Effect of preparation conditions and Co–Pi groups as a noble metal-free redox mediator and hole extractor to boost the photoelectrochemical water oxidation for 1D nanorod α-Fe₂O₃

Abstract

Photoelectrochemical water splitting for hydrogen production is an ideal method for solar energy utilization. Herein, α-Fe₂O₃ was prepared by a hydrothermal process and the precursor concentration, calcination temperature and hydrothermal time were optimized. The Co–Pi groups were deposited on the surface by two-step dipping process. The preparation conditions and loading capacity of CoPi were optimized. SEM showed that the morphology of α-Fe₂O₃ was short nanorods. XPS indicated that CoPi was successfully deposited on the surface of Fe₂O₃. TEM and XRD showed that CoPi was amorphous. The photocurrent density of Fe₂O₃/CoPi increased to 1.95 mA cm⁻² at 1.23 V_RHE, which was 1.91 times that of bare Fe₂O₃. After CoPi modification, the initial potential shifted negatively by 70 mV. It could be inferred that CoPi acted as an effective water oxidation co-catalyst to reduce the activation energy of the oxygen evolution reaction. The chopping I–t curve showed that the transient anodic spike of Fe₂O₃ decreased after CoPi deposition. This indicated that CoPi acted as a hole extractor and redox mediator to reduce the charge accumulation on the Fe₂O₃ surface. Through calculation, the ABPE of Fe₂O₃/Co–Pi was 2.1 times that of Fe₂O₃. The charge injection efficiency of Fe₂O₃/Co–Pi at 1.23 V_RHE was 1.6 times that of Fe₂O₃. This study presented a richer understanding on the functions and promotional effects of the preparation conditions and CoPi groups for Fe₂O₃ photoanodes.

---

Introduction

As a superior process of artificial photosynthesis, photoelectrochemical (PEC) water splitting has become a promising route toward sustainable hydrogen production and solar energy storage.^1,2 However, the efficiency of PEC water splitting is largely limited by water oxidation, which requires high-performance photoanodes to improve the conversion efficiency of solar energy. N-type semiconductors, such as TiO₂,^3–6 BiVO₄,^7,8 and α-Fe₂O₃,^9–11 with band gaps of ∼3.2 eV, ∼2.4 eV and ∼2.1 eV, respectively, have been researched. From the relationship between the semiconductor band gap and light absorption wavelength (E_g (eV) = 1240/λ (nm)), it could be found that the maximum light absorption wavelength of α-Fe₂O₃ would reach 590 nm.¹² Moreover, α-Fe₂O₃ is a non-toxic, stable, and low-cost photoanode material. Its outstanding advantages have attracted a lot of attention in recent years. There are many preparation processes for Fe₂O₃, such as the hydrothermal process, precipitation process, thermal decomposition process, sol–gel process, and others. The most common process is hydrothermal. The advantage of using this process to prepare Fe₂O₃ nanorod arrays on FTO conductive substrates is that the process is easy to control. In addition, due to the large exposed area, the one-dimensional nanorod-like structure facilitates electron transport. This contributes to the charge separation in the process of photoelectrolysis of water. However, α-Fe₂O₃ has a short hole diffusion distance (2–4 nm) and carrier lifetime (10⁻¹² s), which could lead to the easy recombination of electrons and holes, thereby reducing the efficiency of PEC water splitting.¹³ Under simulated sunlight, the theoretical photocurrent density of α-Fe₂O₃ is about 13 mA cm⁻² at 1.23 V_RHE.¹⁴ However, the actual photocurrent density is much lower than the theoretical value. It is important to increase the photocurrent density of α-Fe₂O₃, as well as reduce the overpotential of the water oxidation reaction during the PEC water splitting.

The sluggish oxygen production reaction rate limits the widespread use of α-Fe₂O₃. Therefore, many studies are focused on overcoming the reaction energy barrier and increasing the charge injection efficiency by introducing a co-catalyst to accelerate the water oxidation reaction. Common cocatalysts used to modify α-Fe₂O₃ include BiVO₄,¹⁵ IrO₂,¹⁶ Rh₂O₃,¹⁷ Co–Pi,^18–21 Co₃O₄,²² CoO_x,²³ NiFeO_x²⁴ and transition metal (Fe, Co, Ni) oxyhydroxide.^25,26 Co–Pi is a superior OER co-catalyst, which has the advantages of abundant sources, no pollution, and low preparation cost. The processes of modifying Co–Pi groups on α-Fe₂O₃ mainly include the electrodeposition process,²¹ light-assisted electrodeposition process,²⁷ and immersion process. The thickness of the Co–Pi deposition should be moderate. Too thick will hinder the light absorption of α-Fe₂O₃, while too thin will not be enough to improve the photoelectric performance. When using the electrochemical process to deposit Co–Pi groups, it is necessary to consider whether the applied voltage will cause electrical corrosion of the α-Fe₂O₃ photoanode.

Herein, the α-Fe₂O₃ photoanode was synthesized on the surface of FTO by the hydrothermal process, and the precursor concentration, calcination temperature and hydrothermal time were optimized. Then, CoPi groups were deposited on the surface of α-Fe₂O₃ through an impregnation process. SEM showed that the morphology of Fe₂O₃ prepared by hydrothermal process is a short nanorod. As the calcination temperature increased, the thickness of Fe₂O₃ also increased. XPS detected Co and P elements in the sample. It indicated that CoPi was successfully deposited on the surface of Fe₂O₃. There was no peak of CoPi in the XRD of Fe₂O₃/CoPi, and there was no lattice fringe of CoPi in the TEM of Fe₂O₃/CoPi. These results indicated that the CoPi groups on the Fe₂O₃/CoPi photoanode were amorphous. It was found that the bare Fe₂O₃ photoelectrode had the best photoelectric performance when the Fe precursor concentration, hydrothermal reaction time, and calcination temperature were 6 mM, 4 h, and 600 °C, respectively. The photocurrent density of the optimized Fe₂O₃ photoelectrode at 1.23 V_RHE was 1.02 mA cm⁻², which was about 29 times (0.035 mA cm⁻²) higher than that of the sample without condition exploration. The CoPi groups were supported on the Fe₂O₃ surface through the dipping process, and the optimal immersion times of the electrode in the Co and Pi precursor solutions were both 6 h. The photocurrent density of the optimized Fe₂O₃/CoPi at 1.23 V_RHE was 1.95 mA cm⁻², which was 1.91 times that of the bare sample. The initial potential of Fe₂O₃/CoPi was 0.78 V_RHE, which was 70 mV lower than that of the bare Fe₂O₃. According to the law of conservation of energy, it could be inferred that CoPi reduced the activation energy of OER and accelerated the water oxidation reaction. Therefore, CoPi was an effective OER co-catalyst. The chopping I–t curve showed that the transient anodic spike of Fe₂O₃ decreased after CoPi deposition. It indicated that the charge accumulation on the Fe₂O₃/CoPi surface was reduced. EIS showed that after CoPi modification, the electrode/electrolyte interface resistance of Fe₂O₃ decreased. This showed that CoPi could change the interface performance. Combining the chopping I–t curve and EIS analysis, it could be concluded that CoPi could be used as a hole extractant and redox mediator to promote charge injection on the surface of Fe₂O₃. Through calculation, the applied bias photon-to-current efficiency (ABPE) of the Fe₂O₃/Co–Pi photoanode was 0.17% (2.1 times that of the bare Fe₂O₃). The charge injection efficiency (η_inj) of the Fe₂O₃/Co–Pi photoanode at 1.23 V_RHE was 1.6 times that of Fe₂O₃. In short, the optimization of the preparation conditions of Fe₂O₃ and the modification of the Co–Pi groups as a non-noble metal catalyst and hole extractor effectively promoted the water oxidation reaction, which was beneficial to the design and synthesis of future photoelectrodes.

