The rapid consumption of fossil fuels, together with continuously increasing population and over exploitation of forests, is the major culprit for the high atmospheric CO₂ level. Excessive CO₂ concentration will exacerbate the greenhouse effect, causing average temperature to rise and posing great challenges to global climate and environment [1]. Though notorious, CO₂ is also a potential carbon resource [2-4]. Electrochemical CO₂ reduction reaction (CO₂RR) is a promising technique for efficient CO₂ conversion. CO₂RR technique can utilize electricity generated by clean renewable energy and directly convert CO₂ into value-added chemicals (CO, formate, hydrocarbon, oxygenated compounds, etc.), making it possible to realize an ideal recycling cycle for carbon resources [5-8]. However, the undesirable energy conversion efficiency caused by a high overpotential in initiating CO₂ conversion and limited selectivity due to the competitive hydrogen evolution reaction (HER) in aqueous solution discourage the near-future practical applications of CO₂RR [9-12]. Therefore, intensive research work has been focused on developing highly efficient catalysts. From early well-studied bulk metal catalysts to recent widespread metallic nanostructure/nanomaterials, researchers have done intense work on the investigation of structure-activity relationship towards CO₂RR through constructing active nanostructure for more exposed active sites, modulating nanoparticle size and active facet ratio, introducing second alloy constituent, utilizing metal-oxide interface confinement effect, etc., to optimize the catalytic performance and further confirm that the binding energy and structure of absorbed intermediates are crucial for high selectivity [13-20]. At present, Au-, Ag-, Pd- and Cu-based catalysts have been widely investigated for CO₂RR and exhibit considerable Faradaic efficiencies, yet the low mass activity and scarcity of these noble metal catalysts impede their future applications [9, 21-24]. Alternatively, catalysts containing transition metal-nitrogen (M-N_x) sites, such as transition metal-centered macrocyclic complexes, metal organic frameworks and M-N_x-doped carbon materials, which show comparable or even higher activity and selectivity towards CO₂RR than noble metal counterparts, are emerging as promising candidates [25-27]. In particular, low-cost M-N_x-doped carbon materials synthesized by a facile pyrolysis of precursors containing N, C and transition metal elements, which have been widely explored as the most promising alternatives for noble metal-based catalysts in polymer electrolyte membrane fuel cells [28-30], have shown remarkable activity and selectivity towards CO₂RR and triggered a wave of research in recent years [11, 31-33]. This review summarizes recent advances on the CO₂RR using a variety of M-N_x-containing materials, and practical parameters, such as synthesis protocols, characterization of active sites, and electrocatalytic performance evaluation are included, while we also discuss about the proposed reaction mechanisms and major factors that affect product distribution. We expect this review will provide practical knowledge into the rational design of improved M-N_x-containing catalysts and exploration of optimal process conditions, which are crucial in promoting the practical application of CO₂RR in the near future.

Transition metal-centered macrocyclic complexes, either in solution (homogeneous catalysis) or anchored to carbon materials/electrode (heterogeneous catalysis), have been explored as efficient catalysts for CO₂RR since the 1970s. In particular, metal porphyrins, phthalocyanines and their relative derivatives (Fig. 1) have been widely studied. Examination of product selectivity on gas diffusion electrodes (GDE) impregnated with 17 kinds of metal phthalocyanines by Matsui et al. [34] showed that with transition metals of Group Ⅷ as the coordination center, CO was produced as the major electrolysis product and the Faradaic efficiency on Co and Ni phthalocyanines almost reached 100%; in the case of Sn, Pb or In (Group ⅢA or ⅣA) phthalocyanines, formic acid was the major product; while methane was produced mainly on Ga, Cu and Ti phthalocyanines, with the highest Faradaic efficiencies of 30%–40%. Followed up the research work, Furuya et al. [35] interpreted such discriminative difference qualitatively by Taube's theory, i.e. the electron configurations in the metallic atom affected CO₂ activation and CO desorption, consistent with that proposed by Meshitsuka et al.. Based on earlier studies, more efficient macrocyclic complexes with varying metal centers (mainly Co, Fe, Mn and Zn) and tailorable ligands are developed by structural engineering at the molecular level or constructing hybrid architecture at the nanoscopic scale [36-43].

