Typically, methanol (100 ml) with 2-methylimidazole (3.087 g) was added to methanol (100 ml) with Zn(NO₃)₂·6H₂O (1.4000 g) and stirred at room temperature for 9 h to obtain the ZIF-8. The produce was washed several times with methanol and absolute ethanol, the precipitate was collected by centrifugation and drying in a blast oven at 80°C for 12 h.

Typically, Zn(NO₃)₂·6H₂O (1.4000 g) and AgNO₃ (0.2535 g) were dissolved in methanol (100 ml), then methanol (50 ml) with 2-methylimidazole (3.087 g), methanol (50 ml) with NaBH₄ (0.056745 g) and methanol (20 ml) with PVP (0.1000 g) were simultaneously added to the above solution. The mixture was stirred at room temperature for 6 h to obtain the Ag@ZIF-8. The produce was washed several times with methanol and absolute ethanol, the precipitate was collected by centrifugation and drying in a blast oven at 80°C for 12 h.

The Ag@ZIF-8 were pyrolyzed in a tube furnace at 800°C and 900 °C for 2 h around argon atmosphere, respectively. The produce was denoted as Ag NP@N-C-800, Ag NP@N-C-900 and Ag NP@N-C-1000.

The crystal structure of the sample is detected by XRD (Rigaku D/MAX-2005/PC) using Cu Kα radiation (λ = 1.5406 Å) from 5 to 80° with a scan rate of 4° min-1. The microstructure of the sample was obtained by Scanning electron microscopy (SEM) with a Hitachi S-4800 and FEI Helios G4CX using an accelerating voltage of 20 V–30 kV. TEM and HRTEM images were observed on a Titan ETEM G2 at 300 kV. XPS obtained binding energy and valence on a ThermoFisher X-ray photoelectron spectrometer with Al Kα (1486.71 eV) X-ray radiation (15 kV and 10 mA). The surface areas and pore size distribution was performed on a Micrometrics ASAP2020 analyzer at −196°C (77 K).

The electrochemical tests were carried out on a potentiostat/frequency response analysis system (Bio-logic, VSP) at room temperature. The ORR activity was conducted by a rotating disk electrode (RDE, ALS) equipped with a standard three-electrode system. A glassy carbon (GC) disk with the area of 0.07065 cm², a Pt mesh and Ag/AgCl (3.5 M KCl) as a work electrode, a counter electrode and a reference electrode, respectively. All potentials were given versus RHE (E_(RHE) = E(Ag/AgCl)+ 0.969) (Supplementary Figure S1, Supporting Information). In this work, catalyst inks were prepared as follows: 4 mg sample, 0.8 ml of ultrapure water, 0.2 ml of isopropanol and 10 μL Nafion solution were added to a 5 ml sample tube for ultrasonically dispersing 30 min. Then 2 μL of the well-dispersed catalyst ink was dropped onto the glassy carbon electrode and repeated three times. The electrolyte (0.1 M KOH aqueous solution) was bubbled with pure Ar or O₂ for at least 30 min until the solution was saturated before the ORR tests.

The characterization of ORR performance of catalysts by a self-made reversible zinc air battery. The photo of the self-made reversible Zn-air battery is shown in Supplementary Figure S2. Briefly, a polished zinc plate was used as anode, 6.0 M KOH solution was used as electrolyte and a 1 cm × 1 cm foam nickel coated with catalysts ink was used as air cathode. The air cathode was prepared by placing 3 mg sample, 5 mg acetylene black, 100 mg PTFE emulsion and a small amount of absolute ethanol in a beaker, heated in a 60°C water bath, and the product was coated on foamed nickel and drying at 60°C for 12 h. A nickel strip of 5 cm × 2 mm was taken, and the foamed nickel, the waterproof gas permeable membrane and the nickel strip were pressed together by a tableting machine so that the three did not fall. The galvanostatic charge-discharge tests were performed on a zinc air battery measurement system (Land CT2001A) with a current density of 10 mA cm⁻² at room temperature.

