Cerium-Doped Iron Oxide Nanorod Arrays for Photoelectrochemical Water Splitting

In this work, a simple one-step hydrothermal method was employed to prepare the Ce-doped Fe2O3 ordered nanorod arrays (CFT). The Ce doping successfully narrowed the band gap of Fe2O3, which improved the visible light absorption performance. In addition, with the help of Ce doping, the recombination of electron/hole pairs was significantly inhibited. The external voltage will make the performance of the Ce-doped sample better. Therefore, the Ce-doped Fe2O3 has reached superior photoelectrochemical (PEC) performance with a high photocurrent density of 1.47 mA/cm2 at 1.6 V vs. RHE (Reversible Hydrogen Electrode), which is 7.3 times higher than that of pristine Fe2O3 nanorod arrays (FT). The Hydrogen (H2) production from PEC water splitting of Fe2O3 was highly improved by Ce doping to achieve an evolution rate of 21 μmol/cm2/h.


Introduction
Renewable energy for the sustainable development of the whole society is gaining momentum globally [1]. Hydrogen energy as a clean energy carrier is a non-polluting new energy source with high energy density [2]. The common approaches of electrolysis of water and steam reforming methane for H 2 production either consume large amounts of electricity or fossil energy [3,4]. The photoelectrochemical (PEC) water splitting using solar light has been considered the most promising and environmentally friendly technique for H 2 production [5][6][7][8][9][10]. In the PEC system, photoanodes absorb sunlight and split water directly into H 2 and O 2 . Thus, the efficiency and stability of H 2 production from PEC water splitting highly depend on the photoanodes materials [11][12][13][14].
Metal oxide semiconductors (MOS) have long been developed as photo-active materials for PEC water splitting owing to multiple advantages of superior stability, non-toxicity, and low cost [15,16]. Among various MOS, hematite (α-Fe 2 O 3 ) as photoanodes has outperformed virtues of abundance, chemical stability, and suitable electronic structure. The α-Fe 2 O 3 has a small band gap (2.2 eV) with the absorption of approximately 40% of the incident sunlight [17], which gives it a high theoretical solar-to-H 2 (STH) efficiency. However, hematite has some fatal disadvantages that affect its photocatalytic performance; for instance, low electrical conductivity, high electron-hole complexation rate [12]; hole diffusion length is only 2-4 nm, which is relatively small compared to TiO 2 (100 µm) and WO 3 (100-150 nm) [18], low electron mobility (10 −1 V −1 s −1 ) [19], and short lifetime of the excited state [20]. These have seriously limited the application of hematite for H 2 production from PEC water splitting.
Herein, the Ce-doped Fe 2 O 3 nanorod arrays were prepared by a simple one-step hydrothermal method. The morphology and structure of Fe 2 O 3 (FT) and Ce-doped Fe 2 O 3 (CFT) samples were observed by X-ray diffraction (XRD), Scanning Electron Microscope (SEM), and Transmission Electron Microscopy (TEM). The PEC performances of Fe 2 O 3 and Ce-doped Fe 2 O 3 were comprehensively investigated. The optical properties and photocatalytic activity of Ce-doped Fe 2 O 3 were remarkably improved. The Ce doping effectively promotes the charge transfer and separation of photogenerated electron/hole pairs in Fe 2 O 3 . The applied voltage also plays an important role in exerting the function of Ce doping in Fe 2 O 3 . Therefore, the Ce-doped samples realized high photocurrent density and incident photon-to-current efficiency (IPCE) values. The PEC performance for H 2 production of Fe 2 O 3 was highly improved by Ce doping.