Experimental section

Materials

Acetone, ferric nitrate nonahydrate, cobalt nitrate hexahydrate, potassium hydroxide, dodecahydrate and sodium phosphate were purchased from Sinopharm Chemical Reagent Co., Ltd. Anhydrous ethanol and ammonium fluoride were obtained from Xilong Science Co., Ltd. Sodium sulfite was purchased from Macleans Chemical Reagent Co., Ltd. FTO conductive glass was obtained from Opivite Company. All of the chemicals were of analytical grade, and deionized (DI) water was used for all of the experiments.

Preparation of photoelectrodes

Fabrication of the α-Fe₂O₃ electrode. First, the FTO conductive glass (3 cm × 1 cm) was cleaned ultrasonically with an ethanol–acetone mixture (V_ethanol : V_acetone = 1 : 1) and deionized water for 30 minutes. Then, they were dried in an oven at 60 °C. A 35 mL precursor solution containing 6 mM Fe(NO₃)₃·9H₂O and 0.1 M NH₄F was prepared, and then stirred with a magnet for 10 min to fully dissolve the solids. Second, the FTO conductive glass was put into a polytetrafluoroethylene liner containing 35 mL of the precursor solution, and the reactor was transferred to an oven. The oven was heated at 110 °C for 4 hours. After the hydrothermal deposition, the FTO substrate was taken out and rinsed with ethanol and deionized water. Finally, the sample obtained by the hydrothermal process was put into a tubular resistance furnace for air atmosphere annealing. After annealing at 500 °C for 2 h, the substrate became red-brown. The first two steps in Fig. 1 are a schematic diagram of the preparation process of α-Fe₂O₃.

Fig. 1 Schematic diagram of the preparation process of α-Fe₂O₃ and α-Fe₂O₃/Co–Pi photoelectrodes.

Preparation of the α-Fe₂O₃/Co–Pi photoelectrode

First, 0.1 mol L⁻¹ sodium phosphate (Na₃PO₄) and 0.01 mol L⁻¹ cobalt nitrate (Co(NO₃)₂)solution were prepared. After that, the as-prepared α-Fe₂O₃ nanorod was placed into 0.1 mol L⁻¹ Na₃PO₄ solution for immersion of 6 h. The non-conductive side was close to the inner wall of the beaker; simultaneously, stirring occurred during immersion. After the first step of immersion, the sample was taken out and blow dried with cold air. In the next step, the treated sample was immersed in the 0.01 mol L⁻¹ Co(NO₃)₂ solution for 3 h, 6 h, 9 h, and 12 h to form Fe₂O₃/Co–Pi. Then, the sample was taken out and dried with cold air. Finally, the sample was rinsed with deionized water, and dried at 60 °C. The last two steps in Fig. 1 are a schematic diagram of the preparation process of Co–Pi deposition.

Characterizations

Field emission scanning electron microscope (FESEM) and energy dispersive X-ray spectroscopy (EDS). In order to obtain the morphology and element distribution of the samples, SEM and EDS tests were performed on a field emission scanning electron microscope model (Nova Nano SEM 230). All samples were not sprayed with gold when the front view was taken. The test voltage was 8 kV. Since the conductivity of the FTO substrate was not good enough, all samples were sprayed with gold for 1 minute when the cross-sectional view was taken.

Transmission electron microscope (TEM) and high-resolution transmission electron microscope (HRTEM)

In order to observe the morphology and lattice fringes of the sample, the TEM test was carried out in a JEOL JEM 2100 transmission electron microscope. The powder was scraped gently from the substrate with a knife. The powder was ultrasonically dispersed with ethanol for 10 minutes. The sample was then dropped on a micro-mesh copper net, dried and observed through an electron microscope.

X-ray diffraction (XRD)

The phase analysis of the sample was carried out by XRD test. The model of the polycrystalline diffractometer was DY5261/Xpert3, the test voltage was 40 kV, the current was 40 mA, the radiation source was Cu Kα, and the scanning range was 5–80°.

Raman spectrum

In order to obtain the lattice vibration information of the sample, the Raman measurement was performed on an Invia Reflex laser Raman spectrometer with an excitation wavelength of 532 nm.

X-ray photoelectron spectroscopy (XPS)

The XPS study was used to obtain the element and valence information of the sample, and the instrument was an X-ray photoelectron spectrometer (Thermo Fischer, ESCALAB Xi+, USA). Among them, the vacuum degree of the analysis chamber was 8 × 10⁻¹⁰ Pa, the excitation source was Alka rays (hν = 1486.6 eV), the working voltage was 12 kV, and the filament current was 6 mA. All of the data were calibrated with the standard binding energy (C 1s = 284.80 eV).

Ultraviolet visible absorption (UV-vis) and diffuse reflectance spectrum

The absorbance (A) and reflectance (R) of the sample were tested on a Lambda 750 ultraviolet-visible-near-infrared photometer in the diffuse reflection mode. The scanning range was 200–800 nm. Through the integrating sphere of the UV-vis spectrophotometer, the spectral curve of the reflectance of the sample against the wavelength should be measured first, and then the (Fhv)² − (hv) curve could be calculated by the following equations. The calculated equations were as follows:

(Fhν)ⁿ = A(hν − E_g). (1)

F = (1 − R)²/2R. (2)

hν = 1240/λ. (3)

where R is the reflectivity, h is the Planck constant (6.626 × 10⁻³⁴ J s), c is the speed of light (3 × 10⁸ m s⁻¹), λ is the incident light wavelength (nm), and e is the electron (1.6 × 10⁻¹⁹), and n is determined by the inter-band transition form. For α-Fe₂O₃, the value of n is 2. After calculation, the relationship (Fhν) versus (hν) as plotted to obtain a curve and a tangent line was drawn to the curve. The intersection of the tangent and the abscissa is the band gap (E_g) of the sample.