At the potentials required for CO₂RR, the metal centers of metal phthalocyanines or other macrocyclic complexes are prone to be reduced to a low oxidation state while the macrocyclic ligand framework contributes to maintain chemical stability and store a proton and/or a reducing equivalent. Thus the intrinsic structure of active sites has been controversial all the time and intense efforts have been made to detect reaction intermediates and theoretical protocols are used to shed light on the reaction mechanisms. Savéant et al. [39, 44] have done much work on improving the performance of iron tetraphenylporphyrin (FeTPP) derived molecular catalysts. Cyclic voltammetry results of the redox process of Fe oxidation state with variation of potential revealed that electrogenerated iron(0) porphyrins are active sites for CO₂RR and prepositioned phenol functionalities help improve the activity by acting as both H-bonding stabilizers and high-concentration proton donors (Fig. 2). Similarly, experimental and theoretical studies confirmed that Co center is first reduced to Co(Ⅰ) and then CO₂ is adsorbed and further reduced to CO [41, 42, 45, 46]. In a recent study, Wang et al. [47] discovered that a Cu phthalocyanine based heterogeneous catalyst efficiently converts CO₂ to methane and ethylene. Interestingly, in-situ and operando X-ray absorption spectroscopy (XAS) results in Fig. 3 revealed that ~2 nm metallic copper clusters formed through the reversible oxidation state and structural changes of copper(Ⅱ) phthalocyanine serving as active centers for the production of methane.

Besides the metal center, the ligands are also crucial to improve the CO₂RR performance. Costentin et al. [39] found that the high local concentration of protons derived from all the ortho and ortho' phenolic hydroxyl substituent group of the parent FeTPP (Fig. 2a) remarkably speeded up the CO₂RR, achieving a CO Faradaic efficiency above 90% in a mixed DMF-water solvent. In a recent study, carbon electrodes covered with perfluorinated cobalt phthalocyanine (CoFPc) exhibit much higher CO₂RR performance than the parent CoPc with a maximum CO Faradaic efficiency of 93% as shown in Fig. 4a and 4b [41]. The Co(Ⅱ)/Co(Ⅰ) redox wave and spectroelectrochemical signal of metal-to-ligand transfer confirmed the formation of active Co(Ⅰ)Pc species. They propose that the electron-deprived CoFPc promotes the desorption of associated CO and facilitates the active Co(Ⅰ) regeneration, thus CoFPc is more selective and active than CoPc. Similarly, Wang et al. [42] immobilized CoPc onto carbon nanotubes and found that the introduction of electron-withdrawing cyano groups to the CoPc further improved the CO₂RR activity with much higher current density and CO Faradaic efficiency exceeding 95% (Fig. 4c and 4d).

Most macrocyclic complexes suffer from low conductivity and poor solubility in water, thus anchoring them onto electrode surface or solid supports (usually carbon materials) makes them possible to act as efficient heterogeneous catalysts towards CO₂RR in aqueous electrolytes. Non-covalent adsorption, covalently bonding and polymerization are common protocols that are used to construct hybrid catalysts [25]. Unequivocally, a proper grafting method should take a suitable support material for better electron transfer and avoid blocking catalytic sites or impeding conductivity. In a comparative study, cobalt meso-tetraphenylporphyrin (CoTPP) alone shows a poor CO₂RR performance as a homogeneous catalyst; however, a facile immobilization of CoTPP onto carbon nanotubes remarkably facilitates the CO₂RR, resulting in a substantial enhancement with much improved CO Faradaic efficiency (> 90%) and lower overpotential in aqueous medium [48]. Recently, Li et al. [43] employed an organic-inorganic hybridization strategy to construct cobalt polyphthalocyanine (CoPPc)-sheathed carbon nanotubes (CoPPc/CNT) for highly efficient CO₂RR (Fig. 5a). The thin conformal polymer coating layer achieved by in-situ polymerization of CoPc molecules helps suppress the aggregation of organic molecules and promote the exposure of more Co-N sites, and thus contributes to an excellent performance with a high CO Faradic efficiency of ~90%, considerable turnover frequency (TOF) of 4900 h^–1 at η = 0.5 V, and stability of > 24 h, much superior to CoPc and CoPPc alone (Fig. 5b and 5c).

