Nitrogen doping of carbon nanoballoons by radiofrequency magnetron plasma and evaluation of their oxygen reduction reaction activity

Abstract

Nitrogen‐doped carbon materials exhibit a catalytic activity, such as an oxygen reduction reaction (ORR). In this study, we performed nitrogen doping on a carbon nanoballoon (CNB), which is a nanometer‐sized carbon particle in the form of a hollow sphere made of graphite by radiofrequency (RF) magnetron plasma in a gas mixture of nitrogen and helium. Nitrogen‐doped CNBs (N‐doped CNBs) were prepared by different plasma irradiation conditions: the sample installation positions, input powers, and irradiation times. The samples were examined for chemical state by X‐ray photoelectron spectroscopy (XPS). Hydrodynamic voltammetry was used for the evaluation of the catalytic activity of a N‐doped CNB with a pyridinic N concentration of 0.4‐1.0 at.%. As a result, the onset potential was measured to be 0.13 V versus RHE (reversible hydrogen electrode), which was close to the previously reported data of highly oriented pyrolytic graphite (HOPG) with a pyridinic N concentration of 0.57 at.%, which was prepared by annealing under NH₃.

1 INTRODUCTION

Increasing efforts have been made toward the “hydrogen society” in which hydrogen is used as energy source, thus getting rid of fossil fuel dependence. Fuel cells are considered to play a central role in hydrogen‐based electric power generation, but cost reduction is needed for introduction of fuel cells to residential, transportation, and industrial sectors. “Technology Roadmap Hydrogen and Fuel Cells” compiled by the Council for a Strategy for Hydrogen and Fuel Cells1 sets a goal of making the price of fuel‐cell electric vehicles even with that of hybrid vehicles by 2025.

Research in fuel cells conducted at our laboratory was so far focused on methanol oxidation reaction near the anode of direct methanol fuel cells (DMFC) using methanol as a fuel.2-4 Oxygen reduction reaction (ORR) considered in this study takes place near the cathode, being an important reaction common for fuel cells, metal‐air cells, dye‐sensitized solar cells, and other types (Figure 1). In case of cells fueled by hydrogen, protons and electrons generated by hydrogen oxidation reaction at the fuel electrode (anode) are supplied to the air electrode (cathode) via electrolyte and conductor (load), respectively, and the four‐electron reaction (ORR) advances as shown below.5

(1)

Figure 1

Schematic diagram of fuel cell [Color figure can be viewed at wileyonlinelibrary.com]

This reaction advances if the electrode potential is below 1.23 V versus RHE (Reversible Hydrogen Electrode), and the potential must be set amply lower to obtain certain current. This is called overvoltage. Here “versus RHE” means using electrodes based on electrochemical redox reaction between hydrogen ions and hydrogen gas with reference to measurement of electrode potential.6

Catalysts are employed to reduce overvoltage, that is, to promote the reaction at a higher potential. When platinum is used as a catalyst, ORR onset potential is about 1 V versus RHE.7 It is difficult to cut the cost of fuel cells using platinum and other noble metal catalysts. Thus, reduction of platinum usage or development of alternative catalysts is necessary for popularization of fuel cells, and therefore, for realization of hydrogen society. Metal complexes and other materials were considered so far as alternatives to platinum catalyst.8 In addition, nitrogen‐doped graphene, nitrogen‐doped carbon nanotubes, and other nitrogen‐doped nanomaterials were recently reported to have high ORR activity, being explored as promising platinum alternatives.9

Chemical bonds between carbon atoms and doped nitrogen atoms in a sp²‐bonded graphite structure can involve pyridinic N, pyrrolic N, and graphitic N species, as shown in Figure 2. Pyridinic and pyrrolic N species are nitrogen atoms bonded to two carbon atoms to form a six‐membered ring or a five‐membered ring, respectively. Besides, graphitic N species, also called substitutional N species, are hydrogen atoms bonded to three carbon atoms to form a six‐membered ring. Hydrogen species that become ORR active spots were not known until recently, but Guo et al showed that active spots are carbon atoms adjacent to pyridinic N.10 Thus, in this study we focused on pyridinic N species in carbon nanomaterials.

