NiFe Prussian blue analog cocatalyzed TiO₂/In₂S₃ type-II heterojunction for solar water splitting

Preparation of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP

All reagents were purchased from Aladdin and were used as received without further purification. Deionized water (DI, ~18.25 MΩ cm) was used during the experiments.

Preparation of TiO₂: The fabrication of TiO₂ nanorod arrays was accomplished using a reported hydrothermal method with some modifications^[36]. Fluorine-doped tin dioxide conductive glass substrates (FTO, 2.5 cm × 1 cm, sheet resistance < 10 Ω sq^-1) were ultrasonically cleaned in acetone, ethanol and DI sequentially for 30 min each. FTO glasses were dried with a N₂ gas flow before use. To prepare the precursor, 20 mL of the concentrated HCl (38%) was dropped into 20 mL DI water. Then, 668 μL of titanium isopropoxide was infused into the above solution. The resulting solution was subsequently transferred into a 50 mL Teflon-lined stainless steel autoclave. FTO substrates were gently placed in the autoclave at an angle, ensuring that its conductive sides faced downward. Hydrothermal reactions were performed at 170 °C for 6 h. Following this, the samples underwent annealing in air at 350 °C for 1 h, with a heating rate of 5 °C min^-1.

Preparation of TiO₂/In₂S₃: To obtain the TiO₂/In₂S₃ heterojunction, In₂S₃ was hydrothermally grown on the as-prepared TiO₂. The indium nitrate and thioacetamide were dissolved into the DI water of 40 mL with a stoichiometric ratio of 1:3. The amount of indium nitrate was set at 0.05, 0.1, 0.2, and 0.3 millimoles. TiO₂ was carefully placed in an autoclave with the film surface facing downward. Subsequently, the autoclave was subjected to a hydrothermal procedure at 180 °C for 12 h. After the autoclave had cooled to room temperature, the samples were taken out and washed with DI water. The samples were then transferred to a tube furnace, where they underwent annealing at 350 °C for 30 min under a nitrogen atmosphere. Based on the amount of indium nitrate used in the precursor, the TiO₂/In₂S₃ samples are labeled as TiO₂/In₂S₃-x.

Preparation of TiO₂/In₂S₃/NFP: A layer of NiFe Prussian blue (NFP) was deposited on the surface of TiO₂/In₂S₃ via a chemical bath method^[24]. Firstly, 0.02 M Ni(NO₃)₂·6H₂O and 0.02 M K₃Fe(CN)₆ were mixed under vigorous stirring. After that, TiO₂/In₂S₃ was immersed in the above solution. The optimal temperature for NFP growth is 60 °C. The optimal reaction time is 3 h. After the growth of NFP, the samples were washed with DI water and then dried overnight in a 60 °C oven.

PEC performance measurements

PEC performance tests were performed using a three-electrode system with an electrochemical workstation (CHI 660E). TiO₂/In₂S₃/NFP photoanodes were used as the working electrode; an Ag/AgCl electrode was used as the reference electrode, and a platinum wire was used as the counter electrode. The photoanodes were irradiated with the AM1.5G solar spectrum with an intensity of 100 mW cm^-2 (CELHXF 300). Additionally, 1 M KOH aqueous solution (pH = 13.6) was employed as the electrolyte. Linear sweep voltammetry (LSV) measurements were performed at a scan rate of 20 mV s^-1 with a potential range of 0-1.4 Vvs. RHE. The PEC impedance spectroscopy (PEIS) was recorded at a potential of 0.47 V vs. RHE with a perturbation of 10 mV under AM1.5 illumination. The frequency range is from 0.1 to 10⁵ Hz. The obtained PEIS plots were fitted using Z-View software. Mott-Schottky (M-S) measurements were conducted in the dark at a fixed frequency of 2,000 Hz. Donor density (N_D) and flat-band potential (V_fb) can be obtained from the M-S equation, denoted as:

In this study, the variables are defined as follows: A (electrode area), C (space charge capacitance), V (the applied potential), ε (dielectric constant with a value of 8.85 × 10^-12 F m^-1), ε_r (vacuum dielectric constant, ε_r = 100), e (elementary charge), K_b (Boltzmann constant with a value of 1.38 × 10^-23 J K^-1), and T (temperature).