The density functional theory (DFT) are carried out by VASP code (Kresse and Furthmüller, 1996a). A generalized gradient approximation (GGA) using Perdew-Burke-ernzerhof (PBE) (Kresse and Furthmüller, 1996b; Lai et al., 2016) scheme is used to describe the exchange-correlation energy. The core electrons are represented by the projected enhanced wave (PAW) potential. The kinetic energy cutoff above 500 eV and the K point sampling on the heterojunction interface cell are 5 × 5×1. Structural optimization was performed with a force convergence criterion of 0.01 eV/A and an energy convergence criterion of 10^–5 eV. A metal cluster containing 13 Ag atoms (Ag13) is used. In the free energy calculation, the step with the smallest change is defined as the potential determining step (PDS) which is the *OH → OH step for the considered catalysts.

Figure 1A shows the preparation procedure of Ag NP@N-C-X (X = 800, 900, 1000°C) through a simple solution method. Typically, Zn(NO₃)2·6H₂O and AgNO₃ were dissolved in methanol, then methanol with 2-methylimidazole, NaBH₄ and PVP (polyvinyl pyrrolidone) were simultaneously added to the above solution. The mixture was stirred at room temperature for 6 h to obtain the produce (Ag NP@ZIF-8). The Ag NP@ZIF-8 was pyrolyzed in a tube furnace at 800°C, 900°C and 1000°C for 2 h around Ar atmosphere. The products were denoted as Ag NP@N-C-800, Ag NP@N-C-900 and Ag NP@N-C-1000. The samples without Ag nanoparticles were synthesized by a similar method (in the Experimental Section), which were denoted as N-C-800, N-C-900, N-C-1000. Detailed processes can be found in the experimental section.

X-ray diffraction (XRD) patterns of ZIF-8 and Ag NP@ZIF-8 (Supplementary Figure S3, Supporting Information), which match the pattern of ZIF-8 well, indicating their high crystallinities and similar zeolite-type structures (Lai et al., 2016). It shows the characteristic peaks at (111), (200) and (220) of Ag (JCPDF 04-0783) after polymerization, indicating the loading of Ag nanoparticles (Figure 1B). Figure 1C and Supplementary Figure S4 (Supporting Information) show the typical rhombic dodecahedron morphology of ZIF-8. After pyrolysis at 900°C under Ar atmosphere, the sample is transformed into N-doped porous carbon coated Ag nanoparticle on the surface, and it still maintains the structure of ZIF-8, demonstrating its good thermal stability (Figures 1D, E and Supplementary Figure S5, Supporting Information). In addition, high-magnification TEM image of Ag NP@N-C-900 shows that pyrolysis-generated carbon will coat Ag nanoparticles, which will prevent Ag from agglomerating during pyrolysis and subsequent catalysis (Supplementary Figure S6, Supporting Information). Compared with those of 800°C and 1000°C (Supplementary Figures S7A, B, Supporting Information), the polymerization at 900°C is the best, in which the primitive structure of ZIF-8 is achieved without segregation or collapse. The high-resolution TEM (Figure 1F) image further demonstrates Ag nanoparticles were encapsulated in graphitic carbon shells, with the size of Ag nanoparticles about ∼15.78 nm (Figure 1G), also evidenced by FFT image (Figure 1I). The lattice spacing values are found to be 2.26 nm, 1.49 nm and 2.39 nm along to (111), (220) and (11-1), respectively, which are consistent with those of pure Ag facets. Furthermore, Figure 1H also shows the silver nanoparticles are wrapped in the carbon substrate, accompanied by the existence of nanoscale twin boundaries (TB). In addition, according to Raman spectrum, the ratio of peak D to peak G is greater than 1, which proves that there are a large number of defects in carbon (Supplementary Figure S8, Supporting Information).

