Oxygen Reduction Reactions of Fe-N-C Catalysts: Current Status and the Way Forward

Abstract

Currently, Fe-N-C materials are considered to be among the most important oxygen reduction reaction (ORR) catalysts, because they are potential substitutes for Pt-based catalysts and are therefore promising in the development of non-noble metal-based catalysts. However, challenges such as electron transfer kinetics still exist and need to be improved upon. From a chemical stand point, improvements can be made through the better understanding of mechanisms in Fe-N-C-based ORR catalysis along with a deeper understanding of the chemical origin of active sites on Fe-N-C catalyst surfaces. Based on these, this comprehensive review will focus on the energy conversion, transformation kinetics and electron transfer of the ORR process as catalyzed by Fe-N-C catalysts. And by taking these and other relevant analytical results for Fe-N-C materials into consideration, primary strategies in the improvement in Fe-N-C catalyst activity will be presented.

Graphical Abstract

As the promising Pt substrate for oxygen reduction catalysis, the Fe-N-C materials are active toward the four-electron reduction of O₂ to H₂O. This review focuses on the profound understanding of heterogeneous oxygen reduction reaction on Fe-N-C materials from the following aspects: (1) thermodynamics of energy conversion in ORR processes, (2) kinetics of ORR processes based on Fe-N-C catalysts, (3) the textural features of Fe-N-C and analytic results known as far, (4) fundamental principle for Fe-N-C materials synthesis and (5) practical application for fuel cell and metal–air batteries.

1 Introduction

Environmentally friendly energy storage through the reversible conversion of chemical energy to electrical energy through devices such as fuel cells and metal–air batteries is vital to the abatement of pollution and greenhouse effects in which a shift away from carbon and fossil fuel (coal, oil and gas)-driven economies is required globally [1]. However, a major bottleneck in the commercialization of these devices is the requirement for oxygen reduction reactions (ORR) in their operation, in which high overpotential for ORR reactions due to sluggish kinetics at the cathode exists [2]. Here, platinum (Pt) and its alloys, possessing appropriate oxygen binding energies (Fig. 1a), have been demonstrated to enhance ORR performances [3, 4]. However, the use of Pt possesses substantial economic challenges due to its scarcity and cost (Fig. 1b), making Pt-based applications significantly more costly than earth-abundant first-row (3d) transition metals (Fe, Co, Ni). There is an additional problem, which has to do with the surface chemistry of Pt. Catalysts based on Pt tend to gets deactivated rather quickly due to the inevitable phenomena of methanol crossover (the diffusion of the methanol fuel from the anode to the cathode) [5]. In fact, (1) methanol electro-oxidation of the crossed-over methanol at the cathode and (2) the poisoning effect by CO due to incomplete methanol oxidation are two factors that would reduce the efficiency of ORR reactions. This is known to degrade the performance of the fuel cell in which Pt catalysts are employed. Therefore, significant efforts have been devoted to the identification of cost-effective and methanol/CO-tolerant non-noble metal electrocatalysts, including transition metal and nitrogen co-doped carbon-based catalysts (M-N-C) [6, 7], as well as metal-free heteroatom-doped carbon catalysts [8,9,10].

Fig. 1

(Part a adapted with permission from Ref. [3]; Copyright 2004 American Chemical Society. Part b adapted with permission from Ref. [11]; Copyright 2012 Royal Society of Chemistry.)

a Volcano curve of activity as a function of oxygen binding energy in which Pt is the most active metal for ORR. b Annual production-dependent market prices of VIII metals

Owing to the unique structural and surface stability of Fe-N-C catalysts and their robustness against methanol/CO poisoning [12], these materials have garnered the attention of researchers in ORR applications and there has been an exponential increase in the articles being published on Fe-N-C materials as ORR catalysts (Fig. 2a) based on publication and citation statistics from the Web of Science, along with an exponentially growing citation trend (Fig. 2b), suggesting that this research area is progressively growing. One may note that from the stand point of a larger impact, Fe-N-C ORR materials are also relevant to other areas, such as selective oxidation reaction [13,14,15,16], CO₂ reduction [17, 18] and hydrogenation [19, 20]. Therefore, a timely comprehensive review on this rapidly growing category of Fe-N-C for ORR catalysis is of great significance to a vast body of readers.

Fig. 2

Articles based on Fe-N-C catalysts for oxygen reduction reactions: a histogram of the number of articles and b citations in recent years. Statistics data are based on indexing keywords of oxygen reduction reaction and Fe-N-C on the Web of Science

Till now, several review articles [2, 21,22,23,24,25,26], related or referred to the preparation and electrochemical performance of Fe-N-C materials for ORR catalysis, have been published. A comprehensive review specifically focusing on the structure of Fe-N-C and the activity determined factor as principle for this kind catalysts’ design are definitely necessary to facilitate future research and development in this area. In this review, we engage to the profound understanding of heterogeneous oxygen reduction reaction on Fe-N-C materials from the following aspects: (1) the thermodynamics of energy conversion in the ORR process, (2) the kinetics of the ORR process based on Fe-N-C catalysts, (3) the textural features of Fe-N-C and analytic results known as far, (4) the fundamental principles for Fe-N-C materials synthesis and (5) the practical applications in fuel cells and metal-air batteries. In each of these cases, open challenges and the way forward will be described.

2 Thermodynamic Description of the ORR Process

Thermodynamics, the study of energy transformation, enables the quantitative discussion and prediction of chemical reaction energetics [27]. In fuel cells, the high efficiency of energy conversion (i.e., from chemical to electrical) in terms of the theoretical thermodynamics makes it be a promising technique for an energy-sufficient in future.

For example, in hydrogen fuel cells:

$$ {\text{H}}_{2} + 1/2{\text{O}}_{2} \to {\text{H}}_{2} {\text{O}}_{(\text{l})} $$

(1)

$$ \Delta G_{\text{r}}^{\ominus} = G_{{{\text{H}}_{2} {\text{O}}\left( {\text{l}} \right)}}^{\ominus} - G_{{{\text{H}}_{2} }}^{\ominus} - \frac{1}{2}G_{{{\text{O}}_{2} }}^{\ominus} $$

(2)

Here the product (water) is in liquid form. At standard conditions of 25 °C (298 K) and 1 atm, the thermal energy (ΔH) of the hydrogen/oxygen reaction is 285.8 kJ mol⁻¹, and the free energy available for useful work is 237.1 kJ mol⁻¹. Based on this, the thermal efficiency of ideal fuel cells operating reversibly on pure hydrogen and oxygen at standard conditions is:

$$ \eta_{\text{ideal}} = \frac{\Delta G}{\Delta H} = \frac{237.1}{285.8} = 0.83 $$

(3)

Then, an actual cell thermal efficiency is:

$$ \eta = \frac{{{\text{Useful}}\;{\text{Energy}}}}{\Delta H} = \frac{{{\text{Useful}}\;{\text{Energy}}}}{\Delta G/0.83} = \frac{{{\text{Volts}}_{\text{actual}} \times \,{\text{Current}}}}{{{\text{Volts}} _{\text{ideal}} \times \;{\text{Current}}/0.83}} = \frac{{0.83E_{\text{actual}} }}{{E_{{i{\text{deal}}}} }} $$

At standard state, in which E_ideal = 1.23 V, and hence:

$$ \eta = 0.675E_{\text{actual}} = \, 0.675\left( {E_{\text{cathode}} - E_{\text{anode}} } \right) $$

(4)

In a fuel cell, oxygen (O₂) is electrochemically reduced at the cathode through the mediation of ORR. Taking the overpotential (E_op) into consideration for a real fuel cell, the actual cell efficiency is found to be:

$$ \eta = 0.675\left[ {\left( {E_{\text{ORR}} - E_{\text{op}} } \right) - E_{\text{anode}} } \right] $$

(5)

Therefore, a high E_ORR with low E_op is anticipated, where the E_ORR depends on the process of electron transfer pathway and the E_op is up to the cathodic electrode materials and the electrolyte.