As mentioned above, polymerization of molecular catalysts on support materials is a wise way to improve the density of active sites [43, 49]. However, it is not an optimal way for full accessibility of active sites due to the stacking of densely packed structure. Metal-organic frameworks (MOFs) as a class of coordination polymers with open frameworks, consist of metal ions that are linked together by organic ligands [50]. With the large availability of different metal centers, multiple organic ligands and various building blocks, the structure and functionality of MOFs can be tailored to a large extent for targeted reaction [50, 51]. In addition, such ordered porous structures not only promote the exposure of active sites but also facilitate mass transfer during the catalytic reaction, making MOFs ideal materials for catalysis [52, 53].

Recently, following the urgent demand for efficient CO₂ conversion process and renewed interest in CO₂RR, MOFs containing Cu, Fe, Co, Re and Zn, etc., have been investigated as efficient catalysts for CO₂RR (Fig. 6) [54-59]. Similar to the case in M-N_x-containing macrocyclic complexes, MOFs suffering from low conductivity are always anchored to conductive surfaces or mixed with conductive carbon materials in CO₂RR [54, 56, 59]. Besides, modification or replacement of linked ligands with better electron/proton transfer ones also significantly help improve the charge/electron transfer capability [60-62]. Yamada et al. [63] first reported a Cu-containing MOF dropped on carbon paper for CO₂RR in 2012. A copper rubeanate (CR-MOF) was chosen owing to its excellent electronic and proton conductivity arising from the coalescent of a proton with the nitrogen atom and accompanying of electron transfer from Cu(Ⅱ) ions. CO₂RR measurements were conducted in aqueous electrolyte and the results showed that the CR-MOF electrode endowed much lower overpotential, namely 0.2 V less than that of a Cu metal electrode, and HCOOH was the primary product with a Faradaic efficiency of 30%. Later work from Kulandainathan et al. [64] manifested that Cu₃(BTC)₂ thin film deposited on glassy carbon (GC) electrode reduced CO₂ to oxalate with a Faradaic efficiency of 51% in CO₂-saturated N, N-dimethylformamide containing tetrabutylammonium tetrafluoroborate. They proposed that Cu(Ⅰ) formed at –0.62 V vs. Ag/Ag⁺ in the CO₂-saturated DMF solution catalyzed the reduction and the adsorbed CO₂ gas inside the pores facilitated the CO₂RR.

Yaghi et al. [65] first tried to incorporate CO₂RR-efficient Co porphyrin building units into covalent organic frameworks (COFs, a subgroup of MOFs with better conductivity due to conjugated double bonds) by organic ligands through imine bonds (Fig. 7a and 7b). They reported that COF-366-Co synthesized by the connection of 5, 10, 15, 20-tetrakis(4-aminophenyl) porphinato cobalt [Co(TAP)] with 1, 4-benzenedicarboxaldehyde (BDA) displayed a high CO Faradaic efficiency (90%) when deposited on conductive carbon fabric for CO₂RR measurements. In addition, modifications through modular protocols by appropriate combinations of building units further promote the reaction activity. When biphenyl 4, 4′-dicarboxaldehyde (BPDA) with longer chain was used as the strut substituting for BDA, lattice expanding results in a larger pore radius and allows for a higher CO₂ adsorption capacity and increased accessibility of the active sites. Copper porphyrin, which is catalytically inactive for CO₂RR, was used to dilute cobalt porphyrin active sites and further improve the turnover frequency of active Co sites. Consequently, COF-367-Co (1%) with only 1% Co in all metal sites exhibited an initial TOF_EA of 9400 h^–1, equivalent to a 26-fold higher than the activity of the parent Co(TAP). In another work, the same group took Al₂(OH)₂TCPP-Co (TCPP-H₂ = 4, 4', 4'', 4'''-(porphyrin-5, 10, 15, 20-tetrayl)tetrabenzoate) as the linker units and assembled them into a porous thin film MOF directly grown on a conductive substrate through atomic layer deposition (ALD) technique (Fig. 7c–e) [56]. The thickness of the MOF film was optimized by precise regulation of ALD cycles and such catalytic system displayed a CO Faradaic efficiency exceeding 76% and remained stable over 7 h. Operando spectroelectrochemical measurements showed that the majority of Co(Ⅱ) catalytic centers in the MOF were electro-reduced to Co(Ⅰ) and subsequently catalyzed the CO₂RR. The direct growth of the MOF onto the conductive surface is indeed an effective way for increasing the portion of electroactive sites and enhancing the charge-transport properties for MOFs materials, thus leading to a higher CO₂RR performance.

