Nitrogen-Doped Carbon Nanotubes Encapsulated Cobalt Nanoparticles Hybrids for Highly Efﬁcient Catalysis of Oxygen Reduction Reaction

In this work, a nitrogen doped carbon nanotubes encapsulated cobalt nanoparticles (Co@N-CNTs) catalyst was prepared by assembling the precursors of melamine and cobalt nitrate hexahydrate, followed by gradient temperature carbonization in the presence of glucose as a reducing agent. The as prepared Co@N-CNTs composite exhibited electroactivity and long-term stability toward oxygen reduction reaction (ORR) in both alkaline and acidic solutions. It was found that the gradient temperature synthesis and reduction of the gels in the presence of glucose played an important role in the formation of a thin carbon shell on the Co nanoparticles, which protected it from being destroyed during the electrochemical procedures. These processes also facilitated to the ordered arrangement of nitrogen-carbon precursors and the formation of enriched pyridinic, pyrrolic and quaternary nitrogen functional groups that led to the highly activity of Co@N-CNTs. In addition, the catalyst also presented tolerance to methanol oxidation. This synthetic method and mechanism enlighten the future design of state-of-the-art nanostructured materials as an alternation for precious metal based catalysts for highly efﬁcient ORR.

With the ever increasing demand for energy and environmental protection on the global scale, development of technologies for clean and sustainable energy has been attracting increasingly intense attention. 1 Fuel cells directly generate electricity by electrochemically reducing oxygen and oxidizing fuel into water as the only by-product and is a promising clean and sustainable power. 2 In spite of recent spectacular progress in this technology, a large-scale market introduction of fuel cells continues to face numerous challenges, associated with their high cost and still insufficient performance. 3 The kinetically sluggish oxygen reduction reaction (ORR) at the cathode is the foremost limitation in fuel cells. Platinum-based materials have long been used as active and efficient catalysts for ORR, but they suffered from multiple problems, for instance, the crossover effects, CO poisoning, and poor stability after long-term operation. Furthermore, the high usage cost of Pt due to its limited natural reserves has greatly hindered the large-scale commercialization of fuel cells. 4 Thus, efforts are needed to identify alternative catalysts that are readily available, cost effective, and low-cost non-noble metal catalytic effects for cathodic ORR in fuel cells.
Recently, some efforts have led to the rational design and synthesis of non-precious metal electrocatalysts, such as doped carbon materials, 5-8 transition metal, 9,10 transition metal oxides, [11][12][13] transition metal carbides, 14 transition metal nitrides. 15,16 The theoretical calculations and experimental methodologies indicated that some nonprecious metal electrocatalysts with unique electronic properties, twodimensional or three-dimensional nanostructures had highly efficient electrocatalytic activities which are comparable to noble metal catalyst. By tailoring the morphology and composition of catalysts, the catalytic activities and stability can be remarkably enhanced. Especially, carbon-supported transition metal/nitrogen (M-N-C) materials (M = Co, Fe, Ni, Mn, etc.) have gained increasing attention due to their promising catalytic activity toward the ORR, along with the utilization of abundant and inexpensive precursor materials. [17][18][19][20] In one type of catalyst derived from transition metal and carbon, cobalt = These authors contributed equally to this work. z E-mail: liu-yang@mail.tsinghua.edu.cn; wangzonghua@qdu.edu.cn (Co) lead to the formation of the active centers with higher activity toward the ORR, when compared to other transition metals (e.g., Ni, Mn). 21,22 Furthermore, the Co likely plays different roles in contributing to the active-site formation during the synthesis. 23,24 Normally, the Co nanoparticles supported on the carbon or nitrogen doped carbon materials are synthesized by thermal treatment at high temperature in an inert environment. [25][26][27] For example, Wu et al. reported a synthesis approach for cobalt-nitrogen-doped graphene-sheet-like nanostructures via the graphitization of the polyaniline and cobalt nitrate using multi-walled carbon nanotubes (MWNTs) as a supporting template, which involving the strong acid treatment caused the damage of the morphology. 28 While these methods should synthesize MCNTs as a template firstly. To simplify the procedures, Fu et al. recently reported a one-step pyrolysis method to synthesize cobalt supported on the N co-doped carbon nanotubes (Co-N-CNTs) using dicyandiamide and cobalt chloride as precursors. 9 The as-prepared Co-N-CNTs catalyst exhibited good electroactivity in alkaline solution, while the existence of cobalt oxide may affect the performances and their applications in acid media. Though the progresses have been achieved, it is still a challenge in the synthesis of high efficient electrocatalyst for ORR applications especially possessing excellent activity and stability in both alkaline and acidic media.
