Introduction

The development of next-generation energy conversion and storage devices is the key for the sustainable development of human society due to the rapid consumption of fossil fuels and the concomitant environmental contamination [1]. The oxygen reduction reaction (ORR) is a fundamental reaction in electrochemical devices, such as fuel cells and other renewable energy devices, e.g., metal–air batteries [2, 3]. Although Pt and a series of Pt-based alloys have been well developed and proven to be the most efficient catalysts, the excessive price, scarcity, and the low durability limit their large-scale applications [4]. Therefore, inexpensive and efficient electrocatalysts are urgently needed to enter into widespread commercialization in energy sectors, such as energy-based transportation.

Iron-based carbon materials, as non-precious metal-based catalysts (NMPCs), exhibit high electrocatalytic activity in ORR and high chemical stability in both alkaline and acidic conditions [5]. Nevertheless, most of the highly efficient Fe-based catalysts often contain various Fe species (e.g., Fe–Nx, Fe/Fe3C, Fe3O4@C) [6, 7]. In particular, the active sites, such as FeN4, which are embedded in the sp2 carbon lattice or located at the edges bridging the two graphene planes, have shown great potential, based on both theory and experimental data from X-ray absorption, X-ray photoelectron, and Mössbauer spectroscopy [8, 9]. Rational design and preparation of Fe-based electrocatalysts with homogeneous dispersion of the active sites are highly encouraged aiming to understand the active sites and thus to maximize the catalytic efficiency of the iron atom [10,11,12].

Increasing the number of available active sites and the intrinsic activity of the active site are the two main strategies promoting the increase in catalyst activity [13]. Hierarchically porous carbon (HPC) is widely used in the fields of gas adsorption, gas storage, heterogeneous catalysis, supercapacitors, and batteries [14,15,16]. The HPC, exhibiting a good electrical conductivity, acts as a suitable coordination environment for the iron atoms in which the distribution of Fe sites can be highly homogeneous, even at atomic scale [10, 17]. More active sites are exposed by optimizing HPC structure. The transport of electrons between metal atoms/nanoparticles and a suitable supports results in the charge redistribution, promoting the adsorption of oxygen molecules and reaction intermediates, and thus achieving a highly efficient oxygen reduction reaction [18]. Over the past decade, great efforts have been made to develop HPC supports of various morphologies [19,20,21,22,23]. Abundant studies have revealed that heteroatom doping (e.g., N, B, S, P) can change the electronic structure of the doped carbon material, making the oxygen reduction reaction easier [24,25,26]. Heteroatom doping can also change the electronic structure of Fe/N/C and has a positive effect on ORR performance [27]. Recently, Mu’s group verified the influence of intrinsic pentagon defects in the basal plane on the local electronic redistribution and the contraction of bandgap by theoretical calculations and experiments, which has an impact on the binding affinity of carbon matrix and electrochemical performance [28]. The pentagon defects may have a certain influence on the electronic distribution of Fe–N4; therefore, it is meaningful to explore the effect of the pentagon carbon on the Fe–N4 active center.

Herein, we report a facile dual-templating approach to synthesize a hierarchically structured porous carbon (Fe/N/C-DT) with Fe–N4 active sites by incorporating ferrocenecarboxaldehyde (Fc–CHO) into the polyimide molecular chain directly. Ten nm of silicon dioxide and zinc acetate are used as dual templates. When the temperature was raised to be about 180 °C, Zn(CH3COO)2 was decomposed to be zinc oxide, and then was reduced to be Zn through further annealing at 900 °C. The formed reduced zinc will be evaporated at this high temperature (Eq. S1) [19]. The following HF etching will easily remove SiO2 template. The prepared Fe/N/C-DT exhibits HPC with carbon nanotube structure, which will provide a larger surface area and more active sites. The iron precursor is a stable ferrocene derivative that can maintain an iron center that has the saturated electronic structure of krypton up to 400 °C [29], which may offer a steric hindrance for iron atoms distribution. In Fe/N/C-DT, some inevitable iron nanoparticles induced SiO2 to grow CNTs during pyrolysis, which provides a high mesoporosity and higher conductivity, which enhanced the catalytic activity. The ferrocene with two cyclopentadienyl ligands preferably forms a pentagon carbon around Fe–N4 after calcination. And by comparing with FeCl2/N/C-DT and FeSO4/N/C-DT, the advantage of Fc–CHO as an iron source is highlighted. This is the first report on the incorporation of ferrocene derivatives into the PI polymer aiming to synthesize highly dispersed Fe–Nx/C NMPCs catalysts, and through the calculation and experiments, the synergistic effects of pentagon carbon and Fe–N4 on the catalytic performance were explored.