Structure and Morphology
XRD results confirm the hematite phase of as-prepared samples, FT, 1CFT, 5CFT, and 10CFT, as shown in Figure 1 SEM observed the morphology of the samples FT, 1CFT, 5CFT, and 10CFT, as shown in Figure 2a-d. The homogenous rods with an average length of 350 nm are perpendicular to the substrate, indicating that the morphology of nanorods was not affected by Ce doping. The Fe 2 O 3 nanorods arrays consisted of 5-30 nm nanopores with mesoporous structures, respectively. The TEM measurements were also conducted to study the microstructure of Fe 2 O 3 nanorods further. The hollow pore structures of nanorods in samples FT and 5CFT are directly observed, as shown in Figure 2e,g, respectively. The HRTEM of FT and 5CFT in Figure 2f,h show the lattice stripes with a specular spacing of 0.268 nm and 0.274 nm for the (104) crystal plane of Fe 2 O 3 , respectively. The Ce doping has led to a larger lattice spacing of the (104) crystal plane in Fe 2 O 3 , which is consistent with the XRD results. The HAADF image in Figure 2i and EDS mapping results in Figure 2j  SEM observed the morphology of the samples FT, 1CFT, 5CFT, and 10CFT, as shown in Figure 2a-d. The homogenous rods with an average length of 350 nm are perpendicular to the substrate, indicating that the morphology of nanorods was not affected by Ce doping. The Fe2O3 nanorods arrays consisted of 5-30 nm nanopores with mesoporous structures, respectively. The TEM measurements were also conducted to study the microstructure of Fe2O3 nanorods further. The hollow pore structures of nanorods in samples FT and 5CFT are directly observed, as shown in Figure 2e,g, respectively. The HRTEM of FT and 5CFT in Figure 2f,h show the lattice stripes with a specular spacing of 0.268 nm and 0.274 nm for the (104) crystal plane of Fe2O3, respectively. The Ce doping has led to a larger lattice spacing of the (104) crystal plane in Fe2O3, which is consistent with the XRD results. The HAADF image in Figure 2i and EDS mapping results in Figure 2j-m confirm the homogeneous distribution of Ce elements in Fe2O3 nanorods in the samples of 5CFT. Therefore, the Ce ions are successfully doped in Fe2O3 nanorods without affecting the morphology of Fe2O3 nanorods.

Chemical Composition and Surface Information
For obtaining the chemical composition and surface information of Fe2O3 and Cedoped Fe2O3, XPS measurements were conducted. Except for the pristine FT, the Cedoped samples of 1CFT, 5CFT, and 10CFT consist of C, O, Fe, and Ce, as shown in Figure

Chemical Composition and Surface Information
For obtaining the chemical composition and surface information of Fe 2 O 3 and Cedoped Fe 2 O 3 , XPS measurements were conducted. Except for the pristine FT, the Ce-doped samples of 1CFT, 5CFT, and 10CFT consist of C, O, Fe, and Ce, as shown in Figure 3a. As the Ce content increases, the position of the valence band shifts to the lower energy, as shown in Figure S1a. The valence band position of 5CFT decreases vastly to 1.29 eV from the 1.65 eV in pristine FT. In the spectrum of Fe 2p, there are two peaks at 711.2 eV and 724.5 eV corresponding to Fe 2p 3/2 and Fe 2p 1/2 , respectively, as shown in Figure 3b. In addition, there is a satellite peak at 718 eV, which is the signature of Fe (III) [37]. The Fe 2p peaks in the spectra of the Ce-doped samples shifted significantly towards lower binding energy, which is caused by the interaction between the Fe 3+ and Ce 3+ or Ce 4+ substances [38]. In the spectrum of O 1s, the central peak at around 530 eV in FT corresponds to the lattice oxygen O L , as shown in Figure 3d. The shifting of O L towards lower binding energy after Ce doping is due to the weak electronegativity of Ce ions.  In the Ce 3d spectra, the distinctive peaks of elemental Ce are seen at 889.1 eV, 894.2 eV, 898.5 eV, 903.2 eV, 907.1 eV, 916.5 eV in Ce doped samples as shown in Figure 3c. The peaks at 889.1 eV, 898.5 eV, 907.1 eV, and 916.5 eV originated from the Ce 4+ [39]. Precisely, the former two peaks correspond to the Ce 3d5/2 orbital and the latter to the Ce 3d3/2 orbital. The signal at 894.5 eV originates from the spin-orbit coupling for the Ce 3d5/2 and Ce 3d3/2 states [40]. The spectra of Ti 2p show a standard bimodal curve for titanium dioxide, as shown in Figure S1b. Therefore, the Ce ions are successfully doped in Fe2O3, and the valance band position of Fe2O3 was reduced significantly. In the Ce 3d spectra, the distinctive peaks of elemental Ce are seen at 889.1 eV, 894.2 eV, 898.5 eV, 903.2 eV, 907.1 eV, 916.5 eV in Ce doped samples as shown in Figure 3c. The peaks at 889.1 eV, 898.5 eV, 907.1 eV, and 916.5 eV originated from the Ce 4+ [39]. Precisely, the former two peaks correspond to the Ce 3d 5/2 orbital and the latter to the Ce 3d 3/2 orbital. The signal at 894.5 eV originates from the spin-orbit coupling for the Ce 3d 5/2 and Ce 3d 3/2 states [40]. The spectra of Ti 2p show a standard bimodal curve for titanium dioxide, as shown in Figure S1b. Therefore, the Ce ions are successfully doped in Fe 2 O 3, and the valance band position of Fe 2 O 3 was reduced significantly.

