Enhancing the ORR durability of single atomic Fe-N₄ active sites with implanted SiO₂ nanoparticles as radical and H₂O₂ inhibitors

Abstract

Highly efficient and durable single-atom catalysts (SACs) hold great promise for improving oxygen reduction reaction (ORR) in metal-air batteries and fuel cells. However, their long-term stability is challenged by the byproducts such as H₂O₂ and undesirable radicals. Herein, we report a Fe-N₄ active center-based SAC decorated with SiO₂ nanoparticles (NPs) as a radical scavenger, which was prepared using coffee grounds and industrial spent acid residue. The presence of SiO₂ NPs effectively suppresses the electrochemical H₂O₂ production, significantly improving durability with only a 5 mV half-wave potential loss after 30,000 voltage cycles in alkaline media. Electrochemical evaluations, in-situ characterizations, and density functional theory calculations reveal that the Fe-O-Si binding at the SiO₂–Fe-N₄ interface strengthens the binding of OOH* species, facilitating the 4-electron selectivity in ORR while inhibiting the formation of H₂O₂ and reactive oxygen species. Additionally, the SiO₂ NPs prevent the aggregation of Fe single atomic sites, thereby stabilizing the SAC active sites. Therefore, the incorporation of SiO₂ NP into Fe-based SAC offers a straightforward and effective strategy for enhancing ORR performance.

Introduction

The development of highly durable electrocatalysts for oxygen reduction reaction (ORR) is crucial for advancing fuel cells and metal-air batteries^1,2,3,4. Platinum-group metal-free (PGM-free) catalysts, recognized for their cost-effectiveness and abundant availability, have emerged as promising alternatives to Pt-based materials^5,6,7. Amongst single-atom catalysts (SACs), featuring transition metal-nitrogen coordination moieties on carbon materials (MNC, M = Fe, Co, Cu, etc.), have shown promising activity^8,9,10,11. However, their practical application remains hindered by insufficient catalytic stability, as active sites undergo deactivation and degradation under harsh operating conditions, such as high potentials and acidic environments^12,13,14. Recent advances argued that the ORR performance decline may stem from N-group protonation, demetallation, and electrochemical carbon corrosion^{15,16,17,18,19}.

In Fe–N–C SACs, one of the most extensively studied models, the high ORR activity originates from the dense, O₂-accessible Fe–N_x active sites. However, this is counterbalanced by poor stability due to the generation of reactive oxygen species (ROS), such as •OH and HO₂•, in close proximity to Fe–N_x moieties. These ROS can oxidize the carbon support to CO₂, leading to the demetallation of active metal centers (Fig. 1a, left)^15,20,21. Additionally, accumulated H₂O₂ further exacerbates degradation by producing highly reactive •OH radicals through Fenton-like reactions at Fe–N_x sites^22,23. Also, high local concentrations of H₂O₂ can promote carbon oxidation and Fe–N bond cleavage, further compromising stability^24,25. Therefore, mitigating ROS and H₂O₂ generation is essential for enhancing the durability of Fe–N–C SACs.

Fig. 1: The ORR pathway of proposed structure, radicals characterization, and stability measurements with and without scavengers.

a Schematic illustrations of ORR pathway over FeNC SAC with and without scavengers and (b) the proposed structure of SiO₂/FeNSiC SAC. c EPR spectra of the OH–DMPO• radical and (d) PL intensity of 7-hydroxycoumarins in the electrochemical measurement of SiO₂/FeNSiC and FeNSiC. e UV–Vis spectral variations of ABTS• + radical monitored under various conditions. H₂O₂ yield and n after 1 and 5,000 potential cycles of SiO₂/FeNSiC (f) and FeNSiC (g). ORR polarization curves before and after 30,000 potential cycles (0.6–1.0 V vs. RHE) in O₂-saturated 0.1 M KOH of (h) SiO₂/FeNSiC and (i) FeNSiC without iR-corrected. Source data are provided as a Source Data file.

To tackle the issues of FeNC SACs stability, passive shielding strategy focused on increasing carbon substrate graphitization to improve resistance against H₂O₂-induced damage, albeit often at the cost of Fe–N_x site density^26,27. Very recently, a proactive approach has gained traction, emphasizing the direct elimination of ROS and H₂O₂ to enhance Fe–N–C SAC viability in ORR (Fig. 1a, right). For example, Hu et al. incorporated Ta-TiO_x nanoparticles (NPs) as radical scavengers, significantly improving Fe–N–C SAC durability through suppression of H₂O₂ accumulation and fast decomposition of ROS via homolytic O–O bond cleavage at TaO₂–OH surfaces²⁸; Moreover, Other studies have also shown that CeO₂ NPs improve SAC robustness by leveraging the Ce³⁺/Ce⁴⁺ redox couple and oxygen vacancies to neutralize ROS and buffer oxidative stress^{29,30,31,32,33}. These oxides not only scavenge radicals but also modulate the local electronic environment, suppressing Fe demetalation and Fenton-type degradation through interfacial charge redistribution and interfacial bonding.

In this regime, efficient radical scavenging relies on the uniform dispersion and size control of the oxide NP scavengers; small-sized NPs, uniformly distributed in proximity to the Fe–N₄ center, can efficiently neutralize the free radicals^29,34. Thus, constructing scavenger NP–Fe–N₄ junctions in SACs via a robust manner, to efficiently suppress radicals, is highly desirable and appealing. Silica (SiO₂), bearing high stability and inertness, has been widely employed as a hard template or protective shield material against metal atom agglomeration in SAC fabrication^35,36,37. Typically, the SiO₂ is often removed by hydrofluoric acid (HF) etching to increase SAC porosity and active site exposure. Despite its well-established role in SAC fabrication, the potential influence of SiO₂ on M-N_x active moieties has, to our knowledge, been largely overlooked.

In this work, we demonstrate the previously unrecognized function of SiO₂ in enhancing Fe–N–C SAC stability by acting as a radical scavenger and H₂O₂ inhibitor in ORR. Inspired by the fact that Si is distributed uniformly throughout plant tissues mainly as a hydrated amorphous SiO₂ polymers^38,39,40, a simple hydrolysis process, utilizing biomass of coffee grounds (CGs) and industrial spent acid (ISA), was designed to produce SiO₂ NP-modified Fe–N–Si–C SAC (SiO₂/FeNSiC). The SiO₂ NPs, naturally formed from CG decomposition, were uniformly distributed along the carbon scaffold, ensuring proximity to the active sites. Electrochemical evaluations, dynamic in situ characterizations, and density functional theory (DFT) calculations collectively suggest that embedded SiO₂ NPs play a crucial role in suppressing H₂O₂ and radical generation during ORR.