Figure 2

Nitrogen species doped in sp²‐bonded carbon atoms [Color figure can be viewed at wileyonlinelibrary.com]

There are a number of method of nitrogen doping after synthesis of carbon materials such as (1) introducing nitrogen into a carbon material structure when oxidized materials are reduced through heating in ammonia atmosphere, (2) sintering of solids after drying of carbon materials that were immersed in a nitrogen compound, (3) injection of accelerated nitrogen ions into carbon materials, and (4) treatment of carbon materials with ammonia plasma.11-14

This study is aimed at confirming whether ORR is activated by nitrogen doping to hollow carbon nanoparticles with graphite structure—so called carbon nanoballoons (CNBs), and comparing their performance with previously reported nitrogen‐doped carbon nanomaterials. CNBs are classified with hollow nanoparticles. They have specific surface area of 35 m²/g,15 thus being far from matching the theoretical specific surface area (2630 m²/g) of other carbon nanomaterials such as graphene (stratified substance composed of a single or multiple graphene sheets).16 On the other hand, CNBs have some merits; unlike graphene, CNBs are not likely to agglomerate, while chemical reaction site can be extended from inside to surface due to appropriate oxidation treatment. Due to these features, we used CNBs as an active material for electric double‐layer capacitors with fast charging.15

We employed RF magnetron nitrogen plasma for nitrogen doping to CNBs. The use of plasma treatment is advantageous in that nitrogen gas can be utilized as a more stable nitrogen source as compared to ammonia etc. In addition, electron density in plasma can be increased due to the magnetron effect so that more nitrogen ions and radicals are produced; this makes possible inhibition of CNB damage by accelerated ions and recoil gas atoms during nitrogen doping. There are reports about reducing self‐bias voltage of RF plasma from 1 kV to minus several hundred V due to the use of magnetron.17 Chemical conditions of carbon and nitrogen atoms in fabricated nitrogen‐doped samples were analyzed by X‐ray photoelectron spectroscopy (XPS), while ORR activity of the samples was evaluated by electrochemical measurement, namely, hydrodynamic voltammetry.

2 EXPERIMENTAL METHOD

2.1 Synthesis of carbon nanoballoons15

CNB was discovered in experiments with heat treatment of arc black (AcB). CNB can be obtained by heating AcB—cocoon‐shaped carbon particles with amorphous structure—in Ar atmosphere at 2000°C or higher temperature. In this study, we used CNB obtained by treating AcB at 2600°C. CNB particle is a hollow structure about 50 nm in diameter. In addition, CNB has a spherical graphite structure of several to a dozen layers.

2.2 Nitrogen doping to CNBs

A plasma system used for nitrogen doping is shown schematically in Figure 3. CNBs were fixed on a carbon tape attached to a glass plate to be irradiated with RF magnetron plasma in a position between the cathode (K) and anode (A) (position ①), and in a position on the plasma chamber wall at 9.5 cm from the line connecting A and K (position ②). In position ①, the plate was electrically insulated from the plasma chamber wall. When samples were placed in the middle of A, K, CNB structure was heavily damaged as compared to positions ①, ②, and this installation conditions were excluded in experiments. The damage can be attributed to the fact that A and K where positive ions collide with electrons have a low potential as compared to ①, ②, thus being vulnerable to the effects of these high‐energy particles.18 In addition, helium gas was mixed to nitrogen gas in the plasma chamber in order to increase active particles in plasma, and to prevent CNB disruption. The mixed gas was introduced at a flow rate ratio of N₂:He = 10:1 and a total gas pressure of 40 Pa. Input power was varied at 20, 40, 60 W. In so doing, self‐bias induced at K was about –90, –130, –170 V, respectively. Plasma irradiation time was 5 and 30 minutes.

