Engineering d-band center of FeN 4 moieties for efficient oxygen reduction reaction electrocatalysts

,


Introduction
Fuel cells and metal-air batteries are regarded as viable nextgeneration sustainable energy solutions for effectively mitigating global warming and achieving carbon neutrality [1][2][3][4][5]. Though Pt-based catalysts exhibit excellent O 2 adsorption and catalytic activity for oxygen reduction reaction (ORR), the high cost and low reserve sternly restrict their large-scale applications [6][7][8][9][10]. Exploring earth-abundant and cost-effective non-Pt ORR catalysts with high activity and stability is crucial for promoting the commercialization of these energy technologies [11][12][13]. Single-atom catalysts (SACs) with the structure of isolated transition metal-nitrogen-carbon have been considered the most promising candidate to replace platinum group metals (PGM) for catalyzing ORR, due to their decent catalytic activity and stability in acid/alkaline electrolytes [14][15][16][17]. Fe-Nx/C molecules exhibit overall similar ORR activity compared to other M-Nx/C counterparts because of the adsorption energies of O 2 molecules and thermodynamically favorable 4e − pathway [18]. However, further research revealed that the strong electronegativity of N atoms enabled metal sites to strongly bind with electron-donating intermediates, leading to high energy barriers for the subsequent desorption steps, and thus need further modification. In addition, there is still a need for a better understanding of the modulation of the electronic structures of Fe-N-C catalysts to further reduce the overpotential and improve the kinetic activity [19][20][21][22][23]. d-band center (E d ) is considered an efficient descriptor for the electronic structure of transition metal-based catalysts, which can reflect a general correlation between adsorption energy and the shift of E d . In this case, the improvement of the intrinsic activity as well as the optimization of the interaction with oxygen-containing intermediates can be further facilitated by modulating d-band electronic structures of transition metal atoms. Doping with heteroatoms [23], constructing heterointerfaces [24], and regulating the Fe-spin state [2,25] are generally applied to realize electron modulation. When compared to mono-atom catalysts, dual-atom catalysts have essentially different coordination environments and identified two key types of active sites (Co 2 /Fe-N 10 and Fe-N 4 /Ni-N 4 ) [26,27]. Meanwhile, strong electronic coupling occurs between the double atomic pairs, enhancing the d-electron density near the Fermi energy level and synergistically accelerating the adsorption and dissociation of the intermediates [28]. It is well-known that the performance of an atomic catalyst depends not only on its intrinsic activity, but also on the density and distribution of the exposed active sites [29]. Therefore, synthesizing precisely controllable and highly dispersed diatomic catalysts is still challenging. A deeper understanding of the catalytic process in terms of interactions and kinetic pathways on the microscopic interfacial layers is equally crucial [30].
Herein, we developed a curved nitrogen-doped hollow carbon diatomic catalyst (Fe/Zn-CNHC) that achieved excellent electrocatalytic stability while maintaining a high intrinsic activity. The introduction of Fe/Zn atoms is based on the protonated use of nitrogen-doped carbon substrates with moderate coordination strength to the metal during the confinement heat treatment, resulting in the equilibrium dispersion of the atoms on the substrate. Meanwhile, surfactants boost to regulate the micro-interfacial environment of the ZIF-8 and enhance the stability of the catalyst structure. In addition, the unique coordination structure, revealed by extended X-ray absorption fine structure (EXAFS) and density functional theory (DFT), leads to the left shift of Fe 3d orbitals from projected d-density of states (PDOS) and approaching to E d , which optimizes the binding of the ORR intermediates. Benefiting from these features, Fe/Zn-CNHC shows reinforced ORR activity, overwhelming the 20% Pt/C benchmark. This work opens a new avenue to improve the ORR performance of transition metal-based catalysts, and inspires electronic localization of active sites for regulating catalytic activity.

Catalyst synthesis
Synthesis of ZIF-8@P123 precursors: Zn(NO 3 ) 2 ⋅6H 2 O (1.16 g), P123 (0.2 g), and 2-MeIM (1.32 g) were generally dissolved in 45 ml methanol for the production of ZIF-8@P123. Under stirring, equal volumes of the two solutions were combined. After a few minutes, a milky tint developed, indicating that coordination bonds between Zn 2+ sites and N in 2-MeIM ligands had formed. The as-obtained ZIF-8 product was recovered by filtration after 12 hours of additional reaction at room temperature, twice washed with methanol, and lastly dried at 65 • C for 8 hours.