Results and discussion

Catalyst fabrication and characterization

The Coffee grounds (CGs) primarily consist of cellulose, lignin, and hemicellulose, along with a small proportion of Si content (~3.66 wt%, as determined by inductively-coupled-plasma-optical-emission spectrometry (ICP-OES)). Given their natural abundance, sustainability, and cost-effectiveness, CGs serve as both carbon and silicon sources for SAC synthesis. Thanks to the homogeneous distribution of Si-containing substances in the tissue, small-sized SiO₂ NPs can be in situ formed and evenly dispersed in the carbon matrix upon the carbonization of CGs. Another precursor ISA (Supplementary Table 1), derived from the industrial steel cleaning process, is mainly comprised of hydrochloric acid (HCl) and FeCl₃. Our previous study has verified that the FeCl₃ in ISA, acting as Lewis acid, can favor the formation of nanopores within the carbon matrix, with Fe³⁺ ions embedded inside⁴¹. The noticeable porosity differences between CGs treated with and without ISA were visible through scanning electron microscopy (SEM) observation (Supplementary Fig. 1). Following ISA treatment, the CGs underwent pyrolysis in the presence of urea as a nitrogen source, yielding the final SiO₂/FeNSiC SAC (see schematic structure in Fig. 1b). To elucidate the role of SiO₂ in catalysis, FeNSiC (without SiO₂) and FeNC (lacking both SiO₂ and Si atoms) were also prepared for comparative study (see Materials Section for their synthesis).

To assess the •OH scavenging capability of the as-prepared catalysts with and without SiO₂, electron paramagnetic resonance (EPR) spin trapping experiments were conducted using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as a trapping agent through a Fenton reaction. As displayed in Fig. 1c, the EPR peak intensity is notably suppressed for SiO₂/FeNSiC compared to FeNSiC, indicating enhanced radical scavenging. Additionally, fluorescence-based •OH detection using a coumarin molecular probe revealed a lower photoluminescence (PL) intensity for 7-hydroxycoumarin in the presence of SiO₂/FeNSiC, further confirming its •OH scavenging function (Fig. 1d)²⁸. To evaluate ROS scavenging efficiency, ultraviolet-visible (UV-vis) spectroscopy was performed using 2,2’-azino-bis (3-ethylbenzothiazoline-6-sulfonic acid) diammonium salt (ABTS) as a probe. The SiO₂/FeNSiC catalyst exhibited a significantly lower absorbance at ~417 nm compared to FeNSiC, suggesting enhanced ROS elimination, possibly due to improved surface properties or catalytic activity (Fig. 1e)⁴². The impact of SiO₂ on H₂O₂ selectivity during ORR was further investigated using rotating ring-disk electrode (RRDE) measurements. As illustrated in Fig. 1f of the electron transfer number (n) and H₂O₂ yield before and after 5000 cyclic voltammetry (CV) manner, respectively, the H₂O₂ yield of the SiO₂/FeNSiC remained at a low level during 1 and 5000 potential cycles, with only a small increase from 6.7 to 8.8% at 0.2 V vs. RHE after cycling; meanwhile, the n after cycling is still retained at ~4, implying a good 4-electron transfer pathway during ORR. By stark contrast, the H₂O₂ yield over FeNSiC without SiO₂ is increased to ~14.1% from 7.9% of the initial value (Fig. 1g). Further continuous CV process for 30,000 cycles in an O₂-saturated KOH solution (0.1 M) between 0.6 V and 1.0 V showed the highest durability for SiO₂/FeNSiC after cycling, with a half-wave potential (E_1/2) loss of only 5 mV, much smaller than that of FeNSiC (14 mV), FeNC (18 mV) and Pt/C (28 mV after 5000 cycles) (Fig. 1h, i, Supplementary Fig. 2). The reduced H₂O₂ formation in the SiO₂/FeNSiC system likely contributes to its enhanced durability; notably, Si doping alone was insufficient to improve FeNC stability. Similar trends were observed under acidic conditions (Supplementary Fig. 3), where SiO₂-containing samples exhibited lower E_1/2 degradation (32 mV vs. 70 mV) and reduced H₂O₂ yield (5.0% to 8.9% vs. 7.9% to 15.3% at 0.2 V) after 5000 CV cycles, further validating the stabilizing role of SiO₂.

The structural traits of the SiO₂/FeNSiC were meticulously characterized. As depicted in Fig. 2a, b of the transmission electron microscopy (TEM) images, the sample exhibits a layered architecture interspersed with darker particles ranging from 5 to 20 nm in size. The aberration-corrected high-angle annular dark-field scanning TEM (HAADF-STEM) image (Fig. 2d) reveals numerous bright spots, indicative of heavier metal atoms, confirming the atomic dispersion of Fe sites within the structure. Additionally, energy-dispersive spectroscopy (EDS) mapping (Fig. 2e) demonstrates the uniform distribution of Fe, N, Si, O, and C across the carbon substrate. While HAADF-STEM confirms the atomic distribution of Fe, the composition of the darker particles observed in Fig. 2a, b required further identification. To address this, more precise analysis of electron energy-loss spectroscopy (EELS) was employed to detect the characteristic energy-loss edges of the possible key elements of the particles, which provides significantly higher energy resolution and sensitivity than the EDS to light elements (e.g., C, O, Si, etc.). As illustrated in Fig. 2c, the EELS analysis reveals that these darker particles are primarily composed of Si and O, as evidenced by the distinct O K-edge (~532 eV) and Si K-edge (~1850 eV), confirming that these particles are predominantly SiO₂ rather than carbon NPs or overlapping carbon layer fragments. These findings strongly verify that the sample comprises Fe–N–Si–C single-atomic sites accompanied by homogeneously dispersed SiO₂ NPs within the carbon matrix.

Fig. 2: Structural characterization of the SiO₂/FeNSiC sample.

a, b Images of TEM observation and (c) corresponding EELS analysis (C: red, Si: green, O: blue). d HAADF-STEM images of the SiO₂/FeNSiC (bright spots indicate the single metal atoms). e Images of Dark-field TEM observation and corresponding EDS elemental mapping for Fe (blue), N (cyan), Si (orange), O (yellow) and C (red) in the SiO₂/FeNSiC sample. Source data are provided as a Source Data file.

The chemical state of the SiO₂/FeNSiC, FeNSiC, and FeNC samples (Supplementary Fig. 4; Supplementary Table 2) was explored using X-ray photoelectron spectroscopy (XPS) measurement. The peaks at ~101 and ~103 eV of the Si 2p spectrum indicated by Supplementary Fig. 4a can be assigned to the –C–Si–O– bonds and SiO₂ in the SiO₂/FeNSiC sample, respectively; while the band belonging to the SiO₂ cannot be identified in the FeNSiC sample. The Fe 2p spectra (Supplementary Fig. 4b) agree with the coexistence of Fe²⁺ and Fe³⁺ species, with the Fe 2p_3/2 peak in SiO₂/FeNSiC shifting positively by +0.43 eV relative to unmodified Fe–N–C, suggesting electronic interactions between Fe and SiO₂. After deconvolution, the high-resolution spectra of N 1s (Supplementary Fig. 4c) reveal four peaks at 403.5, 401.0, 400.2, and 398.4 eV, ascribed to oxidation-N, graphite-N, pyrrole-N, and pyridine-N, respectively^43,44. The C 1s spectra (Supplementary Fig. 4d) show the primary chemical bonds of C–C, C–N, C–O, C = O, and C–Si in both the SiO₂/FeNSiC and FeNSiC samples, confirming the presence of Si species⁴⁵.

