Performance of FTO-free conductive graphene-based counter electrodes for dye-sensitized solar cells

Abstract

The attractiveness of graphene arises from its low cost, transparency, high electrical conductivity, chemical robustness, and flexibility, as opposed to the rising cost and brittleness of FTO. In particular, graphene is emerging as a possible substitute for FTO in flexible displays, touch screens, and solar cells. The main goal of our work is to develop new conductive oxide free graphene-based counter electrodes for dye sensitized solar cells (DSSCs). Graphene nanoplates are modified by silane coupling agent to introduce vinyl groups, and then mixed with polyurethane adhesive and cast on glass substrate. The film is irradiated by UV source and heat treated under Ar/H₂. A network graphene film is formed and tightly bonded on glass substrate with enhanced electrical conductivity. The structure of network graphene is investigated by XPS, TGA and SEM. The DSSCs with network graphene counter electrode exhibit power conversion efficiencies of 9.33%, much better than those with FTO electrodes (4.05%).

---

1. Introduction

Since the first archetype was proposed by O'Regan and Grätzel in 1991, dye-sensitized solar cells (DSSCs) have attracted increasing attention due to their relatively high photoelectric conversion efficiency (PCE) and low-cost fabrication.¹ Typical DSSCs have a sandwich structure, including a photoanode, a counter electrode (CE), dye molecules, and redox electrolyte filling between electrodes. As a crucial component, an ideal DSSC CE should possess high conductivity and superior catalytic properties.²

Platinum (Pt) deposited on indium tin oxide (ITO) or fluorine doped tin oxide (FTO) substrates has been widely used as a standard counter electrode for DSSCs.^3,4 However, ITO/FTO has many drawbacks such as brittleness, low adhesion to polymeric materials, a decrease in supply of indium, and the growing cost of indium metal, which have restricted their further applications.^5,6 In addition, the ITO/FTO layer is usually fabricated via a rather expensive vacuum sputtering process. Thus, there is an imperative demand in finding new materials to replace the ITO/FTO film as transparent conductive layers; in this regard, several materials including conductive polymers,⁷ carbon nanotubes (CNT),^8,9 graphene,^10,11 and metal grids¹² have been proposed to substitute ITO/FTO films. For instance, graphene can be a promising candidate due to its high carrier mobility (>2 × 10⁵ cm² V⁻¹ s⁻¹), large specific surface area (2600 m² g⁻¹), and outstanding thermal properties.^13–17 Furthermore, interconnected graphene can capture injected electrons and work as a “speedway” to enhance electron transport rate.^18–20

However, there are only a few reports on a complete replacement of ITO/FTO as the DSSC CE with other materials.^21–23 For example, Veerappan et al. used a graphite film as the CE with a PCE of 6.2%.²⁴ Chung et al. utilized ultraviolet oxidized multi-walled carbon nanotubes (MWCNTs) and glycerol doped PEDOT:PSS films as the CE with PCEs of 5.6% and 4.1%, respectively.²⁵ In this study, we developed a network structure graphene film to replace FTO film for the DSSC counter electrode. For the preparation of network structure graphene film, graphene oxide (GO) was modified by silane coupling agent to introduce vinyl group on graphene surface. Polyurethane (PU) copolymer was used as adhesive to mix with modified graphene (mGr), and the mixture was spin coated on glass substrates. Network structure graphene was then formed and tightly adhered to the glass substrate after UV irradiation, which were characterized by scanning electron microscopy (SEM), X-ray photoelectron spectroscopy (XPS) and thermogravimetric analysis (TGA). The DSSCs with network structure graphene as the CE exhibit an optimal PCE of 9.33%.

