Enhanced oxygen reduction from the insertion of cobalt into nitrogen-doped porous carbons

Abstract

This work addresses unprecedentedly enhanced electrocatalytic activity of oxygen reduction reaction (ORR) by potassium hydroxide (KOH) chemical activation and cobalt doping of nitrogen-doped porous carbon (NPC). Polyacrylonitrile (PAN) and cobalt acetate are used as carbon–nitrogen and cobalt precursors to synthesize the cobalt-doped NPC. The KOH activation primarily enhances the activity of NPCs in the ORR by making them highly porous and forming pyrrolic nitrogen in the carbon network. It is the subsequent cobalt doping process that not only induces more transitions of pyridinic to pyrrolic nitrogen but also generates new active sites of oxide and carbide forms (Co₃O₄ and Co_xC) on the NPCs. A further reduction of hydrogen peroxide can occur on the cobalt-doped NPCs leading more favorably to the four electron pathway. The cobalt-doped NPC has an ORR activity comparable to the commercial Pt catalyst, but it shows higher stability and greater tolerance to methanol poisoning.

---

1. Introduction

As a pollutant-free energy conversion device, a fuel cell has a potential to provide sustainable and portable energy to modern society. For the successful implementation of a fuel cell, effective electrocatalysts for the oxygen reduction reaction (ORR) in the cathode are required due to the slow rate of ORR.¹ Platinum (Pt) and its composites are primarily used for ORR.^2,3 However, they have some obstacles to achieving large-scale commercialization arising from high costs, unsustainable supply and poor durability of Pt. Therefore, many studies have been devoted to finding alternative non-precious metal catalysts and metal-free boron or nitrogen-doped (N-doped) carbon catalysts that have higher durability and lower costs than Pt-based catalysts.^4–10

Among them, hybrids of N-doped carbon and metal oxide catalysts have shown remarkable and interesting performance as alternatives to Pt and its alloys.^11,12 In fact, there are some reasons why their combination is attractive. N-doped carbons themselves can compete with Pt-based catalysts because the existence of nitrogen in the carbon networks effectively increases the number of active sites.^13–16 However, their ORR activity is based on the two-electron oxygen reduction pathway which makes ORR less efficient in fuel cells.¹⁷ In addition, hydrogen peroxide produced by the two-electron oxygen reduction pathway can poison the entire system. Thus, it is necessary to introduce metal oxides to achieve the four-electron path way, which leads to the circumvention of the hydrogen peroxide production. The use of metal oxides is cost effective, but they provide low electrical conductivity. This can be a limitation to the ORR activity but can be overcome by the insertion of nitrogen atoms to the carbon network.¹¹ Therefore, with a hybrid of N-doped carbon and metal oxide, it is possible to compensate for the deficiency of each component and obtain a synergistic effect which can enhance the ORR activity.

This study also focused on the synergistic effect based on hybrids of N-doped carbon and metal oxide. However, herein, we demonstrate our first attempt to elucidate fundamental insights into the enhanced electrocatalytic activity arising from the interaction between nitrogen functional group and cobalt oxide in the highly porous Co–N-doped carbon produced by an unprecedented synthesis route consisting of potassium hydroxide (KOH) activation and subsequent cobalt impregnation. Polyacrylonitrile (PAN) and cobalt acetate (CoAc) are used as precursors for N-doped carbon and cobalt. In the case of PAN, several research groups have investigated its potential as an electrocatalyst for ORR.^18–24 These prior studies on PAN as an electrocatalyst for ORR can be divided into three categories with respect to the synthesis routes: pyrolysis with both carbon support and metal salts,^18–20 formation of PAN-nanowires by electrospinning^22–24 and ordered mesoporous carbon through a silica template.²¹ Although they suggested novel approaches, the overall synthesis steps are complex, and limitations in ORR performance have been revealed. These kinds of limitations are attributed to hydrogen peroxide generation,^14–16 insufficient specific surface area and inefficient active sites.^17–20

Therefore, this study employed facile KOH activation of PAN to produce highly porous N-doped carbon in a single-step process to fully utilize the nitrogen-doped porous carbon (NPC) as a template for cobalt doping. We showed here that KOH activation causes the pore distribution of NPC to extend from micropores to mesopores which is ideal for accommodating new active sites of cobalt oxide. To decouple the synergistic effect arising from the hybridization of NPC and cobalt oxide, we first investigated the contribution of KOH activation of PAN to the ORR activity because the NPC activated by KOH has rarely been considered as an electrocatalyst except for the application of supercapacitor electrode.²⁵ Due to the relative amounts of KOH used for the activation of PAN, the morphological characteristics and existing chemical bonds of the resulting NPC had different aspects, which resulted in different electrocatalyst performances in ORR. In addition, we increased the ORR activity and stability by impregnating cobalt oxide onto the KOH-activated NPC. We investigated how the synergistic effect of the hybridization of NPC and cobalt oxide is reflected in the ORR performance using electrochemical and microscopic measurements. Specifically, we elucidated how cobalt doping leads to enhanced ORR activity by inducing the formation of cobalt oxide (Co₃O₄) and cobalt carbide (CoC_x), and the transition of pyridinic nitrogen to pyrrolic nitrogen. Therefore, this study presents a new avenue for utilizing PAN, which is a cost-effective and non-conductive polymer, as an efficient ORR catalyst by introducing both KOH activation and cobalt doping.