Figure 3

Plasma apparatus and sample installation positions [Color figure can be viewed at wileyonlinelibrary.com]

2.3 Evaluation of sample characteristics

XPS is a method to analyze elemental composition and chemical bonding state of a sample by irradiating its surface with X rays, and measuring energy of photoelectrons emitted from the surface due to the photoelectric effect.19 An XPS system (PHI Quantera SXM‐CI by Ulvac) was employed to analyze the bonding state of nitrogen atoms on the surface of a sample. Special software (PHI MultiPak by Ulvac) was used for peak separation and other spectral analysis. This XPS system is not directly linked to the plasma system, and samples treated by the plasma system were once exposed to the atmosphere before being installed in the XPS system.

ORR activity in samples was estimated through electrochemical measurement, namely, hydrodynamic voltammetry using a rotating disk electrode. First, 5‐wt% nafion solution was diluted to 0.1 wt% in methanol, and 0.5 mL of the diluted solution was added with respect to 1 mg sample; a fluid dispersion was prepared by 20‐minute treatment in a homogenizer while being cooled. Next 1 μL of the fluid dispersion was applied to the rotating disk electrode, and dried to obtain a working electrode. Platinum wire as used as a counter electrode, and RHE was used as a reference electrode. Using 0.1 M H₂SO₄ as an electrolyte solution, the working electrode was rotated at 500 rpm to blow nitrogen gas and oxygen gas in and background background voltammograms were obtained by potential scan at onset potential of 1 V versus RHE and scan speed of 0.01 Vs/ in arrange of 0.05‐1 V versus RHE. Next, voltammograms with blowing of nitrogen gas and oxygen gas were obtained in a similar way.

3 RESULTS AND DISCUSSION

3.1 X‐ray photoelectron spectroscopy of sample surface

XPS survey spectra of CNB sample surface treated and not treated with plasma are shown in Figure 4. Silver (Ag3d) peaks are observed in these spectra, which is because of silver paste applied before measurement, from the sample periphery through the grounded substrate, for the purpose of charge correction. As a result of the plasma treatment, intensity of the oxygen (O1s) peak increased with reference to the carbon (C1s) peak. In addition, nitrogen peak (N1s) appeared. Relative abundance ratio N/C of nitrogen to carbon on the sample surface was found from spectral area ratio. The increase in O1s intensity after the plasma treatment can be explained by defects that appeared in the graphite structure of CNB to promote amorphization. These defects include unsaturated bonds; thus one can assume that exposure to the atmosphere before XPS analysis resulted in chemisorption of oxygen, carbon dioxide, and other species at the defects, which cause intensification of O1s.

Figure 4

Survey spectra of (A) nitrogen‐plasma‐treated and (B) pristine CNBs (experimental conditions for treated CNB are position: ①, input power: 60 W, irradiation time: 30 minutes)

Figure 5 shows N/C ratio versus RF plasma input power for each treatment time. First, as regards sample position, N/C ratio was generally higher in position ① than in position ②. Next, as regards input power, N/C ratio grew with input power in case of 5‐minute treatment in position ① (⬜ in Figure 5A). However, in case of 30‐minute treatment (⬛, ⬤ in Figure 5B), N/C ratio did not monotonely grew with input power: it was the highest at 40 W, and then declined at 60 W. One can assume that with longer treatment, ions in nitrogen plasma promoted sputtering to CNB. On the other hand, no changes were observed depending on the sample position or plasma treatment time at the input power of 20 W (N/C ratio = 0.01‐0.03).

Figure 5

Input power—N/C relation of nitrogen‐plasma‐treated CNB sample's surface placed at different positions (irradiation time is (A) 5 minutes and (B) 30 minutes)

C1s XPS spectra of CNB sample surface after nitrogen plasma treatment are shown in Figure 6. As a result of speak separation in these spectra, peak binding energy in untreated CNB was 284.3 eV. This is indicative of sp² C‐C bonds typical to graphite.20 On the other hand, changes on the high binding‐energy side appeared in C1s XPS spectra of nitrogen‐doped samples. For example, in samples with N/C ratio of 0.15, a relatively broad peak of 285‐290 eV intensified with respect to sp² C‐C peak. This means that spectral width of sp² C‐C expanded due to amorphization of the graphite structure of CNB, that some six‐membered rings formed by nitrogen atoms in CNB changed to sp³ C‐C, or that carbon atoms might form chemical bonds with nitrogen and oxygen. Besides, in samples with the high N/C ratio of 0.33, peaks caused by carbon oxidation, particularly, carboxyl group (O = C‐O) and amide group (O = C‐N), became dominant. O/C ratio (relative oxygen concentration) estimated same as N/C ratio increased with N/C ratio from 0.01 to 0.14. These trends agree with a previous report on nitrogen‐doped carbon films fabricated by electron cyclotron resonance sputtering.21