PEC measurements

The PEC measurements were carried out in a self-made H-shaped electrolytic cell equipped with a quartz glass plate, and the tests used a three-electrode system. The reference electrode was a Hg/HgO electrode, and the light was irradiated from the back of the FTO. The counter electrode was a Pt electrode, and the working electrode was a FTO conductive glass with deposited materials. The light area of the working electrode was corrected to 1 cm², and the electrolyte was 1 M KOH (pH = 13.6) solution. All data were recorded on a CHI750E electrochemical workstation and online computer. The conversion relationship between the reversible hydrogen electrode potential and the Hg/HgO electrode (4) was as follows:

E_RHE = E_Hg/Hg + 0.098 + 0.59pH. (4)

Linear sweep voltammetry (LSV) test

The test was conducted in a three-electrode system with a 300 W PLS SXE300C xenon lamp as the light source. The light intensity was adjusted to 100 mW cm⁻² with a filter. The scanning voltage range was 0.3–1.8 V_RHE with a scanning speed of 10 mV s⁻¹.

Chopping Chronoamperometry (I–t) test

Chopping I–t curves were measured with an external potential of 1.23 V_RHE, under chopped illumination. The illumination parameters during the test were as follows: first light for 40 seconds, then shading for 40 seconds, and so on until 400 seconds.

Electrochemical impedance spectroscopy (EIS) test

The electrochemical impedance test was carried out under simulated sunlight, with an applied voltage of 1.23 V_RHE. The test frequency was 1–10⁶ Hz, and the disturbance was 5 mV. The obtained spectrum was fitted with Zview software.

Flat band potential (Mott–Schottky) test

The test was performed in the dark state, and the test frequency (f) was 10⁴ Hz. The carrier density N_d could be calculated by the Mott–Schottky equation:

N_d = (2/eε₀ε₁)|d(1/C²)dV|. (5)

where e is the charge of the electrons (1.602 × 10⁻¹⁹), ε₀ is the vacuum dielectric constant (8.854 × 10⁻¹² CV⁻¹ m⁻¹), and ε_r is the semiconductor dielectric constant (α-Fe₂O₃ takes 80 CV⁻¹ m⁻¹).

Applied bias photon-to-current efficiency (ABPE)

The test conditions were the same as the linear voltammetry scan under illumination. The calculation formula of ABPE was as follows:

ABPE = J_photo(1.23V_RHE − V_app)/P_light. (6)

where J_photo the photocurrent density, V_app the applied voltage (relative to reversible hydrogen), and P_light the light intensity (100 mW cm⁻²).

Charge separation efficiency (η_sep) and charge injection efficiency (η_inj)

The test was carried out in a three-electrode system. The electrolyte was a 1 M KOH solution containing 0.5 M Na₂SO₃ under light conditions. The values of η_sep and η_inj were obtained by the following eqn (7) and (8):

η_sep = J_Na2SO3/J_abs. (7)

η_inj = J_H2O/J_Na2SO3. (8)

where J_Na2SO3 is the photocurrent density when a hole scavenger (Na₂SO₃) is added to the electrolyte. Assuming APCE = 100%, the J_abs of Fe₂O₃ is calculated to be 11.4 mA cm⁻² by estimating the overlap area between the UV-vis absorption spectrum and the AM 1.5G solar spectrum.²⁸

Results and discussion

Physical and chemical characterization

SEM. Fig. 2 is the FESEM image of α-Fe₂O₃ calcined at various temperatures of 500 °C, 550 °C, and 600 °C, as well as α-Fe₂O₃/Co–Pi photoanodes. Fig. 2a shows the FESEM image of α-Fe₂O₃ calcined at 500 °C. It could be found that a dense film composed of aggregated short nanorods was formed on the surface of the FTO substrate with some small gaps. This might be attributed to the composite of other small units in the film. The diameter of the short nanorods ranged from 40 to 90 nm, with an average diameter of 60 nm. The inset in Fig. 2a shows that the film thickness was 190 nm. Fig. 2b shows the FESEM image of α-Fe₂O₃ calcined at 550 °C. It was found that hematite was still composed of short nanorods (∼40 nm), and the film thickness increased to 230 nm. Fig. 2c indicates that after calcination at 600 °C, the surface of α-Fe₂O₃ was smoother. The average diameter of α-Fe₂O₃ calcined at 600 °C was 30 nm with a film thickness of 320 nm. Fig. S1 (ESI†) is an enlarged SEM image of the cross-section of the Fe₂O₃ photoanode calcined at 600 °C. It can be seen from the dotted frame that Fe₂O₃ was in the shape of short nanorods. Combined with Fig. 2a–c, it was found that as the calcination temperature increased, the thickness of the Fe₂O₃ film increased. Fig. 2d shows the FESEM image of the α-Fe₂O₃/Co–Pi photoanode. It indicated that the morphology of hematite remained unchanged after the modification of Co–Pi groups. The EDS scanning spectra of α-Fe₂O₃ and α-Fe₂O₃/Co–Pi photoanodes are shown in Fig. S2 (ESI†). It showed that all samples had peaks of F, Si, Sn, and Ca elements, which were attributed to the FTO conductive glass. The peaks of the Fe and O elements appeared among the α-Fe₂O₃ and α-Fe₂O₃/Co–Pi photoanodes. It indicated that Fe₂O₃ was successfully prepared on the FTO substrate. There was no obvious peak of Co and P elements for the Fe₂O₃/Co–Pi photoanode. That was mainly because only a small amount of Co–Pi groups could be deposited on the surface of Fe₂O₃ by the immersion process. The scanning depth of EDS had reached the bottom glass layer instead of the surface sample layer, so the surface Co–Pi was not scanned. Table S1 (ESI†) is the contents of the corresponding elements in the α-Fe₂O₃ and α-Fe₂O₃/Co–Pi photoanodes obtained from EDS. It showed that the contents of F, Si, Sn, Ca, O, and Fe in α-Fe₂O₃ were 3.10 wt%, 7.47 wt%, 54.03 wt%, 4.09 wt%, 6.21 wt%, and 25.08 wt%, respectively. The contents of F, Si, Sn, Ca, O, and Fe in α-Fe₂O₃/Co–Pi were 3.13 wt%, 8.32 wt%, 57.17 wt%, 4.37 wt%, 5.97 wt%, and 21.04 wt%, respectively.