2. Experimental

2.1 Materials

Polyacrylonitrile (PAN; average molecular weight: 150 000), KOH, hydrochloric acid (HCl, 37 wt% in water), platinum on graphitized carbon (20 wt% loading) and Nafion® perfluorinated resin solution (5 wt% in lower aliphatic alcohols and water; contains 15–20% water) were purchased from Sigma-Aldrich. Ethanol (with a purity of >99.5%) was obtained from Samchun Chemicals. Cobalt acetate (CoAc) was acquired from Junsei Chemicals.

2.2 Preparation of NPCs

The preparation process for the NPCs can be divided into two steps: stabilization and KOH activation. First, about 2–4 g of PAN were stabilized at 280 °C for 90 min under air (100 mL min⁻¹). Following the stabilization step, the stabilized PAN was mechanically mixed with KOH at a mass ratio of KOH to PAN (KOH/PAN ratio) equal to 0.0 (p-PAN), 1.0 (PANK1), 2.0 (PANK2) and 3.0 (PANK3), respectively. Then, the mixtures were carbonized at 750 °C for 1 h under argon (Ar) (50 mL min⁻¹). Finally, to remove impurities, the synthesized samples were purified by washing and filtering with 5 M HCl, followed by deionized (DI) water and ethanol washing. After the purification, the samples were dried in a vacuum oven at 120 °C.

2.3 Preparation of hybrid catalysts with NPC and cobalt oxide

To synthesize hybrids of NPC and cobalt oxide, each NPC sample was mixed with CoAc (20 wt% of the NPC sample mass) and 10 mL of DI water in a vial. Then, to impregnate the CoAc, ultra-sonication was applied to the mixed sample for 30 min, and the mixture was dried in a vacuum oven at 120 °C. Each dried sample was heat-treated at 850 °C for 1 h under Ar (50 mL min⁻¹). After the heat treatment, the sample was washed in DI water at 90 °C several times and then was dried again in a vacuum oven at 120 °C. The resulting cobalt-doped NPCs were named PANKCo1, PANKCo2 and PANKCo3, based on the starting NPCs PANK1, PANK2 and PANK3, respectively.

2.4 Characterization

The Brunauer–Emmett–Teller (BET) specific surface area and the pore size distribution (PSD) were obtained by nitrogen adsorption at 77 K with three-flex surface characterization (Micromeritics). The microstructural characterization was monitored by scanning electron microscope (SEM, a Magellan 400 UHR-SEM) at 1–2 kV. The X-ray diffraction (XRD) was conducted on D/MAX-2500 with Cu Kα (λ = 0.15418 nm), set at 40 kV and 300 mA. Transmission electron microscopy (TEM) was carried out on a Tecnai G2. X-ray photoelectron spectroscopy (XPS) was done with a Sigma probe (Thermo VG Scientific) to analyse the chemical structures of the NPCs and cobalt-doped NPCs. The XPS peaks were fitted with Avantage software (Thermo VG package), in which the deconvolution of peaks was completed with Gaussian (70%) and Lorentzian (30%) curve fitting procedures, and the binding energy was corrected with reference to C 1s at 284.5 eV.

2.5 Electrochemical analysis

Electrochemical experiments consisted of cyclic voltammetry (CV), rotating disk electrode (RDE) and rotating ring-disk electrode (RRDE) measurements. RDE and RRDE were done with a three-electrode system using a RRDE-3A rotating ring disk electrode rotator by ALS. In both cases, Ag/AgCl, platinum wire and glassy carbon (3 mm diameter) were used as reference, counter, and working electrodes, respectively. A potentiometer (biologic) was used to perform these analyses. For the measurements, the NPC or cobalt-doped NPC samples were deposited onto the working electrode. 5 milligrams of sample were dispersed by ultra-sonication for 15 min in 1 mL of solution consisting of DI water, methanol, and Nafion® perfluorinated resin solution at a volumetric ratio of 15 : 4 : 1. Then, 7.5 μL of the solution with a dispersed catalyst was loaded onto the polished rotating ring disk electrode (disk electrode: glassy carbon; ring electrode: platinum) for the RDE and RRDE experiments. Then, the sample was dried in air. The CV, RDE and RRDE measurements were performed in 1 M NaOH electrolyte saturated with oxygen at several voltage scan rates.

3. Results and discussion

3.1 Morphological characteristics

As shown in Fig. 1, the KOH activation contributes to the generation of porous structures in the PANK series (Fig. 1a) in contrast to the non-porous flat structure in p-PAN (Fig. 1b). At higher magnification, it can be seen that the hive-like porous network consists of smaller pores (Fig. S1†). To confirm the high porosity identified in the scanning electron microscopy (SEM) images, Brunauer–Emmett–Teller (BET) analysis was performed. In the nitrogen sorption isotherm plots (Fig. 2a), the N₂ quantity adsorbed is proportional to the KOH/PAN ratio. As the measured amount of N₂ adsorption becomes higher, the specific surface area of the sample gets larger. Thus, PANK3 shows a specific surface area of 3314.15 m² g⁻¹ dramatically increased from 18.12 m² g⁻¹ for the p-PAN (Table 1).