To evaluate the effect of annealing temperature on ORR performances, the samples of ZIF-8, carbonized at different temperatures, were primarily investigated on the rotating disk electrode (RDE). The cyclic voltammetry (CV) measurements of the Ag NP@N-C-900 were performed in Ar or O₂-saturated 0.1 M KOH aqueous solutions at a scan rate of 50 mV s⁻¹. Typically, as shown in Figure 2A, in an inert atmosphere saturated with argon gas, the catalyst has no characteristic peak in 0.1 M KOH solution, which proves that the catalyst itself has no side reaction. However, the reduction peak exists in the O₂-saturated environment. The comparison shows that the reason for the existence of the peak is the oxygen reduction reaction, which proves that the catalyst has excellent oxygen reduction catalytic activity. The similar CV curves of the Ag NP@N-C-800 and Ag NP@N-C-1000 are also attained in O₂ and Ar-saturated 0.1 M KOH (Supplementary Figures S9, S10, Supporting Information).

The ORR performances of different catalysts were further investigated by linear sweep voltammetry (LSV). As shown in Figure 2B, the LSV curves showed that the Ag NP@N-C-900 possesses prominent electrocatalytic activity with a half-wave potential (E _1/2) of 0.83 V, which is equivalent to commercial Pt/C (0.83 V), and a large diffusion-limited current density of 7.03 mA cm⁻², which are much better than the commercial Pt/C (5.13 mA cm⁻²). Compared to other samples at different temperatures, the Ag NP@N-C-900 sample also shows the best ORR performance.

To further study the reaction mechanism of ORR, the RDE measurements of all samples at different speeds (800–2000 rpm) were carried out. The corresponding kinetic parameters and electron transfer number (n) were calculated by Koutecky-Levich (K-L) equation (Zhou et al., 2016). Basically, the J_L value of all electrocatalysts increases with the increasing of rotation rate due to the decrement of diffusion distance (Figure 2C). The inset of Figure 2C illustrates the corresponding K-L plots. Based on the K-L equation, the value of electron transfer number (n) for Ag NP@N-C-900 is calculated to be approximately 3.94, which is close to the theoretical value of 4.0 for Pt/C and better than that of the Ag NP@N-C-800 (n = 3.87) (Supplementary Figure S9B, Supporting Information) and Ag NP@N-C-1000 (n = 3.9) (Supplementary Figure S10B, Supporting Information), indicating that the ORR process mainly includes a one-step four-electron transfer pathway (O₂ + 2H₂O+ 4e⁻ → 4OH⁻)(Bai et al., 2018). The smaller Tafel slope reveals the better ORR activity (51 mV decade⁻¹ for the Ag NP@N-C-900 and 73 mV decade⁻¹ for the Pt/C, Figure 2D). These results confirm that the catalytic process at the Ag NP@N-C-900 electrode undergoes a four-electron ORR pathway, resulting in a high ORR catalytic efficiency. The stability of Ag NP@N-C-900 is assessed by CV measurements between 0 and 1.2 V at a sweep rate of 50 mV s⁻¹ in O₂-saturated 0.1 M KOH. More importantly, after 1000 continuous cycles (Figure 2E), there is a little reduction in E_1/2 (∼2 mV), demonstrating the Ag NP@N-C-900 had the superb stability. Comparison with other samples of the Ag NP@N-C-800 (Supplementary Figure S9C, Supporting Information) and the Ag NP@N-C-1000 (Supplementary Figure S10C, Supporting Information), the Ag NP@N-C-900 still shows better catalytic performance in combination with outstanding structure stability.

In addition, note that an anode fuel (such as methanol) can frequently pass through the polymer membrane from the anode to the cathode, and the cathode catalyst will be poisoned, reducing its catalytic properties (Zhang et al., 2017). Therefore, the Ag NP@N-C-900 and the commercial Pt/C (20 wt%) catalysts were tested for methanol crossover effect via chronoamperometric responses in O₂-saturated 0.1 M KOH electrolyte. The methanol-tolerance results show that there is limited effect on the Ag NP@N-C-900 sample (Figure 2F). On the contrary, when 3 M methanol is injected into the electrochemical cell, the corresponding current density of the 20 wt% Pt/C instantaneously shifts from a cathodic ORR current to a reversed anodic current owing to the methanol oxidation reaction. The current density fluctuation of Pt/C is mainly due to the clogging of the active center on the Pt nanoparticles by the methanol oxidation. More importantly, as summarized in Figure 2G, the performance of the Ag NP@N-C-900 is comparable or even better than many other ORR Ag-based electrocatalysts reported so far (Zhang et al., 2017; Pech et al., 2015; Wu et al., 2017; Dong et al., 2019; Chala et al., 2020).

