The conversion of solar energy to hydrogen fuel via photoelectrochemical (PEC) water splitting is considered to be a potential solution for the current global energy crisis and environmental problems [1]. The overall efficiency of the water splitting process is limited by the water oxidation process, which involves four electrons [2]. Owing to its favorable bandgap of 2.0-2.2 eV, high durability in aqueous and oxidative environments, and high natural abundance, α-Fe₂O₃ (hematite) stands out as one of the most promising water splitting photoanodes [3, 4]. Theoretically, the photocurrent density of Fe₂O₃ photoanodes could exceed 12 mA cm^-2, when assisted by an applied bias of 0.3-0.4 V [5, 6]. However, in practice, the photocurrent density of as-prepared Fe₂O₃ photoanodes is lower than 3 mA cm^-2. This is mostly attributed to the low mobility of carriers (< 0.1 cm² V^-1 s^-1) [7] and the sluggish oxygen evolution reaction (OER) kinetics, which are affected by the surface states [8, 9]. The PEC performance of Fe₂O₃ could be improved by using nanostructured Fe₂O₃ [10, 11] and foreign element doping [12]. However, the onset potential of doped nanostructured Fe₂O₃ photoanodes could still be as high as ~0.9 V [13, 14], which is much higher than that of other photoanode materials, such as BiVO₄ [15] and TiO₂ [16], and represents a major drawback.

The use of OER catalysts for PEC water splitting has been demonstrated to ameliorate this problem by passivating the photoanode surface and/or accelerate the OER kinetics [17-19]. Most Earth-abundant OER catalyst materials comprise transition metal compounds, such as cobalt phosphate (Co-Pi) [20, 21], nickel borate, Ni or Co oxides [22, 23], Ni or Fe oxyhydroxides [24-26], and their complex oxides/oxyhydroxides [27, 28]. Currently, research efforts are focused on achieving synergistic effects using dual catalysts. For example, Co-Pi/Co₃O₄ [29] and F:FeOOH/FeNiOOH [28] were determined to show a synergistic effect on the lowering of the onset potential of Fe₂O₃ photoanodes. In addition Ni(OH)₂/IrO₂ [17] and C dots/Co₃O₄ dual catalysts [30] were reported to present synergistic effects mostly on the photocurrent rather than the onset potential. Hence, the functional selection and assembly methods of the cocatalysts could be very important to simultaneously achieve the decrease in onset potential and increase in photocurrent.

In this study, Co-Pi and FeOOH were successively deposited on nanostructured Fe₂O₃ photoanodes in an attempt to decrease their onset potential and increase their photocurrent. Furthermore, the effects of Co-Pi combined with layered amorphous (d-FeOOH) and granular (β-FeOOH; h-FeOOH) FeOOH on the performance of the photoanodes were different. A strong synergistic effect was observed when the Co-Pi/h-FeOOH dual catalyst was used, whereas the synergistic effect of Co-Pi/d-FeOOH was negligible. The difference in the functional effects of the combinations of Co-Pi with the two types of FeOOH was explained from a kinetic perspective. This study aimed to analyze the effects of the prospective assembly of dual catalysts on lowering the onset potential of Fe₂O₃ photoanodes.

Sodium hydroxide (NaOH, 98%), hydrochloric acid (37%), cobalt nitrate (98%), ferric chloride (97%), ammonium fluoride (98.0%), monopotassium phosphate (98%), and ferrous sulfate heptahydrate (FeSO₄·7H₂O, 98%), were purchased from Sinopharm Chemical Reagent Co., Ltd., and were used as received without further purification. Fluorine-doped tin oxide (FTO)-coated glass (surface resistivity of ~7 Ω sq^-1) was purchased from Huanan Xiangcheng Tech. Co.

Fe₂O₃ and Co-Pi were synthesized using the procedure outlined in one of our previous reports [21]. A Fe₂O₃ sample was used for the hydrothermal synthesis of h-FeOOH/Fe₂O₃, and the reaction time was optimized to 1 h. Conversely, d-FeOOH was coated on Fe₂O₃ via dip-coating, as described in the literature [31]. In brief, a Fe₂O₃ anode was immersed in a 5 mM FeSO₄·7H₂O aqueous solution, it was maintained at 40 ℃ for 10-30 min, and then it was washed and dried. The dip-coating time was optimized to 10 min.

