Introduction 

Hematite has emerged as a good photocatalyst for effiffifficient solar water splitting due to its favorable optical band gap (2.1–2.2 eV), extraordinary chemical stability in an oxidative environment, abundance, and low cost.1–9 According to the theoretical prediction, the solar-to-hydrogen efficiency of hematite can be 16.8% and the water splitting photocurrent can be 12.6 mA cm 2.4 However, the practical performance of hematite for solar water splitting is far away from the ideal case which has been limited by several factors such as poor conductivity, short lifetime of the excited-state carrier (10 12 s), poor oxygen evolution reaction (OER) kinetics, short hole diffusion length (2–4 nm), and improper band position for unassisted water splitting.5–9 Enormous efforts have been focused on improving the performance of hematite photoelectrodes.3–16 The two most important metrics for the photocurrent curve to evaluate the performance of hematite are the plateau current and the onset potential.9 A high water splitting photocurrent of over 3 mA cm 2 at 1.23 V vs. reversible hydrogen electrode (RHE) and a plateau photocurrent of 3.75 mA cm 2 have been achieved by a cauliower-type hematite nanostructure with IrO2 catalyst.9 By coupling Pt-doping and surface modication with Co–Pi to pristine hematite, a record-breaking photocurrent of 4.32 mA cm ² at 1.23 V vs. RHE was recently reported.5 The results also suggested that a good starting hematite was very important for further modication to achieve high performance.5 The introduction of oxygen vacancies was reported to be an effective way to improve the photocurrent of pristine hematite without elemental doping or surface catalysts.7,10,11 Y. Ling et al. recently reported the highly photoactive hematite nanowire arrays containing oxygen vacancies which showed a photocurrent of 1.82 mA cm 2 at 1.23 V vs. RHE and a maximum photocurrent of 3.37 mA cm 2 at 1.5 V vs. RHE, standing for an excellent photocurrent density achieved by an undoped hematite nanowire.10 We also developed a simple method by coupling Tidoping and oxygen vacancies in hematite nanostructures to prepare photoelectrodes with a high photocurrent of 2.25 mA cm 2 at 1.23 V vs. RHE and a remarkable maximum value of 4.56 mA cm ² at 1.6 V vs. RHE.17 All the methods used to produce oxygen vacancies in those reports were mainly by controlling the oxygen content in the sintering process, which offered an oxygen-decient atmosphere in the formation process of hematite (denoted oxygen-deciency treatment here).10,11,17 Although the reported photocurrent can be signi- cantly improved by the oxygen-deciency treatment in the sintering process, the onset potential of the photocurrent curve in the literature is relatively very high (about 1.0 V vs. RHE).10,11,17 It was also argued that the presence of surface defect states of Fe2+ in hematite would result in a recombination of photoexcited holes with electrons,14 which could be a possible reason for the high onset potential of hematite with oxygen vacancies. Another effective way to produce oxygen vacancies could be sintering hematite in a reductive atmosphere such as H2, which was widely used to improve the water oxidation performance of various materials such as TiO2, WO3, and ZnO.18–20 However, hematite can be easily reduced in H2 to produce magnetite which is photo-inactive,10 hindering the use of H2 treatment to improve the performance of hematite. Here we develop a facile way to prepare H2-treated hematite nanostructures by a simple pyrolysis of NaBH4 in a crucible. The optimized hematite photoelectrode showed a remarkable photocurrent of 2.28 mA cm ² at 1.23 V vs. RHE, which was over 2.5 times higher than that for hematite before the H2 treatment. Moreover, the H2-treated hematite nanostructures showed a low onset potential of about 0.87 V vs. RHE, which was cathodically shied by about 120 mV (from 0.99 to 0.87 V vs. RHE) when compared to the oxygen deciency treated hematite. The cathodic shi of the onset potential should be attributed to the surface effect of H2 treatment compared to the bulk effect by controlling the oxygen content, which was conrmed by X-ray absorption spectroscopy. Our results suggest that the presence of surface defect states of Fe2+ in hematite is not the reason for high onset potential and the optimized oxygen vacancy content in the surface may lower the onset potential by improving the conductivity of hematite and then reducing the recombination of photoexcited holes with electrons. The H2-treated hematite with high efficiency could also be used as a good starting material to achieve better performance