Fig. 1 Scanning electron microscopy images of (a) PANK2 (×15 000) and (b) p-PAN (×15 000).

Fig. 2 (a) Nitrogen sorption isotherms of p-PAN and PANK series and (b) pore size distribution of PANK series based on NLDFT model.

Table 1 Specific surface area and pore volume of p-PAN and PANK series

Sample	BET surface area (m^2 g^−1)	Total pore volume (cm^3 g^−1)	Micropore (<2.0 nm) volume (cm^3 g^−1)	Micropore volume/total pore volume (%)
p-PAN	18.12	0.062	0.0068	11.02
PANK1	2235.74	1.101	0.8899	80.85
PANK2	3055.37	1.715	1.1518	67.15
PANK3	3314.15	2.373	1.1841	49.91

---

The types of isotherm changes are subject to the variation of the KOH/PAN ratio. In Fig. 2a, PANK1 shows a typical Type I isotherm, indicating a microporous material as defined in the IUPAC classification. However, in PANK2 and PANK3, hysteresis loops are identified, which come from capillary condensation phenomena occurring in mesopores. According to the IUPAC classification, these isotherms are defined as a Type IV. Moreover, the hysteresis loop in PANK3 is bigger and clearer than that of PANK2. Therefore, the increase of the KOH/PAN ratio contributes to the dominant formation of mesopores.

The non-local density functional theory (NLDFT) model and Horvath–Kawazoe (HK) method were used to understand the trend obtained in the N₂ sorption isotherms. The plots of the pore size distribution (PSD) based on the NLDFT model are shown in Fig. 2b. With a size under 2 nm (micropores), both PANK2 and PANK3 have a similar distribution pattern. In contrast, in the PANK1 case, most of the pores identified have a size less than 1.4 nm. In addition, pores wider than 2 nm barely exist in PANK1. However, pores greater than 2 nm appear in PANK2 and PANK3. PANK3 has a broader PSD pattern (2.25–5.8 nm) than PANK2 (2.25–4.2 nm). These results are closely associated with the trend in the pore volume based on the HK method. As expressed in Table 1, the contribution of the micropores to the total pore volume gradually decreases and becomes nearly equal to that of the mesopores in PANK3.

These interesting trends in the pore characterization strongly support the SEM image (Fig. 1) of porous PANKs and their N₂ sorption isotherms (Fig. 2). A larger amount of KOH in the activation results in a higher ratio of mesopores to micropores (mesopores/micropores) because more carbons in the PAN react with KOH powders. As seen in the SEM images of the PANK series at a magnification of 15 000 times (Fig. 1a and S2†), the size of the pores from the etching of the KOH activation is proportional to the KOH/PAN ratio. The trend identified in the increasing volume ratio of mesopores to micropores is also in line with the transition of nitrogen configuration in the carbon network, which is later discussed in detail in the compositional characteristic part of this paper.

After investigating the contribution of KOH activation to the PANK series, we performed additional analyses to understand the morphological characteristics caused by cobalt doping. In the SEM image of PANKCo2 (Fig. 3a), a number of spheres, which are considered as cobalt products caused by cobalt doping process, are dispersed well on the NPC support. With SEM analyses, back scattered electron (BSE) image method was applied to effectively visualize the presence of cobalt since the element heavier than carbon is expressed as a brighter dot in the BSE image. As a result, all locations of the cobalt products are identified easily (Fig. 3b and S3†). Furthermore, the energy dispersive X-ray spectroscopy (EDX) result (Table S1†) indicates that the cobalt is clearly incorporated into PANK2. All of the PANKCo series including PANKCo2 retain highly porous structure despite of the impregnation of cobalt (Fig. S4 and Table S2†).

Fig. 3 (a) SEM image of PANKCo2 (×20 000), (b) back scattered electron (BSE) image of PANKCo2 (×20 000), transmission electron microscopy (TEM) image of (c) PANK2 (d) PANK2 (higher magnification) (e) PANKCo2 (f) PANKCo2 (higher magnification).

Next, we used transmission electron microscopy (TEM) images to obtain more detailed morphological information. PANK2 has a clean N-doped carbon support plane (Fig. 3c and d) whereas PANKCo2 has a support plane containing dispersed dark dots (Fig. 3e). The EDX results indicate that the dark dots are cobalt-containing materials (Table S3†). At a higher magnification of PANKCo2 (Fig. 3f), the black dots seem to be encapsulated by the adjacent carbon support, suggesting the existence of interactions between the NPC support and cobalt. To support this fact and identify the chemical state of the cobalt-containing PANKCo2, we carried out XRD and XPS analyses.

