Controllable synthesis and structural characterisation. A synthetic approach combining wet-chemistry impregnation with carbonisation fixation (Supplementary Fig. 1) is developed to precisely control the fabrication of bimetallic Fe/Co SAs on the carbon support. The pre-treated BC, having rich oxygen groups and nanofiber network structures (Fig. 1a, b), was used as the adsorption regulator to controllably impregnate Fe3+ and Co2+. The adsorption experiments were performed using adsorption solutions containing 20 mmol L-1 of Fe3+ or Co2+, and Fe3+/Co2+ mixture solutions having a fixed total Fe3+ and Co2+ concentration of 20 mmol L-1 with different [Fe3+]/[Co2+] ratios of 15/5, 10/10, 5/15 and 1/19. The corresponding Fe3+/Co2+ impregnated BC samples are denoted as Fe3+-20-BC, Co2+-20-BC and Fe3+/Co2+-x/y-BC (x/y: [Fe3+]/[Co2+] ratios in adsorption solutions). As unveiled by Supplementary Fig. 2, for all cases investigated, the adsorption equilibriums can be reached within 4 h, nevertheless, for assurance purpose, an adsorption period of 6 h was selected for all subsequent adsorption experiments. The impregnated Fe3+ and Co2+ contents in Fe3+-20-BC, Co2+-20-BC and Fe3+/Co2+-x/y-BC are summarised in Supplementary Table 2. As shown in Fig. 1c, the impregnated Fe3+ and Co2+ contents on BC are directly proportional to [Fe3+] and [Co2+] in the adsorption solutions. In effect, the slopes of these curves are the distribution factors of Fe3+ (kFe3+ = 0.021 L g-1) and Co2+ (kCo2+ = 0.012 L g-1) that quantitatively define the solid/solution phase distributions of Fe3+ and Co2+. Interestingly, a proportional relationship is also obtained from the plot of Fe3+/Co2+ molar ratios in Fe3+/Co2+-x/y-BC against [Fe3+]/[Co2+] ratios in the adsorption solutions (Fig. 1d). The obtained slope of kFe3+/Co2+ = 1.092 confirms an almost identical ratio between the impregnated Fe3+/Co2+ in Fe3+/Co2+-x/y-BC and [Fe3+]/[Co2+] in adsorption solutions. That is, the quantitative relationships unveiled in Fig. 1c, d can be readily used to precisely guide the impregnation of both Fe3+ and Co2+ contents, and Fe3+/Co2+ molar ratio on BC by simply selecting an adsorption solution with suitable [Fe3+], [Co2+] and [Fe3+]/[Co2+].
Fe3+-20-BC, Co2+-20-BC and Fe3+/Co2+-x/y-BC were then used as precursors and subjected to the thermal treatments to carbothermally reduce the impregnated Fe3+ and Co2+ into metallic Fe and Co, and simultaneously carbonise BC into graphitic carbon. The resultant metallic Fe and Co supported on the carbonised BC (CBC) are denoted as Fe-20-CBC, Co-20-CBC and Fe/Co-x/y-CBC. The scanning electron microscopy (SEM), transmission electron microscopy (TEM) and high resolution TEM (HRTEM) images, and X-ray diffraction (XRD) patterns (Supplementary Figs. 3, 4) confirm that the carbothermal reduction leads to the formation of metallic Fe, Co and Fe/Co alloy nanoparticles (NPs) on CBC34-35. The metal contents in these samples were determined by the inductively coupled plasma atomic emission spectrometer (ICP-AES) and summarised in Supplementary Table 2. These Fe, Co and Fe/Co alloy NPs samples were then subjected to an acid-etching process to remove metal NPs and the remaining Fe, Co and Fe/Co contents in the acid etched samples are summarised in Supplementary Table 2. Fe-O-C and Co-O-C derived from Fe-20-CBC and Co-20-CBC contain 0.27 and 0.22 wt.