Synthesis of Zn-CNHC and Fe/Zn-CHNC catalyst: 200 mg of ZIF-8@P123 precursor was disseminated in 50 ml methanol, and then 50 ml of 0.014 M Fe(acac) 3 methanol solution was added into the ZIF-8 solution above. The aforesaid combination solution was injected with 20 ml of methanol containing 0.462 g 2-MeIM and stirred continuously overnight at room temperature. Then, 0.418 g Zn(NO 3 ) 6 .H 2 O dissolved in 20 ml methanol was added. Finally, the Fe/ZIF-8@P123 was collected by centrifugation at 10500 rpm for 3 minutes and rinsed numerous times with methanol. Fe/ZIF-8@P123 powder was obtained after drying the solid at 65 • C in vacuum for 8 hours. The as-prepared Fe/ZIF-8@P123 and ZIF-8@P123 were mixed with an equal amount of melamine and then thermally maintained at 900 and 1100 • C for 3 hours under flowing Ar atmosphere at a ramp rate of 5 • C min − 1 in muffle furnace. Zn-CNHC and Fe/Zn-CNHC were collected correspondingly and without any treatment after cooling to room temperature naturally.
Synthesis of Fe-CNHC catalyst: The as-prepared Fe/ZIF-8@P123 was thermally maintained at 900 • C for 3 h under flowing Ar atmosphere at a ramp rate of 5 • C min − 1 in tube furnace (120 ml/min).

Materials characterization
The surface morphology of as-prepared catalysts were studied by scanning electron microscope (SEM, TESCA NMIRA3 LMU, CSU, USA). Transmission electron microscopy (TEM, Tecnai G2 F20 S-TWIN TMP, CSU, USA) were used to investigate the microstructures of catalysts. The elementary composition of the catalysts was characterized by Oxford energy-dispersive X-ray spectrometer (EDX), which is attached to TEM equipment. The phase composition of the as-prepared catalysts were tested by X-ray diffraction (XRD, Empyrean, Malvern Panalytical., Shiyanjia Lab, Netherlands) with Cu Kα radiation (λ=1.5406 Å). The Xray photoelectron spectroscopy (XPS, Thermo Scientific K-Alpha, Thermo Fisher Scientific, Shiyanjia Lab, USA) was analyzed on K-Alpha + system, and calibrated by C 1s (284.8 eV). The X-ray absorption spectroscopy were recorded on Beamline TPL14A1 of Taiwan light source at National Synchrotron Radiation Research Center (NSRRC). The content of Zn and Fe in Fe/Zn-CNHC catalysts were measured by inductively coupled plasma optical emission spectrometer (ICP-OES, Agilent 720ES, Shiyanjia Lab, USA). The gas chromatography (GC-2014C, Shimadzu, BUCT, Japan) was employed to detect H 2 quantification and the carrier gas is ultrapure argon (Ar, 99.999%). Raman (Thermo Scientific DXR2, HNU, USA) spectra were obtained using a DXR2 Raman Microscope with a wavelength of 532 nm. The isolated Zn-Fe atom pair were observed through high-angle annular dark-field scanning transmission electron microscope (HAADF-STEM) with a spherical aberration corrector. X-ray absorption find structure (XAFS) was detected at the BL14W1 in Shanghai Synchrotron Radiation Facility (SSRF). The extended X-ray absorption fine structure (EXAFS) data were analyzed using Athena software.

Electrochemical measurements
At room temperature, electrochemical measurements of catalysts were performed using a Gamry Interface 1000 electrochemical workstation equipped with a standard three-electrode system. Dropping the catalyst ink onto a glass-carbon (GC), which is a rotating disk electrode (RDE, 5 mm in diameter). The catalyst ink was made by dispersing 5 mg of each catalyst in 1 mL of a mixed solution including 0.98 mL of ethanol/water (1:1, v/v) and 20 mL of 5% Nafion solution, then ultrasonication for 30 minutes to make a homogeneous black suspension solution. The catalyst ink was then dropped onto the surface of the GC electrode and allowed to dry in the air. The catalyst loading amount was 0.3 mg cm − 2 for both as-prepared catalysts and commercial Pt/C (20 wt %).