Figure 6

C1s spectra of surfaces of nitrogen‐plasma‐treated CNB samples (experimental conditions for CNB samples with N/C = 0.15 were position: ①, input power: 60 W, irradiation time: 30 minutes and with N/C = 0.33: position: ①, input power: 40 W, irradiation time: 30 minutes) [Color figure can be viewed at wileyonlinelibrary.com]

N1s spectrum of samples with N/C ratio of 0.15 is presented in Figure 7. Peak separation was applied to this spectrum, and three types were obtained (from low to high binding energy): pyridinic N, pyrrolic N, and graphitic N. Figure 6 indicates a chemical shift due to bonding between oxygen and carbon atoms; thus we examined presence of chemical bonds (N‐O) between oxygen and nitrogen atoms. According to a previous study,22 peak binding energy of pyridine‐N‐oxide is 404 eV, while no corresponding upsurge is observed in Figure 7. Therefore, N‐O peak was not considered in Figure 7.

Figure 7

N1s spectrum of surface of nitrogen‐plasma‐treated CNB sample with N/C = 0.15 (experimental conditions were position: ①, input power: 60 W, irradiation time: 30 minutes) [Color figure can be viewed at wileyonlinelibrary.com]

Percentage of the three bonding states—pyridinic N, pyrrolic N, and graphitic N—in Figure 7 was 37%, 40%, and 23%. The percentage of pyridinic N in these samples (37%) was close to experimental values for nitrogen doping of graphene flakes by ion injection.13 In addition, percentage of pyridinic N in N1s spectrum plotted against N/C ratio in Figure 8 showed a trend to increase with N/C ratio. This result suggests that defects were produced in CNB structure in the course of nitridization so that pyridinic N species formed easily. Since pyridinic N becomes an active spot in ORR, one can conclude that increase in pyridinic N is an advantage of a nitrogen doping method. On the other hand, if defects underlying formation of pyridinic N are too numerous, then electric conductivity may be lost. There is a report that oxygen reduction reaction was not observed between a carbon film electrode doped with nitrogen at a concentration of 30at% and hexacyanoferric acid ions23; therefore, there is a possibility that samples with N/C ratio of 0.33 lost their electric conductivity.

Figure 8

Relation of N/C—concentration of pyridinic N in nitrogen‐plasma‐treated CNBs

Thus we found that nitrogen‐carbon bonds supposed to have ORR activity were formed in these experiments over a wide percentage range.

3.2 Oxygen reduction reaction

ORR curves of nitrogen‐treated and untreated CNBs obtained through hydrodynamic voltammetry are presented in Figure 9. Current density values plotted on the vertical axes were obtained by subtracting the background measured under saturated conditions of nitrogen gas from values measured under saturated conditions of oxygen and hydrogen gases, and then divided by surface area of the working electrode. One can say that catalytic performance improves with higher ORR onset potential and the absolute value of current density. Comparing potential needed for current density to achieve –0.01 mA/cm², the potential is 0.095 V versus RHE for untreated CNB and 0.13 V versus RHE for nitrogen‐doped CNB (N/C = 0.15); that is, overvoltage was reduced by the nitridization treatment. In addition, comparing current density obtained at the potential of 0.05 V versus RHE, nitrogen‐doped CNB showed 0.042 mA/cm² against 0.018 mA/cm² in untreated CNB, that is, 2.3 times higher. Thus we confirmed that reduction of overvoltage in nitrogen‐doped CNB was only slight, but catalytic performance improved.