MOFs composed of metal nodes and organic linkers potentially combine the advantages of both molecular and heterogeneous catalysts. A precise manipulation of its building units allows for changes on both spatial arrangement (i.e., porous structure, topology of the structure) and electronic environment, offering a great possibility for efficient CO₂RR. Yet, the low conductivity of MOFs still restricted the exposure of electro-active sites, leading to a limited current density.

Pyrolyzed metal-nitrogen-carbon materials have been widely used as promising candidates for the oxygen reduction reaction (ORR) owing to their low cost and earth-abundant components. Progress in the synthesis, characterization, ORR performance and reaction mechanisms is well summarized in related review papers [66-68]. Strasser et al. [11] firstly tried to use porous carbon materials doped with M-N_x sites, derived from the pyrolysis of nitrogen and Fe- and/or Mn-functionalized commercial carbon for CO₂RR, with CO as the major product and a trace amount of CH₄ was detect on Fe-containing catalysts (Fig. 8). Their results confirm that metal-nitrogen-carbon (M-N-C) materials are efficient in CO₂RR and have stimulated an academic research boom on carbon-based metal-nitrogen materials towards CO₂RR. In analogy to the 4-electron ORR, electroreduction of CO₂ to CO or hydrocarbons involves multiple coupled protonation/reduction steps and thereby there are some common strategies for constructing highly effective M-N_x-containing carbon materials towards CO₂RR.

Carbon-based M-N_x-doped materials are always prepared by a facile pyrolysis of mixtures composed of metal-, nitrogen- and carbon-containing precursors. During the pyrolysis, N- and C-containing precursors decomposed and transformed into N-doped carbon, while metal sites coordinated to N atoms were simultaneously embedded into the skeleton. Thus the mixed precursors and appropriate pyrolysis parameters concerning carbonization temperature and duration time are of vital importance for constructing atomically dispersed M-N_x-containing materials [30, 69, 70]. Notably, to avoid the formation of metal nanoparticles which mainly promote the competitive HER, precursors with well-dispersed M-N_x structures or spatial confining metal sites are preferred and acid-leaching process is available for removing these undesirable nanoparticles. Recently, Jiang et al. [71] developed a multistep pyrolysis process to synthesize a class of single-atom catalysts on nitrogen-doped carbon nanotubes (MSA-N-CNTs) with an extraordinary high metal loading. The reconnaissance study revealed that Ni-g-C₃N₄ with Ni single atoms located in the cavity of g-C₃N₄ formed when the mixture was further pyrolyzed at 650 ℃ followed by the first heat treatment at 350 ℃. Thanks to the strong covalent bonding between g-C₃N₄ and Ni atoms, aggregation of Ni atoms was inhibited at higher temperature (700–900 ℃), giving rise to the atomically dispersed NiSA-N-CNTs. MOFs or ZIFs (zeolitic imidazolate frameworks), in which metal nodes are separated by organic linkers (especially N-containing ligands), have been used as ideal precursors to help increase the loading of atomically dispersed metal sites within the final porous carbon materials. In particular, metal residing in MOF/ZIF can be partly exchanged with other metals, thus allowing further increasing of the spatial distance of metal sites. For example, the Zn²⁺ sites in ZIF-8/ZIF-7 are replaceable with Co²⁺, Ni²⁺ and Fe³⁺/Fe²⁺ ions, and subsequently serve as perfect spacer for adjacent Co/Ni/Fe atoms to avoid the agglomeration of metal atoms during the pyrolysis. Typical work on mixed metal ZIF precursors is shown in Fig. 9 [31, 70, 72]. Namely, Zn atoms with low boiling point evaporated away at temperatures over 800 ℃, while Co/Ni/Fe nodes are reduced in situ and anchored in the carbon skeleton with N-coordination. Li's group [70, 73] has done intense work on the synthesis of single-atom catalysts by pyrolyzing ZIF-related materials, and they also found that coordination number could be regulated by pyrolyzing temperature because M–N coordination bonds were vulnerable at higher temperature. Our work on ZIF-derived M-N-C catalysts further highlights the vital role of doping methods. We disclosed that introducing a second transition metal ion (Ni²⁺ or Fe³⁺) during the crystallization of ZIF-8 could maximize the doping concentration of atomically dispersed metal sites compared with the post-synthetic ion exchange method, while post impregnation with specific metal salts (ammonium ferric citrate) helped to confine the Fe³⁺ on the surface of ZIF-8, leading to a high exposure of iron-nitrogen sites [31, 32].

