Doping sp² carbon to boost the activity for oxygen reduction in an acidic medium: a theoretical exploration

Abstract

Metal-free carbon-based nanomaterials have been widely studied as a type of excellent electrocatalyst for the oxygen reduction reaction (ORR) in an alkaline medium due to their great stability, wide availability, environmental acceptability, and strong resistance to poisonous gas. They usually show a much inferior activity in an acidic medium, and a strategy to modify sp² carbon to facilitate the ORR in an acidic medium is critically needed. Herein, by taking carbon nanotubes (CNT) as the platform, and O₂ chemisorption ability as the descriptor of activity in an acidic medium, a series of N, B, BO_x, P and S mono- and multi-doped CNT(5,5) were studied using an ab initio modeling method. Eight configurations resulting from N mono-doping, BO_x doping and B/N, BO_x/N and P/N multi-doping are found to be potential active structures in an acidic medium due to their highly exothermic O₂ chemisorption. These results indicate the great potential of sp² carbon as an ORR electrocatalyst in an acidic medium, and also provide theoretical references for the exploration of carbon-based metal-free ORR electrocatalysts.

---

Introduction

Fuel cells are ideal environmentally-friendly energy conversion devices, which can efficiently extract electricity directly from fuels.^1,2 In a fuel cell, the fuel oxidation and oxygen reduction reaction (ORR) are both catalyzed by Pt-based electrocatalysts at the anode and cathode, respectively. Since the ORR is much more sluggish than the fuel oxidation, high Pt loading is required on the cathode to ensure efficiency. This drives up the cost of fuel cells due to the scarcity and high price of Pt. Besides, Pt-based ORR electrocatalysts have issues of poor durability, low selectivity, agglomeration, and susceptibility to CO poisoning.³ All these facts push intensive research efforts towards developing cheap and stable ORR electrocatalysts.^4,5

Recently, carbon-based metal-free ORR electrocatalysts doped with non-metal heteroatoms such as nitrogen,^6–11 boron,¹² phosphorus,^13,14 and sulfur¹⁵ demonstrated excellent ORR activity with high stability as well as CO and methanol tolerance, thus experiencing great progress.^16–20 In an alkaline medium, these metal-free carbons show high ORR activities, even close to or surpassing Pt-based catalysts.^6,7,13,15 But in an acidic medium, their performances were usually poor with onset potentials of 0.6–0.7 V (vs. NHE).^21,22 Very recently, N-doped graphene nanosheets reached a new record of ∼0.8 V (vs. NHE), indicating the great potential for metal-free carbons to catch up with Pt-based ORR electrocatalysts in an acidic medium.²³ To unlock more potential, the strategy for modifying metal-free carbons to facilitate the ORR in an acidic medium is of great importance. Herein, taking carbon nanotubes (CNT) as the platform, and O₂ adsorption ability as the descriptor of the ORR activity in an acidic medium, a series of N, B, BO_x, S, and P mono- and multi-doping configurations were studied using an ab initio modeling method. Eight configurations from N mono-doping, BO_x doping, and B/N, BO_x/N and P/N multi-doping are found to be potential ORR active structures in an acidic medium due to the considerable exothermic O₂ chemisorption energy. These results stress the importance of the regulation of doping configurations on the activity of carbon-based electrocatalysts, and also provide a strategy to explore metal-free ORR electrocatalysts, which is of great significance for the development of fuel cells.

Methods

Metallic armchair (5,5) single-walled carbon nanotubes (CNT(5,5)) with a length of 5 unit cells and diameter of 6.85 Å were used in the theoretical calculations. To avoid the edge effect, the terminal sites of the nanotubes were saturated with hydrogen atoms. Theoretical calculations were carried out with the Gaussian09 package²⁴ using an unrestricted density functional theory (DFT) method using Becke's hybrid three parameter nonlocal exchange functional combined with the Lee–Yang–Parr gradient corrected correlation functional (B3LYP).²⁵ The 6-31G(d,p) basis set was employed for all the elements. Natural bond orbital (NBO) analyses were performed at the same theoretical level using NBO 3.1, embedded in Gaussian09.^26,27 The spin multiplicity of the lowest energy state was adopted in the calculations. The adsorption energy (E_ads) for triplet O₂ on a nanotube was calculated as follows:²⁸

E_ads = E(NT + O₂) − E(NT) − E(O₂)

where E(NT + O₂), E(NT) and E(O₂) are the energy of the O₂-nanotube system, the relaxed nanotube and isolated triplet molecular O₂, respectively. Basis Set Superposition Error (BSSE) was also taken into account in the adsorption energy.