As seen in Fig. 4, the X-ray diffraction (XRD) patterns reflect the morphological variation caused by the cobalt doping. The cobalt-based materials (PANKCo2) depicted in the previous electron microscopy images turned out to be cobalt(II,III) oxide (Co₃O₄), based on the peaks located at 31.2°, 36.8°, 44.8°, 55.6°, 59.3° and 65.2°, which is consistent with the crystalline spinel structure¹² of Co₃O₄ as shown in the TEM image (a white circle in Fig. 3f). Interestingly, cobalt carbide (Co_xC) peaks²⁶ labeled with the white squares on the XRD plot were also identified. These carbide peaks show the chemical interactions between cobalt and the carbon support as it would be predicted by the dispersion of cobalt in the carbon lattice (TEM images of PANKCo2 in Fig. 3).

Fig. 4 X-ray diffraction (XRD) patterns of PANK2 and PANKCo2.

3.2 Compositional characteristics

During KOH activation, KOH reacts with carbons, and the etching of carbon frameworks occurs, which is the main mechanism to form pores.^27,28 Therefore, in the case in which a carbon support has nitrogen functional groups, the KOH activation step can cause the loss of nitrogen. The X-ray photoelectron spectroscopy (XPS) results of the PANK series show a decreasing trend in the nitrogen contents (Fig. S5 and Table S4†). As the KOH ratio increases, more carbons participate in the activation, and the loss of nitrogen increases. Compared with PANK1, both PANK2 and PANK3 have lower values for nitrogen contents. However, the decreasing relation between PANK2 and PANK3 is not clear, and they have similar values for the nitrogen and carbon ratio (N/C). Because the XPS technique uses signals from surfaces of 1–2 nm in thickness, EDX, which penetrates deeper into the sample, was conducted to clarify the relation between PANK2 and PANK3. The EDX results show a consistent trend of increasing nitrogen loss with the KOH amount used in the PANK series (Table S5;† PANK1: 2.81%; PANK2: 0.92%; PANK3: 0.56%).

In addition, the chemical state of nitrogen existing in the PANK series was identified by deconvoluting the N 1s high resolution spectra. The signals from each sample can be divided into pyridinic nitrogen (N-6; ∼398 eV), pyrrolic nitrogen (N-5; ∼400 eV), graphitic nitrogen (N-Q; ∼401 eV) and pyridinic-N-oxide (N-PO; 402–403 eV) (Fig. 5).²⁹ The π–π* satellite peak, which indicates delocalized electrons in nitrogen-containing aromatic rings, is introduced to deconvolute the spectra, and it is usually apart several eVs from a main nitrogen peak. Interestingly, the amount of N-5, N-PO and π–π* satellite increases as the KOH amount for the activation becomes higher (Fig. 5 and Table 2) whereas the amount of N-6 decreases. The N-6 bonds are dominant in the stabilized PAN, and the bonds are broken because of the loss of adjacent carbon atoms during KOH activation. As a result, new types of nitrogen configuration such as N-5 appear.^22,24,25 Therefore, PANK3, which has the highest amount of mesopores as mentioned in the earlier morphology section, shows the largest N-5 contents and N-Q/N-6 (Table 2) ratios. These variations in the nitrogen configuration are related to the ORR catalytic activity, which is discussed in the subsequent section of electrochemical analyses.

Fig. 5 XPS N 1s high resolution spectra and deconvolutions of (a) PANK1 (b) PANK2 and (c) PANK3.

Table 2 Nitrogen configuration of PANK and PANKCo series analyzed by XPS N 1s deconvolutions

Sample	N-6 (%)	N-5 (%)	N-Q (%)	N-PO (%)	π–π* (%)	N-Q/N-6
PANK1	33.56	14.89	25.83	18.23	7.49	0.77
PANK2	20.74	15.19	33.25	19.16	11.67	1.60
PANK3	14.56	21.39	23.83	22.75	17.46	1.64
PANKCo1	10.78	35.85	22.78	21.45	9.15	2.11
PANKCo2	14.97	20.64	31.32	23.8	9.27	2.09
PANKCo3	16.08	18.84	25.8	28.69	10.59	1.60

---

Fig. 6 and S6† show that XPS analyses were performed to investigate the effect on the compositional characteristics from cobalt doping. To obtain reliable composition data in the XPS survey scans, 4 well-distributed spots on each sample were chosen and analyzed (Fig. S6d†). From the results, the doping ratio of cobalt is the highest in PANKCo3 at 0.82% (Table S6†). Since the amount of cobalt ions that can permeate the porous carbon support is determined by the surface area of the support, PANKCo3 has the highest cobalt content. In Fig. 6b, the O 1s scan of PANKCo2 has a peak at near 530 eV while that of PANK2 does not show. This peak is attributed to cobalt oxides (Co₃O₄), and the 2p_1/2 and 2p_3/2 peaks that play as markers to identify Co₃O₄ are shown in the Co 2p spectrum. The others (PANKCo1 and PANKCo3) also have similar patterns (Fig. S6e and f†) due to the formation of Co₃O₄. All of the XRD and XPS results clearly support the incorporation of Co₃O₄ into the carbon network. Furthermore, it is striking that cobalt atoms give rise to the conversion of N-6 to N-5 during the heat-treatment step, based on the deconvoluted XPS N 1s high resolution curve (Fig. 6a and S7† and Table 2). Nitrogen transition caused by metallic cobalt has been reported by several groups,^22,30 but this study firstly reports the concurrent formation of cobalt oxide with the nitrogen transition. The effect of cobalt oxide formation on the ORR catalytic activity is discussed in the next section.

