1 Introduction

The ever-increasing worldwide demand for clean energy carriers has resulted in increased attention to a “hydrogen economy” as a possible long-term solution for securing clean and renewable energy [1, 2]. Although a hydrogen economy offers a compelling picture of sustainable energy, significant scientific and technical challenges must first be circumvented to allow this vision to be fully implemented. In this context, one of the key factors dictating the overall efficiency of hydrogen-based energy cycles is the performance of the electrochemical energy conversion devices, for example, a fuel cell or a water electrolyzer [3,4,5,6,7].

As a low-temperature type of fuel cell, polymer electrolyte membrane fuel cells (PEMFCs) are zero-emission energy conversion devices that convert hydrogen fuel directly into electricity with high efficiency. The multiple advantages imparted by PEMFCs render them widely applicable for transportation, mobile, and stationary applications [8, 9]. In general, the performance of PEMFCs depends mainly on the efficiency of electrocatalysts for the oxygen reduction reaction (ORR) at the cathode. As the ORR proceeds via a proton-coupled, four-electron transfer, this reaction is sluggish; the intrinsic kinetics for the ORR are approximately million times slower than the hydrogen oxidation reaction taking place at the anode on a Pt catalyst [10]. To overcome the demanding kinetics of the ORR, high-loading Pt-based catalysts have been commonly employed for the PEMFC cathode [11]. However, Pt-based catalysts pose multiple drawbacks including high costs, the scarcity of Pt, declining activity with long-term operation, and susceptibility to poisoning. Indeed, Pt metal is solely responsible for 40–50% of the total cost of the PEMFC stack, which has limited the widespread application of PEMFC systems [12]. To address this issue, the development of nonprecious metal ORR catalysts has recently received increasing attention, and a diverse class of ORR catalysts, based on metal oxide/carbon composites [13,14,15], metal chalcogenides or metal carbides [16–18], transition metal and nitrogen codoped carbon (M–N/C) [19,20,21,22,23], and heteroatom-doped carbon [24, 25], has been pursued. Among these nonprecious metal catalysts, the M–N/C catalysts are particularly noteworthy, primarily due to their high ORR activities.

The field of heterogeneous M–N/C catalysts has a history dating back more than 50 years. In 1964, Jasinski first demonstrated that cobalt phthalocyanine could catalyze the ORR in alkaline media, opening up the possibility of M–N/C catalysts as potential alternatives to Pt-based catalysts [26]. Since then, several types of metallomacrocyclic compounds, such as metalloporphyrins and metallotetraazaannulenes, have been widely explored as new M–N/C catalysts [27, 28]. However, the ORR activity and durability of these molecular catalysts were significantly lower than those of Pt-based catalysts. In the course of overcoming these issues, the preparative chemistry of M–N/C catalysts underwent several stages of breakthroughs. For example, Jahnke et al. suggested that the high-temperature heat treatment of metallomacrocyclic compounds could significantly improve the activity and durability of M–N/C catalysts [29]. The Yeager group demonstrated that a M–N/C catalyst prepared from a mixture of metal, nitrogen, and carbon precursors exhibited a comparable ORR activity to catalysts derived from expensive metallomacrocyclic compounds, thereby representing a more economical route towards M–N/C catalysts [30]. This method suggested a possibility of combining various precursors for each component, allowing for the more flexible design of M–N/C catalysts. However, despite continued research into the development of high-performance M–N/C catalysts prior to 2008, the ORR activities of M–N/C catalysts remained more than two orders of magnitude lower than those of Pt-based catalysts.

In 2009, a major breakthrough in the field of M–N/C catalysts was made by the Dodelet group [20]. They prepared Fe–N/C catalysts by filling microporous carbon black with ferrous acetate and 1,10-phenanthroline, followed by heat treatment under NH3. The optimized Fe–N/C catalyst achieved a PEMFC volumetric current density of 99 A cm−3 at 0.8 V, which was ~ 35 times higher than that of the previously reported best-performing catalyst. Notably, this performance was close to the target of 130 A cm−3 set by the US Department of Energy (DOE) in 2010. Subsequently, in 2011, the Zelenay group developed a M–N/C catalyst based on Fe, Co, and polyaniline, which achieved a highly promising PEMFC durability up to 700 h at 0.4 V along with a high initial current density [21]. The results obtained by the Dodelet and Zelenay groups suggested the practical viability of M–N/C catalysts for PEMFC applications, and triggered a tremendous surge of research interest in this field. Consequently, significant progress has been made in the design and synthesis of M–N/C catalysts, as well as in the elucidation of the catalyst active sites [31,32,33,34,35,36,37,38,39,40,41,42,43,44,45]. The previous studies on the active sites of M–N/C catalysts provided a plausible evidence that the active sites contain atomically dispersed metal coordinated to N atoms (M–N x sites). We note that the metal nanoparticles (NPs) encapsulated in carbon shells were suggested recently to exhibit a considerable ORR activity. However, the catalytic role of such NPs is still a matter of debate and is not discussed in this review.

The advances achieved in this field have been previously presented in a number of review papers. The developments in M–N/C catalysts prior to 2011 have been summarized in reviews by Zhang et al. [46], Dodelet and Zelenay et al. [47], Palacin et al. [48], and Zhang et al. [49]. In addition, recent reviews have focused on more specific aspects of M–N/C catalysts, such as molecular M–N/C catalysts [50], the active sites of M–N/C catalysts [51, 52], the role of transition metals in M–N/C catalysts [53], the catalytic mechanisms over M–N/C catalysts [54], and characterization of the M–N/C catalysts [55].

In this review, we focus on the synthetic strategies and preparative chemistry toward highly active and durable M–N/C catalysts developed since 2011. Although numerous methods are available that can boost the performance of M–N/C catalysts, we selected five major synthetic routes in this review: (i) the exploitation of metal–organic frameworks (MOFs) as hosts or precursors, (ii) the use of sacrificial templates, (iii) the addition of heteroelements, (iv) the preferential generation of active sites, and (v) biomimetic approaches. We explain the advantages of each strategy and present notable examples with their ORR activities and/or PEMFC performances. We also summarize the catalytic activities and measurement conditions of high-performing M–N/C catalysts for the ORR. For a summary regarding the PEMFC performances of M–N/C catalysts, readers are encouraged to refer the recent review by Shao and Dodelet [56]. Finally, we conclude this review with a summary and some suggestions for future studies.

2 Thermal Conversion of Metal–Organic Frameworks

MOFs are crystalline porous materials composed of inorganic species (metal ions or clusters) that are bridged by organic ligands [57, 58]. MOFs generally exhibit large surface areas up to few thousand m2 g−1, and they can be constructed from a myriad of compositions. Although the use of MOFs themselves as electrocatalysts is limited due to their intrinsically low electrical conductivities, the high-temperature pyrolysis of MOFs can convert the organic ligands into porous carbon to endow conductivity to the resulting materials, thereby enabling their utilization as M–N/C catalysts [59]. In the conversion of MOFs to M–N/C catalysts, MOFs can be utilized as microporous hosts for metal and nitrogen precursors, or as metal and nitrogen precursors.

