Engineering band structuring via dual atom modification for an efficient photoanode

Abstract

Efficient carrier separation is important for improving photoelectrochemical water splitting. Here, the morphology modification and band structure engineering of Ta₃N₅ are accomplished by doping it with Cu and Zr using a two-step method for the first time. The initially interstitially-doped Cu atoms act as anchors to interact with subsequently doped Zr atoms under the influence of differences in electronegativity. This interaction results in Cu,Zr_g–Ta₃N₅ having a dense morphology and higher crystallinity, which helps to reduce carrier recombination at grain boundaries. Furthermore, the gradient doping of Zr generates a band edge energy gradient, which significantly enhances bulk charge separation efficiency. Therefore, a photoanode based on Cu,Zr_g–Ta₃N₅ delivers an onset potential of 0.38 V_RHE and a photocurrent density of 8.9 mA cm⁻² at 1.23 V_RHE. Among all the Ta₃N₅-based photoanodes deposited on FTO, a Cu,Zr_g–Ta₃N₅-based photoanode has the lowest onset potential and highest photocurrent. The novel material morphology regulation and band edge position engineering strategies described herein provide new ideas for the preparation of other semiconductor nanoparticles to improve the photoelectrochemical water splitting performance.

---

Introduction

Photoelectrochemical (PEC) or electrocatalytic (EC) water splitting, exploiting sunlight or electricity to convert water into hydrogen and oxygen, are effective ways to respond to growing global energy and environmental issues related to fossil fuel consumption.^1–7 The photoelectrode, which absorbs sunlight to generate carriers, is the core of PEC systems.^8,9 Despite a great deal of effort having been devoted to investigating photoelectrodes with high sunlight-to-electron conversion efficiency, there are still many issues with photoelectrodes.^10,11 Specifically, trap states at grain boundaries,¹² bulk recombination,¹³ high onset potential¹⁴ and low photocurrent¹⁵ hinder the realization of a semiconductor for water splitting. Therefore, the development of photoelectrodes with excellent PEC performance is desirable.

Photoelectrodes for PEC water splitting should possess several important properties, including a small band gap to guarantee long-wavelength light absorption, efficient charge separation and catalytic activity. Tantalum nitride (Ta₃N₅), an n-type semiconductor material, is one of the most promising photoanode materials for PEC water splitting.^16–18 However, pure Ta₃N₅ exhibits inferior PEC water splitting performance, although theoretically it can produce a maximum photocurrent density of 12.9 mA cm⁻² under standard sunlight irradiation.^19–21 Some effective strategies, such as foreign atom doping,^22,23 morphological modification^24,25 or interface engineering,^26–28 have been developed to optimize the performance of PEC water splitting, but the photocurrent of the photoelectrode deposited on FTO is far below the theoretical value.^29–33 This phenomenon is primarily related to poor carrier transport and severe trap states at grain boundaries. In particular, when operating at a low bias potential, carrier transport is highly susceptible to potential barriers within the conduction path and easy recombination at grain boundaries. Therefore, it is crucial to focus on improving the separation efficiency of the carriers. Constructing a built-in electric field is an effective way to improve charge separation.^34,35 In the conventional built-in electric field construction, heterojunctions or homojunctions are mostly effective for enhancing carrier separation. However, there are still some conceivable deficiencies that need to be properly considered.^36,37 On the one hand, it is difficult to create an effective heterojunction due to the mismatched energy levels between two semiconductors without generating a low interface defect density. Although Ta₃N₅ heterojunctions have been reported in the literature, the processes used to synthesize these materials are cumbersome and the inhibition of deep level defects is limited.²⁶ On the other hand, most synthesis methods for doping a material with foreign atoms to obtain homogeneous junctions, such as the dual-source electron-beam evaporation deposit method and nitriding at a high temperature,^34,38 require intricate experimental conditions, which increase costs and the difficulty of synthesis. Additionally, heteroatom doping results in the lattice expansion of the pristine semiconductor, creating lower crystallinity and a smaller grain size, which leads to carrier recombination at grain boundaries.¹² As a consequence, a new strategy needs to be proposed to prevent the grain boundaries from influencing the charge transfer properties and to improve the bulk carrier separation efficiency. To the best of our knowledge, there is no report on modifying the morphology and bandgap structure of a semiconductor through dual-heteroatom doping.

Herein, we report for the first time the preparation of Cu and gradient Zr co-doped Ta₃N₅ (Cu,Zr_g–Ta₃N₅) using a two-step synthesis strategy. Due to the electronegativity difference between Cu and Zr, the interstitially doped Cu serves as an anchor site for subsequent interaction with Zr, which modifies the morphology of Ta₃N₅. Additionally, the incorporation of Zr affects the bandgap position. Therefore, a simplified linear pressure-assisted method is used to achieve gradient Zr doping, which bends the band structure of Ta₃N₅ to form a built-in electric field. Experimental results reveal that the doping of Cu and gradient Zr suppresses the intrinsic defect-related and grain boundary recombination of Ta₃N₅. Additionally, the gradient Zr doping allows electrons to rapid transfer to the surface of Cu,Zr_g–Ta₃N₅, resulting in an onset potential of 0.38 V_RHE, which is approximately 0.27 V_RHE lower than the onset potential of pristine Ta₃N₅ (∼0.65 V_RHE). The photocurrent density at 1.23 V_RHE is 8.9 mA cm⁻², which exceeds that of all Ta₃N₅-based photoanodes deposited on FTO. These results demonstrate that the sequential doping of two types of atoms can change the morphology of the base material, and that gradient doping can alter the bandgap structure and improve the performance of the photoanode.

Experimental

Preparation of pristine Ta₃N₅

Ta₂O₅ (99.99%, Shanghai Adamas Reagent Co., Ltd.) was well mixed with 0.5 mmol of Na₂CO₃ (99.5%, Wako Pure Chemical Industries, Ltd.) via mechanical grinding in an agate mortar for 0.5 h. The mixture was then heated to 900 °C for 15 h at a ramp rate of 10 °C min⁻¹ under an ammonia stream (flow rate: 120 sccm). The nitridation duration of 15 h was determined by the complete transition of the intermediate phase, NaTaO₃ to Ta₃N₅. After being allowed to cool to room temperature naturally, the sample was rinsed with water at 65 °C and dried under vacuum at 40 °C for 5 h.