Figure 9

ORR curves of nitrogen‐plasma‐treated CNB with N/C = 0.15 and pristine CNB (experimental conditions for CNB sample with N/C = 0.15 were position: ①, input power: 60 W, irradiation time: 30 minutes) [Color figure can be viewed at wileyonlinelibrary.com]

3.3 N/C ratio in entire samples

CNB nitridization produced only slight reduction of overvoltage. For example, Yu et al reported that samples with nitrogen concentration about 3.6 at% were obtained by synthesis of nitrogen‐doped carbon nanotube using chemical vapor deposition, and the onset potential was 0.7 V versus RHE.24 N/C ratio was estimated so far only on the sample surface exposed to plasma; on the other hand, this subsection considers N/C in entire samples treated by the plasma system. While citing results of other studies, we overview the role of pyridinic N and other nitrogen‐carbon bonds as well as the role of defects in nitrogen‐free graphite structure, and dwell on possible development of this study.

In order to conduct hydrodynamic voltammetry, plasma‐treated CNB samples were peeled from carbon tape, and dispersed in nafion solution. Thus we wondered whether N/C ratio is greatly different on the sample surface directly irradiated with plasma and in an entire sample containing CNB. We conducted XPS measurement as described below in order to estimate N/C ratio in samples used for voltammetry. Thus obtained results are given in Table 1. N/C ratio and pyridinic N percentage were determined from XPS spectra of plasma‐treated CNB samples peeled from carbon tape and attached to another substrate (B in Table 1) and of CNB samples obtained by evaporating the solvent from nafion dispersion containing CNB used for hydrodynamic voltammetry (C in Table 1). When plasma‐treated were then immersed in nafion solution, N/C ratio decreased 8 times. Many nitrogen‐carbon bonds were formed in CNB samples fixed on carbon tape as compared to outermost surface CNBs exposed to plasma; however, CNB number density is high near the surface, which can explain fewer bonds formed in deeper CNBs. One can assume that active nitrogen ions and atoms (N⁺, N*) supposed to be main factors of nitrogen‐carbon bonds did not penetrate deep in a sample, and only a surface layer was nitridized.25 As suggested by results for B and C representing N/C ratio in an entire sample, the concentration of pyridinic N in doped CNB in this study is 0.4‐1.0 at%. Considering potential of 0.13 V versus RHE needed to achieve current density of –0.01 mA/cm² in the ORR curves in Figure 9, these experimental results offered ORR close to that of HOPG (Highly Oriented Pyrolytic Graphite) doped with pyridinic N at 0.57 at% (approx. 0.12 V vs RHE).10 In order to further boost ORR activity by the method adopted in this study, it would be necessary to modify nitrogen doping so as to obtain uniform N/C ratio not only on the sample surface but in depth as well; one can think of using plasma powder treatment under reduced pressure, or using thinner CNB samples fixed on a large‐area substrate, etc.

Table 1. N/C and pyridinic N concentration of CNB samples prepared by different procedures

	A	B	C
N/C	0.15	0.03	0.02
Pyridinic N (at %)	4.6	0.99	0.36

• A: Surface of nitrogen‐plasma‐treated CNBs fixed on a carbon tape. The experimental conditions: position: ①; input power: 60 W; irradiation time: 30 minutes.
• B: “A” peeled from the carbon tape and pasted on another substrate.
• C: “A” peeled from the carbon tape and dispersed in Nafion® solution. The solvent was removed prior to XPS analysis.

However, according to many reports, metal‐free nitrogen‐doped nanotubes or graphene with nitrogen content as low as several at% showed much higher onset potential than that obtained in this study.9 Among the three nitrogen species analyzed in this study, some species may have high percentage in one case, and all three species may have almost equal percentage in another case.24, 26, 27 Therefore, nitrogen species other than pyridinic N would hardly impede ORR activity, while one can expect for sufficient improvement of ORR activity even with low NC ratio doping as soon as pyridinic N is included at a high level.

On the other hand, ORR activity in the two‐electron reaction shown below was reported to be obtained by forming defects in multiwalled carbon nanotubes, without adding pyridinic N or metal catalysts offering ORR activity.28

(2)

As indicated by XPS analysis in Figure 6, CNB graphite structure was partly amorphized due to nitridization; therefore, it would be reasonable to think that ORR activation in Figure 9 involves not only the four‐elector chemical reaction in Equation 1 but also the two‐electron reaction in Equation 2. Thus, one can assume that with CNB, too, intentional introduction of defects promotes formation of pyridinic N, while ORR activity of the defects themselves can be utilized.

