Recent advances in carbonized non-noble metal–organic frameworks for electrochemical catalyst of oxygen reduction reaction

Abstract

The non-noble metal oxygen reduction reaction (ORR) catalysts prepared by carbonization of metal–organic framework (MOF) have attracted more and more attentions in the fields of fuel cells and metal-air batteries due to their unique intrinsic advantages such as high catalytic activity, low price, simple synthesis and good adaptability. Different from the study of traditional high active noble metal catalysts, this review systematically summarizes recent developments on non-noble metal (Fe, Co, Cu, Ni, Mn and Mo) ORR catalysts prepared by various MOFs carbonization in different metal centers. The effects of synthesis strategies and pyrolysis conditions on the catalyst properties are discussed. Meanwhile, the key parameters of catalytic performances (including active sites, dispersed state and specific surface area) are discussed and the prospect is presented. It is expected that this review will provide effective guidance for future studies on carbonized non-noble MOFs for ORR electrochemical catalyst.

Graphic abstract

1 Introduction

Metal–organic framework (MOF) was first reported in 1995 [1, 2], which is made of various organic ligands and metal ions (Zn²⁺, Cr³⁺, Cu²⁺, Co²⁺, etc.), with a series of advantages such as adjustable porous structure and porosity, low crystal density, ultrahigh specific surface area and highly ordered pore structure. With these functions, MOF can be used for gas storage, nano-separation, biomedicine, proton conduction, nano-reactor, energy conversion and so on [3,4,5,6,7,8]. Many new MOFs with unique characteristics, such as the Materials of Institute Lavoisier (MIL) with large specific surface area (5900 m²·g⁻¹ for MIL-101) and the topologically structured zeolite-like MOF (ZIF), are designed by different metal ions and organic ligands with different sizes [9, 10]. With the continuous development of MOFs, they have become research hotspots in various fields. However, in the research process of MOFs, there are still some inevitable shortcomings, such as low stability in acid and water, and low conductivity, especially the latter greatly restricts their application in the field of electrochemistry [11, 12]. In order to overcome these shortcomings, researchers have developed stable and conductive MOF catalysts through structural design or composite materials with conductive substrates, but their effects are far inferior to other metal-based catalysts [13, 14]. However, due to the large specific surface area and heteroatom doping, MOFs are found to be excellent precursors of carbon-based catalysts [15].

Oxygen reduction reaction (ORR) is one of the most important cathode reactions in fuel cells and metal-air batteries [16, 17]. Usually, these are divided into 2-electron and 4-electron reaction processes (Fig. 1 [18]). Generally, the advantages of 4-electron process include the avoidance of damage to the catalyst by H₂O₂ and a higher theoretical cathode voltage. The ideal ORR requires a small overpotential, which helps to improve the battery’s output voltage and energy efficiency. Therefore, the design of the 4-electron process ORR catalyst with excellent performance has received extensive attention. Currently, the most used and commercially available ORR catalyst is the Pt/C catalyst. A small amount of Pt/C catalyst can achieve satisfactory ORR catalytic performance. However, the Pt/C catalyst is limited by high price, low catalytic stability and cycle efficiency, which indicates that it is urgent to develop low-cost, exciting activity, high stability and green catalysts to replace Pt/C. Compared with noble metal catalysts, non-noble metal catalysts, such as transition metal–nitrogen–carbon (M–N–C, M = Fe, Co, Ni, Cu, etc.) catalysts or transition metal oxide catalysts, have natural electrocatalysis and price advantages [19,20,21]. Although there are many ways to prepare non-noble metal catalysts, the preparation by MOF carbonization has the following advantages: (1) The catalyst performance can be optimized by designing the precursor structure of MOF and controlling pyrolysis conditions; (2) The metal center of MOF and other doped heteroatoms can provide the catalytic active centers; (3) During the pyrolysis of the MOF precursor, the graphene carbon structure or carbon nanotube (CNT) structure can be formed, which helps to improve electrical conductivity. Therefore, the preparation of non-noble metal catalysts by MOF carbonization is an attractive synthetic route [22].

Fig. 1

Schematic diagram of ORR process by MOF-based catalyst. Reproduced with permission from Ref. [18]. Copyright 2019 Elsevier Ltd.

Currently, there are several important parameters to evaluate the ORR performance of catalysts: (1) The onset potential (E_onset) is the voltage at which the ORR occurs and is usually obtained at a current density of 1 mA·cm⁻². The higher the onset potential, the smaller the overpotential and the higher energy conversion efficiency. (2) The half-wave potential (E_1/2) is the electrode potential when the current density of the reaction is equal to half of the limiting diffusion current density. It can qualitatively analyze the catalyst because the value depends on the inherent nature of the catalyst. (3) The Tafel slope reflects the kinetic rate of the ORR process. The smaller the Tafel slope, the faster the ORR kinetics of the catalyst. (4) As mentioned above, it is extremely important for ORR to reveal whether the reaction adopts a 2-electron or 4-electron process. There are two specific calculation methods: one is to calculate the electron transfer number through the Koutecky–Levich equation (Eq. (1)); another method is to monitor the H₂O₂ yield (Eq. (2)) in the catalytic process by rotating ring-disk electrode (RRDE) and calculate the electron transfer number (Eq. (3)). (5) The cyclic stability is mainly used to describe the catalyst durability. In addition, the synthesis method of the catalyst ink, the amount and thickness of electrode coatings, the electrolyte composition and the rate of ORR linear sweep voltammetry (LSV) will also affect the performance of the catalyst [23,24,25].

$$\frac{1}{J} = \frac{1}{{J_{\text{L}} }} + \frac{1}{{J_{\text{K}} }} = \frac{1}{{J_{\text{K}} }} + \frac{1}{{0.62nFC_{0} (D_{0} )^{2/3} \upsilon^{ - 1/6} \omega^{1/2} }}$$

(1)

where J, J_L and J_K are the measured current density, diffusion limiting current density and kinetic current density, respectively. n is the number of electrons transferred, F is the Faraday constant, C₀ is the concentration of O₂ in solution, D₀ is the diffusion coefficient of O₂, υ is the kinetic viscosity of the solution, and ω is the rotation speed of the electrode.

$$w = 200\frac{{I_{\text{R}} /N}}{{I_{\text{D}} + I_{\text{R}} /N}}$$

(2)

$$n = 4\frac{{I_{\text{D}} /N}}{{I_{\text{D}} + I_{\text{R}} /N}}$$

(3)

where w refers to the yield of H₂O₂, I_D and I_R are disk current and ring current, respectively, and N refers to the collection efficiency of the ring.

Many reviews and comments of MOF-based catalysts have been published in recent years [15, 26], but most of them are too broad and do not focus on the synthesis and performance of non-noble metal-based catalysts prepared by MOF carbonization. In this paper, the preparation of ORR catalyst by MOF carbonization is reviewed. The relationship between the preparation method and the catalyst performance is also discussed. According to the classification of different metal centers, the main content is divided into two parts: single-component metal carbon catalysts represented by Fe, Co, etc., and some multi-metal carbon catalysts. This review may benefit the future development of MOF-based non-noble metal ORR catalysts with high performance.

2 MOF-based Fe–N–C catalyst

2.1 Introduction of MOF-based Fe–N–C catalyst

The Fe–N–C catalyst is an amazing ORR cathode catalyst with the advantages of low cost, high activity and good stability. Many studies have shown that the activity of Fe–N–C catalyst mainly comes from the Fe–N_x sites, but for different Fe–N–C catalysts, the x value is different. Some studies have revealed that Fe–N₄ structure is the most active component at the Fe–N_x site, while Fe–N₂ contributes less [27, 28]. Lai et al. [29] prepared the Fe–N–C catalysts by carbonizing MOF-derived precursor to study the activity of Fe–N_x sites (including FeN₅, FeN₄ and FeN₂). According to the density functional theory (DFT) calculation, the energy barrier of N–Fe–N₄ (0.67 eV) in ORR was lower than that of Fe–N₄ (0.75 eV) and Fe–N₂ (1.99 eV). The adsorption energy of N–Fe–N₄ for OH was 2.88 eV, which was close to Pt (111) (2.23 eV), but lower than that of Fe–N₄ (3.07 eV) and Fe–N₂ (4.38 eV). Therefore, N–Fe–N₄ is more likely to convert oxygen into water. In fact, it is difficult to directly prepare N–Fe–N₄ catalyst with high activity, most of which are mixtures of different Fe–N_x structures. In addition, in the pyrolysis of Fe-based MOF, excess iron will combine with adjacent carbon to form FeC₃ particles, which also contributes to the ORR process [30].

The preparation of MOF-based Fe–N–C catalyst can be divided into three strategies: (1) Direct pyrolysis of Fe-MOFs with the greatest advantage of being very simple; (2) Pyrolysis of MOFs with additional iron sources, which has the advantage of various MOF carriers and cheap iron sources; (3) Pyrolysis of heteroatom-doped MOFs, which can enhance the catalytic activity through heteroatoms such as N and S. For all preparation methods, the synthesis and selection of MOF precursors have a significant effect on the catalyst’s performance.

2.2 Preparation of Fe–N–C catalyst by direct carbonization of Fe-MOFs

One-step pyrolysis of Fe-MOFs to prepare Fe–N–C catalyst is the simplest and most effective method, but the key is the synthesis of Fe-MOF precursors. Fe-MOF is mainly synthesized through the coordination of Fe ions and organic ligands. Therefore, the coordination number of Fe ions and the molecular structure of organic ligands play a key role in the micro-morphology and stability of Fe-MOF. For example, Li et al. [31] employed 4,5-dicyanoimidazole (DCI) and iron acetate as raw materials to synthesize an organic DCI-Fe-MOF sheet by one-pot method, which was pyrolyzed in a N₂ atmosphere at 700 °C for 2 h to give a carbonized catalyst Fe-DCI-700 (Table 1 [18, 31,32,33,34,35,36,37,38,39,40,41,42,43]). The design of MOF at the molecular level ensured the uniform distribution of active elements N, C and Fe in the MOF precursor, thus ensuring abundant and well-distributed high active sites in Fe-DCI-700. The carbon shell encased the iron carbide on the N-doped carbon shell structure was beneficial to ORR. The catalyst contained 61.5 wt% Fe, 7.6 at% N, Brunauer–Emmett–Teller (BET) specific surface area of 249 m²·g⁻¹ and the Raman peak intensity ratio of D band and G band (I_D/I_G) was 1.2. The higher I_D/I_G value, the more carbon defects; while the lower the I_D/I_G value, the higher the degree of graphite [44]. Under the catalyst loads of 0.1 mg·cm⁻² and in 0.1 mol·L⁻¹ KOH (aq.), Fe-DCI-700 exhibited a half-wave potential of − 0.152 V vs. Ag/AgCl, which was 17 mV higher than that of commercial Pt/C catalyst. The measured electron transfer number of Fe-DCI-700 was 3.67, manifesting that the ORR process mainly followed the 4-electron mechanism (Table 2 [18, 31,32,33,34,35,36,37,38,39,40,41,42,43]). The result was mainly due to the graphitized shells, the high content of pyridine and pyrrole N (7.67%), and the lamellar porous structure with high-surface area formed by the MOF precursor, which greatly facilitated the transport of the electrolyte and provided rich and accessible active sites.