Results and discussion

Since the discovery of metal-free carbon-based ORR electrocatalysts, the origin of the ORR activity has been extensively studied. In our previous study, we proposed that in an alkaline medium, the ORR activity of carbon-based materials originates from the activated carbon π electrons.¹² Specifically, for N or B mono-doping, the carbon π bonds conjugate with the lone-pair orbital from N or the vacant 2p_z orbital from B. Such conjugation would elevate the highest occupied molecular orbital (HOMO) level of the doped carbon, thus facilitating electron transfer from the doped carbon to O₂.^12,20 Our extended research on regulating a B/N co-doped CNT with different doping microstructures provides further experimental and theoretical evidence for this understanding.²⁹ Moreover, our recent study demonstrated that the intrinsic defects in sp² carbon could also break the integrity of π conjugation and promote ORR activity.³⁰ Essentially, the activation of the carbon π electrons would lower the nanoscale work function, i.e., the minimum energy needed to transfer one electron from the carbon surface to the vacuum. A low work function facilitates the following outer-sphere reaction in the solution phase:^31–33

O₂ + e⁻ → O₂⁻

This reaction occurs without O₂ chemisorption at a potential ∼0.30 V (vs. NHE), just around the onset potential of the ORR in an alkaline medium for most metal-free carbons. From this reaction, the inert pristine O₂ turns into an active O₂⁻ species and reduction starts, leading to a greatly facilitated ORR in an alkaline medium. While in an acidic medium, the onset of the ORR shifts to a much higher potential window (0.6–1.0 V), in which the O₂⁻ species are too difficult to generate.^32,34 Therefore, in an acidic medium, to transfer the first electron to pristine O₂, the orbital overlaps between O₂ and the electrocatalyst are prerequisite. In other words, the chemisorption of pristine O₂ on the electrocatalyst is a necessary step. Therefore, in an acidic medium the transition metal-based electrocatalysts usually possess higher activity than metal-free carbons, due to a much better O₂ chemisorption capability.

Experimentally, CNTs can survive exposure to ambient O₂ at 500 °C, demonstrating a strong chemical inertness towards O₂.³⁵ Theoretically, there is a spin-mismatch between ground-state O₂ and pristine CNTs to prevent them from approaching each other. Fig. 1 shows this spin-mismatch through HOMO contours.¹² It is known that ground-state O₂ is a spin-triplet molecule with two unpaired electrons, while pristine CNT(5,5) is a typical spin-singlet system in which the electrons are all paired up. The π* orbital of O₂ forms an antibonding orbital with the π orbital of CNT(5,5) in one spin direction due to their opposite orbital signs (Fig. 1a), repelling O₂ away from CNT(5,5). This antibonding orbital counteracts the binding effects from the same-sign orbitals in another spin direction (Fig. 1b), ending up with weak physisorption. The physisorbed O₂–CNT(5,5) has a large adsorption distance of 3.52 Å, a small adsorption energy of −0.001 eV, and a negligible amount (0.1e) of charge transferred to O₂ (Table 1). With this understanding, the adsorption could be enhanced by modifying the state of O₂ or the CNT, respectively.

Fig. 1 Contour plots of the spin up (a) and spin down (b) HOMO of a physisorbed O₂–CNT(5,5) system, and the potential energy profiles for triplet (black) and singlet (red) O₂ approaching CNT(5,5) (c).