Photoelectrochemical testing was performed using a CHI 660D (CH Instruments) electrochemical workstation with three-electrode configuration which consisted of Hg/HgO (1 M KOH) and a 1 × 3 cm² Pt sheet as the reference and counter electrodes, respectively; 1.0 M NaOH aqueous solution (pH 13.6) was used as the electrolyte [32, 33]. The polarization curves were obtained at 20 mV s^-1 under simulated sunlight illumination obtained using a Xe lamp (PLS-SXE300D, PerfectLight) equipped with an AM 1.5G filter. The light intensity was measured to be 100 mW cm^-2.

The incident photon-to-current conversion efficiencies (IPCEs) at the bias of 1.23 V vs. the reversible hydrogen electrode (RHE) were calculated by changing the bandpass filters of the Xe lamp in the range of 350-700 nm and using the following equation:

(1)

where J is the photocurrent density, λ is the incident light wavelength, and J_light is the measured power density of the incident light (100 mW cm^-2) [34].

The produced H₂ and O₂ were collected and detected using a 6890D (Agilent) gas chromatograph equipped with a thermal conductivity detector and molecular sieve 5A column. N₂ was used as the carrier gas (flow rate of 36 mL min^-1), and the oven and detector temperatures were 50 and 120 ℃, respectively.

All potentials were referenced to the RHE using the Nernst equation [35]:

(2)

The applied bias photo-to-current conversion efficiency (ABPE) was estimated using the following formula [30]:

(3)

Mott-Schottky (MS) experiments were carried out at the interval of 50 mV under 10 kHz in the dark, and the carrier density (N_d) was calculated as follows:

(4)

where C_S is the capacitance, V is the applied bias, q is the unit electron charge, ε₀ is the vacuum permittivity, and ε is the relative permittivity of Fe₂O₃.

The open circuit potential (OCP) decay curves were obtained under light irradiation before the light was turned off. The OCP values under light irradiation and in the dark were recorded when equilibrium was reached after the light was turned on/off.

Potentioelectrochemical impedance spectroscopy (PEIS) was performed in the range of 100 kHz to 0.1 Hz under AM 1.5G illumination at the amplitude of 20 mV by varying the applied bias.

X-ray diffraction (XRD) patterns were obtained using a D8 Advance, (Bruker) diffractometer equipped with a Cu Kα radiation source. Raman spectra were acquired using a FV-CFR-A (Zolix) laser Raman microscope spectrometer with the excitation wavelength of 532 nm. The morphology of the materials was analyzed using a FEI Nova 200 (NanoLab) field emission scanning electron microscopy (FESEM) instrument and a JEM-2100F (JEOL) transmission electron microscopy (TEM) apparatus. An L750 (PerkinElmer) ultraviolet-visible (UV-Vis) spectrometer in diffuse reflection mode was used to acquire the absorption spectra of the samples. Furthermore, the X-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS) profiles of the samples were obtained using an ESCA PHI 1600 (Physical Electronics) system with Al K_α X-ray excitation source and an ESCALAB 250 Xi (Thermo Scientific) device, respectively.

The crystal phase structure of the electrodes is investigated using XRD and the patterns are compared, as shown in Fig. 1a. The strong peaks of SnO₂ originated from the FTO substrate, and the peaks of the synthesized FeOOH were indexed to the tetragonal β-FeOOH phase (akaganeite-Q, JCPDS No. 34-1266). The peaks of calcined β-FeOOH were indexed to the hexagonal α-Fe₂O₃ (hematite, JCPDS No. 33-0664) phase with (110) preferred orientation. The peaks of α-Fe₂O₃ dominated the XRD patterns of the samples that contained FeOOH and/or Co-Pi. The formation of β-FeOOH was confirmed by the presence of the enlarged peak in the XRD profile of h-FeOOH/Fe₂O₃, which was ascribed to the (301) plane of β-FeOOH (Fig. 1b). However, the characteristic peaks of FeOOH were absent from the XRD profile of the d-FeOOH sample, which indicated that only an insignificantly small amount of FeOOH was present in this sample or that FeOOH formed an amorphous layer. Raman spectroscopy was used to analyze the surface composition of the electrodes more in depth. The typical Raman peaks of FeOOH were present in the spectra of both FeOOH/Fe₂O₃ samples, which confirmed the successful deposition of FeOOH on Fe₂O₃ (Fig. 1c). [36]. The optical properties of the samples were analyzed using their UV-Vis diffuse reflection spectra (Fig. 1d). The absorption edges of β-FeOOH (550 nm) and Fe₂O₃ (600 nm) indicated that their bandgaps were 2.35 and 2.15 eV respectively, as illustrated in the Tauc plots (Fig. S1). The absorption of d-FeOOH was negligible, whereas the light absorption of the h-FeOOH/Fe₂O₃ composite sample slightly increased at 550 nm.

