Tailor-made metal-nitrogen-carbon bifunctional electrocatalysts for rechargeable Zn-air batteries via controllable MOF units

: The majority of chemical syntheses involve the use of catalysts, which play a crucial role in the yield and conversion rates of chemical reactions. In view of the increasing demand for chemical commodities and specialties linked to the growth of the world’s population and the living standards, highly efficient and low-cost catalysts are urgently required. The metal-nitrogen-carbon (M-N-C) catalysts family is one of the most promising candidates. In this work, a series of benzene-1,3,5-tricarboxylate linker based metal organic frameworks (MOFs) were used as self-sacrificial templates and tunable platform for designable preparation of M-N-C catalysts. Changing the pillars between the 2D layers and the nature of the metal ions in the pristine MOFs significantly influenced the structure, chemical composition and catalytic activity of the resulting M-N-C catalysts for the oxygen reduction reaction (ORR). Furthermore, the influence of the MOF units on the catalyst performance, the role of the metals in the M-N-C catalysts and the primary catalytically active sites for ORR were explored by a combination of density functional theory (DFT), in-depth structural and chemical/elemental characterizations, and electrochemical studies. Among the prepared catalysts, Co-BTC-bipy-700 exhibited the highest electrocatalytic activity for oxygen reduction reaction (ORR), which showed a larger limiting current density and similar half-wave potentials with less catalyst degradation and much higher methanol tolerance than the commercial Pt/C catalyst. Meanwhile, as a bifunctional electrocatalyst, Co-BTC-bipy-700 catalyst was also employed for oxygen evolution reaction (OER) and demonstrated a lower overpotential (lowered by 140 mV at a current density of 10 mA cm −2 ) and better durability than IrO 2 . Furthermore, in terms of device performance, the Zn-air battery enabled by Co-BTC-bipy-700 catalyst reached a maximum specific energy as high as 1009.8 Wh kg −1 , which is 76.5% of the theoretical value (1320 Wh kg −1 ), and demonstrated higher discharge potential and lower charge potential than that based on the Pt/C catalyst. Importantly, the presented strategy for tailor-made M-N-C catalysts by controlling the synthesis of the pristine MOFs could offer a guide map for the future design of M-N-C catalysts family not only for electrochemical reactions but also beyond electrochemistry. Catalytic performance: All the electrocatalytic measurements are carried out in a three-electrode cell using a rotating disk electrode (RDE, PINE Research Company). Before measuring, the electrolyte was purged with O 2 (for ORR) or N 2 (for OER) for at least 45 minutes. During the test, the electrolyte was continuously bubbled with O 2 or N 2 . The ORR and OER catalytic performance of the samples were characterized by linear sweep voltammetry (LSV), cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) techniques on an Autolab electrochemical workstation at 25 °C. An automatic thermostat bath was used to control the temperature (Julabo F25, Germany). The supporting electrolyte in all electrochemical experiments is 0.1 M KOH. The currents of the LSV and CV are normalized by the mass of the active materials and the geometrical area of the glassy carbon electrode (GCE). The EIS measurements were recorded in a frequency range from 10 −2 to 10 4 Hz at open circuit potential with a peak-to-peak amplitude of 5 mV. The catalyst covered GCE, a graphite plate (3.5 cm × 5 cm) and a Hg/HgO electrode were used as the working electrode, the counter electrode and the reference electrode, respectively. All potentials are referred to the reversible hydrogen electrode (RHE) potential, which was converted from the Hg/HgO electrode using the following equation:


Introduction
Without catalysts, our lifestyle would be significantly different. More than 80% of all chemicals and pharmaceuticals are made by catalysts. 1 Although tremendous efforts have been made by the chemical industry, noble metal catalysts still dominate the market of energy-related catalysis and fine chemistry. 2 It is urgent to find a suitable alternative due to the prohibitive cost and scarcity of these noble metals. Different types of transition metal based catalysts have been widely studied. 3 Among various alternatives, a new class of metal (or metal compound)-nitrogen-carbon (M-N-C) catalysts have emerged as one of the most promising candidates, and they are Tailor-made metal-nitrogen-carbon catalysis for rechargeable Zn-air batteries via controllable MOF unit design 2 used in a wide range of chemical reactions such as transesterification 4 , hydrogenation [5][6][7] , oxidative degradation 8 , Fischer-Tropsch synthesis 9, 10 and especially for electrochemical reactions (e.g. hydrogen evolution reaction (HER) [11][12][13][14][15] , oxygen evolution reaction (OER) [16][17][18][19][20] , oxygen reduction reaction (ORR) [21][22][23][24][25][26][27][28] and carbon dioxide reduction 29,30 ).