2.1 MOFs as Hosts

When MOFs are employed as hosts, the metal and nitrogen precursors are initially impregnated into the micropores of the MOFs, and are later transformed into active sites through their reaction within the micropores during subsequent pyrolysis. The most widely used MOF for this purpose is zeolitic imidazolate framework-8 (ZIF-8), in which ZnII centers are connected to the N atoms of imidazolate ligands. During pyrolysis of ZIF-8, Zn is eliminated in situ due to the low boiling point of metallic Zn (i.e., 907 °C), thus yielding a highly porous carbon structure after pyrolysis. Furthermore, the imidazolate ligands can serve as an additional source of nitrogen.

The utilization of ZIF-8 as a microporous host was first demonstrated by Dodelet et al. in the preparation of high-performance Fe–N/C catalysts [60]. In their work, ZIF-8 was initially impregnated with ferrous acetate (Fe precursor) and 1,10-phenanthroline (N and C precursor), dried, ball-milled, and finally pyrolyzed under Ar and then under ammonia gas. The resulting catalyst was composed of nitrogen-containing microporous carbon hosting Fe–N x active sites (Fig. 1a). The membrane electrode assembly (MEA) employing the optimized Fe–N/C catalyst at the cathode exhibited a current density of 1.2–1.3 A cm−2 at 0.6 V with a peak power density of 0.91 W cm−2 in PEMFC. The MEA exhibited the highest volumetric activity (230 A cm−3 at 0.8 V) at that time, which exceeded the US Department of Energy (DOE) 2010 activity target of 130 A cm−3 (Fig. 1b). This volumetric current density was ~ 2.3 times higher than the MEA made by their previously reported catalyst [20], which was prepared according to a similar procedure, but without the use of a microporous ZIF-8 host, thereby highlighting the beneficial role of MOFs in preparing advanced M–N/C catalysts.

Fig. 1
figure 1

Reproduced with permission from [60]. Copyright (2011) Nature Publishing Group (a, b); Reproduced with permission from [64]. Copyright (2015) National Academy of Sciences (c, d)

a TEM image of the optimized Fe–N/C catalyst by using ZIF-8 as the host (scale bar: 50 nm). b Comparison of the iR-free polarization curve of the optimized Fe–N/C catalyst (hollow blue stars) with that of the previously reported best-performing catalyst (hollow red circles) in a H2–O2 PEMFC operated at 80 °C. The solid blue stars and red circles indicate the extrapolated volumetric current densities at 0.8 V for each catalyst. The gray star and circle represent the 2020 target (2015 target at that time) and 2010 target set by the US DOE, respectively. c Scanning electron microscopy image of the nanofibrous Fe–N/C catalyst. d iR-free polarization curve of the catalyst in a H2–O2 PEMFC operated at 80 °C. The red diamonds represent the extrapolated volumetric current density at 0.8 V.

Full size image

In the synthesis of Fe–N/C catalysts using ZIFs as the microporous support, the structural properties of ZIFs have significant influences on the PEMFC performance [61,62,63]. To examine this effect, Jaouen et al. prepared Fe–N/C catalysts using nine Zn-based ZIFs with different topologies, Zn/N/C contents, and textural properties [63]. They found that the PEMFC performances of the catalysts correlated linearly with the cavity size and the specific surface area of the parental ZIF supports. In addition, higher activities originated from larger numbers of Fe–N x active sites, as evidenced by Mössbauer spectroscopy. It was suggested that the reaction of Fe and N precursors is promoted inside larger cavities, leading to the formation of more abundant Fe–N x sites [63].

Liu et al. developed a nanofibrous Fe–N/C catalyst from pyrolysis of electrospun polymer mixture containing polyacrylonitrile (PAN), an Fe(II)-based organometallic precursor, and the ZIF (Fig. 1c) [64]. This catalyst design utilizes both the high active site density achieved by the ZIF host and the excellent electronic conductivity of the PAN-derived graphitic carbon fiber. In addition, the macroporous voids generated between the carbon fiber networks could facilitate mass transport. This catalyst exhibited very high PEMFC performance with the highest reported volumetric activity of 450 A cm−3 at 0.8 V, which far exceeded the 2020 DOE target, i.e., 300 A cm−3 (Fig. 1d). These results suggest that the microporous–macroporous hierarchical structural design is critical to enhancing in the MEA performances of Fe–N/C catalysts [64].

In addition to the Zn-based ZIF family, the Zn/aminoterephthalate-based MOF (IRMOF-3) [65], a magnesium/naphthalenedicarboxylate-based anionic MOF [66], and a copper/benzenetricarboxylate-based MOF (HKUST-1) [67] were also employed as porous hosts for metal and nitrogen precursors. However, the organic ligands of these MOFs contain oxygen instead of nitrogen, and therefore may be inappropriate for preparing highly active M–N/C catalysts. For this reason, N-rich imidazolate-based ZIFs are most commonly employed for the preparation of M–N/C catalysts. For oxygen-ligand-based MOFs, ligands are often aminated to impart N-containing functionality, with the MIL-NH2 series being one such example [68,69,70].

2.2 MOFs as All-in-One Precursors

MOFs composed of ORR-active metals (i.e., Fe and Co) and N-containing ligands can serve as all-in-one precursors for the three components of M–N/C catalysts. In this case, a simple pyrolysis of MOFs without the impregnation and drying steps simplifies the preparation process. Importantly, inherent M–N bonding in the parental MOF can be directly translated into active M–N x species in the final M–N/C catalysts. Due to these advantages, this strategy has emerged as a simple yet efficient methodology for preparing high-performance M–N/C catalysts. In addition, owing to the wide tunability of the metal and the ligands, a variety of MOF precursors have been investigated, including Fe-imidazolate frameworks [71,72,73,74,75,76,77,78,79], Co-imidazolate frameworks [80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96], Prussian blue [97,98,99,100,101], and other families of MOFs [102,103,104,105].

The use of MOFs as all-in-one precursors was first realized by Liu et al., who synthesized Co–N/C catalysts by the simple pyrolysis of a Co-imidazolate framework (Co-ZIF) [80, 81]. This Co-ZIF (ZIF-67) is composed of CoII sites tetrahedrally coordinated to four N atoms of the imidazolate ligands (Fig. 2a). They suggested that the thermal activation of ZIF-67 at 750 °C transformed the original Co–N4 structure into ORR-active Co–N x sites (Fig. 2b). Thermal activation and successive acid-leaching resulted in a Co–N/C catalyst with a half-wave potential of 0.68 V (vs. reversible hydrogen electrode, RHE) in 0.5 M H2SO4 [81]. In a similar strategy, the Liu group prepared a porphyrin-based porous organic polymer (POP) with a three-dimensional (3D) network structure and a high surface area of ~ 2300 m2 g−1 [106]. Heat-treatment of the Fe porphyrin-based POP produced a highly active catalyst both in the half-cell configuration and in the PEMFC. This POP-conversion method has therefore provided active M–N/C catalysts in an analogous way to MOF-based synthetic methods [106,107,108,109].

