Aligned d-orbital energy level of dual-atom sites catalysts for oxygen reduction reaction in anion exchange membrane fuel cells

Abstract

The inherent scaling relationships between adsorption energies of oxygen-containing intermediates impose an intrinsic limitation on the maximum oxygen reduction reaction (ORR) activity, which represents one of the bottlenecks for the practical application of anion exchange membrane fuel cells (AEMFCs). To address this challenge, we align the 3dz² orbital energy levels of Fe and Co to selectively customize the dissociative ORR pathway without the formation of OOH* intermediates, circumventing the conventional OH*-OOH* scaling relations. This rational design is achieved by atomic phosphorus(P) substitution, which not only optimizes orbital matching towards O-O cis-bridge adsorption, but also stabilizes the spontaneously adsorbed OH ligand as an electronic modifier. Due to these attributes, the well-designed FeCo-N/P-C catalyst demonstrates ORR performance with a current density of 251 mA·cm^-2 at 0.9 V_iR-free under 1.5 bar H₂-O₂, showing a competitive performance with state-of-the-art Pt-free electrocatalysts and meeting the 2025 DOE target (44 mA·cm^-2). More importantly, the peak power density reaches as high as 0.805 W·cm^-2 under 1.5 bar H₂-air with negligible degradation observed over 10,000 cycles of voltage accelerated stress testing. This work offers a highly competitive electrocatalyst for AEMFCs and opens an effective avenue to bypass the constraints of linear scaling relations for ORR and beyond.

The inherent scaling relationships among the adsorption energies of intermediates limit the efficiency of oxygen electrocatalysis. Here, a dual-atom catalyst with aligned orbital energy level is developed, driving the dissociative pathway to bypass scaling relationships towards enhanced performance.

Introduction

Anion exchange membrane fuel cells (AEMFCs) are a promising alternative technology to proton exchange membrane fuel cells (PEMFCs) due to the potential independence on the costly and limited noble metal electrocatalysts that are considered as the major bottleneck of fuel cells^1–5. Decades of efforts have been devoted to developing high-performance and robust non-precious metal electrocatalysts for oxygen reduction reaction (ORR)^4,6–8. Among the various candidates, atomically dispersed metal-nitrogen-carbon (M-N-C) materials are hailed as the most promising alternatives owing to definite active site structure and their tunable electronic properties towards targeted catalytic performance^9–11. A diversity of strategies, including coordination number regulation¹², heteroatom substitution^13–15, axial ligand grafting^16–18, and second shell engineering^19–21, have been employed to enhance the intrinsic activity of M-N-C catalysts by disrupting electronic and/or geometric symmetry²². However, the preferential associative ORR pathway on the atomic M-N_x site not only imposes a fundamental performance limitation as a result of OH*-OOH* linear scaling relations, but also leads to insufficient long-term stability due to the poisoning effect of inevitable 2-electron byproducts, H₂O₂²³.

In contrast to the associative ORR pathway initiated from the traditional Pauling adsorption model on a single-atom site, a dissociative mechanism that promotes facile O-O Bond cleavage is more favorable on dual-atom sites. This reaction pathway avoids the formation of the OOH* intermediate, thereby circumventing the intrinsic scaling relationship and the potential poisoning effects, addressing both activity and stability associated with the single-atom M-N-C catalysts^24,25. Nevertheless, the dissociative mechanism necessitates strict requirements for the active site structure, favoring a bridge adsorption model²⁶. Naturally, the homonuclear diatomic site seems to be the best choice due to the optimal energy level alignment. The inhibited electron transfer between the same metal atoms, however, is unfavorable for tailoring the adsorption-desorption behaviors^27,28. Heteronuclear diatomic sites offer more opportunities to regulate the catalytic properties, but are criticized for the significant energy level difference between the two metal atoms, which would decrease the oxygen atomic orbital overlapping towards effective O₂ bridge adsorption^26,29. Therefore, minimizing the energy level difference (especially 3dz²) between the two metals is essential for facilitating O₂ bridge adsorption and thus enabling the precise customization of the dissociative pathway.

Herein, we propose a rational strategy that aligns the energy level of the heteronuclear Fe-Co diatomic sites by incorporating the electron-withdrawing P atom to customize the dissociative ORR pathway. The rationale behind the selection of P lies in its superior electron delocalization compared to N, which allows for better tailoring of the electronic states of the adjacent metal sites. This delicate design is accomplished by utilizing a space-confinement pyrolysis approach with a pre-constructed Fe-Co-P precursor, ensuring the preferential formation of the targeted heteronuclear Fe-Co diatomic site coordinated with a P atom. The bridge adsorbed *O-O* intermediate was monitored by in-situ Raman spectroscopy, confirming the dissociative mechanism on the well-designed site. Theoretical calculations further validate that the dissociative pathway is thermodynamically more favorable due to the aligned 3dz² energy level of Fe and Co. In addition to optimizing the orbital energies towards the bridge adsorption model, the P dopant induces spontaneous OH* adsorption, which tailors the adsorption energy of O* and OH*, thus lowering the energy barrier of the rate-determining step towards faster ORR kinetics. Benefitting from these effects, the as-prepared FeCo-N/P-C catalyst demonstrated significantly enhanced ORR activity, achieving a satisfactory current density of 251 mA·cm^-2 at 0.9 V_iR-free in an AMEFC prototype under H₂-O₂ mode. More impressively, it delivered a peak power density of 0.8 W·cm^-2 and negligible performance degradation during 10,000 cycles of accelerated durability tests under H₂-air conditions, demonstrating competitive performance compared to most state-of-the-art catalysts. These results strongly underscore the substantial potential of the developed FeCo-N/P-C catalyst for practical application.