Fig. 2 FESEM images of α-Fe₂O₃ calcined at various temperatures of 500, 550, and 600 °C, as well as α-Fe₂O₃/Co–Pi photoanodes, (a) 500 °C, (b) 550 °C, (c) 600 °C, (d) α-Fe₂O₃/Co–Pi, and the inset is a cross-sectional view.

TEM. TEM and HRTEM were used to further explore the effect of Co–Pi groups on α-Fe₂O₃. Fig. 3a is the TEM image with 50 000× magnification of Fe₂O₃, which showed that Fe₂O₃ was rod-shaped and the nanorods were tightly connected. Fig. S3 (ESI†) is the TEM image with 50 000× magnification of Fe₂O₃ with or without a dashed frame. It could be seen from the dotted frame that Fe₂O₃ had the morphology of short nanorods. The result was consistent with the analyses of FESEM. Fig. 3b is the TEM image with 50 000× magnification of Fe₂O₃/Co–Pi. In this image, the red circle showed some small particles on the top of the nanorods. The particles were densely distributed, with a particle size of about 10 nm. Fig. 3c is the HRTEM image of bare α-Fe₂O₃. There was only one lattice fringe with a lattice spacing of 0.25 nm in it, which corresponded to the (110) crystal plane of α-Fe₂O₃.²⁹ In the HRTEM diagram of α-Fe₂O₃/Co–Pi shown in Fig. 3d, the red dotted line indicated that an amorphous Co–Pi layer was formed on the edge of α-Fe₂O₃.³⁰ It showed that Co–Pi was successfully deposited on the surface of α-Fe₂O₃. It could be found that the thickness of the amorphous Co–Pi layer was only ∼2 nm, which inferred that Co and P could not be detected in the EDS measurements.

Fig. 3 (a) TEM image with 50 000× magnification of α-Fe₂O₃. (b) TEM image with 50 000× magnification of α-Fe₂O₃/Co–Pi. (c) HRTEM image of α-Fe₂O₃, and (d) HRTEM image of α-Fe₂O₃/Co–Pi.

XRD and Raman. XRD measurements were conducted to obtain the phase information of all samples. The XRD spectra for the FTO substrate, Fe₂O₃ calcined at various temperatures of 500, 525, 550, 575, and 600 °C, as well as Fe₂O₃/Co–Pi photoanodes are shown in Fig. S4 (ESI†) and Fig. 4a. Fig. S4 (ESI†) shows that the hematite calcined at various temperatures of 500, 525, 550, 575, and 600 °C had the characteristic diffraction peaks of SnO₂ (JCPDS No. 46-1088) from the FTO substrate. The diffraction peaks located at 35.7° and 64.1° corresponded to the (110) and (330) crystal planes of α-Fe₂O₃ (JCPDS No. 33-0664), respectively. As was shown in Fig. 4a, after the modification of the Co–Pi groups, there was no new diffraction peak in the XRD pattern of Fe₂O₃/Co–Pi. It indicated that Co–Pi was amorphous, which was consistent with the HRTEM results. Raman spectroscopy could obtain the lattice vibration information of the samples. The Raman spectra for the FTO substrate, Fe₂O₃, and Fe₂O₃/Co–Pi photoanodes are shown in Fig. 4b. The Raman spectrum of the FTO substrate had four peaks located at 451, 557, 788, and 1090 cm⁻¹, which were attributed to the peaks of SnO₂. After Fe₂O₃ deposition on the surface of the substrate by the hydrothermal process, the vibration intensity of the FTO substrate became weaker. There were obvious Raman peaks of hematite, located at 413, 503, 613, 663, and 1318 cm⁻¹. The positions of these peaks were close to the standard peaks of α-Fe₂O₃.³¹ The peaks at 413 and 613 cm⁻¹ were attributed to the E_g vibration mode of hematite, and the peak at 613 cm⁻¹ was attributed to the A_1g vibration mode of hematite. At 667 cm⁻¹, an infrared (IR) active longitudinal optical (LO) Eu mode could be detected. It was caused by disorder in the α-Fe₂O₃ lattice.³² The peak at 1320 cm⁻¹ was derived from the dispersion of the two magnons, which was produced by the interaction of the two magnons in the antiparallel closing direction of the spins.³¹ The above results indicated that α-Fe₂O₃ was successfully synthesized without any other impurities. After the deposition of Co–Pi, there was no new Raman peak in Fe₂O₃/Co–Pi. It indicated that Co–Pi has not entered the lattice of Fe₂O₃.

Fig. 4 (a) XRD pattern and (b) Raman spectra of the FTO substrate, bare Fe₂O₃, and Fe₂O₃/Co–Pi photoanodes, SnO₂ (black five-pointed star), α-Fe₂O₃ (red snowflake).

XPS. XPS characterization can confirm the element composition and the oxidation state of the element on the surface of samples. The XPS spectra and fitted spectra of the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes are shown in Fig. 5. Among them, Fig. 5a–d are the spectra of Fe 2p, O 1s, Co 2p, and P 2P, respectively. In Fig. 5a, the XPS spectrum of Fe 2p showed that the binding energies of 2p_1/2 and 2p_3/2 were 724.5 eV and 710.7 eV, respectively. There were two satellite peaks located at 719.3 eV and 733.5 eV. It indicated that there were hematite phase iron oxides in both Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes, while there were no Fe(II) and Fe(0) peaks. This analysis was consistent with the XRD and Raman characterization. The peaks at 529.9, 531.5, and 532.7 eV corresponded to the lattice oxygen (M–O), the adsorbed hydroxyl (M–OH lattice) and the hydroxyl in the adsorbed water molecule in iron oxide in Fig. 5b. After the modification of the Co–Pi groups, the binding energy of the M–O bond was negatively shifted by 0.23 eV, which indicated that the electron density near the Fe site increased. It could be observed that after CoPi modification, the peak intensity of M–OH increased. This was because the sample used in the XPS characterization was photoelectrically tested. This made the CoPi group oxidized, and the oxidized CoPi was composed of CoOOH.³³ After CoPi deposition, the peak intensity also increased. This was due to the fact that the loading of CoPi would cause more water molecules to be adsorbed on the Fe₂O₃ surface. It would promote the water oxidation reaction, thereby improving the PEC performance. The spectrum of Co 2p in Fig. 5c showed the peaks at 780.6 eV and 796.6 eV corresponding to the signals of 2p_3/2 and 2p_3/2, respectively, which were characteristic peaks of Co²⁺ and Co³⁺. The peaks at 787.0 eV and 802.8 eV also confirmed the presence of Co²⁺. In Fig. 5d, the main characteristic peak of P 2p was about 133.4 eV, which corresponded to the phosphate group in PO₄³⁻.²⁸ Table S2 (ESI†) presents the contents of each element for Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes obtained from XPS characterization. After the modification of the Co–Pi groups, the contents of the Co and P elements in the sample were 1.09 at% and 1.38 at%, respectively. It indicated that the Co–Pi deposited by the immersion process was very small.