Fig. 2
figure 2

Reproduced with permission from [81]. Copyright (2011) Wiley-VCH Verlag GmbH & Co. KGaA (a, b); Reproduced with permission from [90]. Copyright (2016) Wiley-VCH Verlag GmbH & Co. KGaA (c, d)

a The tetrahedral Co–N4 unit structure of ZIF-67. b Schematic illustration of the suggested structural evolution of Co–N x active sites during thermal activation of a Co-based ZIF. c Schematic illustration of the preparation of the Co–N/C catalyst (Co SAs/N–C) in the presence of Co–N x sites derived from ZIF-67. d ORR polarization curves of Co SAs/N–C, Co NPs/N–C, and Pt/C in 0.1 M KOH.

Full size image

Recently, the Mai group produced a polyhedral Co–N/C catalyst exhibiting multimodal pores between CNT bundles [96]. During the preparation of this catalyst, the two-step pyrolysis of ZIF-67 was carried out. The first heat-treatment at a relatively low temperature created metallic Co NPs, which catalyzed the formation of the CNT bundles during the second high-temperature pyrolysis step. The final optimized Co–N/C catalyst exhibited a high ORR activity comparable with that of Pt/C in an alkaline electrolyte [96].

In the majority of MOF-derived synthetic routes to M–N/C catalysts, not only catalytically active M–N x species, but high amounts of undesirable large metal-based NPs were also generated. This phenomenon can be attributed to the excess quantities of metals present in MOFs (e.g., ~ 20 wt% Co in ZIF-67), compared to the metallic contents required for preparation of the high-performing M–N/C catalysts (i.e., ~ 0.5–3 wt%). This high density of metal atoms consequently increases the probability of aggregation at elevated temperatures. To address this issue, the Li group introduced Zn2+ ions as “fence” atoms in Co-based ZIFs to expand the adjacent distances of Co atoms, thereby reducing the possibility of Co aggregation [90]. During pyrolysis, the Zn atoms were evaporated, to leave predominantly Co–Nx sites in the pyrolyzed material (Fig. 2c). The resulting active site-rich catalyst exhibited a superior ORR activity to both the Co NP-abundant catalyst and a commercial Pt/C catalyst (Fig. 2d) [90].

Furthermore, Chen et al. prepared Fe–N/C catalysts derived from the thermal conversion of Fe-doped ZIF-8 with various Fe/(Fe + Zn) ratios ranging from 5 to 25 wt% [78]. They found that higher quantities of Fe atoms in the precursor led to the formation of larger quantities of Fe and Fe3C NPs, while the 5 wt%-loaded ZIF-8 produced a catalyst containing only Fe–N x sites. This method greatly improved the ORR activity, with a 130 mV increase in the half-wave potential [78]. The Li group also recently demonstrated that precursor pyrolysis in the presence of a small quantity of Fe (i.e., 0.8 wt% in the precursor) produced exclusively Fe–N x active sites [79]. Upon increasing the Fe loading to 4 wt%, large quantities of Fe and Fe3C NPs were formed, and the ORR activity diminished considerably with a 60 mV negative shift in the half-wave potential [79].

3 Use of Sacrificial Templates

Enlarging the surface area of a catalyst is a direct method to increase the number of active sites on the catalyst surface, and enhance the ORR activity. M–N/C catalysts are typically synthesized via the high-temperature pyrolysis of precursor mixtures, which is critical to creating active M–N x sites as well as to endowing a high electrical conductivity. However, this pyrolysis step commonly results in structurally ill-defined, low surface area catalysts. The use of sacrificial templates can therefore provide a straightforward method for the production of M–N/C catalysts with large surface areas and controlled porosity [110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,38,139]. In addition, well-defined nanostructures can be generated according to the type of template employed. In the template-based method, a precursor in the gas, liquid, or solid phase is initially introduced into the pores or adsorbed onto the surface of a template, and this is followed by carbonization at high temperatures. Finally, the template is removed by an etchant to generate the porous structure. For this purpose, a variety of templates have been examined including silica [110,111,112,113,114,115,116,117,118,119,120,121,122,123,125,125], metal oxides [126,127,128,129,130], nanowires [131,132,133,134,135], and sodium chloride (NaCl) [136,137,138,139].

3.1 Silica Templates

Silica is stable at high temperatures and is easily etched by NaOH or HF. In addition, preparation methods towards various types of silicas, including colloidal silica NPs and mesoporous silica, have been well established. As such, the silica-templating method has been widely investigated for the fabrication of porous M–N/C catalysts. For example, the Atanassov group utilized fumed silica (porous branched silica) as the sacrificial support with a variety of precursors, including macrocyclic compounds [110] and organic molecules [111,112,113] to produce M–N/C catalysts with spherical mesopores of few ten nanometers in diameter. The use of larger templates, such as silica microparticle templates, yielded porous open structured capsule-type M–N/C catalysts with few hundred nanometer pores [114, 115].

Ordered mesoporous silica (OMS) has attracted particular attention because of its periodic, uniform, and tunable pore structure. For example, Joo et al. developed a solid-state hard-templating synthetic method using metalloporphyrins as the all-in-one precursor and OMSs as the templates (Fig. 3a) [117]. In this process, the solid-state mixing of a metalloporphyrin and SBA-15, followed by pyrolysis and silica etching produced in metal-doped ordered mesoporous porphyrinic carbon (M-OMPC, M = Fe, Co, FeCo, etc.). This solid-state infiltration method is simple and rapid compared to the typical wet-impregnation method, as the latter employs an additional drying step that often requires further optimization. Interestingly, the obtained M-OMPC catalysts had hierarchical micropores (~ 1 nm) and mesopores (4–20 nm), and exhibited large surface areas of 1000–1500 m2 g−1, which can expose a high density of catalytically active M–N x sites. Among the family of M-OMPC catalysts, FeCo-OMPC exhibited a particularly high ORR activity in acidic media (0.1 M HClO4) with a half-wave potential at 0.85 V (vs. RHE), which compared favorably to commercial Pt/C catalysts (Fig. 3b). In addition, this solid-state templating method also successfully produced replicated M–N/C structures from KIT-6 (a further type of OMS) and mesocellular-foam silica [117].