In addition, as shown in Figure 2H, the Ag NP@N-C-900 exhibits excellent stability (remains more than 82% of its foremost current density) after 16,000 s of continuous operation, which prevails over that of Pt/C (only 58% retention), suggesting that Ag NP@N-C-900 with an outstanding methanol tolerance outperforms the commercial Pt/C. Finally, a series of zinc-air cells with different samples dispersed nickel foam as air electrode, 6 M KOH electrolyte and zinc plate as an anode was assembled to evaluate the catalytic performance of all the catalysts. As seen in Figure 2I, the battery with the Ag NP@N-C-900 is more stable than the Pt/C counterpart. For the battery with the Ag NP@N-C-900 as the catalyzer, the initial voltage gap during charge–discharge is 0.85 V. After continuous 3600 min operation, the voltage gap only slightly increases 190 mV. Comparatively, the battery with the Pt/C as the catalyzer only cycles for 1300 min, and the voltage gap rises from 0.79 V to 1.87 V. It demonstrates that the zinc–air battery based on the Ag NP@N-C-900 catalyst bestows superior structural stability.

The outstanding electrocatalytic performances and stability of the Ag NP@N-C-900 are mainly related to the well-distributed Ag nanoparticles encapsulated in the carbon layer, hierarchical porous structure and the high concentration of pyridine nitrogen doping.

A carbon layer covered on the surface of Ag nanoparticles prevents their leaching and aggregation during the catalysis process. The hierarchical porous structure of the material is one of the important reasons for high performance of oxygen reduction. It is because the hierarchical porous structure can provide more active sites for oxygen reduction and facilitate electron transport and mass transport. The N₂ adsorption/desorption isotherm was carried out to characterize the surface area and porosity of Ag NP@N-C-X (Figures 3A, B). Differing from the other electrodes, the Ag NP@N-C-900 sample bestows the largest Brunauer–Emmett–Teller (BET) surface area, and then the large specific surface area and porous structure allow for the exposure of active sites and rapid mass transfer kinetics. Moreover, the Ag NP@N-C samples shows typical isotherms with hysteresis loops at high pressure region, generally ascribed to the mesoporous structure (Ramaswamy et al., 2013). The such mesoporous structure can be further verified by the pore size distribution plots (Figure 3B). Furthermore, the doping of carbon materials with N can easily alter the electron neutrality of adjacent carbon atoms to form positively charged sites, which is conducive to O₂ adsorption and reduction (Liu et al., 2017).

To ascertain the surface valence state and element bonding composition of the as-prepared samples N-C-900 and Ag NP@N-C-900, the X-ray photoelectron spectroscopy (XPS) has been performed. As revealed in Figure 3C, the full XPS spectrum of the Ag NP@N-C-900 shows the peaks of C 1s, N 1s, O 1s, Ag and Zn 2p which are different from the N-C-900. The C 1s, N 1s and Ag 3 days spectrum (Figures 3D–F) are further analyzed by deconvolution. The high-resolution C 1s XPS spectrum (Figure 3D) includes four sub-peaks: C-C (284.4 eV), C-N (285.1 eV), C-O (286.5 eV) and O-C=O (289.7 eV) (Shen et al., 2017; Bai et al., 2018). The N 1s XPS spectrum shown in Figure 3E can be fitted by four types of N species: pyridine N (398.2 eV, 38.81%), pyrrolic N (399.9 eV, 19.1%), graphitic N (400.8 eV, 29.1%), and Ag-N (406.6 eV, 12.99%)(Zhou et al., 2017; Huang et al., 2018). It is well known that the catalysts with high pyridine N-based carbon have good oxygen reduction performance because pyridine N increases the initial potential of the catalyst (He et al., 2014). Furthermore, the high-resolution spectra of Ag 3p matched the satellite peaks with Ag 3d_5/2 and Ag 3d_3/2 (Figure 3F) belonged to metallic Ag, suggesting that the Ag⁰ existed and matches well with the XRD result (Lopez et al., 2005). In addition, with the increase of temperature (Supplementary Figures S11, S12, Supporting Information), the content of pyridine nitrogen is decreased, and Ag-N disappeared when the temperature reaches 1000°C, partially accounting for the deterioration of performances.