Photoelectrochemical Performance
The UV-Vis absorption spectra also investigated the optical properties of Fe 2 O 3 photoanodes. The Ce doping specifically improved the light absorption performance of the Fe 2 O 3 photoanode in the region of 800 to 620 nm, as shown in Figure S2a. The band gaps of pristine FT and Ce doped samples, 1CFT, 5CFT, and 10CFT, are calculated to be 1.530, 1.478, 1.441, and 1.453 eV, respectively, as shown in Figure S2b. The sample of 5CFT has the smallest band gap, and the FT sample has the largest band gap. Therefore, the Ce doping vastly increases the light absorption due to the narrowed band gap of Fe 2 O 3 .
A series of photoelectric tests further evaluated the photochemical properties of the Ce-doped Fe 2 O 3 nanorod arrays. Obviously, the Ce-doped Fe 2 O 3 samples, which are 1CFT, 5CFT, and 10CFT, all exhibited much higher photocurrent density than pristine FT, as shown in Figure 4a. Significantly, the photocurrent density of 5CFT (1.47 mA/cm 2 ) at 1.6 V vs. RHE is 7.3 times higher than that of pristine FT (0.2 mA/cm 2 ). The photocurrent density of 5CFT (0.95 mA/cm 2 ) at 1.12 V vs. RHE is 6.3 times higher than that of pristine FT (0.15 mA/cm 2 ). By the Ce doping, the PEC performance of Fe 2 O 3 is highly improved. The 5CFT achieves the maximum ABPE of 1.02% at 1.39 V vs. RHE, which is 7.3 times as that of FT (0.14%), as shown in Figure 4b.
The photocurrent densities of pristine Fe 2 O 3 and Ce-doped Fe 2 O 3 under monochromatic light with different applied bias pressures were measured. As the applied bias increases, the photocurrent densities of these samples increase, as shown in Figure 4c,d and Figure S3. At low bias (1V vs. RHE), the photocurrent response in visible light was faintly negligible due to the high rate of recombination of electron/hole pairs. The high applied bias helps the charge separation of shown in Figure 4a. Significantly, the photocurrent density of 5CFT (1.47 mA/cm ) at 1.6 V vs. RHE is 7.3 times higher than that of pristine FT (0.2 mA/cm 2 ). The photocurrent density of 5CFT (0.95 mA/cm 2 ) at 1.12 V vs. RHE is 6.3 times higher than that of pristine FT (0.15 mA/cm 2 ). By the Ce doping, the PEC performance of Fe2O3 is highly improved. The 5CFT achieves the maximum ABPE of 1.02% at 1.39 V vs. RHE, which is 7.3 times as that of FT (0.14%), as shown in Figure 4b.  The Electrochemical impedance spectroscopy (EIS) results are shown in Figure S4, where the reduction in semicircular response with frequency change can be seen. The inset shows a modified Randles circuit [41] for the EIS. The fitting parameters for each fitted component are given in Table S3. The EIS curves of all samples show a standard semicircle, and the resistance values of the Ce-doped samples are all significantly smaller than the undoped FT, with the 5CFT sample being the smallest, demonstrating that the Ce doping enhances the electron transfer efficiency of Fe 2 O 3 and reduces the resistance values of the samples. In addition, all the curves in the Nyquist plot pass through the origin of the coordinate axis, indicating that the resistance of the solution can be neglected, and the value of Rs is much smaller than Rp, as can be seen from the fitted circuit values.
The Mott-Schottky measurements were conducted to determine the flat band potentials of these samples, as shown in Figure 4g. Using the Mott-Schottky equation, the flat band potentials can be fitted, which are 0.01 V, 0.42 V, 0.58 V, and 0.49 V vs. RHE of FT, 1CFT, 5CFT, and 10CFT, respectively. For n-type semiconductors, the flat-band potential is positive 0.1-0.3 eV compared to the conduction band potential; for p-type semiconductors, the flat-band potential is negative 0.1-0.3 eV compared to the valence band potential. The flat band potential is infinitely near the bottom of the conduction band (CB) of the semiconductor materials, which can indirectly indicate the valence band position of the material. Although a slight positive shift shifts the CB of the Ce-doped sample compared to the FT, it still crosses the reduction potential invited and retains the ability to produce hydrogen, and the Ce-doped samples have much higher carrier concentration than pristine FT, implying high charge transfer efficiency in Ce-doped Fe 2 O 3, as illustrated in Figure S5. Therefore, the Ce ions doping highly promote the photocatalytic performance of Fe 2 O 3 by facilitating the charge transfer and inhibiting the recombination of electron/hole pairs.
The H 2 production performance of Fe 2 O 3 samples was also explored under simulated solar light, as shown in Figure 4h. The H 2 production of all samples increased cumulatively over time. The Ce-doped Fe 2 O 3 showed significantly higher H 2 production than pristine Fe 2 O 3 . After 6 h of the PEC reaction under 4.4 times AM1.5G illumination, the watersplitting H 2 production of 5CFT achieved 126 µmol/cm 2 , which was 32 times higher than that of FT (which was 3.96 µmol/cm 2 ).