To evaluate the charge separation efficiency (η_sep) and charge injection efficiency (η_inj), a mixed electrolyte comprising 1 M KOH and 0.5 M Na₂SO₃ was employed. The calculation formulas for η_inj and η_sep are established as:

The current density of $$J_{{\mathrm{H}}_2\mathrm{O}}$$ was measured in 1 M KOH, while the current density of $$J_{{\mathrm{Na}}_2{\mathrm{SO}}_3}$$ was measured in the presence of a sacrificial agent containing 0.5 M Na₂SO₃. J_abs represents the theoretical photocurrent density calculated based on light absorption, which is calculated based on the UV-vis absorption spectra of the photoanodes.

The transformation from the measured potential (E_Ag/AgCl) to the potential versus reversible hydrogen electrode (E_RHE) is dictated by:

The electrochemical active surface area (ECSA) is typically quantified using the double-layer capacitance measured by cyclic voltammetry (CV). The magnitude of C_dl is calculated to assess the ECSA of the photoanodes. The formula for calculating ECSA is provided as:

Where C_dl represents the electrochemical double-layer capacitance, ΔJ is the recorded current density variation during the potential scan, and ΔE is the corresponding potential change. C_s signifies the specific capacitance, which is considered constant within the same electrolyte.

The transient decay time (τ) is calculated by:

where I_t represents the photocurrent at time t, I_s denotes the photocurrent at steady state, and I_in signifies the photocurrent of the anodic spike. The time at which lnD = -1 is designated as the transient decay time (τ).

Materials characterizations

The crystal structure was investigated using an X-ray diffractometer (XRD, Rigaku Ultima IV, Japan) with a Cu K_α radiation source. Field-emission scanning electron microscopy (FESEM, thermo scientific Apreo 2C, Thermo Fisher) and energy-dispersive X-ray spectroscopy (EDS) were employed for morphological observations and compositional analysis of the prepared photoanodes. Transmission electron microscopy (TEM) was used to obtain atomic-resolution images (TEM, Talos F200S G2 instrument, USA). UV-vis absorption spectra of TiO₂-based photoanodes were recorded in the wavelength range of 300-800 nm (Shimadzu UV-3600 Plus), while the UV-vis absorption spectra of In₂S₃ powder samples were measured on Lambda 1050 UV-Vis-NIR spectrophotometer. Surface elemental information of the samples was obtained using X-ray photoelectron spectroscopy (XPS, Escalab Xi⁺, Thermo Fisher). Ultraviolet photoelectron spectroscopy (UPS, Escalab Xi⁺, Thermo Fisher) and He I (21.22 eV) photon source were employed to obtain information about binding energies. The UPS spectra were analyzed to derive the valence bands (VB) and work functions of the photoanodes. Photoluminescence spectra (PL, Edinburgh FLS980) were obtained under ambient and unbiased conditions with an excitation wavelength of 425 nm.

Properties of TiO₂/In₂S₃/NFP

The morphology of TiO₂-based photoanodes was characterized first. Figure 1A-C shows the top-view SEM images of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP. Figure 1D-F displays the cross-sectional SEM images correspondingly. As shown in Figure 1A and D, we successfully synthesized ordered, vertically aligned TiO₂ nanorod arrays on FTO substrates, with a diameter of approximately 0.15 μm and a length of 2.3 μm. In₂S₃ nanosheets assembled nanostructures can be clearly observed on the surface of TiO₂ nanorods in Figure 1B and E. Cross-sectional EDS mappings of TiO₂/In₂S₃ in Supplementary Figure 1A and B indicate that the coating of In₂S₃ is uniform in the whole TiO₂ film. After the growth of NFP, the surface of TiO₂/In₂S₃ is completely covered with NFP nanoparticles [Figure 1C and F]. As illustrated in Supplementary Figure 1C and D, Fe, Ni, C, and N elements are evenly distributed within the film thickness. EDS pattern of TiO₂/In₂S₃/NFP is shown in Supplementary Figure 2. The mass percentages of TiO₂, In₂S₃, and NFP calculated roughly from EDS pattern are ~70.13%, ~2.86%, and ~2.78%, respectively. Figure 1G and H presents TEM images of TiO₂/In₂S₃ and TiO₂/In₂S₃/NFP, respectively. In₂S₃ exhibits a nanosheet-like structure growing on the surface of TiO₂ nanorods, while NFP nanoparticles appear to cover the TiO₂/In₂S₃ heterojunction. Some sheet-like In₂S₃ nanostructures can also be observed in Figure 1H, suggesting that the growth of NFP did not significantly affect the structure of In₂S₃. In Figure 1I, a high-resolution TEM (HRTEM) image of TiO₂/In₂S₃/NFP shows the formation of heterojunction among TiO₂, In₂S₃, and NFP. Moreover, NFP nanoparticles exhibit a low crystallinity. Figure 1J and K presents an HRTEM image and selected area electron diffraction (SAED) on the TiO₂ region. The interplanar spacings are measured as d₍₁₁₀₎ = 0.322 nm and d₍₁₀₁₎ = 0.237 nm, corresponding to the tetragonal rutile phase of TiO₂. Figure 1L and M exhibits HRTEM and SAED on the region of In₂S₃. The lattice spacings of d₍₁₀₃₎ = 0.621 nm and d₍₁₀₉₎ = 0.33 nm illustrate that the synthesized In₂S₃ is β-In₂S₃ with a tetragonal phase. β-In₂S₃ has a defective spinel structure with the In³⁺ vacancies, which leads to the n-type characteristics of In₂S₃^[35].

