin. Nafion 117 solution (5 wt.%) and methanol (CH₃OH) were purchased from Sigma-Aldrich. Tetra-n-Butylammonium hexafluorophosphate (TBABF₆, 99%) and toluene were obtained from Admas-beta and Sinopharm. Benzaldehyde, benzyl alcohol, benzoic acid, acetonitrile, and triethylamine (Et₃N) were received from Macklin Biochemical Technology Co., Ltd. All chemicals were directly used without any further treatment. Ultrapure water (18.2 MΩ) used in the experiments was supplied by a Millipore System (Millipore Q).

Characterizations

The crystal structures of various electrocatalysts were investigated with an X-ray diffractometer (XRD, Bruker D8 Advance with Cu Kα radiation, 30 kV, 20 mA). The X-ray photoelectron spectrometer (XPS, Thermo Scientific K-Alpha) was utilized to conduct qualitative and quantitative analysis of surface elements and the surface chemical states of various catalysts, in which all binding energies were referenced to the C 1 speak (284.8 eV). The compositions of various electrocatalysts were analyzed by inductively coupled plasma-mass spectrometry (ICP-MS, Agilent 7800). The Brunauer-Emmett-Teller (BET, Micromeritics ASAP 2460) method was applied to evaluate the specific surface area and pore-size distribution. Raman spectra of various catalysts were acquired on a Raman Spectrophotometer (Raman, HORIBA Scientific LabRAM HR Evolution, 532 nm). Fourier Transform Infrared Spectroscopy (FTIR) was performed on a Thermo Fisher Scientific Nicolet iS20. The soft X-ray adsorption experiment was performed at the BL10B beam line in the National Synchrotron Radiation Laboratory (NSRL) in Hefei. The hard X-ray adsorption experiment (XANES and EXAFS) was performed at BL14W1 beam line in Shanghai Synchrotron Radiation Facility (SSRF). The morphology and microstructures of as-synthesized catalysts were examined by transmission electron microscopy (TEM), selected area electron diffraction (SAED), and scanning transmission electron microscopy equipped with an energy-dispersive X-ray spectroscopy (STEM-EDS, FEI Talos F200X TEM, 200 kV). Electron paramagnetic resonance (EPR) spectroscopy was carried out on a Bruker EMXnano spectrometer.

Synthesis of electrocatalysts

Synthesis of zeolitic imidazolate framework 8 (ZIF-8) and N-doped carbon (NC) electrocatalysts

Thirty-two milliliters of an aqueous solution that contained 2.264 g of 2-methylimidazole was mixed with 8 mL of an aqueous solution that contained 108 mg of Zn(NO₃)₂·6H₂O. The mixtures were stirred for 30 min at room temperature (25 °C) and then centrifuged, washed with DI water three times, and dried. The obtained ZIF-8 powders were calcined in a tubular furnace at 800 °C for 2 h under a flowing nitrogen atmosphere (50 mL/min) with a ramping rate of 5 °C min⁻¹ to ensure that the NC electrocatalysts were identical and without ferritin.

Synthesis of Fe@ZIF-8 and atomically dispersed Fe atoms anchored on N-doped carbon (FeSA) electrocatalysts

Initially, 32 mL of an aqueous solution with 2.264 g of 2-methylimidazole was mixed with 80 μL of ferritin (as the Fe³⁺ ion source). Afterward, 8 mL of an aqueous solution that contained 108 mg of Zn(NO₃)₂·6H₂O was added. The mixture was stirred at room temperature for 30 minutes. Next, the resulting products were centrifuged off, washed three times with DI water, and dried at 80 °C for 12 h. The dried precursor was ground evenly, placed in a quartz boat, and calcined in a tubular furnace at 800 °C for 2 h under a flow of nitrogen (50 mL/min) with a ramping rate of 5 °C min⁻¹.