Fig. 5 XPS spectra and fitted spectra of Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes: (a) Fe 2p, (b) O 1s, (c) Co 2p, (d) P 2P.

UV-vis absorption spectrum

For PEC water splitting, the optical property of the semiconductor materials is one of the important factors. Hence, UV-vis absorption spectra were measured to study the light absorption ability of bare Fe₂O₃ with various hydrothermal times of 2, 4, 6, 8, and 10 h, various concentrations of the Fe(NO₃)₃·9H₂O precursor of 2, 4, 6, 8, 10, and 12 mM, various calcination temperature of 500, 525, 550, 575, and 600 °C, as well as Fe₂O₃/Co–Pi photoanodes after immersion in phosphate for various times of 3, 6, 9, and 12 h in Fig. S5–S7 (ESI†) and Fig. 6. Fig. S5 (ESI†) shows that when the hydrothermal time increased from 2 h to 8 h, the light absorption in the 350–800 nm range gradually increased. Fig. S6 (ESI†) indicates that the light absorption in the range of 350–800 nm increased with the increase of the Fe precursor concentration. This was because as the hydrothermal time and concentration of the precursor increased, the thickness of the iron oxide film also increased, thereby increasing the light absorption. Fig. S7 (ESI†) shows that the light absorption of Fe₂O₃ in the range of 350–525 nm increased with the increasing calcination temperature. This was because the enhanced thickness promotes light absorption ability, which was consistent with SEM analysis. In order to obtain the optimal loading of the Co–Pi groups, we changed the immersion time of the α-Fe₂O₃ photoanode in 0.1 mol L⁻¹ Na₃PO₄ solution and 0.01 mol L⁻¹ Co(NO₃)₂ solution to explore the relationship between the loading and α-Fe₂O₃ photoelectric performance. The time of the two-step dipping always retained consistent changes throughout the experiment. Fig. 6 is the UV-vis absorption spectra of the Fe₂O₃/Co–Pi photoanode after immersion in phosphate for various times of 3, 6, 9, and 12 h. It showed that after depositing various amounts of Co–Pi on the surface of α-Fe₂O₃, the light absorption in the 350–550 nm range decreased. As the immersion time increased, the light absorption decreased slightly, which showed that the modification of the Co–Pi groups did not improve the photoelectric performance by enhancing light absorption. Comparatively speaking, when the two-step immersion time was 6 h, the light absorption of the Fe₂O₃/Co–Pi photoanode was the best, so we compared the reflection spectrum and band gap of Fe₂O₃/Co–Pi (6 h) with bare Fe₂O₃. Fig. 7a is the reflectance curve of Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes. Fig. 7b is the (Fhν)² − (hν) curve, which was calculated by the reflectivity of the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes. The band gaps of bare Fe₂O₃ and Fe₂O₃/Co–Pi obtained by extrapolation were 2.11 eV and 2.12 eV, respectively. The data indicated that the modification of Co–Pi groups would not change the internal energy band structure of Fe₂O₃.

Fig. 6 UV-vis absorption spectra of the Fe₂O₃/Co–Pi photoanodes after immersion in phosphate for various time of 3, 6, 9, and 12 h.

Fig. 7 (a) Reflectance-wavelength curve of Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes, (b) (Fhν)² − (hν) curves of Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes.

Photoelectrochemical tests

LSV and chopping I–t curves. The photocurrent densities under constant and transient condition were measured to study the PEC performance in the solar water splitting. The LSV curves of bare α-Fe₂O₃ prepared under various hydrothermal times of 2, 4, 6, 8, and 10 h and various precursor concentration conditions of 2, 4, 6, 8, 10, and 12 mM are shown in Fig. S8 and S9 (ESI†). Fig. S8 (ESI†) shows that when the hydrothermal time increased from 2 h to 4 h, the photocurrent density at 1.23 V_RHE increased from 0.035 mA cm⁻² to 0.084 mA cm⁻². However, when the hydrothermal reaction time continued to increase, the photocurrent density decreased instead. This was due to the fact that the extension of the hydrothermal reaction time would increase the thickness of α-Fe₂O₃, resulting in a longer distance for holes to migrate to the surface. It facilitated the recombination of the holes with the electrons. Fig. S9 (ESI†) shows that in the range of 1.0–1.6 V_RHE, as the concentration of Fe(NO₃)₃·9H₂O increased from 2 mM to 12 mM, the photocurrent density of the samples first increased and then decreased. This was because as the concentration of the precursor increases, the thickness of the Fe₂O₃ film also increased. This increased the light absorption and photocurrent density. However, if the precursor concentration continued to increase, the thickness of the film would be too large. Since the hole diffusion distance of iron oxide is only 2–4 nm, a thickness that is too large could easily recombine electrons and holes.

Fig. 8a is the LSV curves of bare Fe₂O₃ calcined at various temperatures of 500, 525, 550, 575, and 600 °C. It showed that the photocurrent densities of α-Fe₂O₃ calcined at 500 °C, 525 °C, 550 °C, 575 °C and 600 °C at 1.23 V_RHE were 0.012, 0.047, 0.18, 0.38, and 1.02 mA cm⁻², respectively. The current density of all samples in the dark state was almost zero. With the increased calcination temperature, the photocurrent density in the range of 1.1–1.6 V_RHE significantly increased. Fig. 8b is the I–t curves of bare Fe₂O₃ calcined at various temperatures of 500, 525, 550, 575, and 600 °C under chopped light. It showed that the transient photoelectric response at 1.23 V_RHE increased with the increase of the calcination temperature. When the calcination temperature was 600 °C in air condition, α-Fe₂O₃ had the highest transient photocurrent density. There was still an obvious peak current on transient PEC detection, which indicated that the charge recombination happened at the surface of the bare α-Fe₂O₃ semiconductor. The results in Fig. S6, S7 (ESI†), and Fig. 8 indicated that the temperature during the annealing had a greater impact on the PEC performance of Fe₂O₃. The optimum synthesis conditions of bare Fe₂O₃ were a hydrothermal time of 4 h, iron precursor concentration of 6 mM and calcination temperature of 600 °C for 2 h.

Fig. 8 (a) LSV curves of Fe₂O₃ calcined at various temperatures of 500, 525, 550, 575, and 600 °C; (b) Chopping I–t curves of Fe₂O₃ calcined at various temperatures of 500, 525, 550, 575, and 600 °C under chopped light, and the dotted line represents the dark state current density.