As observed from the SEM images (Fig. S2), the as-prepared Fe₂O₃ shows the same coralline-like nanostructure we have previously reported [14, 19]. No apparent changes in surface morphology were identified after d-FeOOH or h-FeOOH were loaded on the surface of Fe₂O₃, and this was likely attributed to the small loading amounts. Co-Pi appeared to fill the interspaces of the nanostructures, which resulted in the rough surfaces of the samples. The high-resolution TEM images of the bulk d-FeOOH/Fe₂O₃, h-FeOOH/Fe₂O₃, Co-Pi/d-FeOOH/Fe₂O₃, and Co-Pi/h-FeOOH/Fe₂O₃ (Fig. 2) revealed that the interplanar fringe distances were 0.368 and 0.270 nm, which corresponded to the (012) and (104) planes of Fe₂O₃, respectively. The surface of bare Fe₂O₃ is also presented in Fig. 2c. The surface of Fe₂O₃ was covered with FeOOH for both FeOOH/Fe₂O₃ samples; however, the surface morphologies of the FeOOH/Fe₂O₃ samples were different: d-FeOOH formed a 3-4 nm thick uniform amorphous layer on the surface of Fe₂O₃ (Fig. 2a), and the appearance of the h-FeOOH deposited on the surface of Fe₂O₃ was granular (Fig. 2b). The lattice fringe spacing of h-FeOOH was 0.333 nm, which coincided with that of the (310) plane of β-FeOOH. In addition, both the Co-Pi/d-FeOOH/Fe₂O₃ and Co-Pi/h-FeOOH/Fe₂O₃ samples presented ~5 nm thick Co-Pi amorphous layers (Figs. 2d and e). The energy-dispersive X-ray spectroscopy maps and Co and P XPS profiles of the Co-Pi/d-FeOOH/Fe₂O₃ sample confirmed the presence of Co-Pi (Fig. S3). In addition, it was revealed that Co²⁺ and Co³⁺ ions coexisted in Co-Pi. The electrochemically active surface areas (ECSAs) of Fe₂O₃, d-FeOOH/Fe₂O₃ and h-FeOOH/Fe₂O₃ were estimated according to their electrochemical double-layer capacitances (C_dl) (Fig. S4). The ECSA of the d-FeOOH layer coating on the surface of Fe₂O₃ was thinner than that of the bare Fe₂O₃ sample, but no differences existed between the ECSAs of the bare and h-FeOOH-coated Fe₂O₃ samples, which indicated that the ECSA of the h-FeOOH nanoparticles was larger than that of the layered d-FeOOH.