To overcome the low signal-to-noise ratio in XPS Fe 2p spectra, ⁵⁷Fe Mössbauer spectroscopy was employed for precise iron species identification (Supplementary Fig. 5). The spectrum of SiO₂/FeNSiC was resolved into two distinct doublets, corresponding to Fe–N_x coordination environments with Fe(III) existing in both high-spin (D1) and low-spin (D2) states⁴⁶. Raman spectra in Supplementary Fig. 6a present the characteristics of defect (D) and graphite (G) carbon of the samples. The I_D/I_G intensity ratios for all samples were calculated to be ~1.0, implying that the SiO₂ NPs scarcely affect the carbon structure in SiO₂/FeNSiC. Crystallographic analysis via X-ray diffraction (XRD) (Supplementary Fig. 6b) reveals no discernible peaks corresponding to Fe-based compounds, suggesting that Fe exists in a monoatomic state within the carbon matrix—consistent with TEM observations.

To further elucidate the atomic dispersion and local coordination environment of Fe in SiO₂/FeNSiC, we implemented extended X-ray absorption fine structure (EXAFS) and X-ray absorption near-edge structure (XANES) measurements. The Fe K-edge XANES spectra for SiO₂/FeNSiC, FeNSiC, and FeNC, alongside reference samples (Fe₂SiO₄, FeO, Fe₂O₃, Fe foil, and iron phthalocyanine (FePc)), are presented in Fig. 3a. The Fe species in SiO₂/FeNSiC exhibit a higher average oxidation state than metallic Fe and FeO, as indicated by the positive shift in the Fe K-edge absorption energy compared to Fe foil and FeO⁴⁷. Furthermore, the Fe–N coordination at the first shell can be clarified by the Fourier-transform EXAFS (FT-EXAFS) spectrum of the SiO₂/FeNSiC, which manifests one dominant peak at ~1.5 Å, with no Fe–Fe scattering visible at ~2.2 Å (Fig. 3b), effectively excluding the possibility of any Fe agglomerates in the sample^48,49.

Fig. 3: X-ray absorption spectroscopy characterization of SiO₂/FeNSiC and other samples.

a Fe K-edge X-ray absorption near-edge structure (XANES) and (b) Fourier transform k³-weighted Fe K-edge extended X-ray absorption fine structure (FT-EXAFS) spectra of SiO₂/FeNSiC, FeNSiC, FeNC, Fe₂SiO₄, Fe foil, FeO, Fe₂O₃, and FePc. c, d Experimental (black line) and fitted (red cycle) Fe FT-EXAFS fitting signals in R space of SiO₂/FeNSiC (the individual paths indicate unrefined theoretical scattering paths from the fitted FEFF model, indicating their baseline radial positions and qualitative relevance). Wavelet transform (WT)-EXAFS of the Fe K-edge of (e) Fe foil, (f) Fe₂O₃, (g) SiO₂/FeNSiC and (h) FeNC from top to bottom. Source data are provided as a Source Data file.

The EXAFS results were fitted to investigate the nearby coordination of the Fe center in SiO₂/FeNSiC (Fig. 3c, d; Supplementary Table 3). Combined with ⁵⁷Fe Mössbauer spectroscopy, the results suggest that Fe atoms in SiO₂/FeNSiC predominantly adopt a Fe–N₄ coordination geometry, characteristic of a classical Fe–N₄ structure^50,51. Interestingly, a secondary peak appears in the light at a high R value (ca. 2.2 Å) in the SiO₂/FeNSiC sample, indicating an additional definite path that should also be accounted for the surrounding coordination of metal centers. Considering the presence of Si and O elements in this sample, we analyzed the possible connection modes of Fe with O and Si atoms; the fitting outcomes in Supplementary Table 3 indicate no direct coordination between Fe and Si atoms in the outer shell of SiO₂/FeNSiC, as the fitted Fe-Si interatomic distance (2.94 Å) is too large to support direct coordination. Instead, the Fe and Si species likely interact through oxygen bridges, forming Fe–O–Si linkages. This conclusion is further supported by complementary XPS and FTIR analyses (Supplementary Fig. 7). The secondary peak at ~2.03 Å corresponds to Fe-O bonding, with a coordination number of ~1.02, reinforcing the presence of Fe–O–Si motifs in the sample^52,53. To further probe the local coordination environment, wavelet transform (WT) analysis of Fe K-edge EXAFS was performed (Fig. 3e–h). SiO₂/FeNSiC exhibits a dominant intensity maximum at ~5.0 Å⁻¹, characteristic of Fe–N scattering. Importantly, no Fe–Fe scattering signal (~9.0 Å⁻¹) is detected, corroborating the absence of Fe aggregates. The WT contour for SiO₂/FeNSiC closely aligns with that of FeNC, indicating a similar Fe–N coordination environment. However, a subtle shift toward the Fe₂O₃ position suggests the presence of Fe–O bonds, further validating the formation of Fe–O–Si linkages in SiO₂/FeNSiC^53,54.

The porosity of all the SAC samples was analyzed using N₂ adsorption/desorption measurement. Supplementary Fig. 8a shows that SiO₂/FeNSiC, FeNSiC, and FeNC samples possess type-I and type-IV nanopore arrangements, with a hysteresis loop at P/P₀ ≈ 0.9, indicating the presence of a well-developed micro-/mesoporous structure. Supplementary Fig. 8b, c present the pore size distribution and cumulative pore volume, calculated using a detailed quenched solid density functional theory (QSDFT) model. These SACs are predominantly occupied by micropores (<2 nm) and narrow mesopores (2–10 nm). Supplementary Table 4 summarizes the substantial surface area and total pore volume of all samples prepared with and without ISA treatment. Apparently, ISA treatment to CGs results in a significant increase of surface area (~1400–1500 m² g⁻¹) by constructing multiple micro-/mesopores, attributed to the important role of Fe³⁺ in ISA as a Lewis acid, which can facilitate the hydrolysis of cellulose chain for pore creation^41,55,56.