Preparation of Cu-doped Ta₃N₅ (Cu–Ta₃N₅)

The starting materials Ta₂O₅ and Cu(NO₃)₂·3H₂O (Kanto Chemical Co., Inc.) were well blended (5.3, 5.8, 6.4, 6.9, 7.5 at% doped). Then, the mixture was annealed in air at 650 °C for 0.5 h to prepare pure oxide precursors. The annealed oxide precursors were mixed with a flux of 0.5 mmol Na₂CO₃via mechanical grinding in an agate mortar for 0.5 h. For nitridation, the mixed powder was placed into an alumina boat in a furnace and was subsequently heated under an NH₃ flow of 120 sccm at 900 °C for 15 h at a ramping rate of 10 °C min⁻¹. After being allowed to cool to room temperature naturally, the sample was rinsed with water at 65 °C and dried under vacuum at 40 °C for 5 h.

Preparation of Zr-doped Ta₃N₅ (Zr–Ta₃N₅)

The starting materials Ta₂O₅ and ZrO(NO₃)₂·2H₂O (Kanto Chemical Co., Inc.) were well blended (0.62, 0.78, 0.90, 1.08, 1.25 at% doped). Then, the mixture was annealed in air at 650 °C for 0.5 h to prepare pure oxide precursors. The annealed oxide precursors were mixed with a flux of 0.5 mmol of Na₂CO₃via mechanical grinding in an agate mortar for 0.5 h. For nitridation, the mixed powder was placed in an alumina boat in a furnace and was subsequently heated under an NH₃ flow of 120 sccm at 900 °C for 15 h at a ramping rate of 10 °C min⁻¹. After being allowed to cool to room temperature naturally, the sample was rinsed with water at 65 °C and dried under vacuum at 40 °C for 5 h.

Preparation of gradient and homogeneous Zr-doped Cu–Ta₃N₅ (Cu,Zr_g–Ta₃N₅; Cu,Zr_h–Ta₃N₅)

The Cu,Zr_g–Ta₃N₅ powder was synthesized by heating Cu–Ta₃N₅ powder with ZrO(NO₃)₂·2H₂O powder (0.90 at% doped). Specifically, the Cu–Ta₃N₅ was well mixed with ZrO(NO₃)₂·2H₂O powder. Then, the mixture was annealed in air at 650 °C for 0.5 h to prepare pure oxide precursors. The annealed oxide precursors were mixed with a flux of 0.5 mmol of Na₂CO₃via mechanical grinding in an agate mortar for 0.5 h.

The mixture was placed in a simplified linear pressure-assisted reactor under an NH₃ atmosphere. Firstly, the temperature was increased to 900 °C (10 °C min⁻¹), and then linearly pressurized from 0.12 MPa at a rate of 0.01 MPa every 10 min for 280 min, resulting in the formation of Cu,Zr_g-doped Ta₃N₅. After being allowed to cool to room temperature naturally, the sample was rinsed with water at 65 °C and dried under vacuum at 40 °C for 5 h. Cu,Zr_h–Ta₃N₅ was synthesized in the same way, but not under simplified linear pressure, and under an NH₃ flow of 120 sccm at 900 °C for 15 h at a ramping rate of 10 °C min⁻¹.

Preparation of photoanodes via an electrophoretic deposition method

All tested photoanodes were prepared via an electrophoretic deposition method (EPD) on conducting FTO glass supports. EPD was carried out in acetone (50 mL) containing the as-prepared samples (40 mg) and iodine (10 mg), which was dispersed by sonication for 10 min. Two FTO electrodes (1.6 × 2 cm²) were immersed parallel in the solution at a distance of 10 mm, and 15 V bias was then applied between the electrodes for 3 min using a potentiostat (PARSTAT 2263, Princeton Applied Research). The coated area was controlled to be ca. 1.6 × 1.0 cm². The electrode was dried in air at room temperature. The average thickness of the as-prepared samples deposited on FTO was 3.0 μm under the EPD conditions.

Preparation and loading of the NiCoFe–B_i co-catalyst

The preparation and loading of the NiCoFe–B_i co-catalyst was based on previous literature.³⁹ Specifically, the as-prepared electrodes were transferred to an Ar-purged solution containing 2 mM NiSO₄·6H₂O (99.99% metals basis, Chengdu Chron Chemical Co., Ltd.), 0.5 mM Co(NO₃)₂·6H₂O (99.99% metals basis, Aladdin) and 0.8 mM FeSO₄·7H₂O (99.95% metals basis, Aladdin) in 0.25 M potassium borate (K₂B₄O₇·4H₂O) buffer at pH 10. The photo-assisted electrodeposition was performed in a three-electrode configuration with Ag/AgCl as the reference electrode and a Pt wire as the counter electrode. The NiCoFe–B_i co-catalyst was electrodeposited onto the electrodes at a constant current density of 20 μA cm⁻² over a period of 10 min under AM 1.5 G simulated sunlight. After the deposition, the electrodes were rinsed with deionized water.

Characterization methods

The morphologies and microstructures of the samples were characterized by field emission scanning electron microscopy (FESEM, JEOL, JSM-7800F, 15 kV) and transmission electron microscopy (TEM; FEI Talos F200X). Energy dispersive X-ray (EDX) spectroscopy elemental mapping was conducted under the TEM with an annular dark-field (ADF) detector. X-ray photoelectron spectroscopy (XPS) analysis was performed using a Thermo Scientific ESCALAB 250Xi instrument with monochromatic Al Kα radiation (225 W, 15 mA, 15 kV). All binding energies were referenced to the C 1s peak (284.8 eV) arising from adventitious carbon. The crystal phase was examined by powder X-ray diffraction (XRD, PANalytical X'Pert Powder with Cu Kα radiation). UV-visible (UV-vis) diffuse reflectance spectra were recorded on a UV-vis spectrophotometer (UV-3600, Shimadzu, Japan). Ultraviolet photoemission spectroscopy (UPS) measurements were carried out on an ESCALAB 250Xi spectrometer with He I resonance lines (21.22 eV). The steady-state photoluminescence (PL) spectra (excited by 420 nm illumination) were recorded on a spectrofluorometer (FLS1000, Edinburgh Instruments) equipped with both continuous (450 W) and pulsed xenon lamps. The time-resolved transient photoluminescence decay (TRPL) spectra were recorded on a spectrophotometer (Fluorolog-3, Horiba Scientific) using a 450 W xenon lamp. The surface photovoltaic properties (SPV) of the samples were tested using a surface photovoltage test system (CEL-SPS1000). The Ta, Cu and Zr concentrations in the Cu- and Cu,Zr_g-doped Ta₃N₅ were determined using inductively coupled plasma-atomic emission spectroscopy (ICP-AES; Icap 6300 Duo, Thermo Scientific). The oxygen and nitrogen contents of the synthesized Ta₃N₅ were determined using an oxygen–nitrogen combustion analyzer (Horiba, EMGA-620W).

