Liquid-based growth of polymeric carbon nitride films and their extraordinary photoelectrocatalytic activity

Abstract

Polymeric carbon nitride films were directly doposited onto different substrates by a fresh liquid-based reaction using thiourea as a precursor. The optical property was tuned by the insertion of barbituric acid. The films grew on the substrates naturally and adhered to them tightly. The obtained g-CN films show a remarkable photocurrent density of about 20 μA cm⁻² at 0.8 V vs. Ag/AgCl in 0.1 M KCl under AM 1.5 illumination.

---

Introduction

In 2009, a polymer-like semiconductor, graphitic carbon nitride (g-C₃N₄) was used as a photocatalyst to split water into hydrogen upon visible light irradiation by Wang and co-workers.¹ Since then, great attention has been focused on the utilization of graphitic carbon nitride in photocatalytic water splitting, organic pollutions degradation and photoelectric conversion.² Having high thermal and chemical stability in air, g-C₃N₄ is also easily prepared by heating carbon and nitrogen-rich precursors, such as urea,³ thiourea,⁴ cyanamide,¹ dicyanamide⁵ and melamine,⁶ etc. A great quantity of work has been done to enhance its photocatalytic activity, for example, metal^7,8 or nonmetal^9–14 elements doping, protonation,¹⁵ morphology control,^16–18 coupled with other semiconductors^19–21 and copolymerization.²² Most of the work only offer g-C₃N₄ powders, which are widely used in photocatalytic water splitting and organic pollutions degradation. When it comes to photoelectric conversion instruments, such as photoelectrochemical cell and photovoltaic cell, a continuous and homogeneous film is required. However, due to the strong van der Waals interaction between the atomic layers, condensed g-C₃N₄ is insoluble in most solvents.²³ Besides, due to the weak adhesion force, it is hard to combine this material with inorganic substrates by heating g-C₃N₄ powders. So it is extremely hard to prepare continuous carbon nitride films by traditional spin coating or blade coating method.

To date, several preliminary attempts have been made to deposit condensed g-C₃N₄ films onto solid substrates. Dong and co-workers have developed a novel in situ approach for effective immobilization of g-C₃N₄ on structured Al₂O₃ ceramic foam using dicyandiamide powder.²⁴ The immobilized g-C₃N₄ was used for photocatalytic removal of 600 ppb NO in air under indoor illumination, with the NO removal ratio reaching up to 77.1%. Yang and co-workers deposited g-C₃N₄ films in situ onto p-type semiconductor CuGaSe₂ substrates by thermal polycondensation of dicyandiamide.²⁵ A cathodic photocurrent of ∼5 μA cm⁻² was detected on the substrates at −0.2 V vs. NHE under 100 mW cm⁻² illumination in 0.1 M H₂SO₄. Shalom's group has reported a liquid-based reaction to deposit g-C₃N₄ film onto various substrates.²⁶ For the first time, a proof-of-concept solar cell was made by using g-C₃N₄ film as the electron acceptor layer, with a high V_oc exceeding 1 V. Recently, Bian and co-workers reported successful preparation of uniform g-C₃N₄ films on various substrates by a thermal vapor condensation (TVC) method.²⁷ The films showed better photoelectrocatalitic activity than those prepared by blade coating or spin coating method. Unfortunately, the films were only several tens of nanometers, which couldn't fully absorb visible light. So efforts are still needed to improve the quality of polymeric carbon nitride films.

Herein, we report a fresh liquid-based reaction to deposit uniform carbon nitride films in situ onto different substrates, including FTO glass and ms-TiO₂ films. Furthermore, the optical absorption of the obtained carbon nitride material is tuned by the insertion of barbituric acid (BA). The morphology, structure and composition of the obtained materials have been carefully characterized by SEM, XRD, XPS, FTIR and UV-Vis. Photoelectrochemical activities and electronic band structures of the products are investigated in a conventional three-electrode cell. Compared with the films made by blade coating method in our previous work,²⁸ carbon nitride films prepared by the liquid-based reaction method show a remarkable photocurrent increasing by a factor of 50. The results indicate a practical way to apply polymeric carbon nitride in PEC cell, DSSC cell and other photoelectric devices.

Experimental

Synthesis process

Unless stated otherwise, all chemicals used in this work were purchased form Sigma-Aldrich without further treatment. F-doped SnO₂ glasses (FTO, Pilkington, TEC7, 7 Ω sq⁻¹) were cleaned with detergent, acetone, ethanol and distilled water in an ultrasonic bath for 15 min, successively.