Experimental

Chemicals

NMP (1-methyl-2-pyrrolidinone), HF (hydrofluoric acid), FeCl2·4H2O, and MA (melamine) were purchased from Sinopharm Chemical Reagent Co., Ltd., and used without any further purification. Fc–CHO (ferrocenecarboxaldehyde), PMDA (pyromellitic dianhydride), zinc acetate, FeSO4·7H2O, and silicon dioxide were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd.

Electrocatalyst preparation

Preparation of Fe/N/C-DT PMDA (3.9658 g, 18 mmol), MA (1.5861 g, 12.45 mmol), Fc-CHO (0.2948 g, 1.35 mmol), and [zinc acetate (0.7 g), 10 nm silicon dioxide (0.45 g)] were added to a 100-mL two-neck round-bottom flask equipped with a reflux condenser. After the addition of 30 mL NMP, the system was degassed by bubbling with nitrogen for 30 min. The homogeneous mixture was refluxed under inert atmosphere at 180 °C for 4 h. The reaction was initially accompanied by a white smoke, and the end product was a black gel. The system was cooled at room temperature under inert atmosphere. The black gel was dried under vacuum at 140 °C for 10 h to obtain a black solid. The black product was simply ground to form a powder, which was poured into a porcelain crucible and placed in a tube furnace. After that, the material was heated at 240 °C with a rate of 5 °C min−1 in N2 for 2 h to achieve the complete imidization between melamine and PMDA. The intermediate product was subjected to a second pyrolysis treatment at 900 °C (rate of 5 °C min−1) in N2 for 2 h. After the tube furnace was cooled at room temperature, the obtained carbon material was taken out to the centrifuge tube. Then, the silica template was removed by soaking in hydrofluoric acid (30 mL) overnight at room temperature. The solid was washed ten times with deionized water to remove the residual impurities. The final product was dried in vacuum oven at 80 °C for 10 h to obtain a dark powder (1.5206 g, yield of 26%). The catalyst was denoted as Fe/N/C-DT. CHNSO element analysis and ICP analysis: Fe 0.565%, N 2.82%, C 85.77%, H 1.254%, Zn 0.093%.

Preparation of Fe/N/C and N/C-DT For the synthesis of Fe/N/C, the protocol was similar to that for Fe/C/N-DT except the use of the dual template. For the synthesis of the sample without iron (Fc–CHO), other steps were unchanged, but the second pyrolysis step was performed at 900 °C for 2 h. The sample was labeled as N/C–DT.

Preparation of FeSO4/N/C-DT and FeCl2/N/C-DT In the synthesis of these two samples, Fc–CHO is replaced by FeSO4·7H2O and FeCl2·4H2O on an equimolar basis, without changing the protocol used for Fe/N/C-DT. The samples were labeled as FeSO4/N/C-DT and FeCl2/N/C-DT, respectively.

Materials characterization

Scanning electron microscopy (SEM) images were recorded on a JEM-7500F. Transmission electron microscopy (TEM) was performed on a JEM-1400Plus at 120 kV. High-resolution transmission electron microscopy (HRTEM) images and elemental mapping were performed on a Talos F200S. Fourier transform infrared spectroscopy (FTIR) was performed on a Nicolet 6700 FTIR spectrometer. Thermogravimetric analysis was performed on a STA449C analyzer in the temperature range from 20 to 1000 °C The inductively coupled plasma optical emission spectrometry (ICP-OES) measurement was conducted on Prodigy 7 elemental analyzer. Elemental analysis (CHNSO) was performed on a Vario EL cube. Powder X-ray diffraction (PXRD) patterns were recorded on a RU-200B/D/MAX-RB diffractometer from 10º to 80º. Raman spectra were recorded on an inVia Renishaw Raman microscope. X-ray photoelectron spectroscopy experiments (XPS) were conducted on a Thermo ESCALAB 250XI. N2 adsorption/desorption measurements were measured on ASAP 2020 M fully automatic surface area and porosity analyzer.