Figure 1. Morphology characterizations of TiO₂-based photoanodes. (A-C) SEM images of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP. (D-F) Cross-sectional SEM images of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP. (G) TEM image of TiO₂/In₂S₃. (H) TEM image of TiO₂/In₂S₃/NFP. (I) HRTEM image of TiO₂/In₂S₃/NFP. (J and K) HRTEM and SAED of TiO₂ region. (L and M) HRTEM and SAED of In₂S₃ region.

XRD patterns shown in Figure 2A reveal the crystal structure of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP. Some XRD peaks from the FTO substrate are marked with stars. The XRD peaks from the TiO₂ film can be well attributed to the rutile phase of TiO₂ (JCPDS No. 21-1276). The XRD peaks at 14.25, 27.43, 33.23, and 47.7° can be attributed to the (103), (109), (0012), and (2212) crystal planes of β-In₂S₃ (JCPDS No. 25-0390). Two XRD peaks at 17.3 and 24.57° correspond to the (200) and (220) crystal planes of NiFe(CN)₆ (JCPDS No. 51-1897), respectively. Moreover, the XRD peaks of In₂S₃ can also be observed on the photoanode of TiO₂/In₂S₃/NFP, suggesting the crystal structure of In₂S₃ is unaffected^[37].

Figure 2. (A) XRD patterns of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP. XPS spectra of (B) Ti 2p, (C) O 1s, and (D) S 2p of TiO₂/In₂S₃. XPS spectra of (E) C 1s, (F) N 1s, (G) Ni 2p, and (H) Fe 2p of TiO₂/In₂S₃/NFP. Inset of (F): the crystal structure of NiFe PBA. (I) Transmission spectra and UV-vis absorption spectra of TiO₂ and TiO₂/In₂S₃ with the different concentrations of indium nitrate in the precursor.