2. Experimental

2.1. Preparation of modified graphene (mGr) and reduced graphene (Gr)

Graphene oxide (GO) was prepared by Hummers method.²⁶ Briefly, 4.0 g of graphite was mixed with 4.0 g of NaNO₃ and 184 mL of 95% H₂SO₄ and the mixture was stirred in an ice bath. 24 g of KMnO₄ was added slowly into the graphite suspension while vigorous stirring and maintaining the temperature at 4 °C for 1 h. The graphite suspension was then stirred at 35 °C for 1 h. When 2 L of distilled water was added to the pasty graphitic suspension, the color of suspension changed to yellowish brown. After vigorous stirring at 98 °C for 1 h, 20 mL of 35% H₂O₂ was added to the suspension resulting in an immediate color change to golden yellow. The oxidized product, GO, was washed several times with distilled water via centrifugation until the solution became acid free followed by filtration and drying under vacuum. GO was finally obtained as gray powder.

The silane coupling agent, triethoxyvinylsilane (VTES, 98%), was purchased from Aladdin (China). 4 g of VTES was added into the dispersion (0.5 g of GO powder dispersed in 50 g of ethanol) followed by immediate adding 1 g of ammonia solution (28%). Excess VTES was beneficial to a higher extent of condensation reaction between hydroxyl groups on GO powder and methoxysilane groups of VTES. The solution was stirred with 300 rpm at 60 °C for 6 h. 5 g of hydrazine hydrate was then added and the reaction was kept at 70 °C for 6 h. After washing with ethanol and distilled water for three times, the mGr was dried under vacuum.

Gr was prepared by GO and hydrazine hydrate. 0.5 g of GO was dispersed in 100 g of distilled water. Then 5 g of hydrazine hydrate was added and the reaction was kept at 70 °C for 6 h to get reduced graphene nanosheets. After washing with distilled water for three times, the Gr was dried under vacuum.

2.2. Preparation of counter electrodes

As shown in Scheme 1, 1 g PU was mixed with 0.1 g mGr and 10 g ethyl acetate, and then stirred ultrasonically for 2 h to get mGr/PU. The mGr/PU mixtures were spin coated on glass substrates at 1200 rpm for 20 s. Some samples were under UV irradiation for 30 min to get CE-1 and CE-4, and CE-2 was without UV irradiation. CE-3 was prepared by the Gr/PU which was mixed with 1 g PU, 0.1 g Gr and 10 g ethyl acetate and without UV irradiation. Then CE-1, CE-2, CE-3 and CE-4 were heat treated under Ar/H₂ at 400 °C for 2 h to remove PU. Except for CE-4, CE-1, CE-2 and CE-3 were drop-coated with an isopropanol solution of 10 mM H₂PtCl₆, followed by a treatment at 400 °C for 30 min. As a comparison, conventional Pt electrodes were fabricated by drop-coating H₂PtCl₆ solution on FTO substrate and treated at 400 °C for 30 min.

Scheme 1 Preparation of CE-1, CE-2, CE-3 and CE-4.

2.3. Characterization

The property of modified graphene was studied with Fourier transform infrared spectroscopy (FTIR, Nicolet iS5, USA). The chemical state of the samples was analyzed using XPS with a spectrometer (EPMA-1600, Shimadzu, Japan) equipped with Al Kα X-ray source. The morphology of the samples was characterized using field-emission scanning electron microscopy (FESEM, JEOL JSM 7000F, Japan). TGA was conducted in a thermo-balance (TASDT Q600, Mettler-Toledo, USA) with approximately 5 mg of sample weighed. The temperature was raised from 100 to 800 °C at a rate of 10 °C min⁻¹ under a nitrogen flow to investigate main decomposition behavior.

TiO₂ photoanodes were supplied by Dalian Heptachroma SolarTech Co., Ltd. They were soaked in an ethanol solution of 0.5 mM ruthenium dye [N-719, cis-di(thiocyanato)-N,N-bis(2,2-bipyridyl-4-carboxylacid-4-tetrabutyl-ammonium carboxylate) ruthenium(II)] for 24 h, resulting in N-719 sensitized TiO₂ photoanode. The photoanode and various CEs were assembled in sandwich-type cell. Liquid electrolyte DHS-E36 was injected into space between the two electrodes. Finally, as prepared DSSCs (0.5 cm² in active area) were tested for photovoltaic characteristics (Scheme 2).