The complete electrochemical ORR involves four net coupled proton and electron transfers. And in general, the mechanisms of ORR follow two pathways (Fig. 3), i.e., the (1) direct 4-electron reduction pathway from O₂ to H₂O (in acidic media) or to OH⁻ (in alkaline media); (2) 2-electron pathway which involves the formation of an intermediate of hydrogen peroxide molecule, depending on the electrode material chosen and pH of the electrolyte [3, 28]. Thermodynamically, the E_ORR for the 4e⁻ reduction pathway is 1.23 V, and the 2e⁻ pathway is 0.68 V, meaning that the 4e⁻ pathway is more efficient and hence desirable.

Fig. 3

ORR pathways in acidic and alkaline solution, in which a 4-electron reduction pathway with high potential is desirable

3 Kinetic Mechanisms of ORR Processes

A catalyst is defined by the IUPAC as a substance that can increase the rate of a reaction without modifying the overall standard Gibbs energy change occurring through the process. That is, the kinetic mechanism of reaction can be changed by a catalyst, and doing so would not affect the thermodynamics at all. As a promising non-precious metal catalysis, studying the kinetics of ORR on Fe-N-C is very useful in understanding the mechanism of ORR and for further improving the catalytic performance. For electrochemical catalysis, the kinetic mechanisms of substance transformation and electron transfer can in fact be investigated individually.

3.1 Substance Transformation

Typically, in acidic solutions, the substance transformation of ORR is best described by using the Damjanovic model [30] in Fig. 4a. The ORR is considered to follow one of the two different reduction pathways; at times both of them can occur during the reduction of O₂ to water. Taking advantage of Damjanovic model and using the method developed by Hsueh et al., the Ohsaka group [31, 32] calculated and compared the reaction rate constants of k₁, k₂ and k₃ by hydrodynamic (rotating ring–disk electrode) voltammetry. It is found that the Fe-N-C formed by Fe introduction is active toward the four-electron reduction of O₂ and small amounts of Fe species leads to a rather large increase in k₁ than in k₂ [32]. A suitably chosen cationic surfactant can modulate the ORR selectivity on a Fe-N-C catalyst by kinetic promotion through Coulombic interaction with the peroxo-Fe complexes [33]. As the first step during ORR, the adsorption and activation of O₂ on catalysts involves associative or dissociative pathways, or a combination of these. Using in situ IR spectroscopy, the associative pathway has been demonstrated. The primary signature of this is the observation of surface-adsorbed superoxide (OOH_ad) and hydroperoxide (HOOH_ad). It may be noted that the breaking of O–O bond only occurs at the peroxo level [34]. The Ohsaka group’s work has addressed the adsorption equilibrium between the adsorbed and desorbed H₂O₂ near the disk surface. And a Wroblowa model (Fig. 4b) is employed to distinct the parallel-sequential path and the sequential paths of the ORR on Fe-N-C materials. The compared results (Fig. 4c) suggest that the Damjanovic model always overestimates the k₁ values and underestimates the k₂ values [29].

Fig. 4

a Damjanovic and b Wroblowa model for ORR. c Variations of the rate constants k₁ and k₂ on Fe-N-C catalysts using Damjanovic (solid lines) and Wroblowa models (dotted lines). (Part a, b and c adapted with permission from Ref. [29]; Copyright 2015 American Chemical Society.)

3.2 Electron Transfer

Mukerjee’s research [35] reveals that there exists two-electron transfer pathway for ORR: (1) the inner-sphere electron transfer (ISET) mechanism (Fig. 5a, inset (1) involves chemisorption of desolvated O₂ onto an active site leading to a direct/series 4e⁻ ORR pathway without the desorption of reaction intermediates, (2) an outer-sphere electron transfer (OSET) mechanism [Fig. 5a, inset (2)], wherein the noncovalent hydrogen bonding forces between specifically adsorbed hydroxyl species (OH_ads acting as an outer-sphere bridge) and solvated O₂ (localized in the outer-Helmholtz plane) promote a 2e⁻ reduction step forming intermediates.

Fig. 5

(Part a, b and c adapted with permission from Ref. [35]; Copyright 2013 American Chemical Society.)

a Schematic illustration of the inner-sphere (inset i) and outer-sphere (inset ii) electron transfer mechanisms during ORR. b ORR activity (I_D) and concomitant ring current (I_R) due to hydrogen peroxide oxidation. c Proposed ORR mechanism on Fe-N-C in alkaline solution

As shown in Fig. 5b for ORR in alkaline media (pH > 12), the characteristic ring–electrode signature at ~ 0.8 V for peroxide-intermediate formation through the outer-sphere process is clearly absent on Fe-N-C. This is because Lewis basic anionic hydrogen peroxide intermediates (HO₂⁻, pK_a ≈ 11.6) can be stably adsorbed and immediately reduced to the 4e⁻ product on Lewis acidic Fe²⁺ active sites via the formation of stabilized Lewis acid–base adduct [35]. This ensures that the catalytic cycle in alkaline medium undergoes complete 4e⁻ transfer (Fig. 5c) to regenerate the active site via the formation of ferric-hydroxyl species. The ORR activity is essentially governed by the dynamic structure associated with the Fe^2+/3+ redox transition (Fig. 5c), rather than the static structure of bare sites [36]. The neutral intermediate H₂O₂ negates its Lewis basic character and undermines its stabilization on Fe²⁺ active sites, facilitating the 2e⁻ ORR pathway and generating H₂O₂ (as in Fig. 5b).

From above, it may be noted that the kinetics of ORR is determined to a large extent by the choice of electrode materials and the electrolyte chosen. Evidently a clear understanding of the kinetic mechanisms involved in ORR on Fe-N-C would facilitate further research to design, synthesize and investigate ORR catalysts.

4 Fe-N-C Structure

The Fe atom in Fe-N-C materials could be coordinated by the pyridine/pyrrole nitrogen, carbon or a combination of C and N [37]. The different coordination environments give rise to the different property and hence lead to varied ORR activities. For the sake of the effective catalyst design, it is hence reasonable to identify and ideally, quantify the Fe-N-C structure–ORR catalytic activity correlations [38, 39].

4.1 Identification

The synthesis conditions used for making Fe-N-C materials, for instance pyrolysis temperature [40, 41] and protected atmosphere [42], result in different models of the Fe-N-C configurations with two- to six-coordinated nitrogen (pyridinic or pyrrolic nitrogen atoms) [43,44,45]. As shown in Fig. 6, it has been reported that ideally in a hypothetical graphene sheet (parent system); the Fe-N-C structure would likely include Fe-N₂ at the edge defects (Fig. 6a); Fe-N₄ in-plane defects of graphene sheet (Fig. 6b–d), Fe-N5 and Fe-N6 with one or two additional pyridinic N perpendicular to graphene sheet is to be expected as well (Fig. 6e–f), respectively [44, 46]. And based on the form of the surrounding carbon atoms, Fe-N configurations can be divided into several sub-categories such as in the case of pyridinic N₄-Fe, sites embedded in an intact graphitic layer and surrounded by ten carbon atoms can form FeN₄C₁₀ in Fig. 6b, or in the case of 8 surrounding carbon atoms, FeN₄C₈ can exist on the edge of micropores to bridge two adjacent armchair-like graphitic edges (Fig. 6c) [38].