The composition and chemical states of the TiO₂-based photoanodes were analyzed with XPS, as presented in Supplementary Figure 3 and Figure 2B-H. All elements of Ti, O, S, In, Ni, Fe, N, and C were successfully identified on the XPS survey spectra. In Figure 2B, Ti 2p_1/2 and Ti 2p_3/2 peaks can be observed at 458.6 and 464.3 eV on the TiO₂ photoanode, respectively. No other peaks were observed at lower binding energy, indicating only the presence of Ti⁴⁺, not Ti^3+[38]. For TiO₂/In₂S₃, Ti 2p_1/2 and Ti 2p_3/2 peaks show a positive shift of 0.4 eV compared to the as-prepared TiO₂. In Figure 2C, O 1s XPS spectra can be fitted with two peaks corresponding to Ti-O and oxygen vacancies (O_v)/S-O. The Ti-O peak also shows a positive shift of 0.35 eV after the growth of In₂S₃. It is important to mention that the S-O peak overlaps with the O_V peak due to similar binding energies^[20,39]. In Figure 2D, the S 2p XPS spectra of In₂S₃ powder exhibit two peaks at 161.61 and 162.82 eV, corresponding to the S 2p_3/2 and S 2p_1/2 peaks, respectively. After the formation of TiO₂/In₂S₃ heterojunction, these two peaks display a negative shift of ~0.15 eV. The opposite shift direction of Ti 2p/O 1s and S 2p powerfully demonstrates the charge transfer from TiO₂ to In₂S₃ at the heterojunction interface and the formation of a built-in electric field oriented from TiO₂ to In₂S₃^[16]. The C 1s spectrum of TiO₂/In₂S₃/NFP in Figure 2E can be deconvoluted into three distinct peaks located at 284.8 (C-C), 285.8 (C≡N), and 288.6 eV (C=O). The N 1s spectrum of TiO₂/In₂S₃/NFP depicted in Figure 2F confirms the presence of C≡N at 397.8 eV. The crystal structure of NFP is shown in the inset of Figure 2F. This demonstrates that the covering of NFP on the surface of TiO₂/In₂S₃ heterojunction is successful. In Figure 2G, the peaks located at binding energies of 873.8 and 856.3 eV correspond to the Ni 2p_1/2 and Ni 2p_3/2 peaks, respectively. The Fe 2p peaks in Figure 2H confirm the presence of Fe³⁺ ions with characteristic binding energies of 708.46 and 721.4 eV^[25]. Figure 2I depicts the transmission spectra and UV-vis absorption spectra of TiO₂ and TiO₂/In₂S₃ prepared with various amounts of indium nitrate. All samples exhibit pronounced UV absorbance, and with an increasing amount of indium nitrate, the light absorption range of TiO₂/In₂S₃ progressively extends into the visible light region. The photographs of TiO₂ and TiO₂/In₂S₃ photoanodes in Supplementary Figure 4A can directly confirm the improved visible light absorption. This demonstrates that the In₂S₃ layer can effectively enhance the visible light absorption for more electron-hole pairs photogeneration.

PEC performance of TiO₂/In₂S₃/NFP photoanode

In₂S₃ and NFP serve a light absorption layer and cocatalyst layer, respectively, which can greatly influence the PEC performance of the TiO₂/In₂S₃/NFP photoanode. The amount of indium nitrate and the reaction time and temperature of NFP growth were optimized first in this work. For the following experiments, 0.05 mmol In(NO₃)₃, 3 h NFP growth time, and 60 °C NFP synthesis temperature were ascertained. The detailed optimization results are shown in Supplementary Figures 5-8.

PEC performances of the TiO₂/In₂S₃/NFP photoanode are shown in Figure 3, compared with the control photoanodes of TiO₂ and TiO₂/In₂S₃. In Figure 3A, the photocurrent density of TiO₂/In₂S₃/NFP is 1.81 mA cm^-2, which is five times higher than that of TiO₂ (0.30 mA cm^-2). The TiO₂/In₂S₃ heterojunction exhibits a photocurrent density of 1.3 mA cm^-2. This clearly demonstrates the critical role of TiO₂/In₂S₃ heterojunction fabrication. The applied bias photon-to-current efficiency (ABPE) was further calculated based on the light LSV curves, and the results are depicted in Figure 3B. TiO₂/In₂S₃/NFP has the highest ABPE efficiency of 0.94% at 0.51 V vs. RHE, surpassing the as-prepared TiO₂ by 5.8 times. As shown in Figure 3C, the normalized photovoltage decay curves show that the TiO₂/In₂S₃/NFP photoanode attenuates the fastest under dark conditions. This suggests an improved separation of photogenerated electron-hole pairs^[40]. Additionally, the photovoltage of TiO₂/In₂S₃/NFP is around 130 mV higher than that of the as-prepared TiO₂. The photovoltage was obtained by calculating the difference between OCP_dark and OCP_light. The photovoltage value can reflect the degree of energy band bending at the semiconductor-electrolyte interface (SEI). A larger photovoltage indicates a more pronounced energy band bending and a stronger built-in electric field generated by the SEI^[41], which can enhance the charge separation. Figure 3D illustrates the I-t curves under the chopped light, which can reflect the transient photocurrent response of photoanodes. Notably, no discernible transient peak can be observed, suggesting the effective separation of the photogenerated electrons and holes^[42]. The origin of the photocurrent decay is believed to be due to the poor charge transport of TiO₂/In₂S₃/NFP, which may be changed by doping TiO₂. The inset of Figure 3D illustrates that the transient decay time (τ) becomes longer after the formation of TiO₂/In₂S₃ heterojunction and the NFP modification. The increased τ of TiO₂/In₂S₃/NFP indicates that the charge recombination has been suppressed efficiently^[43]. Dark LSV curves of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP electrodes are shown in Supplementary Figure 9. At the current density of 10 mA cm^-2, the overpotential of TiO₂/In₂S₃/NFP is significantly negative shift by 166 mV compared to TiO₂/In₂S₃. This confirms the role of NFP overlayer as cocatalyst for water oxidation reaction.