Electrochemical catalytic properties

In subsequent electrochemical catalytic tests, SiO₂/FeNSiC exhibited an E_1/2 of 0.892 V vs. RHE in linear sweep voltammetry (LSV), surpassing FeNSiC (0.875 V), FeNC (0.877 V), and Pt/C (0.863 V) (Fig. 4a). In addition, SiO₂/FeNSiC showed a high kinetic current density of 21.18 mA cm⁻² at 0.85 V (J_k, Fig. 4b), outperforming FeNSiC (11.6 mA cm⁻²), FeNC (16.92 mA cm⁻²) and Pt/C (7.02 mA cm⁻²). The favorable ORR kinetics of SiO₂/FeNSiC were evidenced by its low Tafel slope of 60.3 mV dec⁻¹ (Fig. 4c), smaller than that of FeNSiC (82.6 mV dec⁻¹), FeNC (62.2 mV dec⁻¹) and Pt/C (78.1 mV dec⁻¹). The H₂O₂ yield detected upon SiO₂/FeNSiC is fairly lower (~6%) than those of FeNSiC (7.9%) and FeNC (10.8%), and comparable to that of Pt/C (~5%) at 0.2 V vs. RHE, as illustrated in Fig. 4d, indicating a favorable 4-electron transfer process. The robustness of the SiO₂/FeNSiC was further confirmed through chronoamperometry (CA) tests at a high potential of 0.75 V vs. RHE. As displayed in Supplementary Fig. 9, the SiO₂/FeNSiC shows only a 7.4% reduction in current after 50,000 s, much better than the Pt/C with a 30% decline, indicating a commendable stability of the SiO₂/FeNSiC. Notably, the incorporation of SiO₂ not only enhances stability but also improves the ORR activity of the SACs. Furthermore, the ORR performance of FeNC-based samples, both with and without SiO₂, along with Pt/C, was evaluated in an O₂-saturated 0.5 M H₂SO₄ electrolyte (Supplementary Fig. 10). The J_k@0.8 V, E_1/2 and Tafel slope obtained from the LSV curves reveal that SiO₂/FeNSiC exhibits moderate ORR activity (J_k = 1.86 mA cm⁻², E_1/2 = 0.764 V, 77.3 mV dec⁻¹) compared to FeNSiC (J_k = 0.87 mA cm⁻², E_1/2 = 0.712 V, 126.8 mV dec⁻¹). However, despite these improvements, the performance remains inferior to commercial Pt/C (J_k = 9.26 mA cm⁻², E_1/2 = 0.82 V, 79.6 mV dec⁻¹), indicating further optimization is required for practical applications.

Fig. 4: Electrochemical tests of SiO₂/FeNSiC and other reference catalysts.

a ORR polarization curves, (b) half-wave potential (E_1/2) and kinetic current density (J_k) at 0.85 V vs. RHE, (c) corresponding Tafel plots, (d) peroxide (H₂O₂)% and electron transfer number (n) recorded on SiO₂/FeNSiC, FeNSiC, FeNC, and Pt/C in O₂-saturated 0.1 M KOH at 1600 rpm with scan rate of 5 mV s⁻¹ (pH = 13.0 ± 0.2, the resistance of the solution was 45 ± 5 Ω) without iR-corrected. Source data are provided as a Source Data file.

To investigate the potential dissolution mechanism of SiO₂ under alkaline conditions (0.1 M KOH), inductively coupled plasma (ICP) analysis was performed to quantify SiO₂ leaching during ORR stability tests (Supplementary Fig. 11). The results confirmed that Si dissolution was negligible. Furthermore, post-stability test characterization using TEM, EELS, and XPS (Supplementary Figs. 12, 13) revealed the continued presence of uniformly dispersed SiO₂ NPs, indicating their structural integrity and sustained functionality as reactive oxygen species (ROS) and H₂O₂ scavengers in 0.1 M KOH. To assess the accessibility of active sites, a critical factor in ORR performance, nitrite poisoning experiments were conducted on FeNC catalysts with and without SiO₂ (Supplementary Fig. 14a, b). CV analysis of the poisoned samples (Supplementary Fig. 14c, d) showed a slightly higher stripping charge for the SiO₂/FeNSiC catalyst compared to its SiO₂-free counterpart. The Fe–N_x site density (SD), quantified based on the stripping charge, exhibited comparable values for both catalysts (5.74 × 10¹⁹ site g⁻¹ vs. 5.13 × 10¹⁹ site g⁻¹). This confirms that the Fe–N_x active sites remain fully accessible in the presence of SiO₂, ruling out potential site-blocking effects and underscoring the compatibility of SiO₂ integration with the catalytic functionality of Fe–N_x centers. The SiO₂-modified sample exhibits higher turnover frequencies (TOF), corroborating its enhanced intrinsic activity. Moreover, SiO₂/FeNSiC demonstrates competitive overall performance compared to other advanced ORR electrocatalysts, as summarized in Supplementary Table 5.

In situ characterization for stability enhancement analysis

To gain deeper insights into the ORR mechanism and the behavior of single-atom Fe active sites in FeNC catalysts with and without SiO₂ under ORR conditions, we conducted in situ XAS measurements (Supplementary Fig. 17a). Upon applying ORR-relevant potentials (from 0.9 to 0.3 V vs. RHE), we observed a significant reduction in the white-line peak intensity and a gradual shift of the adsorption edge to lower energy levels. This indicates a continuous reduction in the valence state of the single Fe center as the reaction progresses^57,58. Notably, the near-edge spectrum obtained under open-circuit potential (OCP) closely resembled that of the initial catalyst state (Fig. 5a), suggesting that Fe–N_x sites remain coordinated to oxygen species under ambient conditions and OCP. This finding aligns with the expected thermodynamic stability of Fe–N_x configurations. The spectral evolution under applied potentials indicates that the reduction of Fe centers is primarily driven by cathodic polarization, facilitating the release of oxygen species from Fe–N_x sites^59,60. A comparison of XANES edge behavior between the two samples (Fig. 5a, d) reveals that the SiO₂-free FeNSiC exhibits a larger edge shift of 1.35 eV and a lower potential onset than the SiO₂-modified counterpart. This indicates that redox transitions of Fe sites begin earlier (i.e., at higher overpotentials) in the absence of SiO₂. This behavior is attributed to the stabilizing effect of the SiO₂ matrix, which electronically and spatially passivates Fe–N_x sites, delaying their reduction. In particular, the presence of SiO₂ likely modulates the local electron density and alters oxygen adsorption/desorption kinetics, effectively shifting the redox onset toward lower overpotentials. Such observations are consistent with previous reports on support-induced electronic effects in Fe–N–C systems^46,61,62,63. Further support is provided by DFT calculations presented later, which show that SiO₂ anchoring increases the demetalation energy barrier and modifies the electronic structure of Fe centers. While both samples were synthesized under identical conditions and show similar Fe loading and dispersion, the minor contributions from non-redox-active Fe species cannot be fully excluded. However, Mössbauer spectroscopy, XAS, XRD, and HAADF-STEM confirm that Fe is predominantly atomically dispersed in both samples. Thus, the observed edge-shift differences are more likely attributed to variations in the Fe–N_x environment and its interaction with the SiO₂ matrix, rather than to inactive Fe species.

Fig. 5: In situ XAFS and in situ ATR-IR measurements.

a In situ X-ray absorption near-edge structure (XANES) of the Fe K-edge, (b) the corresponding Fourier transform k³-weighted Fe K-edge extended X-ray absorption fine structure (FT-EXAFS) spectra and (c) In situ ATR-IR spectra were recorded on different potentials in 0.1 M KOH for FeNC SAC with and (d–f) without SiO₂ species (the change of bond length and intensity marked in the shadow area and arrow). Source data are provided as a Source Data file.