Similar to the rules explored in M-N_x-containing macrocyclic complexes, metal centers and their coordination state play an important role in product selectivity. Recent work on M-N_x active sites for CO₂RR revealed that Fe-N, Co-N, Ni-N, Mn-N, Zn-N and Sn-N sites are effective towards CO₂RR [31, 32, 73-82]. Fontecave et al. [74] examined a series of iron-based catalysts containing Fe-N sites for CO₂RR and unveiled that FeN₄ sites facilitated CO₂RR to CO and Fe-based nanoparticles favored the production of undesirable H₂ (Fig. 10a). Thus by synthetically regulating the ratio of Fe-based nanoparticles to isolated FeN₄ sites, products with controlled CO/H₂ ratios can be obtained. In addition to the intrinsic nature of active sites, the mass transfer ability and high exposure of active sites are also crucial for the catalytic performance [83-85]. Recently, our group [32] took a post-synthetic modification strategy to prepare highly exposed Fe-N active sites through surface functionalization of ZIF-8 with ammonium ferric citrate (AFC). The C-AFC©ZIF-8 with highly exposed Fe-N sites exhibited superior selectivity and mass activity towards CO₂RR than its counterpart derived from bulk functionalization of ZIF-8 with similar Fe loading, reaching a maximum FE_CO of 93.0% at –0.43 V (Fig. 10b). We also employed a post-synthetic strategy by a second pyrolysis in ammonia to further improve the performance of Fe-N sites [75]. With aid of the etching effect of ammonium at high temperature, the removal of residual Zn species and vulnerable carbon moieties further maximized the exposure of Fe-N active sites and facilitated mass transfer during the CO₂RR. Therefore, the optimal catalyst after an additional pyrolysis in ammonium showed a remarkable increase in both FE_CO and current density, exceeding other efficient Fe-N-C catalysts (Fig. 10c). Lin et al. [86] developed a composite material composed of pyrolyzed ZIFs and carbon nanotubes for CO₂RR, which manifests nearly 100% CO Faradaic efficiency at an overpotential of 740 mV. Moreover, when little amount of Fe was doped into the catalyst, the overpotential was further decreased to 440 mV. They ascribed such superior activity to the enhanced electron transport and expedited CO₂ transport resulted from the MWCNT support. Chen et al. [76] found that the CO Faradaic efficiency drops with decreasing Co-N_x coordination number. CoPc was loaded onto hollow N-doped porous carbon spheres (HNPCSs) starting from polymer precursors by efficient coordination with abundant N sites in HNPCSs support. EXAFS fitting results revealed that the Co-N coordination number is 5. Co-N₄/HNPCSs and Co-N₃/HNPCSs catalysts were obtained by pyrolyzing Co-N₅/HNPCSs at 400 and 600 ℃, whereas higher temperatures (800 and 1000 ℃) result in the formation of NPs. Electrocatalytic results display that CO Faradaic efficiency remains stable at about 90% from –0.57 V to –0.88 V vs. RHE on Co-N₅/HNPCSs, superior to other counterparts.

As presented above, Fe-N-C/Co-N-C catalysts demonstrated impressive CO Faradaic efficiency at relatively low overpotentials; however, a drawback for future application is that the Faradaic efficiency drops rapidly accompanying an increase in the overpotential for much higher current density [11, 32, 73-75]. Ni-N-C materials were also examined for CO₂RR and the results showed that Ni-N-C catalysts outperformed Fe-N-C/Co-N-C counterparts at slightly larger overpotentials [77-80].