Photoelectrochemical Mechanism for Water Splitting under Solar Light
Based on the above, the PEC mechanism of H 2 production from water splitting was proposed, as shown in Figure 5. Without the application of external bias (left part of Figure 5a), the photogenerated electrons generated by the Fe 2 O 3 require considerable energy to cross the potential barrier and transfer to the electrode surface. Under the applied bias, see the right part of Figure 5a, the Fermi energy level of the Fe 2 O 3 becomes smaller, and a new Fermi energy level is formed with the TiO 2 . The conduction and valence band edges are elevated. Thus, the conduction band position of the Fe 2 O 3 is rendered more negatively. The charge transfer of photogenerated electrons from Fe 2 O 3 to TiO 2 can be facilitated. Therefore, the electron and hole complexation rates are significantly reduced with the help of applied bias, and the photocatalytic performance of Fe 2 O 3 samples was dramatically improved, as shown in Figure 5a. The band structures of pristine Fe 2 O 3 and Ce-doped Fe 2 O 3 nanorod arrays are illustrated in Figure 5b. A smaller band gap reduces the energy barrier that must be crossed to generate photogenerated carriers, resulting in more photogenerated carriers entering the outer circuit to participate in the redox reaction, thus enhancing the photoelectrochemical performance of the sample. Therefore, the electron and hole complexation rates are significantly reduced with the help of applied bias, and the photocatalytic performance of Fe2O3 samples was dramatically improved, as shown in Figure 5a. The band structures of pristine Fe2O3 and Ce-doped Fe2O3 nanorod arrays are illustrated in Figure 5b. A smaller band gap reduces the energy barrier that must be crossed to generate photogenerated carriers, resulting in more photogenerated carriers entering the outer circuit to participate in the redox reaction, thus enhancing the photoelectrochemical performance of the sample.