The catalysts' ORR performance was tested in an O 2 -saturated 0.1 M KOH solution. At a scan rate of 50 mV s − 1 , the cyclic voltammetry (CV) curves were obtained. RDE was used to perform linear sweep voltammetry (LSV) at a scan rate of 5 mV s − 1 with varying rotation speeds (400, 625, 900, 1225, 1600, and 2025 rpm). The standard reversible hydrogen electrode is used to compare all measured potentials (RHE). The onset potential was defined as the potential required for generating a current density of -0.1 mA cm − 2 in LSV curves. The electron transfer number was calculated by Koutecky-Levich (K-L) equation: where J k is the kinetic diffusion current density, JL is the kinetic limiting current density, B is the reciprocal of the slope, is the angular velocity (=2N, N is the rotation speed), n is the transferred electron number, F is the Faraday constant (96485 C mol − 1 ), C0 is the saturated concentration of O 2 (1.210 − 6 mol cm − 3 ), D0 is the diffusion coefficient of O 2 (1.910 − 5 cm 2 s − 1 ), and v is the (0.01 cm 2 s − 1 ). Accelerated degradation tests (ADT) at a scan rate of 100 mV s − 1 for 5,000 cycles between 0.2 and 1.1 V were used to assess the ORR stability of the electrocatalysts in the O 2 -staurated 0. Assembly of Al-air batteries: The polished aluminum plate serves as the anode in most Al-air batteries. In a home-made cell model, the electrolyte is 6M KOH containing 0.0005 M In(OH) 3 , 0.0075 M ZnO, and 0.01 M Na 2 SnO 3 ; the cathode is a gas diffusion electrode with a catalyst loading amount of 2.0 mg cm 2 . The commercial Pt/C was likewise built in a similar manner as a control. The LAND testing system was used to perform battery measurements.

Theoretical calculations
Vienna Abinitio Simulation Package was used to undertake the theoretical study based on density function theory (DFT) computations (VASP). The electron-valence energy was described using the projector augmented wave (PAW) approach and the generalized gradient approximation (GGA) using a Perdew-Burke-Ernzerhof (PBE) functional. For the plane-wave basis set, an energy cutoff of 500 eV was employed. The energy and force convergence thresholds were established at 10 − 5 eV and 0.02 eV/, respectively. The vacuum layer thickness was set to 20 to prevent interaction between two neighboring images. To account for dispersion interactions, a semi-empirical van der Waals (vdW) correction developed by Grimme (DFT-D3) was used.
The free energy of each elementary step in the proton coupled electron transfer reactions was computed using the computational hydrogen electrode (CHE) model for oxygen reduction reaction (ORR). Considering the O 2 molecular is not broken before reduction, the associative 4e − reduction pathway was evaluated to be most feasible for ORR in this work, as follows: where * represents the active site on the corresponding surface.

Catalyst characterization
The electronic structure of an ORR electrocatalyst is known to be intrinsically linked to the adsorption energy of reaction intermediates. Before synthesizing the Fe/Zn-CNHC electrocatalysts, PDOS were carried out to evaluate the charge distribution of these Fe active sites (Fig. 1a). Clearly, the E d energy level of Fe/Zn-CNHC is higher than that of Fe-CNHC, suggesting that the E d energy levels are closer to the Fermi level (E F ) after introducing Zn. The construction mechanism of the Fe/ Zn-CNHC phase-junction electrocatalyst is shown in Fig. 1b. The selective cleavage of the Zn induces the incorporation of the electronwithdrawing status on the active center of FeN 4 , thus forming a unique electronic structure. Especially, the electronic interaction can be seen as a coupling between the valence state of the adsorbate and the D state of the transition metal, leading to the creation of split bonds and the upward shift of E d , which is beneficial for the adsorption of O 2 for accelerating the proton-coupled electron transfer processes [25], thus accelerating entire ORR process.
