Paving the road toward the use of β-Fe2O3 in solar water splitting: Raman identification, phase transformation and strategies for phase stabilization

Abstract Although β-Fe2O3 has a high theoretical solar-to-hydrogen efficiency because of its narrow band gap, the study of β-Fe2O3 photoanodes for water splitting is elusive as a result of their metastable nature. Raman identification of β-Fe2O3 is theoretically and experimentally investigated in this study for the first time, thus clarifying the debate about its Raman spectrum in the literature. Phase transformation of β-Fe2O3 to α-Fe2O3 was found to potentially take place under laser and electron irradiation as well as annealing. Herein, phase transformation of β-Fe2O3 to α-Fe2O3 was inhibited by introduction of Zr doping, and β-Fe2O3 was found to withstand a higher annealing temperature without any phase transformation. The solar water splitting photocurrent of the Zr-doped β-Fe2O3 photoanode was increased by 500% compared to that of the pure β-Fe2O3 photoanode. Additionally, Zr-doped β-Fe2O3 exhibited very good stability during the process of solar water splitting. These results indicate that by improving its thermal stability, metastable β-Fe2O3 film is a promising photoanode for solar water splitting.

In recent decades, iron-based semiconductors with narrow band gaps (e.g. α-Fe 2 O 3 ) have been considered as promising photoanode materials with respect to nontoxicity, cost, stability and theoretical solar-to-hydrogen efficiency [23][24][25][26][27][28][29][30][31][32][33][34]. Recently, a β-Fe 2 O 3 semiconductor with a direct band gap of 1.9 eV was reported by the present authors as a water splitting photoanode under Air Mass 1.5 Global spectrum (AM 1.5 G, 100 mW cm -2 ) illumination [35]. Theoretically, a β-Fe 2 O 3 photoanode for solar water splitting may exhibit a solar-to-hydrogen efficiency of 20.9% because of its narrow band gap of 1.9 eV, whereas its solar photocurrent density is still very low [35]. The phase transformation of metastable β-Fe 2 O 3 to mature α-Fe 2 O 3 may occur initially at the surface or interface during fabrication of photoanode films because metastable β-Fe 2 O 3 cannot withstand the high temperature of annealing.
Phase identification of β-Fe 2 O 3 is very important to avoid the effects of impurities (e.g. α-Fe 2 O 3 ) during the process of solar water splitting. Phase characterization of inorganic materials is frequently achieved with such methods as X-ray C The Author(s) 2020. Published by Oxford University Press on behalf of China Science Publishing & Media Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

RESEARCH ARTICLE
diffraction (XRD), neutron diffraction, electron diffraction, Raman scattering spectroscopy and infrared absorption spectroscopy. Among these methods, Raman spectroscopy is fast and highly sensitive toward material surface structural information. However, Raman spectra of β-Fe 2 O 3 (including Raman peak position and shape) reported in the literature are often contradictory [36,37]. Thus, this study aims to clarify the debates surrounding the Raman spectra of β-Fe 2 O 3 .
The wavelength and power of the excitation laser should be chosen carefully during the process of Raman detection. An appropriate wavelength of laser should be selected to enhance the Raman sensibility and minimize the fluorescence emission of the samples. It is equally important that the power of the laser be examined cautiously to avoid the influences of thermal effects caused by the laser, especially for those materials with poor thermal stability. For example, phase transformation of ε-Fe 2 O 3 and γ -Fe 2 O 3 may take place under laser irradiation [38,39], and β-Fe 2 O 3 may be converted into α-Fe 2 O 3 upon heat treatment [35,40]. Therefore, low-power laser irradiation is required to avoid phase transformation of the β-Fe 2 O 3 samples.
In this study, the influence of laser wavelength and irradiation power on phase transformation of β-Fe 2 O 3 to α-Fe 2 O 3 was investigated, and the Raman vibrational spectrum of β-Fe 2 O 3 was clarified. After being doped with Zr, the particle-assembled β-Fe 2 O 3 photoanodes can withstand higher annealing temperatures during the post-treatment process. Therefore, the solar water splitting photocurrent density for the particle-assembled Zr-doped β-Fe 2 O 3 photoanodes was improved significantly to 1.2 mA cm -2 , which is five times greater than that of pure β-Fe 2 O 3 photoanodes. Thus, this study suggests metastable β-Fe 2 O 3 films as promising photoanode materials for solar water splitting.  [36], four new vibrational peaks at 258, 397, 522 and 635 cm -1 were observed, whereas the vibrational peak at 208 cm -1 was not observed.