Results

Electrocatalyst synthesis and characterization

The synthesis scheme of the FeCo-N/P-C catalyst is shown in Fig. 1a, where a preselected precursor strategy was employed. Initially, Zeolitic Imidazolate Framework-8 (ZIF-8), as the carbon and nitrogen precursor, was prepared. Subsequently, a coating of poly(cyclotriphospazene-co-p-phenylenediamine) (PZP) was applied onto the surface of ZIF-8, resulting in the formation of core-shell structured ZIF-8@PZP³⁰. As depicted in Supplementary Fig. S1, the X-ray diffraction (XRD) patterns of both ZIF-8 and ZIF-8@PZP exhibit similar diffraction peaks, confirming the crystal structure of ZIF-8 is retained after the incorporation of the PZP polymer. Following this, a preselected precursor (C₃₄H₂₈Cl₂CoFeP₂) was encapsulated within the hierarchically porous framework of ZIF-8@PZP, yielding FeCo-ZIF-8@PZP, which maintained the morphology of pristine ZIF-8 and ZIF-8@PZP (Supplementary Fig. S2). The FeCo-N/P-C catalyst was ultimately synthesized through the pyrolysis of FeCo-ZIF-8@PZP at 1000 °C under a flowing N₂ atmosphere. For comparison, we also prepared FeCo-N-C, Fe-N/P-C, and Co-N/P-C catalysts using similar methods. Driven by the Kirkendall effect³¹, phosphorus atoms from the PZP layer would diffuse inward into the polyhedral structure, while Zn²⁺ ions from ZIF-8 migrate outward during the pyrolysis process. Due to the smaller ionic radius of Zn²⁺ compared to phosphorus, the outward diffusion of Zn²⁺ occurs more rapidly than the inward diffusion of phosphorus, leading to the transformation from the solid polyhedral into a hollow structure of the pyrolyzed samples. Nitrogen adsorption-desorption analysis shows that FeCo-N/P-C possesses a significantly higher specific surface area (1088 m²/g) and a more porous structure compared to FeCo-N-C, attributable to the formation of the hollow structure (Supplementary Fig. S3). This hollow structure will enhance the exposure of active sites and facilitate mass transport, thereby potentially improving catalytic performance.

Fig. 1. Synthesis and characterizations of FeCo-N/P-C.

a Synthesis strategy of FeCo-N/P-C catalyst. b TEM image and (c) high-resolution TEM image of FeCo-N/P-C. d Aberration-corrected HAADF-STEM image of FeCo-N/P-C. e Atomic-scale EELS line scans on the selected dual-atom site in FeCo-N/P-C (inset: localized amplification HAADF-STEM image of FeCo sites). f Intensity profile across the dual-atom site in FeCo-N/P-C (inset: localized amplification HAADF-STEM image of FeCo sites). g High-resolution EELS elemental mapping of FeCo-N/P-C. h Intensity profile across Fe and Co elements (inset: mixing diagram of element distribution in Fig. 1g).

Further morphological characterization of FeCo-N/P-C was performed. Scanning electron microscope (SEM) images reveal that FeCo-N/P-C exhibits a more irregular morphology than FeCo-N-C, which is attributed to the etching effect of PZP on ZIF-8 during the pyrolysis process. Similar morphological changes were observed in Fe-N/P-C and Co-N/P-C (Supplementary Fig. S4). Closer observation by transmission electron microscopy (TEM) clearly shows the hollow structure of FeCo-N/P-C (Fig. 1b and Supplementary Fig. S5). No visible metal or metal oxide nanoparticles can be observed in the high-resolution transmission electron microscopy (HR-TEM) image of FeCo-N/P-C, implying the presence of atomically dispersed metal sites (Fig. 1c). Inductively coupled plasma-mass spectrometry (ICP-MS) analysis further shows that the Fe and Co content in FeCo-N/P-C is approximately 1 wt%, with the low metal content primarily existing in the form of single atoms, rather than forming metal clusters (Supplementary Table S1). A similar phenomenon is present in FeCo-N-C, Fe-N/P-C, and Co-N/P-C, and the P-containing samples exhibit a hollow structure (Supplementary Fig. S6–S8). This further validates the crucial role of the PZP coating layer in the formation of a hollow structure. The absence of metal nanoparticles is also confirmed by XRD analysis, where the patterns of FeCo-N/P-C and the reference samples display only two broadened peaks at approximately 24° and 44° (2θ), assignable to amorphous carbon (Supplementary Fig. S9), consistent with the defective carbon detected by Raman spectra (Supplementary Fig. S10). To directly observe the metal sites, we performed aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC HAADF-STEM). The AC HAADF-STEM (Fig. 1d) clearly shows paired bright dots homogeneously dispersed onto the carbon matrix, implying the formation of high-density dual-atom sites. Their composition was then checked by conducting atomic-scale EELS line scans (Fig. 1e) on a randomly selected dual-atom pair. The synchronized Fe L-edge and Co L-edge signals suggest a heteronuclear FeCo site. Atomic spacing analysis (Fig. 1f) reveals a 0.25 nm distance between the atoms. High-resolution EELS elemental mapping (Fig. 1g) shows a uniform distribution of Fe, Co, N, and P at the atomic scale. Element distribution analysis (Fig. 1h) further demonstrates an alternating distribution of Fe and Co, suggesting the predominant formation of heteronuclear Fe-Co diatomic configurations in FeCo-N/P-C catalyst. The precise identification of the active site structure will be obtained through a combination of advanced characterizations. A dual-atom site without P coordination is also confirmed in the sample of FeCo-N-C, whereas only a single-atom site can be observed in the samples of Fe-N/P-C, and Co-N/P-C (Supplementary Fig. S6–S13).

Electronic states and coordination environment

To further elucidate the composition and chemical states of the catalysts, X-ray photoelectron spectroscopy (XPS) was employed. The XPS survey spectrum confirms the presence of Fe, Co, N, P, and C elements (Supplementary Fig. S14). In the C 1 s XPS spectrum, four deconvolved components are located at 283.88 eV, 285.28 eV, 286.48 eV, and 288.98 eV, corresponding to sp²/sp³ carbon (C = C/C-C), C-N, C-P, and C = O, respectively, indicating the successful introduction of P and N atoms into the carbon framework (Supplementary Fig. S15). The formation of M-N species can be verified by the existence of a peak at 398.94 eV in the N 1 s XPS spectrum, where pyridinic nitrogen (397.93 eV), pyrrolic nitrogen (400.50 eV), graphitic nitrogen (402.00 eV) and oxidized nitrogen (404.20 eV) are present as well (Supplementary Table S2). Remarkably, the binding energy of M-N in FeCo-N/P-C shifts negatively by 0.16 eV compared with FeCo-N-C, which might be attributed to the electronic modulation effect of P (Supplementary Fig. S16). The M-P characteristic peaks at 127.9 eV (P 2p_3/2) and 129.3 eV (P 2p_1/2) in the P 2p XPS of FeCo-N/P-C imply the coordination of P with the metal atom (Supplementary Fig. S17 and Table S3). Notably, the P 2p_3/2 and P 2p_1/2 peaks shift to higher binding energies in the P 2p XPS spectrum of FeCo-N/P-C compared to those in Fe-N/P-C and Co-N/P-C, further evidencing the modified electronic structure of M-P in the as-designed dual-atom site. In conclusion, the binding energy shifts of M-N and M-P species coordinated with the metal indirectly suggest that the introduction of P atoms may regulate the electronic structure of the dual-atom sites.