Table 1 Summary of synthesis of MOF-based Fe–N–C catalyst

Table 2 Summary of ORR electrocatalytic activities of MOF-based Fe–N–C catalyst

Besides, the Fe-MOF precursor can be prepared by post-modification of hydrogen-bonded supramolecules. For instance, Zhou et al. [32] first synthesized BA-TAP by hydrogen bond between barbituric acid (BA) and 2,4,6-triaminopyrimidine (TAP), whose N atoms were further coordinated with FeCl₃ to obtain BA-TAP-Fe. After pyrolysis in N₂ atmosphere at 800 °C for 1 h, the final BA-TAP-Fe-800 catalyst was synthesized. BA-TAP-Fe-800 formed a dense crystal structure through hydrogen bonds, which avoided collapse in the pyrolysis process. The multiphase iron was anchored to the multistage N-doped graphitic carbon. The catalyst had a porous structure (BET: 232 m²·g⁻¹) and a high graphitization degree (I_D/I_G: 0.85). It also showed an onset potential of 1.02 V (vs. RHE), which was higher than that of Pt/C catalyst (0.95 V (vs. RHE)). Moreover, it possessed good ORR performance with half-wave potential of 0.85 V (vs. RHE), Tafel slope of − 66 mV·dec⁻¹, and electron transfer number of 3.9. The reaction of organic iron complexes with evaporable metal salts (such as Zn (907 °C)) is another method to prepare Fe-MOF. During pyrolysis, vaporized metals will create more pores in the catalyst. For example, Song et al. [33] used K₃Fe(CN)₆ and ZnCl₂·6H₂O to synthesize Zn₃[Fe(CN)₆]₂·xH₂O (ZnPBA) as precursor. During the pyrolysis of ZnPBA in N₂ atmosphere at 800 °C, Zn would evaporate to form pores and prevented Fe from agglomerating. With the increase of annealing time, the degree of graphitization also increased. At the same time, during the ZnPBA conversion, the cyano group (CN) was decomposed, releasing a large amount of carbon nitride gases, which could be captured by iron particles to catalyze the growth of CNTs. When the heating time was 1 h, the product (Fe/Fe₃C-C@N-C-1) had the maximum surface area of 182.5 m²·g⁻¹, and contained 71.17 wt% iron and 1.61 wt% nitrogen, which was beneficial to the formation of Fe–N_x active sites and the improvement of ORR activity. The catalyst exhibited an onset potential of 0.936 V (vs. RHE), a half-wave potential of 0.804 V (vs. RHE), which was close to that of Pt/C (E_onset = 0.975 V (vs. RHE), E_1/2 = 0.820 V (vs. RHE)). After 1000 cycles, it was found that the half-wave potential of Fe/Fe₃C-C@N-C-1 catalyst had only a negative shift of 15 mV, which was smaller than that of the Pt/C catalyst (20 mV), indicating that its ORR catalytic effect was highly efficient and stable.

2.3 Preparation of Fe–N–C catalyst by carbonization of Fe-loaded MOFs

Although there are many types of MOFs, only MOFs with stable structure and N coordination site are suitable for direct Fe loading, and Fe–N–C catalysts with Fe–N_x active site can be prepared. As a subclass of MOF, ZIF is a zeolite-like porous crystalline material, usually consisting of Zn or Co and imidazole ligands. As for ZIF-8, it is a framework structure constructed by Zn(NO₃)₂·6H₂O and 2-methylimidazole, which is regarded as an excellent carrier with rich N content and large specific surface area [45]. Shah et al. [34] mixed Zn(NO₃)₂·6H₂O, Fe(acac)₂ and 4,5-dicyanoimidazole in one pot to prepare Fe/Zn-dicyanoimidazolate. A series of Fe/Zn-dicyanoimidazolate frameworks with different contents were obtained through different molar ratios of Fe²⁺/Zn²⁺. After pyrolysis, highly active Fe–N_x and Fe₃C ORR sites were formed. By removing Fe₃C and Fe sites using acid etching, the performance of the catalyst was improved. The Fe–N–C catalyst showed excellent ORR performance in acidic and alkaline solutions. In 0.1 mol·L⁻¹ KOH (aq.), it exhibited an onset potential of 0.997 V (vs. RHE), a half-wave potential of 0.91 V (vs. RHE) and 4-electron mechanism. Similarly, in 0.1 mol·L⁻¹ HClO₄ (aq.), it had an onset potential of 0.927 V (vs. RHE), and a half-wave potential of 0.817 V (vs. RHE). Furthermore, the Fe–N–C catalyst displayed good stability in acidic and alkaline electrolytes, proving the excellent ORR catalyst. These results indicated that the coordination between iron ion and cyano group was the key to the synthesis of Fe–N₄ site, and Fe–N_x was more active for ORR than Fe₃C and Fe nanoparticles.

In addition, there are two forms of iron-loaded ZIF-8, namely in situ loading during the synthesis of ZIF-8 and loading by ZIF-8 to adsorb iron ions. Gao et al. [18] mixed Zn²⁺, Fe(acac)₃ and 2-methylimidazole to prepare the Fe-loading Fe@ZIF-8. By controlling the ratio of H₂/Ar concentration in the pyrolysis atmosphere, it was found that when the H₂ content increased from 0% to 10%, the surface roughness of the pyrolysis product Fe–N–C was gradually enhanced, and the pore size enlarged from 1.23 to 2.36 nm. The specific surface area also increased from 961 to 1388 m⁻²·g⁻¹. In order to improve ORR performance, it is very important to regulate the pore structure and active site density of Fe–N–C catalyst. As for the correlation mechanism between the density of the active site and H₂ concentration, hydrogen etching mainly etched unstable C, thus increasing the edge N and adjusting the pore structure. The edge N was then anchored to free iron atoms to form the Fe–N_x active sites. Meanwhile, the hydrogen etched the blocked carbon fragments to enlarge the pore size, leading to the formation of micro/mesopore structures. When H₂ concentration continued to increase, a large number of metal Fe particles that were inactive to ORR were formed, and the proportion of micropores was greatly reduced, resulting in a decrease in specific surface area. Therefore, when the H₂/Ar ratio was 1/9, Fe–N–C-3 catalyst had the best specific surface area (1388 m²·g⁻¹), high graphitization degree (I_D/I_G: 0.886), and high ORR activity. It also had an onset potential of 0.91 V (vs. RHE), a half-wave potential of 0.805 V (vs. RHE), and ultralow H₂O₂ yield (1%). Moreover, one-pot preparation of iron-loaded MOF can achieve the purpose of in situ doping, and iron ions can be well dispersed [46]. For example, Deng et al. [35] took advantage of the easy sublimation of ferrocene (100 °C) to combine iron with ZIF-8 through gas adsorption (Fig. 2 [35]). ZIF-8 and ferrocene were heated in a tube furnace for 2 h to 150 °C, so that the metal–organic gases were fully captured by ZIF-8, and the precursor was then pyrolyzed at 950 °C. As a gaseous doping precursor, ferrocene significantly reduced the content of oxidation state N, induced the formation of Fe–N_x, and increased the proportion of graphite state N. The C-Fe-ZIF-4.44-950 catalyst with a specific surface area of 1255 m²·g⁻¹ maintained the dodecahedral structure of ZIF-8, and iron was highly dispersed in the catalyst. The catalyst had excellent ORR performance in both acidic and alkaline electrolytes, with electron transfer numbers of 3.97 and 3.98, respectively.

Fig. 2

a Schematic illustration of preparation of sample; LSV curves of sample in O₂-saturated b 0.1 mol·L⁻¹ KOH and c 0.1 mol·L⁻¹ HClO₄. Reproduced with permission from Ref. [35]. Copyright 2019 Wiley–VCH

2.4 Preparation of heteroatoms-enhanced Fe–N–C catalysts by MOF carbonization

Heteroatom doping is considered to be an effective method to improve electrochemical performance [47]. For instance, nitrogen atoms can not only improve hydrophilicity, but also change the distribution of charge density in carbon materials, thereby enhancing the catalytic activity. Sulfur can change the spin density, charge density of adjacent carbon atoms, and increase the active area of the material. Owing to the redistribution of charge density and spin density, the co-doping of N and S will produce a synergistic effect. In addition, different N and S sources have different effects on the catalyst. Therefore, according to the type of MOF and catalytic requirements, reasonable heteroatom doping is necessary and can achieve positive effects.

For ZIF, in addition to the necessary Fe source, additional heteroatom doping can effectively enhance the performance of the catalyst. Small molecules or salts containing heteroatoms are common sources of other heteroatoms, such as melamine, NH₃, N₂H₄, (NH₄)₂S₂O₈ and NH₄SCN. For example, Chen et al. [36] used melamine as an additive to Fe-ZIF-8 which was prepared by 2-methylimidazole, Fe(acac)₃ and Zn(NO₃)₂·6H₂O, and then pyrolyzed in an Ar atmosphere at 900 °C. The addition of melamine not only increased the N content of Fe@NMC-1 catalyst to 11.82 at%, but also formed more porous network structure. The specific surface area increased from 705.1 to 844.9 m²·g⁻¹, which was mainly due to the release of gases (such as NH₃) by melamine in the pyrolysis process, which promoted the formation of the inner cavity. Abundant N atoms combined with Fe atoms to form Fe–N_x active sites, which were highly dispersed in a three-dimensional (3D) interconnection network. As a result, the Fe@NMC-1 catalyst exhibited an onset potential of 1.01 V vs. RHE, a half-wave potential of 0.88 V vs. RHE, and a Tafel slope of 54 mV·dec⁻¹ in 0.1 mol·L⁻¹ KOH (aq.), which was superior to Pt/C catalyst (E_onset = 0.97 V (vs. RHE), E_1/2 = 0.85 V (vs. RHE), Tafel slope = 75 mV·dec⁻¹). Moreover, it showed excellent performance under acidic conditions. Jina et al. [37] mixed ZIF-8, FeSO₄, and N₂H₄·H₂O for pyrolysis to prepare a 6% Fe–N–S CNN catalyst with a carbon matrix/nanotube 3D nanostructure (Fig. 3 [37]). FeSO₄ provided sources of iron and sulfur. Hydrazine hydrate used as an alkaline reducing agent was effective in preventing the iron oxidation and structural destruction of ZIF-8. The addition of N₂H₄·H₂O not only increased the N content on the catalyst surface, but also promoted the formation of CNTs, which could improve the conductivity of the catalyst, and provide rich channels of the reactants transport. High specific surface area (974 m²·g⁻¹) and rich Fe and N contents (1.6 at% and 9.65 at%, respectively) were beneficial to forming more Fe–N_x active sites with effective dispersibility. S dopant could also synergistically promote the catalyst’s activity. In addition, the in situ formed CNTs increased the conductivity of the catalyst and enriched the electron transport channels. In 0.1 mol·L⁻¹ KOH (aq.), the half-wave potential of 6% Fe–N–S CNN catalyst was 0.91 V vs. RHE, higher than that of Pt/C catalyst (0.85 V (vs. RHE)). After 12 h operation, the current density only decreased by 5.21%, indicating that it had good stability.

Fig. 3

a Schematic illustration of synthetic process for Fe–N–S CNN; b TEM image of CN; c TEM image; d HRTEM image of 6%Fe–N–S CNN; e C–N content; f pyridinic-N and graphitic-N; g C–S–C content in pyrolysis samples of different catalysts from XPS results. Reproduced with permission from Ref. [37]. Copyright 2019 Elsevier B.V.