Materials and Synthesis Methods
Preparation of Fe 2 O 3 nanorod arrays: The Fe 2 O 3 nanorod arrays were prepared using a previously reported hydrothermal method [42]. The cut titanium pieces (resistance value 1-3 Ω, thickness 0.3 mm, purity 99.99%) are sonicated by acetone (C 3 H 6 O, AR ≥ 99.5%, Sinopharm Chemical Reagent Co., Ltd., Shanghai, China), deionized water (prepared by a laboratory ultrapure water machine, GZY-P10-W, Kertone Water Treatmengt Co., Ltd., Changsha, China), and ethanol absolute (C 2 H 6 O, AR ≥ 99.7%, Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) for 30 min and subsequently dried under a nitrogen stream. A total of 4 mL mixture of 1 M NaNO 3 (AR ≥ 99.0%, Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) and 0.15 M FeCl 3 ·6H 2 O (AR ≥ 99.0%, General-Reagent, Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) was stirred for 2 h, the pH was adjusted to 1.5 by HCl (37%, Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) and transfer to a Teflon vessel, meanwhile the treated titanium pieces were put together in. After holding at 100 • C for 5 h, a yellowish FeOOH film was formed on the surface of the titanium sheet and rinsed using deionized water and annealed at 550 • C for 2 h in a muffle furnace with a ramp rate of 2 • C/min, the film turns reddish brown, and the iron oxide film is successfully prepared.
Preparation of Ce-doped Fe 2 O 3 nanorod arrays: By adding 0.01 M Ce(NO 3 ) 3 (AR ≥ 99.99%, Damas-beta) to the above Teflon vessel and sonicating for 5 min, and otherwise the same as above, we have produced an array of Ce doped iron oxide nanorods of various contents. According to the different feeding ratios, the corresponding samples are named: FT (0 µL/4 mL), 1CFT (6 µL/4 mL), 5CFT (30 µL/4 mL), and 10CFT (60 µL/4 mL). A schematic diagram of the hematite photoanodes preparation procedure is shown in Figure 6. Inductively Coupled Plasma Emission Spectrometry (ICP) and Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) tests were conducted to determine the Ce content in 1CFT, 5CFT, and 10CFT as shown in Table S2. adjusted to 1.5 by HCl (37%, Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) and transfer to a Teflon vessel, meanwhile the treated titanium pieces were put together in. After holding at 100 °C for 5 h, a yellowish FeOOH film was formed on the surface of the titanium sheet and rinsed using deionized water and annealed at 550 °C for 2 h in a muffle furnace with a ramp rate of 2 °C/min, the film turns reddish brown, and the iron oxide film is successfully prepared.
Preparation of Ce-doped Fe2O3 nanorod arrays: By adding 0.01 M Ce(NO3)3 (AR ≥ 99.99%, Damas-beta) to the above Teflon vessel and sonicating for 5 min, and otherwise the same as above, we have produced an array of Ce doped iron oxide nanorods of various contents. According to the different feeding ratios, the corresponding samples are named: FT (0 μL/4 mL), 1CFT (6 μL/4 mL), 5CFT (30 μL/4 mL), and 10CFT (60 μL/4 mL). A schematic diagram of the hematite photoanodes preparation procedure is shown in Figure 6. Inductively Coupled Plasma Emission Spectrometry (ICP) and Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) tests were conducted to determine the Ce content in 1CFT, 5CFT, and 10CFT as shown in Table S2.

Photoelectrochemical (PEC) Testing
The Electrochemical workstation (CHI760e), a high uniformity integrated Xenon light source (PLS-FX300HU, Beijing Perfectlight, Beijing, China) with an AM 1.5G filter (100 mW/cm 2 ) and three-electrode side window electrolytic cell forming the PEC reaction system for testing linear sweep voltammetry (LSV) and chronoamperometry (I-t). An LSV measurement is typically carried out over a potential range of −0.6 to 0.6 V (vs. Ag/AgCl), and the scanning speed is 5 mV s −1 with the light cut off by a 5 s −1 shutter. Chronoamperometry (I-t) was performed under alternating illumination at 0.8 V or 0.6 V (vs. Ag/AgCl). Three-electrode electrolytic cells using a platinum sheet as a counter electrode,

Photoelectrochemical (PEC) Testing
The Electrochemical workstation (CHI760e), a high uniformity integrated Xenon light source (PLS-FX300HU, Beijing Perfectlight, Beijing, China) with an AM 1.5G filter (100 mW/cm 2 ) and three-electrode side window electrolytic cell forming the PEC reaction system for testing linear sweep voltammetry (LSV) and chronoamperometry (I-t). An LSV measurement is typically carried out over a potential range of −0.6 to 0.6 V (vs. Ag/AgCl), and the scanning speed is 5 mV s −1 with the light cut off by a 5 s −1 shutter. Chronoamperometry (I-t) was performed under alternating illumination at 0.8 V or 0.6 V (vs. Ag/AgCl). Three-electrode electrolytic cells using a platinum sheet as a counter electrode, Ag/AgCl (saturated KCl) as a reference electrode, and 1 M KOH (pH = 13.6) as electrolyte. During the test, the working electrode is guaranteed to have an illuminated area of 1 cm 2 and is completely submerged in the electrolyte. The calculated voltage can be changed to a reversible H 2 electrode (RHE) scale by the Nernst equation: where, E 0 Ag/AgCl = 0.1976 V vs. Ag/AgCl at room temperature. The photovoltaic conversion efficiency (ABPE) with external bias can be calculated from the following equation: where J is the photocurrent density, V is the bias voltage applied between the working electrode and reference electrode, and I 0 is the intensity of the incident light (I 0 = 100 mW/cm 2 ). The incident photon-to-current efficiency (IPCE) test system is composed of a 300 W xenon lamp light source (PLS-SXE300D, Beijing Perfectlight), a grating monochromator (7ISU, Shaifan) with filters to remove higher order diffraction and an electrochemical workstation (CS350H, CorrTest), which is calculated by the following equation: 10 of 13 where J is the photocurrent density measured at a specific wavelength, λ is the specific wavelength, and I 0 is the intensity of the incident light. Electrochemical impedance spectroscopy (EIS) and Mott-Schottky (M-S) curves using the Admiral Squidstat Plus electrochemical workstation. The fitted circuit can be obtained from the EIS curve, where the CPE element is calculated. The impedance of the CPE in an AC circuit is: where σ is the prefactor of the CPE, ω is the angular frequency, m is the CPE index (0 ≤ m ≤ 1), and j is an imaginary number j = √ −1 ; if m = 1, then the CPE denotes the ideal capacitor C. The electrochemical analyzer works over a frequency range of 0.01 to 100,000 Hz, with voltage increments of 0.005 V and an AC amplitude of 10 mV. The working electrodes were tested at 500 Hz, 1000 Hz, 1500 Hz, 2000 Hz, and 2500 Hz. The fitted flat-band potentials allow the carrier concentration (N D ) [43] to be calculated, which is calculated by the following equation: where k is the slope of the fitted tangent line, e is the single electron charge, ε is the relative permittivity, ε 0 is the vacuum permittivity, and A 2 is the sample's surface area. Photoelectric coupled H 2 production tests are performed in an automated online trace gas analysis system (Labsolar-6A, Beijing Perfectlight, Beijing, China) and a gas chromatograph (F9790, Fuli Instruments, Wenling, Zhejiang, China) with an external electrochemical workstation (Squidstat Plus, Admiral Instruments, Tempe, AZ, USA) providing bias voltage. The optical power of the xenon lamp (Microsolar 300, Beijing Perfectlight Technology Co., Ltd. Beijing, China) has been tested almost equal to 4.4 times of AM 1.5G. Argon (Ar) was utilized as the carrier gas at a flow rate of 30 mL/min. Electrochemical impedance spectroscopy (EIS) and Mott-Schottky (M-S) curves were also analyzed at this workstation.