Photoelectrochemical measurements

The PEC performance of the samples was investigated via a controllable reaction system (CEL-PAEM-D8, CEAULight, China) with a volume of approximately 150 mL. A commercial solar simulator (S500RE7, CEAULight, China) was used for irradiation of the photoanodes with a light intensity (1 sun at AM 1.5 G) of 100 mW cm⁻². The measurements were carried out in 1.0 M KOH (pH 13.6) aqueous solution under a three-electrode system on an electrochemical workstation (BioLogic SP-200). The temperature of the electrolyte was maintained at 10 °C using a constant temperature water bath during the PEC test. The as-prepared samples were used as the working electrodes (photoanode), Pt wire as the counter electrode, and a Hg/HgO electrode as the reference electrode, respectively. To prevent the back reaction and the light scattering effect of the generated H₂ bubbles, the Pt cathode chamber and the photoanode chamber were separated using a Nafion 117 membrane. The photocurrent was recorded under simulated solar light while sweeping the potential in the positive direction at a scan rate of 0.01 V s⁻¹. All the measured potentials versus a Hg/HgO reference electrode were converted to the potentials versus RHE according to the Nernst equation. The Applied Bias Photon-to-current Efficiency (ABPE) was calculated from the current–potential curves under AM 1.5 G illumination using the equation ABPE = [J × (1.23 − V_app)/P_light] × 100%, where V_app is the applied potential versus RHE, J is the photocurrent density under AM 1.5 G light and P_light is the irradiance of the simulated sunlight (100 mW cm⁻²). The incident photoelectron conversion efficiency (IPCE) for each wavelength was determined using the following formula: IPCE = (1240J)/(λI_light) × 100%, where λ is the wavelength of incident light (nm), J is the photocurrent density (mA cm⁻²) under the illumination wavelength (λ), and I_light is the intensity of incident light (mW cm⁻²). The IPCE spectra were measured using a monochromatic light source (PLS-SXE300) in the wavelength range from 360 to 650 nm with a 10 nm interval at 1.0 V versus RHE in 1 M KOH. The intensity of the monochromatic light was measured using a calibrated reference cell (CEL-FZ-A). Chronoamperometry measurements (steady-state photocurrent curves) were carried out at a potential of 1.0 V versus RHE under AM 1.5 G simulated sunlight (100 mW cm⁻²). Electrochemical impedance spectra (EIS) were collected on the workstation at an open-circuit voltage under light illumination in a frequency range from 0.1 Hz to 100 kHz. Mott–Schottky plots were obtained at an AC frequency of 1.0 kHz at an amplitude of 0.01 V in the dark. Gas production was detected using a gas chromatograph (Aulight GC-7920) equipped with a thermal conductivity detector (TCD) and a flame ionization detector (FID).

Results and discussion

In order to determine the optimal doping levels of Cu and Zr for subsequent experiments, we first measured the oxygen production of photoanodes with different heteroatom doping amounts. The amounts of oxygen produced by NiCoFe–B_i/Cu–Ta₃N₅/FTO and NiCoFe–B_i/Zr–Ta₃N₅/FTO photoanodes, which were held at 1.0 V_RHE in 1 M KOH under AM 1.5 G simulated sunlight for 1 h, are shown in Fig. S1.† The figure displays a volcano-like shape and reveals that the optimal doping level ratios of Cu and Zr are ∼6.4% and ∼0.9%, respectively. Therefore, these two values were used as the doping levels for subsequent experiments. It is difficult to achieve gradient doping using the most commonly employed synthesis method, normal atmospheric nitridation (Fig. S2†). In order to realize gradient doping, a simplified linear pressurized gas-assisted process can be used.^40,41 High-angle annular dark-field scanning transmission electron microscopy (HAADF-TEM) was employed to study the morphology and elemental distribution of Cu,Zr_g–Ta₃N₅. The morphology of Cu,Zr_g–Ta₃N₅ (Fig. 1a) is more compact than that of pristine Ta₃N₅ (Fig. S3†). The elemental mapping of Cu,Zr_g–Ta₃N₅ shows that Ta, N and Cu are uniformly distributed in the nanoparticles, while the distribution of Zr gradually increases from the inside to the outside. The line-scan EDS profile (Fig. 1e) results obtained along the cyan dotted-line from the Cu,Zr_g–Ta₃N₅ nanoparticles (Fig. 1a) are consistent with the elemental mapping images. As shown in Fig. 1b, the observed lattice fringes with an interplanar spacing of 0.366 nm were assigned to the (110) plane of Ta₃N₅.

Fig. 1 (a) HAADF-STEM image of Cu,Zr_g–Ta₃N₅ and the corresponding EDS elemental mappings images of Cu,Zr_g–Ta₃N₅; (b) HRTEM images of Cu,Zr_g–Ta₃N₅ and the corresponding FFT pattern; (c) XRD spectra of Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_g–Ta₃N₅; (d) the zoomed-in image of Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_g–Ta₃N₅ (110) diffraction peaks; (e) the corresponding line-scan EDS profile obtained along the cyan dotted-line from (a); (f) wide-scan survey XPS profiles of Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_g–Ta₃N₅.

As shown in Fig. 1c, the crystal structures of the Cu,Zr_g–Ta₃N₅ nanoparticles and control samples (pristine Ta₃N₅, Cu–Ta₃N₅) were characterized by X-ray diffractometry. All the samples show diffraction patterns consistent with a single phase of Ta₃N₅ (JCPDF no. 79-1533), which indicates that the crystal structure was not changed by Cu or Cu,Zr co-doping. More importantly, Cu₂O, ZrO₂, ZrON₂ and NaTaO₃ impurity phases were not observed in our sample. Furthermore, carefully examining the magnified diffraction peaks of the (110) facet (Fig. 1d) reveals a shift in the diffraction peaks to lower angles for all samples. The shifts in the (110) XRD peak positions of the various Ta₃N₅ are ascribed to the substitution of Ta⁵⁺ (radius: 64 pm⁶) by Ta³⁺ (72 pm⁶)/Cu²⁺ (73 pm⁶)/Zr⁴⁺ (72 pm⁶) as well as the substitution of N³⁻ (146 pm⁶) by O²⁻ (140 pm⁶). Substitutions of atoms with a larger ionic radius would shift the XRD peak position to a lower angle, while a smaller radius would result in a shift to a higher angle, respectively. The lower angle of the (110) peak generated by the pristine Ta₃N₅ compared with that of the standard is primarily due to the presence of Ta³⁺ ions. A shift in the (110) peak to lower angle presented in Cu–Ta₃N₅ compared with pristine Ta₃N₅ is due to the foreign cations (i.e., Cu²⁺ ions) incorporated into Cu–Ta₃N₅. However, the angle shift of Cu–Ta₃N₅ is very small compared to that of pristine Ta₃N₅, which may be due to Cu–Ta₃N₅ containing more O_N defects compared to pristine Ta₃N₅. Cu,Zr_g–Ta₃N₅ displays the same (110) peak position as pristine Ta₃N₅, most probably as a result of the formation of a greater number of O_N defects in the former (Tables S1 and S3†). In addition, studies in the literature have proven that O_N can increase the carrier concentration and promote carrier transfer,^3,39 thus more O_N in samples will improve PEC performance. Meanwhile, the diffraction peaks of Cu,Zr_g–Ta₃N₅ become narrower and of higher intensity than both those of Cu–Ta₃N₅ and pristine Ta₃N₅, suggesting that the gradient incorporation of Zr enhanced crystallinity. The grain size obtained by applying the Scherrer equation to the (110) peak of Cu,Zr_g–Ta₃N₅ is the largest of all the samples (Table S4†).⁴² Studies in the literature have confirmed that the grain size of thin films affects PEC properties.⁴³ However, subsequent PEC testing in this study confirmed that the grain size of nanoparticles also affects the performance. The XPS wide-scan spectrum shows the apparent signals of the elemental Cu and Zr (Fig. 1f). The presence of Cu and Zr ions was verified from the high-resolution XPS spectra of samples (Fig. S4 and S5†). The presence of Ta³⁺ ions was verified from the Ta 4f XPS spectrum of pristine Ta₃N₅ (Fig. S7†).