% of Fe and Co, respectively. While for Fe/Co-O-C-r (r: Fe/Co atomic ratio) derived from Fe/Co-x/y-CBC, the Fe and Co contents are ranged from 0.10–0.22 and 0.08–0.19 wt.%, respectively, corresponding to Fe/Co atomic ratios of 2.9 (Fe/Co-O-C-2.9), 1.5 (Fe/Co-O-C-1.5), 1.0 (Fe/Co-O-C-1.0) and 0.56 (Fe/Co-O-C-0.56). The drastically reduced Fe/Co contents in the acid-etched samples indicate the efficient removal of metallic NPs, which can be further evidenced by the relevant XRD patterns (Supplementary Fig. 4) and TEM images (Supplementary Fig. 5). The aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images (Fig. 1e and Supplementary Figs. 6, 7) disclose the existence of densely populated and atomically dispersed bright dots on all acid-etched samples, suggesting that the remaining Fe and Co contents in these samples are of SAs supported on CBC28,29. The Raman spectra (Supplementary Fig. 8) display distinctive peaks assignable to the D and G bands of graphitic carbon19,20,26, implying that Fe/Co SAs are supported on graphitic carbon. The pore structure measurements (Supplementary Fig. 9) confirm that Fe-O-C, Co-O-C and Fe/Co-O-C-r possess similar micro-/meso-porous structures with Brunauer-Emmett-Teller (BET) specific surface areas ranged from 368.4 to 559.2 m2 g-1 (Supplementary Table 3). The loaded Fe SAs in Fe-O-C, Co SAs in Co-O-C and Fe/Co SAs in Fe/Co-O-C-r (Supplementary Table 2) are subsequently plotted against the impregnated Fe3+ and Co2+ contents in Fe3+-20-BC, Co2+-20-BC and Fe3+/Co2+-x/y-BC (Fig. 1f). The unveiled linear relationships confirm that the impregnated Fe3+ and Co2+ on BC can be proportionally converted into Fe and Co SAs on CBC. The slopes of these curves are the conversion ratios, for which, kFeo/Fe3+ = 0.074 and kCoo/Co2+ = 0.107 infer that 7.4% and 10.7% of the impregnated Fe3+ and Co2+ on BC are converted to Fe and Co SAs on CBC. In fact, the linear relationships with slopes of kFeo/[Fe3+] = 0.0015 L g-1 and kCoo/[Co2+] = 0.0014 L g-1 (Fig. 1g) can be obtained from the plot of the loaded Fe and Co SAs on CBC against [Fe3+] and [Co2+] in the adsorption solutions, disclosing that for a desired loading of Fe, Co and bimetallic Fe/Co SAs on CBC, the required adsorption solution composition can be precisely projected. Remarkably, plotting the bimetallic Fe/Co SAs atomic ratio in Fe/Co-O-C-r against the Fe3+/Co2+ molar ratio in Fe3+/Co2+-x/y-BC gives a linear relationship with a proportional factor of k = 0.772 (Fig. 1h). As expected, a linear relationship is also obtained from the plot of bimetallic Fe/Co SAs atomic ratios in Fe/Co-O-C-r against [Fe3+]/[Co2+] ratios in the adsorption solutions (Fig. 1i). The slope of the curve (k†) is 0.809, which can be used to precisely project the required [Fe3+]/[Co2+] ratio in the adsorption solution for a desired bimetallic Fe/Co SAs ratio in CBC. The quantitative relationships unveiled here can therefore be used to accurately guide the synthesis of Fe, Co and bimetallic Fe/Co SAECs with desirable Fe and Co SAs contents and Fe/Co atomic ratios.