The EXAFS spectrum at the Fe K-edge for the SiO₂-containing sample showed that the Fe–N bond length remained relatively stable, with the main peak centered at a reduced distance (R ≈ 1.41 Å, Fig. 5b). At a lower potential of 0.3 V, the Fe–N peak exhibited a slight negative shift (~0.07 Å), indicating a minor contraction of the Fe–N bond. To further elucidate the structural evolution of single-atom Fe active sites at various operating potentials, we conducted curve fitting analysis of the first shell for the Fe center of the sample with SiO₂. As revealed in Supplementary Fig. 15, the Fe–N₄ configuration can be nicely preserved despite voltage variations. In contrast, the FT-EXAFS spectra of the SiO₂-free sample (Fig. 5e, inset; Supplementary Fig. 16; Supplementary Table 6) exhibited an increasingly pronounced second-shell feature. The corresponding fitting results showed a clear increase in the coordination number of Fe–Fe bonds from 0.18 to 2.5 with decreasing potential, suggesting potential aggregation of Fe centers under reductive conditions, matching the Fe–Fe distance in Fe₂O₃ (marked by the arrow). These results suggest that Fe sites may undergo partial migration or dimerization in the absence of structural stabilization^46,62,63,64. These findings highlight the crucial role of SiO₂ in preventing Fe aggregation and preserving the Fe–N₄ configuration during catalysis, thereby ensuring enhanced stability and performance.

Additionally, we conducted in situ attenuated total reflection-Fourier transform infrared (ATR-FTIR) measurements (see Supplementary Fig. 17b). Figure 5c reveals the emergence and intensification of major peaks at ~1453 cm⁻¹ and ~1295 cm⁻¹ (marked with dashed lines), along with a weak band at ~1373 cm⁻¹ (red arrow), upon the application of negative-going potential steps on SiO₂/FeNSiC (Fig. 5c) and FeNSiC (Fig. 5f). These peaks can be attributed to the O–O stretching mode of adsorbed O₂ molecules, the surface-adsorbed OOH* (OOH_ads), and the OOH bending mode of the surface adsorbed *H₂O₂ (HOOH_ads), respectively^65,66. Notably, compared to the SiO₂-free FeNSiC sample, the SiO₂/FeNSiC system exhibited a lower intensity of HOOH_ads species, indicating a reduced tendency for H₂O₂ formation. Therefore, it can be inferred that SiO₂ plays a critical role in suppressing H₂O₂ production during ORR, the efficiency of the desired 4-electron pathway and enhancing overall ORR kinetics^67,68.

Theoretical insights into the stability improvement

To explore the role of SiO₂ in promoting the stability of the Fe–NC catalyst, DFT calculations were implemented. In our theoretical model, we constructed a Si₁₇O₃₄ cluster and obtained its representative amorphous configuration by the ab initio molecular dynamic (AIMD) simulation at 1173 K (see Supplementary Fig. 18 for details; AIMD trajectories shown in Supplementary Data 1). Supplementary Fig. 19 indicates that there are under-coordinated O atoms on the surface of the SiO₂ cluster, which tend to bind with the Fe center of the Fe–N₄ site. A favorable binding model was attained with a strong binding energy of −2.84 eV between the Fe–N₄ site and the representative configuration of amorphous Si₁₇O₃₄, which was employed in subsequent calculations (Supplementary Fig. 20). As depicted in Fig. 6a and Supplementary Figs. 21, 22, in the absence of a SiO₂ cluster, Fe atoms readily aggregate into Fe₂, a process that is not only exothermic but also nearly barrierless upon disruption of the N₄ coordination site. In contrast, when SiO₂ NPs are present, Fe aggregation becomes an endothermic process with a high energy barrier exceeding 2 eV. This indicates that SiO₂ effectively stabilizes the Fe–N₄ structure by preventing Fe single-atom aggregation, even when the Fe coordination site is perturbed. These results align well with the observations from the in situ XAS investigations.

Fig. 6: DFT calculations to elucidate the stability of samples with and without SiO₂.

a DFT results of the energy change and energy barrier for the two Fe atoms aggregating into the Fe₂ NC with and without a SiO₂ cluster. Two H atoms were added on the damaged N₄ site to mimic the potential site break during the reaction. b The free energy profile for the ORR and the formation of H₂O₂. The projected density of state (PDOS) of OOH species over SAC (c) without and (d) with SiO₂ cluster. Source data are provided as a Source Data file.

The origin of the enhanced stability and activity of the SiO₂/FeNSiC catalyst in ORR was also investigated by DFT. The optimized 3D configurations of various ORR species (O*, O₂*, OH*, and OOH*) on the SAC with and without SiO₂ are depicted in Supplementary Fig. 23 (the atomic coordinates of optimized computational models are provided in Supplementary Data 1). The ORR free energy profiles in Fig. 6b and Supplementary Table 7 show that the free energy change of the H₂O₂ generation step (OOH* → H₂O₂) over the model with SiO₂ is 0.12 eV lower than that without SiO₂, indicating that the SiO₂ suppresses the generation of H₂O₂. Also, the strong binding between OOH* and Fe center in the model with SiO₂ impedes its further conversion to the harmful ROS and H₂O₂, as suggested in previous studies^69,70,71. Inhibiting the formation of attacking radicals and peroxide species could prevent the break of the Fe coordination site in the SAC^20,21,22. Also, the vital step from OH* to H₂O is largely facilitated by the structure with SiO₂ (Fig. 6b), suggesting that the Fe-O-Si binding between SiO₂ NP and Fe center also favors the ORR activity.

To understand why the Fe-O-Si bond at the trans site of OOH induces the strong interaction between OOH* and the Fe–N₄ center, the interfacial electronic structure was dissected by projection density of states (PDOS) analysis. As indicated in Fig. 6c, without SiO₂ the π orbital of OOH overlaps with the Fe 3d orbital, to be more precise, mainly the Fe 3d_yz and 3d_xz orbitals (Supplementary Fig. 24). When the SiO₂ NP binds at the trans site of OOH, the O 2p orbital in the Fe–O–Si bond could have a superposition with the OOH π orbital, the Fe 3d orbital and even the N 2p orbital as displayed in Fig. 6d, that makes the stronger binding of OOH. There is also a stronger electron exchange between the OOH radical and Fe–N₄ site when SiO₂ NP is presented, by comparing the 3D isosurfaces in the circles of Fig. 6c, d. Thus, the SiO₂-Fe binding reinforces the interaction between OOH* species and the Fe–N₄ site, inhibiting the transformation of OOH* species into hazardous radicals or H₂O₂.

