In recent years, the depletion of fossil fuels has led to the impetus to develop new energy sources [1, 2]. Solar water splitting represents a sustainable and environmentally friendly method to produce H₂, which is considered an alternative, clean, and sustainable energy source to fossil fuels [3, 4]. Photocatalytic water splitting has attracted considerable attention as a potential route to produce renewable energy with no reliance on fossil fuels and no carbon dioxide emission [5-7]. In 1972, Fujishima etal. [8] first reported TiO₂ as a photocatalyst for solar water splitting. Since then, TiO₂ has been extensively studied as a promising photocatalyst. Although TiO₂ possesses remarkable advantages, it suffers from low solar light utilization efficiency because of its wide band gap [9]. Some non-TiO₂ photocatalysts, including ferric oxide [10-12], cuprous oxide [13, 14], cadmium sulfide [15, 16], and composite materials, have been used as visible-light photocatalysts to drive water splitting under visible-light irradiation [17-26]. However, these photocatalysts are susceptible to photocorrosion [21, 22]. Thus, a stable visible-light-driven photocatalyst for use in water splitting is still urgently required.

Recently, WO₃ has been proved an attractive candidate for photocatalytic applications in environmental remediation [23-31] and O₂ evolution [32-34] because of its suitable band gap to absorb visible light and excellent stability [35-39]. Although WO₃ possesses an indirect band gap energy of 2.7–2.8 eV, it still cannot directly reduce water to H₂ because of its high conduction band position (0.3 V vs. RHE) [40]. As a result, the photogenerated electrons do not reach the potential of water reduction (0 V vs. RHE) [41]. In addition, WO₃ still suffered from the rapid recombination of photogenerated electron–hole pairs in the photocatalytic process, like other powder photocatalysts [32, 42, 43].

Photoelectrocatalytic (PEC) water splitting has been proved an effective method to solve the problem of the rapid recombination of photogenerated electron–hole pairs [44-46]. In a typical PEC water splitting system, photogenerated holes (or electrons) migrate to the anode (or cathode) surface to participate in water oxidation (or reduction) reactions [47]. The external field induces the photogenerated electron–hole pairs to separate rapidly, thus slowing the recombination rate of charge carriers [48]. Meanwhile, the applied bias voltage allows the photogenerated electrons to reduce water to H₂ [41]. Recently, photoanode-driven PEC systems have been widely used in water splitting [47, 45-51]. Considerable effort has been focused on the development of n-type semiconductor-based photoanodes for solar-light-driven water oxidation in PEC systems[41, 52]. For example, highly crystalline WO₃ with small mesopores produced via a one-step procedure demonstrated high visible-light-driven PEC performance in water oxidation [53]. Pore-rich WO₃ ultrathin nanosheets were fabricated as an effective photoanode for PEC water oxidation, displaying more photogenerated holes, higher carrier migration rate, shorter migration path, and stronger oxidizing capability than normal nanostructured WO₃. More recently, a porous WO₃ photoanode was used for the photoelectrochemical mineralization of emerging contaminants [54].

Regarding overall water splitting, few reports on WO₃-based photoanodes have appeared to date. In the present work, a nanotube array-like monoclinic WO₃/W(NA-WO₃/W) electrode fabricated by an electrochemical anodization method is used as an effective photoanode to realize overall water splitting (stoichiometric H₂/O₂ ratio of 2:1) under visible-light irradiation (λ = 420 nm). We fabricated the first WO₃ nanotube-based electrodes by an anodization route [55, 56]. Various morphologies, including nanopores, grains, and even formless WO₃ thin films and crystals can be produced by adjusting the anodization voltage, time, and electrolyte composition of the anodization process and the annealing temperature. Both the photoelectrochemical properties and PEC activity of the NA-WO₃/W photoanode are investigated using a commercial WO₃/W film as a standard photoanode. The NA-WO₃/W photoanode exhibits stronger visible-light absorption, lower charge recombination rate, and higher stability than the reference photoanode, leading to higher PEC activity in visible-light-driven overall water splitting; the quantum efficiency (QE) of the NA-WO₃/W photoanode in water splitting is approximately 5.23%. Such greatly enhanced PEC water-splitting performance is attributed to the highly crystalline WO₃ nanotube arrays, fast carrier separation and migration rates, and high visible-light absorption capability of the NA-WO₃/W photoanode.