Electrocatalytic testing

Electrochemical performance measurements of all catalysts were performed on a CHI660E electrochemical workstation. Reversible hydrogen electrode (RHE) was used as the reference electrode, carbon electrode was used as the counter electrode, and the glassy carbon electrode (5.0 mm diameter) coated with the catalyst was used as the working electrode. The working electrode was polished with alumina powder of different particle sizes, then washed with deionized water, and dried at room temperature. The test temperature was controlled at 25 °C. For the electrochemical measurement in alkaline electrolytes, N2/O2-saturated 0.1 M KOH solution was used. The catalyst ink was prepared by dispersing 6 mg catalysts into a mixed solution of 1 mL of water/ethanol (Vwater/Vethanol = 9:1) and 20 μL of Nafion (5 wt%) in an ultrasonic ice bath. Then, 22 μL of as-prepared slurry was dropped onto the surface of the polished glassy carbon electrode, and let it dry for 30 min at room temperature. As for the commercial Pt/C ink, 3 mg Pt/C dispersed in the same solution and 10 mL ink were coated on the surface of RDE.

Zn–air battery tests

The liquid Zn–air battery was constructed of carbon paper (as the air cathode), Zn foil (as the anode), and electrolyte (a mixed aqueous solution of 6 M KOH and 0.2 M Zn2+). The Zn foil (10 × 20 mm) was washed with ethanol. The carbon paper (10 × 20 mm) was coated with about 1 mg cm−2 catalyst on an area of 10 × 10 mm. A home-made all-solid-state Zn–air battery with three components was fabricated (Fig. 5d). The PVA membrane was prepared as follows: 4.5 g PVA powder was added to 45 mL deionized water and stirred at 95 °C for at least 40 min. After the mixture was cooled at room temperature, it turned into a transparent gel. Then, 4 mL of 18 M KOH solution and 0.2 M zinc acetate solution were added. After stirring for 30 min, the transparent gel was poured onto a suitable glass dish and freeze-dried in a refrigerator overnight. After heating at 60 °C, a PVA membrane of a 1 mm thick and dimension of 25 × 25 mm was cut. The catalyst (2 mg cm−2) was added dropwise into a 20 × 20 mm central region of the membrane.

Results and discussion

Morphology and structural characterization

A typical route of the synthetic process of Fe/N/C-DT is shown in Fig. 1a. PMDA (pyromellitic dianhydride) reacts with –NH2 in MA (melamine) to form PI (polyimide) [30]. Fc–CHO is then incorporated into the network of the resulted PI in a catalyst free manner and based on the Schiff base chemistry [31]. Fourier transform infrared spectrometry was used to analyze the ferrocene-based polymer structure. In the FTIR spectrum of Fc–MA–PMDA (denoting the synthesized Fe/N/C-DT without pyrolysis and dual-template) (Fig. S1), the characteristic bands at 3469 and 3419 cm−1 of the free amine group in MA and the stretching band of aldehyde groups in Fc–CHO at 1679 cm−1 are invisible. The two bands at 1770 and 1859 cm−1 ascribed to the stretching vibrations of the carbonyl in phase and out of phase in PMDA are not visible, as well. Instead, Fc–MA–PMDA shows characteristic vibration bands of imine ring obtained by the reaction of –NH2 in MA with the anhydride group in PMDA, which appeared at 1787, 1729, and 1360 cm−1. These bands can be ascribed to the in-phase (or symmetric) stretching vibration of two carbonyl groups on the five-membered imine ring of polyimide, corresponding to the reversed (or asymmetric) carbonyl stretching vibration and strong stretching vibration of C–N–C, respectively. Because of the Schiff base chemistry reaction, new stretching vibration bands corresponding to the secondary amine (N–H) and methylimine group (–C=N–) were observed at 3430 and 1625 cm−1, respectively, suggesting the successful condensation reaction of Fc–CHO with MA. In summary, the disappearance of the original functional groups (–NH2, –CHO, carbonyl) and the formation of the new groups (–NH–, imine ring) indicate the successful synthesis of the target product.