For the growth of compact bl-TiO₂ film,²⁹ 50 uL of 0.3 M titanium diisopropoxide dis(acetylacetonate) (Sigma-Aldrich, 75 wt% in isopropanol) in 1-butanol (Sigma-Aldrich, 99.8%) was spin-coated on a FTO glass at 700 rpm for 10 s, 1000 rpm for 10 s and 2000 rpm for 40 s, which was followed by heating at 125 °C for 5 min. Then mesoporous TiO₂ layer was spin-coated onto the pre-treated FTO substrate using TiO₂ paste (Dyesol 18NR-T) diluted in anhydrous ethanol at 1 : 3.5 by weight at 2000 rpm for 30 s. For the growth of carbon nitride films, 2 g of thiourea was put into a crucible, with the FTO or FTO/TiO₂ glass (length 2 cm and width 2 cm) covered on the bottom. The crucible was capped and placed in a muffle furnace, which was heated to 450 °C at a heating rate of 10 °C min⁻¹, and then kept at this temperature for 30 min. After that, the crucible was cooled to room temperature with furnace naturally. Carbon nitride grown on the back side of the FTO glass was scraped off by a tweezer. To modify the optical property of carbon nitride, 0.1 g of barbituric acid was mixed with thiourea, while other processes were the same as what was described above. The film thickness was about 130 μm measured by optical microscopy. The obtained samples were named as g-CN or g-CNB, where B was referred to barbituric acid. It should be noted that the products obtained here were not typical g-C₃N₄ as the polymerization was not complete at 450 °C. When the temperature reached 550 °C or even 500 °C, continuous films couldn't be obtained. Besides, appropriate heating rate was also vital.

Characterization

X-ray diffraction (XRD) patterns were recorded on a Bruker D8 Advance instrument using Cu-Kα radiation source. The bonding structures were measured with a Fourier transform infrared spectrometer (FTIR, PerkinElmer Frontier). X-ray photoelectron spectroscopy (XPS) was measured by a XPS instrument (Thermo ESCALAB 250, USA) with Al-Kα radiation. The peaks were calibrated by C 1s level at 284.8 eV. Surface morphologies of the samples were observed with a scanning electron microscope (SEM, Hitachi, S-4700). The optical properties were analyzed with a UV/VIS/NIR spectrometer (PerkinElmer, Lamda 950). The photoelectrochemical measurements were performed in a conventional three-electrode cell using an electrochemical analyzer (CHI-660D Shanghai Chenhua, China). A platinum wire electrode and an Ag/AgCl (saturated KCl) electrode were used as the counter and reference electrode, respectively. The electrolyte was a 0.1 M KCl aqueous solution, which was purged with nitrogen for 30 min prior to the measurements. The working electrode, FTO/g-CN or FTO/g-CNB was immersed in the electrolyte for 150 s before measurement. And the working electrode was irradiated from the back side. A solar simulator was used to offer the incident light (AM1.5, 100 mW cm⁻², CEL-S500, Ceaulight, China).

Results and discussion

The carbon nitride film growth process is illustrated in Fig. S1.† FTO glass was put on the bottom of a crucible. Thiourea was fully ground before it was transferred into the crucible, covering the FTO glass totally. The crucible was then capped and heated at 450 °C for 30 min in a muffle furnace in air. When the temperature increased to 180 °C, thiourea began to melt.³⁰ Further increasing the temperature, NH₃ and CS₂ were released to form cyanamide, which turned into carbon nitride eventually in the thermal polymerization process.^5,30,31 And optical photographs of the obtained films are shown in Fig. S2.†