Herein, nitrogen doped carbon nanotubes encapsulated Co nanoparticles composite (Co@N-CNTs) was synthesized by the assembling melamine of cobalt nitrate hexahydrate, followed by the gradient temperature carbonization in the presence of glucose, as shown in Scheme 1. The nitrate can indeed trigger protonated melamine to form well-defined three-dimensional networks composed of micro-/nanoscale fibers that could provide a good matrix for gelating water molecules. Then, the obtained product was centrifuged to remove the redundant Co ions, which avoided the damage of the Co@N-CNTs structure by acid treatment. After freeze-drying, the nitrogen doped carbon nanotubes encapsulated Co nanoparticles with high purity was obtained by a gradient temperature calcination procedure with glucose as a reducing agent. The as-prepared Co@N-CNTs catalyst exhibited good electroactivity and high stability toward the ORR in both alkaline and acidic solution, owing to the unique structure, high purity cobalt nanoparticles cores and nitrogen doping in the carbon nanomaterials. In addition, the Co@N-CNTs also showed superior tolerance to methanol oxidation during the ORR procedures, which enlighten the future design of state-of-the-art nanostructured materials as analternative for precious metal based ORR catalysts.

Materials and Methods
Synthesis of the Co@N-CNTs catalysts.-Briefly, melamine (9 mmol) was mixed with Cobalt(II) nitrate hexahydrate (Co(NO 3 ) 2 ·6H 2 O, 9 mmol) and glucose (6 mmol), then the pH of homogeneous mixture was adjusted to 2.5 with HCl solution. It formed a stable gel ( Figure S1). 29,30 After centrifugation, redundant Co ions were removed to avoid the damage of the structure by acidification. After freeze-drying, the xerogel was placed on an alumina boat and heated at a ramp of 5 • C min −1 in a tube furnace and kept at 550 • C for 3 h in Ar atmosphere. Then the temperature in the tube furnace was further raised to 900 • C at a ramp of 5 • C min −1 and kept at 900 • C for 3 h in Ar atmosphere to pyrolyze the Co 2+ -g-C 3 N 4 . 31 Subsequently, the pyrolysis product was washed with 0.1 mol L −1 H 2 SO 4 solution to remove a small amount of cobalt particles on the surface. After washing with water and ethanol for three times, the final product of Co@N-CNTs-1 was obtained after drying. In addition, the product that prepared with the same procedure but in the absence of glucose was assigned as Co@N-CNTs-2. The composite that synthesized with the same procedure as Co@N-CNTs-1 but without the calcination aging treatment at 550 • C for 3 h was named Co@N-CNTs-3.
Characterization.-Field emission scanning electron microscopy (SEM) images were obtained with Hitachi SU-8010. The microscopic features of the samples were observed by high-revolution transmission electron microscope (HRTEM, JEM-2100F) operated at 200 kV. X-ray diffraction (XRD) characterization was carried out on a Bruker D8-Advance using Cu-Kα radiation (λ = 1.5418 Å). Raman spectra were obtained under ambient conditions by using a RM 2000 microscope confocal Raman spectrometer (Renishaw PLC) using 514 nm laser. X-ray photoelectron spectroscopy (XPS) measurements were performed on a PHI Quantera Scanning X-ray Microprobe using a monochromic Al-Kα (1486.7 eV).
Preparation of the electrodes.-The glassy carbon electrode (GCE) was successively polished with 0.3 μm and 0.05 μm α-Al 2 O 3 slurry on an abrasive cloth, followed by cleaning with distilled water and ethanol under sonication. After being dried with nitrogen gas, the catalyst homogeneous ink was prepared by dispersing catalyst (1 mg) in Nafion solution (1 mL, 0.5 wt%, aq.) with at least 30 min sonication. Then 6 μL of the catalyst ink was coated onto a clean GCE (with diameter of 3 mm) for cyclic voltammetry (CV) measurements and 20 μL of the catalyst ink was coated onto GCE (with diameter of 5 mm) for rotating disk electrode (RDE) measurements. All experiments were performed at room temperature.
Electrochemical characterization.-Electrochemical measurements were performed by CHI 1030B electrochemical analyzer (CH Instruments, China) coupled with a RDE system (Princeton Applied Research, Model 616). The electrochemical cell was composed with a conventional three-electrode system using a GCE as working electrode, an Ag/AgCl electrode with saturated solution of KCl as reference electrode, and a Pt wire as counter electrode. The measured potentials vs. Ag/AgCl reference electrode were converted to the reversible hydrogen electrode (RHE) according to the Nernst equation (E RHE = E Ag/AgCl + 0.059 × pH + 0.1976). The RDE measurements were recorded at various of rotating rates from 400 rpm to 2025 rpm in 0.1 mol L −1 KOH solution which was saturated with O 2 before each scanning by the O 2 purging for 30 min.