ZABs and PEMFCs performance and stability

The performance of SiO₂/FeNSiC as the oxygen cathode electrocatalyst was evaluated in a home-built zinc-air battery (ZAB) (Fig. 7a). The open-circuit voltage (OCV) results of the ZABs in Fig. 7b indicate that SiO₂/FeNSiC provides a higher value than Pt/C (1.47 V vs. 1.42 V). The practical potential of the two ZABs in tandem for electronic devices was demonstrated in Fig. 7b inset. They can power a light-emitting display (LED; 1.5 V) consecutively for a week. Figure 7c demonstrates that the rechargeable ZAB based on SiO₂/FeNSiC possesses polarization curves with a smaller voltage gap between charge and discharge, compared to that using Pt/C, indicating a higher energy efficiency and better rechargeable ability. Also, the ZAB assembled with SiO₂/FeNSiC exemplifies a power density of 195.9 mW cm⁻², surpassing that of Pt/C (100.3 mW cm⁻²). The specific capacity of the ZAB, propelled by SiO₂/FeNSiC (Fig. 7d), stands at 765 mAh g⁻¹ at 10 mA cm⁻², comparable to that based on Pt/C (749 mAh g⁻¹). The SiO₂/FeNSiC electrocatalyst was durable and reliable in a mechanically rechargeable battery, only requiring replacement of the electrolyte and consumed zinc anode after each discharge. Noticeably, no significant degradation occurred after 4 rounds throughout 160 h at a current density of 5 mA cm⁻² (Fig. 7e). Furthermore, the SiO₂/FeNSiC-based battery (Fig. 7f and Supplementary Fig. 25) exhibited a consistently lower charge-discharge voltage gap and competitive long-term stability compared to batteries using Pt/C, FeNSiC, and FeNC without SiO₂. In light of potential electrolyte consumption, generated bubbles onto the cathode, and unstable interface arising from Zn anode passivation, we also evaluated the intrinsic activity of the SiO₂/FeNSiC catalyst using a flow ZAB (Fig. 7g and Supplementary Fig. 26). The SiO₂/FeNSiC-based flow battery delivered comparable durability than non-flow ZAB and other recent references (Supplementary Table 8). Notably, it maintained negligible round-trip efficiency loss even after 600 charge-discharge cycles over 200 h, underscoring the catalyst’s operational longevity.

Fig. 7: Performance of the SiO₂/FeNSiC catalyst in zinc-air batteries.

a Schematic illustration of the homemade ZAB at 6 M KOH solution with 0.2 M Zn(OAc)₂. b OCV plots (inset: photo image of an LED light array powered by two ZABs in series). c Polarization charge and discharge plots and power density curves obtained from corresponding discharge polarization curves and (d) discharge plots at a current density of 10 mA cm⁻² of the ZABs assembled with SiO₂/FeNSiC and Pt/C. e The constant current discharge curve of SiO₂/FeNSiC as an air cathode catalyst was used to evaluate the stability of SiO₂/FeNSiC. f Charging-discharging performance of the ZABs assembled with SiO₂/FeNSiC and Pt/C at 5 mA cm⁻². g Charging-discharging performance of the flow ZAB assembled with SiO₂/FeNSiC at 5 mA cm⁻². Source data are provided as a Source Data file.

Beyond ZAB applications, we also evaluated the catalytic performance of FeNC-based catalysts in H₂-air proton exchange membrane fuel cells (PEMFCs). As depicted in Supplementary Fig. 27, the initial polarization and power density curves, along with their corresponding performance metrics following a 100-h durability test, are analyzed. Under 1.0 bar H₂-air conditions, the SiO₂/FeNSiC catalyst achieved a peak power density (P_max) of 379.6 mW cm⁻², surpassing FeNSiC (P_max = 370.5 mW cm⁻²). Furthermore, during the 100-h stability test at a constant voltage of 0.5 V, the FeNSiC catalyst exhibited substantial performance degradation, whereas the SiO₂/FeNSiC catalyst maintained a much more stable electrochemical profile. This highlights the critical role of SiO₂ in enhancing both the stability and catalytic efficiency of the Fe–NC catalyst.

In summary, SiO₂ NP-decorated FeNC SACs have been synthesized using natural coffee grounds and industrial spent acid residues. The small SiO₂ NPs, derived from biomass, were uniformly distributed on the porous carbon substrate. The catalyst exhibits significantly improved stability and activity, thanks to the involvement of SiO₂ NPs in scavenging radicals and regulating the microstructure of the Fe centers. The binding of SiO₂ NP, and Fe sites prevents the aggregation of metal atoms and strengthens the interaction with OOH* species, eventually reducing the production yield of H₂O₂ during ORR, as validated by experiments and DFT calculations. This approach of using SiO₂ NPs as scavengers offers a promising strategy for designing more stable SACs for ORR and other systems involving H₂O₂ and oxidizing radicals.

Methods

Materials

Industrial spent acid (ISA) with a main composition of Fe³⁺ and HCl was obtained from Zhejiang Viersin New Material Co. Ltd., Jiaxing, China. Zinc acetate (Zn(Ac)₂, AR, 98%), ethanol, potassium hydroxide (KOH, AR, 99.99%), potassium chloride (KCl, AR, 99%), HF (AR, >40%), and HCl (AR, ~37%) were obtained from Sinopharm Chemical Reagent Co., Ltd. No additional purification was performed before using any of the chemicals. A Millipore water system with 18.2 MΩ·cm resistivity at 25 °C was used to produce deionized (DI) water. Coffee grounds (CGs) were collected from the waste of commercial products of GROUND ROASTED COFFEE, FULL BODIED, illy blend^TM (100% ARABICA), illycaffè S.p.A, Italy.

Catalyst synthesis

The CGs were first purified by repeated rinsing with DI water and then vacuum-dried at 80 °C for 72 h to remove residual moisture⁴¹. The resulting powder was treated in spent acid for 24 h, freeze-dried, and subsequently carbonized with urea (mass ratio; 1:1) in a N₂-purged tubular furnace at 900 °C for 2 h, with a heating rate of 5 °C min⁻¹. The pyrolyzed product was further leached in HCl for 12 h to dissolve unstable Fe clusters, thoroughly rinsed with DI water, and collected as the SiO₂/FeNSiC composite. Residual silica domains were eliminated by HF etching to obtain FeNSiC. A reference FeNC catalyst was prepared by following the same procedure, except that HF-pretreated CGs were used as the precursor to eliminate Si species prior to pyrolysis.

Electrochemical evaluation

Electrochemical measurements were conducted using a standard three-electrode configuration, with data acquisition performed by a CHI760e electrochemical workstation (CH Instruments, USA). A rotating ring-disk electrode (RRDE; Pine Instrument) with a 5 mm glassy carbon disk and a gold ring (6.5 mm inner, 7.5 mm outer diameter) served as the working electrode. The counter electrode is a graphite rod, while the reference electrode is Ag/AgCl, in which the internal electrolyte is a saturated KCl solution. All glassware and electrodes were rigorously pre-treated (including aqua regia cleaning and ultrapure water rinsing).

The catalyst ink was prepared by ultrasonically dispersing a mixture containing 5 mg of catalyst, 0.3 mL of H₂O, 0.7 mL of ethanol, and 40 μL of 5% Nafion solution for 30 min to achieve homogeneity. The resulting ink was drop-cast onto the electrode surface and dried to form a uniform catalyst layer. Electrochemical measurements were performed using cyclic voltammetry (CV) under N₂ or O₂ atmosphere (scan rate: 50 mV s⁻¹) and linear sweep voltammetry (LSV) in O₂-saturated freshly-prepared 0.1 M KOH electrolyte prepared with ultrapure water and stored in a sealed container at room temperature (pH = 13.0 ± 0.2; the resistance of the solution was 50 ± 3 Ω; scan rate: 10 mV s⁻¹; rotational speed: 1600 rpm). Acidic ORR measurements were performed in O₂-saturated 0.5 M H₂SO₄ (pH = 0.3 ± 0.2) with a solution resistance of 15 ± 2 Ω.