Synthesis of atomically dispersed Fe atoms anchored on N-doped carbon with grafted TEMPO as a HAT mediator (FeSA-T)

To obtain FeSA-T, 100 mg of as-obtained FeSA was dispersed in 12 mL of DMF with 0.5 g of dissolved Et₃N. Afterward, a solution of 100 mg of TEMPO in 8 mL of DMF was added dropwise under vigorous stirring. The reaction was carried out under a N₂ atmosphere for 4 hours to ensure functionalization. The functionalized FeSA electrocatalysts were washed with copious amounts of DMF and centrifuged. Finally, they were dried overnight to obtain FeSA-T at room temperature.

Electrochemical methods

All the electrochemical measurements were performed on an electrochemical workstation (CHI760E). All the potentials reported in this work are not iR corrected unless otherwise noted. Unless otherwise stated, the electrochemical measurements reported in this study were performed independently. Therefore, no standard deviations or error bars are provided for the corresponding performance metrics. The ink of the catalyst was prepared by thoroughly mixing 768 µL of Millipore Q water, 200 μL of ethanol, 32 μL of 5 wt.% Nafion solution, and 4 mg of catalyst under sonication. Afterward, the desired amounts of catalyst were deposited on carbon paper through drop casting and dried under ambient conditions. Electrocatalytic experiments were performed at 25 °C in an H-type electrolytic cell (10 mL, Gaoshiruilian) with a three-electrode configuration: gas-diffusion carbon paper electrodes loaded with 0.2 mg cm⁻² electrocatalysts served as the working electrodes, platinum foil served as the counter electrode, and a Ag/Ag⁺ reference electrode (0.01 M AgNO₃ in 0.1 M TBAPF₆-MeCN) was equipped with a Luggin capillary. The electrolyte consisted of 0.1 M TBAPF₆ in acetonitrile, with a Nafion-117 membrane (1.5 cm × 1.5 cm, thickness: 183 μm) used as a proton-exchange separator between the cathode and anode compartments. Before use, the membrane was sequentially boiled in 3 wt% H₂O₂, deionized water, 0.5 M H₂SO₄, and deionized water for 1 h each and then stored in deionized water until use. The cathodic solution was 0.5 M H₂SO₄. The electrolyte should be prepared and used immediately and stored at room temperature.

A non-aqueous Ag/Ag⁺ reference electrode equipped with a Luggin capillary was used in this work. All potentials reported in this work are referenced to Ag/Ag⁺ unless otherwise specified. For comparison, the reference electrode was calibrated against the ferrocene/ferrocenium (Fc⁺/Fc) redox couple under identical electrolyte conditions (0.1 M TBAPF₆ in acetonitrile). The half-wave potential (E_1/2) of the Fc⁺/Fc redox couple was determined from the average of the anodic and cathodic peak potentials:

$$E_{\frac{1}{2}}\left({\mathrm{Fc}}^{+}/\mathrm{Fc}\right)=\frac{E_{pa}+E_{pc}}{2}$$ 1

The measured E_1/2 (Fc⁺/Fc) was +0.0752 V vs. Ag/Ag⁺ (0.01 M); thus,

$$Evs.{\mathrm{Fc}}^{+}/\mathrm{Fc}=Evs.\mathrm{Ag}/{\mathrm{Ag}}^{+}-0.0752\mathrm{V}$$ 2

Prior to electrocatalytic testing, the electrolyte was saturated with Ar/air/O₂ gas by continuous bubbling for 30 minutes. During electrolysis, constant stirring was maintained to ensure efficient reactant gas mixing and mitigate concentration polarization effects. Cyclic voltammetry (CV) and chronoamperometric measurements were acquired at a scan rate of 50 mV s⁻¹. Electrochemically active surface areas (ECSAs) were determined on the basis of double-layer capacitance (C_dl) values obtained from CV analyses conducted at varying scan rates according to the following equation:

$$\mathrm{ECSA}=\frac{{\mathrm{C}}_{\mathrm{dl}}}{{\mathrm{C}}_{\mathrm{S}}}$$ 3

The linear slope of the double-layer charging current versus the scan rate is equivalent to twice the double-layer capacitance C_dl.