Fig. 6

Atomic structure of Fe-N-C sites supported on graphene with different coordination environments. Gray ball: carbon; Red ball: iron; Blue ball: nitrogen

To unravel the nature of Fe-N-C structure, the techniques of X-ray photoelectron spectroscopic (XPS), electron microscopy, Mössbauer spectroscopic and X-ray absorption fine structure (XAFS) are mostly available. XPS is effective to elucidate the chemical composition and nitrogen bonding configuration for Fe-N-C materials. In Fig. 7a, five differentiate peaks in the deconvoluted N1s spectra of UK 65, a Fe-N-C material pyrolyzed from iron(III) chloride 5, 10, 15, 20-tetrakis(4-methoxyphenyl) porphyrin (FeClTMPP) as the precursor for Fe, N and C in an inert gas, is assigned to pyrrolic (Pyr-N), Fe coordinated (Fe-N), pyrrolic (Pyr-N), graphitic-like (G-like) and oxidized type (Py-N-O) nitrogen, respectively [47]. Drawing from this, Shen et al. [48] have studied the evolution of the N1s spectrum for single-atom dispersed Fe-N-C samples prepared via template casting methods with different amounts of Fe [48]. For iron phthalocyanine (FePc)-derived Fe-N-C materials, the pyrrolic N keeps associating with Fe, resulting in the presence of two peaks for N1s (pyrrolic N with Fe bonding—denoted as N_α and the pyridinic N with carbon bonding—denoted as N_β). Compared with FePc, Deng et al. [14] found that part of the pyridinic species on the outside macrocycle was destroyed during ball milling whereas Fe-N₄ (pyrrolic nitrogen) was well retained. This peak differentiation principle is widely used in the research on structure of the Fe-N-C catalysts and their correlations with the electrochemical performance for ORR [49]. Though the XPS is reasonable to study the Fe–N bonding, it is found to be insufficient to delineate the details of nitrogen coordination which is relevant for ORR catalytic activity.

Fig. 7

(Part a adapted with permission from Ref. [47]; Copyright 2009 American Chemical Society. Part b, c and d adapted with permission from Ref. [50]; Copyright 2018 John Wiley and Sons.)

a N1s narrow scan spectrum of UK-65 (a Fe-N-C material) and its deconvolution. b Fe K-edge, c XANES and d EXAFS spectra of materials with corresponding EXAFS fitting curves

As a sensitive characterization method to identify non-crystalline species, X-ray absorption spectroscopy (XAS) including X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) spectrum recorded at the Fe K-edge is developed to explore the coordination environment of iron. For example, the absorption edge of the XANES of Fe in the Fe-N-C sample (SA-Fe-N-1.5-800) is close to but exhibits a slightly negative shift as compared with reference Fe₂O₃ samples (Fig. 7b), suggesting that the average valence state of Fe in SA-Fe-N-1.5-800 is slightly lower than + 3 [50]. Correspondingly, the Fourier transform (FT) of the Fe K-edge EXAFS spectra, as shown in Fig. 7c is more intuitive and the coordination number for Fe could be obtained through data fitting (Fig. 7d). Theoretically, the bond length of Fe–N is shorter than that of Fe–C and Fe–Fe, and slightly longer than that of Fe–O. The absence of M–M bond as evaluated by XAS indicates the complete dispersion of metal in materials is isolated single atom [51]. Confirmed by EXAFS, as listed in Table 1, the Fe-N-C materials with atomically dispersed Fe-N_x configuration is successfully synthesized via carbonizing Fe, N, C involves precursors (the mixture, polymer or Fe-N₄ contained organic molecule) under 600–1000 °C at inert atmosphere or ammonia along with/without post-acidic treatment.

Table 1 Parameters of EXAFS fitting for Fe-N-C materials

Commonly, pure FeO, Fe₂O₃ or iron phthalocyanine are used as references to clarify that Fe in Fe-N-C materials is coordinated with nitrogen but not oxygen [49, 50, 54]. However, XAS/EXAFS is censorable to distinguish the Fe–O and Fe–N, since their bond length is similar. And therefore the ⁵⁷Fe Mössbauer spectroscopy, based on the recoil-free absorption of γ rays by Fe⁵⁷ nuclei [46], is a powerful tool for the identification of iron species [38, 57], in which different coordination environments result in the change of isomer shift (IS) and a quadrupole splitting (QS). By taking advantage of ⁵⁷Fe Mössbauer spectra, Dodelet group and cooperators achieved a number of instructive results on Fe-N-C structure [42, 58,59,60]. For example, a catalyst containing the highest concentration of FeN₄ sites can be achieved through the purification of porphyrin-based catalysts with a second heat treatment in forming gas (10% H₂ in N₂) followed by a secondary acid-leaching step. Mössbauer spectra in Fig. 8 reveal that the acid-unstable inorganic metal species can be dramatically reduced in some cases, even down to zero through the purification treatment. In this case, only three doublet peaks (Fig. 8c) are observed which are assigned to the electronically different FeN₄ sites (Table 2). And without the interference from inorganic metal species, the ORR activity is tremendously increased in Fig. 8e, f [60]. Unexpectedly, a new doublet with an IS of 0.30 mm s⁻¹ and a QS of 0.63 mm s⁻¹ appears and can be attributed to the Fe₂N or –Fe_2+xN with x = 0.1 [38]. And the new active Fe-N-C site with an IS of 0.60 mm s⁻¹ and a QS of 3.13 mm s⁻¹ is reported. These peaks are assigned to Fe-N₂ moiety and correspond to an average coordination number of 2 for N (obtained from EXAFS fitting) [48].

Fig. 8

(Part a, b, c, d, e and f adapted with permission from Ref. [60]; Copyright 2016 American Chemical Society.)

⁵⁷Fe Mössbauer spectra of (Fe,Fe) and (Fe, Co) catalysts before a, b and after c, d purification treatment. e, f The corresponding ORR current densities at a rotation rate of 900 rpm in 0.5 MH₂SO₄

Table 2 Average Mössbauer parameters as determined for different iron sites. The color code is the same as those used in Fig. 8a–d [60]

Electron microscopy and associated techniques have contributed significantly to the characterization of heterogeneous catalysts and can provide nano- or atomic-scale information on the structure, morphology, composition and electronic state of areas of interest [61]. With advancement in aberration corrections to achieve an image resolution below 0.1 nm and with the rapid development of in situ techniques, advanced electron microscopy is poised to probe fundamental questions of Fe-N-C catalysts [54, 62]. As shown in Fig. 9a–c, the atomic structure of FeN₄ centers in graphene is revealed for the first time through a combining high-resolution transmission electron microscopy/high-angle annular dark-field scanning transmission electron microscopy with low-temperature scanning tunneling microscopy (STM) [14]. The Fe-N-C centers embedded in the plane of a graphene matrix are a desired model for STM. However, for most Fe-N-C materials obtained through high-temperature pyrolysis, possessing complex textural property with rough surfaces is not appropriate for STM. In most of the published work, high-resolution transmission electron microscopy (HRTEM) with elemental mapping is employed for preliminary estimation of dispersion of Fe and N in Fe-N-C materials [16, 52, 56, 63,64,65,66]. In situ electrochemical scanning tunneling microscopy (ECSTM) has been carried out to investigate the iron–phthalocyanine (FePc)-catalyzed oxygen reduction reaction on Au(111) terrace [66]. A tetragonal structure with a bright spot at the center and four additional spots at the corners (as seen in Fig. 9d) is observed, corresponding to iron and the benzene rings, respectively. The transformation between the FePc molecules (the dim spots) and the FePc-O₂ complex (the bright spots) is reversible under applied potential, as it varies from 350 to 50 mV in oxygen-saturated 0.1 M HClO₄. In addition, theoretical calculations were conducted in this study based on the results and indicated that the lowest unoccupied molecular orbital (LUMO) of FePc-O₂ was largely localized at the center of the phthalocyanine framework and further extended to the O-O part (Fig. 9e), while it is homogeneous distributed on the Pc framework (Fig. 9f).