Fig. 6 (a) XPS N 1s high resolution spectrum and its deconvolutions of PANKCo2, (b) O 1s high resolution spectra of PANK2 and PANKCo2, (c) Co 2p high resolution spectrum of PANKCo2. The filled square in (b) indicates the existence of Co₃O₄.

3.3 Electrochemical analyses

Based on cyclic voltammetry (CV) results, both KOH activation and the subsequent cobalt doping enhance the electrocatalytic activity of ORR. As seen in Fig. 7a, p-PAN has a peak potential at −0.312 V. If the value is calibrated against the standard reversible hydrogen electrode (RHE), it can be expressed as 0.688 V (refer to the ESI† for the RHE conversion equation). On the other hand, the peak potential of PANK2 is positively shifted to −0.207 V (0.793 V vs. RHE).

Fig. 7 (a) Cyclic voltammograms of p-PAN, PANK2 and PANKCo2 (dotted line: Ar-saturated, solid line: O₂-saturated 1 M NaOH) (b) peak potentials (columns) and current densities at the peak potentials (red dots) (c) polarization curves of p-PAN, PANK2, PANKCo2 and Pt/C come from the rotating disk electrode (RDE) technique@1600RPM (each open circuit voltage is expressed as the starting point) (d) the summary of half-wave potentials (E_1/2).

The increase in the mass ratio of KOH/PAN made the cathodic peak more positive such that PANK1 and PANK3 were −0.216 V (0.784 V vs. RHE) and −0.206 V (0.794 V vs. RHE), respectively (Fig. S8b and c†). The relationship is clear in Fig. 7b, and the origin of the positive shift in the peak potential is discussed later with the polarization curves. Moreover, the amount of current originating from ORR was easily identified by comparing CV graphs in O₂- and Ar-saturation conditions (Fig. 7a). The cathodic current at peak potentials is also proportional to the KOH/PAN ratio like the trend in the cathodic potential peak. In fact, the value goes up from 0.26 (p-PAN), 0.54 (PANK1), and 0.57 (PANK2) to 0.72 mA cm⁻² (PANK3). The increase of the current density indicates an increment in the amount of catalyzed ORR, which is attributed to the number of active sites exposed to oxygen molecules. Higher ratios of KOH to PAN yield higher surface areas (Fig. 2a and Table 1) and more active sites, thus PANK3 has the highest current density among the PANK series.

In addition, cobalt doping causes a positive shift of the cathodic peaks in the PANK series, which indicates the incorporation of Co₃O₄, and its interaction with the N-doped porous carbon support enhances the rate of ORR catalysis. The cobalt-doped materials showed a peak potential at −0.172 V (0.828 V vs. RHE) (PANKCo1), −0.172 V (0.828 V vs. RHE) (PANKCo2) and −0.178 V (0.822 V vs. RHE) (PANKCo3). In the case of Pt/C (20 wt% loading on graphitized carbon), the cathodic peak was located on −0.135 V (0.865 V vs. RHE) (Fig. S7a†). Thus, the peak potential difference between PANKCo series and Pt/C is only about 37 mV. Current densities at the peak potential also increase after the cobalt doping. The result is summarized in Fig. 7b. Especially, PANKCo1 has the most improved current density at 0.97 mA cm⁻², and all of the cobalt-doped materials are superior to Pt/C (0.59 mA cm⁻²) for the current densities. Despite of more positive cathodic peak potential of Pt/C, PANKCo series show better catalytic activity for ORR at each peak potential and at the whole potential range (Fig. 7 and S8†).

To understand the ORR mechanism of the synthesized catalysts, the rotating ring-disk electrode (RRDE) technique was performed. It provides the polarization curve of each sample under steady state conditions caused by the rotating electrode (Fig. 7c and S9†). In the beginning of the polarization curves, ORR occurs more actively in the PANK2 and PANKCo2 cases than in the p-PAN and even in the Pt/C. After −0.1 V, the ORR occurs in all of the samples, and the half-wave potential (E_1/2), which indicates the starting point of the ORR, is presented in Fig. 7d. The p-PAN shows a poor half-wave potential at −0.318 V (0.682 V vs. RHE), and the PANK series do improve the half-wave potential. As identified in the CV result (Fig. 7b), the catalytic activity of ORR is proportional to the mass ratio of KOH, based on the trend in the half-wave potentials. The relation arises from the larger amount of graphitic nitrogen (N-Q) than that of pyridinic nitrogen (N-6) and the increase in π–π* electrons (XPS results in Fig. 5 and Table 2). The positive effect of a high ratio of N-Q to N-6 on the ORR activity has been reported by several prior studies.^31–33 With this positive effect, high surface areas of the PANK series assist in effectively exposing more active sites to oxygen molecules. Therefore, the electrocatalytic activity of PANK2 and PANK3 is enhanced considerably, compared to that of p-PAN. Furthermore, the surface area contributes to increasing the current density. In the polarization curves (Fig. 7c and S9†), the current density at −0.3 V (0.7 V vs. RHE, which is a common half-cell cathodic potential of a fuel cell³⁴) increased from 0.46 (p-PAN), 2.01 (PANK1) and 2.65 (PANK2) to 2.98 mA cm⁻² (PANK3).