As illustrated in Fig. 2a, a series of Fe/Zn-CNHC catalysts were fabricated via a facile two-step procedure (the detailed process is displayed in the experiment section). Firstly, Fe-doped ZIF-8@P123 (Fe/ ZIF-8@P123) precursors were prepared by precisely controlling Fe doping amount in a simple polymerization process. Subsequently, the as-obtained Fe/ZIF-8@P123 precursors were carbonized using a thermal reaction to synthesize the X-Fe/Zn-CNHC (X stands for temperature) catalysts in the muffle furnace under argon atmosphere with ZIF-8 as both carbon and nitrogen sources (Fig. S1). ZIF-8 forms a continuous carbon frame after thermal reaction, which can not only provide a multidimensional electron transfer pathway but also improve the electrical conductivity of the catalysts [31]. To further analyze the formation mechanism of the Fe/Zn-CNHC hollow structure, we conducted a series of temperature-controlled experiments. We investigated the effect of the temperature, and tracked the evolutionary trajectory of hollow and transmission electron microscopy (TEM) were used to analyze the overall morphology of the prepared catalysts. (Fig. S3) Typical Fe/Zn-CNHC materials with Fe atoms exposed on the surface of the dodecahedron were produced via pyrolysis of ZIF-8@P123 under Ar at 900 • C. In contrast to the standard ZIF-8 synthetic technique, P123 was added to the new system's initial precursor solution as adjusting the favorable aperture.
The planar surfaces of Fe/ZIF-8@P123 are partially etched at 500 • C pyrolysis temperature, while the pristine structure is preserved. Within the framework, the crystal surface is rather smooth, with dominating micropores connected with 3D hydrocarbon networks. The surface and internal regions of ZIF-8@p123 were further removed by pyrolysis at 900 • C, resulting in a curved framework structure with a hollow interior and interconnected edges. This unique structure results in a concave morphology with rough and shrinking surfaces, which differs from ZIF-8 Which is commonly obtained. During the development of hollow structures, the framework offers support to avoid the collapse and fracture of shell layers. The facets of matching carbon particles become less distinct, and the particles become more porous, with many micropores on the surfaces (up to 2 nm). Active areas are thought to lie inside or near the micropores of carbon supports, according to research [32,33]. The related TEM images and EDS mappings in various stages are displayed in Fig. 2d to depict the morphology and distribution of elements, which significantly confirms the preceding findings. The Kirkendall effect is followed in this method [34]. The surfactant P123 and Zn 2+ ions undergo continuous unequal interdiffusion at the core and shell interfaces of the dodecahedron, which leads to the appearance of Kirkendall voids. As the Kirkendall effect process proceeds, the internal ZIF-8 gradually decomposes and forms a uniformly thick hollow structure. The presence of P123 can also further guide the tendency of the shell framework to be curved. When the pyrolysis temperature continues to rise and is maintained at 900 • C for 3 hours, most of the residual zinc in the shell is removed in the form of vapor, thus building the desired complete hollow structure. Then, under the thermal treatment at 1100 • C, the size and shape of ZIF-8 crystals are difficult to maintain efficiency in catalysts. The polyhedral carbon particles become blurred and even partially broken in the catalyst.
The metals were atomically distributed, as demonstrated by aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM). There were numerous, uniformly spaced, isolated bright spots on the carbon matrix, and no metal nanoparticles had formed (Fig. 2e). Many metal-pair locations can be seen more clearly by expanding the magnification of the atomicresolution HAADF-STEM image (Fig. 2f). For accuracy, different forms of atoms are marked with different colored circles. Most of them are in the form of diatomic pairs, with only a small number of single or polyatomic configurations. The electron energy loss spectrum (EELS) (Fig. S4) reals the coexistence of Fe and Zn elements in Fe/Zn-CNHC. Based on the observation of more than 100 atomic pairs with a distance of about 0.25 nm between two bright points, the formation of a double atomic pair is demonstrated. Other individual bright dots marked by blue circles are identified as isolated Fe or Zn single atoms, indicating that the Fe/Zn atomic pair, single Fe and Zn atoms, coexisted on the CNHC. According to the EDS, it was shown that the elemental distribution of Fe, Zn, N, and C is homogeneous in the whole framework (Fig. S5). SEM images and X-ray diffraction (XRD) patterns confirm that the morphology and structures of ZIF-8 remain unchanged after the introduction of P123 (Fig. S6). The XRD peaks of Fe/Zn-ZIF@P123 are consistent with that of ZIF-8@P123 without any impurities. Meanwhile, Fe/Zn-CNHC catalysts show two broad peaks at 25.2 and 43.7 • , assigning to the (002) and (101) crystal planes of carbon, respectively (Fig. S7) [35]. Notably, no diffraction peak ascribed to metallic Fe or other Fe-based species can be detected, indicative of the atomically dispersed Fe single atoms in the Fe/Zn-CNHC. Meanwhile, high-resolution transmission electron microscope (HRTEM) confirmed the phenomenon that only disordered amorphous carbon exists (Fig. S8).