Figure 1
figure 1

a Schematic illustration of the synthesis of Fe/N/C-DT; b SEM image, c high-resolution TEM (HRTEM) image, d TEM images of Fe/N/C-DT with e HAADF-STEM and EDS mapping images

Full size image

Thermogravimetric analysis was performed in a N2 atmosphere between 25 and 1000 °C with heating rate of 5 °C min−1. The recorded TG curve is displayed in Fig. S2 and demonstrates the evolution mechanism of Fc–MA–PMDA–DT (denoting the synthesized Fe/N/C-DT without pyrolysis). TG curve can be divided into several ranges of mass loss. The weight loss of 18% in the temperature range of 100–200 °C is due to the evaporation of the residual solvent NMP (1-Methyl-2-pyrrolidinone) adsorbed in the porous structure and decomposition of zinc acetate. Due to the imidization process and further crosslinking of the polymer, an obviously weight loss of 9% appeared in the range from 200 to 300 °C. A sharp weight loss of 47% occurred in the temperature range from 300–620 °C attributed to the decomposition and carbonization of polymer. Further graphitization leads to a weight loss at the temperature from 620–900 °C. A small weight loss was also observed above 900 °C due to the volatilization of zinc template.

The as-prepared Fe/N/C-DT showed obvious graphite-like structure with a hierarchically porous structure, as demonstrated by the scanning electron microscopy (SEM) images (Fig. 1b and Fig. S3 a, b). The formation of a hierarchically porous structure in Fe/N/C-DT was also confirmed by the corresponding images of transmission electron microscopy (TEM) images (Fig. 1d and Fig. S3 c, d) and high-resolution TEM images (Fig. S4a). In TEM images, Fe/N/C-DT showed the hierarchically porous structure consisting of uniform mesopores with a diameter of ~ 10 nm and carbon nanotubes with a diameter of 12.2 nm crossing the nanostructure (Fig. S4b). Fe/N/C (denoting Fe/N/C synthesized through similar processes but without a dual template) exhibited a porous structure with fish scale-like mesoporous structure, but significant Fe aggregation appeared and no carbon nanotubes were found (Fig. S5a, b). HAADF–STEM and EDS mappings (Fig. 1e) Fe/N/C-DT showed that C, N, Fe, O are uniformly distributed throughout the nanostructure without traces of aggregation of iron nanoparticles.

To gain further insight into the structure of the pyrolysis products, powder X-ray diffraction (XRD) (Fig. S6a) was used. The diffractogram of N/C-DT (denoting N/C-DT synthesized by a similar processes as for Fe/N/C-DT but without Fc–CHO) displays a peak at 26º, which is assigned to the (002) reflection of graphitic carbon, consistent with the corresponding TEM (Fig. S7b). The X-ray patterns of Fe/N/C and Fe/N/C-DT display a sharper and stronger diffraction peak at 26° and a weak peak at 44°, corresponding to the graphite (100)/(101) reflections of graphitic carbon. These changes in the diffractogram suggest a crystallization process induced by evenly dispersed iron incorporation resulting in a higher graphitic degree since iron atoms is an efficient catalyst for the production of graphene. The high-resolution TEM (HRTEM) image (Fig. 1c) further reveals that Fe/N/C-DT has a high graphitization degree with a lattice spacing of 0.35 nm. This improved graphitization degree will greatly accelerate the electron transfer, and thus the ORR performance will be enhanced [32]. Interestingly, no obvious diffraction peaks assigned to crystalline Fe3C and metallic Fe are observed, indicating a high dispersion of Fe species confined in Fe/N/C-DT. By contrary, the diffractogram of Fe/N/C sample with similar low content of iron displayed the peaks at 42.886°, 43.751° and 44.989°, characteristic to iron carbide phase (JCPDS card: No. 85–1317). As shown in Fig. S6b, displaying the Raman spectra of Fe/N/C-DT and Fe/N/C, D and G bands at 1330 and 1580 cm−1, respectively, the intensity ratio of D band and G band (ID/IG) was used to estimate the defects in graphitic carbon [33]. The ID/IG ratio for Fe/N/C-DT is 1.15, but it decreases to 1.12 in Fe/N/C due to the higher graphitization induced by Fe aggregates during the high-temperature pyrolysis, indicating a larger number of defects in Fe/N/C-DT. A broad 2D peak at 2642 cm−1 and D + G peak at 2900 cm−1 were observed in the spectra of Fe/N/C and Fe/N/C-DT samples, which indicate the simultaneous presence of some layered graphene-like architectures and D band.