RESULTS AND DISCUSSION
The theoretical calculations and experimental values of Raman vibrational peaks are listed in Supplementary Table S3. Among the 22 vibrational modes predicted by the theoretical calculations, only 12 significant peaks were observed in the experiments. Three weak vibrational modes are marked with asterisks. The M10 (383 cm -1 ) and M11 (389 cm -1 ) modes merge into one peak (386 cm -1 ) in the experimental spectrum. As given in Supplementary Fig. S1, typical peaks corresponding to vibrational modes 158 cm -1 (T g ), 274 cm -1 (T g ), 368 cm -1 (T g ) and 386 cm -1 (E g +T g ) can be used as a characteristic spectrum for identifying the phase of β-Fe 2 O 3 . In addition, the theoretically calculated infrared vibrational peaks and the experimental infrared absorption spectrum are also given in Supplementary Table S4 and Supplementary Fig. S2, respectively.
The metastable β-Fe 2 O 3 phase may be converted to the mature α-Fe 2 O 3 phase upon laser irradiation of different wavelengths and power. Figure 1 shows the effects of laser wavelength on the Raman spectrum of β-Fe 2 O 3 . The Raman peaks at 227, 293, 411, 499 and 613 cm -1 belonging to α-Fe 2 O 3 appeared after irradiation of the β-Fe 2 O 3 samples with a 532 nm or 633 nm laser. In the case of 785 nm laser irradiation, the experimental and theoretical Raman peaks for β-Fe 2 O 3 match in shape but exhibit a slight shift. These results suggest that a 785 nm laser is ideal for detecting the Raman spectra of metastable β-Fe 2 O 3 .
Raman spectra mapping images shown in Fig. 2 also indicate phase transformation of the particleassembled β-Fe 2 O 3 films without any necking treatment, in which the α-Fe 2 O 3 phase appears when the β-Fe 2 O 3 film samples are irradiated with a 10 mW laser of 532 and 633 nm. Scanning electron microscope (SEM) images also suggest this phase transformation. Even when β-Fe 2 O 3 was irradiated with a laser of 785 nm with 10 mW power, slight damage of the β-Fe 2 O 3 films was observed.
Additionally, the effects of laser power on the Raman spectrum of β-Fe 2 O 3 were investigated. When the power of the 785 nm laser is increased, a phase change of β-Fe 2 O 3 to α-Fe 2 O 3 possibly occurs (Fig. 3). After β-Fe 2 O 3 is irradiated by 0.4 mW 785 nm laser for an extended period of 500 s, there is no signal corresponding to α-Fe 2 O 3 . A significant phase transformation occurs with as little as 2 s irradiation with a 4 mW 785 nm laser. As the laser power increases and the exposure time is prolonged, all the β-Fe 2 O 3 gradually converts to α-Fe 2 O 3 . Figure 4 shows Raman spectra mapping and SEM images of the particle-assembled β-Fe 2 O 3 films without any necking treatment after laser illumination of 785 nm. When the β-Fe 2 O 3 films were irradiated with 785 nm laser at 0.4 W power, no α-Fe 2 O 3 signal appeared. With the increasing power of the 785 nm laser, the signals of α-Fe 2 O 3 were enhanced, suggesting laser-induced phase transformation of β-Fe 2 O 3 to α-Fe 2 O 3 . The SEM images of Fig. 4 also corroborate this result. The phase transformation of β-Fe 2 O 3 to α-Fe 2 O 3 is initiated by laser irradiation, likely as a result of the thermal effects of the laser. Therefore, low-power laser irradiation is required to avoid the phase transformation of metastable materials during Raman characterization to minimize its thermal effects. Thus, a 0.4 mW 785 nm laser is ideal for measuring the Raman spectra of β-Fe 2 O 3 . Figure 5 shows transmission electron microscope (TEM) and selected area electron diffraction (SAED) images of the β-Fe 2 O 3 samples. A simulated SAED image of α-Fe 2 O 3 is shown in Fig. 5 for comparison. The measured SAED image of β-Fe 2 O 3 samples (Fig. 5a) is in good agreement with the simulated SAED image of α-Fe 2 O 3 (Fig. 5d) rather than the simulated SAED image of β-Fe 2 O 3 samples (Fig. 5c) occur. Indeed, after being annealed at 1023 K, the characteristic Raman peaks of α-Fe 2 O 3 appeared in the Raman spectrum of the β-Fe 2 O 3 sample ( Supplementary Fig. S9a), indicating that there was a phase transformation. The Raman spectra of Zr-doped β-Fe 2 O 3 did not change after being annealed at 1023 K ( Supplementary Fig. S9b), indicating that there was no phase transformation. Note that the results of differential scanning calorimetry show that the phase transition temperature of Zr-doped β-Fe 2 O 3 is clearly increased, as shown in Supplementary Fig. S10. The above results show that Zr doping effectively increases the phase transition temperature of metastable iron oxide. SEM images of the particle-assembled β-Fe 2 O 3 photoanodes with and without Zr doping are shown in Supplementary Fig. S11. Figure 6a shows a comparison of the PEC performances of the particle-assembled β-Fe 2 O 3 photoanodes with and without Zr doping under AM 1.5 G irradiation (100 mW cm -2 ). The results show that the photocurrent of Zr-doped β-Fe 2 O 3 photoanodes is much higher than that of pure β-Fe 2 O 3 photoanodes. Zr-doped β-Fe 2 O 3 photoanodes also exhibit much larger monochromatic incident photon-to-electron conversion efficiency (IPCE) than pure β-Fe 2 O 3 photoanodes (Fig. 6b). Compared to the pure β-Fe 2 O 3 , the carrier density of Zr-doped β-Fe 2 O 3 has been improved from 1.8 × 10 20 to 4.6 × 10 20 cm −3 , which can be estimated from the Supplementary Fig. S12a. As shown in Supplementary Fig. S12b, the charge separation efficiency of Zr-doped β-Fe 2 O 3 is improved by approximately four times at 1.6 V versus Reversible Hydrogen Electrode (RHE), compared to that of β-Fe 2 O 3 . The increase in the carrier concentration may favor the charge transport and contribute to the improvement in water splitting photocurrent of the Zr-doped β-Fe 2 O 3 photoanode.
Under AM 1.5 G irradiation (100 mW cm -2 ), both Zr-doped and pure β-Fe 2 O 3 photoanodes exhibit very good photochemical stability for water splitting during the PEC reaction over 10 h, as shown in Fig. 6c. In the case of particle-assembled Zr-doped β-Fe 2 O 3 photoanodes, the Faradaic efficiencies for hydrogen production and oxygen production are 85% and 80%, respectively (Fig. 6d). This result suggests that the photocurrent of the particle-assembled β-Fe 2 O 3 photoanodes is mainly attributed to water splitting. During the process of solar water splitting, β-Fe 2 O 3 as a photoanode material may be exposed to standard simulated sunlight for a long time. To assess the possibility for phase transformation of β-Fe 2 O 3 to α-Fe 2 O 3 under simulated sunlight, the Raman spectra of the samples were measured before and after the photochemical stability testing. As shown in Supplementary Fig. S13, the phase of the β-Fe 2 O 3 photoanodes before and after the reaction of PEC water splitting remains unchanged. In a typical PEC reaction, the intensity of the light used is not sufficient to promote the phase transformation of β-Fe 2 O 3 particle-assembled films. Therefore, β-Fe 2 O 3 is a promising photoanode for water splitting. The PEC performance of β-Fe 2 O 3 may be improved further by not only the rational design of electrocatalysts but also the optimization of preparation methods [41,42].