The structure information of thiourea and carbon nitride are investigated by FTIR spectra. In Fig. 1, the spectrum of thiourea measured in our experiment is in accordance with data reported in related literatures.^32–34 The absorption bands at 3376, 3271, 3167, 3094 cm⁻¹ are assigned to NH₂ stretching vibration, while the band located at 1602 cm⁻¹ is attributed to NH₂ bending vibration. The band at 1471 cm⁻¹ is assigned to N–C–N stretching vibration. The strong bands at 1408 and 1082 cm⁻¹ are attributed to NH₂ rocking vibration, N–C–N and C=S stretching vibrations. The characteristic band at 731 cm⁻¹ is assigned to C=S stretching vibration. As for g-CN, the bands at 731, 1082, 1408 cm⁻¹ corresponding to C=S stretching vibration disappear, which results from CS₂ releasing, supported by TG-FTIR and TG/DTA-MS data.³⁰ The sharp absorption band at 807 cm⁻¹ appears, which can be assigned to the ring-sextant out-of-plane bending vibration characteristic of tri-s-triazine ring.³⁵ The absorption bands in the range of 1200–1600 cm⁻¹ are attributed to the stretching mode of C–N heterocycles, namely tri-s-triazine.³¹ This means the formation of a condensed carbon nitride polymer. The weak absorption bands at 3177, 3319 cm⁻¹ due to ν(N–H) suggest the residual of little hydrogen in g-CN. In the case of g-CNB, weaker ν(N–H) absorption bands are observed, suggesting a more condensed carbon nitride is formed. These results are in accordance with the next XRD data, as shown in Fig. 2.

Fig. 1 FTIR spectra of thiourea, g-CN and g-CNB.

Fig. 2 XRD patterns of g-CN and g-CNB.

As for g-CN, the XRD pattern is similar with that of related literature.³¹ The main peaks around 27.1° and 13.6° appear, while impure peaks around 12.6°, 22° and 25.3° also exist. This means the formation of graphitic-like networks corresponding to the main peaks, but the condensation is not complete. In the case of g-CNB, the sharp peak around 27.2° indicates the formation of typical graphitic-like layered structure, indexed as (002) peak, with the corresponding d-spacing to be 0.327 nm. The peak around 13.8° corresponding to in-plane ordering of tri-s-triazine forming 1D melon strands is weak, which means the formation of a low in-plane order. This results from the replacement of some ring nitrogen atoms by CH groups, which weakens the H-bonding network due to the insertion of barbituric acid into classical graphitic carbon nitride condensation process.²²

The chemical composition and structure of the products are further revealed by XPS analysis, as shown in Fig. S3.† Obviously, the characteristic peaks of C, N and O elements are observed for both g-CN and g-CNB. High-resolution XPS spectra of S 2p indicate no sulfur element is detected on the surface of g-CN, while little sulfur element exists in our g-CNB product. The binding energy of S 2p at 165.1 eV can be assigned to S–N bond resulting from the replacement of C by S in tri-s-triazine.³⁶ The atomic ratio of C : N : O : S of g-CN and g-CNB is determined to be 1 : 1.12 : 0.15 : 0 and 1 : 1.06 : 0.05 : 0.02 from XPS, respectively. The lower nitrogen content of g-CNB is due to the insertion of barbituric acid into classical graphitic carbon nitride condensation process. The S 2p binding energy of thiourea around 162.3 eV is not observed, indicating a complete conversion from thiourea to carbon nitride, which is in accordance with the previous FTIR analysis.³⁷ Based on the above analysis, the formation mechanism of g-CN and g-CNB is proposed, as shown in Fig. 3.

Fig. 3 Thermal polymerization reaction pathway for the formation of (a) g-CN and (b) g-CNB.

The optical properties of the products are evaluated by the UV-vis DRS. In Fig. S4,† the absorption edge of g-CNB exhibits a remarkable redshift relative to that of g-CN. As indirect band gap material, its optical band gap decreases from 2.67 eV to 1.81 eV. This is due to the co-doping of C and S atoms, which changes the optical and electronic properties of g-CNB.²² The wavefunction of the valence band for polymeric carbon nitride is a combination of the HOMO levels of the melem monomer. The conduction band is assigned to the LUMO of the melem monomer. The HOMO of melem is derived from heterocyclic nitrogen P_z orbitals, and LUMO mainlyconsists of carbon P_z orbital.¹ As for g-CNB, carbon doping together with sulfur doping affects the band structure of carbon nitride by replacing related atoms in the tri-s-triazine rings. In principle, the properties of carbon nitride can be tuned by organic protocols according to different requirements.