W foil (20.0 × 30.0 × 0.1 mm, 99.95% purity, Chuang Qi, Baoji, China) was degreased by ultrasonication in acetone, deionized water, and ethanol in sequence for 15 min each and then dried in a nitrogen (N₂) stream. The WO₃/W nanostructures were fabricated through constant-voltage anodization in a two-electrode electrochemical cell at room temperature using the cleaned W foil as the anode and Pt foil (20.0 × 20.0 × 0.1 mm) as the cathode. The W foil was anodized in ammonium fluoride (NH₄F; 0.15 mol/L) aqueous electrolyte solution containing glycerol (glycerol/water = 50:50 vol %) under a DC oxidation voltage of 40 V for 1 h. The as-obtained WO₃/W film was annealed at 400 ℃ for 3 h in an air atmosphere after heating at a rate of 2 ℃/min to form NA-WO₃/W. The effect of the electrochemical parameters, including the concentration of NH₄F (0.10–0.20 mol/L), oxidation voltage (30–70 V), and oxidation time (0.5–2.0 h), and the annealing temperature (300–500 ℃) on the PEC activities of NA-WO₃/W photoanodes during visible-light-driven water splitting were investigated.

To aid comparison, a reference WO₃/W photoanode was also prepared by coating commercial WO₃ powder on W foil. In a typical process, polyethylene glycol (PEG; average MW 20000, 20 mg) was dissolved in ethanol (100 µL). Commercial WO₃nanoparticles (< 200 nm, 99.9% metals basis, Aladdin, 100 mg) were dispersed in the PEG solution by grinding for 15 min. The as-obtained mixture was coated on W foil (20.0×30.0×0.1 mm, 99.95% purity, Chuang Qi) by the doctor-blade method to give an active area of 15.0×25.0 mm. Film thickness was controlled using plastic tape with a thickness of 20 µm. The WO₃ electrode was annealed at 400 ℃ for 3.0 h in an air atmosphere following heating at a rate of 5 ℃/min [57]. The as-prepared reference WO₃ electrode is called standard WO₃.

Wide-angle X-ray diffraction (XRD) measurements were carried out in a parallel mode (ω = 0.5°, 2θ from 10° to 80°) using a Rigaku D/max-3C Advance X-ray diffractometer with Cu K_α radiation (λ = 1.5406 Å ). The morphologies of the products were observed by field-emission scanning electron microscopy (FESEM; S-4800, Hitachi, Japan). Surface electronic states were analyzed by X-ray photoelectron spectroscopy (XPS; PHI 5000, Perkin-Elmer, USA). All binding energy values were calibrated using the C 1s peak at 284.6 eV as a reference. Ultraviolet-visible diffuse reflectance spectra (UV-vis DRS) of the samples over a range of 200–800 nm were collected on Dilor Super Lab Ram II and MC-2530 spectrometers using barium sulfate as a reference. Transmission electron microscopy (TEM; JEM-2010, JEOL, Japan, 200 kV) and high-resolution transmission electron microscopy (HRTEM; JEOL-2010F, JEOL, 200 kV) were used to analyze the structure of the photoelectrodes.

Photoelectrochemical measurements were carried out in a standard three-electrode system with a custom-made single-compartment quartz cell on an electrochemical work station (CHI 660D, Shanghai Chenhua Instrument Inc., China). A WO₃/W electrode with an active area of 5 cm² served as the photoanode and a Pt sheet (4 cm²) served as the cathode. A saturated calomel electrode (SCE) was used as the reference electrode. An aqueous solution of sodium sulfate (Na₂SO₄; 0.5 mol/L, 60 mL) was used as the supporting electrolyte. A light-emitting diode (LED; λ = 420 nm, 15 mW/cm²) was used as the visible-light source. The NA-WO₃/W photoanode was positioned 5.0 cm away from the light source. Impedance measurements were collected under simulated solar illumination in 0.5 mol/L Na₂SO₄ solution at open-circuit potential over a frequency range from 10⁶ to 10⁻² Hz with an AC voltage of 5 mV. Mott-Schottky plots were obtained at a fixed frequency of 1 kHz to determine flat-band potentials (E_FB). Transient photocurrents were measured using a 20 s on/off cycle at a bias voltage of 0.6 V vs. SCE. Incident photon-to-current efficiency (IPCE) measurements were conducted in a standard three-electrode system using an electrochemical workstation (Zahner-Elektrik) with an applied potential of 0.6 V vs. SCE. The light wavelength range was from 367 to 720 nm. The absorbed photon-to-current efficiency (APCE) was obtained by dividing the IPCE by the light harvesting efficiency at each wavelength using a reported calculation method [58].