for practical applications with further modications such as surface catalysts or elemental doping. Experimental section Preparation of a-Fe2O3 photoanodes Hematite nanostructures were prepared on a uorine-doped SnO2 (FTO, Nippon Sheet Glass, Japan, 14 ohm sq 1 ) glass by a hydrothermal method.6 A Teon-lined stainless steel autoclave was lled with 100 ml aqueous solution containing 0.15 M ferric chloride (FeCl3$6H2O, Sinopharm Chemical Reagent Co., Ltd.), 1 M sodium nitrate (NaNO3, Sinopharm Chemical Reagent Co., Ltd) and 100 ul HCl (45.3–45.8 wt%). The FTO glass was cleaned with acetone, ethanol and deionized water. The cleaned FTO glass slide (35 mm  50 mm  2 mm) was put into the autoclave and heated at 95  C for 4 h. A uniform layer of yellow color lm (FeOOH) was formed on the FTO substrate. The FeOOHcoated substrate was washed with deionized water and subsequently sintered in air at 550  C for 2 h and then 750  C for 10 min. The prepared hematite photoanodes were labeled as pristine hematite. The H2 treatment of pristine hematite was performed by the pyrolysis of NaBH4 (Sinopharm Chemical Reagent Co., Ltd.) in a crucible. Various amounts of NaBH4 (0, 2, 5, 8, 10 mmol) were sintered at 500  C for 30 min in a crucible (10 cm3 ), which was covered by the FTO-hematite (pristine hematite on FTO glass) slide (35 mm  50 mm  2 mm) with the pristine hematite side facing down (the experimental set up can be found in ESI, Fig. S1†). The pyrolysis temperature of NaBH4 is about 400  C and H2 will be released in the pyrolysis process which can be used for the surface treatment of pristine hematite. The amounts of NaBH4 can be used to adjust the H2 concentration. The sintering temperature was optimized to be 500  C. For PEC measurements, the H2-treated FTO-hematite slide was cut into 10 mm  20 mm  2 mm pieces. The oxygen-deciency treated hematite sample was prepared by a similar method but the FeOOH-coated substrate was sintered in a tube furnace with a controlled oxygen-decient atmosphere (Ar + air) at 550  C for 2 h. The oxygen-decient atmosphere was achieved by applying vacuum to the system down to a pressure of 3.0  10 2 Torr, and then relling with ultra-high purity Ar and air. The ow-in rates of Ar and air were controlled by a mass ow controller (MFC) and the total pressure in the tube furnace was kept to be 1 Torr. The partial oxygen pressure (pO2) was estimated by the equation: pO2 ¼ 21%  (air ux)/(air ux + Ar ux)  total pressure (1 Torr). The partial oxygen pressure was optimized and the sample with best performance was used for comparison in this work. Structural characterization SEM images of hematite nanostructures and energy dispersive X-ray analysis (EDX) spectroscopy were taken on a FEI-quanta 200 scanning electron microscope with an acceleration voltage of 20 kV. TEM and High-Resolution TEM (HRTEM) images were obtained with a FEI/Philips Techai 12 BioTWIN transmission electron microscope and a CM200 FEG transmission electron microscope, respectively. X-ray Diffffraction (XRD, PANalytical, Empyrean) and X-ray photoelectron Spectrometry (XPS, Kratos AXIS UltraDLD) were also used for structure characterization. Xray Absorption Near Edge Structure (XANES) and Extended X-ray Absorption Fine Structure (EXAFS) data were collected on beamline 14W at the Shanghai Synchrotron Radiation Facility. X-ray absorption spectroscopy (XAS) experiments at the O Kedge were performed at the So X-ray Spectroscopy station of the Beijing Synchrotron Radiation Facility (BSRF). The spectra were recorded at room temperature with a resolution of 0.3 eV at the O K-edge with the total electron yield (TEY) detection mode. PEC measurements Hematite photoanodes on the FTO substrate were covered by non-conductive hysol epoxy except for a working area of 0.1 cm2. All PEC measurements were carried out using a CHI 660D electrochemical workstation in a three-electrode electrochemical cell with a Pt wire as a counter electrode and an Ag/ AgCl electrode as a reference. The electrolyte was an aqueous solution of NaOH with a pH of about 13.6, bubbled with N2 for 20 min before the measurement. The measured potentials were converted into the RHE scale according to the Nernst equation.6 In a typical experiment, the potential was swept from 0.6 V to 1.8 V vs. RHE at a scan rate of 50 mV s 1 . Xenon high brightness cold light sources (XD-300) coupled with a lter (AM 1.5G) were used as the white light source and the light power density of 100 mW cm 2 (spectrally corrected) was measured with a power meter (Newport, 842-PE). IPCE was measured using a Xenon lamp (CEL-HXF300/CEL-HXBF300, 300W) coupled with a monochromator (Omni-l3005). 