Fig. 3
figure 3

Reproduced with permission from [117]. Copyright (2013) Nature Publishing Group (a, b); Reproduced with permission from [118]. Copyright (2013) American Chemical Society (c–f)

a Schematic illustration of the synthesis of M-OMPC catalysts by the solid-state hard-templating method using OMS, and a suggested model of FeCo-OMPC. The yellow, cyan, red, blue, grey, and white spheres represent Co, Fe, O, N, C, and H atoms, respectively. b Polarization curves of FeCo-OMPC and a commercial Pt/C catalyst measured in 0.1 M HClO4. TEM images of Co–N/C catalysts templated from c montmorillonite, d SBA-15, and e silica colloids using vitamin B12 as a precursor. f Linear correlation between the BET surface areas of the templated Co–N/C catalysts and their ORR activities.

Full size image

Similarly, other precursors, such as metal phthalocyanines [120, 123, 124], metal-polyaniline complexes [116, 122], and metal-phenanthroline complexes [121] were successfully employed as precursors for mesoporous M–N/C catalysts. Müllen et al. used vitamin B12 as the Co, N, and C precursor and montmorillonite (a layered clay), SBA-15, and silica colloids as sacrificial templates [118]. The three obtained catalysts exhibited well-developed pore structures, with the silica colloid-based catalyst exhibiting the largest surface area, followed by the SBA-15- and montmorillonite-templated catalysts (Fig. 3c–e). The ORR activities of the catalysts correlated linearly with their Brunauer–Emmett–Teller (BET) surface areas, thereby highlighting the importance of surface area in the electrocatalytic process (Fig. 3f) [118].

3.2 Metal Oxides as the Reactive Templates

Metal oxide nanostructures based on ORR-active metals such as Fe and Co can serve both as sacrificial templates and as a source of the metal species during the preparation of M–N/C catalysts [126,127,128,129,130]. For example, Chen et al. took advantage of Fe(OH)3 as a volatile template containing an Fe source [126]. They combined 2-fluoroaniline and FeCl3 to form a poly(fluoroaniline)-Fe(OH)3 hybrid, and the subsequent pyrolysis of this composite produced a porous Fe–N/C catalyst free from Fe-based particles. They suggested that Fe(OH)3 is dehydrated to form FeO(OH) which subsequently undergoes reductive transformation to Fe2O3 → Fe3O4 → Fe in situ by adjacent carbon atoms at high temperatures. The reduced Fe then reacts with the HCl produced from the Cl− ions bonded to poly(fluoroaniline) to generate FeCl3 that was subsequently sublimated. The resulting catalyst exhibited a half-wave potential of 0.80 V (vs. RHE) in 0.1 M KOH [126]. In addition, Zhang et al. employed a similar approach using porous Fe3O4 microspheres as the template [127]. In this case, the Fe3O4 microspheres were coated with polypyrrole with the aid of cetyltrimethylammonium bromide. Subsequent pyrolysis of the composite at 950 °C produced the desired impurity-free porous Fe–N/C microspheres, which exhibited a high BET surface area (674 m2 g−1) and a half-wave potential of 0.86 V (vs. RHE) in 0.1 M KOH [127].

Recently, Dong et al. prepared a porous cubic graphitic carbon framework using Fe3O4 nanocube superlattice [130]. In this system, pre-synthesized Fe3O4 nanocubes capped with an oleic acid surfactant were self-assembled to form a superlattice structure. During the thermal annealing step, oleic acid was converted into a graphitic carbon framework and the Fe atoms from the Fe3O4 nanocubes diffused into the carbon layer. Subsequent acid-leaching and heat-treatment under NH3 produced the desired Fe- and N-doped carbon framework structure (Fig. 4a). Thus, in this protocol, the Fe3O4 nanocubes served as both the sacrificial template and the Fe source. The resulting Fe–N/C catalyst exhibited a highly ordered porous network with a large BET surface area of 1180 m2 g−1 (Fig. 4b). In addition, the active sites of the catalyst were predominantly Fe–N x (~ 3 wt% Fe), as evidenced by X-ray diffraction (XRD), transmission electron microscopy (TEM), and X-ray absorption spectroscopy (XAS). The high surface area and active site density of this material contributed to an excellent ORR activity with a half-wave potential of 0.883 V in 0.1 M KOH (Fig. 4c) [130].

Fig. 4
figure 4

Reproduced with permission from [130]. Copyright (2017) American Chemical Society

a Schematic illustration of the preparation of Fe- and N-doped carbon frameworks using an Fe3O4 nanocube superlattice as both the template and the Fe precursor. b TEM image of the Fe- and N-doped carbon framework catalyst (Fe–N–SCCFs). c ORR polarization curves for Fe–N–SCCFs, the same catalyst without N, a disordered Fe–N/C catalyst, and a Pt/C catalyst measured in 0.1 M KOH.

Full size image

3.3 Other Sacrificial Templates

The exploration of new types of sacrificial templates has allowed the production of unique nanostructured M–N/C catalysts. For example, tellurium nanowires (Te NWs) have been demonstrated as useful supports for the construction of tubular fiber M–N/C catalysts. Due to the relatively low boiling temperature of Te (~ 450 °C [140]), Te NW templates are removed in situ during the pyrolysis stage without any additional process. The Yu group demonstrated the Te NW-templated synthesis of porous fiber Fe–N/C catalysts [131]. A mixture of glucosamine/ferrous gluconate was hydrothermally treated with Te NWs to polymerize the components onto the surface of the NWs. They emphasized the importance of employing an FeII precursor rather than an FeIII precursor, as the latter rapidly oxidizes Te, leading to the collapse of the fibrous structure of the product [131]. Similarly, Lin et al. developed a Fe-glucosamine-derived nanotube (NT) catalyst with well-dispersed Fe–N x sites [132]. The Manthiram group utilized Te NTs as the sacrificial template and ZIF-8 as the microporous host for the precursors [135]. In this case, ZIF-8 was grown on the porous Te NTs, prior to coating with Fe-polydopamine. Pyrolysis at 950 °C vaporized the majority of Zn present in the ZIF-8 and the Te NTs, leading to a highly porous and tubular Fe–N/C catalyst with a BET surface area of 1380 m2 g−1 (Fig. 5a, b) [135]. This strategy was also demonstrated for Co-doped ZIF precursor to yield Co–N/C catalysts [134].

Fig. 5
figure 5

Reproduced with permission from [135]. Copyright (2017) Wiley-VCH Verlag GmbH & Co. KGaA (a, b); Reproduced with permission from [138]. Copyright (2016) The Royal Society of Chemistry (c); Reproduced with permission from [139]. Copyright (2011) American Chemical Society (d)

a Schematic representation of the synthesis of porous and tubular Fe–N/C catalysts using Te NWs as templates and ZIF-8 as the microporous host, and b TEM image of the obtained catalyst. c Schematic illustration of the NaCl-templated synthesis of Fe–N/C catalyst constructed with CNT–carbon nanosheet hybrids. d Schematic illustration of the preparation of the 2-dimensional Fe–N/C catalyst using Fe phthalocyanine as the precursor and NaCl as the template.