Owing to its large specific surface area and high stability, MIL is also an alternative precursor [48], which is a framework composed of trivalent transition metal ions (such as Cr³⁺, Fe³⁺, Al³⁺) and carboxylic acid ligands (such as glutaric acid, succinic acid and other dibasic acids). Since these ligands having no N atoms, an additional N source is required in order to prepare the Fe–N–C catalysts from MIL, while the addition of iron source depends on the metal center of MIL. Guo et al. [38] replaced Cr³⁺ with Fe³⁺ and coordinated with 1,3,5-benzenetricarboxylic acid to prepare MIL-101-Fe [49] (832 m²·g⁻¹), and then encapsulated hexamethylenetetramine (HMT) into the pores of MIL-101-Fe by rotary evaporation. The bottle was under negative pressure when evaporated, which allowed HMT to easily enter the pores of MIL-101-Fe for N loading. When the product was pyrolyzed at 700, 800, 900 and 1000 °C, simple Fe–N₂-700 (128 m²·g⁻¹), Fe–N₂-800 (138 m²·g⁻¹), Fe–N₂-900 (151 m²·g⁻¹) and Fe–N₂-1000 (167 m²·g⁻¹) were obtained, respectively. However, Fe–N₂-800 showed the outstanding ORR activity with an onset potential of − 0.077 V (vs. SCE) and a half-wave potential of − 0.25 V (vs. SCE), which were close to that of Pt/C catalyst (E_onset = −0.033 V (vs. SCE), E_1/2 = −0.175 V (vs. SCE)). This was because the content of pyridinic-N and graphitic-N was the highest in Fe–N₂-800, and the coexistence of Fe/Fe₃C nanoparticles and Fe–N_x active sites played an important role in promoting ORR activity. Therefore, the proper pyrolysis temperature had a key effect on the performance. Besides, Gao et al. [39] took MIL-101-Fe [50] as the precursor, coated its surface with polyaniline through in situ polymerization, and pyrolyzed it at 900 °C to prepare NC@Fe₃O₄-900-1.5 catalyst with specific surface area of 698.52 m²·g⁻¹. Polyaniline was not only the source of N, but also could prevent particle aggregation and structure collapse during pyrolysis. Furthermore, it also played an important role in the formation of ORR active sites, such as Fe–N_x and pyridinic-N. The catalyst contained a large number of oxygen-containing groups like –COOH and –OH, and had natural hydrophilicity of C–N bonds (contact angle of 10.6°), which was conducive to the complete entry of aqueous electrolytes into its pores. It exhibited excitation stability with an onset potential of 0.058 V (vs. Hg/HgO), which was close to Pt/C catalyst (0.071 V (vs. Hg/HgO)).

In addition to the doping of N atoms, S doping will also lead to the improved catalyst performance. For instance, Yang et al. [40] mixed MIL-53 [51] (iron ion center), (NH₄)₂S₂O₈ and melamine by ultrasonication, then heated the mixture at 240 °C for 2 h and 900 °C for 1 h, and finally obtained the N, S co-doped Fe_0.2N_0.2M_0.2-900 catalyst. The Fe, N and S contents of Fe_0.2N_0.2M_0.2-900 were 0.62 at%, 3.25 at% and 0.82 at%, respectively. (NH₄)₂S₂O₈ is a weak acidic salt. Adding melamine can adjust the pH, which is beneficial to control the morphology of the catalyst. Therefore, F_0.2N_0.2M_0.2-900 not only maintained the spindle morphology of the MOF precursor, but also formed the graphene structure, and a large amount of CNTs, which was helpful for electron transfer. The F_0.2N_0.2M_0.2-900 catalyst had excellent stability and the measured onset potential (0.970 V (vs. RHE)) and half-wave potential (0.873 V (vs. RHE)) which were higher than those of Pt/C catalyst (0.942 V (vs. RHE) and 0.807 V (vs. RHE), respectively). Besides, Yang et al. [41] prepared the Fe, N, S tri-doped core–shell carbon catalyst (HCSC catalyst) with MIL-88 as the sacrifice template and iron source (Fig. 4 [41]). First, MIL-88, aniline, and HClO₄ were stirred in an ice bath for 30 min, then aniline adhered to the surface of MIL-88, while H⁺ etched the tip of the MIL-88 polyhedron into an elliptical sphere. An initiator (NH₄)₂S₂O₈ was added to induce aniline polymerization to form a porous polyaniline shell on the outer surface of MIL-88. Meanwhile, MIL-88 further dissolved in equilibrium to form a nucleus after sufficient reaction time. In this process, MIL-88 served as a sacrificial template to promote Fe, N, S dispersion in the carbon skeleton. After further pyrolysis at 240 °C for 2 h in N₂ atmosphere, and then at 800 °C for 1 h, a porous core–shell HCSC-IV catalyst with a specific surface area of 90.5 m²·g⁻¹ and an I_D/I_G ratio of 0.93 was obtained. The special nucleus of HCSC was similar to the vein of a leaf and was important to the electrocatalytic process by facilitating charge transfer to the entire shell nanostructure. Hence, the catalyst exhibited an onset potential of 0.918 V (vs. RHE), a half-wave potential of 0.794 V (vs. RHE), a diffusion limiting current density of 7.35 mA·cm⁻², and good stability.

Fig. 4

a Schematic illustration of u-HCSC-(I, II, III, IV and V) nanostructures prepared by “polymerization-dissolution” strategy; b–e TEM images of u-HCSC-(I, II, III, IV and V) nanostructures; f–g HAADF-STEM images of HCSC-IV; h polarization curves of materials of different samples in O₂-saturated 0.1 mol·L⁻¹ KOH. Reproduced with permission from Ref. [41]. Copyright 2019 Elsevier Ltd

Other types of MOF can also be used as precursors of Fe–N–C catalysts by heteroatom doping. For instance, Huang et al. [42] prepared porous carbon-containing copper nanoparticles (Cu@PC) by using Hong Kong University of Science and Technology-1 MOF (HKUST-1) as precursor and pyrolyzed at 800 °C for 2 h. Cu@PC was dispersed in saturated FeCl₃ aqueous solution, and copper particles were etched to obtain porous carbon (PC). PC was further mixed with FePC, and then annealed in ammonia gas to obtain the Fe₂N@NPC-500 catalyst (381 m²·g⁻¹). The uniform distribution of Fe₂N nanoparticles and their close contact with the shell promoted the formation of Fe–N/C bond and the increase of active sites. Owing to the uniform distribution of NPs, large specific surface area, large number of mesopores and active Fe–N/C binding, Fe₂N@NPC-500 catalyst showed an onset potential of − 0.038 V (vs. Ag/AgCl), and a half-wave potential of − 0.175 V (vs. Ag/AgCl), which was close to commercial Pt/C catalyst. In addition, the stability of the catalyst was higher than that of Pt/C catalyst. Zheng et al. [43] encapsulated the hydrophilic NH₄SCN and FeCl₃ into the University of Oslo-66-NH₂ MOF (UIO-66-NH₂) by a double-solvent method, making Fe and S highly dispersed in the derived carbon. Compared with the impregnation method, the double-solvent method realized the atomic-level dispersion of active sites on the mesoporous structure and maintained the form of derived carbon, which was beneficial to the catalytic performance of the MOF-derived carbon catalyst. The resulting carbon material perfectly inherited the morphology of UIO-66-NH₂ precursor, forming large surface area, clear mesoporous structure and atomic dispersion of doped elements. After heating, the I_D/I_G ratio of the Fe/N/S-PC catalyst was 1.19, indicating that there were more carbon defects in the catalyst and more active sites were exposed. Moreover, the addition of S can reduce the content of oxidized-N (10.71%), increase the content of pyridinic-N (19.49%), metal-N (9.11%) and graphitic-N (45.64%). Pyridinic-N can activate the ORR process and reduce ORR overpotential, while the presence of graphitic-N facilitates electron conduction and increases the catalytic current of ORR [52, 53]. Fe/N/S-PC contained 0.31 at% iron, 4.93 at% nitrogen, 1.49 at% sulfur, and high specific surface area (1589 m²·g⁻¹) as well high porosity (77.22%), resulting in excellent ORR performance in acidic or alkaline electrolyte. The large specific surface area was mainly credited to the addition of Fe, which played an important role in the formation of mesoporous structure and porosity. In 0.1 mol·L⁻¹ KOH (aq.), Fe/N/S-PC catalyst exhibited an onset potential of 0.97 V (vs. RHE) and a half-wave potential of 0.87 V (vs. RHE). It was superior to Pt/C (E_onset = 0.94 V (vs. RHE), E_1/2 = 0.84 V (vs. RHE)). In 0.1 mol·L⁻¹ HClO₄ (aq.), it showed an onset potential and a half-wave potential of 0.89 V (vs. RHE) and 0.785 V (vs. RHE), respectively, and followed the 4-electron mechanism. In conclusion, as long as the methods and strategies are reasonable and appropriate, various heteroatom sources and doping methods can use different types of MOF as precursors for catalysts preparation.

2.5 Summary of MOF-based Fe–N–C catalyst

The electrochemical properties of catalysts are determined by a variety of factors, such as the structure and dispersion of active sites, specific surface area, porosity, conductivity and hydrophilicity. Therefore, the influence of precursor structure and heat treatment method on Fe–N–C catalyst is very crucial. The direct carbonization of Fe-MOF to prepare Fe–N–C catalyst has the following advantages: (1) The target structure of MOF can be designed from cheap iron sources and organic ligands; (2) Fe-MOF precursor has high iron content and atom dispersibility, which is beneficial to obtaining highly dispersed Fe–N_x active sites; (3) One-pot preparation and one-step pyrolysis are relatively simple. However, in order to avoid structural collapse and particle aggregation during pyrolysis, the Fe-MOF structure needs to be finely designed.

By selecting suitable MOF carrier to load the iron source, Fe–N–C catalyst with high performance can also be prepared. The specific surface area, pore size and coordination number of MOF are the focus of adjustment. The porous network structure and proper aperture of MOF enable Fe ions to penetrate easily and bind to abundant coordination sites (such as N) for in situ loading. The iron source can be selected according to the characteristics of the carrier, such as molecular size and chemical properties. The chemical adsorption method is mainly used to achieve a stable load. By adjusting the ratio of MOF to iron source, pyrolysis temperature, atmosphere and time, the optimal loading effect and ORR catalytic activity can be achieved.

The Fe–N–C catalysts prepared by doping different heteroatoms can effectively improve the morphology after pyrolysis, thus promoting ORR catalysis, especially for N-free MOF precursors. At present, the most common heteroatom doping is mainly electronegative atoms (such as N) or larger-size atoms (such as S). The advantages of heteroatom doping are: (1) the heteroatom raw materials, such as melamine and polyaniline, are very cheap; (2) the heteroatom source is easy to compound, with obvious doping effect, which can effectively improve the ORR effect; (3) under controllable reaction conditions, the optimal element ratio can be adjusted by the ratio of MOF precursor to additive.

3 MOF-based Co–N–C catalyst and cobalt oxide catalyst

3.1 Introduction of MOF-based Co–N–C catalyst and cobalt oxide catalyst

Owing to excellent electrochemical performance, Co–N–C catalysts and Fe–N–C catalysts have been widely studied to replace noble metal catalyst of ORR [54]. Similar to Fe–N–C catalyst, the active center of Co–N–C catalyst is mainly Co–N_x [55]. Yan et al. [56] found that the atomic Co with tetrahedral coordination played an extremely important role in the ORR process. In addition, the effect of metallic Co NPs on the activity of Co–N–C catalyst was also studied. Zhao et al. [57] found that Co NPs coating on the catalyst surface and carbon layer had little effect on the activity of catalyst and even affected the high yield of H₂O₂. Therefore, it is of great significance to pyrolyze the cobalt ions in the precursor into active sites that are favorable for catalysis rather than cobalt metal particles.