We synthesized Cu,Zr_g–Ta₃N₅ nanoparticles using a simplified linear pressure-assisted method. The Cu,Zr_g–Ta₃N₅ nanoparticles exhibited a distinct morphology compared to pristine Ta₃N₅, Cu–Ta₃N₅, Zr–Ta₃N₅ and Cu,Zr_h–Ta₃N₅. As presented in Fig. 2a–e, pristine Ta₃N₅, Cu–Ta₃N₅, Zr–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ show porous structures compared to Cu,Zr_g–Ta₃N₅. It is worth noting that the morphologies of Zr–Ta₃N₅ and Cu–Ta₃N₅ are similar to that of pristine Ta₃N₅, indicating that doping Cu or Zr alone do not change the morphology of Ta₃N₅. The morphology of Cu,Zr_h–Ta₃N₅, however, is different to those of pristine Ta₃N₅ and Ta₃N₅ doped with Cu or Zr alone, and it can be seen that the number of holes decreases (Fig. 2e). This is because interstitially doped Cu acts as an anchor point, and after Zr doping, interaction between Cu and Zr occurs due to the difference in electronegativity.⁴⁴ Nevertheless, the morphology of Cu,Zr_h–Ta₃N₅ is also different from that of Cu,Zr_g–Ta₃N₅. The morphology disparity observed between Cu,Zr_g–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ particles can be attributed to the gradient doping of Zr, which gradually enhances the interaction between Zr and Cu from the outside in. This results in Cu,Zr_g–Ta₃N₅ exhibiting a denser morphology compared to the other samples. The Brunauer–Emmett–Teller (BET) test results further corroborate that Cu,Zr_g–Ta₃N₅ has fewer pores and its morphology is different from the other samples (Fig. 2f).

Fig. 2 SEM images of (a) Cu,Zr_g–Ta₃N₅; (b) pristine Ta₃N₅; (c) Cu–Ta₃N₅; (d) Zr–Ta₃N₅ and (e) Cu,Zr_h–Ta₃N₅; (f) BET surface area of pristine Ta₃N₅, Cu–Ta₃N₅, Zr–Ta₃N₅, Cu,Zr_h–Ta₃N₅, and Cu,Zr_g–Ta₃N₅. In the same size red circle, (b–d) exhibit more pores than (e), but (e) shows more pores than (a).

High-resolution X-ray photoelectron spectroscopy (XPS) data were employed to elucidate the valence states and chemical environment of the as-synthesized materials. The high-resolution Cu 2p spectra are exhibited in Fig. S4.† Notably, Fig. S4† reveals an interesting finding that Cu is present in the sample not only in the form of Cu²⁺, but also as Cu⁺. The presence of Cu⁺ in all Cu-doped Ta₃N₅ can be explained by the oxidation of Ta³⁺ ions by Cu²⁺, which is why Ta³⁺ disappears when Ta₃N₅ is doped with Cu. However, the content of Cu⁺ is reduced and Cu²⁺ increased when Zr is incorporated into Cu–Ta₃N₅, regardless of uniform or gradient doping. This phenomenon can be explained by the difference in electronegativity, which causes electrons to transfer from Cu⁺ (1.163) to Zr⁴⁺ (1.610).⁴⁵ Electron transfer was further confirmed by the presence of Zr³⁺ in the Zr 3d XPS spectrum etched for 0 s (Fig. S5a†). We further performed XPS depth-profile analysis of Cu,Zr_g–Ta₃N₅ by means of argon sputtering, as shown in Fig. S5b and S6.† With an extension in the sputtering time, the peak area ratio of Zr³⁺ first increases and then decreases, while the peak area of Zr³⁺ in Cu,Zr_h–Ta₃N₅ remains almost unchanged (Fig. S8†). This proves that the interaction between Cu and Zr is continuously enhanced. The O 1s XPS spectra in Fig. S7a† show two typical peaks at approximately 530.2 and 531.4 eV, which are attributed to lattice oxygen (TaO_x species) and a surface-absorbed oxygen species, respectively.³⁴ The Ta 4f peak could be fitted to three doublets (Ta 4f_7/2–Ta 4f_5/2) with a spin–orbit splitting of 1.9 eV and a fixed area ratio of 4 : 3 (Fig. S7b†). The binding energy of the Ta 4f_7/2 components were 23.6, 24.5, 26.0 eV, corresponding to Ta³⁺, Ta₃N₅ and tantalum oxynitride (TaO_xN_y).⁴⁶ The proportions of Ta³⁺, Ta₃N₅ and TaO_xN_y among the total Ta species in pristine Ta₃N₅ were estimated to be 12.5%, 85.8% and 1.7%, respectively. However, with Cu or Cu/Zr doping, the Ta³⁺ completely disappears, and the proportion of Ta₃N₅ is considerably increased (Table S3†). The N 1s peak at ∼396.6 eV in Fig. S7c† corresponds to Ta–N binding. Additionally, a shift in the peaks of Cu–Ta₃N₅ and Cu,Zr_g–Ta₃N₅ is observed, demonstrating Cu or Cu/Zr have successfully doped.