A custom-made quartz reactor with two chambers each with a volume of 120 mL (Fig. 1) was used to evaluate the photoelectrochemical water-splitting activity of the photoanodes. Each chamber of the reactor was filled with 0.5 mol/L Na₂SO₄ solution (75 mL). A WO₃/W electrode was used as the photoanode, Pt foil as the counter electrode, and SCE as the reference electrode. An external bias voltage (0.5 V vs. SCE) was applied to the three-electrode system. An LED (λ = 420 nm, 15 mW/cm²) was used as the visible-light source. The system was purged with a high-purity N₂ gas to remove dissolved O₂ before reaction. To quantify the amounts of evolved H₂ and O₂, 0.5 mL of the gas in the reactor was sampled through the septum after 1 h of reaction. Both O₂ and H₂ were analyzed by gas chromatography (GC9800, Shanghai Kechuang Chromatograph Instruments Co. Ltd, China) using N₂ (99.999%) as the carrier gas. O₂ was detected using a 5 A molecular sieve column and H₂ using a TDX-1 column. The reproducibility was checked by repeating the measurements ten times. QE was calculated using Eq. (1),

where n is the number of electrons transferred; R is the yield of production; N_A is the Avogadro constant; t₁ is the time (h); E is the monochromatic light intensity; t₂ is the time (s); A is the irradiation area; λ is the wavelength; h was the Planck constant; and c is the speed of light.

The NA-WO₃/W thin film electrode was fabricated via an anodic oxidation route. During electrode fabrication, anodization was carried out by a DC-regulated power supply with a two-electrode setup [59] using W foil as the anode and Pt foil as the cathode. To investigate the effect of different oxidation conditions on the nanostructure of the WO₃/W film, parameters including the oxidation voltage, time, and concentration of fluoride ions were adjusted. The morphology of the WO₃/W electrode was characterized by FESEM. Fig. 2(a) and (b) show FESEM images of the NA-WO₃/W thin film. The average diameter of the nanotubes was about 120 nm and their wall thickness was about 10 nm. The HRTEM image in Fig. 2(c) of broken grains obtained from WO₃ nanotubes reveals clear lattice fringes of 0.375 and 0.380 nm derived from the (020) and (202) planes of monoclinic WO₃ crystals, respectively [60].

To further confirm the crystal structure of the NA-WO₃/W thin film, it was investigated by XRD. As shown in Fig. 3(a), the characteristic diffraction peaks of the WO₃/W sample suggested the formation of pure monoclinic WO₃ (PDF#43-1035) on the W foil, consistent with the HRTEM results. XPS data were recorded to determine the chemical composition and electronic binding energies of W and O of the sample. The survey spectrum of the WO₃/W sample over a large energy range at low resolution is presented in Fig. 3(b). Two peaks were observed at 35.7 and 37.9 eV, which are assigned to W 4f _7/2 and W 4f _5/2, respectively [61]. These results indicate that monoclinic WO₃ crystals were present on the surface of the NA-WO₃/W photoelectrode, which was supported by the XRD analysis.

For further investigate the surface electronic state of the samples, UV-vis spectra were recorded, as shown in Fig. 4(a). The band gap of the NA-WO₃/W electrode estimated from a Tauc plot was about 2.35 eV, which was slightly smaller than the band gap of standard WO₃/W of about 2.45 eV (Fig. 4(a)) [62, 63]. The NA-WO₃/W photoanode exhibited broad visible-light absorption. The weak absorption of NA-WO₃/W at longer wavelength (λ > 520 nm) was ascribed to the light scattering by pores or cracks in the nanotube arrays [64]. IPCE was measured at 0.6 V vs. SCE. The NA-WO₃/W and standard WO₃/W electrodes exhibited IPCE values of 0.5% and 0.36% at 420 nm, respectively (Fig. 4(b)).