The effect of cobalt doping, previously confirmed in the CV results, was verified in the polarization curves (Fig. 7c and S9†) as well. In all of the cobalt-doped materials, the half-wave potential becomes more positive, and PANKCo1 shows the most positive half-wave potential, 0.839 V vs. RHE. In a view of half-wave potential, PANKCo series show slightly higher values than the cobalt oxide–graphene oxide composite catalyst in the previous study (0.83 V vs. RHE),³⁴ but have lower half-wave potentials than that of Pt/C (0.857 V vs. RHE). In terms of the current density at −0.3 V (0.7 V vs. RHE), PANKCo2 has the larger current density (4.09 mA cm⁻²) than that of Pt/C (3.30 mA cm⁻²). PANKCo1 also has larger current density (Fig. S9a;† 3.84 mA cm⁻²) than the current density of Pt/C. Based on the summarized results of previous studies,³³ the PANKCo series show better performance overall than other precious and non-precious metal catalysts in ORR catalytic activity^33,35–41 (Table 3).

Table 3 Comparison of ORR catalytic activity of PANKCo2 with other precious metal and non-precious metal catalysts in previous literatures

Catalyst	EORR (V) at J = −3 mA cm^−2	Reference
PANKCo2	0.813	In this study
Co/N–C-800	0.74	33
Ir/C (20 wt%)	0.69	35
Ru/C (20 wt%)	0.61	35
Mn-oxide	0.73	35
LaNiO3/NC	0.64	36
NiCo2S4@N/S–rGO	0.76	37
Co3O4/2.7Co2MnO4	0.68	38
NiCo2O4/G	0.54	39
NiCo2O4	0.75	40
NCO–N1	0.72	41

---

Most notably, cobalt doping (the formation of Co₃O₄) affects the structure of N-doped carbon support, especially nitrogen containing groups. In Table 2, the N-Q/N-6 ratio increases after cobalt doping (PANK1: 0.77 → PANKCo1: 2.11; PANK2: 1.60 → PANKCo2: 2.09), and the ratio does not change in the PANKCo3 case, probably due to the low concentration of nitrogen (0.56%). With these positive effects, the formation of new active sites as Co₃O₄ causes the enhanced ORR activity in the PANKCo series. Interestingly, the transition in nitrogen configuration was not considered as the factor to improve ORR activity in related previous studies.^33,34,42,43 Therefore, the formation of Co₃O₄ accompanying the transition of nitrogen configuration is the main mechanism for the origin of enhancement of ORR activity.

To exclude the effect of different cobalt impregnation temperatures between PANK2 (750 °C) and PANKCo2 (850 °C), the control group named PANKCG2 was prepared by heating PANK2 at 850 °C without the cobalt precursor. CV was performed to evaluate the ORR catalytic activity of PANKCG2. As seen in the CV graphs (Fig. S10†), PANKCG2 has a cathodic peak at 0.793 V (vs. RHE) which is equivalent to PANK2. In other words, the enhanced ORR catalytic activity of PANKCo2 is attributed to the cobalt doping but not to the thermal effect.

In addition, from the RRDE analyses, information on the variation of the current on ring and disk electrodes for each sample was obtained. The number of electrons transferred and the percentage of peroxide (%HO₂⁻) during ORR was calculated from these data (see the ESI† for details of the calculation). As a result, from −0.3 V to −0.4 V (0.6–0.7 V vs. RHE), the electron transfer number of p-PAN was 2.20–2.22 while that of PANK2 ranged from 2.81 to 2.95. The others were 2.86–2.93 (PANK1) and 2.86–2.97 (PANK3) (Fig. 8a and b and S11†). The 2 electron pathway is dominant in p-PAN; however, both the 2 and 4 electron pathways are present in the PANK series. This result is attributed to the existence of pyrrolic nitrogen (N-5), which is the origin site to catalyze ORR as the direct 4 electron pathway.⁴⁴ In addition, Co₃O₄ formed in the PANKCo series acts as a new type of active site to decrease the generation of hydrogen peroxide, which is the product from the 2 electron pathway, as suggested in the dual-site mechanism theory.⁴⁵ Therefore, the electron transfer numbers of the PANKCo series were 3.38–3.44 (PANKCo1), 3.6–3.61 (PANKCo2) and 3.24–3.26 (PANKCo3). In a practical potential range (0.6–0.7 V vs. RHE), the overall ORR system in the PANKCo series favors more the four electron pathway compared to the PANK series.