The polarization (J-V) curves of the bare and modified Fe₂O₃ photoanodes are compared (Fig. 3a), and the turn-on voltage (V_on) is estimated using the traditional linear extrapolation method [37]. The V_on of the bare Fe₂O₃ anode was as high as 0.92 V vs. RHE and its photocurrent density was 0.85 mA cm^-2 at 1.23 V vs. RHE. After optimization, d- and h-FeOOH loadings contribute to shifting V_on by 60 and 90 mV, respectively, toward the cathode side. Moreover, the photocurrent of the h-FeOOH/Fe₂O₃ anode was slightly higher than that of the d-FeOOH/Fe₂O₃ one. The photocurrent densities of the h-FeOOH/Fe₂O₃ and d-FeOOH/Fe₂O₃ anodes obtained at different FeOOH loading amounts are compared (Fig. S5). The photocurrent densities of both anodes increased and then decreased as the loading amount of FeOOH increased. Owing to the effect of Co-Pi on the catalytic performance of the Fe₂O₃ anode, V_on decreased by 110 mV. More importantly, when FeOOH and Co-Pi were co-loaded on Fe₂O₃, V_on could be further reduced. However, the PEC performance of the Co-Pi/h-FeOOH/Fe₂O₃ anode was superior compared to that of the Co-Pi/d-FeOOH/Fe₂O₃ one. For the Co-Pi/h-FeOOH/Fe₂O₃ anode, V_on was as low as 0.65 V and the photocurrent reached 1.31 mA cm^-2 at 1.23 V, and therefore its V_on shifted by 270 mV and its photocurrent increased by 54% compared to those of the bare Fe₂O₃ anode. In addition, in the dark, the currents of all electrodes were constant and their values were almost the same. The derived maximum ABPE of Fe₂O₃ was 0.07%, and that increased to 0.10% and 0.12% after d- and h-FeOOH coating, respectively (Fig. S6). The maximum ABPE values of the Co-Pi/Fe₂O₃ and h-FeOOH/Fe₂O₃ anodes were the same but at a smaller applied bias. The maximum ABPE values of the Co-Pi/d-FeOOH/Fe₂O₃ and Co-Pi/h-FeOOH/Fe₂O₃ anodes were 0.13% and 0.20% at 1.0 and 0.94 V, respectively. The IPCE values are in agreement with the measured photocurrents (Fig. 3b). The dual catalyst-loaded photoanodes realized higher IPCE values than the FeOOH or Co-Pi loaded photoanodes. The long-term photocurrents of the Fe₂O₃, h-FeOOH/Fe₂O₃, and Co-Pi/h-FeOOH/Fe₂O₃ photoanodes were measured to investigate their stability (Fig. 3c). While the bare Fe₂O₃ and h-FeOOH/Fe₂O₃ photoanodes exhibited favorable stability after 12 h, the current of the Co-Pi/h-FeOOH/Fe₂O₃ photoanode decreased with time, which indicated that Co-Pi negatively affected the stability of the photoanode. However, the Co 2p XPS profile and the SEM and TEM images of the Co-Pi/h-FeOOH/Fe₂O₃ photoanode presented insignificant changes before and after the long-term stability test (Figs. S7a-c), which confirmed the good adhesion of Co-Pi. The stability of the Co-Pi/Fe₂O₃ photoanode was subsequently evaluated (Fig. S7d), and the decrease in its photocurrent was similar to that of the Co-Pi/h-FeOOH/Fe₂O₃ photoanode. However, during the second run of the stability test, which was performed without turning off/on the light or changing the electrolyte, the photocurrent density of the Co-Pi/Fe₂O₃ photoanode fully recovered and then decreased just as it did during the first run. Similar results have been reported in the literature [35]. Therefore, the decrease in the photocurrent of the Co-Pi/h-FeOOH/Fe₂O₃ photoanode should be attributed to the changes in the properties of the as-deposited Co-Pi during the water oxidation reaction. The Faraday efficiencies of O₂ of all analyzed samples were approximately 90% (Fig. S8). The amounts of H₂ and O₂ gases generated during the first 3 h of the steady-state I-t test were quantified and the results are presented in Fig. 3d. In addition, the H₂/O₂ molar ratio was ~2.03, which was larger than the stoichiometric ratio, and should be attributed to the ineluctable O₂ dissolution at normal pressure and temperature. The H₂ and O₂ evolution analysis demonstrated that the photogenerated carriers mostly participated in the water splitting process in the system. The onset characteristic performance of the Co-Pi/h-FeOOH/Fe₂O₃ photoanode synthesized using this strategy was comparable to those of the previously reported state-of-the-art dual catalyst-modified Fe₂O₃ composite photoanodes summarized in Table S1. The higher photocurrent in the report contributes to Ti doping in Fe₂O₃.

The MS plots (C^-2 vs. V) of the electrodes were obtained in the dark. The slope of the MS plot of the Fe₂O₃ electrode shows a positive (Fig. 4a), which is typical for n-type semiconductors. In addition, the donor concentration of the h-FeOOH/Fe₂O₃ electrode was higher than that of the Fe₂O₃ electrode, which was indistinctive for the d-FeOOH/Fe₂O₃ electrode. The slope of the MS plot of β-FeOOH is negative (Fig. 4b), which confirms its previously reported p-type semiconductor behavior [38]. Thus, β-FeOOH and Fe₂O₃ formed a p-n junction, which explained the higher donor concentration in the FeOOH/Fe₂O₃ electrodes. The UPS-determined valence band positions of β-FeOOH and Fe₂O₃ are determined to be 7.19 and 6.88 eV relative to the vacuum level, respectively (Fig. 4c). The band alignment and formation of the p-n junction between Fe₂O₃ and β-FeOOH were determined using their bandgaps and MS plot-derived flat-band potentials, and the results are presented in Fig. 4d. Consequently, the transfer of charge carriers was driven by the built-in electric field between β-FeOOH and Fe₂O₃.