To investigate the impact of Cu or Cu/Zr doping on the band gap structure, ultraviolet photoelectron spectroscopy (UPS) and ultraviolet-visible (UV-vis) spectroscopy were conducted to manifest the band edge positions of pristine Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅. By subtracting the cut-off energies for the secondary electrons from the He I excitation energy (21.22 eV), the Fermi levels (E_F) of pristine Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ are −3.92, −4.10 and −4.37 eV, respectively (Fig. 3a). The UPS spectra reveal that the valence bands of pristine Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ are 2.05, 1.89 and 1.94 eV below their Fermi levels, respectively (Fig. 3b). The UV-vis results show that the absorption of Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ is weaker than that of pristine Ta₃N₅ when the wavelength exceeds 600 nm (Fig. 3c). This is indicative of a lower density of defects caused by the reduced Ta⁵⁺ species or nitrogen vacancies.⁴⁷ The Tauc plots of the UV-vis spectra in Fig. 3d manifest that the optical band gaps of the pristine Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ are 2.06, 2.08 and 2.10 eV, respectively. The low-valence cations (Cu²⁺ and Zr⁴⁺) doped in Ta₃N₅ would require the formation of V_N and/or O_N (Table S1†) to compensate for the imbalanced charge.^48,49 Therefore, the valence band may partially be hybridized with O 2p orbitals with a lower electronic potential, leading to the enlarged band gap after Cu or Cu/Zr doping.⁵⁰ Combined with the UV-vis, UPS and Tauc data, the energy band positions of the pristine-Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ are shown in Fig. 3e, which are consistent with the M–S plot results (Fig. S9†). Detailed values are summarized in Table S2.† As shown in Fig. 3e, the Fermi energy level of Cu, Zr_h–Ta₃N₅ is much lower than those of pristine Ta₃N₅ and Cu–Ta₃N₅. This finding confirms that gradient doping can tailor the band structure, thereby enhancing carrier separation.^34,35 The photoluminescence (PL) emission spectra of pristine Ta₃N₅, Cu,Zr_h–Ta₃N₅ and Cu,Zr_g–Ta₃N₅ are shown in Fig. 3f. The pristine Ta₃N₅ exhibits a quenched characteristic PL peak between 700 and 850, which is attributed to a strong defect-related recombination.²⁹ Notably, the disappearance of defect-related recombination in Cu,Zr_h–Ta₃N₅ and Cu,Zr_g–Ta₃N₅ confirms the positive effect of the Cu/Zr doping. The intensity of the PL emission spectrum, however, is significantly reduced after Zr gradient doping, indicating that the gradient Zr bends the band structure to facilitate the separation of photogenerated charges. Time-resolved PL (TRPL) spectroscopy was performed to study the carrier kinetics in various photoanodes (Fig. 3g). The fitted parameters for different Ta₃N₅ samples are listed in Table S5,† where τ₁ and τ₂ are the defect-related photogenerated carrier trapping and electron–hole recombination from the conduction band to the valence band, respectively.⁵¹ As shown in Table S5,† the PL decay of the pristine Ta₃N₅ is mainly caused by defect-related recombination, while the defect-related recombination is slightly suppressed in Cu,Zr_h–Ta₃N₅. In contrast, both the τ₁ and τ₂ values of Cu,Zr_g–Ta₃N₅ are greater than those of pristine Ta₃N₅ and Cu,Zr_h–Ta₃N₅. Importantly, the decreased τ₁ ratio of Cu,Zr_g–Ta₃N₅ implies that gradient-doping of Zr further inhibits the defect-related recombination. In addition, the electron–hole recombination lifetime of Cu,Zr_g–Ta₃N₅ is much longer than that of pristine Ta₃N₅ and Cu,Zr_h–Ta₃N₅, suggesting that gradient doping with Zr improves carrier separation capability. We also used surface photovoltaic (SPV) spectral data to advance the understanding of the surface separation/transfer ability of carriers for different samples (Fig. 3h). The signal of Cu,Zr_g–Ta₃N₅ is markedly enhanced relative to those of Cu,Zr_h–Ta₃N₅ and pristine Ta₃N₅, indicating accelerated carrier separation/transfer of Cu,Zr_g–Ta₃N₅. The rapid carrier separation and transfer is conducive to the negative shift in the onset potential, which was confirmed in the subsequent PEC tests.

Fig. 3 (a) UPS spectra of Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅; (b) the zoomed-in image of the gray dashed line in (a); (c) UV-vis absorption spectra of Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅; (d) Tauc plots of the UV-vis absorption spectra of Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅. α, absorption coefficient; ℏ, Planck's constant; v, photon frequency. The dotted lines show the extrapolation of the linear portion of the absorption edges; (e) band diagrams of Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ determined from UPS and UV-vis absorption measurements. CB, conduction band; VB, valence band; E_F, Fermi level; (f) room temperature PL emission spectra of Ta₃N₅, Cu,Zr_h–Ta₃N₅ and Cu,Zr_g–Ta₃N₅; (g) TRPL decay curves of Ta₃N₅, Cu,Zr_h–Ta₃N₅ and Cu,Zr_g–Ta₃N₅; (h) SPV spectra of Ta₃N₅, Cu,Zr_h–Ta₃N₅ and Cu,Zr_g–Ta₃N₅.