Fig. 9a is the LSV curve of the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes. It showed that the photocurrent density of Co–Pi modified Fe₂O₃ at 1.23 V_RHE increased from 1.02 mA cm⁻² to 1.95 mA cm⁻². The turn-on potential was reduced from 0.85 V_RHE to 0.78 V_RHE, which was negatively shifted by 70 mV. Based on the law of the energy conservation

Photocatalysis: E_g ≥ E° + E_a (9)

Photoelectrocatalysis: E_g + E_v + E° + E_AOP (10)

where E° is the chemical energy at the standard potential for O₂ evolution (1.23 V), E_a is the activation energy, and E_AOP is the activation energy in PEC water oxidation; the activation overpotential is denoted as AOP; E_g is the band gap of Fe₂O₃; E_V is the externally applied electric energy at the onset potential.

Fig. 9 (a) LSV curve of the Fe₂O₃ and Fe₂O₃/Co-Pi photoanodes under dark and light condition. (b) Schematic diagram of Fe₂O₃/Co–Pi reducing the activation energy of the water oxidation reaction.

For the water splitting reaction, the activation energy of the hydrogen evolution reaction (HER) is very low. Thus, here we mainly discussed the activation energy of water oxidation. After Co–Pi was deposited, the initial potential of all samples shifted negatively by about 70 mV. This indicated that less E_V was applied during the PEC water oxidation process. Co–Pi deposited on the Fe₂O₃ electrode could reduce the E_AOP of PEC water oxidation, thereby increasing the rate of OER. Fig. 9b is the schematic diagram of Fe₂O₃/Co–Pi reducing the activation energy of the water oxidation reaction. This result showed that the Co–Pi groups effectively reduced the water oxidation overpotential of Fe₂O₃ and accelerated the reaction.

Fig. 10 is the chopping I–t curve of the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes. It showed the transient current–time curve of the samples before and after modification under visible light. During the illumination period, the photocurrent density of Fe₂O₃ was about 0.9 mA cm⁻², and the photocurrent density of Fe₂O₃ modified with Co–Pi groups was about 1.8 mA cm⁻². The dotted frame in Fig. 10 was the peak-shaped current curve at the moment of turning on the light. With illumination, sharp anodic spikes were observed in the case of bare Fe₂O₃. This was due to the accumulation of photo-generated holes at the photoanode/electrolyte interface and were subsequently used in the recombination process. These transient anode spikes were reduced in the presence of CoPi. This meant that the holes in the Fe₂O₃ valence band (VB) were rapidly consumed at the semiconductor/electrolyte interface, so that the electron–hole pair recombination process on the electrode surface was reduced.

Fig. 10 Chopping I–t curve of the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes. The dotted line represents the current density in the dark state, and the dotted frame in figure is the peak-shaped current curve at the moment of turning on the light.

Applied bias photon-to-current efficiency (ABPE), charge separation efficiency and charge injection efficiency analysis

Fig. 11 is the ABPE curves of the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes. It showed that within the voltage range of 0.8–1.23 V_RHE, the ABPE of Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes both showed a trend of first increasing and then decreasing. The Fe₂O₃ photoanode had the largest ABPE at 1.05 V_RHE with a value of 0.08%. The Fe₂O₃/Co–Pi photoanode had an ABPE of 0.17% at 1.05 V_RHE, which was 2.1 times that of Fe₂O₃.

Fig. 11 Applied bias photon-to-current efficiency (ABPE) curves of Fe₂O₃ and Fe₂O₃/Co-Pi photoanodes.

Fig. 12a and b compares the charge separation efficiency (η_sep) and charge injection efficiency (η_inj) of the photoanodes before and after modification. Fig. 12a shows that the η_sep of Fe₂O₃ at 1.23 V_RHE was 20.39%. After Co–Pi modification, Fe₂O₃ had a similar η_sep with a value of 25.77%. Fig. S11a and b (ESI†) are the LSV curves of bare Fe₂O₃ and Fe₂O₃/CoPi without and with 0.5 M Na₂SO₃. From the curves, the η_inj of Fe₂O₃ and Fe₂O₃/Co–Pi could be calculated. The η_inj of Fe₂O₃ and Fe₂O₃/Co–Pi at 1.23 V_RHE shown in Fig. 12b were 41.76% and 66.67%, respectively. It could be seen that through the modification of Co–Pi, the η_inj of Fe₂O₃/Co–Pi was 1.6 times that of bare Fe₂O₃. Combining Fig. 12a and b, it could be obtained that the η_inj of Fe₂O₃ was improved even more after CoPi modification. This indicated that CoPi groups were more often used as a water oxidation promoter to accelerate the water oxidation reaction (OER).

Fig. 12 (a) Charge separation efficiency and (b) charge injection efficiency of bare Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes.

Fig. 13a is the LSV curve of the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes obtained after immersion in phosphate for various times of 3, 6, 9, and 12 h. It showed that after depositing various amounts of Co–Pi groups on the surface of Fe₂O₃, the photocurrent density in the entire test voltage range was higher than that of the bare Fe₂O₃. As the immersion time increased from 3 h to 6 h, the photocurrent density gradually increased. However, when the immersion time continued to increase to 12 h, the photocurrent density of Fe₂O₃ decreased instead. This result was consistent with the phenomenon shown in Fig. 6 that the immersion time was inversely proportional to the light absorption. Fig. 13b is the current–time curve of Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes obtained after immersion in phosphate for various times of 3, 6, 9, and 12 h. As shown, the transient photocurrent density at 1.23 V_RHE showed a trend of first increasing and then decreasing with the increase of Co–Pi loading. This trend was consistent with the change of the photocurrent density of the LSV curve. When the immersion time in Na₃PO₄ solution and Co(NO₃)₂ solution was 6 h, α-Fe₂O₃/Co–Pi had the best photoelectric performance. Its photocurrent density at 1.23 V_RHE was 1.95 mA cm⁻², and its on-potential was 0.78 V_RHE.

Fig. 13 (a) LSV curve of the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes obtained after immersion in phosphate for various time of 3, 6, 9, and 12 h and (b) current–time curve under chopped wave, and the dotted line represents the current in the dark state density.