Materials Characterization
A Scanning Electron Microscope (SEM, Apreo S LoV ac, Thermo Fisher Scientific, Waltham, MA, USA) with an operating voltage of 10 kV was used to observe the morphology of the samples. X-ray diffraction (XRD) data were collected using a Miniflex 600 X-ray diffractometer, with CuK radiation and a measurement range of 20-80 • . Transmission Electron Microscopy (TEM), High-Resolution Transmission Electron Microscopy (HRTEM), and Selected Area Electron Diffraction (SAED) were carried out on a TF20 (FEI) operating at 200 kV, Waltham, MA, USA. X-ray photoelectron spectroscopy (XPS, Thermo Scientific K-Alpha, Waltham, MA, USA) was performed to characterize the atomic composition and state at the surface of the samples, using Al Ka rays as the excitation source. UV-vis absorption spectra were gathered by a Cary 5000 spectrophotometer (Agilent Technologies, Santa Clara, CA, USA). The elemental content of Fe was determined using inductively coupled plasma emission spectrometry (ICP; UI. TIMA2, HoribaJobin Yvon S.A.S), while the content of Ce, Ti was determined using Inductively coupled plasma-Mass Spectrometry (ICP-MS; AGILENT 8800, Agilent Technologies, Santa Clara, CA, USA).

Conclusions
In conclusion, the ordered nanorod arrays of Ce-doped Fe 2 O 3 were prepared by a simple one-step hydrothermal method for the photocatalytic H 2 production from H 2 O. SEM and TEM images show that Fe 2 O 3 nanorod arrays consisted of 5-30 nm nanopores with mesoporous structures. With Ce doping, the band gap structure of Fe 2 O 3 was successfully modulated and narrowed with enhanced light absorption. Moreover, both the charge transfer and separation of photogenerated electron/hole pairs in Fe 2 O 3 are highly improved by Ce doping due to their superior electrical conductivity. With the help of applied voltage, the doping effect of Ce doping can be further enhanced. Therefore, the Ce doped samples, 5CFT, realized a high photocurrent density of 1.47 mA/cm 2 at 1.6 V vs. RHE, which is 7.3 times higher than that of pristine FT (0.2 mA/cm 2 ). The IPCE of 5CFT reaches a maximum value of 21% at 357 nm with bias voltage 1.6 V vs. RHE, which is 12.4 times increase to the FT sample (1.69% IPCE in the same situation). Owing to the optimized band structure and enhanced PEC performance, the H 2 production of Fe 2 O 3 was highly improved by Ce doping to achieve an evolution rate of 21 µmol/cm 2 /h.