Cobalt oxide, including Co₃O₄ and CoO with a rock-salt structure, has been reported to have ORR catalytic activity [58, 59]. Owing to the good dispersibility of cobalt oxide particles and favorable conductivity of catalysts, the activity can be effectively improved by combining cobalt oxide with the carbon matrix. CoO and Co₃O₄ were used as active centers to prepare high-performance catalysts. For example, Liu’s group [60] studied the effect of the geometric occupancy and oxidation state of cations in cobalt oxide on ORR activity. They found that the Co²⁺ $\left( {{\text{Co}}_{\text{Oh}}^{2 + } } \right)$ and Co³⁺ in the octahedral site $\left( {{\text{Co}}_{\text{Oh}}^{3 + } } \right)$ were the main active sites of CoO and Co₃O₄, respectively; while the activity of $\left( {{\text{Co}}_{\text{Oh}}^{2 + } } \right)$ was higher than that of $\left( {{\text{Co}}_{\text{Oh}}^{3 + } } \right)$. In the ORR, the $\left( {{\text{Co}}_{\text{Oh}}^{2 + } } \right)$ sites in CoO tended to be converted into $\left( {{\text{Co}}_{\text{Oh}}^{3 + } } \right)$ and Co₃O₄ was formed, indicating that the reaction was unstable. Therefore, the corresponding cobalt oxide catalyst can be prepared by MOF pyrolysis for different active sites.

3.2 Preparation of Co–N–C catalyst and cobalt oxide catalyst by carbonization of Co-MOFs

Co-MOFs can be prepared by different cobalt compounds and organic ligands. For example, by solvothermal method, Co(NO₃)₂·6H₂O and 2-methylimidazole can prepare ZIF-67 with rhombic dodecahedron structure with large specific surface area [61]. Liu et al. [62] took p-phenylenediamine (PPD) as the organic ligand and Co(NO₃)₂ to prepare the rosette spherical Co-PPD complex composed of cross-linked slices through simple hydrothermal reaction (Table 3 [62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79]). After pyrolysis in an Ar atmosphere at 800 °C for 2 h, some Co ions were reduced to Co(0) and diffused into the graphene shell, thus producing a Co@NPC-acid catalyst with 1.72 wt% Co (Table 4 [62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79]). The inert substances on the catalyst surface, such as cobalt oxide and cobalt, can be removed by soaking in an acidic solution for activation. Meanwhile, compared with non-acid-etched Co@NPC (specific surface area: 242.6 m²·g⁻¹; pore volume: 0.34 cm³·g⁻¹), Co@NPC-acid had a larger contact area (859.5 m²·g⁻¹, 1.82 cm³·g⁻¹), which provided more opportunities for the contact between active sites and reactive substances. In 0.1 mol·L⁻¹ KOH (aq.), the half-wave potential of Co@NPC-acid catalyst with Co–N_x active sites was 0.12 V vs. Hg/HgO, and the diffusion limiting current density was 5.3 mA·cm⁻², which was higher than that of Pt/C catalyst. After 10 h of continuous testing, the Co@NPC-acid catalyst maintained 91% of the initial current, showing excellent stability. Ma et al. [63] prepared rod-shaped Co-MOF using sodium dicyandiamide, pyrazine and Co(NO₃)₂ (Fig. 5 [63]). Dicyandiamide was used as carbon and nitrogen source for Co-MOF pyrolysis in a N₂ atmosphere at 900 °C for 2 h to induce the formation of graphite structures. The prepared Co@BNCNT-900 catalyst had high-density bamboo-like N-doped CNT structures, which was beneficial to improve conductivity and provide accessible active sites. The onset and half-wave potentials of the Co@BNCNT-900 catalyst was 0.93 V (vs. RHE) and 0.82 V (vs. RHE), respectively. The electron transfer number was 3.9, which proved the excellent ORR activity. In addition, Xu et al. [64] selected 2,2′-bipyridyl-5,5′-dicarboxylic acid (H2bpydc) and cobalt acetate to synthesize the Co-MOF precursor. In the carbonization process, H2bpydc would release CO₂ and H₂O, forming mesoporous N-doped graphitic carbon layer. At the same time, cobalt tetrahydrate was reduced to metallic cobalt NPs that coordinated with graphite carbon, and it was embedded into 3D N-rich mesoporous carbon foam, exposing more active sites. After pyrolysis at 700 °C in N₂ atmosphere, the obtained CoN-CF-700 catalyst contained 3.73 at% Co, 4.13 at% N and the BET specific surface area was 593 m²·g⁻¹. The core–shell structure of Co NPs coated with graphitic carbon layer had unique structural advantages, which not only improved ORR performance through physical and chemical protection, but also played a role in the conduction pathway and reaction site in the electrocatalysis process. CoN-CF-700 catalyst also displayed an onset potential of 0.94 V (vs. RHE), and a half-wave potential of 0.85 V (vs. RHE). Besides, the core–shell structure of Co NPs enclosed by graphitic carbon layers can reduce the corrosion of active sites and play a long-term stabilizing role in ORR catalysis.

Table 3 Summary of synthesis of MOF-based Co–N–C catalyst and cobalt oxide catalyst

Table 4 Summary of ORR electrocatalytic activities of MOF-based Co–N–C catalyst and cobalt oxide catalyst

Fig. 5

a Schematic illustration of preparation of Co@BNCNTs as an electrocatalyst for oxygen reduction reaction; b SEM and c HRTEM images of Co@BNCNTs; d LSV curves of Co@BNCNTs-700, 800, 900, 1000 and Pt/C in O₂-saturated 0.1 mol·L⁻¹ KOH (aq.). Reproduced with permission from Ref. [63]. Copyright 2018 Royal Society of Chemistry

In addition to traditional organic ligands, biomass ligands can also coordinate with Co²⁺ to construct Co-MOF. Guo et al. [65] used adenine, a biomass material with high N content (51.8%) and rigid molecular structure, to prepare Co-Ade-MOF with cobalt/adenine molar ratio of 1/3. It was further pyrolyzed at 700 °C in N₂ atmosphere to obtain Co@N-CNTF-2 catalyst. The Co-MOF was synthesized by adenosine ligands and pyrolyzed without any other precursors to form a N-doped carbon/CNT skeleton containing Co NPs. On the one hand, Co NPs not only catalyzed the formation of CNTs and hollow carbon structures, but also improved the conductivity of the entire hybrid structure. On the other hand, less graphitized shells can prevent the acid corrosion, oxidation and aggregation of Co NPs during electrocatalysis, especially under the high oxidation potential of strong acidic electrolyte, so that Co NPs had good durability and stability in harsh environments. The Co-doped carbon core–shell structure provided an effective way to adjust the electronic structure of the catalyst, thus greatly reducing the adsorption free energy and local work function, and promoting the adsorption and desorption of reaction intermediates. The N content of the prepared catalyst was 6.5 at%, in which pyridinic-N and graphitic-N accounted for 90%, indicating high catalytic activity. The pore size was about 0.5 nm, which facilitated the efficient transport of electrolyte ions and gaseous substances. Therefore, in 0.1 mol·L⁻¹ KOH (aq.), Co@N-CNTF-2 catalyst exhibited an onset potential of 0.91 V (vs. RHE), and a half-wave potential of 0.81 V (vs. RHE), which was close to that of Pt/C catalyst. The electron transfer number was 3.96, indicating its admirable ORR activity.

For cobalt oxide catalyst, the Co-MOF precursor requires a mild pyrolysis temperature, because cobalt oxide can be reduced by carbon at high temperatures. Secondly, compared with Co–N–C catalyst, the pyrolysis atmosphere needs a certain amount of O₂, which is conducive to the formation of active sites. Ding et al. [66] prepared Co₃O₄ NPs by two-step pyrolysis with ZIF-67 as the precursor. The pyrolysis of ZIF-67 in H₂/Ar (H₂:Ar = 1:9) at 800 °C produced the Co NPs. Then, Co NPs were oxidized at 250 °C for 40 min under O₂/Ar atmosphere (O₂:Ar = 3:97). They were finally converted into yolk@shell Co@Co₃O₄ NPs and a small amount of hollow Co₃O₄ NPs through the nanoscale Kirkendall effect. The oxidation time had an important influence on the morphology and structure of the products. When the oxidation time reached 90 min, all Co₃O₄ NPs were converted into hollow Co₃O₄ NPs. With the change of the oxidation time, the specific surface area of the product decreased from 246 m²·g⁻¹ (no oxidation) to 176 m²·g⁻¹ (with 90 min oxidation). The optimized yolk@shell Co@Co₃O₄ NPs had the following advantages: (1) Large surface area increased the density of exposed active sites and accelerated the charge transfer; (2) With good dispersive yolk@shell Co@Co₃O₄ NPs, the high electrochemical activity and structural stability were obtained; (3) There were enough N species as supplementary active sites for ORR and hollow porous carbon shells, which can improve conductivity, promote electron transfer and mass diffusion. In 0.1 mol·L⁻¹ KOH (aq.), the half-wave potential of Co₃O₄/HNCP-40 catalyst for 40 min of oxidation was 0.843 V relative to RHE, the Tafel slope was 69 mV·dec⁻¹, and the diffusion limiting current density was 5.88 mA·cm⁻². It was close to commercial Pt/C catalyst (E_1/2 = 0.84 V (vs. RHE), J_L = 5.81 mA·cm⁻²). This was mainly due to the excellent electrochemical activity provided by highly dispersed yolk@shell Co@Co₃O₄ NPs, and the high conductivity provided by the hollow N-doped porous carbon, resulting in excellent ORR activity. Qi et al. [67] prepared the Co(ATZ)(BDC)_0.5 precursor from 3-amino-1,2,4-triazole (ATZ), 4-dicarboxylic acid (BDC) and Co(NO₃)₂·6H₂O. The Co@Co₃O₄/NC-2 catalyst was prepared by the pyrolysis of Co(ATZ)(BDC)_0.5 precursor in N₂ at 700 °C for 2 h, and in air atmosphere at 250 °C for 1.5 h. During heat treatment, the Co ions from the MOF framework were easily transformed into metallic Co NPs, which could catalyze organic ligands to form N-doped graphitic carbon layers. When CoNC-700 was further oxidized in air at 250 °C, Co NPs in CoNC-700 were gradually oxidized to Co₃O₄. Co@Co₃O₄/NC-2 had abundant micropores and mesopores, with a specific surface area of 78.27 m²·g⁻¹. During the ORR process, the layered microporous/mesoporous structure facilitated the transport of reactants/electrolytes and adequate exposure of active sites. In 0.1 mol·L⁻¹ KOH (aq.), the Co@Co₃O₄/NC-2 catalyst showed a half-wave potential of 0.15 V relative to Ag/AgCl, and the H₂O₂ yield was less than 6%, representing a 4-electron reaction. Besides, Zhang et al. [68] rationally designed a new enantiotopic 3D MOF-C2 from N-(4-(1H-imidazol-1-yl)phenyl)-4-(pyridin-4-yl)-N-(4-(pyridin-4-yl)phenyl)aniline (MIDPPA), Co(NO₃)₂·6H₂O and 1,2,4-benzenetricarboxylic acid. The C-MOF-C2-900 catalyst with graded rod-like structure can be obtained by pyrolyzing MOF-C2 in N₂ atmosphere at 900 °C for 1 h, which contained a series of ORR active Co sites, such as Co(0), Co–N_x and CoO_x. The catalyst showed a half-wave potential of 0.82 V (vs. RHE), which was higher than that of Pt/C catalyst (0.817 V (vs. RHE)). Meanwhile, it showed good stability and methanol resistance, indicating good ORR catalytic activity. Li et al. [69] mixed Co(NO₃)₂·6H₂O with polyvinylpyrrolidone (PVP), and added 1,4-benzenedicarboxylic acid (H2bdc) to hydrothermal synthesis of Co-MOF. Then, it was pyrolyzed at 600 °C in N₂ for 5 h to yield Co@C catalyst containing Co/CoO_x NPs. When the pyrolysis temperature increased from 600 to 900 °C, the average size of Co/CoO_x NPs changed from 25 nm to over 100 nm, while the specific surface area of Co@C decreased from 96.8 to 44.1 m²·g⁻¹, and the atomic ratio of Co(0):Co²⁺:Co³⁺ changed from 0.32:0.35:0.33 to 0.18:0.42:0.46. The Co@C-600 catalyst was prepared at an optimum temperature of 600 °C. In 0.1 mol·L⁻¹ KOH (aq.), the onset potential of Co@C-600 catalyst (vs. RHE) was 0.87 V, and the diffusion limiting current density was 4.9 mA·cm⁻², which proved its good ORR performance. In a word, by controlling the structure and pyrolysis conditions of Co-MOF precursor, an excellent cobalt oxide catalyst can be obtained.