Fig. 9

(Part a, b and c adapted with permission from Ref. [14]; Copyright 2015 American Association for the Advancement of Science. Part d, e and f adapted with permission from Ref. [67]; Copyright 2016 John Wiley and Sons.)

Structural analysis of a FeN₄ (FeN₄/GN) catalyst. a High-resolution transmission electron microscopy (HRTEM) image. b High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image. Inset is the electron energy loss spectroscopy (EELS) atomic spectra of elemental Fe and N from the bright dots. c Low-temperature scanning tunneling microscopy (LS-STM) image. d High-resolution STM image of a FePc monolayer on Au (111) surface. Top views of the electron density distribution in e FePc-O₂ and f FePc

It is to be noted that hitherto the complex Fe-N-C structure is not completely identified; it is a result that ought to be corroborated further. This hence presents a rather interesting puzzle for both catalyst researchers and those who work on advanced characterization techniques.

4.2 Quantification

Turnover frequency (TOF), calculated as the derivative of the number of turnovers of the catalytic cycle with respect to the time per active site, is a direct evaluation parameter for intrinsic activity comparisons on different active sites [68], in which active site quantifications are key to TOF calculations. And unlike precious metal (Pt, Pd, Ru) [69,70,71]-based catalysts, titration by under potential deposition (UPD) of copper, H₂ or CO stripping titration is inactive for Fe-N-C materials. However recently, significant improvements have been made for Fe-N-C quantification, making it relevant for the further development and rational design of Fe-N-C catalysts.

Ideally, without interference of inorganic metal sites, the densities of pristine Fe-N-C moieties on materials can be quantified from the information of either Fe or N. Through peak differentiation on deconvolution of N1s XPS spectra (Fig. 10a), the amount N in the form of Fe-N is thus determined by Shen et al. [48, 73]. As the average coordination number of N is 2.0 confirmed by EXAFS, the amount Fe-N-C therefore is half of that of N at. % present in Fe-N. Since low Fe concentrations are not detected by routine XPS, Jason et al. [74] digested the sample using a mixture of nitric acid and hydrochloric acid in 5:1 ratio and measured the Fe content by inductively coupled plasma optical emission spectroscopy (ICP-OES). The total reflection X-ray fluorescence spectroscopy (TXRF) is also a reliable means for iron content determination in the Fe-N-C catalysts [72]. Another relevant approach would be neutron activation analysis (NAA) as is introduced by Kramm [60], to determine the concentrations of mononuclear Fe-N₄ moieties. However, we cannot regard all the Fe in the catalyst as active sites for a large proportion is likely to be inactive phases or in-accessably buried in the carbon skeleton [42, 75]. Similar to CO adsorbed on platinum, Daniel [39, 72] found that nitrite anion can strongly interact with Fe-N-C to form a stable poisoned catalyst adduct 3, as shown in Fig. 10b. The as-formed adduct 3 can be reduced and stripped at very low potential (Fig. 10c), along with complete recovery of ORR activity (Fig. 10d). This stripping process makes the in situ quantification of Fe-N-C sites for ORR reasonable. So that the gravimetric site density (MSD) of active sites can be calculated as:

$$ {\text{MSD}} = \frac{{Q_{\text{strip}} }}{{n_{\text{strip}} F}} $$

(6)

where Q_strip is the excess Coulometric charge associated with the stripping peak (Fig. 10c), n_strip is the number of electrons associated with the reduction in one adsorbed nitrosyl per site, F is the Faraday constant. On the other hand, if one is to simply count the amount of Fe using the nitrite stripping, it would result in an underestimation of active sites in view of that the ORR activity of poisoned Fe-N-C material is still much higher than metal-free N-C material (Fig. 10d).

Fig. 10

(Part a adapted with permission from Ref. [48]; Copyright 2017 Elsevier. Part b adapted with permission from Ref. [72]; Copyright 2016 American Chemical Society. Parts c and d adapted with permission from Ref. [39]; Copyright 2016 Nature Publishing Group.)

a Deconvoluted N1s of XPS. b Proposed reaction mechanism of Fe-N-C active sites after treatment with a nitrite containing solution and subsequent treatment under different pH conditions for the hybrid catalysts. c Homogeneous reduction in aqueous nitrite and excess current associated with the reductive stripping of intermediates. d ORR performance of catalyst layers before, during and after nitrite adsorption

5 Activity of Fe-N-C

High activities can be achieved for Fe-N-C catalysts in alkaline media, and the nature of the active sites has been progressively investigated in recent studies [16, 54, 76]. However, in acidic medium, Fe-N-C catalysts still exhibit higher overpotentials and lower stabilities, despite significant improvements. This is because as has been discussed in Sect. 3.2, in acidic medium, the neutral intermediate H₂O₂ negates the Lewis basic character and undermines stabilization on Fe²⁺ active sites, driving the 2e⁻ ORR pathway and generating H₂O₂ (as in Fig. 5b) [35]. For ORR on Fe-N-C materials, to improve the activity (i.e., for facilitating substance transformation and to enhance electron transfer rate), similar to the traditional heterogeneous catalyst, there are two strategies: (1) increasing the intrinsic activity of Fe-N-C or (2) increasing the number of active site on a given electrode (e.g., through increased loading or tuned catalyst structure to expose more active sites per gram) [77]. Besides, the activity of electrochemical catalyst can also be enhanced by the improvement in conductivity for fast electron transport.

5.1 Improve Intrinsic Activity of Fe-N-C

Since the adsorption of O₂ on carbon and nitrogen atoms is rather weak, the O₂ molecule is mainly activated on the Fe atom of Fe-N-C catalysts. The spin-polarized DFT computations reveal that the 2p orbitals of O₂ are hybridized with the 3d orbitals of Fe in both spin-up and spin-down channels and that the O₂–2π* state can be partially occupied in the spin-down channel due to the charge transfer from the Fe-N-C monolayer to the O₂ molecule [78]. The d ²_z orbital of Fe, which can be controlled through the incorporation of functional groups, would split into bonding and anti-bonding states before and after the adsorption of O₂. Subsequently the relative energy position of the d ²_z orbital with respect to the Fermi level can induce an Fe redox couple potential shift and modulate the catalytic activity toward ORR [79]. Hence, the intrinsic activity of Fe-N-C relies on the coordination environments toward Fe, including the coordination number of N, the property of the coordinated N and any other invasive factors.

As a stable structure, Fe-N₄ with different states of N exhibits varied performance for ORR. One important attempt to correlate bonding in the solid with observed ORR activity has come through the work by Xu et al. [37]. They took the number of valence electrons in the d orbital (θ_d) and electronegativity into consideration together, to arrive at a corresponding descriptor (φ) for activity prediction:

$$ \varphi = \theta_{\text{d}} \times \frac{{E_{\text{M}} + \alpha \times \left( {n_{\text{N}} \times E_{\text{N}} + n_{\text{C}} \times E_{\text{C}} } \right)}}{{E_{{{\text{O}}/{\text{H}}}} }} . $$

(7)

where θ_d is the number of valence electrons in the occupied d orbital of the metal element, E_M represents the electronegativity of metal element, E_N and E_C represent the electronegativity of nitrogen and carbon elements, n_N and n_C represent the number of nearest-neighbor N and C atoms, and α is the correction coefficient.