The specific surface area and pore-size distribution of the pyrolysis products were investigated by N2 adsorption/desorption measurement (Fig. 2a, b). The N2 isotherms of Fe/N/C-DT and N/C-DT are a combination of type I and IV isotherms, revealing the existence of micropores and mesopores [34]. The BET surface areas of Fe/N/C-DT, N/C-DT, and Fe/N/C are 484, 586, and 35 m2 g−1, respectively. The total pore volume of pores Fe/N/C-DT, N/C-DT, and Fe/N/C are 0.503, 0.673, and 0.097 cm3 g−1, respectively. Pore-size distribution curves show that Fe/N/C-DT and Fe/N/C possess more mesopores about 3 nm than N/C-DT, indicating that the reaction between Fc–CHO and MA introduced a large number of mesopores about 3 nm. Due to the addition of 10 nm SiO2, many mesopores of 10 nm are present in Fe/N/C-DT and N/C-DT, in line with the TEM results (Fig. 1d and Fig. S7a). This hierarchically porous structure will benefit the enhancement of the electrocatalytic activity to ORR and OER. The rich microporosity of the as-synthesized catalyst will expose a large number of accessible active sites for electrocatalytic reaction; while the high mesoporosity is essential for the efficient mass transport in the catalyst layer [20]. Compared with Fe/N/C, Fe/N/C-DT with more active sites and a higher efficient mass transport efficiency will surely realize much improved electrochemical performance.

Figure 2
figure 2

a N2 adsorption–desorption isotherms and b pore-size distribution of Fe/N/C-DT, Fe/N/C, and N/C-DT; c N1s XPS spectra and d Fe2p XPS spectra of Fe/N/C-DT, Fe/N/C, and N/C-DT

Full size image

X-ray photoelectron spectroscopic (XPS) measurements were further carried out to explore the elemental composition and chemical state at the sample surface. XPS survey spectrum of Fe/N/C-DT, Fe/N/C, N/C-DT (Fig. S8a) confirms the coexistence of carbon, oxygen, and nitrogen elements on the surface. On the other hand, ICP and EDS analyses showed a very low content of Fe in Fe/N/C-DT (0.565%) and Fe/N/C (0.65%) (Table S1 and Fig. S9), which explain the very small Fe2p peaks observed in survey spectra of all catalysts. The high-resolution N1s spectra of the three sample (Fig. 2c) were deconvoluted and fitted with four components (i.e., 398.5, 400.1, 401.2, and 403.2 eV) assigned to pyridinic N, pyrrolic N, graphitic N, and oxidized N, respectively. Interestingly, the total content of pyridinic N and pyrrolic N species in Fe/N/C-DT dramatically increased in comparison with Fe/N/C and N/C-DT samples (Fig. S8b), which may improve ORR performance [35]. The high-resolution Fe2p spectra of Fe/N/C-DT and Fe/N/C were deconvoluted and fitted with six peaks (Fig. 2d), which correspond to Fe0 (707.8 eV), Fe2p3/2 (Fe2+, Fe3+ 710.7–714 eV), satellite peak (717.8 eV), and Fe2p1/2 (Fe2+, Fe3+ 728-723.2 eV), demonstrating the coexistence of Fe0, Fe2+, and Fe3+.