3.3 Preparation of heteroatoms-enhanced Co–N–C catalyst and cobalt oxide catalyst by carbonization of Co-MOFs

Encapsulation of small or large molecules into the MOF precursor will cause some changes in the final structure of the catalyst, which affects the catalytic performance. In addition to some commonly used small molecule dopants (such as melamine and ammonium sulfate), vaporizable metal ions, such as Zn²⁺ or Mg²⁺ can also be used as dopants to modify ZIF-67, because Zn and Co have the same coordination number for 2-methylimidazole [80]. Li et al. [70] reported a variety of self-assembled bimetallic ZIF (Co_0.6Zn_0.4-ZIF) in aqueous solution. The 2D stable network was formed by the H-bonds between N atoms in 2-methylimidazole and H₂O (Fig. 6 [70]). Through the coordination of metal ions with 2-methylimidazole, these 2D networks were connected layer by layer to form ZIF nanoleaves. Owing to the same size and coordination number of Co and Zn atoms, the morphology of ZIF nanoleaves was not damaged. The Co_0.6-N/C-800 catalyst with optimum content of pyridinic-N and graphitic-N was synthesized by pyrolysis in a N₂ atmosphere at 800 °C for 2 h, and retained the lateral size around 3 μm and thickness of about 160 nm. In addition, by adjusting the molar ratio of Co and Zn, a series of N/C-800 (924.7 m²·g⁻¹, 0.553 cm³·g⁻¹), Co_0.2-N/C-800 (311.8 m²·g⁻¹, 0.244 cm³·g⁻¹), Co_0.6-N/C-800 (283.2 m²·g⁻¹, 0.421 cm³·g⁻¹), Co–N/C-800 (179.1 m²·g⁻¹, 0.377 cm³·g⁻¹) with different specific surface area and pore volume were available. This result indicated that Zn vaporization could prevent the damage of micropores in ZIF. However, among the four samples, Co_0.6-N/C-800 had the most excellent ORR performance, indicating that cobalt doping could effectively enhance the activity of the catalyst while the proper pores were necessary. In 0.1 mol·L⁻¹ KOH (aq.), Co_0.6-N/C-800 catalyst exhibited an onset potential of 0.916 V (vs. RHE) and a half-wave potential of 0.825 V (vs. RHE), which were comparable to that of Pt/C catalyst (E_onset = 0.926 V (vs. RHE) and E_1/2 = 0.811 V (vs. RHE)). These results revealed that the Co_0.6-N/C-800 catalyst had excellent ORR activity, which was mainly attributed to the rapid mass transfer of the 2D porous foliation structure and abundant active site exposure. Therefore, by adjusting the ratio of Co and Zn in ZIF, catalysts with different properties can also be produced. Since non-toxic and low-cost MgO residue could be simply removed by acid or high-temperature evaporation, Mg was also widely used as a sacrificial agent. For instance, Wu et al. [71] introduced light metal Mg to synthesize Mg–Co bimetallic MOF, and then pyrolyzed at 600 °C. The BET surface of the CNT-containing nanoporous carbon was 712.78 m²·g⁻¹, much larger than that of the Co-MOF (305.4 m²·g⁻¹). After post-synthetic modification with N/S heteroatoms at 900 °C, the resulting hierarchical carbon catalyst showed higher activity in ORR than the commercial Pt/C catalyst.

Fig. 6

a Schematic illustration of synthesis process of ZIF nanoleaves in aqueous solution and its subsequent conversion to Co_x-N/C nanoleaves; b SEM image of Co_0.6Zn_0.4-ZIF nanoleaves; c TEM image of Co_0.6-N/C-800 (inset: magnification image of red marked area; typical pore structure is marked by red circle); d LSV curves of N/C-800, Co_0.2-N/C-800, Co_0.6-N/C-800 and Co–N/C-800. Reproduced with permission from Ref. [70]. Copyright 2019 Royal Society of Chemistry

Moreover, large molecules like PANI, PVP, PS, graphene are also commonly used additives. According to the expected catalyst effects, such as improved surface roughness and conductivity, Co-MOF and large molecules can be targeted to increase activity. For instance, Liang et al. [72] prepared Co-embedded N-enriched meso/microporous carbon materials through the auxiliary pyrolysis of ZIF-67 wrapped by PVP. The addition of PVP was beneficial to the increase of the N content and the formation of the interface structure, thus further promoting the enhancement of ORR activity. The C=O group in PVP had a strong coordination effect with the metal center of ZIF-67, which prevented the disintegration of ZIF-67 during pyrolysis. The pyrolysis conditions were in a N₂ atmosphere at 800 °C for 2 h. After adding 0, 2, 4, 6 and 8 mmol PVP, the specific surface area of the prepared P-ZIF catalyst was firstly increased and then decreased. When the dosage of PVP was 4 mmol, the maximum specific surface area was 548 m²·g⁻¹. In addition to a higher specific surface area, P-Co-NC-4 catalyst had a higher N content (1.88 at%), which was conducive to combining with Co atoms embedded in C to form Co–N_x active sites and enhance the catalytic activity. It also showed an onset potential of 0.90 V (vs. RHE), a half-wave potential of 0.85 V (vs. RHE), and a 4-electron mechanism. Guan et al. [73] synthesized a single-hole cobalt/N-doped carbon hollow particle, which was mainly based on PVP modified PS sphere as template, and then a MOF (ZIF-67) shell was grown on the surface of the PS sphere. During the pyrolysis process, PS spheres were decomposed into hydrocarbon gas, which escaped from the interior and formed holes on the surface of the carbon shell. The size of surface pores was related to the heating rate. Especially when the heating rate was 10 °C·min⁻¹, strong gas emissions would completely destroy the hollow spheres. When the heating rate was 5 °C·min⁻¹, single-hole hollow spheres were formed. The hollow spheres with large holes not only helped to make full use of the internal catalytic sites, but also promoted the transport of electrolyte and the formation and release of catalytic products. The single-hole Co–N–C hollow particle catalyst exhibited an onset potential of 0.98 V (vs. RHE), a half-wave potential of 0.87 V (vs. RHE), and a Tafel slope of 102 mV·dec⁻¹. Moreover, after continuous operation for 24 h, the catalyst retained 92% of the original current, while the Pt/C catalyst lost about 36.5% after 16 h, indicating good stability. Chen et al. [74] synthesized hollow porous carbon spheres co-doped with ultra-small Co NPs by pyrolyzing a core–shell precursor composed of PS core and bimetallic Zn-Co-ZIFs (BMZIFs) shell (Fig. 7 [74]). The Zn ions in BMZIFs could not only spatially separate Co species to prevent Co NPs from aggregating, but also evaporate during pyrolysis to provide additional pores, and the PS core would decompose to create a unique hollow structure as well. Owing to the large surface area (737.4 m²·g⁻¹), high porosity (pore volume = 0.81 cm³·g⁻¹), tunable particle size (d = 4.28 nm), high graphitization degree (I_D/I_G = 0.95), and rich in highly active Co–N_x sites, the Co-HNCS-0.2 catalyst exhibited a remarkable ORR activity with the onset and half-wave potentials of 0.94 V (vs. RHE), and 0.82 V (vs. RHE), respectively, which was higher than that of commercial 20% Pt/C catalyst (0.93 V (vs. RHE) and 0.80 V (vs. RHE), respectively).

Fig. 7

Schematic illustration of procedures for synthesis of a HNCS, b Co-HNCS, and c Co-HNCS-x. Reproduced with permission from Ref. [74]. Copyright 2017 Elsevier B.V.

Owing to the outstanding conductivity and processability, graphene occupies an important position in the field of material composites and catalysis [81, 82]. In the preparation of cobalt oxide, in addition to using oxygen or air for oxidation, graphene oxide (GO) can also be used as an oxygen source, as it has rich oxygen-containing groups that can improve the conductivity of the catalyst. For example, Cai et al. [75] mixed GO and 2-methylimidazole, and then added Co²⁺ to self-assemble Co-MOF layer by layer on the GO surface. As a substrate, GO was negatively charged in water, while 2-methylimidazole was positively charged in water, and they were held together by electrostatic attraction. The subsequent addition of Co²⁺ would coordinate with 2-methylimidazole to form Co-MOF on the GO surface. Different layers of Co-MOF/GO were obtained by repeated impregnation. After pyrolysis in a N₂ atmosphere at 600 °C for 5 h, the Co-MOF was converted into a Co/N-doped porous carbon catalyst with BET specific surface area of 252 m²·g⁻¹ and I_D/I_G ratio of 0.99, providing porous channels and high active sites for the catalyst. Co–N_x and CoO_x can be used as catalytic active centers to improve electrocatalytic performance. GO can be reduced to graphene (rGO), thus effectively improving the conductivity of the catalyst and preventing the aggregation of MOF precursors with metal particles. In 0.1 mol·L⁻¹ KOH (aq.), the Co–N–C/rGO-6-600 catalyst exhibited an onset potential of 0.963 V (vs. RHE), a half-wave potential of 0.87 V (vs. RHE), a Tafel slope of 54 mV·dec⁻¹, and a 4-electron mechanism. The stability test showed that the current of CO–N–C/RGO-6-600 maintained 97% under the long-term operation of 10,000 s, while the current loss of Pt/C catalyst was 7%. Guo et al. [76] used thermally shock exfoliated GO (EGO) as a substrate to grow modified ZIF-67 crystals in situ [83], and prepared Co₃O₄/Co@N-G-450 as an ORR catalyst. EGO had a high specific surface area (~ 700 m²·g⁻¹) and a soft porous network (~ 2 to 200 nm in size). The surface contained a large number of oxygen functional groups and defects, which was conducive to the anchoring and growth of ZIF-67. During the synthesis process, it was very important to control the concentration of precursors, since ZIF overgrowth could lead to particle aggregation and dissociation. In addition, the control of pyrolysis temperature was also very important, because the oxygen functional groups of EGO would dissociate at high temperature, leading to the etching of ZIF-67 and the oxidation of Co center. When ZIF was loaded with 60 wt%, the pyrolysis yield was as high as 65% after pyrolysis in N₂ at 450 °C for 6 h. The Co₃O₄/Co@N-G-450 catalyst exhibited an onset potential of 0.962 V (vs. RHE), a half-wave potential of 0.808 V (vs. RHE), and a Tafel slope of 62 mV·dec⁻¹. Long-term stability test showed that after continuous operation for 50 h, the Co₃O₄/Co@N-G-450 catalyst remained 75% activity. Additionally, it was found by Yi et al. [77] that carbon black (CB) could not only improve the conductivity of Co-MOF, but also regulate the N type. The Co-MOF was prepared with 3,5-dimethyl-4-carboxypyrazole and Co²⁺, and then the mixture of MOF and CB was pyrolyzed to prepare Co–Co_xO/N-doped carbon catalyst. The results revealed that when the mass ratio of MOF to CB was 3:1, CB effectively increased the total content of pyridinic-N and graphitic-N. It accounted for more than half of the total N and showed the best ORR catalytic activity. Moreover, CB could not only effectively increase the conductivity of the catalyst and promote electron transfer, but also improve the dispersibility of Co NPs, thus enhancing the catalytic activity. The BET specific surface area of MOF(Co)/CB (3:1)-500 catalyst was 203.5 m²·g⁻¹. It also had an onset potential of 0.92 V (vs. RHE), a half-wave potential of 0.8 V (vs. RHE), and a Tafel slope of 51 mV·dec⁻¹. The measured electron transfer number was 3.91–3.98 (0.06 to 0.56 V (vs. RHE)), suggesting a favorable ORR catalyst.