A volcano relationship (Fig. 11a) between onset potentials for ORR and the descriptor φ is observed in both theoretical and corresponding experimental verification [80]. The volcano curve indicates that the intrinsic activity of Fe-N₄C₁₀ with four pyridinic nitrogen atoms is higher than that of Fe-N₄C₁₂ (Fig. 6d) with four pyrrolic nitrogen atoms. Furthermore for the Fe-N₄ moiety with four pyridinic nitrogens, Liu et al. [36, 80] took advantage of DFT formalism and conducted calculations to reveal that the activation energy for O–O bond breakage in FeN₄C₈ is about 0.20 eV, less than that for FeN₄C₁₀ of 0.56 eV (Fig. 11b). This suggests that the intrinsic catalytic activity for ORR on FeN₄C₈ is higher than FeN₄C₁₀ [36, 81]. And Koslowski et al. show that a C-Fe-N₂ center where iron is bonded to two nitrogen atoms and one carbon atom, favor a 2-electron ORR that yields hydrogen peroxide and that it is less active than an in-plane Fe-N₄ center [50, 82]. Compared with Fe-N₄ in Fig. 11c, the edge site Fe-N₂ (iron is not directly bonded to carbon) achieves superior intrinsic activity due to a lower interaction with *O₂ and *OH intermediates and enhanced electron transport [48, 83,84,85]. In the work of Dodelet and co-workers, both Fe-N₂ and Fe-N₄ are shown to coexist in the carbon-supported materials with a ratio determined by preparation conditions [43, 84].

Fig. 11

(Part a adapted with permission from Ref. [37]; Copyright 2018 Nature Publishing Group. Part b adapted with permission from Ref. [80]; Copyright 2017 American Chemical Society. Parts c and d adapted with permission from Ref. [48]; Copyright 2017 Elsevier.)

a Theoretical and corresponding experimental onset potentials for ORR versus the descriptor φ. b Calculated free energy evolution diagrams for O₂ reduction at the active sites of FeN₄-C₁₀, FeN4-C₁₂ and FeN₄-C₈. c Free energy diagram of the ORR on FeN₂ and FeN₄ sites. d Synthesis procedure for the Fe-N₂ catalyst (FeN₂/NOMC)

To summarize, the intrinsic activity of different active sites on Fe-N-C can be listed in the order of: Fe-N₂ > Fe-N₄C₈ > Fe-N₄C₁₀ > Fe-N₄C₁₂ > C-Fe-N₂. And hence, a catalyst consisting of abundant Fe-N-C site with high intrinsic activity is desirable; in particular there would be advantages to have sufficient Fe-N₂ species on the edge of carbon support. For example in Fig. 11d, a nominal Fe-N2-based catalyst (FeN₂/NOMC) is synthesized using an intriguing template casting strategy. In the sample, ⁵⁷Fe Mössbauer spectroscopy confirmed that Fe is only coordinated by N atoms, and EXAFS at Fe K-edge X-ray absorption spectroscopy reveals that the average coordination number of N is 2.0. The well-defined structure exhibits superior performance with a more positive half-wave potential for ORR compared to Pt/C catalyst [48].

In addition to the above pointers to improve the ORR activity of Fe-N-C, it may be noted that there is scope for further improvement through the tuning of electronic structure of Fe in Fe-N-C configuration. This in fact may be possible via use of a suitable catalysts supporting either carbon with heteroatom doping or other nanoparticles. Among them, it is reported that S-doped Fe-N-C catalyst can exhibit better ORR activity than undoped Fe-N-C in both alkaline and acidic electrolyte [86,87,88,89,90,91]. This is consistent with another report [73] wherein thiophene-like structure (C-S-C) formed by S incorporation reduces the electron localization around Fe centers and improves interaction with oxygenated species, thus lower ORR activation barriers on Fe-N-C and boosts the ORR activity. In addition, Sun et al. [92] demonstrate that embedding of oxygen into the Fe-N-C structure is favorable for boosting durability and methanol endurance of the catalyst through the elongation of the inter-atomic atom distance (from the active core) [92]. Not only electronegative elements S and O, the electropositive element B [93] has also been introduced to modify the property of Fe-N-C site. As reported by Kai et al., boric acid is selected to be mixed with other precursors [94]. Then after direct pyrolysis and acid leaching, B-decorated porous carbon frameworks with atomically dispersed Fe-N-C species (denoted as FeBNC) are obtained. And the B atom around the Fe-N-C structure is more favorable to the adsorption of chemical species and catalytic activity for ORR [95]. The electropositive effect also exists on the Fe-N₄ treated by NH₃ and with higher coordinated N to enhance the intrinsic ORR activity [38, 44]. Factually, without post-acid leaching, the magnetic Fe/Fe₃C nanoparticle is ubiquitous as a by-product in the preparation of Fe-N-C catalysts through high-temperature pyrolysis [73, 96,97,98]. Although the Fe/Fe₃C is also active toward ORR [99,100,101], the activity is much less than Fe-N-C. The vicinity of Fe/Fe₃C nanocrystals is not required per se [60, 72, 97, 102], but it could enable the enhancement of catalystic performance prepared via some specific preparation routes. For instance, an Fe-based material, synthesized by pyrolysis of thiourea and agarose containing α-Fe₂O₃ nanoplate as Fe precursor, displays excellent catalytic performance for ORR with more positive onset potential (1.1 V vs. the reversible hydrogen electrode) [103]. Likewise Jiang et al. [49] reported a facile and easily scalable method to prepare Fe-N-C materials via pyrolysis of the mixture of carbon nanotube, glucose, sodium dodecyl sulfate and nitrate nonahydrate. The resulting sample contains Fe-N-C configurations and graphene-encapsulated Fe/Fe₃C nanocrystals. It is proven that the metallic iron nanoparticles reduces the charge density of central Fe atoms in neighboring Fe-N-C configurations and favors the adsorption of oxygen molecule to deliver high ORR catalytic performance.

5.2 Increase the Number of Fe-N-C Sites

Catalytic activity relies on the number of accessible active Fe-N-C sites on the surface. Here, our previous study found that ORR activity is linearly proportional to the concentration of Fe-N-C sites (Fig. 12a). Furthermore as is deduced from Butler–Volmer formulation, the half-wave potential (E_1/2), a characteristic of ORR activity, is independent of the substrate concentration and is directly related to the number of active sites:

$$ \begin{aligned} &E_{1/2} = {\text{aln}}\left[ {\text{Fe} -{ \text{N}} -{ \text{C}} \, \left( {{\text{at.}}\% } \right)} \right] + b \hfill \\ &a = \frac{RT}{\alpha F},\;b = \frac{RT}{\alpha F}\ln (\frac{{4mFAk_{3}^{0} C_{c} (0,t)}}{{J_{L} }}) + E^{\theta } \hfill \\ \end{aligned} $$

(8)

where A is the coefficient constant of current to ORR reaction rate, S is the number of active site, E^θ is equilibrium potential, F is the Faraday constant, k₀ is the standard rate constant, and α is the transfer coefficient.

Fig. 12

(Part a adapted with permission from Ref. [48]; Copyright 2017 Elsevier. Part b adapted with permission from Ref. [41]; Copyright 2014 American Chemical Society. Part c adapted with permission from Ref. [6]; Copyright 2016 American Chemical Society. Part d adapted with permission from Ref. [104]; Copyright 2015 American Chemical Society.)

a Correlation between ORR current density and Fe-N-C concentration at potentials of 0.89, 0.84 and 0.79 V. b Dependence of H₂O₂ yield at 0.70 V on Fe-N-C catalyst loading at the electrode surface. Synthesis scheme for c the “silica-protective-layer-assisted” approach and d the “shape fixing via salt recrystallization” method

Thus, a high content of Fe-N-C active sites is essential to deliver high ORR catalytic performance. So far, several advanced strategies have been established to improve Fe-N-C sites during catalyst preparation and electrochemical measurement.