In response to the issue of the sluggish oxygen evolution kinetics of Ta₃N₅, a NiCoFe–B_i oxygen evolution co-catalyst layer was deposited on the surface of the photoelectrodes.^39,52 We deposited pure NiCoFe–B_i on FTO to demonstrate its excellent oxygen evolution reaction (OER) properties. As shown in Fig. S10,† NiCoFe–B_i/FTO exhibits excellent OER performance and stability. To advance our understanding of the importance of NiCoFe–B_i in the enhanced OER performance of Cu,Zr_g–Ta₃N₅, electrochemical-impedance-spectroscopy (EIS) and chopped I–T tests at 1.0 V_RHE were carried out. Obviously lower resistance between photoanode and electrolyte was observed, indicating that co-catalyst modification is necessary to enhance the charge transfer kinetics between the photoanode and electrolyte (Fig. S11†). Furthermore, the greatly enhanced photocurrent demonstrates the importance of NiCoFe–B_i in improving this property (Fig. S12†).^53,54 The SEM images of Cu,Zr_g–Ta₃N₅ deposited on FTO and NiCoFe–B_i adhering to Cu,Zr_g–Ta₃N₅ are shown in Fig. S13.† It can be seen that Cu,Zr_g–Ta₃N₅ is uniformly deposited on FTO, while NiCoFe–B_i is also uniformly adhered to Cu,Zr_g–Ta₃N₅. Due to the gradient bandgap structure resulting from the gradient doping of Zr in Cu,Zr_g–Ta₃N₅, electrons are able to transfer quickly to the surface of the photoanode to generate a photocurrent at a low bias (depicted in Fig. 4a). Steady-state photocurrent testing under low bias conditions was conducted (Fig. 4b). A steady photocurrent density of ∼20 μA cm⁻² was generated at 0.38 V_RHE, demonstrating a low onset potential for the Cu,Zr_g–Ta₃N₅ photoanode. To the best of our knowledge, this is better than the results of other Ta₃N₅-based photoanodes deposited on FTO (Fig. 4f). Linear-sweep voltammetry (LSV) was employed to study the PEC activity of Cu,Zr_g–Ta₃N₅, and the corresponding control sample results are also shown in Fig. 4c and S14a.† Apparently, the doping of foreign elements enhances the photocurrent density of Ta₃N₅. The fill factor of the J–V curves was also substantially improved. This phenomenon indicates that foreign ion doping significantly inhibits defect-related recombination, which is consistent with the PL and TRPL results. Compared to pristine Ta₃N₅ with a value of ∼2.2 mA cm⁻², Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ exhibit values of ∼6.5 mA cm⁻² and ∼6.7 mA cm⁻², respectively; however, Cu,Zr_g–Ta₃N₅ achieves the highest photocurrent density of ∼8.9 mA cm⁻² at 1.23 V_RHE. To the best of our knowledge, a photocurrent density of ∼8.9 mA cm⁻² is the highest value reported upon depositing Ta₃N₅ on FTO (Fig. 4f). Additionally, as shown in Fig. 4c and S14a,† compared to pristine Ta₃N₅, Ta₃N₅ doped with foreign ions exhibits a lower onset potential. The onset potential of Cu,Zr_h–Ta₃N₅ (∼0.43 V_RHE) is negatively shifted compared to Cu–Ta₃N₅ (∼0.5 V_RHE), which is due to the incorporation of Zr.⁴⁹ Notably, when Zr is gradient-doped into Cu–Ta₃N₅, not only is there a cathodic shift in the onset potential, but also the photocurrent density is significantly enhanced relative to that of Cu,Zr_h–Ta₃N₅. This can be ascribed to the gradient doping of Zr forming a gradient bend band structure that enhances carrier separation. Owing to the high photocurrent density, low onset potential and high fill factor, the Cu,Zr_g–Ta₃N₅ photoanodes reached a maximum ABPE of 3.5% (Fig. 4d and S14b†), which compares favorably with the previous literature on Ta₃N₅-based photoanodes, to the best of our knowledge.^34,55,56 This demonstrates the exceptional performance of Cu,Zr_g–Ta₃N₅ for PEC applications. Fig. S15† shows the IPCEs of pristine Ta₃N₅ and Cu,Zr_g–Ta₃N₅. The IPCE onset wavelength of pristine Ta₃N₅ and Cu,Zr_g–Ta₃N₅ are around 600 nm, corresponding to a band gap of ∼2.1 eV. The value of Cu,Zr_g–Ta₃N₅ (maximum value 72%) is significantly higher than that of pristine Ta₃N₅, implying that gradient Zr doping enhances the charge carrier separation/transfer efficiency. The oxygen production of the Cu,Zr_g–Ta₃N₅ photoanode at 1.0 V_RHE under AM 1.5 G was determined by gas chromatography (Fig. 4e). The continuous evolution of oxygen with a faradaic efficiency (FE) of close to 100% manifests that the high and stable photocurrent is primarily used for oxygen production rather than for other side reactions.

Fig. 4 (a) Schematic diagram of carrier migration and separation under simulated sunlight and bias; (b) steady-state photocurrent of the NiCoFe–B_i/Cu,Zr_g–Ta₃N₅/FTO photoanode under low applied potentials; (c) J–V curves of Ta₃N₅, Cu,Zr_h–Ta₃N₅ and Cu,Zr_g–Ta₃N₅ photoanodes with a NiCoFe–B_i co-catalyst in 1 M KOH under AM 1.5 G simulated sunlight; (d) ABPE of the photoanodes calculated from the J–V curves in (c); (e) amount of oxygen evolved from the NiCoFe–B_i/Cu,Zr_g–Ta₃N₅/FTO photoanode under an applied potential of 1.0 V versus RHE; (f) comparison of the photocurrent and onset potential between our NiCoFe–B_i/Cu,Zr_g–Ta₃N₅/FTO and other Ta₃N₅-based photoanodes deposited on FTO. The left vertical axis represents the photocurrent at 1.23 V_RHE, while the right vertical axis represents the onset potential of the comparison sample; (g) steady-state photocurrents of Ta₃N₅, Cu,Zr_h–Ta₃N₅ and Cu,Zr_g–Ta₃N₅ with a NiCoFe–B_i co-catalyst at 1.0 V versus RHE under AM 1.5 G simulated sunlight.

Photoanode stability is another crucial criterion by which to evaluate the PEC performance. The stability of the Cu,Zr_g–Ta₃N₅ photoanode in 1 M KOH was tested under 1 sun at 1.0 V_RHE (Fig. 4g and S14c†). The steady-state photocurrent density was stable for 180 min for Cu,Zr_g–Ta₃N₅, which is consistent with the J–V curve at 1.0 V_RHE shown in Fig. 4c. The photocurrent density of Cu,Zr_g–Ta₃N₅ still remained at 86% of the initial value after 10 h (Fig. S16†). Although pristine Ta₃N₅, Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ demonstrated stable currents initially, their performance was not sustainable over time. After only 30 min, the photocurrent of pristine Ta₃N₅ abruptly dropped, while the stabilities of Cu–Ta₃N₅ and Cu,Zr_h–Ta₃N₅ were also limited to just 67 and 88 min, respectively. The stability of the Cu,Zr_g–Ta₃N₅ photoanodes after 3 h was studied in more detail. As shown in Fig. S17a,† there was no obvious change in XRD patterns, indicating that Cu,Zr_g–Ta₃N₅ exhibits excellent stability. The SEM image illustrates that Cu,Zr_g–Ta₃N₅ retains its morphology after stability testing (Fig. S17b†). The XPS results after the durability tests were not significantly different from those before the reaction, which also provided strong evidence for the stability of the photoanode (Fig. S18†). The outstanding structural stability of the Cu,Zr–Ta₃N₅ photoanodes may be attributed to two reasons: (1) Zr is co-doped with O_N;⁴⁴ (2) gradient doped Zr and interstitial doped Cu generate a gradually increasing interaction. Contrastingly, the XPS and SEM results of pristine Ta₃N₅ exhibit a clear chemical state change and structural collapse (Fig. S19 and S20†).