Capacitance was derived from the electrochemical impedance obtained at each potential with 10 000 Hz frequency in the dark. Mott–Schottky plots were generated from the capacitance values. Results and discussion H2-treated hematite nanostructures were prepared by the pyrolysis of NaBH4 in a crucible covered by a uorine-doped SnO2 glass (FTO). The pyrolysis temperature of NaBH4 is about 400  C and at the experimental temperature of 500  C H2 will be released in the pyrolysis process for the H2 treatment of pristine hematite. The amounts of NaBH4 in the crucible can be used to precisely adjust the H2 concentration for oxygen vacancies before the hematite is converted to magnetite. The detailed experiments are described in the Experimental section. Illustration of the experimental setup can be found in ESI, Fig. S1.† Scanning electron microscopy (SEM) images of pristine and H2- treated (optimized with 8 mmol NaBH4) hematite nanostructures are shown in Fig. 1. The SEM images of hematite samples before and aer the H2 treatment are very similar which show a thin lm of vertical nanorods covered on the FTO substrate. The H2 treatment did not change the morphology of the as-prepared hematite sample. The thickness of the H2- treated hematite lm is about 320 nm measured in the SEM cross-sectional image and the result is shown in ESI, Fig. S2.† The dark eld image of transmission electron microscopy (TEM) and the corresponding elemental mappings for H2- treated hematite nanostructures are shown in Fig. 2, which clearly indicate the uniform distribution of Fe and O in the hematite. No obvious carbon coating on the hematite nanorods can be observed from the high resolution TEM images. XRD data for pristine and H2-treated hematite nanostructures with various amounts of NaBH4 are shown in ESI, Fig. S3,† which can be indexed to the characteristic peaks of the typical hematite structure (JCPDS 33-0664) aer subtracting the SnO2 peaks from the FTO substrate.

Fig. 1 SEM images of (a) pristine and (b) H2-treated (8 mmol NaBH4) hematite nanostructures.

Photoelectrochemical (PEC) measurements were performed in a 1 M NaOH electrolyte (pH about 13.6) using a three-electrode electrochemical cell with hematite nanostructures on FTO as the working electrode, a platinum coil as the counter electrode, and a reference of Ag/AgCl. The photocurrent density– applied potential (J–V) scans for hematite nanostructures were measured with the AM 1.5G simulated solar light at 100 mW cm ² . In Fig. 3 we show the J–V scans for hematite nanostructures before (black) and aer (green) the H2 treatment (8 mmol NaBH4). For comparison, the J–V scans for hematite nanostructures with oxygen vacancies (red) treated in an oxygen-decient atmosphere during the sintering process (sintered in a partial oxygen pressure of 2.1  10 2 Torr with the optimized performance) are also shown in Fig. 3. Obviously, the curve for pristine hematite before the H2 treatment shows a relatively low photocurrent density of 0.88 mA cm ² at 1.23 V vs. RHE, while for the H2-treated sample, the photocurrent density at 1.23 V vs. RHE drastically increases to 2.28 mA cm ² , which is more than 2.5 times higher than that for pristine hematite. The enhanced performance suggests that the effffective H2 treatment by our facile method may introduce oxygen vacancies and then improve the photocurrent. The performance of hematite nanostructures with typical oxygen vacancies but treated in an oxygen-decient atmosphere shows an enhanced performance