A standard photoelectrochemical (PEC) cell in aqueous solution is frequently used to characterize the photoactivity of a new semiconductor material.³⁸ In this manner, photocurrent response of different photoelectrodes are investigated. As a base line for comparison, photocurrent response of bare FTO glass is also measured, which is inappreciable, as shown in Fig. S5.† Surface and cross-sectional SEM images of different photoelectrodes are shown in Fig. 4. As for FTO/bl-TiO₂, a layer of smooth and compact TiO₂ is grown on the surface of FTO glass. Its thickness is about 90 nm according to Fig. 4(b). Then a layer of mesoporous TiO₂ is deposited on the FTO/bl-TiO₂ and the total thickness of bl-TiO₂ and ms-TiO₂ is about 220 nm according to Fig. 4(d) and (f). The mesoporous structure make it easier for the growth of g-CN film. When thiourea begins to melt, the liquid will infiltrate into the mesopores. In the following thermal polymerization process, solid g-CN will fill in the mesopores. For FTO/bl-TiO₂/ms-TiO₂/g-CN photoelectrode, the surface of g-CN film is very rough. The layers of TiO₂/g-CN can be distinguished easily. The g-CN film grows on the substrate naturally and combines with it strongly. They won't fall off from the substrate even after they are immersed in the electrolyte for 1 h. The immobilization of g-CN film on the substrate can be ascribed to some physical and chemical interaction between g-CN and FTO or ms-TiO₂ formed during the growth process.

Fig. 4 Surface and cross-sectional SEM images of different photoelectrodes. (a) and (b) FTO/bl-TiO₂; (c) and (d) FTO/bl-TiO₂/ms-TiO₂; (e) and (f) FTO/bl-TiO₂/ms-TiO₂/g-CN.

It should be noted that continuous g-CN film can't be obtained on the FTO/bl-TiO₂ substrate due to the poor adhesion with the smooth TiO₂ layer. The case is different for FTO or FTO/bl-TiO₂/ms-TiO₂ substrate due to the rough surface of FTO or ms-TiO₂.

As for FTO/g-CN photoelectrode, a typical anodic photocurrent curve is observed. This indicates that g-CN contains n-type semiconductor characteristic, which is further supported by Mott–Schottky plots with a positive slope as shown in Fig. S6.†^39–41 The case is the same for FTO/g-CNB and FTO/bl-TiO₂/ms-TiO₂ photoelectrodes in Fig. 5. Due to the barrier of preparing homogeneous polymeric carbon nitride films, Zhang and co-workers made carbon nitride photoelectrodes by frequently-used blade coating technique.²³ The electronic band structure and photocurrent generation of carbon nitride were studied in a standard PEC cell for the first time. For modified rpg-C₃N₄, the photocurrent density (λ > 420 nm, 150 W Xe lamp) at 0.4 V vs. Ag/AgCl in 0.1 M KCl was about 0.5 μA cm⁻², which was the highest among the different kinds of carbon nitride photoelectrodes. In our previous work, the photocurrent density at 0.4 V vs. Ag/AgCl in 0.1 M KCl was only about 0.1 μA cm⁻² for the pristine carbon nitride photoelectrode made by blade coating technique.²⁸ In addition to its low conductivity, the poor PEC activities were mainly caused by grain boundary effects and weak adhesion with the substrates, which limited the separation and transport of photo-generated carriers. Fortunately, the situation can be improved by the liquid-based deposition technique. The photocurrent density increases to 5 μA cm⁻² at 0.4 V vs. Ag/AgCl in 0.1 M KCl for FTO/g-CN photoelectrode, as shown in Fig. S7.† When the potential increases to 0.8 V, the photocurrent density is as high as 20 μA cm⁻². As for FTO/bl-TiO₂/ms-TiO₂/g-CN photoelectrode, the photocurrent response is further enhanced by the formation of TiO₂/g-CN heterojunction, which accelerates the separation of photo-generated carriers. According to the electrochemical Mott–Schottky plots in Fig. S6,† the flat band potential of g-CN is about −1.2 V. Combined with the above optical band gap, the energy scheme of g-CN and related mechanism are shown in Fig. 6.

Fig. 5 Current-potential curves of (a) FTO/g-CNB, (b) FTO/g-CN, (c) FTO/TiO₂ and (d) FTO/TiO₂/g-CN in 0.1 M KCl.

Fig. 6 The energy scheme of g-CN/TiO₂ system and proposed charge transfer path in 0.1 M KCl (pH = 7).