Table 1 Adsorption parameters for the O₂–nanotube systems

System	Eads^a (eV)	d(tube–O2)^b (Å)	d(O=O)^c (Å)	QO2^d (e)
a The adsorption energy.b The distance between the absorbed O2 and the nanotube.c The O=O bond length of the absorbed O2.d The charge transferred from the nanotube to the absorbed O2.
CNT(5,5)	−0.001	3.520	1.215	−0.100
g-NCNT(5,5)	0.067	1.577	1.310	−0.347
p-NCNT(5,5)	0.834	1.456	1.499	−0.567
g-N3CNT(5,5)	−1.152	1.349	1.448	−0.789
BCNT(5,5)	0.137	1.553	1.321	−0.452
BOCNT(5,5)	−0.032	1.543	1.319	−0.454
BO2CNT(5,5)	−0.516	1.494	1.328	−0.486
BNCNT(5,5)	0.124	1.526	1.331	−0.497
BN2CNT(5,5)	−0.296	1.501	1.338	−0.521
BN3CNT(5,5)	−1.622	1.465	1.510	−0.771
BONCNT(5,5)	0.159	1.496	1.335	−0.536
BO2NCNT(5,5)	−0.211	1.438	1.468	−0.744
PCNT(5,5)	0.828	1.759	1.341	−0.539
PNCNT(5,5)	−0.695	1.610	1.556	−1.067
PN2CNT(5,5)	−1.548	1.608	1.566	−1.079
PN3CNT(5,5)	−0.563	1.602	1.574	−1.068
SCNT(5,5)	0.631	1.530	1.314	−0.319
SNCNT(5,5)	0.708	1.523	1.316	−0.322
SN2CNT(5,5)	0.536	1.430	1.481	−0.727

---

The first way is to change the spin state of the adsorbate. Fig. 1c shows the potential energy profiles of triplet and singlet O₂ approaching CNT(5,5), which indicates that the chemisorption of O₂ on CNT(5,5) could be achieved by exciting the triplet O₂ to the spin-singlet state. Consequently, the spin-mismatch is eliminated, by which the antibonding orbital in Fig. 1a turns into a bonding one, and O₂ chemisorbs readily on CNT. But this spin-flip requires an activation energy of ∼1.5 eV (Fig. 1c) and radical methods (e.g. high temperature or ultraviolet (UV) light). Since fuel cells should be working under mild and controllable conditions, activating O₂ to enhance chemisorption on CNTs is out of consideration.

The second way is to change the spin state of the adsorbent. The spin mismatch could be resolved by doping CNTs with heteroatoms. The unpaired electrons could change the local spin states and interact with the pristine O₂. In the following, by tuning the N, B, BO_x, P, and S mono- and multi-doping configurations, we demonstrate the creation of active sites towards O₂ chemisorption, which would be potential ORR active sites in an acidic medium.

The N, B, BO_x, P, and S mono- and multi-doped CNT(5,5) are shown in Fig. 2, with the detailed doping configurations sketched in the insets. For N doped CNT(5,5), three types of doping configurations are modelled, i.e., graphitic N (g-NC₃), pyridinic N (p-NC₂), and one C atom surrounded by three graphitic N atoms (g-N₃C). The respective models are denoted as g-NCNT(5,5), p-NCNT(5,5) and g-N₃CNT(5,5) (Fig. 2a–c). For B and BO_x doping, the models with BC₃, BC₂O and BCO₂ structures are denoted as BCNT(5,5), BOCNT(5,5) and BO₂CNT(5,5), respectively (Fig. 2d–f). For B/N and BO_x/N multi-doping, the models with the configurations of BNC₂, BN₂C, BN₃, BONC, and BO₂N are denoted as BNCNT(5,5), BN₂CNT(5,5), BN₃CNT(5,5), BONCNT(5,5) and BO₂NCNT(5,5), respectively (Fig. 2g–k). For P mono-doping and P/N multi-doping, the models with PC₃, PNC₂, PN₂C and PN₃ are denoted as PCNT(5,5), PNCNT(5,5), PN₂CNT(5,5) and PN₃CNT(5,5), respectively, as shown in Fig. 2l–o. For S doping, since the atomic radius of an S atom is much larger than a C atom, the S dopant is more likely to stay at the edge of the CNT rather than in the plane. Hence, the SC₂ configuration is modelled by substituting an edge C atom with a S atom, as shown in SCNT(5,5) (Fig. 2p). Moreover, S/N multi-doping structures, i.e., SNC₂ and SN₂C, are modelled as SNCNT(5,5) and SN₂CNT(5,5) (Fig. 2q and r). The O₂ adsorption sites are also displayed in Fig. 2, which were scanned over all the possible sites around the dopants to ensure the lowest total energy of the O₂–nanotube system. The calculated results of the adsorption parameters are listed in Table 1.

Fig. 2 O₂ adsorption on doped CNT(5,5). Insets are the detailed doping configurations. (a) g-NCNT(5,5); (b) p-NCNT(5,5); (c) g-N₃CNT(5,5); (d) BCNT(5,5); (e) BOCNT(5,5); (f) BO₂CNT(5,5); (g) BNCNT(5,5); (h) BN₂CNT(5,5); (i) BN₃CNT(5,5); (j) BONCNT(5,5); (k) BO₂NCNT(5,5); (l) PCNT(5,5); (m) PNCNT(5.5); (n) PN₂CNT(5,5); (o) PN₃CNT(5,5); (p) SCNT(5,5); (q) SNCNT(5,5); (r) SN₂CNT. O: red, C: dark, H: gray, N: blue, B: pink, P: purple, S: yellow.