Electrochemical results

In order to better understand the nature of the interface between the electrode and the electrolyte, we conducted electrochemical impedance tests on the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes. The EIS Nyquist diagram of Fe₂O₃ calcined at various temperatures of 500, 525, 550, 575, and 600 °C is shown in Fig. S12 (ESI†). When the calcination temperature increased from 500 °C to 575 °C, the radius of the arc in the high-frequency region of the Nernst diagram gradually decreased. It indicated that the charge transfer resistance decreased. When the calcination temperature continued to rise to 600 °C, the charge transfer resistance increased instead. This was mainly due to the fact that too high temperature would destroy the SnO₂ layer on the surface of the conductive substrate, thereby reducing the conductivity of the substrate. Fig. S13 (ESI†) is the digital photo of bare Fe₂O₃ calcined at 600 °C. Due to the high calcination temperature, the FTO substrate in the picture had been slightly deformed. Combining the SEM analysis, it could be seen that when the calcination temperature increased from 500 °C to 600 °C, the film thickness increased by 110 nm. This increased the thickness of the electric double layer, resulting in an increase in resistance during the charge transfer process. The EIS Nyquist plot for the Fe₂O₃/Co–Pi photoanode after immersion in phosphate for various time of 3, 6, 9, and 12 h is shown in Fig. S14 (ESI†). The figure showed that the Co–Pi modified sample had a reduced arc radius in the high frequency region of the Nernst plot. It indicated that depositing the right amounts of Co–Pi groups would reduce the charge transfer resistance.

Fig. 14a is the EIS spectrum of the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes. The curve in the figure showed that the radius of the arc of Fe₂O₃/Co–Pi in the high frequency region was smaller than that of bare Fe₂O₃. It indicated that the deposition of CoPi changed the electrode/electrolyte interface performance. After the holes in the VB reach the surface of Fe₂O₃, they were extracted by the CoPi covered on the surface. During the subsequent catalytic water oxidation, CoPi undergoes proton-coupled electron transfers along with cyclic valency changes between Co(II/III) and Co(III/IV). As an effective hole extractor and redox mediator, CoPi could inhibit the recombination of photo-generated carriers at the semiconductor/electrolyte interface and improve the charge transport capability.³⁴Fig. 14b is the Mott–Schottky plots of the bare Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes, which showed that the slopes of all samples were positive. It inferred that the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes were both n-type semiconductors and the majority carriers were electrons. The carrier density of the Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes could be calculated by the slope of the M–S curve, which were 1.03 × 10¹⁷ cm⁻³ and 3.48 × 10¹⁶ cm⁻³, respectively. The data indicated that the modification of the Co–Pi groups did not increase the carrier density.

Fig. 14 (a) EIS Nyquist plots of bare Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes measured at a bias of 1.23 V_RHE under simulated solar light illumination, (b) Mott–Schottky plots of bare Fe₂O₃ and Fe₂O₃/Co–Pi photoanodes.

Mechanism analysis

Fig. 15 is the schematic diagram of the mechanism of Co–Pi enhancing the photoelectric performance of the Fe₂O₃ photoanode. According to the transient I–t curve, the transient anodic spike of Fe₂O₃ decreased after CoPi deposition. This indicated that CoPi extracts holes on the surface of Fe₂O₃, thereby inhibiting the accumulation of holes on the electrode surface. The holes transferred to the promoter layer had strong oxidizing ability. This made the Co element in the low-valence state (II/III) oxidized to the high-valence state (III/IV). Then, the high-valence Co got electrons from the water and was re-reduced to the low-valence Co. In the above process, CoPi acted as a redox mediator. According to the LSV curve, after CoPi deposition, the initial potential of Fe₂O₃ moved negatively from 0.85 V_RHE to 0.78 V_RHE. According to the law of conservation of energy and the Arrhenius equation, the following conclusions could be derived. CoPi deposited on the Fe₂O₃ electrode could reduce the E_AOP of PEC water oxidation, thereby increasing the rate of OER. CoPi not only acted as a hole extractor and redox mediator, but also a common co-catalyst to promote OER. Through the synergistic effect of CoPi, the PEC performance of Fe₂O₃ photoanode was improved. The ABPE of Fe₂O₃/Co–Pi photoanode was 1.51% (4.87 times that of Fe₂O₃). The η_inj of Fe₂O₃/Co–Pi was 66.67%, which was 1.6 times that of Fe₂O₃.

Fig. 15 Schematic diagram of the mechanism of Co–Pi enhancing the photoelectric performance of the Fe₂O₃ photoanode.

Conclusions

In this paper, a hydrothermal process was used to prepare bare α-Fe₂O₃, and Co-Pi groups were deposited on the surface of the Fe₂O₃ photoanode through a two-step dipping process. The effects of the hydrothermal time, concentration of the Fe precursor, annealing temperature, as well as loading of Co–Pi on the performance of the α-Fe₂O₃ photoanode were studied. SEM showed that Fe₂O₃ is composed of a short nanorod, and the thickness of the Fe₂O₃ layer increased with the increase of the calcination temperature. There were no peaks of Co–Pi groups in the XRD of Fe₂O₃/CoPi, and there were no lattice fringes of Co–Pi groups in the TEM of Fe₂O₃/CoPi. The Co and P elements signals in XPS indicated that the Co–Pi groups were indeed deposited on the surface of Fe₂O₃. These measurements showed that the Co–Pi groups on the Fe₂O₃/CoPi photoanode were amorphous. When the Fe precursor concentration was 6 mM, the hydrothermal reaction time was 4 h, and the calcination temperature was 600 °C, the bare Fe₂O₃ photoanode had the best photoelectric performance. The photocurrent density of the optimized sample at 1.23 V_RHE was 1.02 mA cm⁻², which was 29 times that of the sample without conditional exploration. The photocurrent density of Fe₂O₃/Co–Pi at 1.23 V_RHE was 1.95 mA cm⁻². CoPi made the initial potential of Fe₂O₃ negatively shifted by 70 mV. According to the law of conservation of energy, it could be deduced that CoPi acted as an effective co-catalyst to reduce the activation energy of OER. The transient J–t curve showed that the transient anodic spike of Fe₂O₃ decreased after CoPi deposition. It indicated that the charge accumulation on the surface of Fe₂O₃/CoPi was reduced. EIS showed that the electrode/electrolyte interface resistance of Fe₂O₃/CoPi decreased. It showed that CoPi could change the interface performance. Combined with the transient I–t curve and EIS analysis, it could be obtained that CoPi could be used as a hole extractant and a redox mediator to promote charge injection. Through the synergy of CoPi, the ABPE of the Fe₂O₃/Co–Pi photoanode was 0.17% (2.1 times that of bare Fe₂O₃). The η_inj of the Fe₂O₃/Co–Pi photoanode at 1.23 V_RHE was 1.6 times that of bare Fe₂O₃. It was hoped that this work could provide more insights into the synthesis of the Fe₂O₃ photoanodes, and the role of CoPi on Fe₂O₃ and other semiconductor photoanodes.

Conflicts of interest

The authors declare no competing financial interest.