X-ray absorption fine structure (XAFS) spectroscopy analysis was then carried out to determine the valence state and local coordination environment of Fe and Co in FeCo-N/P-C and reference samples. The X-ray absorption near-edge structure (XANES) spectra of Fe -K edge (Fig. 2a and Supplementary Fig. S18a) show a significant increase in the intensity of the pre-edge peak at 7118 eV, resulting from a 1 s → 3 d electronic transition after P incorporation, implying a broken symmetry of the D_4h plane of Fe sites in the FeCo-N/P-C analyst³². In addition, the Fe absorption edge (E₀) in FeCo-N/P-C is between those of FeO and Fe₂O₃, illustrating that the valence state of Fe is between +2 and +3 (Supplementary Fig. S18a). Quantitative linear fitting analysis (Supplementary Fig. S18b) shows that the valence state of Fe is +2.54, +2.32, and +2.17, respectively, in FeCo-N/P-C, FeCo-N-C, and Fe-N/P-C. This result confirms that P doping withdraws electrons from Fe in FeCo-N/P-C. Correspondingly, Fe L-edge also demonstrates increased oxidation state of Fe in FeCo-N/P-C, as the Fe L₃ edge (2p_3/2 → 3 d) displays a substantial positive shift compared to those of FeCo-N-C and Fe-N/P-C (Supplementary Fig. S18c)³³. In the Co K-edge XANES spectra (Fig. 2b and Supplementary Fig. S19a), P doping similarly increases the intensity of the pre-edge peaks, suggesting a significant disruption of the structural symmetry around Co atoms in FeCo-N/P-C. The Co absorption edge of FeCo-N/P-C lies between those of Co foil and Co₃O₄, indicating that the Co oxidation state is between 0 and +2.67. In sharp contrast to Fe, P incorporation decreases Co valence state in FeCo-N/P-C (+1.3), compared with the P-free FeCo-N-C counterpart (+1.8) and the single-atom Co-N/P-C counterpart (+2.1) (Supplementary Fig. S19b). Due to the partial substitution by P, the pyridinic N signal at 402.3 eV exhibits a slight negative shift (− 0.1 eV) compared to the reference samples, as shown in the N K-edge XANES spectra (Supplementary Fig. S19c). The aforementioned results indicate that P atoms exert the strongest electronic regulation on the FeCo diatomic sites. P doping extracts electrons from Fe and transfers them to Co, effectively demonstrating that P doping significantly modulates the 3 d orbital energy levels of both Fe and Co atoms.

Fig. 2. Chemical state and atomic local structure of FeCo-N/P-C.

a Fe K-edge XANES for FeCo-N/P-C and reference catalysts. b Co K-edge XANES. c k²-weighted Fe K-edge FT-EXAFS spectra for FeCo-N/P-C and reference catalysts. d k²-weighted Co K-edge FT-EXAFS spectra for FeCo-N/P-C and reference catalysts. e FT-EXAFS R-space fitting curves of Fe in FeCo-N/P-C. f FT-EXAFS R-space fitting curves of Co in FeCo-N/P-C. Wavelet transform of the k²-weighted EXAFS data for the Fe element of FeCo-N/P-C (g), FePc (h) and the Co element of FeCo-N/P-C (i) and CoPc (j).

Extended X-ray absorption fine structure (EXAFS) spectroscopy, that can provide atomic local structure, was then conducted. As shown in Fig. 2c, the k²-weighted Fourier transform EXAFS (FT-EXAFS) spectra of Fe K-edge reveal that FeCo-N/P-C, FeCo-N-C, and Fe-N/P-C exhibit a dominant peak at around 1.47 Å, corresponding to the Fe-N bond in the first coordination shell. Notably, this peak is broader in FeCo-N/P-C compared to FeCo-N-C, likely due to the contribution of the Fe-P bond. In addition to the Fe-N peak, a minor peak around 2.4 Å assigned to the Fe-Co bond is observed in the spectrum of FeCo-N/P-C, while the metallic Fe-Fe/Fe-Co peak at 2.15 Å is absent, confirming the exclusion of nanoparticles/nanoclusters in the sample. Similarly, the k²-weighted FT-EXAFS spectra at the Co K-edge (Fig. 2d) display a predominant peak at approximately 1.65 Å, indicative of the Co-N bond in the first coordination shell. The Co-N bond length in FeCo-N/P-C exhibits a notable shift (+0.35 Å) compared to FeCo-N-C, suggesting that the incorporation of P atoms alters the interaction between Co and N atoms. To obtain precise atomic structural information, FT-EXAFS fitting was performed for both Fe and Co K-edges. The fitting results (Fig. 2e, f, Supplementary Fig. S20 and Supplementary Table S4) reveal that Fe atoms are coordinated with three N atoms, one P atom, and one Co atom, with respective Fe-N, Fe-P, and Fe-Co bond lengths of 1.94 Å, 2.28 Å, and 2.47 Å. The presence of the Fe-Co bond was further confirmed by fitting the Co-Fe scattering path with coordination number and bond length of 0.9 and 2.48 Å, respectively, in the Co K-edge FT-EXAFS spectrum, consistent with the dual-atom distance observed in the AC HAADF-STEM images. Based on the fitting analysis, the atomic configuration of the dual-atom site was determined to be FeCoN₅P. To validate the EXAFS fitting results of FeCo-N/P-C, we performed time-of-flight secondary ion mass spectrometry (TOF-SIMS) analysis to examine the mass fragments released when FeCo-N/P-C is bombarded with an ion beam. Supplementary Fig. S21 shows the signal for the dual-atom fragment [FeCoN₅P]^- at m/z = 215.9, further confirming the presence of the Fe-Co bonded FeCoN₅P dual-atom sites. The P incorporation breaks the structural symmetry of FeCo-N/P-C and induces an asymmetric electron distribution at the FeCoN₅P active site. This perturbation in charge density at the Fe and Co sites leads to a significant redistribution of electron density within the 3 d orbitals of the Fe and Co atoms. The near-range coordination structure of the FeCo-N-C, Fe-N/P-C, and Co-N/P-C counterparts was identified as FeCoN₆, FeN₃P, and CoN₃P, respectively, according to the FT-EXAFS fitting results (Supplementary Fig. S22–S24 and Supplementary Tables S5–S7). Furthermore, wavelet transform (WT) analysis was applied to examine the EXAFS oscillations of Fe and Co atoms in FeCo-N/P-C and the reference samples. As shown in Fig. 2g–j and Supplementary Fig. S25, the WT spectra of FeCo-N/P-C, FeCo-N-C, Fe-N/P-C, and Co-N/P-C reveal a maximum at ~ 4 Å⁻¹, corresponding to Fe-N/Co-N interactions, while the Fe foil and Co foil exhibit a higher maximum at ~ 7 Å⁻¹ that is assignable to metallic Fe-Fe/Co-Co scattering. These results strongly confirmed that Fe and Co atoms are atomically dispersed in these samples³⁴.