Characterization of Co@N-CNTs.-The
Co@N-CNTs-1 was synthesized by the assembly of melamine and cobalt nitrate. After freeze-drying procedure, the precursors were calcinated at 550 • C for 3 h in Ar atmosphere, and pyrolysis to 900 • C and kept at 900 • C for 3 h in Ar atmosphere, and the final products were obtained. SEM image ( Figure 1A) shows that the products consist of carbon nanotubes with several micrometers long and approximate 100 nm of diameter. Figure 1B exhibits TEM image of the product, and some nanoparticles dispersed inside the nanotubes especially on the end point of the nanotubes. Figure 1C shows the enlarged TEM image of the Co@N-CNTs-1, and the corresponding elemental mapping pictures of carbon and cobalt were presented as presented in Figures 1D and 1E, respectively. Homogeneous distribution of carbon element on the carbon nanotube was observed, and the cobalt element was located on the nanoparticles. Moreover, the HRTEM pattern ( Figure 1F) shows that the nanoparticles are coated with carbon shells with a thickness of about 10 nm. The lattice spacing of the carbon shell is 0.356 nm which is slightly larger than that of graphite of 0.335 nm, suggesting that there is graphitic carbon with turbostratic structure in the stacking of graphitic carbon coating. 15 Besides, the carbon networks of the graphitic layers were disordered, which may be attributed to the doping of hetero-atoms N that could lead to the stacking disorder and a turbostratic structure of the graphitic planes. The structure and components of Co@N-CNTs-1 nanocomposites were further studied by XRD as shown in Figure 1G. The XRD peaks are in agreement with the structural feature of graphite carbon and the metallic cobalt (JCPDS: 15-0806). The similar components were observed for Co@N-CNTs-3( Figure S2a), while a complex composite consist of Co, Co 3 O 4 and graphite carbon were appeared in Co@N-CNTs-2 ( Figure S2b), which revealed that glucose may act as a reducing agent and carbon source in the process of pyrolysis. Figure 2A shows the XPS spectra of the Co@N-CNTs-1 composite, and the elements of C, N, and Co in Co@N-CNTs-1 are observed. The quantification results showed that the atomic concentrations of Co and N in the Co@N-CNTs-1 composite are 0.24% and 1.75%, respectively. In addition, the cobalt was assigned to metallic cobalt according to the Co 2p, which is similar to that of XRD result ( Figure S3). Figure  2B shows the high resolution XPS of nitrogen in the Co@N-CNTs-1. The N 1 s spectra of Co@N-CNTs-1 present three types of nitrogen species at 398.4 eV, 399.3 eV and 401.1 eV, corresponding to pyridinic nitrogen (N1), pyrrolic nitrogen (N2) and quaternary nitrogen (N3), respectively. The contents of the three types of nitrogen in the products were summarized (Table S1). Especially, the atomic ratios of N2/N total and N3/N total in the Co@N-CNTs-1 are 26.7% and 59.3%, respectively. The results suggested that the formation of pyrrolic nitrogen and quaternary nitrogen is suppressed in the absence of glucose. Furthermore, the calcination aging at 550 • C facilitates the transfor-mation of pyridinic nitrogen to pyrrolic nitrogen. Figure 2C shows the Raman spectra of Co@N-CNTs. Three remarkable peaks at approximately 1330 cm −1 , 1600 cm −1 and 2700 cm −1 in the Raman spectra of Co@N-CNTs ( Figure 2C) are corresponding to the D, G and 2D bands of CNTs, respectively. The D band is often referred to as the disorder or defect band, which originates from hybridized vibrational mode associated with the edges of CNTs, and the G band is related to the tangential oscillations and vibrations of all the sp 2 carbon atoms in the CNTs. The intensity ratio between D and G bands (I D /I G ) could generally reflect the defect density in CNTs. In addition, the 2D band, also known as G' band, is a two-phonon resonance Raman bands that reflect the thickness of the carbon film directly. 32,33 The I D /I G of the Co@N-CNTs-1 is calculated to be 1.05. As a comparison, the Co@N-CNTs-3 was 0.88 (curve b), and the Co@N-CNTs-2 was 0.98 (curve c). The highest I D /I G value for Co@N-CNTs-1 means that more defects formed in the carbon of Co@N-CNTs-1. In addition, the sharp 2D peak of Co@N-CNTs-1 means that a thinner carbon film formed than that of Co@N-CNTs-3.