The traditional methods for preparing M-N-C catalysts can roughly be divided into two categories. The first one is the two-step method in which N-doped carbon materials are subsequently loaded with metals. For example, Shaabani et al. synthesized three-dimensional nitrogen-doped graphene frameworks by hydrothermal and freeze-drying process, and then immobilized nickel nanoparticles on the prepared substrate. 31 The second method is the one-step method in which a mixture of carbon, nitrogen and metal precursors is pyrolized. Metal/N 4 macrocycle complexes together with some carbon source are often used for fabricating M-N-C catalysts. 26,32,33 Common inorganic salts (sulfates, acetates and chlorides) and organic compounds (including dicyandiamide, melamine and tripyridyl) have also successfully been employed as M-N-C catalyst precursor materials. [34][35][36] Moreover, in order to improve the mass transport properties and increase the exposure of the active sites, templates were introduced into the synthesis process, such as SBA-15 37,38 and SiO 2 nanoparticles 25,39 .
In view of the above, metal-organic frameworks (MOFs), constructed from metal ions or clusters bridged by organic ligands, have been considered as ideal self-sacrificial templates for M-N-C catalysts due to their highly ordered cavities, open channels and rational composition and structure. 2,40 These unique characteristics of MOFs allow them to serve as both template and precursor materials (metal, carbon and nitrogen). Furthermore, the highly ordered structure and chemical composition result in a homogeneous distribution of heteroatom and metal particles, which could provide more active sites after pyrolysis. Therefore, different kinds of MOFs have attracted attention as precursors for preparing M-N-C catalysts, such as cobalt-4′-(4-pyridyl)-4,2′,6′,4′-terpyridine 41 , cobalt-imidazolate 42 , cobalt-benzimidazole 43,44 , ZIF-67 [45][46][47] , ZIF-8 48,49 , MIL-101 50 , MIL-100 51 , MIL-88B-NH 3 52 , tip of the iceberg as more than 20,000 different MOFs have been created and the number is still increasing 56,57 . The large quantities of MOFs offer extensive possibilities for the M-N-C catalyst family but at the expense of endless lab work. The previous studies often suffer from the unclear influence of different metal nodes and linkers on the resulted catalysts. Similar studies for conventional method (mix metal salts with carbon resource, which could be easily adjusted) have been well discussed recently. However, it is hard to achieve this goal for MOF-derived materials due to the complexity of MOF structures. The separated studies and unclear influence of MOF units hampered scientific and industrial progress of MOF-derived catalysts. Therefore, for the sake of minimizing unnecessary lab work and to avoid both the consumption of time and resources, the critical problem is to gain rational and comprehensive insight into the influence of MOF units (i.e. organic ligands, metal units and interactions of these two parts) on key parameters of catalyst performance via experiments and computational chemistry. Unfortunately, the related research in this aspect is still very scarce. Due to the complexity of MOF structures, the role of metal and linkers in M-N-C structures based on MOFs is much more difficult to study than that of M-N-C structures prepared via common or conventional methods which mix precursors of metals, carbon, and nitrogen. In order to further improve the performance of MOF-derived materials, it is urgent to find a suitable platform enabled by rational methodologies.