The recombination of the photogenerated electrons in Fe₂O₃ photoanodes is schematically illustrated in Fig. 5a. In addition to the recombination of electrons with the holes in the bulk (Step 1, > 0.001 s) and in the depletion region (Step 2, 0.01-0.1 s), electrons could be trapped by surface trap states (Step 3, seconds to minutes) [39, 40]. The time scale of the normalized transient decay profile of the open-circuit potential (OCP) ranges from 0.01 s to minutes (Fig. 5b), and thus it provides information on the charge recombination in both the depletion region and the surface states (Steps 2 and 3, respectively). Given that the current that resulted from the assumed redox reactions did not pass through, the catalytic and kinetic effects were excluded under the open circuit conditions [41]. As light was switched off, the OCP of bare Fe₂O₃ increases by only 20% within 0.1 s, which indicated that electrons were mostly trapped by the surface states other than being recombined at the surface. For the FeOOH/Fe₂O₃ photoanodes, the decrease in OCP slowed down. The inhibited surface charge recombination that occurred during Step 2 could be explained by the formation of the p-n junction between FeOOH and Fe₂O₃, and the slower Step 3 of h-FeOOH/Fe₂O₃ than that of d-FeOOH/Fe₂O₃ could be caused by the larger electrochemical surface area of h-FeOOH compared to that of d-FeOOH, which hindered charge accumulation. The decrease in the OCP of the Co-Pi-coated anodes was completed within 1 s, which implied that Co-Pi efficiently interfered with the surface states via passivation. During the first second, the OCP of the Co-Pi/d-FeOOH/Fe₂O₃ and Co-Pi/h-FeOOH/Fe₂O₃ anodes decreased faster than those of the d-FeOOH/Fe₂O₃ and h-FeOOH/Fe₂O₃ ones. The enhancement in surface charge recombination partly resulted from the increase in the amount of spatial charges in the depletion region with enlarged band bending [42-44]. The photovoltage (V_ph) of the anodes was calculated as the difference between the equilibrium OCP values under illumination and in the dark (OCP_light and OCP_dark, respectively). Owing to the Fermi level pinning effect, OCP_dark of the Fe₂O₃ anode equaled the peak position of the surface states and it is determined to be 0.85 V in this study (Fig. 5c) [27, 45]. The decrease in the OCP_dark values of the two FeOOH/Fe₂O₃ anodes revealed that FeOOH shifted the surface states toward a higher position. The OCP_dark of the Co-Pi/Fe₂O₃ anode was 1.1 V, which implied that the surface states were incompletely passivated. The addition of FeOOH between Co-Pi and Fe₂O₃ in the Co-Pi/FeOOH/Fe₂O₃ anodes induced a decrease in its OCP_dark that was similar to that observed for the bare Fe₂O₃ anode. However, the V_ph values of the Co-Pi/h-FeOOH/Fe₂O₃ and Co-Pi/d-FeOOH/Fe₂O₃ anodes were higher and lower, respectively, than that of the Co-Pi/Fe₂O₃ one. This indicated that the associated amount of photogenerated electron-hole pairs of the Co-Pi/h-FeOOH/Fe₂O₃ anode was higher than that of the Co-Pi/Fe₂O₃ one.