Electrochemical performance

The electrocatalytic performances of the as-prepared catalysts were first examined in O2/N2-saturated 0.1 M KOH solution on the rotating disk electrode (RDE) in a typical three-electrode system. As can be seen in Fig. 3a and Fig. S10a, Fe/N/C-DT displays an obvious strong oxygen reduction peak at 0.88 V. The two reduction peaks’ current density at 0.74 and 0.76 V for Fe/N/C and N/C-DT is much lower. Linear sweep voltammetry (LSV) curves illustrated in Fig. 3b show that the half-wave potential of N/C-DT shifts from 0.78 to 0.902 V after the incorporation of Fe species, and the onset potential of Fe/N/C-DT (1.043 V) is 143 mV higher than that of N/C-DT (0.9 V). The Fe/N/C shows an inferior onset potential (Eonset) of 0.9 V and the half-wave potential of 0.731 V (Fig. 3b). Moreover, Fe/N/C-DT exhibits a higher diffusion-limited current density (5.93 mA cm−2 at 0.4 V) than that of Pt/C (5.81 mA cm−2 at 0.4 V). Koutecky–Levich (K–L) plots obtained from the LSV polarization curves at different rotation speeds are shown in Fig. 3c. The electron transfer number was calculated based on the Koutecky–Levich (K–L) equation. The average electron transfer number of Fe/N/C-DT is ~ 4.5, indicating a quasi-four-electron pathway for ORR. Furthermore, a low Tafel slope of 63 mV dec−1 for Fe/N/C-DT (Fig. 3d), compared with that of Fe/N/C (114 mV dec−1), N/C-DT (140 mV dec−1), and 20% Pt/C (66 mV dec−1), indicating that Fe/N/C-DT will show a smaller overpotential at the same current density, further proved that Fe/N/C-DT exhibits superior activity in ORR. Methanol-resistant activity of Fe/N/C-DT was also examined. There is no obvious change in the normalized current on Fe/N/C-DT working electrode after an injection with 9 mL of methanol while Pt/C dropped immediately under the same conditions (Fig. 3e). Amperometric i–t tests were carried out to evaluate the electrochemical stability of the catalysts. Fe/N/C-DT exhibited significantly better stability (retention rate of 90%) than the benchmark Pt/C catalyst (retention rate of 73.6%) after 40000 s (Fig. 3f). We attributed the better performance of Fe/N/C-DT to the porosity of the obtained nanostructure and more active Fe sites.

Figure 3
figure 3

ORR performance of Fe/N/C-DT. a the cyclic voltammetry (CV) curves of Fe/N/C-DT, Fe/N/C, and N/C-DT in 0.1 M KOH saturated with O2 at a sweep rate of 20 mV s−1; b ORR polarization curves of Fe/N/C-DT, Fe/N/C, N/C-DT, and 20% Pt/C at 1600 rpm and a sweep rate of 10 mV s−1; c polarization curves at different rotating rates (800–2400 rpms) at a sweep rate of 10 mV s−1 for Fe/N/C-DT (inset: corresponding K–L plots(J−1 versus W−1/2) at various potentials); d Tafel plots of Fe/N/C-DT, Fe/N/C, N/C-DT, and 20% Pt/C derived by the mass transport correction from the corresponding LSV data; e methanol-resistant test of Fe/N/C-DT and 20% Pt/C with 9 mL methanol injection; f The i–t chronoamperometric responses of Fe/N/C-DT and 20% Pt/C at a constant voltage of 0.6 V (inset: retention rate after 40000 s)

Full size image

As shown in Fig. S10b, the OER (oxygen evolution reaction) performances of Fe/N/C-DT in 1 M KOH solution were also tested using IrO2 as a comparison. The operating potential delivering a current density of 10 mA cm−2 (Ej=10) has usually been described as the OER activity benchmark [36]. The potential of Fe/N/C-DT (1.66 V) at 10 mA cm−2 is just 77 mV higher than that of the commercial IrO2 catalyst (1.58 V). The Tafel slope of Fe/N/C-DT is 170 mV dec−1, which is higher than that of IrO2 (56 mV dec−1) (Fig. S10c). Nevertheless, the OER performance of Fe/N/C-DT is better than most of the previously reported NMPCs electrocatalysts (Table S2).