Analysis methods

For the quantitative analysis of the benzyl alcohol, benzaldehyde, and benzoic acid contents in the reaction product, the product was collected and analyzed using high-performance liquid chromatography (HPLC). The HPLC system was configured with an Eclipse Plus C18 column (250 mm × 4.6 mm, 5 µm particle size) and a UV‒Vis detector that was operated at 254 nm. Chromatographic separation was achieved using a mobile phase that consisted of 40% acetonitrile and 60% aqueous acetic acid (0.1% v/v) delivered at a flow rate of 0.5 mL min⁻¹. The toluene conversion (%), product selectivity (%) and Faradaic efficiency (FE) were determined according to the following equations:

$$\mathrm{Conversion}\left(\%\right)=\frac{\mathrm{moles}\mathrm{of}\mathrm{toluene}\mathrm{consumed}}{\mathrm{moles}\mathrm{of}\mathrm{initial}\mathrm{toluene}}\times 100\%$$ 4

$$\mathrm{Selectivity}\left(\%\right)=\frac{\mathrm{moles}\mathrm{of}\mathrm{product}}{\mathrm{total}\mathrm{moles}\mathrm{of}\mathrm{intermediates}}\times 100\%$$ 5

$$\mathrm{FE}\left(\%\right)=\frac{\mathrm{zFN}}{\mathrm{Q}}\times 100\%$$ 6

where z represents the number of electrons transferred to form one product molecule, F represents the Faraday constant (96,485 C mol⁻¹), N represents the amount of product formed (mol), and Q represents the total charge passed during electrolysis (C).

Temperature-dependent magnetization measurements

Temperature-dependent magnetization measurements were performed using a Magnetic Property Measurement System (MPMS) Superconducting Quantum Interference Device (SQUID) magnetometer. Zero-field-cooled (ZFC) magnetization curves were recorded in the temperature range from 2 to 400 K under an applied magnetic field of 1 kOe. The magnetic field strength was kept well below 3 T to ensure measurements within the linear response regime.

Operando spectroscopic methods

Operando electrochemical ATR-SEIRAS investigations

were performed using a Bruker INVENIO S FTIR spectrometer equipped with an MCT/B detector (EC-ATR-H, Beijing Scistar Technology Co. Ltd.) cooled by liquid nitrogen. All the spectral data were acquired in transmittance mode. Catalyst inks were deposited onto carbon paper substrates and allowed to dry prior to measurements. A three-electrode configuration that comprised an Ag/Ag⁺ reference electrode and a Pt counter electrode was used. The working electrode potential was initially maintained at the open-circuit potential (OCP) in 0.1 M TBAPF₆/acetonitrile electrolyte, followed by background spectrum acquisition. Potentiostatic measurements were conducted at 1.8 V (vs. Ag/Ag⁺) with 10-minute intervals between successive potential steps. Infrared spectra were continuously recorded in the 650–2000 cm⁻¹ range with 4 cm⁻¹ resolution and 64 scan coadditions per spectrum during each potential hold.

In situ Raman spectroscopic profiles

were obtained at ambient temperature using a Horiba Jobin Yvon HR Evolution system. The experiments were performed in an electrochemical cell under an ambient oxygen atmosphere without stirring. Samples were excited using a 532 nm continuous-wave (CW) diode laser that was delivered through a 50× Olympus microscope objective (0.50 NA) at a power of 10.0 mW. Ex situ laser parameters were preoptimized to prevent thermal damage to the samples, with spectral acquisition settings of an exposure time of 10 s and 2× signal accumulation. The system was equipped with a multichannel CCD detector (1024 × 256 pixels) and 1200 lines/mm grating. To ensure data reproducibility, multiple spectra were collected at distinct sample locations ( ≥ 3 positions per sample), and independent replicate samples with identical compositions were analyzed under equivalent experimental conditions.