Conclusions

In summary, Ta₃N₅ nanoparticles with Cu and gradient Zr doping were prepared using a two-step method. We found that due to the difference in electronegativity, interstitially doped Cu acts as an anchor site to interact with Zr. Therefore, the morphology of the Cu,Zr_g–Ta₃N₅ becomes more compact, and the compact morphology reduces the combination rate of carriers at the grain boundaries, which is conducive to improving the PEC performance. In addition, as doping Zr can alter the band edge positions, a Zr gradient-doping method was proposed to modify the band structure of Ta₃N₅, which enhanced the separation of carriers. Due to the gradual increase in the doped Zr content from the inside out, the interaction between Zr and Cu increased from the outside in, resulting in a denser morphology of Cu,Zr_g–Ta₃N₅ compared to that of Cu,Zr_h–Ta₃N₅. The resultant Cu,Zr_g–Ta₃N₅ possesses superior structural stability and exhibits extraordinary PEC performance. The use of co-doping to modify the morphology and engineer the band structure paves the way for other nanoparticle semiconductor light absorbers to improve PEC performance.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Author contributions

Xiaodong Wang and Huijuan Zhang conducted the project and contributed equally to this work. Chuanzhen Feng performed the photoelectrochemical tests. Yu Wang supervised the project and conceived the idea.

Conflicts of interest

The authors declare no conflict of interest.

Acknowledgements

This study was supported by the Fundamental Research Funds for the Central Universities (0301005202017, 2018CDQYFXCS0017, 106112017CDJXSYY0001), the Thousand Young Talents Program of the Chinese Central Government (Grant No. 0220002102003), the National Natural Science Foundation of China (NSFC, Grant No. 22371022, 22271029, U19A20100, 21971027, 21373280, 21403019), the Beijing National Laboratory for Molecular Sciences (BNLMS) and Hundred Talents Program at Chongqing University (Grant No. 0903005203205), the State Key Laboratory of Mechanical Transmissions Project (SKLMTZZKT-2017M11), the Natural Science Foundation of Chongqing (Grant No. cstc2019jcyj-msxmX0426), and the Science and Technology Research Project of Education Agency of Chongqing (Grant No. KJZD-K201800102).