Electrochemical oxygen reduction performance

The intrinsic activity and stability of FeCo-N/P-C and reference catalysts for ORR were evaluated using a rotating ring-disk electrode (RRDE) in an O₂-saturated 0.1 M KOH. The half-wave potential of pure N/P-C (0.85 V) is significantly lower than that of metal-containing samples, indicating that metal centers are the main contributors to ORR activity (Supplementary Fig. S26). As shown in Fig. 3a, linear sweep voltammetry (LSV) results demonstrate that FeCo-N/P-C exhibits an impressive ORR activity, with a half-wave potential (E_1/2) of 0.913 V (vs. reversible hydrogen electrode, RHE) and a diffusion-limited current density of 5.5 mA·cm⁻². This performance significantly surpasses those of FeCo-N-C (0.890 V, 5.0 mA·cm⁻²), Fe-N/P-C (0.896 V, 5.14 mA·cm⁻²), Co-N/P-C (0.845 V, 4.8 mA·cm⁻²), and 20 wt% Pt/C (0.874 V, 5.16 mA·cm⁻²). The high ORR activity of FeCo-N/P-C was further verified by showing a minimum Tafel slope of 42.1 mV·dec⁻¹ and the highest kinetic current density (8.6 mA·cm⁻² at 0.9 V), which is 1.9, 1.8, and 16.5 times higher than that of FeCo-N-C, Fe-N/P-C, and Co-N/P-C, respectively (Figs. 3b–d and Supplementary Table S8). Cyclic voltammetry (CV) curves in N₂-saturated electrolyte further reveal that FeCo-N/P-C possesses the largest electrochemical active surface area (ECSA) among the tested catalysts (Supplementary Fig. S27), which may consist of one of the factors contributing to the enhanced ORR activity. To exclude the effect of active site density on catalytic performance, we conducted the reversible nitrite poisoning experiments for quantitatively estimating the intrinsic activity of per active site (Fig. 3e and Supplementary Fig. S28)^35,36. As shown in Fig. 3f, the turnover frequency (TOF) of FeCo-N/P-C reaches as high as 5.40 s⁻¹at 0.9 V, much higher than FeCo-N-C (4.10 s⁻¹), Fe-N/P-C (3.97 s⁻¹) and Co-N/P-C (0.25 s⁻¹), reflecting the enhanced capacity of the FeCoN₅P site to catalyze ORR (Supplementary Table S9). We further examined the ORR pathway on the FeCo-N/P-C catalyst using the Koutecky-Levich (K-L) equation. As expected, the ORR proceeds through the ideal 4e^- reduction pathway on the FeCo-N/P-C electrode with electron-transfer number (n) of 3.9, exceeding the reference catalysts (Fig. 3b and Supplementary Fig. S29). This is consistent with the results obtained from RRDE testing, where a lower HO₂^- yield of less than 2% was detected in the range of 0.2–0.9 V, corresponding to the 4e^- reaction selectivity and first-order reaction kinetics of FeCo-N/P-C (Supplementary Fig. S30). To obtain a more accurate measurement of the electron transfer number, we further conducted RRDE tests at catalyst loadings of 0.3 mg/cm² and 0.2 mg/cm² (Supplementary Fig. S31). Our results show that the low-loading FeCo-N/P-C catalyst also demonstrates high 4e^- selectivity and a low HO₂^- yield. The high 4e^- selectivity and minimal HO₂^- yield are likely due to the direct O₂ dissociation on the well-aligned Fe-Co dual atom site during the reduction process, which we will discuss later. As well known, the desirable O₂ dissociative pathway would not only accelerate ORR kinetics, but also enable high stability due to the elimination of the poisoning effect of HO₂^-. To test the hypothesis, the catalytic stability of FeCo-N/P-C was investigated by accelerated stress testing (AST). Impressively, FeCo-N/P-C exhibits satisfactory long-term stability. After 30,000 continuous cycles, there is a negligible decrease in half-wave potential (ΔE_1/2 = 4 mV) and diffusion-limited current density (Fig. 3g). The ORR LSV curves on FeCo-N/P-C after different continuous cycles demonstrate that the insignificant performance degradation occurs at the initial 6500 cycles and then remains stable. By contrast, the 20 wt% Pt/C benchmark undergoes a serious performance decay with a negative shift (47 mV) in E_1/2 after 30000 cycles of operation (Fig. 3h and Supplementary Fig. S32). Structural characterization of FeCo-N/P-C after stability assessment reveals structural robustness, which contributes to the satisfactory ORR electrocatalytic stability. Both the metal coordination configuration and electronic states exhibit significant structural preservation, with no detectable degradation in active site geometry or charge distribution characteristics (Fig. 3i and Supplementary Fig. S33 and S34a).

Fig. 3. Electrocatalytic ORR performance evaluation in 0.1 M KOH.

a ORR polarization curves of FeCo-N/P-C, FeCo-N-C, Fe-N/P-C, Co-N/P-C, and commercial Pt/C in O₂-saturated 0.1 M KOH, measured at 5 mV·s⁻¹ and 1600 rpm. b ORR polarization curves of FeCo-N/P-C at various rotational speeds and K-L plots in O₂-saturated 0.1 M KOH at a scan rate of 5 mV.s⁻¹ (inset: K-L plot of FeCo-N/P-C). c Comparison of E_1/2 and kinetic current density at 0.9 V. d Tafel plots of FeCo-N/P-C and reference catalysts. e CV curves before and after nitrite adsorption in a 0.5 M acetate buffer at pH 5.2. f Comparison of TOF at 0.9 V. g stability of polarization curves of FeCo-N/P-C before and after 30 k potential cycles in O₂-saturated 0.1 M KOH. h Comparison of stability of FeCo-N/P-C and 20 wt% Pt/C. i Dissolution potential of FeCoN₅P sites.