N has one electron more than C, and N-doping increases the electron density of the C π system. Thus, the nanoscale workfunction is lowered and the generation of active O₂⁻ species is facilitated, which boosts the ORR activity in an alkaline medium. However, as listed in Table 1, pristine O₂ has a weak chemisorption around g-NC₃ and p-NC₂, indicating the poor activity with these two configurations in an acidic medium. Very recently, p-NC₂ was designated as an active structure for the ORR in an acidic medium, based on activity tests at very low potentials (<0.3 V vs. NHE).²³ At such low potentials, the O₂⁻ generated in the solution phase is responsible for the ORR instead of the p-NC₂. We also found that O₂ chemisorption become strongly exothermic by −1.15 eV for the g-N₃C configuration. The chemisorbed O₂ extracts electrons to get reduced from the nanotube with this C atom as a bridge. The chemisorbed O₂ obtains −0.79e from g-N₃CNT(5,5), and the transferred charge weakens the O=O bond with an elongation from 1.21 Å in the gas phase to 1.45 Å in the absorbed state (Table 1). These results indicate that a combination of N doping configurations could greatly improve the adsorption ability of N-doped carbon, and create ORR active sites in an acidic solution. Thus, the experimentally observed ORR activity of N-doped carbons in acidic solutions could be understood from this perspective.

B has one electron less than C, leaving one unpaired electron in the substitutionally doped BCNT(5,5). As shown in Table 1, the chemisorption of O₂ on BCNT(5,5) is endothermic, with adsorption energy of ∼0.14 eV. Since B is very reactive towards oxidation, X-ray photoelectron spectroscopy (XPS) measurements of experimentally prepared BCNTs always demonstrate a signature of BC₂O and BCO₂ structures.^12,36,37 It is necessary to check the activity of these B oxide groups. BC₂O and BCO₂ emerge from the oxidation of a BC₃ structure, with one or two O atoms occupying the bridge position between B and C, as shown in Fig. 2e and f. Since O has a much larger electronegativity of 3.44 compared to 2.55 for C and 2.04 for B, B atoms are more positively charged in BO_xCNT(5,5) than in BCNT(5,5). As shown in Table 1, O₂ chemisorption enhances synchronously with increasing B oxidation state, and the charge transferred from the nanotube to the chemisorbed O₂ increases as well. As so, chemisorbed O₂ gets more and more reduced with an increase in the B oxidation state. For BO₂CNT(5,5), the chemisorbed O₂ extracts −0.486e, indicating good ORR activity.

BNCNT(5,5) has quite poor O₂ chemisorption, with chemisorption energy of 0.12 eV (Table 1). This is because with N as an electron donor and B as an electron accepter, the BNCNT(5,5) model is a closed shell system with no unpaired electrons, similar to pristine CNT(5,5). Such a closed shell is broken when more N atoms bond with one B. The chemisorbed O₂ obtained −0.52e from BN₂CNT(5,5) and −0.77e from BN₃CNT(5,5), with rather exothermic chemisorption energies of −0.30 and −1.62 eV, respectively. Since the chemisorbed O₂ is more reduced on BO_x groups (Table 1), a combination of BO_x and N dopants were also calculated for the BONCNT(5,5) and BO₂NCNT(5,5) models. The O₂ has endothermic chemisorption on BONCNT(5,5) with chemisorption energy of 0.159 eV. In contrast, the O₂ chemisorbs on BO₂NCNT(5,5) in a parallel style, with each O atom bonding to active B and C respectively. The chemisorption energy is −0.21 eV, and the chemisorbed O₂ obtained −0.74e from BO₂NCNT(5,5). The O=O bond length elongates to 1.47 Å, and O₂ is already partially reduced.

For the P mono-doped case, PCNT(5,5) has quite endothermic O₂ chemisorption, at ∼0.83 eV (Table 1). When P and N are multi-doped into CNT(5,5), the O₂ affinity gets greatly enhanced. As listed in Table 1, O₂ chemisorption on the P/N multi-doped configurations is quite exothermic with chemisorption energies of −0.70, −1.55 and −0.56 eV and transferred charges of −1.07e, −1.08e and −1.07e for PNCNT(5,5), PN₂CNT(5,5) and PN₃CNT(5,5), respectively, indicating the high ORR activity of P/N multi-doped CNT(5,5).