The electron transfer number (n) and the yield of the intermediate (HO₂⁻) during the ORR process were obtained from rotating ring-disk electrode measurements and calculated by the following equations:

$${{\rm{n}}}=\frac{4{I}_{{{\rm{d}}}}}{{I}_{{{\rm{d}}}}+\left(\frac{{I}_{{{\rm{r}}}}}{{{\rm{N}}}}\right)}$$

(1)

$$\%{{{\rm{H}}}}_{2}{{{\rm{O}}}}_{2}=\frac{200\left(\frac{{I}_{{{\rm{r}}}}}{{{\rm{N}}}}\right)}{{I}_{{{\rm{d}}}}+\left(\frac{{I}_{{{\rm{r}}}}}{{{\rm{N}}}}\right)}$$

(2)

where I_d, I_r, and N denote disk current, ring current and collection efficiency of Pt ring (0.37), respectively. All the recorded potentials were corrected to RHE using the following equation without iR-compensation:

$${E}_{{{\rm{RHE}}}}={E}_{{{\rm{Ag}}}/{{\rm{AgCl}}}}+0.059\times {{\rm{pH}}}+0.197$$

(3)

where ${E}_{{{\rm{RHE}}}}$ and ${E}_{{{\rm{Ag}}}/{{\rm{AgCl}}}}$ refer to the potential relevant to a reversible hydrogen electrode (RHE) and Ag/AgCl electrode, respectively.

The Fe–N_x site density in FeNC-SACs was quantitatively determined through nitrite (NO₂⁻) adsorption and reduction analysis following Kucernak’s methodology⁷². The experiments were conducted in three stages: (1) ORR LSV curves and nitrite reduction CV were measured without NO₂⁻ to obtain background data; (2) the same measurements were performed with NO₂^⁻ present; (3) additional experiments without NO₂⁻ were carried out to study catalyst recovery for ORR. The site density (SD) and turnover frequency (TOF) were calculated using the following equations^72,73:

$${{\rm{SD}}}\left({{\rm{site}}}\,{{{\rm{g}}}}^{-1}\right)=\frac{{Q}_{{{\rm{strip}}}}\times {N}_{{{\rm{A}}}}}{{n}_{{{\rm{strip}}}}\times F\times {m}_{{{\rm{cat}}}}}$$

(4)

$${{\rm{TOF}}}\left({{{\rm{s}}}}^{-1}\right)=\frac{{n}_{{{\rm{strip}}}}\times \Delta \;{j}_{{{\rm{k}}}}\left({{\rm{mA}}}\,{{{\rm{cm}}}}^{-2}\right)}{{Q}_{{{\rm{strip}}}}\left({{\rm{C}}}{{{\rm{g}}}}^{-1}\right)\times {L}_{c}\left({{\rm{mg}}}\,{{{\rm{cm}}}}^{-2}\right)}$$

(5)

$${j}_{{{\rm{k}}}}=\frac{{j}_{{\mathrm{lim}}}\times j}{{j}_{{\mathrm{lim}}}-j}$$

(6)

$$\Delta {j}_{{{\rm{k}}}}={j}_{{{\rm{k}}}\left({{\rm{unpoisoned}}}\right)}-{j}_{{{\rm{k}}}\left({{\rm{poisoned}}}\right)}$$

(7)

where ${Q}_{{{\rm{strip}}}}$ (C g⁻¹) is the excess coulometric charge from the stripping voltammogram, ${N}_{{{\rm{A}}}}$ denotes Avogadro constant (N_A = 6.02 × 10²³), F stands for the Faraday constant (F = 96485 C mol⁻¹), ${n}_{{{\rm{strip}}}}$ indicates the electron transfer number involved in NO₂⁻ reduction per active site (n_strip = 5), ${m}_{{{\rm{cat}}}}$ (g) is the mass of the catalyst, ${L}_{c}$ is the catalyst loading during the reversible nitrite poisoning experiments (0.27 mg cm^–2). To quantify the active site density, all measurements were performed in a 0.5 M acetate buffer at pH = 5.2.

Characterization

A JEOL-7800F SEM and Talos F200X TEM were used to image the morphology of the samples. HAADF-STEM was conducted at Shanghai, ThermoFisher Scientific. N₂ adsorption/desorption analysis was measured at 77 K on an AUTOSORB IQ Instrument, Quantachrome. XRD patterns were characterized using a Bruker D8 with the X-ray source being Cu-Kα radiation operated at 3 kV. ICP-OES was tested with Agilent 720ES. The stationary XAFS test was carried out at the BL08B2* of SPring-8 (8 GeV, 100 mA) in Japan, where the X-ray beam was mono-chromatized with a water-cooled Si (111) double-crystal monochromator and focused with two Rh-coated focusing mirrors with the beam size of 2.0 mm in the horizontal direction and 0.5 mm in the vertical direction around the sample position. XPS measurements were performed on ESCALAB Xi+. Raman spectra were collected with a Thermo Fisher DXR with a wavelength of 532 nm. PL spectra were conducted on an FS5 Spectrophotometer. The EPR test was determined using a Bruker A-300 spectrometer using DMPO as the trapping agent. In situ electrochemical ATR-FTIR was conducted on a Shanghai Linglu Instruments ElectroChem IR cell equipped with a Pike VeeMAX^TM III ATR. Data were collected with a Nicolet iS50 FT-IR spectrometer using a mercury-cadmium-telluride (MCT) detector cooled by liquid N₂, and Au-modified single bounce silicon served as a cathode. In situ XAFS spectroscopy was carried out using the RapidXAFS 2 M (Anhui Absorption Spectroscopy Analysis Instrument Co., Ltd.) by transmission (or fluorescence) mode at 20 kV and 20 mA, and the Ge (620) spherically bent crystal analyzer with a radius of curvature of 500 mm was used for Fe. Room-temperature ⁵⁷Fe Mössbauer spectra were recorded using a ⁵⁷Co/Rh source and a triangular velocity waveform spectrometer equipped with a NaI detector, with velocity calibration performed using an α-Fe foil. The optical absorption properties were characterized by UV–vis spectroscopy using a Shimadzu UV-3600i Plus spectrophotometer.

X-ray absorption spectra were processed using the ATHENA software package. For XANES analysis, the pre-edge background was removed, and spectra were normalized by extrapolating a quadratic polynomial fitted to the post-edge region, following the AUTOBK algorithm⁷⁴. EXAFS signals, χ(k), were extracted from the normalized spectra using standard procedures. To enhance contributions from higher-k components, the data were weighted by k³ before Fourier transformation, which was carried out using a Hanning window function to minimize truncation effects. Quantitative EXAFS analysis was performed using ARTEMIS⁷⁵, with theoretical scattering paths generated by the FEFF9 code⁷⁶. The Fe–N ([FeN_mp-6988_computed] N1.1) path was calculated using FeN crystal, the Fe–O ([Fe₂SiO₄] O24.1_) and Fe–Si ([Fe₂SiO₄] Si9.1_) paths were calculated using the fayalite (Fe₂SiO₄) crystal structure model, as it serves as a representative structural model for the Fe–O–Si coordination motif. These models provided reliable theoretical references for simulating the local atomic environments around Fe atoms. The fitting was performed over a k-range of 2.5–12 Å⁻¹. The amplitude reduction factor (S₀²) was fixed at 0.79 for Fe, determined from fitting metallic Fe reference foils, and coordination numbers were constrained based on the respective crystal structures used for each path.