Understanding of the catalytic enhancement mechanism

To shed light on the origins of the enhanced activity and stability of the FeCo-N/P-C catalyst, we employed in situ Raman spectroscopy to investigate the reaction mechanism by monitoring oxygen-containing intermediates during ORR (Supplementary Fig. S35). For comparison, the ORR dynamic processes on FeCo-N-C and Fe-N/P-C were also examined in the potential range from 1.0 V to 0.05 V (vs. RHE). Notably, a pronounced peak at 1145 cm⁻¹ attributable to bridge-adsorbed O = O appeared as the potential decreased in the Raman spectra of FeCo-N/P-C and FeCo-N-C in the oxygen-saturated electrolyte (Fig. 4a, c). This peak is absent in the spectra acquired in a nitrogen-saturated electrolyte environment (Fig. 4b), testifying the peak indeed stems from the vibrations of the oxygen molecule. We confirm that this peak arises from the vibrations of O = O adsorbed in the bridge configuration^37,38. Besides, the absence of the 1145 cm⁻¹ peak in FeN₃P spectra (Fig. 4d) indirectly verifies that this peak is generated by O = O vibrations at the bridging sites of the FeCo dimer. Further analysis revealed that the typical peak of bridge adsorbed O = O vibrations on FeCo-N/P-C is stronger than that on FeCo-N-C, suggesting a more favorable oxygen bridge adsorption on the FeCoN₅P site. These observations indicate that our engineered FeCo dimer structure can promote the bridge-adsorption model towards the ideal dissociative ORR pathway. It is noteworthy that a distinct peak at 845 cm⁻¹ was observed in the in-situ Raman spectra of FeCo-N/P-C upon immersing into the electrolyte and its intensity remained unchanged despite variations in potential (Fig. 4a). More interestingly, this peak is present in a 0.1 M nitrogen-saturated KOH solution, regardless of potential changes (Fig. 4b). Previous studies show that OH ligands spontaneously adsorb on FeCo diatomic sites^28,39,40. Therefore, the observed peak is attributed to this spontaneous OH adsorption. On the contrary, neither FeCo-N-C nor Fe-N/P-C exhibited the characteristic peak of OH adsorption (Fig. 4c, d), demonstrating the indispensable role of P substitution and dual-atom configuration in inducing and stabilizing OH adsorption. Such a unique active site structure can customize a desirable dissociative ORR pathway and stabilize the OH ligand towards tailored adsorption-desorption behaviors of ORR intermediates (Fig. 4e).

Fig. 4. In situ Raman characterizations for ORR catalytic mechanism under alkaline conditions.

a The Raman spectrum of FeCo-N/P-C obtained in a 0.1 M oxygen-saturated KOH solution. b The Raman spectrum of FeCo-N/P-C obtained in a 0.1 M nitrogen-saturated KOH solution. The Raman spectra of FeCo-N-C (c) and Fe-N/P-C (d) obtained in a 0.1 M oxygen-saturated KOH solution. e Schematic diagram of the oxygen reduction pathways for FeCo-N/P-C, FeCo-N-C, and Fe-N -C.

Furthermore, we carried out theoretical calculations to unveil the rationales of the tailored dissociative pathway and promoted OH adsorption, as well as their contribution to the enhanced ORR performance. The incorporation of P atoms significantly disrupts the symmetry of the electronic structure, thus resulting in a decrease in electron density at the Fe site and an increase at the Co site in FeCoN₅P (Fig. 5a and Supplementary Fig. S39). This alteration results in the localization of electrons within the Fe-Co-P triangular region, which is probably able to align the energy level of Fe and Co atoms. This hypothesis is subsequently verified by projected density of states (PDOS) analysis. As shown in Fig. 5b, PDOS reveals that P doping significantly enhances the density of states overlap between the Fe and Co 3 dz² orbitals, effectively reducing the energy gap between 3dz² orbitals of Fe and Co. Since the 3dz² orbital plays the dominant role in the hybridization between the metal 3d orbitals and O 2p orbitals (Supplementary Fig. S40–S42)^41–43, We further conducted an in-depth analysis of the interaction between the Fe/Co 3dz² orbitals and the O₂ 2p orbitals to elucidate the role of Fe/Co 3dz² energy levels in stabilizing the bridge adsorption configuration of O₂. In the undoped FeCoN₆ system (Fig. 5c and Supplementary Fig. S43), the energy level of O₂ π* antibonding orbital aligns closely with the Fe 3dz² orbital but resides lower than that of the Co 3dz² orbital. This leads to higher energy states for the σ* antibonding orbitals compared to the Fe-O counterpart (Fig. 5d). The imbalance between the Co-O interaction and the Fe-O interaction disrupts the cooperative adsorption on the dual-metal sites, thereby disfavoring the bridge configuration. Upon the introduction of P atoms (Fig. 5c, d, and Supplementary Fig. S43), the energy level of the Co 3dz² orbital downshifts towards the O₂ π* orbital. This adjustment effectively regulates the Co-O hybridization and promotes a more balanced adsorption of O₂ across the Fe and Co sites, thereby stabilizing the bridge adsorption configuration. This orbital energy tuning significantly reduces the O₂ adsorption energy (ΔE = − 5.24 eV) (Fig. 5e and Supplementary Fig. S44) and elongates the O = O bond length from 1.37 Å to 1.68 Å, facilitating the occurrence of an ideal dissociative pathway. This regulation directly facilitates the occurrence of the ideal dissociative pathway.

Fig. 5. Elucidation of the enhanced ORR activity.

a Charge density difference of FeCo-N/P-C. Projected density of states for FeCoN₅P-OH and FeCoN₆P-OH without O₂ adsorption (b) and with O₂ adsorption (c). d Schematic diagram of orbital hybridization of the O₂ π* orbitals and 3dz² orbitals of Fe and Co in FeCoN₆-OH and FeCoN₅P-OH. e Oxygen binding energy for FeCo-N/P-C and FeCo-N-C. f Adsorption energies of spontaneously adsorbed OH at different sites. g Free energy step diagram of the oxygen reduction reaction with and without OH adsorption. h Free energy step diagrams for FeCo-N/P-C and reference samples.

P doping not only promotes O₂ bridge adsorption but also significantly enhances the stability of spontaneously adsorbed OH (Fig. 5f and Supplementary Fig. S45). The adsorption energy of OH on the FeCoN₅P site (either Fe or bridge site) is more negative than that of the FeCoN₆ site, corresponding to higher thermodynamic stability of the OH ligand on the P modified site. Additionally, both the Gibbs free energy diagrams and the Pourbaix diagrams for FeCo-N/P-C and FeCo-N-C further confirm that the P incorporation is more effective to induce OH spontaneous adsorption and stabilize it under oxygen reduction reaction conditions (Supplementary Figs. S45 and S46). The OH ligand would tailor the adsorption-desorption properties of ORR intermediates on the FeCoN₅P site, switching the last electron transfer step (OH-H₂O) from endothermic to exothermic (Fig. 5g). Benefitting from the OH ligand modification and O₂ bridge adsorption model, the FeCoN₅P-OH site exhibits the lowest energy barrier (0.4 eV) of the rate-determining step (RDS) for ORR compared with P-free FeCoN₆ (0.99 eV), FeN₃P (1.83 eV) and FeN₄ (0.66 eV) sites (Fig. 5h). This analysis is consistent with the experimentally observed highest intrinsic ORR activity.