Coordinatively unsaturated Ni-N_x active sites embedded in porous carbon were synthesized by facile heat treatment of Zn/Ni bimetallic ZIF-8 at 900 and 1000 ℃. Nickel was doped during the crystallization process and the maximum loading of atomically dispersed Ni reached 5.44 wt%. CO₂RR measurements in 1 mol L^–1 KHCO₃ electrolyte showed that FE_CO was kept between 92.0% and 98.0% in a higher overpotential window ranging from –0.53 V to –1.03 V (vs. RHE), and thereby CO current density increases with the applied potential and reaches 71.5 ± 2.9 mA cm^–2 at –1.03 V (vs. RHE) (Fig. 11a and 11b), breaking the limit in Fe-N-C/Co-N-C materials [31]. Based on the unsaturated coordination structure of Ni-N_x sites obtained from XAS analysis, density functional theory calculations stimulated three Ni-N active sites (NiN₃, NiN₃V and NiN₂V₂) and free energies of major intermediates were employed to unveil the activity trend. Specifically, theoretical results revealed that the free energies of ^*COOH (G_^*COOH) on coordinatively unsaturated NiN₃, NiN₃V and NiN₂V₂ were significantly lower than those on NiN_4, and the *H blockage was relatively weak on NiN₃V and NiN₂V₂ structures, confirming that the high CO₂RR performance resulted from those coordinatively unsaturated Ni-N sites (Fig. 11c and 11d).

As mentioned above, various transition metal-functionalized carbon/nitrogen precursors and different approaches were employed to construct carbon-based metal-nitrogen materials, leading to discrepancies in carbon architectures (porosity, composition and structure), M-N coordination structure and nitrogen doping state (contents and configurations). In order to understand the role of metal centers and identify a champion metal for the construction of optimal M-N_x-containing carbon materials towards CO₂RR, atomically dispersed M-N_x materials with different transition metals should be prepared by the same method, minimizing the effects of carbon and nitrogen variables. Strasser et al. [87] synthesized a family of porous carbon materials doped with active M-N_x moieties (M = Mn, Fe, Co, Ni, Cu) with transition metal salts and bipyridine-based coordinated polymers as initial precursors for more profound understanding of reactivity trends towards CO₂RR (Fig. 12a). The CO₂RR performance greatly depended on the transition metals, both in terms of CO Faradaic efficiency and overpotential, among which Fe-N-C and Ni-N-C outperformed other metal-related porous carbon showing much improved activity while the maximum CO Faradaic efficiency was obtained at a lower overpotential on Fe-N-C (V_RHE = –0.55 V, FE_CO = 65%) than on Ni-N-C (V_RHE = –0.78 V, FE_CO = 85%) (Fig. 12b). DFT simulations of reaction pathways on these catalytically active M-N_x moieties agreed well with the experimental results (Fig. 12c). Consistent with the experiments, Co-N_x sites holding downhill energetics for hydrogen suffered from severe energetic barriers to CO production; while larger overpotentials were required for Ni-N_x sites to start the reaction resulting from the low H^* and CO^* binding energy and at larger overpotentials hydrogen evolution was hindered on Ni-N_x catalysts with much less energetic barrier for CO production. Reported work from Daasbjerg's and Li's groups [33, 88] also confirmed the notable CO Faradaic efficiency of Fe-N-C and Ni-N-C with Ni as the champion metal center even at much higher overpotentials.