EPR and FL spectroscopy for radical scavenging evaluation

Hydroxyl radicals (•OH) were generated through the Fenton reaction using 9 M of H₂O₂ and FeSO₄ as reactants, with 100 mM DMPO serving as the spin-trapping agent for radical identification. The reaction system was prepared by sequential addition of deionized water, H₂O₂, DMPO, and FeSO₄, followed by pH adjustment to ~3 using H₂SO₄. Comparative experiments were conducted without and with the SiO₂ scavengers. EPR measurements were immediately performed after reaction initiation using capillary-loaded samples. Parallel radical detection was achieved through a non-fluorescent coumarin-based molecular probe that undergoes conversion to highly fluorescent 7-hydroxycoumarin upon •OH oxidation. In a typical experiment, the 5 mL of coumarin solution (0.5 mM, dissolved in 0.1 M H₂SO₄) was introduced to acidic electrolytes after the chronoamperometric method (i–t) test, the temporal fluorescence evolution of coumarin derivatives was then quantified by monitoring intensity variations relative to baseline values using a fluorescence spectrophotometer. The radical scavenging activity of FeNC-based catalysts was assessed using ABTS as a probe. Catalysts (4 mg) were dispersed in 5 mL of 0.1 M H₂SO₄ by ultrasonication. ABTS• + radicals were initiated by combining 5 mL of 2 mM ABTS with 5 mL of 10 mM H₂O₂ in 0.1 M H₂SO₄ for 10 min, followed by 50-fold dilution. The catalyst suspension was then added to 2 mL of the diluted ABTS• + solution and reacted for 5 min. Radical scavenging efficiency was quantified through absorbance measurement at 417 nm using UV-vis spectroscopy.

Zinc-air battery measurements

The ZAB was constructed using 0.3 mm of polished zinc foil (1 × 3 cm) as the anode. The air cathode was fabricated by coating carbon paper with a catalyst ink containing 10 mg catalyst, 200 μL of 5% Nafion solution, 1.5 mL of ethanol, and 0.5 mL of H₂O, followed by 30 min ultrasonic dispersion and hot-pressing at 80 °C to achieve a catalyst loading of 2 mg cm⁻². For reference, commercial 20 wt% Pt/C catalysts were deposited onto carbon paper at a mass loading of 1.0 mg cm⁻² and employed as the air electrode in the ZABs. The electrode assembly was vacuum-dried at 60 °C for 2 h before testing. A mixed electrolyte of 6 M KOH and 0.2 M ZnCl₂ was employed. Polarization curves were acquired via LSV at 10 mV s⁻¹, while cycling stability was evaluated through galvanostatic charge-discharge tests at 5 mA cm⁻² (10 min charge/10 min discharge per cycle) using an electrochemical workstation (CS 2350 bipotentiostat; Wuhan Corrtest Instrument Co., Ltd).

The power density was determined using the following equation:

$${{\rm{power}}}\; {{\rm{density}}}={{\rm{current}}}\; {{\rm{density}}}\times {{\rm{voltage}}}$$

(8)

The specific capacity was derived from the following equation:

$${{\rm{specific}}}\; {{\rm{capacity}}}=\frac{{{\rm{current}}}\times {{\rm{test}}}\; {{\rm{hours}}}}{{{\rm{consumed}}}\; {{\rm{zinc}}}\; {{\rm{plate}}}\; {{\rm{mass}}}}$$

(9)

PEMFC tests

The Fe–N–C catalyst ink was formulated by dispersing 5 wt% of Nafion (Sigma-Aldrich) in isopropanol/deionized water (2:1, v/v) followed by sequential 2 h ultra-sonication and magnetic stirring to achieve homogeneity, then brush-coated onto carbon paper (5 cm², GDS2240, Ballard) with a loading of 3 mg cm⁻² as the cathode and vacuum-dried at 80 °C for 2 h before blade-coating onto a Nafion NR211 membrane, while the anode was fabricated with commercial 40 wt% Pt/C (0.2 mg_Pt cm⁻²). Fuel cell tests were conducted on an 850e station (Scribner) under controlled conditions: 2-h membrane hydration with humidified N₂ (100% relative humidity, 0.5 L min⁻¹) at 80 °C, followed by H₂/air operation (300/500 mL min⁻¹, 1 bar) at 80 °C with 100 h durability assessment at 0.5 V constant voltage.

Computational details

The DFT calculations were carried out with the Vienna ab initio simulation package (VASP) under periodic boundary conditions⁷⁷. The interactions between ions and core electrons were treated using the projector-augmented wave (PAW) approach, and the exchange–correlation energy was described by the generalized gradient approximation with the Perdew–Burke–Ernzerhof (PBE) functional^78,79. Long-range dispersion interactions were accounted for using the DFT-D3 correction proposed by Grimme⁸⁰. The simulation cell dimensions were 17.22 Å × 17.04 Å × 20.00 Å, for which Γ-point sampling of the Brillouin zone was adopted. A plane-wave cutoff of 400 eV was applied. Electronic self-consistent iterations⁸¹ were converged to 1 × 10⁻⁷ eV, and atomic relaxations were performed until the maximum force on each atom was below 0.05 eV Å⁻¹. Transition-state (TS) searches and migration barrier calculations were performed using the climbing-image nudged elastic band (CI-NEB) method⁸².

For the ORR process, based on the following steps, the four-electrons pathway as follows^83,84:

$${{{\rm{O}}}}_{2}\left({{\rm{g}}}\right)+\ast \to {{{\rm{O}}}}_{2}^{*}$$

(10)

$${{{{\rm{O}}}}_{2}^{*}+{{\rm{H}}}}_{2}{{\rm{O}}}\left({{\rm{l}}}\right)+{{{\rm{e}}}}^{-}\to {{{\rm{OOH}}}}^{*}+{{\rm{OH}}}$$

(11)

$${{{\rm{OOH}}}}^{*}+{{{\rm{e}}}}^{-}\to {{{\rm{O}}}}^{*}+{{\rm{OH}}}$$

(12)

$${{{{\rm{O}}}}^{*}+{{\rm{H}}}}_{2}{{\rm{O}}}\left({{\rm{l}}}\right)+{{{\rm{e}}}}^{-}\to {{{\rm{OH}}}}^{*}+{{\rm{OH}}}$$

(13)

$${{{\rm{OH}}}}^{*}+{{{\rm{e}}}}^{-}\to {{{\rm{OH}}}}^{*}+\ast$$

(14)

The equation of (10–14) represent the four successive electron-transfer steps involved in the ORR. The reaction free energy change (∆G) for each elementary step was evaluated using

$$\Delta G=\Delta H-{{\rm{T}}}\Delta S-{qU}+{{\rm{kBT}}}{ln}10\times {{\rm{pH}}}$$

(15)

where ∆H is the enthalpy change obtained from the reaction energy (∆E) corrected by zero-point energy (ZPE) within DFT calculations; The term T∆S accounts for entropy contributions; U is the applied electrode potential, and q denotes the number of electrons transferred per step (–|e|).