In closing, consider the effect of sample's electric conductivity. In a study by Kamata et al,23 redox reaction activity was evaluated while varying nitrogen concentration in samples from 0 to 30.4 at%; maximum activity was obtained at 20.2 at%. This was attributed to the effect of electric conductivity that declined, though number density of pyridinic N (a factor contributing to the four‐electron reaction) increased. Similarly, results of Figure 9 suggest that with CNB obtained in this study, an increase in onset potential was observed due to nitrogen doping or amorphization of graphite structure, but growth of current density might be inhibited by decline in electric conductivity.

4 CONCLUSION

CNBs—a carbon nanomaterial with spherical graphite structure—were nitrogen‐doped by irradiation with RF magnetron nitrogen‐helium mixture plasma. XPS analysis showed that CNBs were partly amorphized, and nitrogen was doped into thus produced effect sites. In order to examine ORR activity in nitrogen‐doped CNBs, hydrodynamic voltammetry was applied to plasma‐treated samples with N/C ratio of 0.15 on the surface; thus improvement of activity due to nitrogen doping was confirmed. Pyridinic N content in the nitrogen‐doped CNBs fabricated in this study is estimated at 0.4‐1.0 at%.

ACKNOWLEDGMENTS

This study was in part conducted at the Knowledge Hub Aichi facilities. We also express our gratitude to Ms. Moeko Furukawa from the University of Tsukuba for guidance in hydrodynamic voltammetry measurement.

Biographies

• 

  Ryota Takahashi, nonmember. In 2016 graduated from Toyohashi University of Technology (Fac. of Eng., Dept. of Electrical and Electronic Information Eng.), 2018 completed 1st term of doctorate at the University (Electrical and Electronic Information Eng.).

• 

  Toru Harigai, member. Assistant professor at Toyohashi University of Technology (Electrical and Electronic Information Eng.). Doctor of Eng. Research in plasma processes and renewable energies. Membership: JSAP, SSSJ.

• 

  Tsuyoshi Tanimoto, member. Assistant professor at Toyohashi University of Technology (Electrical and Electronic Information Eng.). Doctor of Eng. Research in plasma science and laser fusion. Membership: JSAP, JSPF, JACT.

• 

  Hirofumi Takikawa, member. In 1986 completed master's course at Toyohashi University of Technology (Grad. School of Eng., Electrical and Electronic Eng.). Then researcher at Sherbrooke University (Canada), assistant at Toyohashi University of Technology, now professor at the University (Electrical and Electronic Information Eng.). Doctor of Eng. Research in generation control and applications of plasma and nanomaterials, effective utilization of ecological energies. Membership: IEEE, JSAP, FNTG, JSES, JSER.

• 

  Toshiya Setaka, nonmember. In 1989 completed master's course at Nihon University (Grad. School of Sci. and Tech., Industrial Chem.). Now section head in Fuji Research Labs, Tokai Carbon Co., Ltd.

• 

  Junji Nakamura, nonmember. In 1988 completed doctorate at Hokkaido University (Grad. School of Sci.). Then postdoc fellow at Indiana University, assistant professor at University of Tsukuba (Dept. of Eng.), and other positions. Now professor at University of Tsukuba (Pure and Appl. Sci.). Doctor of Sci. Research in catalytic chemistry, surface science. Membership: ACS, FNTG, Catalysis Soc. Japan, JPI, Carbon Soc. Japan, Chem. Soc. Japan, SSSJ.

• 

  Yoshiyuki Suda, senior member. In 1997 completed master's course at Hokkaido University (Grad. School of Eng.). Then was employed by Paloma Co., Ltd., assistant at Hokkaido University (Grad. School of Eng.), and other positions. Now adjunct professor at Toyohashi University of Technology (Electrical and Electronic Information Eng.). Doctor of Eng. Research in synthesis and applications of carbon nanomaterials. Membership: JSAP, FNTG, ESJ.