BET surface areas of Fe/Zn-CNHC, Zn-CNHC, and Fe-CNHC are 1351, 835, and 796 m 2 g − 1 , respectively, based on N 2 sorption isotherms. Due to the addition of P123, the produced catalysts had a wider micro-pore size distribution (Fig. S9). Meanwhile, the interaction of different metals also affects the morphology of the catalyst. Through inductively coupled plasma mass spectrometry (ICP), the Fe and Zn loading are 0.98 wt% and 2.87 wt%, respectively. Additionally, the metal contents of the other catalysts are shown in Table S1. The overall N doping determined by XPS is summarized in Fig. 3a. The four types of N species found in all catalysts are pyridinic N (398.5 eV), pyrrolic N (399.5 eV), graphitic N (400.7 eV), and oxidized N (402.7eV) [36]. Through Zn capture, the total content of pyridinic N and pyrrolic N increases in the Fe/Zn-CNHC catalyst (Fig. 3b). Meanwhile, based on the previous reports, pyridinic N can not only operate as active sites for ORR activity, but it can also create Fe/Zn-CNHC moieties by covalently coordinating with the neighboring metal atoms [15,16]. Changing the gas partial pressure environment at high temperatures affects the metal-nitrogen content, but the nitrogen content is not the key factor affecting the activity. The four types of C species found in all catalysts are C=O, C-O, C-N, and C-C group (Fig. S10). Carbon-oxygen species are regarded as defects that facilitate ORR in alkaline media. sp 2 C also improves ORR activity due to its high electrical conductivity. Furthermore, due to the little amount of Fe in each catalyst, the Fe 2p signal could not be identified in all of them, as shown in Figs. S11 and S12 [19]. This is consistent with the ICP results.
X-ray absorption near edge structure (XANES) and X-ray absorption fine structure (EXAFS) spectra were used to identify precise electronic structures and local coordination environments of the Fe/Zn-CNHC. We carried out EXAFS characterization to verify the atomic dispersion of Fe species and their coordination structure. According to the energy of the absorption edge, the oxidation state of the Fe atoms in Fe/Zn-CNHC is near +3 (Fig. 3d). Moreover, the pre-edge peak of Fe/Zn-CNHC located at 7113 eV is very close to that of Fe 2 O 3 , further confirming the +3 valence state of the Fe atoms in Fe/Zn-CNHC [37]. From the R space of Fe/Zn-CNHC catalyst, only a Fe-N peak at 2.36 Å is observed, confirming its atomic dispersion [38]. The subsequent least-squares EXAFS fit analysis demonstrated that the Fe atomic site was coordinated to four N atoms and the bond length of Fe to N was 2.02 Å (Fig. 3e,f and Table S2). Fig. 3g-i show electronic structures and local coordination environments of the Zn atoms in Fe/Zn-CNHC. Based on the observation that the position of the absorption edge of zinc in the K-edge XANES spectrum is between that of the Zn foil and ZnO, its valence state was shown to be between 0 and +2. The peak at 1.5 Å is assigned to the scattering path of Zn-N. The Zn-Zn and Zn-O-Zn bonds are not monitored. The fitted coordination number between the Zn atom and N is 4, and its bond length is 2.03 Å (Fig. 3i and Table S2), respectively. Therefore, we suggest that the Fe/Zn diatomic sites are homogeneous in the form of atomically dispersed FeN 4 and ZnN 4 structures. Fig. S14 shows the k-space spatial oscillations of Fe and Zn atoms, respectively, which provide data support for the fitting of the coordination numbers and bond lengths of the above-mentioned coordination metal atoms, and the R-space data can be extracted by Fourier transform. Plots of R and K spatial energy oscillations of Fe and Zn foils as comparison are presented in Fig. S13. The wavelet transform (WT) analysis of the EXAFS oscillations of Fe/Zn-CNHC was performed to visualize the dispersion of metal atomic sites at radial distances and K-space in high-resolution maps (Fig. 3c). These results further confirm that the atomic dispersion of the Fe and Zn sites with M-N 4 conformation are well-established.