Despite of the intrinsic activity of the catalysts, reaction parameters such as pH value of the electrolyte and reactant CO₂ pressure also affect the selectivity and activity of CO₂RR. Sonoyama et al. [89] found that metal-meso-tetraphenylporphyrin (MTPP) supported on GDEs exhibited much enhanced activity with less overpotential and higher selectivity at 20 atm compared with at ambient pressure. Higher concentration and faster diffusion rate of CO₂ may account for such enhancement. Koper et al. [90] carried out a detailed pH-dependent study for an immobilized cobalt protoporphyrin catalyst, showing that at pH = 1 hydrogen evolution reaction dominated and little CO and CH₄ formed compared with at pH = 3, and further CO electroreduction to CH₄ prefers a more acidic environment (at pH = 1), as shown in Fig. 13a. The observation that CO₂ electroreduction to CO becomes much more dominate at a high pH, indicating that CO₂ activation is not sensitive to the proton concentration, which is obviously different from the competing HER. This result agrees well with the assumption that the formation of a catalyst-bound CO₂^·- radical anion possessing a strong Br nsted-base character is crucial for initiating CO₂ reduction, and therefore the proton donor may be water rather than H⁺ (Fig. 13b). Analogously, Strasser et al. [91] studied the effect of pH on the CO₂RR performance of Fe-N-C catalysts, especially focusing on the selectivity of H₂, CO and methane to investigate the role of proton concentration on the catalytic reactions. CO₂RR measurements were performed in CO₂-saturated electrolyte at different pH values between 1 and 7.25. When plotting production rate versus the applied potential (vs. NHE) at different pH values (Fig. 14a), it was obvious that H₂ and CH₄ formation depended on the pH, but CO formation was independent of pH. The pH dependence observed for CH₄ and H₂ production was consistent with a rate-determining step involving a coupled-proton-electron transfer. In contrast, the unveiling pH independence for CO formation indicated a decoupled electron-proton transfer process (DEPT) in which CO₂^·- anion adduct was a key intermediate (Fig. 14b and 14c) similar with the mechanism proposed by Koper et al. [90, 92] in Co(Ⅱ)-N-C complexes. Based on the guidance of the above reaction mechanism, important catalyst design rules bearing the DEPT mechanism involve: (ⅰ) Reducing the barrier for the formation of the initial catalyst-bound CO₂^·- radical anion by enhancing the interaction between active sites and intermediate is an effective way to decrease the overpotential of CO₂RR; (ⅱ) A suitable adjustment of local pH by employing electrolyte with appropriate bulk pH and buffer capacity helps to tune the product ratio leading to optimal Faradaic efficiency; (ⅲ) To increase the selectivity of methane or other products that requires further electroreduction of CO, a stronger binding of CO is essential [90].

Recent advances on a variety of M-N_x sites-containing transition metal-centered macrocyclic complexes, metal organic frameworks and M-N_x-doped carbon materials towards efficient CO₂RR are summarized, including both experimental and theoretical studies. Unequivocally, these atomically dispersed metal nitrogen sites constituted of earth abundant elements with maximum atom-utilization efficiency exhibit great potential as promising alternatives for well-performed metal catalysts (Au, Ag, Cu, Pd, etc.). Despite of the encouraging progress, the catalytic performance still needs to be optimized to meet the industrial requirements, and studies on unveiling the intricate reaction mechanisms are highly required for the exploration of optimal catalysts and reaction conditions. To be specific, high current density and stability are essential requirements for future application of CO₂RR, thus active sites should hold considerable turnover frequency and competitive HER should be inhibited at higher potentials in order to achieve a high performance and stability towards CO₂RR.

Though combined with conductive substrates contributes to the improved activity and selectivity towards CO₂RR on M-N_x-containing macrocyclic complexes and MOF materials, poor conductivity which leads to limited electro-active sites still remains the main challenge. In contrast, carbon-based metal-nitrogen materials derived from pyrolysis of N-, C- and transition metal-containing precursors endow excellent conductivity with improved charge/electron transfer capability. However, unlike macrocyclic complexes or MOF materials with precise active site structures, there are several functionalities (i.e., multiple N configurations and various metal nitrogen coordination states) in carbon-based metal-nitrogen catalysts, making it difficult to extract the structure-activity relationship. Therefore, careful control of synthetic strategies for structural uniformity of active sites, advanced in situ/operando characterization methods and profound theoretical simulations are the key for understanding the role of each functionality. As illustrated above, CO and formate are major products on carbon-based M-N_x sites with few studies showing hydrocarbon or oxygenates as reduction products. Novel catalytic systems concerning M-N_x active sites that can effectively reduce CO₂ into more valuable products involving multiple proton/electron transfers still remain challenging. In addition to the intrinsic nature of active sites, a suitable adjustment of reaction conditions (pH, buffer capacity, pressure, and set-up configurations) contributes to a higher selectivity to a certain product, and a deep understanding of the CO₂RR mechanism provides inductive guidance for optimized catalytic processes in return. In brief, highly efficient, robust and cost-effective M-N_x sites manifest great potential in promoting the industrial application of CO₂RR.