Coordination configuration of bimetallic Fe-Co sites. It has been well-documented that the catalytic activities of SACs are determined collectively by the nature of SA, physiochemical properties of support, importantly, the coordination bonds that anchor SAs to support5,6. The coordination configurations of the bimetallic Fe-Co sites in Fe/Co-O-C-r were therefore thoroughly investigated using Fe/Co-O-C-1.0 as a representative. The HAADF-STEM and corresponding EDX elemental mapping images reveal the homogeneously distributed C, O, Fe and Co in Fe/Co-O-C-1.0 (Supplementary Fig. 10). The presence of Fe/Co SAs is further evidenced by the aberration-corrected HAADF-STEM image took from different locations of a Fe/Co-O-C-1.0 sample (Supplementary Fig. 11). The existence of rich O-containing groups and the formation of Fe/Co-O-C bonds are unveiled by the X-ray photoelectron spectroscopy (XPS, Supplementary Fig. 12)19. The undetectable Fe/Co elements in Co 2p and Fe 2p XPS spectra (Supplementary Fig. 12d) are due to the low Fe/Co contents31.
The synchrotron-based X-ray absorption near-edge structure (XANES) and the extended X-ray absorption fine structure (EXAFS) spectra were obtained to further validate the Fe/Co-O-C coordination configurations. The Fe K edge XANES spectra (Fig. 2a) reveal that the valence states of Fe species in Fe/Co-O-C-1.0 situate in between Fe0 and Fe3+ (ref. 28,29). The observed shoulder peak (A) in the pre-edge region indicates the formation of Fe-O bond19. Notably, the characteristic peak at ~2.20 Å is absented in the Fourier-transformation k3-weighted Fe K edge EXAFS spectrum of Fe/Co-O-C-1.0 (Fig. 2b), confirming the absence of Fe-Fe metallic bonds, while the presence of the peak at ~2.41 Å suggests the existence of the bimetallic Fe-Co bonds28,29. The prominent EXAFS peak at ~1.69 Å observed from Fe/Co-O-C-1.0 further endorses the presence of Fe-O coordination bonds19. According to the Co K edge XANES spectra (Fig. 2c), the valence states of Co species in Fe/Co-O-C-1.0 is in between the metallic Co and the valence states of Co species in Co3O4. As unveiled by Fig. 2d, the absented peak at ~2.20 Å and the presented peak at ~2.41 Å in the EXAFS spectrum of Fe/Co-O-C-1.0 confirm the absence of Co-Co metallic bonds and the presence of the bimetallic Fe-Co bonds28,29. Additionally, the observed prominent Co K edge EXAFS peak at ~1.61 Å from Fe/Co-O-C-1.0 confirms the existence of Co-O coordination bonds. The Fe and Co K edge wavelet transform (WT)-EXAFS spectra were also obtained to further depict the bonding states of the adjacent metal atoms in Fe/Co-O-C-1.0. As displayed in Fig. 2e, the WT contour peak at ~8.4 Å-1 related to the metallic Fe-Fe bond is absented from Fe/Co-O-C-1.0. Also, comparing to the reference samples, the WT-EXAFS spectrum of Fe/Co-O-C-1.0 shows a sole contour peak with a maximum intensity at ~6.8 Å-1, evidencing the existence of Fe-O coordination bonds28,29. As unveiled in Fig. 2f, the characteristic contour peak at ~8.2 Å-1 corresponding to the metallic Co-Co bonds is absented from the Co K edge WT-EXAFS spectrum of Fe/Co-O-C-1.0. Importantly, the observed contour peak at ~12.0 Å-1 from Fe/Co-O-C-1.0 signifies the formation of Fe-Co bimetallic bonds28,29, while the observed intensity maximum at ~9.8 Å-1 confirms the existence of Co-O coordination bonds36. The above experimental results unambiguously confirm the presence of the atomically-dispersed bimetallic Fe-Co in Fe/Co-O-C-1.0 and both Fe and Co species are anchored to the graphitic carbon via oxygen-coordinated bridge bonds. The fitting of EXAFS spectra was performed to deduce the coordination number and geometric configuration. As shown in Fig. 2g, h, the fitting curves match well with the experimental spectra of Fe/Co-O-C-1.0 at both R- and K-spaces. As summarised in Supplementary Table 4, Fe species in Fe/Co-O-C-1.0 has two coordinating interactions, one at 2.00 Å with a coordination number of 2.80 and another at 2.13 Å with a coordination number of 1.00, corresponding to Fe-O and Fe-Co, respectively. Similarly, Co species in Fe/Co-O-C-1.0 also has two coordinating interactions, one at 2.05 Å with a coordination number of 2.80 corresponding to Co-O and another at 2.13 Å with a coordination number of 0.80 assignable to Fe-Co. The above results suggest that the atomically dispersed Fe and Co species in Fe/Co-O-C-1.0 are anchored to CBC via a bimetallic Fe-Co arrangement that coordinates with the graphitic carbon through six O-bridging bonds, for which, the [(O-C2)3Fe-Co(O-C2)3] is the most likely bimetallic unit in Fe/Co-O-C-1.0 with a possible configuration as shown in the inset of Fig. 2g.