Applications in the PEMFC system

To evaluate the practical application of the as-prepared FeCo-N/P-C catalyst, we integrated it into a AEMFC with FeCo-N/P-C, PtRu/C and anion exchange membrane as cathode, anode and solid electrolyte, respectively (Fig. 6a). The FeCo-N/P-C-based membrane electrode assembly (MEA) demonstrates an impressive current density of 251 mA·cm⁻² at 0.9 V_iR-free under 1.5 bar H₂-O₂ conditions, exceeding the U.S. Department of Energy’s 2025 target of 44 mA·cm⁻² by more than five times and demonstrating competitive performance in AEMFC cathode catalysts. Due to the modified geometric and electronic structure of the P-doped FeCo dual-atom site, the hybridization between the O₂ 2p orbitals and Fe/Co 3 d orbitals is significantly enhanced towards a more desirable bridge configuration. This modulation directly facilitates oxygen activation, leading to faster ORR kinetics. As a result, FeCo-N/P-C as the cathode in the MEA not only shows a significant increase in current density at 0.9 V_iR-free but also achieves an impressive maximum power density of nearly 1 W·cm⁻² (Figs. 6b–d). Furthermore, the hollow structure of catalyst promotes the transport of O₂ and H₂O, further enhancing the performance of FeCo-N/P-C assembled in the AEMFC under H₂-air conditions, achieving a peak power density of 0.805 W·cm⁻², even exceeding the Pt/C benchmark (Fig. 6e). Furthermore, the tailored ORR pathway and stabilized OH ligand impart the better stability to the catalyst. There was no viable activity degradation after 10,000 cycles of AST under H₂-air conditions, indicating satisfactory long-term durability (Fig. 6f). Supplementary Fig. S47 further compares the performance of FeCo-N/P-C with the state-of-the-art non-precious metal-based electrocatalysts ever reported, clearly highlighting its advantage in terms of higher current density and peak power density, and underscoring the effectiveness of our energy level tuning strategy.

Fig. 6. Performance of AEMFC.

a Schematic illustration of the AEMFC. b, c H₂-O₂ polarization curves and power density of the MEAs with FeCo-N/P-C (b) and commercial Pt/C (c) as cathode catalysts (without iR-correction, cathode catalysts: 2.5 mg/cm²FeCo-N/P-C or 0.2 mg_Pt/cm²Pt/C, anode catalysts: 0.4 mg_Pt/cm²PtRu/C for both MEAs) measured at 80 ^oC, 100% RH, and 1.5 bar. d Tafel plot for FeCo-N/P-C and at 0.9 V_iR-free. e AEMFC tests under 1.5 bar H₂-air at 80 °C and 100% RH. f H₂-air fuel cell performance at begin-of-life (BOL) and end-of-life (EOL) of the FeCo-N/P-C. The AEMFC tests under 1.5 bar H₂-O₂ at 80 °C and 100% RH. Anode catalyst: Pt-Ru/C. Cathode catalyst: 20 wt% commercial Pt/C.

Discussion

In summary, we developed an orbital energy matching strategy through partially P substitution to enhance ORR kinetics. This strategy not only promotes O₂ bridge adsorption, but also stabilizes spontaneously adsorbed OH groups, as directly verified by in-situ Raman spectroscopy. These modifications facilitate a more favorable dissociative ORR pathway and optimize the adsorption energy of oxygen intermediates during ORR, thus leading to significantly enhanced intrinsic catalytic activity. As a result, the FeCo-N/P-C catalyst exhibits high performance under both H₂-O₂ and H₂-air conditions, achieving a peak power density of nearly 1 W·cm⁻² and 0.805 W·cm⁻² under H₂-O₂ and H₂-air, respectively. Notably, it delivers a current density of 251 mA·cm⁻² at 0.9 V_iR-free, surpassing the U.S. Department of Energy’s 2025 target by over five times. Furthermore, the catalyst maintains its high performance even after 10,000 cycles of AST, with virtually no degradation observed, highlighting its impressive long-term durability. These results not only validate the efficiency of our energy level tuning strategy but also position the FeCo-N/P-C catalyst as a highly promising candidate for future AEMFC applications, offering both satisfactory performance and long-term stability in real-world operating conditions.

Methods

Materials

All chemicals were purchased and used without further purification. Zinc nitrate hexahydrate (AR, 99%, Aladdin), 2-methylimidazole (98%, Aladdin), Cobalt nitrate hexahydrate (99%, Aladdin), Ferric nitrate nonahydrate (99%, Xilong scientific), [1,1’-bis(diphenylphosphino)ferrocene]dichlorocobalt(II) (98%, Aladdin), Phosphonitrilic chloride trimer, Sodium borohydride (98%, Aladdin), p-Phenylenediamine (AR, 97%, Aladdin), KOH (EL, 99.99%).

Synthesis of ZIF-8

In a typical synthesis of ZIF-8, Zn(NO₃)₂.6H₂O (3 g) and 2-methylimidazole (6.5 g) were dissolved in 40 mL and 80 mL methanol, respectively. Then, two solutions were mixed with vigorous stirring at room temperature. Subsequently, slowly add NaBH₄ (2 g) to the mixed solution. Finally, the mixture was maintained at room temperature for 10 min. The resulting precipitate was centrifuged, washed with ethanol and dried at 60 °C in a vacuum oven.

Synthesis of ZIF-8@PZP and FeCo-ZIF-8@PZP

In a typical synthesis of ZIF-8@PZP, 600 mg of the as-obtained ZIF-8 and 224 mg of p-phenylenediamine were dispersed in 105 mL and 150 mL tetrahydrofuran (THF), respectively. Then, two solutions were mixed with stirring at room temperature. Subsequently, 240 mg of phosphonitrilic chloride trimer was slowly added into the above mixed solution. Afterward, 12 mL triethylamine was slowly added into the suspension, followed by stirring at 40 °C for 4 h. Finally, the as-obtained ZIF-8@PZP was collected, washed with ethanol and dried. For the synthesis of FeCo-ZIF-8@PZP, 1 g of ZIF-8@PZP was dispersed in THF via ultrasonic treatment. Then, 24.4 mL [1,1’-bis(diphenylphosphino)ferrocene]dichlorocobalt(II) THF solution (1 mg/mL) was added to the suspension, followed by ultrasonic treatment for 2 h. After stirring for 8 h, the FeCo-ZIF-8@PZP was obtained by centrifuging and drying in a vacuum oven at 60 °C.

Synthesis of FeCo-N/P-C, FeCo-N-C, Fe-N-C, Co-N-C

For the synthesis of FeCo-N/P-C, the as-obtained powder of FeCo-ZIF-8@PZP was placed in a quartz boat and then kept at 1000 °C for 1 h in a tube furnace with a heating rate of 5 °C min⁻¹ under a flowing nitrogen atmosphere. FeCo-N-C, Fe-N-C and Co-N-C was prepared with the same synthesis procedure as for FeCo-N/P-C, except that ZIF-8@PZP were replaced by ZIF-8, and [1,1’-bis(diphenylphosphino)ferrocene]dichlorocobalt(II) was replaced by Fe(NO₃)₃.9H₂O and Co(NO₃)₂.6H₂O.