In this work, rechargeable metal-air batteries, which are one of the most promising energy storage devices for practical domestic/industrial applications because of their high theoretical energy density, environmental benignity, safety and low cost, [58][59][60][61][62] were chosen as the targeted application. For rechargeable metal-air batteries, OER and ORR are involved in the charge-discharge process. Therefore, designing an effective bifunctional catalyst is essential for rechargeable metal-air batteries. According to previous studies, 16

Tuned properties of M-N-C catalysts via MOFs units design (including thermal stability, carbon structure, N content, distribution, structure and electrochemical double-layer capacitances)
A series of benzene-1,3,5-tricarboxylate (BTC) based MOFs, which act as a tunable platform due to the well-defined two-dimensional (2D) layers, have been successfully prepared by anodic electrodeposition (Figure 1).  67 Compared with solvothermal methods, the visible and adjustable electrodeposition process is easier to control the reaction process and avoids the influence of anions from salts. 68, 69 SEM images of the as-prepared electrodes by anodic electrodeposition of Co-BTC-bipy at 60 °C and 120 °C are presented in Figure S4. Additionally, the feasibility of large scale production is important for practical industrial applications. Herein, when two electrodes connected in series, there is no difference for electrodeposition of Co-BTC-bipy ( Figure S5), which shows that this method could be used for preparing multi-electrodes simultaneously. The same cobalt plate was used 10 times as the substrate for anodic electrodeposition of Co-BTC-bipy. After each time, the cobalt plate was dipped into dilute HCl solution and polished with sand paper. After 10 times, the Co-BTC-bipy covered cobalt plate looks the same as that after the first time (the inset of Figure S5). This proves that our procedure has the potential for large scale production.
In order to get M-N-C catalysts, the as-prepared MOFs were pyrolyzed under N 2 atmosphere at different temperatures. Based on the thermogravimetric analysis (TGA) results ( Figure S6), it is concluded that the onset temperature of carbonization of the framework for Ni-BTC-DMF, Ni-BTC-bipy and Co-BTC-bipy are 395 °C , temperature of Co-BTC-bipy is lower than Ni-BTC-bipy. Before decomposition of the framework, there is one weight loss event each for Ni-BTC-bipy and Co-BTC-bipy, which corresponds to the release of adsorbed molecules (around 15% and 10% weight loss, respectively). In addition, when the linker is changed, there is no obvious decomposition temperature changes are observed between Ni-BTC-DMF and Ni-BTC-bipy. However, multi-steps weight loss events can be observed for Ni-BTC-DMF, with around 45% weight loss, which is probably caused by the loss of DMF molecules in the structure of Ni-BTC-DMF. 70 Based on the composition of Ni-BTC-DMF, DMF was considered as N source for M-N-C catalysts. Therefore, the loss of DMF molecules from the structure of Ni-BTC-DMF could lead to the low N content for Ni-BTC-DMF derived from Ni-BTC-DMF.
The composition and structure of M-N-C catalysts derived from MOFs were investigated by Raman spectroscopy, CHN analysis, XPS and TEM. Raman spectroscopy was used to analyze the degree of structural order with respect to a perfect graphitic carbon structure. Generally speaking, the D-band at ~1350 cm − 1 and a G-band at ~1585 cm − 1 are reflection of ideal and disordered graphitic lattices, respectively. The intensity ratio of D and G bands (I D /I G ) can be regarded as a measure of the carbon crystalline order of carbon materials. 71  The difference of I D /I G ratio between Co-BTC-bipy and Ni-BTC-bipy based catalysts could be caused by the different catalytic ability of different metals for carbon graphitization. The detailed mechanism of metal nanoparticles in catalytic graphitization is still not clear, which are probably related to carbon solubility and carbon diffusion coefficients in the different metal. 73 It is worth mentioning that Co-BTC-bipy-700 (I D /I G = 0.84) and Co-BTC-bipy-900 (I D /I G = 0.72) show even lower I D /I G ratio than multi-walled carbon nanotubes (MCNTs), whose I D /I G ratio is 0.92 ( Figure S10).