Fig. 8 (a) Electron transfer number (n) and (b) the amount of produced peroxide (HO₂⁻) calculated from RRDE data (c) current–time chronoamperometric responses of PANKCo2 and Pt/C (20 wt% loaded) in O₂-saturated 1 M NaOH. Rotation speed of electrode: 1200 rpm (d) polarization curves of PANKCo2 and Pt/C (20 wt% loaded) in O₂-saturated 1 M NaOH (80 mL) with 3 M methanol (5 mL). Rotation speed of electrode: 1600 rpm.

Stability of an electrode is also an important factor to commercial fuel cell systems, and the stability test was performed with the chronoamperometry technique. A constant voltage of 0.7 V (vs. RHE) was applied to the electrode, and the resultant responses are shown in Fig. 8c. PANKCo2 shows remarkable durability at 98.0%, compared to that of Pt/C (88.9%). PANKCo2 maintained almost a constant current density, which is considered as contribution of the cobalt oxide surrounded by the carbon network in the form of cobalt carbide. In case of the methanol crossover test, PANKCo2 also shows greater tolerance to methanol poisoning effects than Pt/C (Fig. 8d). As the huge oxidation current density appears in the polarization curve, Pt/C has weak tolerance and catalyzes electrooxidation of methanol. However, PANKCo2 maintains ORR catalytic activity subject to the methanol exposure, judging from the fact that the half-wave potential is 0.824 V (vs. RHE) (0.836 V without methanol). Based on these electrochemical analyses, it was concluded that cobalt-doped NPC has a comparable electrocatalytic activity and superior durability to Pt/C.

4. Conclusions

This study showed that by using a novel approach of both KOH activation of PAN and cobalt doping, it dramatically increased the ORR activity of the resulting cobalt- and N-doped porous carbon. High porosity caused by KOH activation increased the number of active sites exposed to oxygen molecules, which raises the activity for ORR. Therefore, the N-doped porous carbon activated by larger amounts of KOH showed better ORR activity. Additionally, the existence of pyrrolic nitrogen (N-5) formed during KOH activation provided the active site for ORR. Cobalt doping induced another kind of active sites, Co₃O₄ and Co_xC, on the N-doped carbon support. Due to the dual sites which had compensating roles in ORR catalysis, the entire process was efficient for the four electron pathway. In addition, cobalt doping caused the conversion of pyridinic to pyrrolic nitrogen atoms, which promoted the ORR activity. This finding can support the enhancement of ORR catalytic activity in previous literatures which proposed composites of cobalt and N-doped carbon. As a result, the combination of KOH activation and Co₃O₄ impregnation of the N-doped porous carbon increased the ORR rate and enhanced the ORR activity, durability and tolerance to methanol poisoning effects.

Acknowledgements

This authors are grateful for the financial support from the Korea CCS R & D Center funded by the Ministry of Science, ICT, and Future Planning (NRF-2014M1A8A1049297).