To further confirm the positive effect of Fc-CHO as an iron source for electrocatalysis, FeCl2/N/C-DT and FeSO4/N/C-DT using FeCl2 and FeSO4 as iron sources, respectively, were prepared for comparison. It was found that FeCl2/N/C-DT, and FeSO4/N/C-DT all exhibit almost the same hierarchically porous structure with the interpenetrated carbon nanotubes as Fe/N/C-DT. However, the carbon nanotube diameters are 12.2 nm for Fe/N/C-DT, 59 nm for FeCl2/N/C-DT, and 151 nm for FeSO4/N/C-DT (Fig. S4b, S11). It is well known that the diameter of CNTs is directly proportional to the size of the iron catalyst particle, which can indirectly confirm the positive effect of Fc–CHO on the uniform dispersion of iron atoms [37], and smaller carbon nanotubes diameters can carry more FeNx active sites [38]. It is attributed to the steric hindrance effect of ferrocene in the polymer, which can largely alleviate the severe iron agglomeration. Since the structure and composition of the three samples are very similar, there are many similarities on XRD and XPS analysis (Fig. S12, S13, Fig. 4b). However, since the agglomeration of iron nanoparticles in FeCl2/N/C-DT and FeSO4/N/C-DT is more pronounced with respect to Fe/N/C-DT, the number of carbon nanotubes produced by iron-catalyzed silicon is higher and the diameter of the carbon nanotubes is larger, which indirectly leads to a decrease in the ID/IG. The ID/IG ratio decreased from 1.15 for Fe/N/C-DT to 1.05 and 1.02 for FeCl2/N/C-DT and FeSO4/N/C-DT, respectively, as shown in Fig. 4a. This indicates that Fe/N/C-DT exhibits higher defect density and more active sites than FeCl2/N/C-DT and FeSO4/N/C-DT.

Figure 4
figure 4

a Raman spectra, b N1s XPS spectra and c Fe2p XPS spectra of Fe/N/C-DT, FeCl2/N/C-DT, and FeSO4/N/C-DT. d The cyclic voltammetry (CV) curves, e LSV curves, and f the corresponding Tafel plot of Fe/N/C-DT, FeSO4/N/C-DT, and FeCl2/N/C-DT in O2/N2-saturated 0.1 M KOH

Full size image

Although the iron content is low and the noise of Fe2p XPS spectra is large, a series of smooth fitting curves of Fe/N/C-DT, FeSO4/N/C-DT, and Fe/N/C by the Savitzky–Golay smoothing method as shown in Fig. 4c. The signal at around 710 eV can be ascribed to Fe in Fe–Nx configuration [39]. For the FeSO4/N/C-DT and Fe/N/C samples, a significant positive shift of the signal of Fe–Nx was noticed. Higher binding energy means lower charge density on central Fe atoms, and large degree of interaction between Fe–Nx and Fe3C leads to a decrease in the charge density of central Fe atom [6]. From this point of view, it can be proved that the degree of interaction between Fe–Nx and Fe3C in Fe/N/C-DT is not as prominent as in FeSO4/N/C-DT and Fe/N/C (perhaps because of the low content of Fe3C, it is not easy to be detected in XRD), so the use of ferrocene and the dual-templating approach facilitates the formation of highly dispersed electrocatalysts, even at the atomic scale.

As can be seen from cyclic voltammetry (CV) curve (Fig. 4d), electrocatalysts with three different iron sources displayed a well-defined and similar ORR peak at potentials between 0.80 and 0.96 V in 0.1 M KOH. The linear sweep voltammetry (LSV) curves tested on RDE (Fig. 4e) show that the onset potentials of the three materials are nearly the same, but the half potential of Fe/N/C-DT is 26 mV higher than those of FeCl2/N/C-DT and FeSO4/N/C-DT. Fe/N/C-DT (63 mV dec−1) also shows a significantly lower Tafel slope than those of other two reference catalysts (75 and 72 mV dec−1, respectively) (Fig. 4f). Just as heteroatoms (N, B, O, P, S) doped with carbon, pentagon defects can also cause electron modulation to change the charge distribution and electronic properties of carbon skeletons, which in turn affects their interaction with oxygen molecules and reaction intermediates for ORR [24,25,26]. Grafting the iron source to the polymer makes the active sites in the synthesized catalyst more uniformly dispersed. Under the same iron content, compared to FeCl2/N/C-DT and FeSO4/N/C-DT, Fe/N/C-DT has higher catalytic activity, undoubtedly.

DFT calculation and performance of Zn–air batteries

Calculations based on density functional theory (DFT) in Dmol3 code were performed on the four theoretical model structures (Fig. S14a, b, c, d), irregular graphene with all the common hexagon (C6), irregular graphene with a Fe–N4 site, irregular graphene with a pentagon defect carbon on the edge, and irregular graphene with a Fe–N4 site and a pentagon defect carbon (labeled as I-Gra, I-FeN4, I-Pen, and I-FeN4-Pen, respectively). The adsorption energies of oxygen, Eads, of the four theoretical models showed that I-FeN4-Pen exhibits a higher oxygen adsorption energy (3.265 eV) than that of I-Gra (1.034 eV), I-FeN4 (2.340 eV), and I-Pen (1.496 eV) (Fig. S14e, f, g, h, Fig. S15). The introduction of a pentagon defect-rich carbon can enhance the energy of O2 adsorption, and the synergy between the pentagon carbon and Fe–N4 strongly enhances the adsorption energy than Fe–N4 active site alone. These results reveal that I-FeN4-Pen accelerates the oxygen adsorption process and thus achieves an outstanding oxygen reduction performance [2, 40].