Full size image

NaCl can also function as a structure directing agent, and is particularly useful in the preparation of M–N/C catalysts with two-dimensional (2D) sheet morphologies. This NaCl-based templating method has a number of advantages: (i) the synthesis is scalable and economic, as NaCl is inexpensive and can be easily collected and recycled; (ii) this method allows the formation of Fe-based particles to be suppressed due to the in situ generation and volatilization of FeCl x species in the presence of NaCl; and (iii) template removal is simple and can be carried out in neutral conditions, thereby avoiding degradation of the active sites In this context, Hu et al. synthesized carbon nanosheet-nanotube composite catalysts [138], in which the pyrolysis of an Fe(NO3)3-glucose-melamine mixture in the presence of a NaCl template led to the formation of CNT–carbon nanosheet hybrids (Fig. 5c). The resulting carbon nanosheet-nanotube catalyst contained a larger number of Fe–N x sites and a better-developed pore structure than the untemplated catalyst, thereby resulting in a high ORR activity with a half-wave potential of 0.87 V (vs. RHE) in 0.1 M KOH [138]. Furthermore, Sung et al. employed FeII phthalocyanine as a precursor and NaCl as a template to produce a nanosheet Fe–N/C catalyst with well-dispersed and structurally defined Fe–N4 sites, enabled by low temperature treatment [139]. This was achieved through polymerization in the presence of 10,10ʹ-dibromo-9,9ʹ-bianthryl at 450 °C, and subsequent removal of NaCl by washing with H2O (Fig. 5d) [139]. However, mechanism of the formation of the sheet-like morphology with NaCl-templated catalysts remains unclear, and further studies are required to investigate these points in greater detail.

Alternatively, Wei et al. used NaCl as a “shape fixing” template [136]. They created Fe-polyaniline complex nanostructures and fully sealed the precursor in NaCl via repeated recrystallization. During the pyrolysis stage, NaCl acted as a nanoreactor to prevent the collapse of the initial polymer nanostructures. In addition, gaseous products evolving during pyrolysis step generated the desired porosity, and N-containing gases reacted with Fe to increase the active site density. The use of this trapped synthetic system also reduced weight loss during the heat-treatment, thereby indicating that NaCl also could facilitate carbonization [136].

4 Addition of Heteroelements

Although the origin of synergistic effect is still unclear, heteroatom-doping has been effective to enhance the ORR activity of M–N/C catalysts. It is often the case that a bimetallic M1M2–N/C outperforms each monometallic catalyst [109, 111, 117, 141,142,143,145,145]. The other promotion effect could be found when sulfur (S) is doped into M–N/C catalysts [146,147,148,149,150,151,152,153,154,155]. This section introduces representative recent works that demonstrated the activity improvement by the addition of secondary metal or S.

4.1 Addition of Secondary Transition Metal

In the context of bimetallic synergy, Atanassov et al. investigated the impact of secondary metal for FeCo–, FeCu–, FeNi–, and FeMn–N/C catalysts [111]. They found that all catalysts possessed similar surface areas and pore structures, and that synergistic increase in the ORR activity was only shown for the FeMn–N/C catalyst (Fig. 6a). The other bimetallic catalysts exhibited similar or less active ORR activity than its monometallic counterpart, whereas the 4-electron selectivity was always enhanced in bimetallic systems [111]. In the case of M-OMPC catalysts (introduced in “Sect. 3.1”), FeCo-OMPC exhibited an improved ORR activity compared to both Fe- and Co-OMPC in acidic media [117]. Müllen et al. also observed a Fe–Co promotion effect when the catalysts were prepared by the pyrolysis of polymeric porphyrin [141]. In their catalysts, although the active site densities in Fe–N/C and FeCo–N/C were comparable, the FeCo–N/C catalyst exhibited a 40 mV positive half-wave potential for the ORR in acidic media (Fig. 6b) [141]. In contrast to the observations made by the Atanassov group, this bimetallic catalyst was less selective to the 4-electron ORR than the Fe–N/C catalyst. We note that the co-existence of Fe and Co in M–N/C catalyst does not always enhance the ORR activity [156].

Fig. 6
figure 6

Reproduced with permission from [111]. Copyright (2012) Elsevier (a); Reproduced with permission from [141]. Copyright (2015) The Royal Society of Chemistry (b)

a ORR polarization curves of monometallic and bimetallic (Fe, Mn)–N/C catalysts derived from 2-aminoantipyrine, measured in 0.5 M H2SO4. b ORR polarization curves of monometallic and bimetallic (Fe, Co)–N/C catalysts obtained by pyrolyzing porphyrin polymers measured in 0.5 M H2SO4.

Full size image

To date, the bimetallic synergy effect has not yet been clearly understood. This can be attributed to the heterogeneity of the active sites in M–N/C catalysts, which are formed mainly through high-temperature pyrolysis. Nevertheless, a number of explanations have been tentatively suggested, including a change in the catalyst structures during pyrolysis caused by the presence of a secondary metal, and the tuned electronic and local geometric structures of the active sites. In this regard, Strasser et al. prepared a polyaniline-based FeMn–N/C catalyst [143], which exhibited an enhanced ORR activity compared to its respective monometallic catalysts in an alkaline electrolyte, but a poorer activity compared to the Fe–N/C catalyst in an acidic electrolyte. They also conducted CO chemisorption experiments at low temperatures as well as temperature-programmed desorption (TPD) experiments, to quantify the active site density and to investigate the interactions between the M–N x sites and CO molecules. The observed peak shift in the TPD profile suggested a stronger binding between CO molecules and the Fe–N x sites in FeMn–N/C than those in the Fe–N/C. Given Mössbauer spectroscopy evidenced no alloy-like interactions between Fe and Mn, the TPD result suggested structural modification of the Fe–N x active sites by addition of Mn. This modification led to 20 and 150% higher turnover frequencies (TOFs) of the bimetallic catalyst compared to the Fe–N/C catalyst in alkaline and acidic media, respectively [143].

4.2 Addition of Sulfur

The addition of sulfur has also been proven effective in enhancing the ORR activity of M–N/C catalysts, with the majority of S-promoting effects being demonstrated for Fe–N/C catalysts. The most prominent studies were performed by Herrmann and Kramm et al., who systematically investigated the role of elemental S as an additive during the pyrolysis of CoTMPP and FeTMPPCl with Fe oxalate (a pore-forming agent) [146,147,148]. They revealed that the addition of S exerts a number of positive effects: (i) the favorable formation of acid-leachable FeS x species instead of insoluble FeC x @C species; (ii) the generation of an amorphous carbon matrix (i.e., higher catalyst surface area) with extended graphene layers; and (iii) a change in the electronic structure of the active Fe–N x sites.