The most straightforward method for Fe-N-C sites enhancement is to increase the materials loading during testing. As presented in Fig. 12b, Wang et al. [41] found that the H₂O₂ yields on Fe-N-C are inversely proportional to the catalyst loading. Once the catalyst loading increase from 0.06 to 0.6 mg cm⁻², the intermediate H₂O₂ yield drops from 16 to < 1%. The estimated rate constants based on the Wroblowa model (Fig. 4b) reveal that at high loading density (> ca. 200 μg cm⁻²), k₃/k₅ is larger than unity, k₁ ≫ k₂ and k₆ value decreases. These indicate that the intermediate H₂O₂ undergoes further reduction to H₂O rather than desorption. Whereas, the lower loading density leads to the 2-electron reduction of O₂, which corresponds to k₁ < k₂ and k₃/k₅ is less than unity [29]. However, the higher loading of catalysts, the thicker of modified layer, brings about the more intensive normal diffusion, especially for the three-dimensional porous materials [8].

On the other hand, significant efforts have been devoted to improve Fe-N-C active sites, either by increasing the density of Fe-N-C or enhancing the accessible area of materials. It is generally recognized that the avoidance of carbide formation during pyrolysis represents a promising way to enhance the density of ORR active sites on Fe-N-C catalysts [97]. A general “silica-protective-layer-assisted” approach in Fig. 12c is developed to preferentially generate catalytically active Fe-N-C sites in catalysts by suppressing the formation of large Fe-based particles [6]. Yang et al. propose a surfactant-assisted method to increase the active site with a small amount of iron [105]. In addition, the Fe-based particles are more often removed by post-acid etching, since iron, iron carbides or iron oxides are acid-unstable [96, 106]. The inorganic metal species in Fe-N-C catalysts can be drastically reduced by performing a second heat treatment of as-synthesized catalysts in forming gas or NH₃ with a subsequent acid leaching [38, 59]. The method is very efficient since the inorganic Fe nanoparticle covered or buried in the carbon can be exposed to acid after the carbon removed via methanation under forming gas or NH₃ treatment [42, 107]. Using a semi-embed strategy by pre-anchoring Fe precursor on the surface of template (SBA-15), the strong anchoring effect of SBA-15 prevents the Fe-based species to penetrate into the carbon skeleton and make the formed Fe species be accessible on the surface. In this way, different amounts of Fe-N-C sites are introduced by simply regulating the amount of impregnated Fe precursor [47]. However, it is noteworthy here that it is impossible to continuously increase Fe-N-C content by monolithically adding Fe precursor, since the transition metals such as Ni, Co and Fe are common catalysts to enhance the graphitization of carbon [108, 109] and hence lead to the evolution of doped nitrogen as well as lower specific area. Therefore, keeping balance of the invasive Fe precursor and doped nitrogen is a director to optimize the Fe-N-C materials. Interestingly, a “shape fixing via salt recrystallization” method in Fig. 12d is employed to efficiently synthesize N-doped carbon nanomaterials with a high density of Fe-N-C sites as an ORR catalyst, where the NaCl crystal functions as a fully sealed nanoreactor and facilitates the N incorporation and graphitization [104].

Currently large surface carbon with a mass of defect is used as support for Fe-N-C materials preparation. Benefiting from pore-forming technology, in addition to the typical nanocarbons such as carbon nanotube- [49, 82], nanofiber- [110] or graphene [14, 88, 111]-based catalysts, numerous novel Fe-N-C materials with tunable porous structure can be fabricated by bottom-up strategy [112]. Using the method of hard template sacrifice, the sample after carbonization can inversely replicate the porous structure of template and maintain its morphology, where the surface area is significantly determined by the selected template. For examples, silica nanoparticles, ordered mesoporous silica and montmorillonite are used as templates for achieving mesoporous structures, which reveals a strong correlation between the activity and the apparent BET surface area [113, 114]. Song et al. introduced ultrathin tellurium nanowires as template with biomass-derived d(+)-glucosamine hydrochloride and ferrous gluconate as precursors to synthesize Fe-N-C carbon nanofibers aerogels as efficient non-precious metal catalysts for ORR (Fig. 13a) [115]. In Fig. 13b [116], Wang et al. [116] report a 3D simple cubic carbon frameworks (SCCFs) with the surface decorated homogeneously with Fe-N-C moieties. It is synthesized via in situ carbonization of the surface-coating ligands on Fe₃O₄ 3D simple cubic superlattices along with acid etching and NH₃ activation. Benefiting from the homogeneously dispersed and fully accessible Fe-N-C sites, highly graphitic nature and enhanced mass transport achieved through the porous nature of the structure, these SCCFs outperformed Pt/C for ORR. It is observed that the specific surface area of N-doped carbon spheres doubles through CO₂ activation treatment. It hence improves the availability of accessible N-functional groups to form high-density and homogeneous Fe-N-C moieties for boosting ORR electrocatalysis [117]. Appealingly, metal–organic frameworks (MOFs), as a novel porous material, have received wide attention in recent years and have been brought in Fe-N-C materials fabrication [53]. After rational carbonation, MOFs derived or decorated materials inherit the porous structure of corresponding MOFs with high specific surface area and uniform active sites, particularly with the zeolitic imidazolium framework (ZIFs), which has transition metal ions and nitrogen-containing organic ligands, such as ZIF-7 [76] and ZIF-8 [56, 118, 119], as the most applicable for self-sacrificed precursor/template for the preparation of nanoporous carbons. As shown in Fig. 13c [76], the 3D structure of ZIF-7 is transformed to 2D by surface functionalization with AFC, and after pyrolysis, a 2D mesoporous carbon doped with Fe-N-C active sites can be realized with interconnected and mesopore-dominated pore structure for high performance toward ORR in alkaline medium. Figure 13d [56] represents a highly stable isolated Fe-N-C catalyst with excellent ORR reactivity which is prepared via a cage encapsulated precursor pyrolysis strategy using ZIF-8 with cavity diameter of 11.6 nm and pore diameter of 3.4 nm as molecular-scale cages. Due to the confinement effect, the trapped Fe precursor in the molecular-scale cages turns into isolated single iron atoms anchored on nitrogen species during carbonization directly without acid treatment. Furthermore, a “semi-closed system” pyrolysis strategy using ZnCl₂/KCl eutectic salts is adopted to control the precursor carbonization and pore-making process [120]. The molten eutectic salts create an ionic liquid-confined space for preventing the fast decomposition, cracking, shrinking and sintering of precursors. This approach also suppresses the large weight loss and prevents N evaporation of the precursor during pyrolysis. It thus contributes to the unique three-dimensional porous graphene-like structure with a high surface area and graphitization degree, as well as high density of the Fe-N-C sites. Accordingly, the as-prepared Fe-N-C catalyst exhibits superior performance with E_1/2 of 0.803 V and 0.918 V in acidic and alkaline media.

Fig. 13

(Part a adapted with permission from Ref. [115]; Copyright 2016 John Wiley and Sons. Part b adapted with permission from Ref. [116]; Copyright 2017 American Chemical Society. Part c adapted with permission from Ref. [76]; Copyright 2017 American Chemical Society. Part d adapted with permission from Ref. [56]; Copyright 2017 John Wiley and Sons.)

Schematic illustration of the synthesis of Fe-N-C catalysts using templates of a ultrathin tellurium nanowires, b Fe₃O₄ cubic super lattices, c ZIF-7 and d ZIF-8

5.3 Introduce Secondary Active Sites

A Fe-N-C catalyst with less H₂O₂ yield during ORR is crucial for displacing Pt [121]. The fundamental understanding of ORR mechanisms on Fe-N-C catalysts in acid medium is as shown in Fig. 14a. Choi et al. proposed that the desirable 4-electron reduction pathway for ORR can proceed via a direct 4e- or a consequential 2e⁻× 2e⁻ mechanism on a single active site. Likewise a bifunctional 2e⁻+ 2e⁻ mechanism on two active sites is plausible too [57]. Thus, the introduction of a secondary active sites to catalyze the peroxide reduction reaction (PRR) is a rational approach to alleviate the issue associated with H₂O₂. Since the Fe particles encapsulated in N-doped carbon layers (Fe@N-C) are active toward the intermediate H₂O₂ in acidic solution. However, in moderation, it can play the role of secondary active sites to lower the yield of H₂O₂ during ORR through 2e⁻+ 2e⁻ mechanism. Therefore, a material containing Fe-N-C moieties and Fe@N-C exhibits high O₂ reduction performance with enhanced selectivity and durability [57].