Scheme 2 Diagram of DSSC with graphene-based counter electrode.

3. Results and discussion

Silane coupling agent VTES can be anchored on the surface of modified graphene (mGr) via condensation reactions between –COOH, –OH and epoxy groups present on the mGr surface and silanol groups formed by the hydrolysis of alkoxysilanes in VTES. The progress of condensation reaction can be identified by simultaneous disappearance of characteristic bands assigned to methoxysilane groups.²⁷ Therefore, FTIR was used to understand the formation of anchored VTES on mGr surface and its chemical reduction via hydrazine hydrate. Fig. 1 shows FTIR spectra of GO, mGr and VTES. As shown in Fig. 1, characteristic GO absorption bands (curve c) appear at 3010–3680 cm⁻¹ (–OH stretching), 1732 cm⁻¹ (–C=O stretching), 1220 cm⁻¹ (C(O)–OH bending), and 1045 cm⁻¹ (C–O–C stretching), which provide evidence of carboxylic acid, epoxide, and hydroxide functional groups.²⁸ For the VTES (curve a), absorption peaks at 2840 cm⁻¹ and 2948 cm⁻¹ correspond to stretching vibrations of –CH₃ and –CH₂, respectively. The peak at 1076 cm⁻¹ is related to Si–O–C₂H₅ linkages and disappeared after modified reaction (curve b). The stretching vibration of vinyl groups at 1602 cm⁻¹ was observed for the mGr (curve b). On the other hand, C=O stretching vibrations in carboxyl group of GO (curve c) decreased after chemical reduction (curve b). After chemical reduction and VTES addition, new absorptions appeared due to the introduction of Si–O–Si groups (1169 cm⁻¹) onto the mGr surface, which indicates successful conduct of modified reaction. As a result, VTES was linked to the surface of mGr and resulted in UV curable vinyl groups on the mGr surface.

Fig. 1 FTIR spectra of VTES (a), mGr (b), and GO (c).

Reduced graphene flakes in polymer matrix tend to aggregate and stack in multi-layers due to van der Waals interactions. These aggregates usually have different physical and chemical properties from exfoliated monolayers.^29–31 As shown in Fig. 2, for reduced graphene solution (Gr/PU, Fig. 2b and d), Gr aggregated and settled at the bottom of bottle (Fig. 2d). In contrast, no aggregations appeared for mGr solution (mGr/PU, Fig. 2a and c) after standing for 24 h, which shows that mGr can provide better dispersion in polymer matrix due to modified functional groups on the mGr surface. The CEs were prepared by these mixture solutions. As shown in Fig. S1 and S2 in ESI,† the film thickness of CE-1 is about 25 μm which is much thinner than that of CE-2 (49 μm) and CE-3 (55 μm). The photographs (Fig. S3†) show that the film of CE-1 was sufficiently transparent that the letters below the film could be seen clearly. However a lot of aggregations of graphene layers appeared on the surface of CE-2 and CE-3, which caused them to have low transparency. The surface of CE-1 is rather uniform and tightly bonded on glass substrate, even after immersing in acetonitrile solution for 24 h.

Fig. 2 Images of mGr/PU (a and c) and Gr/PU (b and d) solutions before and after standing at room temperature for 24 h.

XPS measurements were conducted to determine surface chemical states, functional groups and elemental composition of CE-1, CE-2 and CE-3 and to investigate the change of functional groups induced by surface functionalization and UV-curable treatment. As presented in Fig. 3a, in the survey spectrum, the peaks centered on binding energies of 285 eV and 530 eV can be ascribed to C 1s and O 1s core-level emissions, respectively. And the binding energy of 400 eV corresponds to N 1s due to PU residues. The signals at 73 eV, 315–330 eV, and 519 eV can be assigned to Pt 4f, Pt 4d and Pt 4p, respectively. In the survey spectrum of CE-1 and CE-2, the binding energy of 103 eV corresponds to Si 2p due to the introduction of silane coupling agent.