Figure 3. (A) LSV curves, (B) ABPE plots, (C) normalized photovoltage decay and inset OCP, (D) transient photocurrent curves and inset transient decay time (τ) curves of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP. (E) PEC stability curves of TiO₂/In₂S₃ and TiO₂/In₂S₃/NFP photoanodes at 1.23 V vs. RHE under AM1.5G illumination. Inset: SEM images of the tested TiO₂/In₂S₃ and TiO₂/In₂S₃/NFP photoanodes.

To reveal the PEC stability of TiO₂/In₂S₃/NFP, a long-term stability test was performed nonstop for 12 h under AM1.5G illumination at 1.23 V vs. RHE, as depicted in Figure 3E. After 12 h, TiO₂/In₂S₃/NFP exhibits an approximate 95% retention of photocurrent after an initial transient decrease, which is an 11% improvement over TiO₂/In₂S₃ (84%). The high stability of TiO₂/In₂S₃/NFP should be attributed to the robust NFP, which can also block the photocorrosion of In₂S₃ by physically separating the semiconductor from the electrolyte. The slight decrease of TiO₂/In₂S₃/NFP should be attributed to the imperfect coating of NFP on the surface of TiO₂/In₂S₃. It is worth noting that the photocurrent of TiO₂/In₂S₃/NFP shows significant improvement from 1.81 to 3.1 mA cm^-2 through the LSV measurements [Supplementary Figure 10]. As shown in Supplementary Figure 10C, ECSA was used as a key indicator for evaluating the catalytic performance of photoanodes^[44]. After the stability test, TiO₂/In₂S₃/NFP (5.06 × 10^-2 mF cm^-2) is 1.57 times the as-prepared photoanode (3.22 × 10^-2 mF cm^-2). This increase is consistent with the improvement of photocurrent, illustrating that the photocurrent enhancement is due to the enhanced ECSA of TiO₂/In₂S₃/NFP^[45]. This is consistent with the reported light-induced restorative effect of the Prussian blue material^[33]. The inherent mechanism should be further investigated for answering this open question. The morphology and XPS characterizations of the tested TiO₂/In₂S₃ and TiO₂/In₂S₃/NFP photoanodes were performed, and the results are shown in Supplementary Figure 11 and the inset of Figure 3E. The NFP overlayer can greatly slow down the photocorrosion of In₂S₃. After the stability test, the Ni 2p peaks of TiO₂/In₂S₃/NFP are similar to the peaks of Ni(OH)₂^[46]. This suggests that the as-prepared NFP cocatalyst is partially transformed into the active species of Ni(OH)₂, which is consistent with the reported results^[47]. As per the previous findings, the in-situ generated Ni(OH)₂ from NFP can create NiOOH_2-x that contains Ni⁴⁺ ions, which are highly active sites for water oxidation reactions. In addition, Supplementary Table 1 summarizes the comparison of the PEC performance of this work with the reported related photoanodes. The PEC performance of TiO₂/In₂S₃ achieved in this study is comparable to or surpassing that of the reported photoanodes. This work further demonstrates the effectiveness of NFP as a cocatalyst for TiO₂-based photoanodes. In the absence of any sacrificial agents, the photocurrent density of the TiO₂/In₂S₃/NFP photoanode reaches 1.81 mA cm^-2, which is six times that of the bare TiO₂. In addition, we measured the photocurrent of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP photoanodes under various light intensities, such as 40, 70, 100, 130, 160, 190, and 200 mW cm^-2. There is a linear growth of the photocurrent with the light intensity in the measured range [Supplementary Figure 12]. Moreover, TiO₂/In₂S₃ and TiO₂/In₂S₃/NFP photoanodes show a higher slope of ~0.02 mA mW^-1 than that of TiO₂ (0.004 mA mW^-1). This feature suggests that the TiO₂/In₂S₃ heterojunction and the NFP cocatalyst overlayer can effectively facilitate the charge separation without limiting the hole transfer to the electrolyte^[48].