Knowing that Fe/N/C-DT presents an outstanding activity in ORR and considerable OER performance, and an O2 electrode potential gap △E(△E = Ej=10,OER − E1/2,ORR) of 0.758 V, the possibility to use the oxygen electrodes for the reversible redox reaction is obvious [41]. Liquid Zn–air battery and all-solid-state Zn–air battery are assembled separately. In the liquid Zn–air battery, catalysts are loaded on the carbon paper acting as an air cathode, Fe/N/C-DT and the Pt/C catalysts were carried out under the same conditions. Fe/N/C-DT displays a higher open-circuit voltage (1.48 V) than that of Pt/C (1.41 V) (Fig. S16). In discharge polarization curves and power density curves (Fig. 5a), Fe/N/C-DT displays a peak power density of 220 mW cm−2, which far exceeds that of Pt/C (125 mW cm−2). By replacing the zinc anode and electrolyte after a continuous discharge current density of 5 mA cm−2 for 20 h, the same electrode continues discharging at 20 mA cm−2 for 20 h. The output voltage of Fe/N/C-DT at a discharge current density of 5 and 20 mA cm−2 is more stable and higher than that of Pt/C (Fig. 5b). Due to the small potential gap, △E, a discharge–charge test at 5 mA cm−2 was conducted. A gradual increase in discharge–charge potential gap is observed for both Fe/N/C-DT air electrode and Pt/C air electrode (Fig. 5c). However, Fe/N/C-DT air electrode shows much better rechargeability. An all-solid-state Zn–air battery was constructed with Fe/N/C-DT catalyst coated nickel mesh as the air electrode (Fig. 5d). Remarkably, Fe/N/C-DT exhibits a high open-circuit voltage of 1.451 V. In addition, three such batteries in series can illuminate a 3-V light-emitting diode (LED) lamp.

Figure 5
figure 5

Performance of “rechargeable” Zn-air batteries. a Discharge polarization and power density curves of Fe/N/C-DT and Pt/C; b The galvanostatic discharge curves of the liquid Zn–air batteries at 5 mA cm−2, followed by discharge at 20 mA cm−2. c Galvanostatic discharge–charge cycling curves at a current density of 5 mA cm−2 with 400 s cycle. d Schematic diagram of the all-solid-state zinc–air, a high open-circuit voltage of 1.451 V, three batteries in series illuminated a 3-V red and blue LED lamp

Full size image

Conclusions

In summary, we successfully developed a Fe/N/C-DT electrocatalyst via pyrolysis Fc–MA–PMDA polymer. The material was prepared by the dual-templating approach and incorporating the Fc–CHO to the PI polymer chain directly. The hierarchically porous carbon structure, resulted from the dual-templating approach, facilitated the uniform dispersion of the iron nanoparticles or even single Fe atom dispersion. The well-dispersed iron nanoparticles and the well-conducting carbon support promoted the electron exchange at the interface of metal atoms/nanoparticles and carbon supports, improving the adsorption energy of oxygen and the intermediate product, which further enhanced the electrocatalytic activity of the oxygen. DFT calculations showed that the synergistic effect of pentagon carbon and Fe–N4 had a positive effect on catalyst performance in ORR. The prepared Fe/N/C-DT exhibited excellent ORR activity, superior stability, and impressive OER performance in alkaline media. Fe/N/C-DT driven by the liquid Zn–air battery possessed a higher peak power density, a higher open-circuit voltage, and a more stable discharge than those of Pt/C. And all-solid-state Zn–air battery showed a high open-circuit voltage and feasible practical application. The novel design implying the incorporation of iron source with steric hindrance effect into the polymer could inspire the design of new electrocatalysts with the active sites dispersed at the atomic scale and boost the exploration of the impact of the pentagon defect-rich carbon on Fe–N4.