Fig. 3 (a) XPS survey spectra of CE-1, CE-2 and CE-3 film, and C 1s core level XPS spectrum of CE-1 film (b), CE-2 film (c) and CE-3 film (d).

The C 1s spectrum of CE-2 (Fig. 3c) can be decomposed into five components: C=C in aromatic rings (284.3 eV), C–C in non-oxidized ring (284.5 eV), carbon in C–O bonds (286.1 eV), C=C vinyl group (284.6 eV) and C–C atoms in defective structure of graphene lattice (285.6 eV). In comparison, the changes in the functional groups of the C 1s emission of CE-1 can be seen in Fig. 3b. The C–C in aliphatic (284.8 eV) was observed and the vinyl group disappeared after UV irradiation. This indicates that network structure graphene film was successfully formed on glass substrate from the vinyl group of mGr, which was shown in Fig. S1, S2(a), S3 (a and d)† and 4(a). There are four components in C 1s spectrum of CE-3 (Fig. 3d), which include C=C in aromatic rings, C–C atoms in defective structure of graphene lattice, C–C in non-oxidized ring (284.5 eV) and C–O(H) (288.5 eV). This indicates that Gr was effectively reduced during the reduction process.

Fig. 4 SEM images of the film surface of CE-1 (a) and CE-3 (b).

The atomic percentage of three samples is summarized in Table 1. The Pt percentage of CE-1 was much higher than those of CE-2 and CE-3, which suggests that network structure graphene can effectively enhance the deposition of more Pt nanoparticles into modified surface and prevent the aggregation of Pt nanoparticles.^32–35

Table 1 Atomic percentages for CE-1, CE-2 and CE-3 as determined using XPS analyses

Samples	Atomic percentages (%)
	C	O	Si	Pt	N
CE-1	80.57	5.35	0.73	2.30	11.05
CE-2	80.49	6.89	0.75	0.33	11.54
CE-3	85.45	5.26	—	0.26	9.03

---

As shown in Fig. 4, graphene sheets in Gr samples (CE-3, Fig. 4b) physically assembled and the film was not uniformly covered with graphene, which is undesirable for the deposition of Pt particles and their catalytic activity.^36,37 Instead, as given in Fig. 4a, graphene sheets in mGr samples (CE-1) exhibit much wrinkled surface and are connected to each other to form network structure. The crumpled surfaces can provide more sites for the deposition of Pt particles, and the network structure can affiliate carrier transport during catalytic processes. This clearly demonstrates that the introduction of silane coupling agent and UV-curable process can effectively modify graphene surfaces and result in the formation of network structure graphene.

TGA was used to compare thermal decomposition behavior of CE-1, CE-2, and CE-3. At 800 °C, the residue of CE-3 is about 91%, which is higher than that of CE-1 and CE-2. It indicates that reduced graphene has high thermal resistance due to sp² structure. As given in Fig. 5, modified graphene (mGr in CE-1 and CE-2) demonstrates an initial weight loss of 2.5% and 2.3% between 100 and 400 °C, respectively, which are higher than that of Gr (CE-3). These initial weight loss is due to the presence of more functional groups on the surface of mGr. In comparison to CE-2, CE-1 shows higher thermal stability from 400 to 800 °C. It was reported that network structure of UV-curable films could improve thermal resistance.^38,39 Therefore, good thermal resistance of CE-1 indicates that network structure graphene can be formed via UV-curable treatment.

Fig. 5 TGA curve of CE-1, CE-2 and CE-3.

Table 2 summarizes photovoltaic parameters, such as open-circuit voltage (V_oc), short-circuit current density (J_sc), fill factor (FF), and power conversion efficiencies (η) of all DSSCs which were prepared in this work. The photocurrent density–voltage (J–V) characteristics of the DSSCs with various CEs are shown in Fig. 6.