The ECSA of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP was further studied. CV curves [Supplementary Figure 13] were measured at scan rates of 5-100 mV s^-1 in 1 M KOH, within a potential window of 0.625-0.725 V vs. RHE. The relationship between ΔJ and scan rate is illustrated in Figure 4A. The fitted dashed line provides the double-layer capacitance of TiO₂-based photoanodes, determined by their slopes. After the coating of the NFP overlayer, the TiO₂/In₂S₃/NFP photoanode presents a two-fold ECSA in comparison with TiO₂. This increase is lower than the improvement of five times on the photocurrent. This implies a significantly synergistic effect among TiO₂, In₂S₃, and NFP, not just the increase of ECSA. In order to investigate the charge transfer processes at the SEI interface, we conducted PEIS experiments under AM1.5G illumination at 0.47 V vs. RHE. The Nyquist plot was fitted using an equivalent circuit model, as shown in the inset of Figure 4B. It consists of a series resistance (R_s), a trapping resistance (R_trap), a bulk phase constant phase element (CPE_bulk), a charge transfer resistance (R_ct,trap), and a trapping CPE (CPE_trap). The fitting parameters used are summarized in Table 1. Compared with TiO₂/In₂S₃, the TiO₂/In₂S₃/NFP photoanode shows a significant decrease in R_trap (475.15 Ω) and R_ct,trap (5711 Ω). This demonstrates that the NFP cocatalyst layer can boost the charge transfer kinetics at SEI. M-S plots were acquired at a fixed frequency of 2,000 Hz in the dark, as depicted in Figure 4C. All samples exhibit the positive slopes, illustrating their n-type conductivity. Evidently, TiO₂/In₂S₃/NFP exhibits a more negative flatband potential, indicating the position of the Fermi level of the photoanode in relation to the potential of the reference electrode. The negatively shifted flatband potential is typically associated with a lowered Fermi level^[49]. The donor density of TiO₂/In₂S₃/NFP is 1.33 times of that of TiO₂/In₂S₃, indicating the good conductivity of the NFP overlayer for charge transport. To further elucidate the kinetics of the TiO₂-based photoanodes, we measured charge transfer efficiency (η_inj) and charge separation efficiency (η_sep), as shown in Figure 4D and E, respectively. At 1.23 V vs. RHE, TiO₂/In₂S₃/NFP exhibits the highest η_inj of 93.86% and the highest η_sep of 93.55%. TiO₂/In₂S₃ has a moderate η_inj and η_sep, corresponding to 84.58% and 80.67%, respectively. TiO₂ only achieves a very low η_sep of 21.80%, indicating that the TiO₂/In₂S₃ heterojunction plays a major role for the charge separation. The NFP cocatalyst overlayer can promote the charge injection at SEI and then facilitate the charge separation in the depletion region. Steady-state PL spectra of TiO₂-based photoanodes measured with a light excitation of 420 nm are shown in Figure 4F. PL spectra reveal the radiative recombination of the photogenerated electrons and holes. The PL peaks located at 535.61 and 682.07 nm can be attributed to the excitonic PL process, in which the radiative transition from the sub-bands to the VB top occurs^[50]. The sub-bands mainly result from the surface defects of the TiO₂ photoanode. Therefore, the suppressed PL signal can be observed after the formation of TiO₂/In₂S₃ heterojunction. The heterojunction can efficiently boost the process of the charge separation because of the built-in electric field. The modification of the NFP overlayer cannot further suppress the PL signal. This is because the PL measurements were conducted in air, not in the electrolyte. The charges transferred to the NFP overlayer cannot be migrated from NFP to the redox substance in the air. In other words, in the electrolyte, the NFP overlayer can migrate the charges from NFP to water molecules and promote the charge separation.

Figure 4. (A) Determination of capacitance (C_dl) values through ΔJvs. scan rate plots for TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP. (B) Nyquist plots of TiO₂-based photoanodes measured at 0.47 V vs. RHE under AM1.5G illumination. Inset shows the equivalent circuit for alternating current (AC) impedance fitting. (C) M-S plots of TiO₂-based photoanodes measured at a fixed frequency of 2,000 Hz in the dark. (D) Surface charge injection efficiency (ɳ_inj) of TiO₂-based photoanodes. (E) Bulk charge separation efficiency (ɳ_sep) of TiO₂-based photoanodes. (F) PL spectra of TiO₂, TiO₂/In₂S₃, and TiO₂/In₂S₃/NFP.