with a photocurrent of 1.75 mA cm ² at 1.23 V vs. RHE than that for the pristine sample, which is in good agreement with the literature.10 However, the photocurrent for oxygen-deciency treated hematite is much lower than that for the H2-treated sample and a signicant increase of the onset potential can be observed as well. The oxygen-deciency treated hematite sample exhibits a water oxidation onset potential at about 0.99 V vs. RHE (here we use the potential at the intersection point of dark current and the tangent at the maximum slope of photocurrent),12,13 while the H2-treated sample shows an onset potential of 0.87 V vs. RHE. The onset potential shied cathodically by up to 120 mV and the photocurrent increases at potentials more negative than 1.4 V vs. RHE. Although both methods, the oxygen-deciency treatment in the sintering process and the post-H2 treatment, may produce oxygen vacancies in hematite and then improve the performance for solar water oxidation, the detailed processes are different and a signicant difference of the onset potential can be obtained. For the oxygen-deciency treated hematite sample, the oxygen vacancies can be generated in the whole sample involving both the surface and the bulk at an oxygen-decient atmosphere. However, for a controlled H2 treatment in our experiment, only the surface of hematite lm could be affected by hydrogen and the oxygen vacancies exist mainly on the surface. The surface effect of H2 treatment should be responsible for the cathodic shi of the onset potential. It is well known that oxygen vacancies can signicantly improve the photocurrent by enhanced conductivity. However, the reported photocurrent curves showed a high onset potential (about 1.0 V vs. RHE) when introducing oxygen vacancies by the oxygen-deciency treatment.10,11,17 It was also argued that the presence of surface defect states of Fe2+ in hematite would result in a recombination of photoexcited holes with electrons,14 which could be a possible reason for the high onset potential of hematite with oxygen vacancies. However, our results suggest that the oxygen vacancies on the surface of hematite are not the reason for high onset potential. In contrast, the optimized oxygen vacancy content in the surface by the H2 treatment will improve the conductivity of hematite and then electrons can be rapidly transferred, which may reduce the recombination of photoexcited holes with electrons. Thus a low onset potential and a high photocurrent can be observed. The oxygen vacancies in the bulk may act as the traps of electrons and then result in the high onset potential in the literature. Another possible reason for the improvement of the onset potential could be the surface reaction of H2. Recently a surface corrosion by HCl was reported to be an effective way for the cathodic shi of the onset potential of hematite by suppressing the back reaction.13 Here the H2 treatment is also a surface treatment which may result in similar surface corrosion when taking oxygen atoms away from the hematite. The H2-treated hematite nanostructures show a high photocurrent and a low onset potential which could be used as a good starting material for further modications to achieve excellent performance for practical applications. The effect of H2 concentration on the performance of hematite has been investigated via adjusting the amounts of NaBH4 in the crucible. J–V scans for H2-treated hematite nanostructures with various amounts of NaBH4 (2, 5, 8, 10 mmol) are shown in Fig. 4a. Also J–V scans for pristine hematite nanostructures are shown for comparison. For H2-treated hematite samples the photocurrent increases with the amount of NaBH4 (the H2 concentration), which means more surface oxygen vacancies lead to better performance. However, the best performance shown at 8 mmol NaBH4 and 10 mmol NaBH4 with more H2 may produce magnetite which is photo-inactive leading to worse performance.10 The incident photon-to-current conversion efficiency (IPCE) data of hematite nanostructures before (black) and aer (green) the H2 treatment (8 mmol NaBH4) are measured at 1.23 V vs. RHE as a function of the incident light wavelength as shown in Fig. 4b. The H2-treated sample shows enhanced IPCE values compared to the pristine sample at all measured wavelengths, which are consistent with the J–V curves. The highest IPCE measured for the H2-treated sample is about 53% (at a wavelength of 370 nm). Based on the IPCE data, the integrated photocurrent of the H2-treated sample (the integrated curves can be found in ESI, Fig. S4†) is also calculated which reaches 2.03 mA cm ² at 1.23 V vs. RHE. The integrated photocurrent is in accord with the PEC measurement. H2 treatment conditions were optimized to obtain the best performance. The sintering temperatures from 450 to 550  C were tested and a favorable temperature of 500 C with best performance has been achieved. The suitable sintering time of 30 min was also obtained. The performance stability tests of pristine and H2-treated samples were carried out and the curves of photocurrent vs. time at 1.23 V vs. RHE are shown in ESI, Fig. S5.† The stable photocurrents in a long run suggest that it is a true effffect of water oxidation ruling out the sacrice of the hematite nanostructures themselves. Mott–Schottky curves of pristine and H2-treated (8 mmol NaBH4) hematitesamples are plotted in Fig. 5. The slopes from