The bulk N content (at.%) of catalysts were explored by CHN analysis (Figure 2b). With increase of the pyrolysis temperature, the N content decreases for all the samples. Comparing with Ni-BTC-DMF and Ni-BTC-bipy, Ni-BTC-bipy derived catalysts (5.9 at.% for 700 °C and 2.8 at.% for 900 °C ) show much higher N content than Ni-BTC-DMF (1.9 at.% for 700 °C and 1.3 at.% for 900 °C ), which is related to the loss of DMF molecules in the structure of Ni-BTC-DMF during pyrolysis, which can be proved by TGA results (Figure S6).
These results reveal that the MOF ligand has a big influence on the N content of the as-prepared catalysts. When two MOFs which have similar N content in the structural unit are pyrolized, the resulting M-N-C catalyst show higher N content when the metal atoms are directly coordinated to the N atoms. In addition, the type of metal ion has some effect on the N content as well, which is probably caused by the difference in thermal stability and the ability to catalyze carbon growth between Ni-BTC-bipy and Co-BTC-bipy. A similar influence of metal particles on nitrogen content of N-C materials was observed by Kudashov et al. as well. 74 The N contents of Co-BTC-bipy based catalysts are 5.0 at.% and 2.2 at.% at the pyrolysis temperature of 700 °C and 900 °C , respectively.
Furthermore, the metal content was calculated from the CHN analysis and metal oxide residues (based on phase diagrams of Ni-O 2 and Co-O 2 binary systems shown in Figure S11), and the results are displayed in Table S2. Ni-BTC-DMF derived materials exhibit higher capacitance than Ni-BTC-bipy based materials. It is speculated that the intermolecular short contact interactions between coordinated pillar (organic solvent) and coordinated carboxylate group in Ni-BTC-DMF is weaker than that in Ni-BTC-bipy, leading to obviously release of DMF pillars. Therefore, more pores and space will be produced during pyrolysis. Concerning different metals, the detailed function of metal particles during pyrolysis is still unclear. It is well known that different metals have different catalytic ability for carbon growth, which could eventually lead to different carbon net structures.
Furthermore, the near-surface elemental compositions (atomic percentage) based on full scan survey XPS spectra for different catalysts are shown in Table S3. There is a huge difference between the full scan survey XPS spectra and CHN analysis for metal contents, because the metal nanoparticles are covered by a carbon layer while for XPS tests, the information depth is less than 10 nm. 19,75,76 Furthermore, the high-resolution XPS spectra of the Ni 2p  Table S5, the formation energies of N graphite -G (0.62 eV) structure are smaller than both of Ni-N graphite -C (4.20 eV) and Co-N graphite -C (5.34 eV). 89 However, the formation energies of N pyrid -G (3.32 eV) and N pyrro -G (10.21 eV) are dramatically reduced after forming the M-N-C structure. The formation energies of Co-N pyrid3 -C, Co-N pyrid4 -C, Co-N pyrro2 -C and Co-N pyrro3 -C are 2.61 eV, 1.31 eV, 7.70 eV and 6.59 eV, respectively. Therefore, the N would be enriched around metal particles and more for pyridinic N and pyrrolic N, which is in good agreement with XPS results as well. In addition, all the Ni-based M-N-C structures show smaller formation energies than corresponding Co-based materials. This might be the reason why Ni-based materials show higher N content than Co-based materials even with the same structure of the pristine MOFs and using the same pyrolysis condition. Meanwhile, based on the STEM-EELS mapping, the O signal seems to be always accompanied by C and N signals rather than metal signal, meaning that the metal nanoparticles are mainly present in their metallic form and M-N x compounds.