Physical characterization

Electron microscopy and spectroscopic techniques were utilized to analyze the samples. Scanning electron microscopy (SEM) observations were conducted using an XL 30 ESEMFEG field emission microscope. The Philips TECNAI G2 electron microscope was employed for TEM, HRTEM, and elemental mapping analysis. X-ray diffraction (XRD) analysis was performed to identify the crystal structures using a Shimadzu XRD-6100 instrument. In addition, HAADF-STEM images were captured with a Titan 80–300 scanning/transmission electron microscopy operated at 300 kV. Surface electronic states were probed by XPS, utilizing a Thermo Fisher Scientific ESCALAB 250Xi with an Al-Kα X-ray source. This was carried out with Cu Kα radiation (wavelength = 1.54 Å) at a setting of 40 kV voltage and 30 mA current. Raman spectral data were acquired using a J-Y T64000 spectrometer with a 514.5 nm laser. Nitrogen adsorption/desorption isotherm measurements at 77 K with a Micromeritics ASAP 2020 instrument provided the specific surface areas and pore size distributions. X-ray Absorption Fine Structure (XAFS) spectra were recorded at the BL17B1 beamline of the Shanghai Synchrotron Radiation Facility (SSRF), Shanghai Institute of Applied Physics, China. The raw XAFS data were normalized using the ATHENA program. Soft X-ray Absorption Spectroscopy (SXAS) spectra were recorded at the BL08U1A beamline of the Shanghai Synchrotron Radiation Facility (SSRF), Shanghai Institute of Applied Physics, China.

In situ Raman measurements

The in-situ Raman spectroscopy measurements were performed using a J-Y T64000 spectrometer with a 514.5 nm laser. In this experiment, the catalyst was loaded onto a roughened gold substrate, which served as the working electrode. In-situ electrochemical testing was conducted in a 0.1 M KOH solution saturated with O₂, using a Pt electrode as the counter electrode and an Ag/AgCl electrode as the reference electrode. The applied potential range was from 1 V to 0.05 V, with each potential held for approximately 10 minutes. Spectra were collected under open-circuit potential (OCP) conditions without any additional bias. All potentials were calibrated relative to the reversible hydrogen electrode (RHE) using the Nernst equation (E_RHE = E_Ag/AgCl +0.0592 pH +0.197 V).

Electrochemical measurements

All electrochemical tests were conducted in a three-electrode system using a CHI 760E electrochemical workstation (CH Instrument) in 0.1 M KOH electrolyte (pH = 13.0 ± 0.1) at room temperature. Directly plot the obtained data polarization curve, cyclic voltammetry curve, etc. 0.1 M KOH electrolyte was prepared by adding solid KOH to a 1 L volumetric flask (for use within one month) and stored at room temperature in a sealed polytetrafluoroethylene (PTFE) bottle. The counter electrode utilized a graphite rod, while the reference electrode employed a saturated calomel electrode (SCE). The rotating disk electrode (RDE) with a diameter of 5 mm was used as the working electrode. The SCE electrode was calibrated for a reversible hydrogen electrode (RHE) with a Pt wire in a high-purity H₂-saturated electrolyte. The preparation of all catalyst inks involved ultrasonically dispersing 5 mg catalyst in a mixture of 50 μL Nafion (5 wt%) and 450 μL ethanol. The ink was then dispersed onto the rotating disk electrode (RDE) to create a catalyst layer with a loading of 500 μg.cm⁻². For comparison, a commercial 20 wt% Pt/C catalyst ink was also prepared using the same method. This ink was applied to a glassy carbon surface to achieve a Pt loading of 20 μg.cm⁻². ORR polarization curves were corrected by iR-compensation in 0.1 M KOH solution. The solution resistance can be directly obtained by using the iR calibration program of the CHI 760E electrochemical workstation. The solution resistance was found to be 40 ± 5 Ω. All reported potentials in the study were calibrated versus the reversible hydrogen electrode (RHE).

$$E_{RHE}=E_{SCE}+0.059pH+0.244$$ 1

The linear sweep voltammetry (LSV) curves were tested in O₂-saturated 0.1 M KOH with a scan rate of 5 mV/s and a rotation speed of 1600 rpm. Durability tests were carried out by continuously cycling the potential from 0.6 V to 1.0 V at 100 mV/s for 6500, 10,000 and 30,000 cycles in 0.1 M KOH, respectively. Polarization curves for the oxygen reduction reaction (ORR) were recorded using a scan rate of 5 mV/s and various rotational speeds of the electrode (from 400 to 2500 rpm). The number of electrons (n) transferred per oxygen molecule during the ORR was calculated based on the Koutechy-Levich (K-L) equation:

$$\frac{1}{J}=\frac{1}{J_L}+\frac{1}{J_k}=\frac{1}{B{\omega }^{0.5}}+\frac{1}{J_k}$$ 2

$$B=0.62nF{C}_0{D}_0^{2/3}{v}^{-1/6}$$ 3

Where J is the measured current density and ω is the electrode rotating rate (rad/s). B is determined from the slope of the Koutecky-Levich (K-L) plot. J_L and J_K are the diffusion-limiting and kinetic current densities, n is the transferred electron number, F is the Faraday constant (F = 96485 C/mol), C₀ is the O₂ concentration in the electrolyte (C₀ = 1.2 × 10 − 6 mol.cm⁻³), D₀ is the diffusion coefficient of O₂ (D₀ = 1.93 × 10 − 5 cm².s⁻¹), and v is the kinetic viscosity (v = 0.01 cm².s⁻¹).

Measurements using a rotating ring-disk electrode (RRDE) were conducted to monitor the production of HO₂^- at the disk during the oxygen reduction reaction. These measurements took place in 0.1 M KOH at a scan rate of 5 mV/s and a rotation speed of 1600 rpm. The potential at the ring was maintained at 1.3 V versus RHE, which is high enough to oxidize any HO₂^- produced at the disk. The peroxide percentage and the electron transfer number (n) were determined by the following equations:

$$H{O}_2^{-}=200\times \frac{I_R/N}{I_D+I_R/N}$$ 4

$$n=4\times \frac{I_D}{I_D+I_R/N}$$ 5

Where I_D is disk current, I_R is ring current, and N is current collection efficiency of the Pt ring. N is 0.37.