Fe/Zn-CNHC shows the highest C dl of 4.63 mF cm − 2 and the largest ECSA of 115.75 m 2 g − 1 compare to Fe-CNHC (1.86 mF cm − 2 /46.5 m 2 g − 1 ) and Zn-CNHC (1.42 mF cm − 2 / 35.5 m 2 g − 1 ) ( Fig. S18 and S19). In addition, to exclude the contribution of ECSA to the ORR performance, all samples of j k from the measured were subjected to ECSA normalization. j k curves normalized by ECSA showed the same trend as the corresponding LSV curves (Fig. S20), again indicating the outstanding ORR intrinsic activity of Fe/Zn-CNHC [40]. Moreover, the kinetic current density was normalized by loading the mass of catalyst to give the kinetic mass specific activity (j m ) [41,42]. As shown in Fig. S21, the j m of Fe/Zn-CNHC is determined to be 36.52 A g − 1 , which is 1.8 and 9.1 times higher than those of Fe-CNHC and Zn-CNHC, respectively. This result suggests that the utilization of active sites for Fe/Zn-CNHC is much higher under the same mass loading, further demonstrating the introduction of ZnN 4 can facilitate the intrinsic activity and boost ORR kinetics. To understand the origin of the observed high durability of the Fe/Zn-CNHC catalyst, we carried out XPS measurements to monitor changes for Fe/Zn-CNHC electrocatalysts before and after accelerated stress tests ( Figure S23-25). The high-resolution Zn 2p spectrum exhibits two main peaks at 1021.4 eV (2p3/2) and 1044.6 eV (2p1/2), attributing to the oxidation state of Zn single atom. Notably, the Zn species have no significant change during the ex-situ test, except for the slightly positive shift. It can be ascribed to the electron transfer from Zn to Fe during the reaction, as well as the adsorption of oxygen containing intermediates [43]. Beyond that, the Fe/Zn-CNHC catalyst exhibited a relatively small decrease in pyridinic-N content from 56.8% to (32.6-52.6) % after accelerated stress tests. In general, Fe/Zn-CNHC catalysts largely retained their ZnN 4 and FeN 4 active sites after the long-term operation. The performance of the long-term stability of Fe/Zn-CNHC was effectively evaluated by high-speed cyclic voltammetry tests. After 5,000 CV cycles, the LSV curve showed no significant attenuation and a negative shift of only 7mV (Fig. 4e). The outstanding stability is also confirmed by chronoamperometry. After 10 hours at 0.7 V vs. RHE in an O 2 -saturated electrolyte, Fe/Zn-CNHC can maintain 93.2% of the initial current density. However, under the identical conditions, commercial Pt/C maintained 85.6% retention. At the same time, to further demonstrate the Zn-Fe atomic pair as the real active site, the introduction of SCN − in the test system resulted in a significant negative shift of 130 mV in the half-wave potential, a significant decrease in the diffusion-limited current density (=85% retention), and a significant poisoning of the metal atoms, while no significant changes were found in the prepared porous NC, indicating that the Fe-N 4 site constitutes the key active center of the ORR (Fig. 4f).
The effect of Zn-load amounts on the ORR activity of Fe/Zn-CNHC catalysts are also investigated. As shown in Fig. S26, When the atom % Zn is ca.2.87, the Fe/Zn-CNHC shows the highest positive E 1/2 and lowest activity. Both the lower and higher amounts of Zn-load will decrease its ORR activities. The reason may be that the E d energy levels of Fe are tailored by the Zn-load amounts, which is closely linked with the adsorption and desorption capabilities for the ORR intermediates.