The HRTEM micrographs of Co-BTC-bipy-700 and Co-BTC-bipy-900 ( In this work, the performances of different potential microstructures as an ORR catalyst were evaluated theoretically by computing their free energy diagram for ORR with each elementary steps (four-electron way) at a fixed overpotential (η = 0.4 V), which is based on real conditions. The different N-heteroatom species and structures have a significant influence on the catalytic activity of overall ORR steps (Figure 6a). Firstly, all reaction steps on the Ni-N graphite -C surfaces are downhill except for the OH* formation, which is 1.44 eV uphill. This step is the rate-determining step (RDS) of the overall ORR for the Ni-N graphite -C catalyst. When Co replaces Ni, the OH* formation is thermodynamically favorable. However, the last step, OH* desorption, is 1.89 eV uphill. Hence, a four-electron oxygen reduction mechanism should be thermodynamically unfavorable on both Ni-N graphite -C and Co-N graphite -C surfaces. There are four kinds of potential configurations of M-N pyrid -C microstructures (Figure 6b). For the different Co-N pyrro -C microstructures, a four-electron ORR can only happen favorably on the surface of Co-N pyrro2 -C with all ORR steps going downhill. Therefore, the active sites of Co-N pyrid3 -C, Co-N pyrid4 -C and Co-N pyrro2 -C could be the primary catalytically active site for ORR in Co-N-C catalysts. It is worth stressing that Co-N pyrid4 -C is not the only active site in Co-N-C catalyst. In fact, Co-N pyrid3 -C and Co-N pyrro2 -C can play an important role as well, which is proven by DFT calculations for the first time. This useful information would clearly guide the further improvement of Co-N-C catalysts. Based on the DFT calculation results, it is concluded as follows: (i) the primary catalytically active sites for ORR in Co-N-C catalysts include Co-N pyrid4 -C, Co-N pyrid3 -C and Co-N pyrro2 -C; (ii) the N content of pyridinic N and pyrrolic N could play an important role in the ORR for both Co-N-C catalyst and Ni-N-C catalyst; (iii) Co-N-C catalysts could show better ORR catalytic performance than Ni-N-C catalysts. However, in the real conditions, the situation will be much more complicated. A lot of other factors would have influenced the ORR performance of the M-N-C catalysts. For example, there is no perfect graphene framework for loading M-N-C structure. In the following parts, we will discuss the present M-N-C catalysts for the ORR under practical conditions by electrochemical methods.

Enhancement of catalytic performance based on MOFs units design
The catalytic performance of as-prepared catalysts was explored by LSV (Figure 7). Compared to the blank electrode, all catalysts show obvious catalytic performance for ORR. Firstly, the effect of the pyrolysis temperature is studied (Figure 7a−7c). graphitization of these materials. Therefore, the real structure might be close to the DFT configurations.
Co-BTC-bipy-700 shows better performance than Co-BTC-bipy-900 because Co-BTC-bipy-700 has a higher N content (especially, pyridinic N and pyrrolic N are almost 5 times higher) and larger electrochemical surface area (i.e. higher electrochemical double-layer capacitance). These results are in good agreement with DFT and XPS results. As in this situation, the number of effective active sites might dominate the catalytic performance, which can be denoted as "N-dominated behavior". The Ni-BTC-bipy-700 shows slightly higher limiting current density than Ni-BTC-bipy-900 (Figure 7b). This is caused by the change of N content, which is similar to the Co-BTC-bipy based materials. On the other hand, this N-dominated behavior effect is not as strong as for Co-BTC-bipy based materials. This is possible because of the high I D /I G ratio of Ni-BTC-bipy. When the pyrolysis temperature increases, the I D /I G ratio decreases significantly, which would benefit the ORR. A similar phenomenon is also observed in Ni-BTC-DMF based materials. Therefore, for Ni-BTC-bipy based materials, the structure of the carbon, the electrochemical surface area and the N content show a synergistic effect for the ORR, which can be classified as "synergistic effect behavior".
Compared to Ni-BTC-DMF based catalysts, Ni-BTC-bipy based catalysts show better performance at both pyrolysis temperatures, although Ni-BTC-DMF based catalysts show smaller I D /I G ratio and higher electrochemical double-layer capacitance than Ni-BTC-bipy. This is caused by the large difference in the N content. With increasing pyrolysis temperature, the differences in catalytic ability between Ni-BTC-bipy-900 and Ni-BTC-DMF-900 are smaller than at low pyrolysis temperature. This is probably due to the decrease of the N content for Ni-BTC-bipy based catalyst at high pyrolysis temperatures and the lower I D /I G ratio of Ni-BTC-DMF based catalyst. Therefore, when the metal atoms are directly connected to N atoms in the pristine MOF, the resulting M-N-C material has more favorable catalytic behavior for the ORR. Co-BTC-bipy-700 exhibits better catalytic activity for ORR than Ni-BTC-bipy-700 (Figure 7e). This could be attributed to two reasons: (1) The metal ions of MOFs will lead to different active sites (Co-N x -C show better catalysts performance than Ni-N x -C), which is proven by DFT results as well; (2) Different metals have different ability to catalyze carbon growth, which will have an influence on the structure of the carbon after pyrolysis and on the electrochemical surface area.