CONCLUSION
In conclusion, the Raman spectrum of β-Fe 2 O 3 under excitation by a 785 nm laser with 0.4 W power shows 12 significant vibrational modes corresponding to β-Fe 2 O 3 . The phase transformation of metastable β-Fe 2 O 3 to mature α-Fe 2 O 3 may be observed under laser irradiation, electron irradiation and heating. Zr doping was introduced to particleassembled β-Fe 2 O 3 films, thus not only increasing the carrier concentration but also suppressing the phase transformation of β-Fe 2 O 3 . The PEC performance of the Zr-doped β-Fe 2 O 3 photoanode was vastly boosted, in comparison with that of the pure β-Fe 2 O 3 photoanode. This study demonstrates that metastable β-Fe 2 O 3 remains stable during the PEC reaction and is a promising photoanode material for decomposing water, thus paving the road toward the use of β-Fe 2 O 3 in solar water splitting (Supplementary Scheme S1).

METHODS
In this study, β-Fe 2 O 3 powder was synthesized by calcining the mixture of NaFe(SO 4 ) 2 (or 1.5% ZrSO 4 doped NaFe(SO 4 ) 2 ) and NaCl in a muffle furnace at 450 • C for 1 h [43]. As-prepared β-Fe 2 O 3 powder was then deposited on fluorinedoped tin oxide (FTO) glass by electrophoretic deposition to prepare the particle-assembled films. In the case of necking treatment for the particle-assembled β-Fe 2 O 3 films, 0.2 μmol of TiCl 4 in methanol solution was dropped on the particle-assembled β-Fe 2 O 3 films. Afterward, the  films were annealed in a muffle furnace at 600 • C (or 650 • C) for 1 h. Raman spectra of the samples were characterized with a confocal laser Raman spectrometer (Japan, Horiba, LabRAM Aramis, calibrated with silicon). Except for Fig. 1, all of the Raman spectra data acquisition used a 785 nm laser as the excitation source. Except for Fig. 3, all of the Raman spectra data acquisition used a 0.4 mW power laser as the excitation source. Raman imaging was obtained using a 785 nm, 0.4 mW laser. All laserinduced phase-change samples were confirmed using a 785 nm laser as the excitation source before laser irradiation to ensure that the samples contained no α-Fe 2 O 3 . The phosphor spectrum calculations were calculated in Material Studio with the Local Density Approximate, using the norm-conserving situation.

RESEARCH ARTICLE
Density functional theory calculations on Zr-doped β-Fe 2 O 3 were implemented in the VASP (Vienna Ab-initio Simulation Package) with a projected-augmented-wave method in the scheme of generalized-gradient approximation, whereas the strong on-site Coulomb repulsion among the localized Fe 3d electrons was described with the generalized-gradient approximation +U approach ('U' is 'the strength of the on-site Coulomb interaction').
SEM images were obtained on an SEM (Germany, Zessis, Ultra 55) and SAED images of the samples were obtained on a TEM (Japan, JEOL, Ltd. JEM2100).