As listed in Table 1, SCNT(5,5), SNCNT(5,5) and SN₂–CNT(5,5) all demonstrate highly endothermic O₂ chemisorption with chemisorption energies of 0.63, 0.71 and 0.54 eV, indicating poor ORR activities.

So far, we can see that the conventional B, N, P, and S mono-doped CNTs show poor O₂ chemisorption (Fig. 3). Thus the experimental facts can be understood, most mono-doped metal-free carbons usually demonstrate low ORR activity in acidic solution. The first electron cannot be efficiently transferred to O₂ without chemisorption at low pH. For some recently reported N-doped carbons showing certain activity in acidic solution, the active sites would be certain special configurations, such as a C site surrounded by three graphitic N atoms, due to the great O₂ chemisorption capability. When B is in a high oxidation state or N is multi-doped with B, BO_x, and P, certain active sites are generated with quite large O₂ chemisorption energies (Fig. 3). From these theoretical results, it can be deduced that, experimentally, B/N and P/N multi-doped CNTs should be promising to obtain good ORR activity in acidic solution. More detailed theoretical works, such as spin density calculations, electronic structure analysis, and ORR free energy profile calculations, are needed to further judge the potential of these active structures. Furthermore, experimental tests of the theoretical understandings are highly expected.

Fig. 3 O₂ adsorption energy of all doping configurations.

Conclusions

In conclusion, the exothermic chemisorption of pristine O₂ should be the predetermination method for the ORR capabilities in an acidic medium of metal-free carbons. With this criterion, the O₂ chemisorption for a series of B, N, P, S and BO_x mono- or multi-doped CNT(5,5) were systematically studied. It is found that O₂ chemisorption on g-N₃C, BN₂C, BN₃, BCO₂, BO₂N, PNC₂, PN₂C, and PN₃ configurations is an exothermic process accompanied by considerable amounts of charge transferred to O₂ for reduction. The preferable O₂ chemisorption ensures the first electron transfer to O₂, and greatly lowers the overpotential of the ORR in an acidic medium. Thus, a strategy to further improve and develop carbon-based metal-free ORR electrocatalysts in an acidic medium should focus on enhancing the O₂ chemisorption. Experimentally, this could be achieved by creating g-N₃C and BCO₂ configurations through controlled N or B mono-doping, or B/N and P/N multi-doping. All these results indicate the great potential of doped sp² carbon in developing advanced practical electrocatalysts for fuel cells.

Acknowledgements

This work was jointly supported by NSFC (51232003, 21573107, 21473089, 21373108, 21173115), the “973” program (2013CB932902), and Suzhou science and technology plan projects (ZXG2013025).