3.4 Preparation of Co-based catalyst by carbonization of Co-loaded MOF

In addition to being a pyrolysis precursor for catalysts, MOF can also be used as a synthesis template to provide a favorable structure. For example, Park et al. [78] designed a mixed MOF (ZIF-L@ZIF-67) with leaf core–shell structure as the precursor to prepare porous carbon leaves Co, N-PCLs catalyst after pyrolysis (Fig. 8 [78]). ZIF-L was synthesized hydrothermally with Zn²⁺ and 2-methylimidazole as the core, and then a ZIF-67 layer made of Co²⁺ and 2-methylimidazole was grown on the core surface as the shell. After 5 h pyrolysis in N₂ atmosphere at 800 °C, the Co, N-PCLs catalyst was produced, which maintained the original form of ZIF-L and formed more CNTs on the surface. Compared with C-ZIF-L (38 m²·g⁻¹ and 0.03 cm³·g⁻¹) and C-ZIF-67 (254 m²·g⁻¹ and 0.18 cm³·g⁻¹), the Co, N-PCLs catalyst had higher specific surface area (319 m²·g⁻¹) and porosity (0.22 cm³·g⁻¹), which were conducive to electrolyte penetration and electron transfer. Besides, it showed an onset potential of 0.992 V (vs. RHE), a half-wave potential of 0.91 V (vs. RHE), and a diffusion limiting current density of 5.22 mA·cm⁻², which was similar to Pt/C catalyst but higher than C-ZF-L (E_1/2 = 0.592 V (vs. RHE), J_L = 1.86 mA·cm⁻²) and C-ZF-67 (E_1/2 = 0.831 V (vs. RHE), J_L = 3.56 mA·cm⁻²). Meanwhile, the electrochemical stability of the Co, N-PCLs catalyst was higher than that of Pt/C. Bhattacharyya et al. [79] prepared NCD-1, a N-doped carbon dot, through hydrothermal synthesis of anionic Mg-MOF, composite pyrolysis of NCD-1, and 5,10,15,20-tetrakis (4-methoxyphenyl)-21H,23H-porphine cobalt (II) as a cobalt source. In 0.1 mol·L⁻¹ KOH (aq.), the onset potential of Co@CN catalyst was 0.8 V (vs. RHE), which was higher than that of NCD-1 catalyst (0.36 V (vs. RHE)), and the yield of H₂O₂ was 4%, far lower than that of NCD-1 (40.8%). The loading of Co had a positive effect on improving the catalytic performance of N-doped C materials, because Co and N combined to form the active sites of Co–N_x, while cobalt oxide could promote the ORR catalytic reaction. Compared with the direct synthesis of Co-MOF, the selection of suitable supported Co catalyst for the preparation of Co-based catalyst is a little more complex, but can enhance the advantages of Co-supported catalyst.

Fig. 8

a Schematic representation for formation of leaf-shape ZIF-L and leaf-shape core–shell type ZIF-L@ZIF-67, and its transformation to thin Co, N-PCL; b, c SEM images of Co, N-PCL (4); d LSV in O₂-saturated 0.1 mol·L⁻¹ KOH electrolyte of C-ZIF-L, C-ZIF-67 and Co, N-PCL (4). Reproduced with permission from Ref. [78]. Copyright 2019 Elsevier B.V.

3.5 Summary of cobalt-based catalyst

As a short conclusion, there are various synthesis strategies and methods for Co-MOF. According to the coordination number of Co ions and the coordination form of organic ligands, Co-MOF with different morphologies can be designed to further prepare high-performance ORR catalyst. Hybrid doping provides more possibilities for Co-based catalysts. The addition of heteroatoms can not only change the pyrolysis behavior of Co-MOF, increase the specific surface area, and adjust the pore size, but also enhance the activity of the catalyst. Whichever method is used to prepare the highly active co-catalyst, the following factors are worth considering: (1) Determine the active site of Co-based catalyst and take it as the starting point of the design materials; (2) Promote the formation of active sites, e.g., heteroatom doping and material composite; (3) High conductivity and porous network structure are conducive to electron transfer and reaction rate.

4 Other MOF-based non-noble metal (Cu, Mn, Ni, Mo) catalysts

4.1 Cu-based catalysts

In addition to Fe–N–C and Co–N–C catalysts with excellent performance, many non-noble metal catalysts with other metal ions are also worth for further exploration and development. Copper is an abundant and cost-effective metal on earth with high conductivity (close to silver), which can promote electron transfer and exhibit great potential in the ORR field. The research progress of Cu-based ORR catalysts in recent years was reported. Wang et al. [84] used N-containing pyrazine and P-containing 1-hydroxyethylidene-1,1-diphosphoonic acid as organic ligands, and then self-assembled with Cu²⁺ to prepare Cu–NPMOF by in situ doping of N and P (Fig. 9 [84]). The reasonable selection of MOF with multiple heteroatoms (such as N, P, S) as single-source precursor is a good strategy to obtain catalyst with uniform doping, higher component repeatability, high activity and stability. Cu–NPMOF was pyrolyzed at 650 °C in a N₂ atmosphere to prepare Cu@NPPC (Table 5 [84,85,86,87,88,89,90,91,92,93]). Then, it was ground and mixed with NaH₂PO₂·H₂O for secondary pyrolysis (250 °C, N₂). Finally, the graded porous carbon-containing Cu₃P NPs (Cu₃P@NPPC-650 catalyst) was produced. The specific surface area of Cu₃P@NPPC-650 (1004.7 m²·g⁻¹) was much higher than that of Cu@NPPC-650 (402.5 m²·g⁻¹). This was because NaH₂PO₂·H₂O could form pores during pyrolysis after decomposition. The porous carbon shells with high specific surface area, the active sites of Cu₃P NPs, and the heteroatoms doping were the main reasons for its high performance. In 0.1 mol·L⁻¹ KOH (aq.), Cu₃P@NPPC-650 catalyst had a half-wave potential of 0.78 V (vs. RHE), only 29 mV lower than Pt/C catalyst (Table 6 [84,85,86,87,88,89,90,91,92,93]), and its stability was better than Pt/C catalyst. Li et al. [85] prepared Cu-MOF/KB by mixing Cu²⁺, 1,4-benzenedicarboxylic, triethylenediamine and HNO₃-functionalized KB (Ketjenblack carbon, EC-300J). After pyrolysis at 800 °C, the final CuNC/KB-400 catalyst was yielded. Using Cu-MOF as template, KB was modified to obtain a novel hybrid catalyst which could form Cu/Cu₂O NPs and amorphous CuN_xC_y. The conductivity of CuNC/KB-400 was 5.62 S·cm⁻¹ and the specific surface area was 394.3 m²·g⁻¹. In CuNC/KB-400, Cu/Cu₂O NPs and CuN_xC_y could synergistically improve ORR performance, leading to higher half-wave potential (0.82 V (vs. RHE)) and diffusion limiting current density (6.05 mA·cm⁻²) than Pt/C catalyst.

Fig. 9

a Schematic illustration of preparation of Cu₃P@NPPC; b LSV curves of Cu₃P@NPPC-600, 650, 700 and Pt/C in O₂-saturated 0.1 mol·L⁻¹ KOH solution, inset being corresponding mass activities and error bar of Cu₃P@NPPC catalyst; c endurance test of Cu₃P@NPPC-650 catalyst. Reproduced with permission from Ref. [84]. Copyright 2017 Wiley–VCH

Table 5 Summary of synthesis of MOF-based non-noble metal (Cu, Mn, Ni, Mo) catalysts

Table 6 Summary of ORR electrocatalytic activities of MOF-based non-noble metal (Cu, Mn, Ni, Mo) catalysts

Obviously, doping Cu²⁺ into MOF precursor is a common strategy to prepare Cu-based catalyst. Xie et al. [86] used adsorption method to combine Cu²⁺ to ZIF-8 framework, and after further pyrolysis at 1000 °C to produce Cu–N/C catalyst with Cu NPs and Cu(II)-N active sites (the Cu content was 0.69 wt%). Although Cu doping would make the total N content (1.21 at%) and pyridinic-N content (0.18 at%) in Cu–N/C lower than that in N/C (3.80 at% and 1.08 at%, respectively), the ORR activity of Cu–N/C catalyst was much higher than that of N/C catalyst. The half-wave potential of Cu–N/C was 0.813 V (vs. RHE), similar to Pt/C (0.821 V (vs. RHE)), but higher than that of N/C (0.775 V (vs. RHE)). Furthermore, compared with Pt/C (84 mV·dec⁻¹) and N/C (120 mV·dec⁻¹), the Tafel slope of Cu–N/C was lower (81 mV·dec⁻¹), indicating that the presence of Cu species could effectively improve the ORR kinetic of N-doped carbon. After treating Cu in Cu–N/C with 1 mol·L⁻¹ HNO₃, it was found that the half-wave potential reduced from 0.813 to 0.764 V (vs. RHE), and J_L decreased from 6.0 to 4.38 mA·cm⁻². This result showed that Cu (0) NPs could stabilize the adjacent Cu(II)–N (i.e., N–Cu(II)–Cu(0)). Lai et al. [87] also prepared Cu-ZIF-8 by adsorption method, and then obtained Cu–N/C catalyst through pyrolysis at 900 °C under N₂ atmosphere. By controlling the copper load and effectively regulating the states of Cu (II) and Cu (0), the highly active hybrid coordination sites of N–Cu(II)–Cu(0) were constructed. For the 0%, 25% and 40% Cu–N/C samples, the Cu content was 2.96 wt%, 5.51 wt% and 9.90 wt%, as well the N content was 11.20 wt%, 11.04 wt% and 9.05 wt%, respectively. For the 40% Cu–N/C sample with high Cu doping content, after pyrolysis, Cu in Cu-ZIF-8 was easier to form Cu NPs than Cu–N sites. On the contrary, when Cu–N/C content was 10%, Cu was mainly distributed in carbon matrix in the form of atomic size. Therefore, Cu content played an important role in the formation of Cu–N sites, and adjacent metallic Cu NPs would stabilize the Cu–N sites. The 25% Cu–N/C catalyst showed the best ORR activity. Its onset potential (0.914 V (vs. RHE)), half-wave potential (0.813 V (vs. RHE)), and J_L (5.50 mA·cm⁻²) were very close to 30 wt% Pt/C (0.936 V (vs. RHE), 0.810 V (vs. RHE), 5.46 mA·cm⁻², respectively). Through the current time test, the current attenuation of 25% Cu–N/C was only 12.6% after continuous 10,000 s operation, while that of 30 wt% Pt/C described by 21.9%.