The commercial Pt/C (10 %) catalyst was used as reference sample at same mass loading as Co-BTC-bipy-700.
loading of 20 μg cm −2 ). Interestingly, the onset potential of Co-BTC-bipy-700 is lower than Pt/C catalyst (Figure   8a), but with a larger limiting current density and similar half-wave potentials. While long-term stability is an important parameter for high-performance catalysts, Co-BTC-bipy-700 catalyst again outperforms the Pt/C catalyst.
After 2000 CV cycles (in the range of 0.5 V−1 V vs. RHE), the LSV curve of Co-BTC-bipy-700 catalyst shows only minor changes whereas the limiting current density of the Pt/C catalyst decreased a lot.
Furthermore, another advantage of Co-BTC-bipy-700 catalyst is the excellent tolerance to methanol (Figure   8b). With the addition of 1 M methanol, Pt/C catalyst shows an obvious peak of methanol oxidation and the cathodic peak corresponding to the ORR disappears. On the contrary, there is no change for the Co-BTC-bipy-700 catalyst, revealing its potential applications in direct methanol fuel cells. Moreover, considering economic and environmental impact, the recycling of catalysts is a big issue. Magnetic separation was recognized as a gentle and economical way to solve this issue. [94][95][96] The recycling of Co-BTC-bipy-700 catalyst by magnetic separation in 1 M KOH solution proves that magnetic separation is an effective way to recycle Co-BTC-bipy-700 catalyst ( Figure   S23 a−b). Even after one month, the Co-BTC-bipy-700 catalyst can be easily recycled by magnetic separation (Figure S23 c-d), which can be attributed to the protection of the cobalt by the carbon layer.
To know the electron transfer process of the ORR involving the as-prepared samples, LSV on RDE were recorded at rotation speeds from 400 rpm to 2000 rpm as shown in Figure S24a (Figure 9b). Generally speaking, lower Tafel slopes imply faster kinetics. 97 The Tafel slopes calculated from Figure 9b are given in Table S6. All the MOF derived catalysts show low Tafel slopes, ranged from 53 mV dec −1 to 64 mV dec − 1 . Obviously, the Tafel slope of Co-BTC-bipy-700 catalyst (53 mV dec −1 ) is the lowest. Compared with Pt/C whose Tafel slope is 74 mV dec −1 , the Co-BTC-bipy-700 catalyst exhibit more desirable ORR kinetics.
In order to figure out the role of the metal in Co-BTC-bipy-700 catalyst for ORR, cyanide anions (CN − ) was used as a molecular probe to explore active ORR catalytic sites on Co-BTC-bipy-700 catalyst as they readily coordinate with the Co center. 98 In addition, the control electrode of cobalt-covered RDE was prepared by electrodeposition. The cobalt covered RDE did not show catalytic activity for ORR (Figure S32a), indicating that metallic cobalt is not the active site for ORR, justifying that the contribution of metallic phase to catalytic performance can be ignored. After the addition of CN − , the onset potential and the diffusion-limiting current density of the ORR polarization curve of the Co-BTC-bipy-700 electrocatalyst decrease significantly (Figure S32b Another key reaction for rechargeable metal-air battery is the OER, 47,99 corresponding to the charging process. The OER catalytic activity of Co-BTC-bipy-700, Pt/C and IrO 2 were explored by LSV on RDE at a rotation speed of 1600 rpm in 0.1 M KOH. Compared to Pt/C and IrO 2 catalysts, Co-BTC-bipy-700 catalyst shows lower onset potential (1.54V vs. RHE) and lower potential at a current density of 10 mA cm −2 (1.63 V vs. RHE) as shown in Figure 10a. Based on Tafel plots obtained from Figure 10a, the Tafel slopes of Co-BTC-bipy-700 and Pt/C and IrO 2 are 77, 98 and 298 mV dec −1 , respectively, indicating that the OER was kinetically faster on the Co-BTC-bipy-700 catalyst. To assess the OER stability of catalysts, chronoamperometric test is carried out at 1.7 V Figure 10c, IrO 2 catalyst suffers a rapid current loss at the initial stage and the total current loss was about 66% after 24,000 s. On the contrary, more than 95% current can be remained for Co-BTC-bipy-700 catalyst, indicating a high stability for OER. After stability tests, there is no obviously change on the LSV for Co-BTC-bipy-700 catalyst (Figure 10d).