References

1. B. C. H. Steele and A. Heinzel, Nature, 2001, 414, 345–352.
2. M. Winter and R. J. Brodd, Chem. Rev., 2004, 104, 4245–4269.
3. J. Wu and H. Yang, Acc. Chem. Res., 2013, 46, 1848–1857.
4. H. A. Gasteiger and N. M. Marković, Science, 2009, 324, 48–49.
5. M. K. Debe, Nature, 2012, 486, 43–51.
6. K. P. Gong, F. Du, Z. H. Xia, M. Durstock and L. M. Dai, Science, 2009, 323, 760–764.
7. W. Wei, H. Liang, K. Parvez, X. Zhuang, X. Feng and K. Müllen, Angew. Chem., Int. Ed., 2014, 53, 1570–1574.
8. J. Liang, Y. Zheng, J. Chen, J. Liu, D. Hulicova-Jurcakova, M. Jaroniec and S. Z. Qiao, Angew. Chem., Int. Ed., 2012, 51, 3892–3896.
9. P. Chen, T. Y. Xiao, Y. H. Qian, S. S. Li and S. H. Yu, Adv. Mater., 2013, 25, 3192–3196.
10. L. T. Qu, Y. Liu, J. B. Baek and L. M. Dai, ACS Nano, 2010, 4, 1321–1326.
11. S. Chen, J. Bi, Y. Zhao, L. Yang, C. Zhang, Y. Ma, Q. Wu, X. Wang and Z. Hu, Adv. Mater., 2012, 24, 5593–5597.
12. L. Yang, S. Jiang, Y. Zhao, L. Zhu, S. Chen, X. Wang, Q. Wu, J. Ma, Y. Ma and Z. Hu, Angew. Chem., Int. Ed., 2011, 50, 7132–7135.
13. Z. W. Liu, F. Peng, H. J. Wang, H. Yu, W. X. Zheng and J. Yang, Angew. Chem., Int. Ed., 2011, 50, 3257–3261.
14. D. S. Yang, D. Bhattacharjya, S. Inamdar, J. Park and J. S. Yu, J. Am. Chem. Soc., 2012, 134, 16127–16130.
15. Z. Yang, Z. Yao, G. Li, G. Fang, H. Nie, Z. Liu, X. Zhou, X. A. Chen and S. Huang, ACS Nano, 2011, 6, 205–211.
16. L. Dai, Y. Xue, L. Qu, H. J. Choi and J. B. Baek, Chem. Rev., 2015, 115, 4823–4892.
17. Y. Zheng, Y. Jiao, M. Jaroniec, Y. Jin and S. Z. Qiao, Small, 2012, 8, 3550–3566.
18. J. P. Paraknowitsch and A. Thomas, Energy Environ. Sci., 2013, 6, 2839–2855.
19. D. W. Wang and D. Su, Energy Environ. Sci., 2014, 7, 576–591.
20. L. J. Yang, Y. Zhao, S. Chen, Q. Wu, X. Z. Wang and Z. Hu, Chin. J. Catal., 2013, 34, 1986–1991.
21. J. I. Ozaki, N. Kimura, T. Anahara and A. Oya, Carbon, 2007, 45, 1847–1853.
22. D. S. Yu, Q. Zhang and L. M. Dai, J. Am. Chem. Soc., 2010, 132, 15127–15129.
23. D. Guo, R. Shibuya, C. Akiba, S. Saji, T. Kondo and J. Nakamura, Science, 2016, 351, 361–365.
24. M. J. Frisch, Gaussian 09. Revision A.02, Gaussian, Inc., Wallingford CT, 2009.
25. A. D. Becke, J. Chem. Phys., 1993, 98, 5648–5652.
26. E. D. Glendening, C. R. Landis and F. Weinhold, Wiley Interdiscip. Rev.: Comput. Mol. Sci., 2012, 2, 1–42.
27. E. D. Glendening, A. E. Reed, J. E. Carrpenter and F. Weinhold, NBO, Version 3.1, Theoretical Chemistry Institute, Univ., Wisconsin, Madison WI, 2001.
28. Y. X. Feng, F. F. Li, Z. P. Hu, X. G. Luo, L. X. Zhang, X. F. Zhou, H. T. Wang, J. J. Xu and E. G. Wang, Phys. Rev. B: Condens. Matter Mater. Phys., 2012, 85, 155454.
29. Y. Zhao, L. Yang, S. Chen, X. Wang, Y. Ma, Q. Wu, Y. Jiang, W. Qian and Z. Hu, J. Am. Chem. Soc., 2013, 135, 1201–1204.
30. Y. Jiang, L. Yang, T. Sun, J. Zhao, Z. Lyu, O. Zhuo, X. Wang, Q. Wu, J. Ma and Z. Hu, ACS Catal., 2015, 5, 6707–6712.
31. B. B. Blizanac, C. A. Lucas, M. E. Gallagher, M. Arenz, P. N. Ross and N. M. Markovic, J. Phys. Chem. B, 2004, 108, 625–634.
32. B. Blizanac, P. Ross and N. Markovic, Electrochim. Acta, 2007, 52, 2264–2271.
33. P. Quaino, N. Luque, R. Nazmutdinov, E. Santos and W. Schmickler, Angew. Chem., Int. Ed., 2012, 51, 12997–13000.
34. N. Ramaswamy and S. Mukerjee, J. Phys. Chem. C, 2011, 115, 18015–18026.
35. Y. J. Lee, H. H. Kim and H. Hatori, Carbon, 2004, 42, 1053–1056.
36. L. G. Jacobsohn, R. Schulze, M. M. da Costa and M. Nastasi, Surf. Sci., 2004, 572, 418–424.
37. T. Shirasaki, A. Derré, M. Ménétrier, A. Tressaud and S. Flandrois, Carbon, 2000, 38, 1461–1467.

---