The photocurrent response decreases a lot for FTO/g-CNB photoelectrode, although it shows enhanced light absorption and polymerization degree. This may originate from higher recombination center concentration caused by C and S co-doping, which increases the recombination of photo-generated carriers.⁴² Electrochemical impedance technique provides a powerful mean for the study of charge transfer process at semiconductor/electrolyte interfaces.⁴³ To further reveal the difference of photoelectrocatalytic activities between g-CN and g-CNB, electrochemical impedance spectra are measured in the dark and under light at 0.4 V vs. Ag/AgCl in the frequency region of 0.1–10⁵ Hz with AC amplitude of 10 mV. For FTO/g-CNB photoelectrode, the diameter of the semicircular Nyquist plot is larger than that of FTO/g-CN photoelectrode whether there is light or not. This means that the charge transfer resistance is higher from FTO/g-CNB photoelectrode to the electrolyte, which is in accordance with lower photocurrent response.

Conclusions

We have demonstrated a fresh liquid-based reaction to deposit uniform carbon nitride films in situ onto different substrates by the thermal polymerization of thiourea. The carbon nitride films grow on the substrates naturally and adhere to them tightly. The optical property can be tuned by the insertion of barbituric acid. The pristine g-CN films on FTO glass show a remarkable photocurrent density of about 20 μA cm⁻² at 0.8 V vs. Ag/AgCl in 0.1 M KCl, which is much larger than that reported in our previous work. The enhanced photoelectrocatalytic activity results from reduced grain boundary effects and strong adhesion with the substrates, which limit the separation and transport of photo-generated carriers. Besides, it is proved that the formation of TiO₂/g-CN heterojunction can accelerate the separation efficiency of photo-generated carriers. We believe that the pristine results indicate a practical way towards applying polymeric carbon nitride in PEC cell, DSSC cell and other photoelectric conversion devices.

Acknowledgements

The authors would like to thank Dr Jiantao Bian and Dr Anjun Han for the helpful discussions. This work was supported by International S&T Cooperation Program of China (2015DFA60570) and Key project of Zhangjiang National Innovation Demonstration Zone Special Development Fund (ZJ2015-ZD-001).