Electrocatalytic N2 reduction performance. All experiments were performed strictly following the protocols recommended in the published literatures19,26,37-46. The experimental system setup and analytical methods are illustrated in Supplementary Figs. 13-16 and detailed in Method section. The NRR performance of Fe/Co-O-C-1.0 was firstly examined. Fig. 3a shows the dependence of NH3 yield rate (RNH3) and faradaic efficiency (FE) on applied potentials. The reported RNH3 and FE in Fig. 3a are derived from the recorded chronoamperometric curves (Supplementary Fig. 17a) under different potentials over a 2 h reaction period with the yielded NH3 being determined by the indophenol blue method (Supplementary Fig. 17b). It should note that under the experimental conditions, NH3 is the sole NRR product and N2H4 is undetectable (Supplementary Fig. 18). Although the observed steady-state cathodic current densities are increased with the applied cathodic potentials (Supplementary Fig. 17a), the determined RNH3 and FE (Fig. 3a) are increased initially with cathodic potentials, peaked at -0.30 V (vs. RHE) and rapidly decreased with the further increased cathodic potentials due to the favoured conditions for hydrogen evolution reaction under high cathodic potentials47. An superb RNH3 of 574.8 ± 35.3 μg h-1 mgcat.-1 (185.4 ± 11.4 mg h-1 mgFe+Co-1) with an exceptional FE of 73.2 ± 4.6% are attained at -0.30 V (vs. RHE). To the best of our knowledge, the achieved RNH3 and FE by Fe/Co-O-C-1.0 are the highest among all reported NRR SAECs (Supplementary Table 1).
To confirm whether the yielded NH3 is exclusively resulted from Fe/Co-O-C-1.0 catalysed NRR, the isotopic labelling validation experiments were performed using 15N2 and 14N2 saturated 0.1 M Na2SO4 electrolyte at -0.30 V (vs. RHE) over a 2 h reaction period. The yielded 15NH4+ and 14NH4+ concentrations were quantified by both the indophenol blue method and 1H nuclear magnetic resonance (NMR) analysis. As shown in Fig. 3b and Supplementary Fig. 19, the yielded 14NH4+ and 15NH4+ concentrations determined by 1H NMR method are 39.7 ± 2.1 and 39.2 ± 2.4 μg mL-1, corresponding to NH3 yield rates of 535.7 ± 28.7 and 531.4 ± 32.8 μg h-1 mgcat.-1, respectively, very closely approximated to those determined by the indophenol blue method (41.2 ± 2.8 μg mL-1 for 14NH4+ and 40.7 ± 2.6 μg mL-1 for 15NH4+, corresponding to NH3 yield rates of 555.9 ± 37.8 and 551.4 ± 35.2 μg h-1 mgcat.-1, respectively). Such closely approximated NH3 yield rates from 15N2 and 14N2 confirmed by the two analytical methods infer that the yielded NH3 is indeed originated from the Fe/Co-O-C-1.0 catalysed NRR. The control experiments were subsequently conducted to eliminate potential environmental interferences. As unveiled in Supplementary Fig. 20, only ignorable NH3 concentrations can be detected when the experiments were carried out using N2-saturated 0.1 M Na2SO4 without electrocatalyst (blank), with electrocatalyst but without applied potential (open-circuit) and with electrocatalyst in Ar-saturated 0.1 M Na2SO4 at -0.30 V (vs. RHE) for 2 h. These control experimental results eliminate any noticeable environmental interference and further confirm that the yielded NH3 is resulted exclusively from the Fe/Co-O-C-1.0 catalysed NRR.