Electrocatalytic activities of Co@N-CNTs toward the ORR.-
The electrocatalytic activity of Co@N-CNTs-1 for ORR was measured by CV in N 2 and O 2 saturated KOH solution. As shown in Figure 3, there is no redox peak within the potential range for the N 2 -saturated solution (curve a). When the solution was enriched with O 2 , a well-defined peak with high current density at 0.82 V vs. RHE was observed, and the onset potential (E onset ) was 0.94 V (curve b), suggesting the excellent electrocatalytic activity of the Co@N-CNT-1 for ORR. For a comparison, the CVs of multi-walled carbon nanotubes (curve c) and commercial Pt/C (20 wt%) (curve d) modified electrode were also measured. It was observed that both the onset potential and the current density for Co@N-CNT-1 were superior to that of multi-walled carbon nanotubes (curve c), and was comparable to Pt/C catalyst (curve d). In addition, the electrocatalytic activity of Co@N-CNTs-2 was also studied as shown in Figure S4, and its electrocatalytic activ-  ity was obviously lower than that of Co@N-CNT-1, which could be attributed the co-existence of Co 3 O 4 in the products that impressed the electrocatalytic activity.
To better understand the electrocatalytic performance of Co@N-CNTs catalysts during the ORR process, linear sweep voltammetry (LSV) curves were measured using RDE in O 2 saturated 0.1 mol·L −1 KOH solution at a rotation rate of 1600 rpm. As shown in Figure  4A, The ORR E onset of the Co@N-CNTs-1 (0.975 V, curve c) is very close to the commercial Pt/C (1.01 V, curve d) and is better than that of Co@N-CNTs-2 (0.963 V, curve b) and Co@N-CNTs-3 (0.970 V, curve a). Moreover, the limiting current density at 0.4 V of Co@N-CNTs-1 (2.697 mA cm −2 , curve c) is close to that of the commercial Pt/C catalyst, which is much larger than that of Co@N-CNTs-2 (2.318 mA cm −2 , curve b) and Co@N-CNTs-3 (2.004 mA cm −2 , curve a). These results confirmed that the good electrocatalytic activity of Co@N-CNTs-1 toward the ORR. ORR activity of the Co@N-CNTs-1 was also gleaned from the smaller Tafel slope of 49 mV/decade ( Figure 4B, curve c) at low over-potentials compared with Co@N-CNTs-2 (65 mV/decade, curve a) and Co@N-CNTs-3 (59 mV/decade, curve b). It's known that the activity of nitrogen doped carbon is associated with the types of nitrogen. As it was described in the XPS spectra, the content of pyrrolic nitrogen in Co@N-CNTs-1 was higher than that of Co@N-CNTs-3. The Co@N-CNTs-2 had the highest content of pyridinic nitrogen. The results are similar to those reported that the content of the three types nitrogen achieved a balance contributes to more active site for ORR. The facts suggested that glucose in the experiments facilitates the formation of pyrrolic nitrogen and quaternary nitrogen. Furthermore, the calcination aging at 550 • C increased the transformation of pyridinic nitrogen to pyrrolic nitrogen, which may be resulted in the good catalytic activity toward the ORR. These results were also confirmed by Raman spectrum that after the calcination aging at 550 • C, the Co@N-CNTs-1 has the higher I D /I G (1.05) and the sharper 2D peak than the Co@N-CNTs-3. These results implied that Co@N-CNTs-1 has a thinner carbon film therefore has more significant structural defects and N dopants which has more active site for ORR. RDE voltammetry was further performed to study the kinetics of the electrochemical catalytic ORR for Co@N-CNTs-1. The polarization curves were obtained by scanning the potential from 0.3 to 1.0 V at a scan rate of 5 mV s −1 in O 2 -saturated 0.1 mol L −1 KOH solution with a various of rotation rates ranging from 400 rpm to 2025 rpm. As shown in Figure 5, the limiting current density increased with increasing rotation rate. The corresponding Koutecky-Levich (K-L) plots showed good linearity with parallelism over the potential range  Figure S5). To get insight into the superior activity of Co@N-CNTs-1 as well as the relationship between the structure and properties, some factors including the content of melamine, the added amount of glucose, the synthesis of materials by pyrolysis of the Co 2+ -g-C 3 N 4 at different temperatures were investigated. As shown in Figure S6A, with the increasing of melamine, the ORR current density increased first and then decreased, and the maximal current density was obtained with the molar ratio of cobalt nitrate and melamine is 1:1. In addition, a similar peak type tendency was observed by changing the glucose content in the precursor. As shown in Figure S6B, the maximal current density was obtained when the glucose content was 6 mmol, which proved added an appropriate amount of reducing agent and carbon source is necessary to increase the ORR activity. Moreover, a series of materials were obtained by pyrolyzing Co 2+ -g-C 3 N 4 at 700, 800, 900 and 1000 • C, and their electrocatalytic activities were also investigated. Figure S6C shows the CV curves of the products synthesized at different calcination aging temperatures. The ORR catalytic activity of the materials was first increased as the pyrolysis temperature increased from 700 • C to 900 • C, and then decreased after 900 • C. The highest ORR current density was obtained at 900 • C. The electrocatalytic performance of Co@N-CNTs in acidic medium was also investigated. The LSV curves were obtained at different electrodes with an electrode rotation rate of 1600 rpm in O 2 -saturated 0.5 mol L −1 H 2 SO 4 solution. As shown in Figure 6, all these Co@N-CNTs catalysts show larger current densities and more positive E onset compared with the MCNTs in acidic medium (curve a). Particularly, the E onset of the Co@N-CNTs-1 (curve d) is 0.77 V, which is more positive than the Co@N-CNTs-3 (0.710 V, curve b) and the Co@N-CNTs-2 (0.720 V, curve c).