the Mott–Schottky plots are used to estimate the carrier densities using the following equation: Nd ¼ (2/e0330)[d(1/C2 )/dV] 1 where e0 is the electron charge (1.602  10 19 C), 3 is the dielectric constant of a-Fe2O3 (80),16 30 is the permittivity of vacuum (8.854  10 14 Fm 1 ), Nd is the donor density and V is the potential applied at the electrode. Capacitances were derived from the electrochemical impedance obtained at each potential with 10 000 Hz frequency in the dark. The Mott– Schottky slopes of pristine and H2-treated hematite samples are positive which indicates that they are n-type semiconductors with electrons as majority carriers. The carrier densities are estimated from the Mott–Schottky plots.16 The results show that the donor density of the H2-treated hematite sample is 1.88  1020 cm 3 , which is more than one order higher than that of the pristine hematite sample (7.47  1018 cm 3 ). The increased photocurrent for H2-treated samples can be attributed to the increased donor density with oxygen vacancies on the surface.10,17 The increased donor density

could improve the conductivity and then the surface electrons can be rapidly transferred, which may reduce the recombination of photoexcited holes with electrons and favor for the low onset potential. In order to reveal the mechanism of H2 treatment, X-ray photoelectron spectroscopy (XPS) was employed to elucidate the electronic structure and surface state of H2-treated hematite nanostructures. In Fig. 6 Fe and O signals are strong for both pristine and H2-treated hematite samples. Additional C and Sn signals can also be observed for both samples. The C signal can be attributed to the adsorbed carbon contaminants in the sample preparation process, while the Sn signal stands for Sndoping originated from the Sn diffusion from the FTO substrate.6 Sn-doping in hematite may help for the performance.6 The Fe 2p XPS spectrum of H2-treated hematite nanostructures was also recorded and a small peak for Fe2+ at about 716 eV was observed for the sample standing for oxygen

vacancies (ESI, Fig. S6†).10 No B or Na signals can be observed for the H2-treated sample ruling out the effect of B or Na doping from the pyrolysis of NaBH4. To further investigate the electronic structure of H2-treated hematite nanostructures, XAS experiments were performed. XAS is an element-specic spectroscopic technique involving the excitation of electrons from a core level to an empty orbital, and is particularly powerful to obtain structural information of different species. In the near edge region (from just below the edge to  50 eV above the threshold) it is oen referred to as XANES which probes densities of states and local symmetry to be distinguished from EXAFS in the extended region (50 eV up to as much as 1000 eV above the threshold) which contains information about the inter-atomic distance and the local dynamics of the system. XAS spectra of pristine and H2-treated hematite samples at the O K-edge were recorded on the so X-ray beamline at the Beijing Synchrotron Radiation Facility and are shown in Fig. 7a. XANES and EXAFS data at the Fe K-edge of hematite samples have been collected on beamline 14W at the Shanghai Synchrotron Radiation Facility. The Fourier transform of the EXAFS data in R space is shown in Fig. 7b.