Acknowledgements

This work was financially supported by the National Natural Science Foundations of China (No. 21805040/B0605), the Natural Science Foundation of Fujian Province of China (No. 2018J05015), and the Foundation of Fujian Educational Committee (No. JAT170131). We gratefully acknowledge the help of the Dalian Institute of Chemical Physics, Chinese Academy of Sciences.

References

1. Y. Zhao, C. Ding, J. Zhu, W. Qin, X. Tao, F. Fan, R. Li and C. Li, Angew. Chem., Int. Ed., 2020, 59, 9653–9658.
2. C. Xu, W. Sun, Y. Dong, C. Dong, Q. Hu, B. Maa and Y. Ding, J. Mater. Chem. A, 2020, 8, 4062–4072.
3. X. Chai, H. Zhang, Q. Pan, J. Bian, Z. Chen and C. Cheng, Appl. Surf. Sci., 2019, 470, 668–676.
4. F. Feng, C. Li, J. Jian, X. Qiao, H. Wang and L. Jia, Chem. Eng. J., 2019, 368, 959–967.
5. P. S. Basavarajappa, S. B. Patil, N. Ganganagappa, K. R. Reddy, A. V. Raghu and C. V. Reddy, Int. J. Hydrogen Energy, 2020, 45, 7764–7778.
6. S. Bai, H. Chu, X. Xiang, R. Luo, J. He and A. Chen, Chem. Eng. J., 2018, 350, 148–156.
7. Q. Sun, T. Cheng, Z. Liu and L. Qi, Appl. Catal., B, 2020, 277, 119189.
8. M. Zhao, T. Chen, B. He and X. Hu, J. Mater. Chem. A, 2020, 8, 15976–15983.
9. L. Yu, Y. Zhang, J. He, H. Zhu, X. Zhou, M. Li, Q. Yang and F. Xu, J. Alloys Compd., 2018, 753, 601–606.
10. R. L. Spray and K. S. Choi, Chem. Mater., 2009, 21, 3701–3709.
11. X. Wang, Y. Tuo, Y. Zhou, D. Wang, S. Wang and J. Zhang, Chem. Eng. J., 2021, 403, 126297.
12. J. Wang, N. H. Perry, L. Guo, L. Vayssieres and H. L. Tuller, ACS Appl. Mater. Interfaces, 2019, 11, 2031–2041.
13. R. Abe, J. Photochem. Photobiol., C, 2010, 11, 179–209.
14. K. Sivula, F. F. Le and M. Graetzel, ChemSusChem, 2011, 4, 432–449.
15. J. Li, J. Li, H. Yuan, W. Zhang, Z. Jiao and X. S. Zhao, Chem. Eng. J, 2020, 398, 125662.
16. S. D. Tilley, M. Cornuz, K. Sivula and M. Gratzel, Angew. Chem., 2010, 49, 6405–6408.
17. K. Maeda, K. Teramura, D. Lu, N. Saito, Y. Inoue and K. Domen, Angew. Chem., 2006, 45, 7806–7809.
18. B. Klahr, S. Gimenez, F. Fabregat-Santiago, J. Bisquert and T. W. Hamann, J. Am. Chem. Soc., 2012, 134, 16693–16700.
19. J. Y. Kim, G. Magesh, D. H. Youn, J. W. Jang, J. Kubota, K. Domen and J. S. Lee, Sci. Rep., 2013, 3, 2681–2688.
20. Y. Hou, Z. H. Wen, S. M. Cui, X. L. Feng and J. H. Chen, Nano Lett., 2016, 16, 2268–2277.
21. Q. J. Bu, S. Li, K. Zhang, Y. H. Lin, D. J. Wang, X. X. Zou and T. F. Xie, ACS Sustainable Chem. Eng., 2019, 7, 10971–10978.
22. J. M. Feckl, H. K. Dunn, P. M. Zehetmaier, A. Müller, S. R. Pendlebury, P. Zeller, K. Fominykh, I. Kondofersky, M. Döblinger, J. R. Durrant, C. Scheu, L. Peter, D. Fattakhova-Rohlfing and T. Bein, Adv. Mater. Interfaces, 2015, 2, 1500358.
23. F. Li, Z. Zhao, H. Yang, D. Zhou, Y. Zhao, Y. Li, W. Li, X. Wu, P. Zhang and L. Sun, Dalton Trans., 2020, 49, 588–592.
24. C. Du, X. G. Yang, M. T. Mayer, H. Hoyt, J. Xie, G. McMahon, G. Bischoping and D. W. Wang, Angew. Chem., Int. Ed., 2013, 152, 12924–12927.
25. Q. Yu, X. Meng, T. Wang, P. Li and J. Ye, Adv. Funct. Mater., 2018, 25, 2686–2692.
26. C. H. Liu, Y. Xu, H. Luo, W. C. Wang, Q. Liang and Z. D. Chen, Chem. Eng. J., 2019, 363, 23–32.
27. T. Wang, X. Long, S. Wei, P. Wang, C. Wang, J. Jin and G. Hu, ACS Appl. Mater. Interfaces, 2020, 12, 49705–49712.
28. S. N. S. Moraes, D. L. Morgado and M. Nalin, Vib. Spectrosc., 2019, 100, 57–63.
29. L. Hu, X. Zeng, X. Wei, H. Wang, Y. Wu, W. Gu, L. Shi and C. Zhu, Appl. Catal., B, 2020, 273, 119014.
30. C. G. Shull, W. A. Strauser and E. O. Wollan, Phys. Rev., 1951, 83, 333–345.
31. C. H. Li, C. L. Huang, X. F. Chuah, D. S. Raja, C. T. Hsieh and S. Y. Lu, Chem. Eng. J., 2019, 361, 660–670.
32. L. Yu, H. Zhou, J. Sun, F. Qin, F. Yu, J. Bao, Y. Yu, S. Chen and Z. Ren, Energy Environ. Sci., 2017, 10, 1820–1827.
33. N. M. Randall, J. Qiu and F. Laskowski, ACS Energy Lett., 2018, 2286–2291.
34. S. Alam, T. K. Sahu and M. Qureshi, ACS Sustainable Chem. Eng., 2021, 9, 5155–5165.

---

Footnotes

† Electronic supplementary information (ESI) available: Including SEM, TEM and XRD images of Fe₂O₃, the content of each element of Fe₂O₃ and Fe₂O₃/CoPi obtained from EDS and XPS, the ultraviolet-visible absorption spectra, stability test, EIS spectra of Fe₂O₃ and Fe₂O₃/CoPi, as well as LSV curves of Fe₂O₃ and Fe₂O₃/CoPi with Na₂SO₃, etc. Full experiment details and the additional figures are present See DOI: 10.1039/d1nj04516d

‡ Ai Qin and Wanqing Fang contributed equally to this work.

---