PEIS is a powerful tool that could be used to elucidate the kinetics of the photoelectrodes under illumination. The typical Nyquist plots and proposed equivalent circuit used to interpret the PEIS data are presented in Fig. S9. The derived bulk capacitance (C_bulk) and resistance (R_bulk) as well as the trapping capacitance (C_trap) and resistance (R_{ct, trap}) were plotted as functions of the applied potential. The C_bulk and R_{ct, bulk} values (Fig. S10) mostly depended on the bulk Fe₂O₃, and were insignificantly affected by the FeOOH or Co-Pi coatings. However, C_trap and R_{ct, trap} varied significantly (Fig. 6). The C_trap plot of the bare Fe₂O₃ electrode presented convex shape and C_trap peaked at 1.1 V. In the voltage range of 0.65-0.8 V, the C_trap values of d-FeOOH/Fe₂O₃ were slightly higher than those of bare Fe₂O₃, whereas the C_trap values of h-FeOOH/Fe₂O₃ increased more significantly. Furthermore, the C_trap values at potentials that exceeded 0.8 V were constant for both d-FeOOH/Fe₂O₃ and h-FeOOH/Fe₂O₃. Fe₂O₃ presented both OH- and O-terminated surface states and the potential of the O-terminated surface states was higher than that of the OH-terminated ones [46]. Accordingly, it was concluded that the O-terminated surface states of bare Fe₂O₃ behave dominantly, and FeOOH introduced more OH-terminated surface states. The h-FeOOH nanoparticles induced higher C_trap values than the d-FeOOH layer probably owing to h-FeOOH inducing a larger amount of photogenerated holes than d-FeOOH did. The decrease in the C_trap values of Co-Pi/Fe₂O₃ at potentials exceeding 0.9 V reflected the passivated surface states. Furthermore, the C_trap of Co-Pi/Fe₂O₃ peaked at 0.65 V. The co-loading of d-FeOOH and Co-Pi did not affect this peak C_trap value; however, the co-loading of h-FeOOH and Co-Pi significantly increased the C_trap values, which indicated that the Co-Pi/h-FeOOH/Fe₂O₃ presented higher hole capture ability than Co-Pi/d-FeOOH/Fe₂O₃. The R_{ct, trap} values of bare Fe₂O₃ decreased as the potential increased and were significantly reduced when FeOOH and Co-Pi were co-loaded on Fe₂O₃ (Fig. 6b). The R_{ct, trap} values in the potential range of 0.65-0.8 V decreased as follows: bare Fe₂O₃ > d-FeOOH/Fe₂O₃ > h-FeOOH/Fe₂O₃ > Co-Pi/d-FeOOH/Fe₂O₃ > Co-Pi/h-FeOOH//Fe₂O₃.

Ponomarev and Peter used charge transfer/recombination processes, and proposed a phenomenological model for estimating the kinetic rate constants [47]. The model had been previously used to estimate the kinetic parameters of the Fe₂O₃ electrode assuming that the capacitance of the semiconductor was much smaller than that across the Helmholtz layer [48, 49]. The charge transfer rate (k_ct) only depended on the time constant of the surface trapping states, and could be expressed as follows:

(5)

The relationship between the charge recombination rate (k_rec) and k_ct could be written as:

(6)

The charge transfer efficiency (Φ_ct) could be calculated as the charge transfer distribution ratio:

(7)

The dependence of the calculated kinetic parameters on the potential is depicted in Fig. 7. As the potential increased from 0.5 to 1.1 V, the k_ct value of bare Fe₂O₃ increased by two orders of magnitude (Fig. 7a). This dependence of k_ct on the applied potential is typical for Fermi level pinning [49, 50]. Both d- and h-FeOOH contributed to the increase in k_ct. However, d-FeOOH only increased k_ct at biases lower than 0.9 V whereas h-FeOOH prompted the permanent increase in k_ct at all recorded applied biases. Both d- and h-FeOOH could accelerate the charge transfer kinetics; however, their mechanisms might be different. A similar increase in k_ct was reported for a FeOOH-covered Ti:Fe₂O₃ anode [51]. For the Co-Pi-coated photoanodes, d-FeOOH co-loading led to a significant increase in k_ct; however, the change in k_ct associated with the co-loading of h-FeOOH and Co-Pi was negligible. The initial charge recombination rate of bare Fe₂O₃ was higher than its charge transfer rate, which prevented the oxidation of water (Fig. 7b). Both d- and h-FeOOH co-loading led to the slight decrease in k_rec with no impact on the variation tendency. The changes in k_rec of Co-Pi/Fe₂O₃ were flattened and at potential lower than 0.95 V k_rec is reduced. The less dependency of potential is caused by surface states passivation and the reduction in charge recombination at less positive potentials is partly owing to the expanded depletion region (Fig. 5c). The k_rec of the Co-Pi/d-FeOOH/Fe₂O₃ photoanode was not significantly different than that of d-FeOOH/Fe₂O₃ By contrast, the k_rec of the Co-Pi/h-FeOOH/Fe₂O₃ anode was significantly lower than that of h-FeOOH/Fe₂O₃ Therefore, Co-Pi worked synergistically with d-FeOOH to slightly accelerate the charge transfer, but its combination with h-FeOOH could significantly delay the charge recombination. The calculated Φ_ct values and J-V results are compared, as shown in Fig. 7c. The Φ_ct values of all samples were consistent with the J-V results, and thus justified the applicability of this model for the photoanodes in this study.