Generally, Cu–N/C is prepared by the pyrolysis of Cu-doped ZIF (without considering the pyrolysis atmosphere), and the Cu species can exist in the following states: (i) Cu (0) NPs; (ii) CuO_x; (iii) Cu–N_x (including Cu⁺–N₂ and Cu²⁺–N₄). Li et al. [94] etched Cu–N/C with 0.5 mol·L⁻¹ H₂SO₄ and 1.3 mol·L⁻¹ HNO₃, respectively, and obtained Cu–N/C-H₂SO₄ catalyst containing Cu⁺–N₂, and Cu–N/C-HNO₃ catalyst containing Cu²⁺–N₄. Then, the effects of Cu⁺–N₂ and Cu²⁺–N₄ on ORR process were discussed. Cu–N/C was prepared by pyrolysis of Cu-ZIF-L at 800 °C in N₂ atmosphere. The low concentration of H₂SO₄ removed the remaining zinc and possible metal oxides, and had little effect on the Cu–N structure. However, HNO₃ could not only remove metals and metal oxides, but also oxidize Cu⁺ to Cu²⁺. X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) spectra confirmed the presence of Cu⁺–N₂ active sites in Cu–N/C-H₂SO₄, and Cu²⁺–N₄ active sites in Cu–N/C-HNO₃. For Cu–N/C-H₂SO₄ and Cu–N/C-HNO₃, the Cu contents were 23.9 wt% and 7.9 wt%, respectively, and the specific surface areas were 774.3 and 895.3 m²·g⁻¹, respectively. The E_onset (0.987 V (vs. RHE)) and E_1/2 (0.860 V (vs. RHE)) of Cu–N/C-H₂SO₄ catalyst were significantly higher than those of Cu–N/C-HNO₃ catalyst (E_onset = 0.927 V (vs. RHE), E_1/2 = 0.802 V (vs. RHE)) and Cu–N/C catalyst (E_onset = 0.985 V (vs. RHE), E_1/2 = 0.800 V (vs. RHE)). After 5000 cyclic voltammetry (CV) cycling test, the ORR activity of Cu–N/C-H₂SO₄ catalyst showed no obvious change, while the E_1/2 of Cu–N/C-HNO₃ and Pt/C catalyst displayed negative shift of 11 and 20 mV, respectively, indicating that Cu–N/C-H₂SO₄ had good catalytic stability. Li et al. [95] calculated the adsorption energy for O₂ and OOH at the Cu–N₂ and Cu–N₄ sites. The former had much higher O₂ and OOH adsorption energies, and the O–O bond stretching of O₂ and OOH was more prone to break (O–O bond length were 0.01455 and 0.01460 nm, respectively). Cu–N₄ showed relatively low adsorption energies for O₂ and OOH, 0.74 and − 0.23 V, respectively. The O–O bonds of O₂ and OOH were less stretched, at 0.01289 and 0.01435 nm, respectively. The higher adsorption energies of active sites and O–O stretching degree were helpful to promote the ORR process, which implied that Cu–N₂ had strong catalytic activity. For the preparation of Cu-based catalysts, it is noteworthy that the content and form of copper may maintain the predominance of precursor structures and the formation of active sites.

4.2 Other M-based catalysts (M = Ni, Mn, Mo)

Ni-based catalysts are often studied in the field of catalysis, but their catalytic performance in ORR is always inferior to that of Fe–N–C and Co–N–C catalysts [96,97,98]. Wang et al. [99] prepared electrocatalyst by self-assembly using metal ions (Fe, Co, Ni) and glycerin. The products were Fe₃C-CNTFs (210.610 cm²·g⁻¹, 1.92 nm), Co-CNTFs (268.739 cm²·g⁻¹, 2.21 nm) and Ni-CNTFs (284.416 cm²·g⁻¹, 2.30 nm), which had similar specific surface area and average pore size. The electrocatalytic activity was in descending order: Co-CNTFs > Fe₃C-CNTFs > Ni-CNTFs. Their electron transfer numbers were 3.9, 2.9 and 2.7, respectively. Ni-CNTFS mainly covered the crystal surfaces of Ni (111) and Ni (200). By DFT calculation, Ni (111) had a strong binding energy (− 0.149 eV·atom⁻¹) to OOH, leading to a 2-electron dominated ORR process. However, Co (111)–OOH and Fe₃C (031)–OOH were unstable and prone to produce free OH⁻, so they proceeded through the 4-electron reaction pathway. In addition, Zhao et al. [88] prepared the self-assembled Ni-FA from Ni²⁺ and folic acid in the presence of hydrazine. After pyrolysis at 600 °C, Ni-FA was converted into Ni@NiNC-600 catalyst with high N content (N: 11.45 at%). Echinops-like Ni@NiNC-600 contained a large number of regularly dispersed CNTs with a diameter of ~ 200 nm, and Ni NPs were enclosed in graphite shells composed of CNTs. Higher N content in Ni@NiNC-600 contributed to the formation of more Ni–N active sites, while CNTs could promote the conductivity. Owing to the synergistic effect of unique hierarchical porous nanostructure composed of numerous CNTs, Ni@NiNC was fixed on the microsphere, exposing abundant Ni–N active sites, and showing superior ORR performance in alkaline media. The catalyst’s onset potential and half-wave potential (E_onset = 0.88 V (vs. RHE), E_1/2 = 0.85 V (vs. RHE)) were higher than Ni@NiNC-350 (E_onset = 0.59 V (vs. RHE), E_1/2 = 0.65 V (vs. RHE)) and commercial Pt/C (E_onset = 0.86 V (vs. RHE), E_1/2 = 0.84 V (vs. RHE)).

The addition of heteroatoms can effectively improve the electrocatalytic performance of Ni-based catalysts. For example, Yan et al. [89] reported a N, S double-doped rod catalyst (Ni-NSPC-6) prepared from nickel acetate and 2-mercapto-5-methyl-1,3,4-thiadiazole. After removing Ni and excess nickel sulfide with 6 mol·L⁻¹ HCl, the remaining Ni₃S₂ NPs (520 nm) were wrapped in the amorphous carbon layer, preventing the dissolution and aggregation in the catalytic process. As the co-doping of N, S played a positive synergistic effect, the one-dimensional structure facilitated the electrode–electrolyte contact and promoted the transport of ions and electrons. The onset potential of Ni-NSPC-6 catalyst was 0.08 V (vs. Ag/AgCl), and its stability was better than that of Pt/C catalyst. Tyagi et al. [90] synthesized Ni-GT precursor using nickel nitrate hexahydrate and guanidine thiocyanate containing N and S, and then heated the precursor and treated it with acid to produce Ni-GT-750-A (Fig. 10 [90]). When the guanidine thiocyanate was decomposed at high temperature (~ 1000 °C), Ni might bond with sulfur to form Ni_xS_y (NiS and Ni₃S₂). The Ni/Ni_xS_y NPs and N, S-doped carbon substrates had good dispersion, which jointly promoted the ORR, so the NI-GT-750-A catalyst exhibited positive onset potential in acidic and alkali electrolytes (0.89 V (vs. RHE), and 0.91 V (vs. RHE), respectively). Therefore, through the reasonable design and preparation, Ni-based catalyst can also have good ORR performance.

Fig. 10

a Pictorial scheme for illustration of synthesis procedure of Ni-GT-750-A; b HAADF-STEM image of Ni-GT-750-A; c LSV curves of Ni-GT-750-A in 0.1 mol·L⁻¹ KOH (aq.). Reproduced with permission from Ref. [90]. Copyright 2020 Elsevier Inc.

Manganese oxide (MnO_x) is a common non-noble metal catalyst and can be applied into the fields of OER and ORR [100,101,102]. Through effective means, MnO_x can be designed into various forms (such as nanocrystalline particles, nanorods and nanotubes); however, the ORR activities of MnO_x in different valences, and crystal structures are different [103]. Furthermore, the weak conductivity of MnO_x greatly limits its development. Combined with large scale, high porosity and high conductivity, the catalytic performance of MnO_x can be promoted. For instance, Ji et al. [91] synthesized a bimetallic N-rich tetrazolate-based MOF (Cd/Mn-MOF) and used it as a precursor to prepare the MnO@NC-1100 catalyst by pyrolysis at 1100 °C. During the pyrolysis process, Cd (II) evaporated (the boiling point is 765 °C), and tetrazolate would explode, leading to the formation of pores. After thermal annealing, MnO@NC-1100 was mainly composed of carbon nanorods and a small amount of MnO particles wrapped in graphene-like carbon layers. The MnO@NC-1100 catalyst showed significant ORR activity in 0.1 mol·L⁻¹ KOH (E_onset = 0.90 V (vs. RHE), E_1/2 = 0.74 V (vs. RHE)), and its activity did not decrease greatly after 2000 cycles. Najam et al. [92] took the N-doped carbon polyhedrons (NCP) prepared by the pyrolysis of ZIF-8 as template, and then hydrothermally reacted with Mn(NO₃)₂ to obtain Mn₃O₄@NCP catalyst, which contained the oriented growth of Mn₃O₄ nanotubes and porous NCP (Fig. 11 [92]). Although the specific surface area of Mn₃O₄@NCP (613 m²·g⁻¹) was only half that of NCP (1364 m²·g⁻¹), it had very high pyridinic-N (34.51 at%) and graphitic-N (30.76 at%) content. The presence of Mn may be responsible for nitrogen protection. The mesoporous carbon structure with high N content (pyridinic-N and graphitic-N) reduced the adsorption energy barrier of O₂ and improved the ORR performance. The Mn₃O₄@NCP catalyst and Pt/C catalyst had the same E_onset (0.92 V (vs. RHE)) and similar diffusion limiting current density (5.24 and 5.0 mA·cm⁻², respectively). The H₂O₂ yield of Mn₃O₄@NCP was less than 3%, lower than that of Pt/C (5%) and NCP (30%), indicating that it was an excellent ORR catalyst.

Fig. 11

a Graphical expression of synthetic method; b TEM image of Mn₃O₄@NCP; c LSV curves of Mn₃O₄@NCP in 0.1 mol·L⁻¹ KOH solution. Reproduced with permission from Ref. [92]. Copyright 2020 Elsevier Ltd.

Recently, 2D metal sulfide material, MoS₂, has attracted wide attention in the field of electrocatalysis with excellent electrical and optical properties [104,105,106,107]. Amiinu et al. [93] prepared N-doped porous carbon matrix (N/C) by the pyrolysis of ZIF-8, and then the MoS₂ nanosheets were in situ generated on the N/C. After heat treatment at 800 °C, Mo–N/C@MoS₂ catalyst was obtained. The semiconductor properties of MoS₂ and the formation of Mo–N sites were beneficial to improve the catalytic activity of Mo–N/C@MoS₂. It presented similar onset and half-wave potential (E_onset = 0.90 V (vs. RHE), E_1/2 = 0.81 V (vs. RHE)) with Pt/C catalyst (E_onset = 0.92 V (vs. RHE), E_1/2 = 0.81 V (vs. RHE)). Mo–N/C@MoS₂ displayed larger diffusion limiting current density (J_L = −5.3 mA·cm⁻²) than Pt/C (J_L = −4.9 mA·cm⁻²). The ORR activity of Mo–N/C@MoS₂ catalyst mainly came from the synergistic effect of Mo–N and N/C phase, especially the combination of N doping and high concentration of pyridine-N, which enhanced the adsorption and reduction of oxygen. So far, there have been few studies on the preparation of M-based catalysts (M = Ni, Mn, Mo) using MOF as the precursor. Admittedly, the ORR performance of these non-noble metals is similar to that of Pt/C when properly designed and prepared, and the unique advantages of different metal centers suggest that there is still much room for improvement.