The sulfur promotion effect was also demonstrated with the Zelenay group’s polyaniline-based Fe–N/C catalyst, for which ammonium peroxydisulfate, (NH4)2S2O8, was added to oxidatively polymerize the aniline monomers [149]. The ORR activity was improved upon the addition of an optimal amount of (NH4)2S2O8. For S-containing catalysts, lower quantities of FeC x and higher quantities of porphyrin type Fe–N4 active sites were detected by XAS than S-free catalysts [149]. In addition, Cho et al. explained the enhanced ORR activity of S-doped Fe–N/C catalysts by the lowered work function, which facilitates electron transfer from the catalyst to O2 [155, 157].

The selection of suitable precursors containing both S and N atoms could result in high-performance S-doped Fe–N/C catalysts. This enables more intimate contact between the Fe, N, and S atoms, and therefore maximizes the S-promotion effect compared to when the S source is separately mixed. For example, He and Liu et al. designed a thiolated polyacrylonitrile telomer; that is fluidic at room temperature, thereby enabling good contact with the Fe ions [150]. S-containing catalyst exhibited an improved ORR activity than S-free catalyst. They revealed the presence of FeN x C x species without any Fe–S species using time-of-flight secondary ion mass spectrometry, indicating the S-promotion effect did not originated from Fe–S bond [150]. In addition, the Strasser group used a S- and N-containing ionic liquid as a precursor for S-doped Fe–N/C catalysts [151]. They systematically investigated the role of other heteroatoms (i.e., phosphorus and boron) in the ORR activity of doped Fe–N/C catalysts, but found that only S had a positive influence on the activity (Fig. 7a), resulting a high ORR activity with a half-wave potential of 0.88 V (vs. RHE) in 0.1 M KOH, which is greater than a commercial Pt/C catalyst [151].

Fig. 7
figure 7

Reproduced with permission from [151]. Copyright (2014) American Chemical Society (a); Reproduced with permission from [152]. Copyright (2015) Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim (b)

a ORR polarization curves of the ionic liquid-derived heteroatom-doped Fe–N/C catalysts measured in 0.1 M KOH. b ORR polarization curves of the Fe–N/C catalysts prepared using FeIII thiocyanate (denoted as Fe/N/C–SCN) and FeIII chloride (denoted as Fe/N/C–Cl) measured in 0.1 M H2SO4.

Full size image

Finally, Sun et al. investigated the promotion effect of S-doping when the chloride ions in the FeCl3 precursor were replaced by the thiocyanate (SCN−) anion [152]. In this case, an excellent ORR activity was achieved in acidic media with a half-wave potential at 0.84 V (vs. RHE) (Fig. 7b) in addition to the highest reported PEMFC performance in terms of the maximum power density (1.03 W cm−2) [152].

5 Preferential Formation of M–N x Active Sites

Since the beneficial effects of high-temperature heat treatments on the ORR activities and the stabilities of M–N/C catalysts were demonstrated [29, 30], the pyrolysis step has been considered crucial and the majority of M–N/C catalysts are now prepared by the pyrolysis of macrocycles or precursor mixtures. During the pyrolysis step, catalytically active M–N x sites dispersed on a carbon support are generated, thereby endowing catalytic activity towards the ORR. However, high-temperature heat treatment also induces the aggregation of metal atoms into less active large metal and/or metal carbide NPs, leading to the loss of M–N x active sites in the resulting catalysts and an accompanying loss in ORR activity. As such phenomena are generally uncontrollable, the majority of heat-treated products possess a significant portion of inactive NPs. Furthermore, these NPs catalyze the Fischer–Tropsch reaction in situ at high temperatures, and so become coated with graphitic carbon shell that prevents the NPs from being acid-leached [17, 18].

In order to remove NP-based impurities, several protocols based on multiple heat treatments (sometimes under an NH3 atmosphere) or repeated acid-/heat-treatments have been developed [20, 21]. In this regard, Kramm et al. developed a novel purification method based on secondary heat-treatment under 10% H2/N2 gas (forming gas) followed by acid-leaching (Fig. 8a) [158]. As confirmed by Mössbauer spectroscopy, these purified Fe–N/C catalysts contained no particulate species and consisted exclusively of active Fe–N x sites (Fig. 8b, c). This method was therefore suitable for the preparation of Fe–N/C catalysts with a high density of active sites (i.e., over 3 wt%). In addition, the ORR activity of the purified catalyst was enhanced 3–10 times in terms of the mass activity in acidic media, when compared to the catalyst that had not undergone secondary heat treatment (Fig. 8d). This strategy was found to be widely applicable to other metal-based M–N/C and bimetallic FeM–N/C catalysts (M = Co, Sn) (Fig. 8e–g) [158].

Fig. 8
figure 8

Reproduced with permission from [158]. Copyright (2016) American Chemical Society

a Schematic illustration of the catalyst purification process through heat-treatment under 10% H2/N2 (forming gas). Mössbauer spectra of the Fe–N/C catalysts, b before and c after purification. Signals D1–D3 represent structurally distinctive Fe–N4 sites, while the Sing signal is attributed to the presence of superparamagnetic alpha-Fe. ORR polarization curves of the d Fe–N/C, e FeSn–N/C, f FeCo–N/C, and g CoSn–N/C catalysts before and after purification measured in 0.5 M H2SO4.

Full size image

As an alternative strategy to preferentially generate M–N x active sites, Joo et al. developed the “silica-protectivel-layer-assisited” method to preserve Fe–N x sites during high-temperature pyrolysis [159]. For this process, 5,10,15,20-tetrakis(4-methoxyphenyl)porphine iron(III) chloride (FeIITMPPCl) was mixed with CNTs and annealed at 400 °C to allow the adsorption of FeTMPPCl on the CNT surfaces. A silica layer was then overcoated onto the CNT/FeTMPPCl composite using tetraethyl orthosilicate as the silica source. Pyrolysis and subsequent silica removal resulted in a catalyst composed of a CNT coated with a thin porphyrinic carbon layer (CNT/PC) (Fig. 9a). The silica-coated CNT/PC was composed primarily of active Fe–N x sites, whereas the catalyst prepared without a silica overcoating (i.e., CNT/PC_w/o SiO2) exhibited large Fe and Fe3C NPs (Fig. 9b, c). Mössbauer spectroscopy and XAS revealed that the ratios of Fe–N x sites to Fe/Fe3C species were approximately 3:1 and 1:1 for the CNT/PC and CNT/PC_w/o SiO2 catalysts, respectively. In addition, temperature-controlled in situ XAS suggested the presence of interactions between the silica layer and the Fe–N4 sites of FeTMPPCl, which could stabilize the Fe–N4 sites and suppress the formation of inactive Fe-based particles. The resulting CNT/PC catalyst exhibited high ORR activities with half-wave potentials of 0.88 and 0.79 V in 0.1 M KOH and 0.1 M HClO4, respectively. CNT/PC catalyst prepared with the silica coating strategy exhibited 3–4 times higher ORR activities than the catalyst prepared with silica coating step both in acidic and alkaline electrolytes (Fig. 9d, e). When CNT/PC was employed as the cathode catalyst in MEA for an alkaline anion exchange membrane fuel cell (AEMFC), very high current density at 0.6 V and peak power density (i.e., 498 mA cm−2 and 380 mW cm−2) were achieved, which were greater than those of previously reported AEMFC MEAs based on non-precious metal catalysts. The CNT/PC-based MEA also exhibited an excellent PEMFC performance with a volumetric activity of 320 A cm−3, which compared favorably to the 2020 US DOE target of 300 A cm−3 [159]. Similarly, Zhang et al. prepared mesoporous silica-coated Co-doped ZIF-8 to prevent particle aggregation during pyrolysis [160]. In this case, the mesoporous silica shell performed dual roles during pyrolysis of the MOF precursor, i.e., prevention of Co-based particle formation and MOF particle fusion, leading to an increase in the BET surface area by a factor of two. These effects resulted in an enhancement in the ORR activity by approximately 6 times [160].