References

1. Y. He, P. Ma, S. Zhu, M. Liu, Q. Dong, J. Espano, X. Yao and D. Wang, Joule, 2017, 1, 831–842.
2. M. Volokh, G. Peng, J. Barrio and M. Shalom, Angew Chem. Int. Ed. Engl., 2019, 58, 6138–6151.
3. J. Fu, F. Wang, Y. Xiao, Y. Yao, C. Feng, L. Chang, C.-M. Jiang, V. F. Kunzelmann, Z. M. Wang, A. O. Govorov, I. D. Sharp and Y. Li, ACS Catal., 2020, 10, 10316–10324.
4. T. Higashi, H. Nishiyama, V. Nandal, Y. Pihosh, Y. Kawase, R. Shoji, M. Nakabayashi, Y. Sasaki, N. Shibata, H. Matsuzaki, K. Seki, K. Takanabe and K. Domen, Energy Environ. Sci., 2022, 15, 4761–4775.
5. Y. Qiu, Z. Liu, A. Sun, X. Zhang, X. Ji and J. Liu, ACS Sustain. Chem. Eng., 2022, 10, 16417–16426.
6. L. Zhang, L. Li, J. Liang, X. Fan, X. He, J. Chen, J. Li, Z. Li, Z. Cai, S. Sun, D. Zheng, Y. Luo, H. Yan, Q. Liu, A. A. Alshehri, X. Guo, X. Sun and B. Ying, Inorg. Chem. Front., 2023, 10, 2766–2775.
7. J. Chen, L. Zhang, J. Li, X. He, Y. Zheng, S. Sun, X. Fang, D. Zheng, Y. Luo, Y. Wang, J. Zhang, L. Xie, Z. Cai, Y. Sun, A. A. Alshehri, Q. Kong, C. Tang and X. Sun, J. Mater. Chem. A, 2023, 11, 1116–1122.
8. K. Sivula and R. van de Krol, Nat. Rev. Mater., 2016, 1, 15010.
9. V. Nandal, Y. Pihosh, T. Higashi, T. Minegishi, T. Yamada, K. Seki, M. Sugiyama and K. Domen, Energy Environ. Sci., 2021, 14, 4038–4047.
10. W. Yang, R. R. Prabhakar, J. Tan, S. D. Tilley and J. Moon, Chem. Soc. Rev., 2019, 48, 4979–5015.
11. L. Jin, F. Cheng, H. Li and K. Xie, Angew. Chem., Int. Ed., 2020, 59, 8891–8895.
12. B. A. Pinaud, P. C. K. Vesborg and T. F. Jaramillo, J. Phys. Chem. C, 2012, 116, 15918–15924.
13. M. Xiao, Z. Wang, B. Luo, S. Wang and L. Wang, Appl. Catal., B, 2019, 246, 195–201.
14. L. Wang, X. Zhou, N. T. Nguyen, I. Hwang and P. Schmuki, Adv. Mater., 2016, 28, 2432–2438.
15. H. Hajibabaei, D. J. Little, A. Pandey, D. Wang, Z. Mi and T. W. Hamann, ACS Appl. Mater. Interfaces, 2019, 11, 15457–15466.
16. M. Xiao, B. Luo, M. Lyu, S. Wang and L. Wang, Adv. Energy Mater., 2017, 8, 1701605.
17. T. Higashi, H. Nishiyama, Y. Suzuki, Y. Sasaki, T. Hisatomi, M. Katayama, T. Minegishi, K. Seki, T. Yamada and K. Domen, Angew Chem. Int. Ed. Engl., 2019, 58, 2300–2304.
18. P. Wang, C. Ding, Y. Deng, H. Chi, H. Zheng, L. Liu, H. Li, Y. Wu, X. Liu, J. Shi and C. Li, ACS Catal., 2023, 13, 2647–2656.
19. Y. Li, L. Zhang, A. Torres-Pardo, J. M. Gonzalez-Calbet, Y. Ma, P. Oleynikov, O. Terasaki, S. Asahina, M. Shima, D. Cha, L. Zhao, K. Takanabe, J. Kubota and K. Domen, Nat. Commun., 2013, 4, 2566.
20. G. Liu, J. Shi, F. Zhang, Z. Chen, J. Han, C. Ding, S. Chen, Z. Wang, H. Han and C. Li, Angew Chem. Int. Ed. Engl., 2014, 53, 7295–7299.
21. P. Wang, P. Fu, J. Ma, Y. Gao, Z. Li, H. Wang, F. Fan, J. Shi and C. Li, ACS Catal., 2021, 11, 12736–12744.
22. Y. Kado, R. Hahn, C.-Y. Lee and P. Schmuki, Electrochem. Commun., 2012, 17, 67–70.
23. Y. Kado, C. Y. Lee, K. Lee, J. Muller, M. Moll, E. Spiecker and P. Schmuki, Chem. Commun., 2012, 48, 8685–8687.
24. S. S. Ma, T. Hisatomi, K. Maeda, Y. Moriya and K. Domen, J. Am. Chem. Soc., 2012, 134, 19993–19996.
25. Y. W. Kim, S. Cha, I. Kwak, I. S. Kwon, K. Park, C. S. Jung, E. H. Cha and J. Park, ACS Appl. Mater. Interfaces, 2017, 9, 36715–36722.
26. J. Fu, Z. Fan, M. Nakabayashi, H. Ju, N. Pastukhova, Y. Xiao, C. Feng, N. Shibata, K. Domen and Y. Li, Nat. Commun., 2022, 13, 729.
27. M. Zhong, T. Hisatomi, Y. Sasaki, S. Suzuki, K. Teshima, M. Nakabayashi, N. Shibata, H. Nishiyama, M. Katayama, T. Yamada and K. Domen, Angew. Chem., Int. Ed., 2017, 56, 4739–4743.
28. S. Chen, S. Shen, G. Liu, Y. Qi, F. Zhang and C. Li, Angew. Chem., Int. Ed., 2015, 54, 3047–3051.
29. L. Pei, B. Lv, S. Wang, Z. Yu, S. Yan, R. Abe and Z. Zou, ACS Appl. Energy Mater., 2018, 1, 4150–4157.
30. L. Pei, H. Wang, X. Wang, Z. Xu, S. Yan and Z. Zou, Dalton Trans., 2018, 47, 8949–8955.
31. Z. Wang, Y. Qi, C. Ding, D. Fan, G. Liu, Y. Zhao and C. Li, Chem. Sci., 2016, 7, 4391–4399.
32. M. Liao, J. Feng, W. Luo, Z. Wang, J. Zhang, Z. Li, T. Yu and Z. Zou, Adv. Funct. Mater., 2012, 22, 3066–3074.
33. M. Higashi, K. Domen and R. Abe, Energy Environ. Sci., 2011, 4, 4138.
34. Y. Xiao, C. Feng, J. Fu, F. Wang, C. Li, V. F. Kunzelmann, C.-M. Jiang, M. Nakabayashi, N. Shibata, I. D. Sharp, K. Domen and Y. Li, Nat. Catal., 2020, 3, 932–940.
35. F. F. Abdi, L. Han, A. H. Smets, M. Zeman, B. Dam and R. van de Krol, Nat. Commun., 2013, 4, 2195.
36. Y. Lin, Y. Xu, M. T. Mayer, Z. I. Simpson, G. McMahon, S. Zhou and D. Wang, J. Am. Chem. Soc., 2012, 134, 5508–5511.
37. Y. Wu, X. Liu, H. Zhang, J. Li, M. Zhou, L. Li and Y. Wang, Angew Chem. Int. Ed. Engl., 2021, 60, 3487–3492.
38. C. Wang, T. Hisatomi, T. Minegishi, M. Nakabayashi, N. Shibata, M. Katayama and K. Domen, Chem. Sci., 2016, 7, 5821–5826.
39. Y. Xiao, Z. Fan, M. Nakabayashi, Q. Li, L. Zhou, Q. Wang, C. Li, N. Shibata, K. Domen and Y. Li, Nat. Commun., 2022, 13, 7769.
40. W. Fu, Y. Wang, J. Hu, H. Zhang, P. Luo, F. Sun, X. Ma, Z. Huang, J. Li, Z. Guo and Y. Wang, Angew Chem. Int. Ed. Engl., 2019, 58, 17709–17717.
41. C. Feng, Y. Wu, Y. Fu, H. Zhang and Y. Wang, Appl. Surf. Sci., 2022, 605, 154810.
42. Y.-R. Chen, Y. Chen, P.-H. Chiu, P.-H. Hsiao and C.-Y. Chen, ACS Appl. Nano Mater., 2022, 5, 8962–8972.
43. S. Hwang, S. H. Porter, M. Li, R. Thorpe, A. B. Laursen, H. Gu, A. Safari, M. Greenblatt, E. Garfunkel and G. C. Dismukes, J. Phys. Chem. C, 2022, 126, 5970–5979.
44. J. Wang, Y. Jiang, A. Ma, J. Jiang, J. Chen, M. Li, J. Feng, Z. Li and Z. Zou, Appl. Catal., B, 2019, 244, 502–510.
45. K. Li and D. Xue, J. Phys. Chem. A, 2006, 110, 11332–11337.
46. J. Xiao, J. J. M. Vequizo, T. Hisatomi, J. Rabeah, M. Nakabayashi, Z. Wang, Q. Xiao, H. Li, Z. Pan, M. Krause, N. Yin, G. Smith, N. Shibata, A. Bruckner, A. Yamakata, T. Takata and K. Domen, J. Am. Chem. Soc., 2021, 143, 10059–10064.
47. Z. Wang, Y. Inoue, T. Hisatomi, R. Ishikawa, Q. Wang, T. Takata, S. Chen, N. Shibata, Y. Ikuhara and K. Domen, Nat. Catal., 2018, 1, 756–763.
48. Y. Wang, D. Zhu and X. Xu, ACS Appl. Mater. Interfaces, 2016, 8, 35407–35418.
49. J. Seo, T. Takata, M. Nakabayashi, T. Hisatomi, N. Shibata, T. Minegishi and K. Domen, J. Am. Chem. Soc., 2015, 137, 12780–12783.
50. T. Takata, G. Hitoki, J. N. Kondo, M. Hara, H. Kobayashi and K. Domen, Res. Chem. Intermed., 2007, 33, 13–25.
51. X. Wang, X. Liu, Y. Wu, Y. Fu, H. Zhang, M. Zhou and Y. Wang, Appl. Catal., B, 2023, 323, 122182.
52. C. Feng, F. Wang, Z. Liu, M. Nakabayashi, Y. Xiao, Q. Zeng, J. Fu, Q. Wu, C. Cui, Y. Han, N. Shibata, K. Domen, I. D. Sharp and Y. Li, Nat. Commun., 2021, 12, 5980.
53. S. Wei, S. Chang, J. Qian and X. Xu, Small, 2021, 17, 2100084.
54. K. Zhang, B. Jin, C. Park, Y. Cho, X. Song, X. Shi, S. Zhang, W. Kim, H. Zeng and J. H. Park, Nat. Commun., 2019, 10, 2001.
55. G. Liu, S. Ye, P. Yan, F. Xiong, P. Fu, Z. Wang, Z. Chen, J. Shi and C. Li, Energy Environ. Sci., 2016, 9, 1327–1334.
56. Y. Pihosh, T. Minegishi, V. Nandal, T. Higashi, M. Katayama, T. Yamada, Y. Sasaki, K. Seki, Y. Suzuki, M. Nakabayashi, M. Sugiyama and K. Domen, Energy Environ. Sci., 2020, 13, 1519–1530.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3sc05420a

‡ These authors contributed equally to this work.

---