The photoelectrochemical properties of the electrodes were then examined in 0.5 mol/L Na₂SO₄ solution containing 25 vol % methanol as a hole scavenger [65, 66]. The oxidation of methanol is faster than that of water both thermodynamically and kinetically. Therefore, the photocurrent measured for methanol oxidation enabled investigation of the photoelectrochemical properties of the NA-WO₃/W electrode independent of its poor water oxidation kinetics [65]. To evaluate the photoelectrochemical response of the electrodes under visible-light irradiation, their photocurrent density was also measured under light pulses with a duration of 20 s. At an applied potential of 0.6 V vs. SCE, the NA-WO₃/W electrode exhibited a high and stable photocurrent, while that of the standard WO₃/W sample was much smaller and unstable, indicating that the photogenerated carriers in the standard WO₃/W sample recombined easily (Fig. 5(a)). The photocurrent response of the NA-WO₃/W electrode presented good reproducibility for numerous on/off cycles. The enhanced photocurrent density of the NA-WO₃/W electrode indicated that the photogenerated electrons could be transferred to the cathode and the holes could be injected into the 0.5 mol/L Na₂SO₄ solution much more readily by the NA-WO₃/W electrode than by the standard WO₃/W one. This is because the tubular structure of the NA-WO₃/W electrode allowed multiple reflections and offered better proton channels and lower electron transfer resistance than the standard WO₃/W sample. In terms of the photocurrent of standard WO₃/W, it decreased for a time before becoming steady, which was attributed to the backflow of photogenerated electrons from the conduction band of WO₃ to its valence band. Thus, a small voltage was used to inhibit this phenomenon in the PEC process.

The interfacial properties between the electrode and electrolyte were also investigated by electrochemical impedance spectroscopy (EIS), as shown in Fig. 5(b). The semicircle at high frequency in a Nyquist plot may illustrate the charge-transfer process; the charge-transfer resistance depends on the diameter of the semicircle. The capacitance arc for the NA-WO₃/W electrode under simulated sunlight irradiation was considerably smaller than that of the standard WO₃/W electrode. This implies that the conductivity of the NA-WO₃/W electrode was enhanced because the recombination of photogenerated electron–hole pairs was lowered, which in turn enhanced the photocurrent density [67].

Typical photocurrent-potential (J-V) curves for methanol oxidation with the NA-WO₃/W and standard WO₃/W electrode are shown in Fig. 6(a). The NA-WO₃/W electrode exhibited a higher photocurrent and faster increase in photocurrent than standard WO₃/W in the range of 0.13 to 0.6 V vs. SCE, representing a much higher fill factor. The photocurrent density of the NA-WO₃/W electrode was 0.38 mA/cm² (0.6 V vs. SCE) [58]. The photocurrent density obtained for methanol oxidation was used to calculate the electron–hole separation yield (ϕ_sep) using Eq. 2,

(2)

where J_PEC is the photocurrent density and J_abs is the photon absorption rate expressed as current density, which was calculated assuming 100% APCE (refer to the literature for calculation details, materials, and methods [62, 53-71]). J_abs of the NA-WO₃/W and standard WO₃/W electrodes under LED irradiation (λ = 420 nm, 15 mW/cm²) were calculated to be 2.53 and 1.83 mA/cm², respectively [72]. ϕ_sep was also the yield of photogenerated holes that reached the surface and ϕ_ox was the yield of holes that reached the surface and were injected into the solution species [73].

For methanol oxidation, which has extremely fast kinetics, surface recombination was negligible and ϕ_ox was about 1. Therefore, ϕ_sep was obtained by dividing J_PEC by J_abs (Fig. 6(a), inset). The results showed that the NA-WO₃/W electrode achieved ϕ_sep = 0.15 at 0.6 V, and 0.25 at 1.2 V vs. SCE, which were about three times those of the standard WO₃ electrode. Such enhancement of ϕ_sep for the NA-WO₃/W photoanode could be attributed to its special nanotube array-like morphology, which may be favorable for carrier separation and migration to the electrode surface. Fig. 6(b) displays the Mott-Schottky plots of 1/C² as a function of the applied potential. A positive slope was observed for the sample, suggesting that NA-WO₃ is an n-type semiconductor. Furthermore, the plots were extrapolated to 1/C² = 0 to estimate E_FB, which were 0.87 and 0.65 V vs. SCE for the NA-WO₃/W and standard WO₃/W electrodes, respectively.