Table 2 Photovoltaic performance of DSSCs with different counter electrodes. J_sc: short-circuit current density, V_oc: open-circuit voltage, FF: fill factor, and η: photoelectric conversion efficiency

Electrodes	Jsc (mA cm^−2)	Voc (V)	FF (%)	η (%)
CE-1	25.46	0.632	58.0	9.33
CE-2	6.47	0.129	44.3	0.36
CE-3	2.56	0.193	26.3	0.13
CE-4	15.12	0.447	53.3	3.60
Pt/FTO	20.86	0.646	30.1	4.05
Bare FTO	1.7208	0.231	24.6	0.098

---

Fig. 6 Photovoltaic characterizations of DSSCs with different counter electrodes.

DSSCs made with bare FTO performed poorly owing to the impotency of FTO substrate for I₃⁻ reduction. The Pt- and FTO-free CE-4 substrate shows a low resistivity of 0.74 Ω cm (as measured by the four-point probe method), a good solar conversion efficiency of 3.6% and fill-factor of 53.5%. This is mainly due to that high specific surface area and many chemical defects of network structure graphene can provide high catalytic activity toward iodine reduction. On the other hand, the relatively low solar conversion efficiency of 0.36% and 0.13% obtained for CE-2 and CE-3 respectively is attributed to two-dimensional surface structure of no-UV curable graphene. The observations above indicate that the optimal material for catalytic activity and conductivity is network structure graphene, CE-4.

An improved efficiency of 9.33% was obtained by using CE-1 which is higher than that of η = 4.05% with Pt/FTO electrode. As photovoltaic measurements were tested under the same conditions and with the same photoanode, the increase in J_sc can be ascribed to the network structure of CE-1 since the network can serve as a conduction pathway. The large surface area of CE-1 can result in high conductivity and improved catalytic activity of the electrode.^40–42 In comparison to CE-4, the increased efficiency of CE-1 can result from synergistic effects of network structure graphene and Pt embedded within graphene matrix (Fig. S4†).

Therefore, graphene structure can enhance active surface area for electrochemical reactions and offer efficient channels for electrolyte diffusion, meanwhile, an electronic conducting network created by graphene improves inter-layer conductive connection, which is beneficial to electron transfer. It is clearly found that the cells applying FTO as CE substrate generally exhibit unsatisfactory photovoltaic performance mainly due to their low fill factor; instead, by applying graphene-based substrate, the fill factor turns to be remarkably improved, the photocurrent J_sc of CE-1 only shows minor increased compared to those with FTO-based CE. Besides the advantages of graphene discussed above, it is deduced that the superior band matching of graphene coupled with CE of DSSCs can contribute to the cell performance improvement. In general, the work function of Pt and FTO is 5.65 eV and 4.4 eV, respectively; however, due to the extremely thin Pt layer with nanometer scale, people usually neglect the potential barrier caused by Pt-FTO structure for electrons coming from external circuit. The working function of graphene, on the other hand, could be tuned from 4.7 eV to 5.0 eV, depending on different synthetic processes. In this case, Pt–graphene structure can provide unobstructed channel for electrons to take part into the redox reaction with electrolyte; in other words, the energy barrier of Pt–graphene is much lower than that of Pt-FTO, which probably leads to efficient electron transfer on the interface of CEs, and finally generates outstanding fill factor observed from graphene-based DSSCs. As given in Table S1 and Fig. S4 and S5,† electrochemical impedance spectroscopy (EIS) measurements demonstrate that good photovoltaic performance and charge transfer characteristics of DSSCs with the network structure graphene as counter electrode can be attributed to fast diffusion of electrolyte species.