DFT calculations
With the information of the composition and structure of Fe/Zn-CNHC active sites, the DFT calculation was conducted to explore the detailed electronic and geometric structures of the active sites for ORR. Fig. S27 reveals the optimized geometric model of simulated Zn-CNHC, Fe-CNHC and Fe/Zn-CNHC electrocatalysts. The electron charge density distribution shows that the non-uniform charge distributions are intensified after introducing Zn. Fig. 5a-c shows that the electron density of Fe centered structures in Fe/Zn-CNHC is higher than that of Fe-CNHC. In comparison to Zn-CNHC, Fe/Zn-CNHC has a lower electron density. In the Fe-CNHC and Zn-CNHC model, the charge numbers of Fe and Zn sites were 1.06 eV and 1.35, respectively, while this value is promoted to 1.43 eV of the Fe/Zn-CNHC , revealing a valence state transition from Fe 3+ (3d 5 ) to Fe 2+ (3d 6 ) [26,44]. Additionally, DOS was employed to estimate the binding strength among the metal center and non-metal atoms (C and N) [45]. As illustrated in Fig. 5d-f, an obvious energy negative shift for O 2 adsorption of single-metal atom center (Fe-N 4 : -1.90eV, Zn-N 4 : -1.61eV) can be observed on the dual-metal atom centered structures (-2.10 eV). Interestingly, the bimetal structure loses more electrons while C gains more electrons, which is consistent with the DOS results (Fig. S28). Based on previous reports, the introduction of Zn spontaneously causes the generation of spin-polarized electrons at the Fermi energy level. More specifically, the 3d band is intrinsically narrow for Fe-CNHC and broadened when combined with Zn, illustrating an upshift of E d [28,46]. Based on the PDOS analyses (Fig. S29) and d-band theory, an upshift of the d-band center for the electrocatalysts (Fig. 5g) is attributed to the charge transfer between the metal sites. The d-band of Fe/Zn-CNHC shifts close to Fermi level, and enhances the capture of the O 2 molecule that is the initial step of ORR [47].
ORR processes on different sites were estimated at 1.23 V and compared to further confirm the electrocatalytic active center, as shown in Fig. 5h (Fig. 5i), which is in good agreement with the fact that G *OH decreases in Fe/Zn-CNHC. Fe/Zn-CNHC has the smallest energy difference among the O* and OH* intermediates, indicating good ORR activity. Furthermore, when compared to Fe-CNHC (0.655 eV) and Zn-CNHC (0.697 eV), the formation energy of intermediates on Fe/Zn-CNHC (0.372 eV) is much lower. However, when the distance between Fe and Zn atoms exceeded 5 Å, the adsorption potential of oxygen-containing intermediates increased, which is in agreement with the HAADF-STEM results that the Zn-Fe atom pair at the appropriate distance is favorable for ORR activity (Fig. S30).

Aluminum-air battery performance
In addition, the performance of the air cathode prepared from Fe/Zn-CHNC catalyst was evaluated in the flow cell system depicted in Fig. 6a, using 20% Pt/C as a comparison [50,51]. The designed flow cell consists of an aluminum anode, an electrolyte, and an air cathode [52]. As expected, the power density of the Fe/Zn-CNHC anode (100 mW cm − 2 ) was substantially higher compared to Pt/C (80 mW cm − 2 ) (Fig. 6b). Then, we evaluated the discharge performance of the catalyst at different discharge current densities (Fig. 6c). It is clear that Fe/Zn-CNHC exhibits better rate capability and fast dynamic response compared to Pt/C. The open circuit potentials of Zn/Fe-CNHC and Pt/C are -1.80 and -1.75 V, respectively. Fe/Zn-CNHC also exhibits superior operating voltage for high current density operation. The discharge curves shown in Fig. 6d indicate that the Al-air cell assembled with Fe/Zn-CNHC can provide a specific capacity of 856 mAh g − 1 at a high current density of 100 mA cm − 2 , which is greater than that of the cell assembled with Pt/C (782 mAh g − 1 ). Its stability as an air positive electrode was further evaluated with excellent cycling stability (100 h) at a current density of 25 mA cm − 2 . After 100 h of operation, the voltage gap decreased by only 0.06 V, with an extremely low decay rate of 0.01% per hour (Fig. 6e).

Conclusion
In summary, a novel mechanism with the introduction of atomic Zn into Fe-N 4 coordination is unraveled to optimize the electronic structures and performance of as-designed electrocatalysts for ORR. Due to the lower electronegativity of Zn, it is easier to donate its outer electron to Fe. The optimized d-band electronic structure leads to the spontaneous filling of the Fermi layer with spin-polarized electrons and promotes their trapping of oxygen-containing intermediates and binding to O 2 . As expected, the Fe/Zn-CHNC exhibits exceptional ORR performance in alkaline media, such as a small half-wave potential of 0.91 V, and high stability (only 12 mV loss of half-wave potential after 5,000 LSV scans). Interestingly, the co-presence of Zn and Fe species not only promotes the formation of hollow structures, they can also reduce the adsorption potential of the reaction intermediates. Benefiting from these features, this finding establishes a facile way to obtain diatomic catalysts on a larger scale and can be further applied to the simultaneous atomic dispersion of multiple metals on the substrates, providing valuable insights for the preparation of high-performance and stable diatomic catalysts.