Notes and references

1. M. K. Debe, Nature, 2012, 486, 43–51.
2. V. R. Stamenkovic, B. Fowler, B. S. Mun, G. F. Wang, P. N. Ross, C. A. Lucas and N. M. Markovic, Science, 2007, 315, 493–497.
3. V. R. Stamenkovic, B. S. Mun, M. Arenz, K. J. J. Mayrhofer, C. A. Lucas, G. F. Wang, P. N. Ross and N. M. Markovic, Nat. Mater., 2007, 6, 241–247.
4. R. Bashyam and P. Zelenay, Nature, 2006, 443, 63–66.
5. Y. Zheng, J. Liu, J. Liang, M. Jaroniec and S. Z. Qiao, Energy Environ. Sci., 2012, 5, 6717–6731.
6. J. S. Zhang, A. Byeon and J. W. Lee, J. Mater. Chem. A, 2013, 1, 8665–8671.
7. A. Byeon and J. W. Lee, J. Phys. Chem. C, 2013, 117, 24167–24173.
8. J. Zhang, A. Byeon and J. W. Lee, Chem. Commun., 2014, 50, 6349–6352.
9. A. Byeon, J. Park, S. Baik, Y. Jung and J. W. Lee, J. Mater. Chem. A, 2015, 3, 5843–5849.
10. S. Baik and J. W. Lee, RSC Adv., 2015, 5, 24661–24669.
11. Y. M. Tan, C. F. Xu, G. X. Chen, X. L. Fang, N. F. Zheng and Q. J. Xie, Adv. Funct. Mater., 2012, 22, 4584–4591.
12. Y. Liang, Y. Li, H. Wang, J. Zhou, J. Wang, T. Regier and H. Dai, Nat. Mater., 2011, 10, 780–786.
13. K. P. Gong, F. Du, Z. H. Xia, M. Durstock and L. M. Dai, Science, 2009, 323, 760–764.
14. L. T. Qu, Y. Liu, J. B. Baek and L. M. Dai, ACS Nano, 2010, 4, 1321–1326.
15. J. Maruyama, N. Fukui, M. Kawaguchi and I. Abe, J. Power Sources, 2009, 194, 655–661.
16. Z. W. Chen, D. Higgins, A. P. Yu, L. Zhang and J. J. Zhang, Energy Environ. Sci., 2011, 4, 3167–3192.
17. T. P. Fellinger, F. Hasche, P. Strasser and M. Antonietti, J. Am. Chem. Soc., 2012, 134, 4072–4075.
18. S. Gupta, D. Tryk, I. Bae, W. Aldred and E. Yeager, J. Appl. Electrochem., 1989, 19, 19–27.
19. M. C. M. Alves, J. P. Dodelet, D. Guay, M. Ladouceur and G. Tourillon, J. Phys. Chem., 1992, 96, 10898–10905.
20. D. Ohms, S. Herzog, R. Franke, V. Neumann, K. Wiesener, S. Gamburcev, A. Kaisheva and I. Iliev, J. Power Sources, 1992, 38, 327–334.
21. G. Liu, X. G. Li, P. Ganesan and B. N. Popov, Appl. Catal., B, 2009, 93, 156–165.
22. D. Shin, B. Jeong, B. S. Mun, H. Jeon, H. J. Shin, J. Baik and J. Lee, J. Phys. Chem. C, 2013, 117, 11619–11624.
23. J. Yin, Y. J. Qiu and J. Yu, Chem. Lett., 2013, 42, 413–415.
24. D. S. Yang, S. Chaudhari, K. P. Rajesh and J. S. Yu, ChemCatChem, 2014, 6, 1236–1244.
25. K. T. Cho, S. B. Lee and J. W. Lee, J. Phys. Chem. C, 2014, 118, 9357–9367.
26. L. L. Chen, T. Mashimo, C. Iwamoto, H. Okudera, E. Omurzak, H. S. Ganapathy, H. Ihara, J. D. Zhang, Z. Abdullaeva, S. Takebe and A. Yoshiasa, Nanotechnology, 2013, 24, 045602.
27. Y. Zhu, S. Murali, M. D. Stoller, K. J. Ganesh, W. Cai, P. J. Ferreira, A. Pirkle, R. M. Wallace, K. A. Cychosz, M. Thommes, D. Su, E. A. Stach and R. S. Ruoff, Science, 2011, 332, 1537–1541.
28. M. A. Lillo-Rodenas, D. Cazorla-Amoros and A. Linares-Solano, Carbon, 2003, 41, 267–275.
29. P. H. Matter, L. Zhang and U. S. Ozkan, J. Catal., 2006, 239, 83–96.
30. R. L. Arechederra, K. Artyushkova, P. Atanassov and S. D. Minteer, ACS Appl. Mater. Interfaces, 2010, 2, 3295–3302.
31. H. Niwa, K. Horiba, Y. Harada, M. Oshima, T. Ikeda, K. Terakura, J. Ozaki and S. Miyata, J. Power Sources, 2009, 187, 93–97.
32. B. Zheng, J. Wang, F. B. Wang and X. H. Xia, Electrochem. Commun., 2013, 28, 24–26.
33. Y. Su, Y. Zhu, H. Jiang, J. Shen, X. Yang, W. Zou, J. Chen and C. Li, Nanoscale, 2014, 6, 15080–15089.
34. Y. Y. Liang, Y. G. Li, H. L. Wang, J. G. Zhou, J. Wang, T. Regier and H. J. Dai, Nat. Mater., 2011, 10, 780–786.
35. Y. Gorlin and T. F. Jaramillo, J. Am. Chem. Soc., 2010, 132, 13612–13614.
36. W. G. Hardin, D. A. Slanac, X. Wang, S. Dai, K. P. Johnston and K. J. Stevenson, J. Phys. Chem. Lett., 2013, 4, 1254–1259.
37. Q. Liu, J. T. Jin and J. Y. Zhang, ACS Appl. Mater. Interfaces, 2013, 5, 5002–5008.
38. D. D. Wang, X. Chen, D. G. Evans and W. S. Yang, Nanoscale, 2013, 5, 5312–5315.
39. D. U. Lee, B. J. Kim and Z. W. Chen, J. Mater. Chem. A, 2013, 1, 4754–4762.
40. C. Jin, F. L. Lu, X. C. Cao, Z. R. Yang and R. Z. Yang, J. Mater. Chem. A, 2013, 1, 12170–12177.
41. M. Prabu, K. Ketpang and S. Shanmugam, Nanoscale, 2014, 6, 3173–3181.
42. Z. Y. Wu, P. Chen, Q. S. Wu, L. F. Yang, Z. Pan and Q. Wang, Nano Energy, 2014, 8, 118–125.
43. B. Zheng, J. Wang, F. B. Wang and X. H. Xia, J. Mater. Chem. A, 2014, 2, 9079–9084.
44. S. M. Unni, S. Devulapally, N. Karjule and S. Kurungot, J. Mater. Chem., 2012, 22, 23506–23513.
45. T. S. Olson, S. Pylypenko, P. Atanassov, K. Asazawa, K. Yamada and H. Tanaka, J. Phys. Chem. C, 2010, 114, 5049–5059.

---

Footnote

† Electronic supplementary information (ESI) available: Additional SEM, BET, XPS, CV, RDE and RRDE data. See DOI: 10.1039/c5ra15635a

---