Probable mechanism of PEC improvement

To determine the energy band positions of TiO₂ and In₂S₃, Tauc plots were generated using UV-vis absorption spectra [Figure 5A]. UPS spectra of TiO₂ and In₂S₃ were further measured to determine the work functions and the edge of the VB. During the UV-vis absorption spectra and UPS measurements of In₂S₃, the In₂S₃ powder, which was collected from the hydrothermal residues, was used. The photograph of the In₂S₃ powder is shown in Supplementary Figure 4B. The bandgaps (E_g) of TiO₂ and In₂S₃ can be derived from the Tauc plots, corresponding to 3.01 and 1.92 eV, respectively. The two bandgap values are consistent with the reported results^[20]. The work functions of 4.39 eV for TiO₂ and 4.08 eV for In₂S₃ can be obtained from the UPS spectra in Figure 5B and C, respectively. Furthermore, the energies from the VB edge to the Fermi level are 2.78 and 1.61 eV for TiO₂ and In₂S₃, respectively. These experimental data offer detailed insights into the energy band structures of TiO₂ and In₂S₃, benefitting our understanding of the electron structure and PEC processes of the TiO₂/In₂S₃ photoanode.

Figure 5. (A) Tauc plots of TiO₂ and In₂S₃. (B) UPS spectrum of TiO₂. (C) UPS spectrum of In₂S₃. (D) Schematic diagrams of the energy band structures for TiO₂ and In₂S₃ before and after their contact. (E) Mechanism diagram of TiO₂/In₂S₃/NFP photoanode for solar water splitting.

The schematic diagram of the energy band structure before and after the contact between TiO₂ and In₂S₃[Figure 5D] illustrates that, initially, there is a disparity in the Fermi levels between TiO₂ and In₂S₃, with In₂S₃ exhibiting a higher Fermi level. When In₂S₃ was grown on the surface of TiO₂, electrons began to migrate from In₂S₃ to TiO₂, gradually aligning the Fermi levels of both materials. Finally, a classical type-II heterojunction was formed with TiO₂ and In₂S₃. An internal built-in electric field was generated at the interface of TiO₂/In₂S₃ heterojunction. The electric field direction is from TiO₂ to In₂S₃. The energy band structure of TiO₂/In₂S₃ type-II heterojunction facilitates the separation of the photogenerated charges due to the electric field force. Specifically, it is energy-favorable that the photogenerated electrons are transported from the conduction band (CB) of In₂S₃ to the CB of TiO₂. For the photogenerated holes, the large VB offset at the interface of TiO₂/In₂S₃ can provide tremendous driving force and push the holes to the side of In₂S₃. The separated holes can be migrated to the NFP cocatalyst overlayer. The NFP catalyst captured holes rapidly, preventing the recombination of electrons and holes at the TiO₂/In₂S₃ interface and thus enhancing the utilization of photogenerated charge carriers. More importantly, the NFP catalyst can promote the hole participation in the water oxidation reaction at SEI^[16]. As shown in Supplementary Figure 14, the NFP cocatalyst can remarkably improve the electrochemical activity and provide the active species through the state variation of nickel and iron elements. Therefore, thanks to the type-II heterojunction of TiO₂/In₂S₃ and the efficient NFP cocatalyst, a higher photocurrent density can be achieved with TiO₂/In₂S₃/NFP photoanodes. The theoretical photocurrent density (2.03 mA cm^-2) of TiO₂/In₂S₃/NFP can be estimated from its UV-vis absorption spectrum shown in Supplementary Figure 15. The experimental photocurrent of 1.81 mA cm^-2 is ~90% of the above theoretical value. In brief, as shown in Figure 5E, TiO₂ and In₂S₃ can absorb the UV and visible light, respectively, and form the type-II heterojunction. The type-II heterojunction can promote the charge separation and suppress the charge recombination due to the built-in electric field at the interface of TiO₂/In₂S₃. A NFP layer, as cocatlyst and protector, can enhance the water oxidation reactions through the transition of Ni²⁺/Ni³⁺ and Fe²⁺/Fe³⁺ and protect the In₂S₃ layer from the photocorrosion, respectively.