The stability of Fe/Co-O-C-1.0 was examined using N2-saturated 0.1 M Na2SO4 electrolyte at -0.30 V (vs. RHE) over a 72 h period (Fig. 3c). The recorded chronoamperometric profile exhibits no noticeable change in current density over the entire testing period, demonstrating an excellent stability. The cycling stability was examined using N2-saturated 0.1 M Na2SO4 electrolyte at -0.30 V (vs. RHE) with 1 h as a testing cycle for 10 consecutive cycles (Supplementary Fig. 21). The nearly identical chronoamperometric profiles adding to the almost unchanged RNH3 and FE obtained from the cycling tests demonstrate a superior cycling stability. The superb stability of Fe/Co-O-C-1.0 can be attributed to its structural stability as evidenced by the well retained atomically-dispersed Fe/Co species (Supplementary Fig. 22) without aggregated metallic Fe/Co NPs (Supplementary Fig. 23) for Fe/Co-O-C-1.0 after 72 h stability test. Interestingly, the XANES and EXAFS spectra (Supplementary Fig. 24 and Supplementary Table 4) unveil that after 10 NRR testing cycles, the bimetallic Fe-Co in Fe/Co-O-C-1.0 could be fitted into a coordination structure of [(O-C2)3Fe-Co(O-C)C2], which might be the actual catalytic active sites, deserving a further investigation.
The NRR performances of Fe-O-C with Fe SAs, Co-O-C with Co SAs and Fe/Co-O-C-r with bimetallic Fe-Co SAs including Fe/Co-O-C-2.9, Fe/Co-O-C-1.5, and Fe/Co-O-C-0.56 were examined. For comparison purpose, CBC without Fe/Co content (Supplementary Fig. 25 and Table 3) was synthesised and its NRR performance was evaluated. Supplementary Fig. 26 shows the recorded chronoamperometric profiles from all targeted electrocatalysts in N2-saturated 0.1 M Na2SO4 electrolyte under -0.30 V (vs. RHE) over 1 h reaction period. The derived RNH3 and FE are summarised in Fig. 3d, e. Interestingly, both RNH3 and FE follow the same trend of CBC < Co-O-C < Fe-O-C < Fe/Co-O-C-2.9 < Fe/Co-O-C-0.56 < Fe/Co-O-C-1.5 < Fe/Co-O-C-1.0. Obviously, Fe/Co-O-C-r with bimetallic Fe-Co SAs outperforms CBC without Fe/Co, and Fe-O-C and Co-O-C with only Fe or Co SAs. This alludes to a bimetallic Fe-Co site induced synergistic effect28-33 as the total Fe and Co SAs contents in Co-O-C and Fe-O-C are similar to those in Fe/Co-O-C-r (Supplementary Table 2). Notably, the total Fe and Co SAs contents in Fe/Co-O-C-r are also closely approximated each other, therefore, for a given total Fe/Co content, the NRR performance of Fe/Co-O-C-r is likely determined by their Fe/Co atomic ratio. In fact, the NRR performance of Fe/Co-O-C-r increases as the Fe/Co atomic ratios closer to the unity, and Fe/Co-O-C-1.0 with a Fe/Co atomic ratio of 0.99 (Supplementary Table 2) exhibits the best NRR performance. This could be due to that Fe/Co-O-C-r with a Fe/Co atomic ratio approaching the unity possesses higher density of the bimetallic Fe-Co sites that dictate the NRR performance. Nonetheless, to our knowledge, no existing analytical technique is capable of quantitatively determining the density of such atomically-dispersed bimetallic Fe-Co sites. Fortunately, the controllable synthetic approach used in this work enables us to precisely control both the content and the atomic