Fig. 9
figure 9

Reproduced with permission from [159]. Copyright (2016) American Chemical Society

a Schematic illustration of the “silica-protective-layer-assisted” strategy to preferentially generate Fe–N x active sites while suppressing the formation of Fe-based less-active particles. CNT/PC and CNT/PC_w/o SiO2 denote the silica-protected and unprotected catalysts, respectively. TEM images of b CNT/PC and c CNT/PC_w/o SiO2. ORR polarization curves of CNT/PC and CNT/PC_w/o SiO2 measured in d 0.1 M KOH and e 0.1 M HClO4. The insets show the kinetic current densities of the catalysts at 0.9 V (vs. RHE) in 0.1 M KOH and 0.8 V (vs. RHE) in 0.1 M HClO4.

Full size image

6 Non-pyrolyzed M–N/C Catalysts with Well-Defined Active Site Structures

The naturally occurring enzyme cytochrome c oxidase (CcO) is a key component in the respiratory cycle of organisms, selectively reducing oxygen molecules to H2O without producing reactive oxygen species such as H2O2. The structure of CcO consists of a bimetallic center comprising an iron porphyrin with an axially coordinated histidine and a trihistidine-coordinated Cu in the distal pocket (Fig. 10a) [161]. Inspired by the elaborate structural features of CcO, significant of effort has been made to synthesize CcO-mimicking molecules, with the most successful example being reported by the Collman group (Fig. 10b) [161]. Although such biomimetic molecules are useful in understanding the functions of CcO under physiologically relevant conditions (e.g., at pH ~ 7), they are unsuitable for PEMFC applications due to their instability under the harsh operating conditions required for PEMFCs, such as very low pH and working temperatures around 80 °C.

Fig. 10
figure 10

Reproduced with permission from [161]. Copyright (2007) American Association for the Advancement of Science

a The active site of cytochrome c oxidase (CcO). The red, green, black, blue, and white spheres represent the Fe, Cu, N, O, and C atoms, respectively. b Synthetic analogue that has a similar structure and function to CcO.

Full size image

One would therefore expect that incorporation of the CcO-like structural moiety into heterogeneous electrocatalysts could in principle significantly boost their ORR activities. However, the high-temperature pyrolysis step commonly required for preparing heterogeneous ORR electrocatalysts would unavoidably cause the destruction and distortion of the M–N x sites [162,163,164]. Thus, as a compromise between molecular and heterogeneous catalysts, namely heterogenized molecular catalysts, have been developed by immobilizing CcO-mimicking molecules onto a conductive carbon support under ambient conditions, which can improve the stability of this moiety while preserving its catalytic functions.

A straightforward approach to the preparation of such catalysts involves the use of square planar M–N4 structures, such as molecular phthalocyanines or porphyrins, which can be easily stabilized on carbon supports through π–π interactions [165,166,167,168,169,170]. For example, the Campidelli group reported the preparation of CNT–Co porphyrin layer core–sheath nanostructures [169]. They initially physisorbed designed Co porphyrin molecules onto CNT surfaces, then carried out polymerization to produce a porphyrin network layer on the CNTs (Fig. 11a). The resulting core–sheath catalyst exhibited an enhanced ORR activity compared to the catalyst bearing the physisorbed Co porphyrin. In this catalyst, the bifacial structure between the stacked layers of the Co porphyrin was suggested to be responsible for the near 4-electron selectivity, which is similar to the bimetallic interplay observed in CcO [161], in binuclear Cu-triazole–dipyridine complexes [170], and in bifacial Co-based porphyrins [171]. In addition, interaction of the Co-porphyrin units with the CNTs produced the catalysts with high stabilities in acidic media, as confirmed by a minimal decrease in their ORR activities under such conditions [169].

Fig. 11
figure 11

Reproduced with permission from [169]. Copyright (2014) American Chemical Society (a, b); Reproduced with permission from [173]. Copyright (2014) Wiley-VCH Verlag GmbH & Co. KGaA (c–e)

a Schematic illustration of the polymeric Co porphyrin layers stacked on CNTs (MWNT–CoP). b ORR polarization curves of MWNT–CoP (red curve) and MWNT/CoP (black curve, physical mixture) measured in 0.5 M H2SO4. c Schematic representation of the heme-like active sites grafted onto a CNT through a coordination bond to the imidazolic N atom [(DFTPP)Fe-Im-CNTs]. d ORR polarization curves measured in 0.1 M HClO4 and e electron transfer numbers during the ORR for (DFTPP)Fe-Im-CNTs and a simple physical mixture [(DFTPP)Fe-CNTs], measured using rotating ring disk electrode (RRDE).

Full size image

Cho et al. developed a Fe–N/C catalyst with a biomimetic active center [172]. They covalently attached 4-aminopyridine onto a CNT surface, and anchored FeII phthalocyanine (FePc). The resulting heterogenized molecular catalyst exhibited a high activity, with a half-wave potential of 0.92 V (vs. RHE) and near 4-electron selectivity, which was superior to that of Pt/C in 0.1 M KOH. In this case, the strong coordination bond between the FePc molecule and pyridine formed a stable Fe–N4–Naxial active site structure, which is highly analogous to the structure of CcO. They suggested that Fe 3d rehybridization with the axial ligand orbital significantly modified the electronic structure of the Fe–N5 sites compared to the square planar Fe–N4 sites, thereby leading to increased ORR activity through the facile cleavage of the O–O bond, as supported by DFT calculations [172].