In Fig. 7a the XAS spectrum of pristine hematite shows typical hematite features such as two separated prepeaks A1 and A2 and the main peak B. The prepeak can be attributed to transitions to anti-bonding O 2p states hybridized with the 3d metal states (t2g and eg orbital symmetry), mainly localized at the Fe site.21,22 Feature B can be generally attributed to oxygen 2p states hybridized with iron 4s and 4p states.21 Compared to the spectrum of pristine hematite, the O K-edge spectrum of H2-treated hematite with oxygen vacancies shows signicant differences such as the decrease of the prepeaks (A1 and A2) and the increase of the shoulder A3. According to the literature, a low oxidation state of Fe such as Fe0 or Fe2+ may lead to the decreased intensity of the prepeaks compared to that of the main peak B.21 The spectral differences between pristine and H2-treated hematite samples can be assigned to the contribution of FeO with less oxygen content.21 Peak A3 could also be assigned to the appearance of Fe2+ in hematite or the carbon contamination in the sample preparation process.23 The O K-edge XAS spectra are recorded in the so Xray range with the TEY detection mode, which is a surfacesensitive method. The XAS results clearly indicate that aer the H2 treatment Fe2+ exists on the surface of hematite with oxygen vacancies. Compared to the surface-sensitive XAS measurement at the O K-edge, the XAS measurement at the Fe K-edge is in the hard X-ray range detecting the uorescence signal, which is a bulksensitive method. Thus the XANES and EXAFS data at the Fe Kedge mainly re ect the bulk information of hematite samples. The Fourier transform of the EXAFS data in R space of different hematite samples is shown in Fig. 7b for comparison. In Fig. 7b the peak around 1.5 ˚ A can be attributed to Fe–O bonds while the peak around 3 ˚ A to Fe–Fe bonds.24,25 Interestingly, the curve for H2-treated hematite shows almost identical features compared to that for pristine hematite, which suggests that aer the H2 treatment the bulk of hematite is not affected by oxygen vacancies. For a controlled H2 treatment in our experiment, only the surface of hematite was affected by hydrogen and the oxygen vacancies exist mainly on the surface. The surface-sensitive so X-ray absorption spectroscopy conrms the electronic structure change of the surface of H2-treated hematite, while the bulk sensitive hard X-ray spectroscopy revealed that aer the H2 treatment the bulk hematite remains in its own state. The Fourier transform of the EXAFS data in R space of the oxygendeciency treated hematite sample is also shown in Fig. 7b (red curve). Compared to H2-treated hematite, the oxygen-deciency treated hematite sample has oxygen vacancies in both the surface and the bulk. An obvious decrease of the intensity of Fe– O bonds compared to that for H2-treated hematite (green curve) can be observed, indicating the existence of oxygen vacancies in the bulk. The XAS results clearly conrm that the oxygen vacancies only exist on the surface of the H2-treated hematite samples, suggesting that the onset potential shi between H2- treated hematite and oxygen-deciency treated hematite should be attributed to the surface effffect by the H2 treatment. Our facile way for the H2 treatment stands for a good strategy for the improvement of both the photocurrent and the onset potential for solar water oxidation.

Conclusions 

We present the preparation of H2-treated hematite nanostructures by a simple pyrolysis of NaBH4 in a crucible. The H2- treated hematite photoelectrode shows a high photocurrent of 2.28 mA cm² at 1.23 V vs. RHE, which is over 2.5 times higher than that for pristine hematite. Moreover, when compared to the hematite photoelectrode with typical oxygen vacancies treated in the oxygen-decient atmosphere during the sintering process, a cathodic shi of the onset potential has been achieved by about 120 mV (from 0.99 to 0.87 V vs. RHE). The cathodic shi of the onset potential is attributed to the surface effect of H2 treatment. Our results suggest that the presence of surface defect states of Fe2+ in hematite is not the reason for high onset potential. The H2-treated hematite with high effi- ciency can be used as a good starting material to achieve better performance for practical applications. 

Acknowledgements 

We thank Q. Sun for XPS experiments. We thank C. Hong for the support during runs at the BSRF. We thank S. Zhang, Z. Jiang and Y. Huang for the support during runs at the SSRF. We acknowledge the National Basic Research Development Program of China (2010CB934500, 2012CB825800), the National Natural Science Foundation of China (11275137, 11179032, 91333112) and the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD). This is also a project supported by the Fund for Innovative Research Teams of Jiangsu Higher Education Institutions. 

Notes and references 

1 A. Fujishima and K. Honda, Nature, 1972, 238, 37.

2 M. Gr¨atzel, Nature, 2001, 414, 338.

3 K. Sivula, F. L. Formal and M. Gr¨atzel, ChemSusChem, 2011,4, 432.

4 A. B. Murphy, P. R. F. Barnes, L. K. Randeniya, I. C. Plumb, I. E. Grey, M. D. Horne and J. A. Glasscock, Int. J. Hydrogen Energy, 2006, 31, 1999. 