ratio of Fe and Co. As such, if we assume that for a given Fe/Co-O-C-r, the bimetallic Fe-Co sites are homogeneously distributed and the maximum possible bimetallic Fe-Co site density is determined by the lower Fe or Co content in a sample, then according to Supplementary Table 5, the maximum density of the bimetallic Fe-Co sites can be estimated as 6.72 ± 0.24, 8.87 ± 0.28, 10.93 ± 0.31 and 13.31 ± 0.51 nmol cm-2 for Fe/Co-O-C-2.9, Fe/Co-O-C-0.56, Fe/Co-O-C-1.5 and Fe/Co-O-C-1.0, respectively. Excitingly, the plot of RNH3 against the density of bimetallic Fe-Co sites unveils a linear relationship (Fig. 3f), signifying that the trend of the bimetallic Fe-Co site density matches the trend of the experimentally determined NRR performance. This provides us with a reasonable confidence that the density of the bimetallic Fe-Co sites is a decisive factor for NRR performance of Fe/Co-O-C-r, nevertheless, further elucidation on the NRR activity origin of the bimetallic Fe-Co sites is needed.
Mechanistic studies. The experimentally identified Fe-Co sites configurations were used to construct the bimetallic Fe-Co sites structural model for the density functional theory (DFT) calculations. Fig. 4a shows the DFT optimised [(O-C2)3Fe-Co(O-C2)3] configuration on graphitic carbon. For N2 adsorption on the bimetallic [(O-C2)3Fe-Co(O-C2)3] unit, we considered three possible adsorption sites: Fe, Co and Fe-Co sites in [(O-C2)3Fe-Co(O-C2)3]. Our DFT calculations indicate that N2 can adsorb separately on Fe and Co sites via end-on approach (Supplementary Fig. 27) or concurrently adsorb on Fe-Co site via side-on adsorption (Fig. 4b). With side-on adsorption, the calculated charge density difference (Supplementary Fig. 28) discloses that the electrons in d orbitals of Fe and Co are transferred to the empty π* orbitals of N2. As a result, the adsorbed N2 gains 0.72 e- from the bimetallic site, and the N-N bond is elongated from 1.120 Å (free gaseous N2 state) to 1.225 Å, suggesting that the adsorbed N2 is activated. Notably, the calculated reaction free energy of the first hydrogenation step (*N2 + H+ + e− → *NH-N) for side-on adsorption on Fe-Co site is 0.61 eV, which is approximately half of that for the end-on adsorption on Fe (1.16 eV) and Co (1.30 eV) sites, confirming that the side-on adsorbed N2 on the bimetallic Fe-Co site is favourable for NRR.
The catalytic activity origin and NRR pathway on the bimetallic [(O-C2)3Fe-Co(O-C2)3] sites were subsequently investigated in details. Supplementary Fig. 29 shows the projected density of states (PDOS) before and after side-on adsorption on the bimetallic Fe-Co site. The broadened Fe-3d and Co-3d states after N2 adsorption addicting to the overlapped Fe, Co and N states imply the hybridized Fe/Co 3d orbitals with N 2p orbital. The overlapped PDOS between Fe-3d and *N2-2p, and Co-3d and *N2-2p are distributed below and above the Fermi energy, signifying a back-bonding between the bimetallic Fe-Co site and *N248. These suggest that the electrocatalytic NRR activity of Fe/Co-O-C-r electrocatalysts is resulted from the effective hybridisation of N 2p orbital with the synergistically configured bimetallic Fe-Co site.