The electric double-layer capacitance (C_dl) was determined by conducting cyclic voltammetry (CV) at different scan rates. Specifically, the capacitive current (i_c) of each sample was plotted as a function of the scan rate (ν). The slope of the resulting linear fit of the i_c-ν plot yields the C_dl. The electrochemically active surface area (ECSA) is positively correlated with C_dl.

$$i_c=C_{dl}\times v$$ 6

Fuel cell MEA tests

The performance of the anion exchange membrane fuel cell (AEMFC) with a FeCo-N/P-C cathode was evaluated using the 850e fuel cell testing system (Scribner Associates). The synthesized catalyst was dispersed in an isopropanol/water (1:1) solution containing FAA ionomer (Fumion FAA-3-Solut-10). The ink was ultrasonicated for 2 h and subsequently uniformly sprayed onto the gas diffusion layer (XGL-R-055). The prepared cathode had an active area of 4 cm² with a catalyst loading of 2.5 mg/cm². The electrode and anion exchange membrane (PiperIOn A20-HCO₃, 20 μm thick) were immersed in a 1 M KOH solution at 60 °C for 4 h to facilitate ion exchange. Afterward, the exchanged electrode and membrane were assembled into a fuel cell fixture (HEPHAS). The anode was fabricated using PtRu/C as the catalyst. During the membrane electrode assembly (MEA) tests, the cell and gas humidifier temperatures were maintained at 80 °C with 100% relative humidity. The backpressure of hydrogen and oxygen or air was set to 150 kPa, with gas flow rates of 0.5 L/min for hydrogen and 1.5 L/min for oxygen, and 0.5 L/min for hydrogen and 1.3 L/min for air, respectively. The open circuit voltage of AEMFC with FeCo-N/P-C as cathode is 1.04 V under such H₂-O₂ conditions. The voltage accelerated stress test was conducted under a flow rate of 0.3 L/min for H₂ and 0.3 L/min for N₂.

Theoretical calculation method

Spin-polarized density functional theory (DFT) calculations were conducted using the Vienna Ab-initio Simulation Package (VASP)^44–48. The core electrons of atoms were treated by Blöchl’s all-electron-like projector augmented wave (PAW) method^48,49. The Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional within the generalized gradient approximation (GGA) framework was employed to describe exchange and correlation effects⁵⁰. A plane-wave cutoff energy of 400 eV was set for the calculation. Electronic occupancies were treated using a Gaussian smearing method with a smearing width of 0.1 eV. The Brillouin zone was sampled using a Monkhorst-Pack k-point grid of 3 × 3 × 1⁵¹. The energy (converged to 1.0 × 10⁻⁶eV/atom) and force (converged to 0.01 eV/Å) were set as the convergence criterion for geometry optimization.

To evaluate the catalytic activity for the oxygen reduction reaction (ORR), the computational hydrogen electrode (CHE) model was employed⁵². The change in free energy (ΔG) for each elementary step of the ORR was computed using the following relation:

$$\mathrm{\Delta }G=\mathrm{\Delta }E+\mathrm{\Delta }ZEP-T\mathrm{\Delta }S+\mathrm{\Delta }{G}_U+\mathrm{\Delta }{G}_{pH}+\mathrm{\Delta }{G}_{field}$$ 7

Here, ΔE is the DFT-calculated total energy difference between the initial and final states, ΔZPE is the change in zero-point energy, and T is room temperature (298.15 K). ΔS is the change in entropy. ΔG = −eU accounts for the effect of electrode potential U, referenced to the standard hydrogen electrode (SHE), with e representing the number of electrons transferred.

$$\mathrm{\Delta }{G}_{pH}=k_{B}Tln10\times pH$$ 8

where k_B is the Boltzmann constant^53,54. The free energy correction due to the electrochemical double layer, $\Delta G_{field}$, was neglected in accordance with prior studies.

The reference state for water was chosen as gas-phase H₂O at a partial pressure of 0.035 bar, corresponding to equilibrium with liquid water at 298.15 K. The free energy of O₂ was derived from the reaction (O₂ +2H₂ → 2H₂O), which is − 4.92 eV at 298.15 K and a pressure of 0.035 bar. According to the CHE model proposed by Nørskov et al., the free energy of (H⁺+e^-) in solution at U = 0 and pH = 0 is equivalent to that of 1/2H₂. Entropy values for H₂ were sourced from the National Institute of Standards and Technology (NIST) database, while those of ORR intermediates were estimated from calculated vibrational frequencies.

The theoretical ORR working potential, U_ORR, defined as the potential at which all elementary steps are thermodynamically favorable (exergonic), was calculated as.

$$U_{ORR}=mini\left[-\mathrm{\Delta }Gx\right]/ne$$ 9

where n is the number of electrons involved in each step, and e is the elementary charge^55,56.

Model

A single layer of graphene model with supercell of 6 × 6 (72 carbon atoms) was employed as the support for catalyst. The Fe-CoN₅P was doped in the graphene sheet to model the Fe-CoN₅-Gra in experiment. The vacuum layer of 20 Å was set along the c direction normal to the graphene sheet to avoid periodic interactions. Meanwhile the model of Fe-CoN₆ and FeN₄ were also set as the same size to estimate the ORR activity for comparison.

The surface Pourbaix diagram⁵⁷ of FeCoN₅P was estimated by the free energy of a given surface structure at different potentials and pH values as shown in E_q of ΔG = ΔG [U = 0, pH = 0] - neU_SHE - kT*ln10*pH. Firstly, the free energy change study the dependence with the potential at pH = 0 was investigated. For intermediate species on the surface, the free energy change referenced to H₂O can be described by a straight line, in which the slope is the transferred electron number n (n = 0 for clean substrate surface, n = 1 for *OH adsorbed surface, n = 2 for *O adsorbed surface, n = 3 for *OOH adsorbed surface, n = 4 for O₂). Secondly, the pH variable is introduced by the follows: U_RHE = U_pH=0 - kT*ln10*pH.

Author contributions

Y.Z.Z. performed the catalyst synthesis, structural characterization, electrochemical tests, and fuel cell measurements, and analyzed the data. Y.Z.Z. wrote the manuscript. M.L.X., M.H.S. and J.B.Z. supervised the entire project and revised the overall language and framework of the manuscript. K.L. conducted the theoretical calculations. X.W. and W.Q. assisted in the analysis of the results. L.L.L. guides the testing of X-ray absorption spectroscopy. C.P.L., F.X. and W.X. provided valuable advice and assistance. The results of the manuscript were discussed by all authors.

Peer review

Peer review information

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

Data availability

The data that support the findings of this study are available within the article and its Supplementary Information files. All other relevant data supporting the findings of this study are available from the corresponding authors upon request. Source data are provided with this paper. The source data can be obtained through the following link 10.6084/m9.figshare.28943621 Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-63322-4.