Operando Fe K-edge XANES spectra

were obtained on beamline BL20U1 at the SSRF. In situ XAS measurements were performed using an electrochemical workstation (CHI 660E) coupled with a custom-designed photoelectrochemical (PEC) cell (Fig. S31). The PEC cell consisted of a carbon fiber paper (CFP) working electrode that was mounted on a copper frame, a platinum plate counter electrode, and an Ag/Ag⁺ reference electrode. For operando XAS measurements, the CFP electrode was assembled such that the side covered with a thin Kapton film faced the incident X-ray beam, whereas the catalyst-coated side was in direct contact with the electrolyte. The edges of the CFP were tightly fixed to the copper frame to ensure good electrical contact. To minimize X-ray attenuation and interference from the electrolyte and gas bubbles during data acquisition, the electrode–electrolyte interface was immobilized using an additional Kapton film, which was gently pressed toward the bare side of the CFP. Finally, the Kapton film was sealed using a flat compression plate with a coaxial elliptical aperture and secured by four screws to prevent electrolyte leakage during operando measurements.

Spin-polarized density functional theory (DFT) calculations

were carried out employing projector augmented-wave (PAW) method and the Vienna Ab initio Simulation package (VASP) code^75,76. The Perdew-Burke-Ernzerhof (PBE) and generalized gradient approximation (GGA) were utilized⁷⁷. The energy cutoff for the plane-wave basis was set to 400 eV. The TEMPO-Fe^IIN₄C₁₀ and Fe^IIN₄C₁₀ were used to model the FeN₄ center embedded in graphite surfaces. The optimized atomic coordinates derived from DFT calculations are provided in Supplementary Data 1.To simulate the surface, a 15 Å vacuum layer was added in the z-direction to avoid interaction between periodic images. The Brillouin zone was integrated on a 1 × 1 × 1 gamma-centered k-point mesh. Structural optimization assumed geometric convergence with a max atomic force <0.01 eV/Å. Bader charge analysis was used to analyze the electron gain and loss^78,79. The Van der Waals interactions between the intermediate and the electrocatalyst were described by using the DFT-D3 method with Becke-Jonson damping. The DFT + U method was applied to the 3 d orbitals of Fe to alleviate the unphysical electronic self-interaction errors. Based on Duan’s previous hybrid-functional-calibrated calculations⁷², the value of U_eff for the 3 d orbital of Fe was set to 3.29 eV.

In order to obtain different spin states, the occupation matrix control plugin for VASP was used to generate initial structures based on different metastable spin states⁸⁰. The FeN₄ spin state was first constrained by setting the d-electron occupation matrix and then relaxed. After that, the constraint was removed for the system to relax to the target metastable state. The relaxation was done using (ISMEAR = 0 and SIGMA = 0.05) and followed by a final single-point energy calculation without smearing (ISMEAR = −5). The orbital occupancy is extracted from the single-point calculation without smearing.

The Gibbs free energy change (ΔG) of every elemental step was calculated. The ΔG value can be determined as follows: ΔG = ΔE + ΔZPE - TΔS + ΔG_pH, where ΔE is the electronic energy difference directly obtained from DFT calculations, ΔZPE is the change in zero-point energies, T is the temperature (T = 298.15 K), and ΔS is the entropy change. Furthermore, when considering the effect of the pH of the solution on the reaction free energy (ΔGpH), the correction equation was ΔG_pH = 2.303 × k_BT × pH (or 0.059 × pH), where k_B is the Boltzmann constant.

Author contributions

X.T., Y.M. and Y.Q. conceived the original idea. Y. M. and Y. Q. directed the project. X.T., Y.L.Y.M. and Y.Q. wrote the manuscript. X. T. carried out the experiments and characterization. D. C. and C. S. assisted with the material fabrication. W. G. and Q. T. assisted with the electrochemical measurements. Y. L. performed the DFT calculations. Z. Du. assisted with the DFT calculations. Z. Da. performed the experiments and analysis of X-ray absorption. All the authors discussed the results and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The source data generated in this study are provided in the Source Data file. All the data that support the findings of this study are available within the article, its Supplementary Information, and the Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-72801-1.