The resistance to crossover effects and stability of the catalyst materials are important considerations for their practical application in fuel cells. Chronoamperometric responses of Co@N-CNTs-1 were measured to evaluate the methanol crossover and stability ( Figure 7A). It could be seen that the Co@N-CNTs-1 exhibited a consistent ORR curve after the addition of 3 mol L −1 methanol to an O 2 -saturated 0.1 mol L −1 KOH solution, whereas the ORR curve of the commercial Pt/C suffered a sharp decrease after the addition of methanol. These results indicate that Co@N-CNTs-1 exhibit high ORR selectivity and have a good ability to avoid crossover effects. The durability of Co@N-CNTs-1 and commercial Pt/C were also tested ( Figure 7B). The Co@N-CNTs-1 were held at 0.81 V for 25000 s in an O 2 -saturated 0.1 mol L −1 KOH solution. It could be seen that the chronoamperometric response for Co@N-CNTs-1 exhibited a slight decrease of less than 17%, which was much better than that of commercial Pt/C of approximately 40% decline after 25000 s. These results demonstrated that Co@N-CNTs-1 have superior stability and have potential use in direct methanol and alkaline fuel cells. In addition, the durability of Co@N-CNTs-1 in O 2 -saturated 0.5 mol L −1 H 2 SO 4 solution was also evaluated. As shown in Figure S7, the chronoamperometric response for Co@N-CNTs-1 exhibited a decrease of less than 30%, while the commercial Pt/C lost approximately 75% after 25000 s, indicating a better stability of the Co@N-CNTs-1 than the commercial Pt/C in acidic medium. The above results demonstrate that the Co@N-CNTs-1 also presents excellent the ORR catalytic activity in acidic medium.
As far as the excellent performances on ORR and the unique structure features of Co@N-CNTs-1, there are three reasons should be responsible for its superior electrocatalytic performance for ORR: (i) The assembling process of the precursors to form novel superstructures, adding glucose as a reducing agent and the gradient temperature synthesis mechanism are indispensable for catalyzing the ordered assembled N-doped nanocarbon. (ii) The calcination aging temperature of 550 • C facilitates the formation of pyridinic nitrogen and pyrrolic nitrogen at a suitable ratio which are valuable active sites for ORR.
(iii) Glucose added during calcination procedures acts as a reducing agent and carbon source for the formation of carbon encapsulated Co metallic nanoparticles, avoiding the formation of cobalt oxides and the exposure of Co nanoparticles during electrochemical reduction procedures. These synergies effects are attributed to the excellent electrocatalytic activity toward the ORR.

Conclusions
In summary, the nitrogen doped carbon nanotubes encapsulated cobalt nanoparticles (Co@N-CNTs-1) was fabricated by assembling the precursors of melamine, glucose and cobalt nitrate hexahydrate. The as-prepared Co@N-CNTs-1 catalyst exhibited activity with highly positive onset potential and preferable current density, excellent tolerance to methanol oxidation and high stability in both alkaline and acidic medium. The glucose added during gradient temperature plays an important role in formation of a thin carbon shell on the Co nanoparticles, which protects it from being destroyed during the electrochemical procedures, and facilitates to the ordered arrangement of nitrogen-carbon precursors and the formation of enriched pyridinic, pyrrolic and quaternary nitrogen, resulting in the highly activity of Co@N-CNTs. The strategy enlightens the future design of state-of-the-art nanostructured materials as an alternation for precious metal-based ORR catalysts.