Home-made Zn-air battery performance
A home-made Zn-air battery was used to study the catalyst performance under real conditions. A Zn plate Clearly, Co-BTC-bipy-700 shows higher polarization current density than Pt/C-IrO 2 (Figure 11a), which benefits from better ORR and OER performance of Co-BTC-bipy-700. Moreover, the corresponding power density and current density during the discharging is shown in Figure 11b. The maximum power density of the assembled zinc-air battery based on Co-BTC-bipy-700 is determined to be 336 mW cm −2 , which is higher than that of the zinc-air battery based on Pt/C-IrO 2 (270 mW cm −2 ). In order to compare with previous works, the Tafel slop for each reaction and the overvoltage between ORR and OER ( OER@10 mA cm −2 − ORR@3 mA cm −2 base on the three-electrode system) is used as a descriptor for evaluating the bifunctional electrocatalytic activity of a catalyst ( Table S8). The Co-BTC-bipy-700 catalyst shows that the value of OER@10 mA cm −2 − ORR@3 mA cm −2 reaches 0.84 V, which is comparable with the highest level of bifunctional electrocatalysts reported recently (Table S8) Figure S34. Even if Co-BTC-bipy-700 is not the best ORR electrocatalyst, it surpasses most bifunctional electrocatalysts reported to date considering both ORR and OER, leading to the high Zn-air battery performance. Compared with Pt/C-IrO 2 catalyst, the Co-BTC-bipy-700 catalyst enables the Zn-air battery a higher voltage platform during the discharge process (benefiting from the high catalytic activity for ORR) and a lower voltage platform during the charge process (benefiting from the high catalytic activity for OER) at all the tested current densities (Figure 11c). At a current density of 3 mA cm −2 , the specific energy density of the Zn-air battery based on the Co-BTC-bipy-700 catalyst equals 1009.8 Wh kg −1 (based on the mass of Zn), which is around 76.5% of the theoretical value (1320 Wh kg −1 ). Meanwhile, the specific energy density of the Zn-air battery based on the whole system (including anode, electrolyte, cathode, and all other components of the device) is 4.37 Wh kg −1 . Furthermore, long term charge-discharge tests were carried out with different catalysts. As shown in Figure S35, after 45 cycles, the degradation of the discharge voltage platform for Zn-air batteries based on Pt/C-IrO 2 and Co-BTC-bipy-700 catalysts are around 8%. The decrease of catalyst performance is probably caused by the passivation of the zinc plate side and the corrosion of the carbon structure or the carbonate precipitation on the gas diffusion electrode side. [100][101][102][103] The Zn-air battery based on Co-BTC-bipy-700 catalyst shows lower charging voltage and higher discharging voltage at all cycles than Pt/C catalyst. In addition, the design diagram of the cell and the demonstration of the Zn-air battery (based on Co-BTC-bipy-700) with an OCV of around 1.51 V are given in Figure S36.  Table S1.