5 J. Kim, G. Magesh, D. Youn, J. Jang, J. Kubata, K. Domen and J. Lee, Sci. Rep., 2013, 3, 2681. 

6 Y. Ling, G. Wang, D. A. Wheeler, J. Z. Zhang and Y. Li, Nano Lett., 2011, 11, 2119. 

7 J. J. Deng, X. X. Lv, J. Gao, A. W. Pu, M. Li, X. H. Sun and J. Zhong, Energy Environ. Sci., 2013, 6, 1965. 

8 J. J. Deng, J. Zhong, A. W. Pu, D. Zhang, M. Li, X. H. Sun and S.-T. Lee, J. Appl. Phys., 2012, 111, 084312. 

9 S. D. Tilley, M. Cornuz, K. Sivula and M. Gr¨atzel, Angew. Chem., Int. Ed., 2010, 49, 6405. 

10 Y. C. Ling, G. M. Wang, J. Reddy, C. C. Wang, J. Z. Zhang and Y. Li, Angew. Chem., Int. Ed., 2012, 51, 4074. 

11 T. Y. Yang, H. Y. Kang, Y. J. Lee, J. H. Lee, B. J. Koo, K. T. Nam and Y. C. Joo, Phys. Chem. Chem. Phys., 2013, 15, 2117. 

12 D. Zhong, M. Cornuz, K. Sivula, M. Gr¨atzel and D. Gamelin, Energy Environ. Sci., 2011, 4, 1759. 

13 D. Cao, W. Luo, J. Feng, Y. Zhao, Z. Li and Z. Zou, Energy Environ. Sci., 2014, 7, 752. 

14 T. Hisatomi, F. Formal, M. Cornuz, J. Brillet, N. T´etreault, K. Sivula and M. Gr¨atzel, Energy Environ. Sci., 2011, 4, 2512. 

15 F. Formal, N. T´etreault, M. Cornuz, T. Moehl, M. Gr¨atzel and K. Sivula, Chem. Sci., 2011, 2, 737. 

16 I. Cesar, K. Sivula, A. Kay, R. Zboril and M. Gr¨atzel, J. Phys. Chem. C, 2009, 113, 772. 

17 A. Pu, J. Deng, M. Li, J. Gao, H. Zhang, Y. Hao, J. Zhong and X. Sun, J. Mater. Chem. A, 2014, 2, 2491. 

18 G. Wang, H. Wang, Y. Ling, Y. Tang, X. Yang, R. Fitzmorris, C. Wang, J. Zhang and Y. Li, Nano Lett., 2011, 11, 3026. 

19 G. Wang, Y. Ling, H. Wang, X. Yang, C. Wang, J. Zhang and Y. Li, Energy Environ. Sci., 2012, 5, 6180. 

20 X. Lu, G. Wang, S. Xie, J. Shi, W. Li, Y. Tong and Y. Li, Chem. Commun., 2012, 48, 7717. 

21 Z. Y. Wu, S. Gota, F. Jollet, M. Polllak, M. G. Soyer and C. R. Natoli, Phys. Rev. B: Condens. Matter Mater. Phys., 1997, 55, 2570. 

22 A. Braun, K. Sivula, D. K. Bora, J. F. Zhu, L. Zhang, M. Gr¨atzel, J. H. Guo and E. C. Constable, J. Phys. Chem. C, 2007, 116, 16870. 

23 A. Braun, D. Bayraktar, S. Erat, A. S. Harvey, D. Bechel, J. A. Purton, P. Holtappels, L. J. Gauckler and T. Graule, Appl. Phys. Lett., 2009, 94, 202102. 

24 I. Arcon, M. Mozetic and A. Kodre, Vacuum, 2005, 80, 178. 

25 F. Jiao, A. Harrison, J. C. Jumas, A. V. Chadwick, W. Kockelmann and P. G. Bruce, J. Am. Chem. Soc., 2006, 128, 5468.