The charge transfer and recombination processes for the bare and hybrid Fe₂O₃ photoanodes are schematically illustrated in Scheme 1. The holes photogenerated in Fe₂O₃ transferred from the O-terminated surface state (SS1) participated in the water oxidation reaction (Scheme 1a). The locations of the surface states determined the onset potential [52, 53]. Moreover, significant charge recombination occurred at the surface states and limited the maximum photocurrent. OH-terminated surface states (SS2) were present in the d-FeOOH-loaded Fe₂O₃ photoanode (Scheme 1b), and hole transfer and recombination occurred at both SS1 and SS2. The potential of SS2 was higher than that of SS1, and thus it allowed for a lower onset potential. Furthermore, owing to the formation of the p-n junction for the h-FeOOH/Fe₂O₃ photoanode (Scheme 1c), the transfer of the photogenerated electrons and holes between h-FeOOH and Fe₂O₃, and consequently the bulk recombination, was inhibited, which explained the photocurrent density of h-FeOOH/Fe₂O₃ was slightly higher than that of d-FeOOH/Fe₂O₃. Most surface states of the Co-Pi coated Fe₂O₃ were passivated (Scheme 8d), and therefore the pinned quasi-Fermi level (E_{f, p}) was relieved. The remaining surface states were not considered for simplification. Thus, the holes transferred directly from the valence band to Co-Pi and participate in the oxidation of water via several intermediate processes that involved Co⁴⁺ ions [54]. In addition, the electrons in the conduction band recombined with the Co species of Co-Pi, and thus, inhibited the catalytic effect [50]. Upon the co-loading of d-FeOOH and Co-Pi, the amorphous d-FeOOH acted more like a hole-transfer media and an electron-blocking layer, and thus, accelerated the charge transfer to some extent. By contrast, the holes in the valence band of Fe₂O₃ migrated toward β-FeOOH under the force of the electric field of the p-n junction, and eventually transferred to Co-Pi together with those photogenerated in β-FeOOH. Consequently, the amount of accumulated surface electrons was reduced and the charge recombination rate was lowered.

In this study, nanostructured Fe₂O₃ was synthesized hydrothermally and used as a benchmark photoanode for PEC water oxidation. The co-loading of Co-Pi and hydrothermally synthesized h-FeOOH nanoparticles was demonstrated to present synergistic effect on the lowering of the onset potential and enhancement of the photocurrent of the Fe₂O₃ photoanode. By contrast, the synergy between Co-Pi and amorphous FeOOH, which was synthesized via dip-coating, was negligible. The onset potentials of the Co-Pi/h-FeOOH and Co-Pi/d-FeOOH dual catalysts cathodically shifted by 270 and 170 mV, respectively. In addition, the photocurrent density of the Co-Pi/h-FeOOH/Fe₂O₃ anode reached 1.31 mA cm^-2 at 1.23 V vs. RHE; moreover, the onset potential of this anode was as low as 0.65 V vs. RHE, which was comparable to those of the previously reported dual catalyst-modified Fe₂O₃ composite photoanodes. The kinetics of charge transfer and recombination processes were investigated using photoelectrochemical and impedance measurements. When d-FeOOH was added between Co-Pi/Fe₂O₃, the charge transfer was slightly accelerated, but the charge recombination rate was barely affected, which implied that the amorphous d-FeOOH acted as a hole-transfer medium and weakly promoted the charge transfer kinetics. By contrast, the charge recombination of the Co-Pi/ h-FeOOH/Fe₂O₃ photoanode was significantly inhibited, which was attributed to the reduced accumulation of surface electrons owing to the formation of the p-n junction between the p-type β-FeOOH and n-type Fe₂O₃ semiconductors. Therefore, the addition of a full-contact layered and granular p-type cocatalyst to photoanode materials could be a suitable strategy to obtain photoanodes with excellent PEC performance.

Supplementary data associated with this article can be found in the online version.