Conclusions
Material characterizations: X-ray diffraction (XRD) patterns of the samples were recorded on a Bruker AXS D8 diffractometer using Cu Kα radiation (λ = 0.15405 nm) and Ni filter with 2θ ranging from 5° to 30° and a step size . For comparison, multi-walled carbon nanotubes (MCNTs, Nanocyl, Belgium), Pt/C catalyst (10% of platinum, Alfa Aesar Germany) and IrO 2 (P40V020, particles from Premetek Co.) were conducted on the same electrochemical tests with the same mass loading. Here, the mass loading Pt is around 50 μg cm −2 , which is higher than the "standard" loading (20 μg cm −2 ). 36 In addition, the cobalt-covered electrode was electrodeposited at a current density of -40 mA cm −2 in CoCl 2 (0.9 mol L −1 ) aqueous solution with LiClO 4 (0.1 mol L −1 ) as the supporting electrolyte at room temperature until a mass loading of 0.50 mg cm −2 (based on Faraday's law) was obtained on the GCE. The details were shown in previous work. 70,104 As the catalysts in this study are composite materials, electrochemical surface area (ECSA) measured by electrochemical double-layer capacitance or Brunauer-Emmett-Teller (BET) might lead to errors for current normalization. 105 Therefore, the geometrical surface area of the GCE and constant catalyst mass loading for all catalysts were used as current normalization method.  where i is the current and R is the uncompensated ohmic electrolyte resistance measured by EIS at high frequency.
Electrode kinetic data were calculated based on Koutecky-Levich (K−L) equation:  Rechargeable Zn-Air battery tests: For the full-cell test of our catalysts, we prepared a zinc plate as an anode, and 6 M KOH with 0.2 M zinc acetate solution was used for the electrolyte (Figure S36). The different air cathodes were prepared by a mixture of different catalysts with 5 wt.% Nafion solution with the same ratio for half reaction test (ORR and OER), and a mass loading of 3.5 mg cm −2 on a gas diffusion electrode (carbon fiber paper). All the Zn-air batteries were tested under ambient atmosphere without a separator for anode and cathode. An assembled full-cell was characterized by chronopotentiometry (CP) at several charge and discharge currents.
First-principles calculations: All calculations were performed using plane-wave density functional theory (DFT) employing periodic boundary conditions as implemented in the VASP code 110,111 . The electronic exchange-correlation energy was modeled using the Perdew-Burke-Ernzerhof (PBE) functional 112 within the generalized gradient approximation (GGA). The projector augmented wave (PAW) method was used to describe the ionic cores 113,114 . For the plane-wave expansion, a 550 eV kinetic energy cut-off was used after testing a series of different cut-off energies. To model metal-nitrogen-carbon (M-N-C) structures, the (5×5×1) supercell M, N co-doped graphene with lattice parameters of a = b = 12.28 Å was constructed by a periodic boundary condition, and the vacuum layers were set to be larger than 20 Å to avoid periodic interaction. A Monkhorst-Pack 3 × 3 × 1 k-point grid was used to sample the Brillouin zone 115 . The convergence criterion for the electronic structure iteration was set to be 10 -4 eV, and that for geometry optimizations were set to be 0.01 eV/Å on force. To better describe the dispersion interaction within water adsorption systems, vdW correction was considered by adopting the Grimme's D2 scheme 116 .
In order to understand the mechanism and the active site of the oxygen reduction pathway on the M, N co-doped graphene, we use the free energy change (G) for each ORR step as the criterion for assessing the ORR activity.
Specifically, we considered the four-electron reaction pathway for ORR in alkaline media, as listed in Supplementary equations (2) to (5), which is the dominated mechanism for doped graphene catalysis 117  where ΔE, ΔZPE, and ΔS are the changes of DFT total energy, zero-point energy, and entropy from the initial state to the finial state, respectively; T is temperature; U is the electrode potential; e is the charge transfer. ΔZPE, and ΔS can be obtained by the thermodynamics table for gaseous molecules 119 and by calculating the vibrational frequencies for the oxygenated intermediate (see Table S7), respectively.    maps of C (at 94.0 ± 2.9 %) and N (at 6.1 ± 2.9%) for Co-BTC-bipy-700 and the relative composition maps of C (at 94.9 ± 11.6 %) and N (a: 3.7 ± 2.3%) for Co-BTC-bipy-900.     Co-BTC-bipy-700 catalysts at different current densities after 30 cycles.