We consequently calculated the Gibbs free energy diagram of side-on adsorption NRR pathway and intermediates structures on [(O-C2)3Fe-Co(O-C2)3] corresponding to each reaction step (Fig. 4c). The 1st hydrogenation requires an energy input of 0.614 eV to form *NH-N and its N-N bond is elongated from 1.225 to 1.286 Å. The 2nd hydrogenation to form *NH-NH* is more favourable than the formation of *NH2-N*, as the former is exothermic and the latter requires an energy input of 0.869 eV. In this process, the N-N bond of *NH-NH* is elongated to 1.339 Å. The formation of *NH2-NH* requires an energy input of 0.209 eV and the N-N bond is further elongated to 1.451 Å. During the 4th hydrogenation step, the N-N bond breaks to form *NH3 + *NH and/or *NH2 + *NH2, which are thermodynamically favoured by 3.19 and 2.49 eV, respectively. For the last hydrogenation step, the formation of *NH3 from *NH2 is uphill by 0.631 eV, which is the likely determining step.
Surprisingly, at the end of the initial NRR cycle (Fig. 4c), [(O-C2)3Fe-Co(O-C2)3] is transformed into [(O-C2)3Fe-Co(O-C)C2], because the latter is 2.176 eV favourable in energy, consistent with the observed structural change by the X-ray absorption analysis of Fe/Co-O-C-1.0 after 10 NRR testing cycles (Supplementary Fig. 24). The DFT calculations were then conducted to depict whether [(O-C2)3Fe-Co(O-C)C2] is the actual bimetallic active site of Fe/Co-O-C-r. Fig. 4d shows the DFT optimised structure of [(O-C2)3Fe-Co(O-C)C2] at the end of the initial NRR cycle. We found that N2 can adsorb separately on Fe and Co sites via the end-on adsorption (Supplementary Fig. 30) and concurrently on Fe-Co site via the side-on adsorption (Fig. 4e), respectively. The calculated charge density differenceunveils that the side-on adsorbed N2 gains 0.62 e- from the Fe-Co site in[(O-C2)3Fe-Co(O-C)C2] (Supplementary Fig. 31). Deferring from [(O-C2)3Fe-Co(O-C2)3], the PDOSs of [(O-C2)3Fe-Co(O-C)C2] (Supplementary Fig. 32) uncover that the side-on adsorption on the Fe-Co site leads to spin-down/beta orbital of Fe-3d shifting away slightly from the Fermi level, suggesting a less effective charge transfer with *N2. Similar to the case of [(O-C2)3Fe-Co(O-C2)3], for the 1st hydrogenation step, the side-on adsorption on Fe-Co site in [(O-C2)3Fe-Co(O-C)C2] requires a lower reaction free energy (0.767 eV) than that of end-on adsorption on Fe (0.922 eV) and Co (0.906 eV) sites. The calculated Gibbs free energy diagram of side-on adsorption NRR pathway and intermediates structures on [(O-C2)3Fe-Co(O-C)C2] corresponding to each reaction step are shown in Fig. 4f. As can be seen, the unveiled NRR pathway is identical to the case of [(O-C2)3Fe-Co(O-C2)3]. However, differing from [(O-C2)3Fe-Co(O-C2)3], where the last hydrogenation step with a uphill energy of 0.631 eV is the likely determining step, the 1st hydrogenation step on [(O-C2)3Fe-Co(O-C)C2] requires an energy input of 0.767 eV to form *NH-N, which is the highest among all the hydrogenation steps. Importantly, our calculations confirm that the [(O-C2)3Fe-Co(O-C)C2] configuration can be regenerated at the end of a NRR cycle upon the desorption of *NH3 (Fig. 4f) and readily available to catalyse the next NRR cycle, inferring that [(O-C2)3Fe-Co(O-C)C2] is likely the actual active site of Fe/Co-O-C-r under the operando conditions and responsible for the attained superb NRR performance.