4. Conclusions

The network structure of graphene was generated from modified graphene after UV irradiation. We have successfully employed network structure graphene as an efficient counter electrode for DSSCs. The high conductivity of graphene suggests that FTO-free DSSCs may be developed using graphene. Advantageously, graphene layer may act as a conductive substrate so that it can successfully replace FTO. The charge transfer resistance and device performance of network structure graphene layers as a CE showed promising results. The low-cost FTO-free graphene based counter electrode optimized in the present work showed an energy conversion efficiency of 9.33%.

Acknowledgements

This work was partially supported by the International Science & Technology Cooperation Program of China (2014DFA60150) and the National Natural Science Foundation of China (51172113 & 51373086). L. F. Dong thanks financial support from the Malmstrom Endowed Fund.

Notes and references

1. B. O'regan and M. Grätzel, Nature, 1991, 353, 737–740.
2. A. Hagfeldt, G. Boschloo, L. Sun, L. Kloo and H. Pettersson, Chem. Rev., 2010, 110, 6595–6663.
3. A. Klein, C. Körber, A. Wachau, F. Säuberlich, Y. Gassenbauer, S. P. Harvey, D. E. Proffit and T. O. Mason, Materials, 2010, 3, 4892–4914.
4. S. S. Jeon, C. Kim, J. Ko and S. S. Im, J. Mater. Chem., 2011, 21, 8146–8151.
5. Y. Leterrier, L. Medico, F. Demarco, J. A. Månson, U. Betz, M. F. Escola, M. Kharrazi Olsson and F. Atamny, Thin Solid Films, 2004, 460, 156–166.
6. D. R. Cairns, R. P. Witte II, D. K. Sparacin, S. M. Sachsman, D. C. Paine, G. P. Crawford and R. R. Newton, Appl. Phys. Lett., 2000, 76, 1425–1427.
7. J. H. Burroughes, D. C. Bradley, A. R. Brown, R. N. Marks, K. Mackay, R. H. Friend, P. L. Burns and A. B. Holmes, Nature, 1990, 347, 539–541.
8. T. Y. Tsai, C. Y. Lee, N. H. Tai and W. H. Tuan, Appl. Phys. Lett., 2009, 95, 013107.
9. A. Abdelhalim, A. Abdellah, G. Scarpa and P. Lugli, Carbon, 2013, 61, 72–79.
10. L. Gao, W. Ren, J. Zhao, L. P. Ma, Z. Chen and H. M. Cheng, Appl. Phys. Lett., 2010, 97, 183109.
11. D. D. Nguyen, N. H. Tai, Y. L. Chueh, S. Y. Chen, Y. J. Chen, W. S. Kuo, T. W. Chou, C. S. Hsu and L. J. Chen, Nanotechnology, 2011, 22, 295606.
12. Y. T. Lai and N. H. Tai, ACS Appl. Mater. Interfaces, 2015, 7, 18553–18559.
13. W. Wei, K. Sun and Y. H. Hu, J. Mater. Chem. A, 2014, 2, 16842–16846.
14. X. Du, I. Skachko, A. Barker and E. Y. Andrei, Nat. Nanotechnol., 2008, 3, 491–495.
15. H. Wang and Y. H. Hu, Energy Environ. Sci., 2012, 5, 8182–8188.
16. H. Wang, K. Sun, F. Tao, D. J. Stacchiola and Y. H. Hu, Angew. Chem., 2013, 125, 9380–9384.
17. S. Unarunotai, Y. Murata, C. E. Chialvo, N. Mason, I. Petrov, R. G. Nuzzo, J. S. Moore and J. A. Rogers, Adv. Mater., 2010, 22, 1072–1077.