5 MOF-based non-noble multi-metal catalyst

Compared to single-metal catalysts, multi-metal composites are used to improve the ORR performance of catalysts, because the multi-metal components usually exhibit synergistic effects to improve the ORR performance. Multi-metal MOF is an excellent precursor for the synthesis of multicomponent non-noble metal catalysts, especially Fe or Co-centered metal MOF with catalytic potential [108]. For example, Wu et al. [109] utilized Co/Zn-ZIF with unique porous structure as a cage to capture Fe(III) to synthesize FeCo-NCs catalyst by pyrolysis at 600 °C (Table 7 [109,110,111,112,113]). Fe and Co can be well dispersed in the carbon matrix and form Fe–N and Co–N active sites with N atoms, thereby synergistically promoting electrocatalytic performance. The FeCo-NC catalyst had a high specific surface area of 647.6 m²·g⁻¹ and lamellar pores, showing an activity (E_1/2 = 0.855 V (vs. RHE)) comparable to Pt/C (E_1/2 = 0.827 V (vs. RHE)) (Table 8 [109,110,111,112,113]). Moreover, Wang et al. [110] incorporated the synthesized Fe and Co into P-doped porous carbon catalyst (FCPA-900), which was derived from Fe/Co-MOF precursor by in situ carbonization. The specific surface area of the sample after carbonization at 900 °C was relatively high (1646 m²·g⁻¹), which was conducive to the transportation of reactants. Since P doping changed the symmetry of the spin density of carbon atom, and the P–O–Fe or P–O–Co bond changed the electronic structure, the catalyst showed excellent ORR performance and stability.

Table 7 Summary of synthesis of MOF-based non-noble multi-metallic catalyst

Table 8 Summary of ORR electrocatalytic activities of MOF-based non-noble multi-metallic catalyst

In addition to Fe and Co bimetallic catalyst, Fe or Co with other non-noble multi-metallic catalysts are also promising for ORR activity. Dong et al. [111] prepared Fe₃Mn₁/NCNTs-100 catalyst by synthesizing MOF-74 (Fe and Mn as the metal ions) and adding melamine as the N source for pyrolysis. By adjusting the molar ratio of Fe and Mn, it was found that the addition of Fe could significantly reduce the particle size of the sample, and the presence of Mn ions and melamine could promote the formation of CNTs. By comparing the concentration of melamine, it was found that the higher the concentration, the larger the diameter of CNTs. Fe₃Mn₁/NCNTs-100 was a typical core–shell structure, containing a small amount of NPs (average diameter of ~ 26.0 nm) wrapped by CNTs (~ 80 nm in diameter). Owing to the unique microstructure of Fe–Mn alloy coated with multilayer N-doped CNTs, high pyridinic-N content (57.32%) and Fe–N_x active sites, Fe₃Mn₁/NCNTs-100 catalyst showed the best performance at the ORR activity with E_1/2 of 0.865 V (vs. RHE) and J_L of − 5.33 mA·cm⁻², compared with commercial Pt/C catalyst (0.855 V (vs. RHE) and − 5.27 mA·cm⁻²). Besides, Xia et al. [112] synthesized a 2D-hexagonal-leaf-like ZIFs lamella (ZIF-L) precursor by solution-mediated method, and then pyrolyzed to prepare a novel Ceria@2D-hexagonal-leaf-like hierarchical porous carbon nanosheets (Ce-HPCN). Since N atoms in 2-methylimidazole tended to connect with H atoms in H₂O to form N–H–N, adding a small amount of aqueous solution could induce the conversion of Co-Zn-ZIF to 2D ZIF-L. Ce-HPCN showed good electrocatalytic performance (E_1/2 = 0.831 V (vs. RHE)), close to 30% Pt/C catalyst (E_1/2 = 0.846 V (vs. RHE)). The result was due to the synergistic effect of Co–N_x and CeO₂, especially the coexistence of Ce⁴⁺ and Ce³⁺ redox couple was conducive to O₂ adsorption. Moreover, Xu et al. [114] used NiCl₂·6H₂O, Co(NO₃)₂·6H₂O, 2-methylimidazole, and terephthalic acid as raw materials to prepare 2D multi-metallic MOF (Co, Ni MOF), and then further transferred to the Co, Ni-embedded N-doped porous carbon (CoNi-NC) catalyst with a high specific surface area of 362 m²·g⁻¹ through pyrolysis at 900 °C for 2 h. The prepared bimetallic CoNi-NC retained the nanoleaf morphology and contained some graphitic nanotubes in the nanoleaf. The CoNi-NC catalyst exhibited high-efficiency ORR catalytic performance. Its onset potential was 1.038 V (vs. RHE), and the half-wave potential was 0.866 V (vs. RHE), which were higher than that of Pt/C catalyst (E_onset = 1.006 V (vs. RHE), E_1/2 = 0.850 V (vs. RHE)). Li et al. [113] prepared PNi-MOF and DNi-MOF with different morphologies by reaction of nickel nitrate with tris-1,10-phenanthroline and dimethylglyoxime (Fig. 12 [113]). Then, ZIF-67 was epitaxially grown on the surface of PNi-MOF and DNi-MOF. After pyrolysis at 750 °C for 3 h, a two-atom Co/Ni dual sites was obtained in N-doped porous carbon catalysts (CoPNi-N/C and CoDNi-N/C). The measured specific surface area for CoPNi-N/C was 446.8 m²·g⁻¹, and 398.3 m²·g⁻¹ for CoDNi-N/C. Their V_mic/V_t (the ratio of micropore volume to total pore volume) values were 41.72% and 27.06%, respectively, much higher than that of Co–N/C (354.9 m²·g⁻¹ and 22.4%). It was worth noting that in the Co–N/C system, the two N atoms can obtain 1.18 and 1.11 electrons; while in the CoXNi–N/C (X = P or D) system, two N atoms can obtain 1.23 and 1.20 electrons, indicating that the Ni–N bond can increase the electron activity of N atoms. Under similar conductivity condition, the increase of electron concentration and hole concentration of the catalyst plays a more important role in the catalyst performance [115]. Therefore, compared with single metal Co–N/C catalyst, the addition of Ni can cause positive changes in the onset potential and reduction peak potential. The electrocatalytic activity analysis showed that the CoPNi-N/C catalyst was superior to CoDNi-N/C catalyst under both alkaline and acidic conditions. In 0.1 mol·L⁻¹ KOH and 0.1 mol·L⁻¹ HClO₄ aqueous solution, the CoPNi-N/C catalyst exhibited an onset potential of 0.93 V and 0.86 V (vs. RHE), a half-wave potential of 0.84 V (vs. RHE), and 0.73 V (vs. RHE), and a Tafel slope of 43 mV·dec⁻¹ and 61 mV·dec⁻¹, respectively. Meanwhile, the ORR activity of CoDNi-N/C catalyst was only slightly lower than that of CoPNi-N/C catalyst, because CoPNi-N/C had 41.72% micropores, higher than that of CoDNi-N/C, which was favorable for the absorption of oxygen molecules, thus reducing the ORR activation energy. The superior ORR activity made it the best electrocatalyst reported at that time. For the preparation of multicomponent catalysts with multi-metallic MOF, the selection and ratio of metals, as well as the existing form after pyrolysis, have an important effect on the performance of catalysts. The synergy between different metal phases can effectively improve the ORR activity, but the synergistic effect of non-noble multi-metal catalysts still need systematical study.

Fig. 12

Schematic illustration of synthesis process for a CoPNi-N/C and b CoDNi-N/C; plane averaged electron density difference of c Co–N/C and d CoXNi-N/C models. Reproduced with permission from Ref. [113]. Copyright 2019 Elsevier B.V.

6 Summery and outlook

This review mainly summarized the preparation methods of non-noble metal ORR catalysts by MOF carbonization in recent three years from a synthesis perspective. Through the design of MOF precursor and carbonization conditions, the improvement of catalytic performance was discussed from the aspects of preparation, dispersion and exposure area of specific active sites, and the promotion strategy was proposed. In general, there are four different methods for synthesizing non-noble metal catalysts from MOF. (1) Direct pyrolysis of metal MOF using target metal-centered atoms is simple and effective. (2) MOF is used as the carrier to adsorb the target metal and then undergoes pyrolysis. Therefore, the choice of carrier and the doping amount of target metal have an important influence on the activity of catalyst. (3) The porous topological structure of MOF is used as template to synthesize the target product, and the morphology of the target product was optimized to increase the interface contact. (4) MOF is introduced in situ or mixed with heteroatom sources to further enhance the ORR catalytic activity. For most metal-based catalysts (M = Fe, Co, Cu, etc.), the main catalytic active centers are mainly M–N_x, which means that the pyrolysis precursor must have high N content and metal dispersibility. Similarly, the porosity, hydrophilicity and conductivity of carbon base also have important effects on the activity of the catalyst. These characteristics are mainly related to the precursor structure, pyrolysis conditions and post-treatment methods.

Designing the MOF structure at the molecular level is conducive to the dispersion of metals and heteroatoms, and the formation of effective active sites during pyrolysis. For non-noble metal ORR catalysts, in addition to the active sites of metal-N, the ratio and content of pyridinic-N, pyrrolic-N and graphitic-N also have a certain impact on the performance of catalysts. However, the type and content of N, and the formation of graphite carbon are mainly related to the pyrolysis temperature. Excessive pyrolysis temperature contributes to the formation of graphite carbon and specific surface area, but it will reduce the content of N and increase the diameter of the metal particles, which weaken the catalytic activity. Therefore, in the preparation of catalysts, it is indispensable to adjust the formulation ratio and pyrolysis temperature effectively to optimize the performance of catalysts. A good catalyst does not require a high metal content, but a relatively high nitrogen content, as this will facilitate the formation of the M–N_x site. At the same time, the structure of porous network is also an important part. However, larger specific surface area does not imply higher catalytic activity, which is related to the dispersion and exposure of active sites. This means that the active region of the catalyst can more intuitively characterize the activity of the catalyst than the specific surface area. Hence, in the design of MOF, the influence of multiple factors should be considered to determine the metal type and active site and make structure adjustments such as heteroatom doping, material composite, solution etching and pyrolysis condition, to prepare hydrophilic, conductive, porous and highly active catalysts.

Although many excellent non-noble metal ORR catalysts have been prepared by MOF carbonization, some problems still need to be solved. First of all, the preparation of non-noble metal catalysts by MOF carbonization is a better method, but how to shorten the synthesis cycle of catalysts and improve the pyrolysis yield is still a huge challenge. Secondly, studies on the preparation of non-noble metal catalysts by MOF pyrolysis mainly focus on iron, cobalt, nickel and copper, etc., while other non-noble metal ORR catalysts need further exploration. Thirdly, how to prepare catalysts with high activity, stability and corrosion resistance under acidic conditions is worth for further study. Fourthly, the synergistic effect of non-noble multi-metal catalysts should be studied systematically. Last but not least, it is undeniable that the doping of heteroatoms can improve the activity of catalysts. However, most of the studies on heteroatom doping focus on N and S, and few on P, B, F and other heteroatoms. It is worth discussing what role the heteroatoms play and in what ways.