To further elucidate the reaction mechanism during electrochemical catalytic process, the ORR elementary steps have been calculated in terms of density functional theory (DFT). Taking into account the large content of graphite N and pyridine N, four models, i.e., Ag-C, Ag-gN-C, Ag-pN-C (Vacancy I) and Ag-pN-C (Vacancy II) (Supplementary Figure S13, Supporting Information), were constructed to compare distinguish the effects of different types of N and vacancy size. Among them, gN represents the doped form of graphitic nitrogen, pN represents the doped form of pyridine nitrogen. The structure of Ag-N-C the graphite nitrogen is simulated where it mainly contains pyridine nitrogen in Ag-N-C (vacancy I) and Ag-N-C (vacancy II). Among them, vacancy I represents a carbon defect lacking one carbon atom, and vacancy II represents a carbon defect lacking two carbon atoms. The result indicates that the excellent ORR catalytic performance of the Ag NP@N-C can be well-understood by the Ag-pN-C (vacancy II) model, wherein the interaction between the metal and support is strengthened by the cooperative interaction of pyridine nitrogen and carbon vacancy, as demonstrated by the shortest Ag-N distance (2.281 Å) among the models (Supplementary Figure S10, Supporting Information). The enhanced interaction can effectively inhibit the migration of Ag nanoparticles in the carbon surface, in good accordance with the well-dispersion of the Ag nanoparticles observed from the TEM images (Figures 1E, F, Supplementary Figure S14, Supporting Information).

The free-energy change of the reaction pathways and the key reaction intermediates are shown in Figures 4A, B for Ag-pN-C (vacancy II) (Supplementary Figures S15–S17, Supporting Information). Specifically, the initial step of the ORR process is to capture O₂ moleculars. In our calculations, the Ag-C, Ag-gN-C and Ag-pN-C (Vacancy I) are characterized with strong adsorption ability with the E_ads(O₂) −1.234 eV, −1.106 eV and −1.122 eV, respectively. However, the binding of O₂ and Ag-pN-C (Vacancy I) is greatly alleviated, whereby the E_ads(O₂) is only -0.031 eV (Supplementary Figure S18, Supporting Information), indicating a moderate interaction which is beneficial to the reversible ad/desorption in the ORR process (Lopez et al., 2005; Sun et al., 2019; Wang et al., 2021). After the adsorption, the catalytic process is divided into four elementary steps, accompanied with one electron transferring in each step(Chen et al., 2017): i * O 2 + H 2 O l + e − → * O O H + O H − i i * O O H + e − → * O + O H − i i i * O + H 2 O l + e − → * O H + O H − i v * O H + e − → O H − + * where the * represents the adsorption site. It is seen that the first three elementary reactions (steps i to iii) are downhill for all the catalysis, illustrating they are exothermic and spontaneous. In the last step, whereby the * OH is desorbed from the catalyst, the uphill trend on the Ag-C, Ag-gN-C and Ag-pN-C (Vacancy I) elucidates an endothermic reaction, restricting the ORR efficiency. By contrast, it remains downhill in Ag-pN-C (Vacancy II) due to the mild binding interaction between the reaction intermediates and the catalyst which enables facile desorption of OH. Therefore, the significance of pyridine nitrogen and pores in the carbon support not only facilitates a favorable uniform particle dispersion, but also guarantees a moderate binding strength for key intermediates, which improved the free energy change for the potential determining step.