Fig. 14

(Part a adapted with permission from Ref. [57]; Copyright 2017 John Wiley and Sons. Part b adapted with permission from Ref. [126]; Copyright 2018 Royal Society of Chemistry. Part c adapted with permission from Ref. [73]; Copyright 2017 John Wiley and Sons.)

a ORR mechanisms of Fe-N-C catalysts in acidic media and the promotion of Fe particles encapsulated in N-doped carbon layers (Fe@N-C) toward Fe-N-C catalysis through the 2e⁻+2e ⁻pathway. b Relationship between the ratio of the amount of pyridinic and Fe-N-C to the amount of pyrrolic N and the ORR half-wave potential of the catalysts. c Kinetic current densities of Fe-N-C catalysts with and without S-doping in O₂-saturated 0.1 M KOH at a scan rate of 10 mV s⁻¹

To bring in extra active sites is also effective to further enhance the activity of Fe-N-C materials. As the member of M-N-C materials, the single-atom dispersed Co-N-C catalyst also shows excellent performance for O₂ reduction [122,123,124]. An optimized Fe + Co-N-C sample is prepared by one-step pyrolysis of Fe/Co bimetal zeolitic imidazolate framework (Fe,Co-ZIF). It exhibits a positive half-wave potential (E_1/2) of 0.875 V and 0.764 V in the alkaline and acidic electrolyte with superior methanol tolerance and electrochemical stability [125]. Actually, the site with low activity, such as remainder pyridinic, pyrrolic nitrogen or others in Fe-N-C sample, can also contribute to the ORR activity. For example, it is confirmed that the ORR activity is proportional to the ratio of the amount of pyridinic and Fe-N-C to the amount of pyrrolic N [126], because the carbon atoms with Lewis basicity next to pyridinic N are the main active sites in N-doped carbon materials [127]. As shown in Fig. 14b [126], the sample with higher ratio of the amount of pyridinic and Fe-N-C to the amount of pyrrolic N present results in a more positive E_1/2. On the other hand, compared with S free Fe-N-C sample (Fe/NC), an improved ORR kinetic currents density (J_k) from C-S-C site on support is observed for the S-doped Fe-N-C catalyst (Fe/SNC) in alkaline solution (Fig. 14c) [73].

5.4 Enhance the Conductivity of Fe-N-C Materials

A material with high conductivity is approved for electrochemical catalysis. For carbon materials, the higher degree of graphitization, the better of its conductivity. Previous studies reveal that the pyrrolic N is the most unstable nitrogen, which is mainly formed at low temperatures (200–350 °C) in the nitrogen-doped carbons. When temperature reaches about 550–750 °C, the pyrollic N would transform to pyridinic N and further to graphitic-N (above 750 °C) [129,130,131,132,133,134,135,136,137]. It is hence inferred that a higher temperature gives rise to higher degree of graphitization. Besides, the introduction of non-3d high-valency Mo can not only prevent Fe species from aggregation through the dilution effect of Mo, but also enhance the graphitization degree of carbon skeleton to resulting in a high electrical conductivity of Fe-N-C catalysts [138]. However, it is indeed a double-edged sword since the graphitizing process gives rise to the sacrifice of surface area, evolution of heteroatom and destroys the active Fe-N-C structure. Otherwise, the conductivity could also be improved by structural combination. Such as in Fig. 15a [128], a porous Fe-N-C hybrid material composed of hierarchically ordered porous carbon (OPC) microblocks interlinked with CNTs is developed for catalyzing ORR. The abundant graphitic CNTs maximize the conductivity and results in the outstanding catalytic activity. Lately, a high surface area of 1380 m² g⁻¹ and homogeneous distribution of Fe, N-embedded hierarchical carbon framework wired onto 1D porous carbon nanotubes is rationally designed from a nucleation and growth of ZIF-8 on the surface of porous tellurium nanotubes (Fig. 15b) [65]. The 1D structures enhance the electron transport and the high surface supply enabling host for abundant access to Fe-N-C sites, contributing to a E_1/2 of 0.867 and 0.818 V in alkaline and acidic media, respectively.

Fig. 15

(Part a adapted with permission from Ref. [128]; Copyright 2014 John Wiley and Sons. Part b adapted with permission from Ref. [65]; Copyright 2017 John Wiley and Sons.)

Schematic illustration of the synthesis of a Fe-N-C NT-OPC catalysts and b Fe,N-embedded interconnected MOF-derived porous carbon nanotubes using tellurium nanotubes as a sacrificial template

6 Application for Energy Conversion Systems

6.1 Proton Exchange Membrane Fuel Cells

Membrane electrode assembly fuel cells (MEAFC, Fig. 16a) are able to efficiently convert hydrogen/oxygen chemical energy into electricity at low temperatures (normally < 90 °C) and are ideal clean-energy systems for sustainable economic development. MEAFCs can be classified into two categories based on the type of polymer electrolyte membrane (PEM) used in which one is the alkaline polymer electrolyte fuel cell (APEFC) in which the PEM conducts hydroxide ions and the other is the proton exchange membrane fuel cell (PEMFC) in which the PEM conducts protons in acidic environments [139]. Of these two, APEFCs possess several advantages over PEMFCs, including faster oxygen reduction reactions, less corrosive environments, decreased fuel crossover rates and mitigated CO poisoning [140]. However, disadvantages of APEFCs include system complexities due to water removal requirements at the cathode, the occasional need to replenish KOH electrolytes and the requirement for pure H₂ and O₂ gas (APEFCs cannot tolerate even atmospheric levels of CO₂). And as a result, PEMFCs are currently considered to be the most suitable candidates for application in portable and automotive applications [141]. And currently, the dominant limiting factor in PEMFCs, unlike APEFCs, which is the PEM, is the electrocatalyst [140], and therefore, the focus of this review will be on the application of Fe-N-C catalysts for PEMFCs.

Fig. 16

(Part a adapted with permission from Ref. [144]. Part b adapted with permission from Ref. [107]; Copyright 2009 American Association for the Advancement of Science. Part c adapted with permission from Ref. [145]; Copyright 2018 John Wiley and Sons. Part d adapted with permission from Ref. [146]; Copyright 2011 Nature Publishing Group.)

a Schematic of a membrane electrode assembly fuel cell (MEAFC). b Comparison of the polarization curves from H₂-O₂ fuel cells tested (p_O = p_H2 = 1.5 bar, 80 °C) using Fe-N-C catalysts with different mass loading and a Pt-based catalyst. c Polarization and power density curves (80 °C, absolute pressure = 2.5 bar) of Fe2-Z8-C, Z8-Fe2-C, and Pt/C(20%) with cathode loadings of 2.8, 2.7 and 0.33 mg cm⁻², respectively. d Tafel plots of the volumetric activity of cathodes made from Fe-N-C materials