18. F. Xu, J. Chen, X. Wu, Y. Zhang, Y. X. Wang, J. Sun, H. C. Bi, W. Lei, Y. R. Ni and L. T. Sun, J. Phys. Chem. C, 2013, 117, 8619–8627.
19. H. Wang, S. L. Leonard and Y. H. Hu, Ind. Eng. Chem. Res., 2012, 51, 10613–10620.
20. X. Zhang, Y. Yang, S. Guo, F. Hu and L. Liu, ACS Appl. Mater. Interfaces, 2015, 7, 8457–8464.
21. J. Kwon, V. Ganapathy, Y. H. Kim, K. D. Song, H. G. Park, Y. Jun, P. J. Yoo and J. H. Park, Nanoscale, 2013, 5, 7838–7843.
22. G. Wang, S. Zhuo and W. Xing, Mater. Lett., 2012, 69, 27–29.
23. K. S. Lee, H. K. Lee, D. H. Wang, N. G. Park, J. Y. Lee, O. O. Park and J. H. Park, Chem. Commun., 2010, 46, 4505–4507.
24. G. Veerappan, K. Bojan and S. W. Rhee, ACS Appl. Mater. Interfaces, 2011, 3, 857–862.
25. D. J. Yun, Y. J. Jeong, H. Ra, J. M. Kim, T. K. An, M. Seol, J. Jang, C. E. Park, S. W. Rhee and D. S. Chung, J. Mater. Chem. C, 2015, 3, 7325–7335.
26. S. Bose, T. Kuila, M. E. Uddin, N. H. Kim, A. K. Lau and J. H. Lee, Polymer, 2010, 51, 5921–5928.
27. F. Bauer, H. J. Gläsel, E. Hartmann, H. Langguth and R. Hinterwaldner, Int. J. Adhes. Adhes., 2004, 24, 519–522.
28. D. C. Marcano, D. V. Kosynkin, J. M. Berlin, A. Sinitskii, Z. Z. Sun, A. Slesarev, L. B. Alemany, W. Lu and J. M. Tour, ACS Nano, 2010, 4, 4806–4814.
29. Z. S. Wu, D. W. Wang, W. Ren, J. Zhao, G. Zhou, F. Li and H. M. Cheng, Adv. Funct. Mater., 2010, 20, 3595–3602.
30. A. K. Sundramoorthy, Y. Wang, J. Wang, J. Che, Y. X. Thong, A. Chee, W. Lu, B. Mary and C. Park, Sci. Rep., 2015, 5, 10716.
31. R. B. Rakhi, W. Chen, D. K. Cha and H. N. Alshareef, J. Mater. Chem., 2011, 21, 16197–16204.
32. Z. Bo, D. Hu, J. Kong, J. Yan and K. Cen, J. Power Sources, 2015, 273, 530–537.
33. M. Wang, X. Song, Q. Yang, H. Hua, S. Dai, C. Hu and D. Wei, J. Power Sources, 2015, 273, 624–630.
34. S. L. Yang, Y. W. Huang, W. Zhu, B. C. Deng, H. Wang, Z. H. Zhang, P. F. Bao and G. Z. Wang, Int. J. Hydrogen Energy, 2014, 39, 15063–15071.
35. Y. Zhu, H. Cui, S. Jia, J. Zheng, Z. Wang and Z. Zhu, Electrochim. Acta, 2015, 178, 658–664.
36. L. Kavan, J. H. Yum and M. Grätzel, ACS Nano, 2010, 5, 165–172.
37. D. W. Zhang, X. D. Li, H. B. Li, S. Chen, Z. Sun, X. J. Yin and S. M. Huang, Carbon, 2011, 49, 5382–5388.
38. B. Pang, C. M. Ryu, X. Jin and H. I. Kim, Appl. Surf. Sci., 2013, 285, 727–731.
39. A. Asif, L. Hu and W. Shi, Colloid Polym. Sci., 2009, 287, 1041–1049.
40. D. W. Zhang, X. D. Li, H. B. Li, S. Chen, Z. Sun, X. J. Yin and S. M. Huang, Carbon, 2011, 49, 5382–5388.
41. L. Kavan, J. H. Yum and M. Grätzel, ACS Nano, 2010, 5, 165–172.
42. J. D. Roy-Mayhew, D. J. Bozym, C. Punckt and I. A. Aksay, ACS Nano, 2010, 4, 6203–6211.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra02854c

---