To evaluate the PEC water-splitting performance of the photoanode, the amounts of both H₂ and O₂ evolved under visible-light irradiation were detected. The influence of the NA-WO₃/W electrode synthesis parameters on water-splitting performance was investigated. Fig. 7(a) presents the results for a series of electrodes produced with different anodization times at the same anodization voltage and fluoride ion concentration (DC 40 V, 0.15 mol/L NH₄F). The sample produced after 1.0 h of anode oxidization exhibited the highest PEC water-splitting performance under visible-light irradiation of the electrodes. Fig. 7(b) shows the PEC performance of samples fabricated with various anodization voltages (30–70 V) and the same anodization time and fluoride ion concentration (1.0 h, 0.15 mol/L NH₄F). The results indicate that the sample oxidized at a DC voltage of 40 V gave the highest H₂ and O₂ evolution rates under visible-light irradiation of the electrodes. The effect of fluoride ion concentration in the electrolyte solution during anode oxidation was also investigated with the same anodization voltage and anodization time (40 V, 1.0 h) (Fig. 7(c)). The results showed that the sample prepared in the electrolyte with 0.15 mol/L of fluoride ions exhibited the highest PEC performance of the electrodes. Meanwhile, Fig. 7(d) revealed that the sample calcined at 400 ℃ exhibited the highest PEC performance for water-splitting of the electrodes. The NA-WO₃/W photoanode fabricated under the optimal conditions (DC 40 V, 1.0 h, 0.15 mol/L of fluoride ions, and 400 ℃) exhibited the highest PEC water-splitting performance under visible-light irradiation of the samples, with evolution rates of H₂ and O₂ of 3.93 and 1.96 µmol/cm²/h, respectively. This may be caused by the high electron–hole separation rate resulting from both the one-dimensional nanotube array-like structure of WO₃ and the high crystallinity of monoclinic WO₃. The QE of the optimal NA-WO₃/W photoanode under LED irradiation (λ = 420 nm, 15 mW/cm²) reached 5.23%.

To aid comparison, standard WO₃/W was used as a reference. Fig. 8(a) shows the H₂ and the O₂ evolution performance of both NA-WO₃/W and standard WO₃/W photoanodes in the PEC process for overall water splitting. The results indicated that the NA-WO₃/W photoanode exhibited an activity of about 3.0 µmol/cm²/h, which was much higher and more stable PEC activity for water splitting than that of the standard WO₃/W photoanode. No obvious decrease of activity was observed in the water-splitting process over 10 h. The XRD pattern of the NA-WO₃/W photoanode used for 10 h in Fig. 8(b) was similar to that of the unused photoanode, confirming the phase stability of the NA-WO₃/W film during PEC water splitting. An FESEM image further confirmed the nanotube structure was maintained after 10 h of PEC reaction (Fig. 8(c)). The high activity and stability of the NA-WO₃/W photoanode might be attributed to the formation of tubular structures of WO₃ directly on the W sheet. This tubular structure consisting of highly crystalline monoclinic WO₃ could allow preferable light absorption through multiple reflections in the WO₃ nanotubes, as well as rapid hole transport in the highly crystalline pore walls of WO₃.

A proposed mechanism for the PEC process for water splitting using NA-WO₃/W as the photoanode is illustrated in Fig. 9. Electrons and holes were generated upon irradiating the NA-WO₃/W photoanode with visible light. The photogenerated electrons were transferred to the cathode to evolve H₂ through the external circuit with the aid of the bias voltage. Meanwhile, the photogenerated holes in the valence band of the NA-WO₃/W photoanode could oxidize water to evolve O₂ because the valence band position of WO₃ was sufficiently positive [74].

In summary, a NA-WO₃/W photoanode was obtained via an electrochemical anodization method. The NA-WO₃/W photoanode exhibited excellent PEC activity for overall water splitting with a QE of approximately 5.23% under LED irradiation (λ = 420 nm, 15 mW/cm²). The high PEC performance of the NA-WO₃/W electrode was attributed to the one-dimensional nanotube-like structure of WO₃, highly crystalline monoclinic WO₃, and strong interaction between WO₃ and W foil. These features allowed both high visible-light absorption capability and fast electron–hole separation, thus leading to the high photocurrent density under visible-light irradiation. Our photoanode design could be extended to other semiconductorelectrodes based on intrinsic metal substrates for photoelectrochemical applications.