Liu et al. also designed a bio-mimetic Fe–N5 active site on a CNT surface [173], where they used an Fe porphyrin and an imidazolate axial ligand to enhance O2 activation (Fig. 11c) [174, 175]. This catalyst exhibited the highest ORR activity among the previously reported non-pyrolytic M–N/C catalysts, with a half-wave potential of 0.88 V (vs. RHE) in an acidic solution (Fig. 11d). In addition, it was highly selective towards the 4-electron ORR (> 99.5% over the full potential range investigated, Fig. 11e) [173]. Although the Cho and the Liu groups demonstrated the high-performance of M–N/C catalysts through the use of heterogenized molecular catalysts, there is much room for improvement in the ORR activity/selectivity by tuning of the macrocyclic molecules, ligand structures, and ligand-carbon support interactions.

The concept of “heterogenized molecular catalysts” based on low-temperature syntheses enables the identification of new active site structures for the ORR. Joo and Park et al. exploited the assembly of an archetypical organometallic compound, CoII acetylacetonate [Co(acac)2], with N-doped graphene to prepare a Co–N/C catalyst [176]. Simple stirring of a mixture of ammonia-reduced graphene oxide (A–rG–O) with Co(acac)2 resulted in the formation of a hybrid catalyst (CoII–A–rG–O). Solid-state nuclear magnetic resonance (SSNMR) measurements revealed that CoII–A–rG–O contained alkyl groups, which were not detected for A–rG–O support, implying that the signal corresponding to the alkyl groups originated from the methyl groups of the Co(acac)2 sites. This SSNMR result was supported by XAS observations, where the Co atoms in CoII–A–rG–O exhibited similar spectral features to Co(acac)2, with the exception of slightly elongated interatomic distances. In addition, density functional theory calculations suggested that the square planar Co(acac)2 molecule has the highest affinity for bonding with the pyridinic N atoms in N-doped graphene. Based on these analyses, a novel type of active site was identified, namely the Co–O4–N site, where Co ions are coordinated to four equatorial oxygen atoms from two acac ligands and to an axial N atom from A–rG–O (Fig. 12a). The creation of this active site in CoII–A–rG–O increased the onset potential by 90 mV compared to A–rG–O (Fig. 12b). In contrast, the increase in the ORR activity was negligible when Co(acac)2 was added to N-free reflux-reduced graphene oxide (Re–G–O), highlighting the significance of Co–N coordination in the ORR. It was deduced that the Co–O4–N site was responsible for promoting the 4-electron ORR, while the Co–O4–O and Co–O4–Ph (Ph = phenyl ring) species played an auxiliary role in peroxide reduction (Fig. 12d) [176]. It is noteworthy that the pentacoordinated Co–O4–N sites resemble the geometric local structure of CcO.

Fig. 12
figure 12

Reproduced with permission from [176]. Copyright (2015) Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

a Schematic description of ORR electrocatalysis on a Co–O4–N active site. Polarization curves of b A–rG–O, CoII–A–rG–O, and Pt/C and c Re–G–O, CoII–Re–G–O, and Pt/C measured in 0.1 M KOH. d Electron transfer number for A–rG–O, CoII–A–rG–O, Re–G–O, and CoII–Re–G–O measured using the RRDE technique.

Full size image

7 Summary of the ORR Activities of Reported M–N/C Catalysts

The ORR activities of selected M–N/C catalysts measured in acidic and alkaline electrolytes are summarized in Tables 1 and 2, respectively. These tables provide the ORR activity parameters [i.e., half-wave potential, kinetic current density (j k ), and mass activity (j m )], in addition to some selected experimental conditions, such as the electrolyte, catalyst loading, and type of counter electrode employed, all of which have a significant influence on the ORR activity.

Table 1 Summary of the experimental conditions and the ORR activity parameters of reported M–N/C catalysts in acidic media
Full size table
Table 2 Summary of the experimental conditions and the ORR activity parameters of reported M–N/C catalysts in alkaline media
Full size table

It should be noted that the ORR activities of reported M–N/C catalysts are generally high in alkaline media, with the majority of catalysts exhibiting half-wave potentials > 0.8 V, and on average ~ 0.85 V (vs. RHE). In acidic electrolytes, however, the catalysts show ~ 100 mV lower half-wave potentials (Fig. 13). Recently, it was found that this decline in the ORR activities of these catalysts in acidic media coincides with the protonation of pyridinic N sites, which may provide a clue to such a discrepancy [187]. However, the mechanism responsible for the differences in ORR activities between acidic and alkaline media remains unclear, and requires further systematic studies.

Fig. 13
figure 13

Histogram of the half-wave potentials of previously reported M–N/C catalysts measured in acidic (indicated as red bars) and in alkaline media (indicated as blue bars). Detailed ORR activity parameters of the examined catalysts are summarized in Tables 1 and 2

Full size image

Finally, we note that the majority of research groups (> 60% of the literature examined) employ Pt-based counter electrodes in their studies. However, the platinum is prone to be anodically dissolved, and can be redeposited on the nonprecious catalyst layer, leading to an enhancement in the ORR activity. Such an effect would become more apparent in long-term durability/stability tests. Indeed, the activation behavior caused by the Pt dissolution–redeposition process has recently been investigated in the context of electrocatalytic hydrogen evolution [188, 189]. The accidental incorporation of Pt prevents a fair comparison of the ORR activity, and therefore the use of carbon-based materials as the counter electrode is highly recommended.

8 Conclusions and Outlook

This review presented the recently emerged synthetic strategies toward high-performance M–N/C catalysts for the ORR, and summarized the ORR activities of highly active M–N/C catalysts. While many methods have been developed, we identified five major strategies for enhancing the catalytic activity for the ORR: the exploitation of MOFs as hosts or precursors, the use of sacrificial templates, the addition of heteroelements, the preferential generation of active sites, and a biomimetic approach. We summarized the advantages of each preparation method for M–N/C catalysts and presented representative examples for each strategy.

The advances in the preparation of M–N/C catalysts, along with the progress in the identification of their active sites for the ORR, indeed led to highly active catalysts, with some catalysts showing very high ORR activities that can rival that of a Pt/C catalyst. However, such performances are in most cases attained with the rotating disk electrode (RDE) measurements in a half-cell configuration [190]. Compared to the RDE measurements, only limited examples of MEA-based single cell performances have been reported. Furthermore, it is often the case that excellent RDE performances of M–N/C catalysts are not fully translated in MEA measurements. Hence, to realize M–N/C catalysts as a true alternative to the state-of-the-art Pt/C catalyst, the performance improvement of MEAs based on M–N/C catalysts is essential. In this sense, very recent result of the Zelenay group [45], who demonstrated highly promising MEA performances for H2-air PEMFC employing a newly developed Fe–N/C catalyst, sheds light on the practicality of this class of catalysts into the fuel cell market.