To achieve high-performance O₂ electrochemical reduction, Fe-N-C catalysts are currently the most promising substitutes for Pt-based catalysts in membrane cathode fabrication, in which the current density of a MEA using Fe-N-C electrocatalysts of 5.3 mg cm⁻² at the cathode is equivalent to that of a Pt-based cathode with a loading of 0.4 mg Pt cm⁻² at a cell voltage of ⩾ 0.9 V (Fig. 16b) [107]. And recently, high-power densities of 1.14 and 1.11 W cm⁻² at 2.5 and 2.0 bar H₂/O₂ were achieved using a MEA fabricated with a Fe-N-C catalyst (Fe2-Z8-C, Fig. 16c), in which 280 mW cm⁻² at 0.8 V was achieved, which is higher than that of the 2020 target (250 mW cm⁻²) of the US Department of Energy (DOE) [142]. However, in comparison with Pt/C, the nominally high power density of the reported Fe-N-C catalyst was related to higher mass loadings. Another crucial target for MEAs using non-Pt catalysts, as per DOE guidelines, is a kinetic activity per volume of 300 A cm⁻³@0.8V_iR-free (intersection of the extrapolated Tafel slope with the 0.8 V_iR-free axis, Fig. 16d) [142, 143]. Here, ORR volumetric performance is related to catalytic activity because mass transport limitations are related to electrode thicknesses, in which a thick MEA will lower mass transport and decrease power density [107]. Therefore, it is important to attain the same kinetic current density on NPMC cathodes as Pt-based cathodes without exceeding thickness; however, the volumetric current densities of state-of-the-art Fe-N-C catalysts are still far from the DOE target (Table 3). This can be explained by the facts that Fe-N-C catalysts are carbonized at high temperatures (> 700 °C) and are more hydrophobic, and that although the porous structure of Fe-N-C catalysts can contribute to higher surface areas for active site dispersion and mass transport enhancement, it also lightens bulk density and cause these catalysts to be much thicker than Pt/C at same mass loading. Therefore, the fabrication of thin Fe/N/C catalyst layers with fast mass transfer is a key issue in future studies [91].

Table 3 State-of-the-art PEMFC performances of published Fe-N-C materials

6.2 Metal-Air Batteries

Metal-air batteries (MABs) are attracting increasing attention from researchers due to advantages such as their cost-effectiveness and environmental compatibility as well as their high energy density [151]. In practical applications, the discharge performance of metal-air battery systems [152, 153] with either two-electrode (Fig. 17a) or tri-electrode configuration (Fig. 17b) relies on the oxygen reduction reaction at the air electrodes. Therefore, a Fe-N-C material that can facilitate ORR through the direct four-electron reduction pathway is highly preferred. There are various MABs according to the metals used [154,155,156,157,158]. Among them, Li-air and Zn-air batteries with theoretical energy density of 5200 and 1084 Wh kg⁻¹ (including oxygen) are the most promising ones [152].

Fig. 17

(Parts a and b adapted with permission from Ref. [152]; Copyright 2018 American Association for the Advancement of Science. Parts c and d adapted with permission from Ref. [161]; Copyright 2018 Elsevier. Part e adapted with permission from Ref. [94]; Copyright 2018 American Chemical Society. Part f adapted with permission from Ref. [162]; Copyright 2018 Royal Society of Chemistry.)

a Tri-electrode and b two-electrode systems of rechargeable metal-air batteries. c Comparison of the polarization and power density of Zn-air batteries (ZAB) between Fe-N-C and commercial Pt/C at a catalyst loading of 2.0 mg cm⁻². d ORR activity of Fe-N-C and commercial Pt/C measured using rotating disk electrode (RDE) at 1600 rpm. Polarization curves and corresponding power density plots of ZABs using Fe-N-C doped with e boron and f sulfur as ORR catalysts

In Shui et al. work, atomically dispersed Fe-N-C sample is for the first time used as the cathode catalyst to improved performance of the rechargeable Li-air battery (LABs). Compared with the benchmark, which is metal oxide catalyst (α-MnO₂/XC-72); LABs with Fe-N-C as the catalyst exhibit high cyclability and low overpotential during both discharge and charge processes [159]. However, the reversibility is achieved only at the smaller current densities of 0.045 mA cm⁻² (making it far away from the practical applications). Using the air cathode with Fe-N-C which is uniformly dispersed on the graphene sheets, a high discharge capacity of 6762 mAh g⁻¹ and 50 discharge–charge cycles is also achieved for LABs [160].

Accordingly, the Fe-N-C materials show promising applications for the booming activities in Zn-air battery (ZABs). An efficient ORR catalyst constructed by anchoring Fe-N4 moieties on the carbon matrix with high Fe loading (up to 3.8 wt %) exhibits 121.8 mW cm⁻² for ZABs at catalyst loading of 2.0 mg cm⁻², comparable to that of the Pt/C-based air cathode (Fig. 17c) [161]. It is worth noting that the ZABs performance is assessed at a static station, and the diffusion process is the rate determining step and can impact the ORR activity (Fig. 17d) during rotating disk electrode (RDE) testing under forced convection. Although the ORR activity of Fe-N-C materials is much higher than Pt/C assessed on RDE in alkaline solution, there still exists the problem of mass transport limitations for ZABs. It is confirmed that the incorporation of heteroatom can adjust the surface polarities and electronic properties of Fe-N-C materials to elevate ORR activity. In comparison with boron-free Fe-N-C catalyst in Fig. 17e, the B doped one shows 19.1% increment in peak power density during ZABs discharge process [94]. Using S-promoted Fe-N-C catalyst as air cathode for ZABs, power densities of 250 mW cm⁻² (Fig. 17f) are achieved at 5 mA cm⁻², double that of the untreated Fe-N-C materials [162]. In this sample, sulfur is employed as a “promoter” to increase the specific surface area via affording interpenetrating holes, enhance the dispersion of Fe to form evenly dispersed Fe-N-C site and maintain a litchi-like structure of carbon framework.

Promisingly, the Fe-N-C materials are also adopted to the rechargable ZABs in two-electrode system as bifunctional catalyst in the reports [96, 163]. However, the fraught deactivation is observed during charge and discharge cycles. It is due to the damage/oxidation on ORR activity site of Fe-N-C during charge process as Fe-N are thermodynamically less stable and can be in situ oxidized at high potential [164].

7 Challenges and the Way Forward: a Perspective

In recent years, significant progress has been made in developing Fe-N-C materials for oxygen reduction reactions in both alkaline and acidic media. Due to the developments in characterization techniques at high resolution, as well as theoretical simulations, reasonable progress has been made to elucidate these complex real catalytic systems. The potential of Fe-N-C materials is evident; however, there remain many scientific and technological challenges.

Most of the work relating to Fe-N-C for ORR being reported recently is focused on novel methods for material preparation; however, the activity seems to be stagnate [102, 165]. Especially in acidic media, the activity of Fe-N-C materials is much less than Pt/C with ~ 50 mV lower of E_1/2. Is it the extreme value for actual Fe-N-C materials, although the activity of Fe-N-C materials can be further theoretically enhanced by the structural optimization?

The verified Fe-N-C-based catalysts have activities that are primarily due to their Fe-N₄ active centers. However, this is not the most active site according to theoretical predictions [73, 83]. Fe-N₂ sites, the most active sites for ORR still unconfirmed experimentally with solid evidence. Also fundamental questions remain, such as the coordination state of Fe in these systems remains unanswered.

A key limitation for further activity would be alleviating the rapid decay in fuel cell performance, which is usually observed with Fe-N-C catalysts at the initial testing due to the loss of iron located in the open micropore network or radical oxygen species produced by Fenton reactions [122, 166, 167].

Despite the many challenges, there is reason to be hopeful about Fe-N-C for ORR since advancements in situ and ex situ characterization techniques will continue to give insights of relevance to progressively move toward champion catalysts. The gap between theoretical and experimental value is narrowing over time and mechanistic insights are certainly getting more robust. The key reason to continue this activity may eventually be to do for its sweeping relevance for ORR, and perhaps beyond!