References

1. X. C. Wang, K. Maeda, A. Thomas, K. Takanabe, G. Xin, J. M. Carlsson, K. Domen and M. Antonietti, Nat. Mater., 2009, 8, 76.
2. Y. J. Zhang, T. Mori and J. H. Ye, Sci. Adv. Mater., 2012, 4, 282.
3. Y. W. Zhang, J. H. Liu, G. Wu and W. Chen, Nanoscale, 2012, 4, 5300.
4. F. Dong, Y. J. Sun, L. W. Wu, M. Fu and Z. B. Wu, Catal. Sci. Technol., 2012, 2, 1332.
5. A. Thomas, A. Fischer, F. Goettmann and M. Antonietti, J. Mater. Chem., 2008, 18, 4893.
6. Y. H. Zhang, Q. W. Pan, G. Q. Chai, M. R. Liang, G. P. Dong, Q. Y. Zhang and J. R. Qiu, Sci. Rep., 2013, 3, 1943.
7. X. C. Wang, X. F. Chen, A. Thomas, X. Z. Fu and M. Antonietti, Adv. Mater., 2009, 21, 1609.
8. B. Yue, Q. Y. Li, H. Iwai, T. Kako and J. H. Ye, Sci. Technol. Adv. Mater., 2011, 12, 034401.
9. Y. J. Zhang, T. Mori, J. H. Ye and M. Antonietti, J. Am. Chem. Soc., 2010, 132, 6294.
10. Y. Wang, J. S. Zhang, X. C. Wang, M. Antonietti and H. R. Li, Angew. Chem., Int. Ed., 2010, 49, 3356.
11. Y. Wang, Y. Di, M. Antonietti, H. R. Li, X. F. Chen and X. C. Wang, Chem. Mater., 2010, 22, 5119.
12. S. C. Yan, Z. S. Li and Z. G. Zou, Langmuir, 2010, 26, 3894.
13. G. Liu, P. Niu, C. H. Sun, S. C. Smith, Z. G. Chen, G. Q. Lu and H. M. Cheng, J. Am. Chem. Soc., 2010, 132, 11642.
14. Z. W. Zhao, Y. J. Sun, F. Dong, Y. X. Zhang and H. Zhao, RSC Adv., 2015, 5, 39549–39556.
15. Y. J. Zhang, A. Thomas, M. Antonietti and X. C. Wang, J. Am. Chem. Soc., 2009, 131, 50.
16. J. S. Zhang, F. S. Guo and X. C. Wang, Adv. Funct. Mater., 2013, 23, 3008.
17. Y. H. Ishida, L. Chabanne, M. Antonietti and M. Shalom, Langmuir, 2014, 30, 447.
18. J. Li, C. B. Cao and H. S. Zhu, Nanotechnology, 2007, 18, 115605.
19. J. Fu, Y. L. Tian, B. B. Chang, F. N. Xi and X. P. Dong, J. Mater. Chem., 2012, 22, 21159.
20. Y. L. Tian, B. B. Chang, J. L. Lu, J. Fu, F. N. Xi and X. P. Dong, ACS Appl. Mater. Interfaces, 2013, 5, 7079.
21. J. S. Zhang, M. W. Zhang, R. Q. Sun and X. C. Wang, Angew. Chem., 2012, 124, 10292.
22. J. S. Zhang, X. F. Chen, K. Takanabe, K. Maeda, K. Domen, J. D. Epping, X. Z. Fu, M. Antonietti and X. C. Wang, Angew. Chem., Int. Ed., 2010, 49, 441.
23. Y. J. Zhang and M. Antonietti, Chem.–Asian J., 2010, 5, 1307.
24. F. Dong, Z. Y. Wang, Y. H. Li, W. K. Ho and S. C. Lee, Environ. Sci. Technol., 2014, 48, 10345–10353.
25. F. Yang, M. Lublow, S. Orthmann, C. Merschjann, T. Tyborski, M. Rusu, M. Kanis, A. Thomas, R. Arrigo, M. Hävecker and T. Schedel-Niedrig, ChemSusChem, 2012, 5, 1227.
26. J. S. Xu, T. J. K. Brenner, L. Chabanne, D. Neher, M. Antonietti and M. Shalom, J. Am. Chem. Soc., 2014, 136, 13486.
27. J. C. Bian, Q. Li, C. Huang, J. F. Li, Y. Guo, M. Zaw and R. Q. Zhang, Nano Energy, 2015, 15, 353.
28. X. Q. Wei, Y. Qiu, W. Y. Duan and Z. X. Liu, RSC Adv., 2015, 5, 26675.
29. N. Ahn, D. Y. Son, I. H. Jang, S. M. Kang, M. Choi and N. G. Par, J. Am. Chem. Soc., 2015, 137, 8696.
30. J. Madarász and G. Pokol, J. Therm. Anal. Calorim., 2007, 88, 329.
31. G. G. Zhang, J. S. Zhang, M. W. Zhang and X. C. Wang, J. Mater. Chem., 2012, 22, 8083.
32. P. Bombicz, I. Mutikainen, M. Krunks, T. Leskelä, J. Madarász and L. Niinistö, Inorg. Chim. Acta, 2004, 357, 513.
33. A. Yamaguchi, R. B. Penland, S. Mizushima, T. J. Lane, C. Curran and J. V. Quagliano, J. Am. Chem. Soc., 1958, 80, 527.
34. M. M. El-Etri and W. M. Scovell, Inorg. Chem., 1990, 29, 480.
35. B. V. Lotsch, M. Döblinger, J. Sehnert, L. Seyfarth, J. Senker, O. Oeckler and W. Schnick, Chem.–Asian J., 2007, 13, 4969.
36. J. D. Hong, X. Y. Xia, Y. S. Wang and R. Xu, J. Mater. Chem., 2012, 22, 15006.
37. R. A. Walton, Coord. Chem. Rev., 1980, 31, 183.
38. Y. J. Zhang, T. Mori, L. Niu and J. H. Ye, Energy Environ. Sci., 2011, 4, 4517.
39. M. Long, W. M. Cai and H. Kisch, J. Phys. Chem. C, 2008, 112, 548.
40. M. A. Alpuche-Aviles and Y. Y. Wu, J. Am. Chem. Soc., 2009, 131, 3216.
41. J. S. Zhang, J. H. Sun, K. Maeda, K. Domen, P. Liu, M. Antonietti, X. Z. Fu and X. C. Wang, Energy Environ. Sci., 2011, 4, 675.
42. J. S. Xu, I. Herraiz-Cardona, X. F. Yang, S. Gimenez, M. Antonietti and M. Shalom, Adv. Opt. Mater., 2015, 3, 1052.
43. W. H. Leng, Z. Zhang, J. Q. Zhang and C. N. Cao, J. Phys. Chem. B, 2005, 109, 15008.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra16699g

---