Improved Charge Separation in WO3/CuWO4 Composite Photoanodes for Photoelectrochemical Water Oxidation

Porous tungsten oxide/copper tungstate (WO3/CuWO4) composite thin films were fabricated via a facile in situ conversion method, with a polymer templating strategy. Copper nitrate (Cu(NO3)2) solution with the copolymer surfactant Pluronic®F-127 (Sigma-Aldrich, St. Louis, MO, USA, generic name, poloxamer 407) was loaded onto WO3 substrates by programmed dip coating, followed by heat treatment in air at 550 °C. The Cu2+ reacted with the WO3 substrate to form the CuWO4 compound. The composite WO3/CuWO4 thin films demonstrated improved photoelectrochemical (PEC) performance over WO3 and CuWO4 single phase photoanodes. The factors of light absorption and charge separation efficiency of the composite and two single phase films were investigated to understand the reasons for the PEC enhancement of WO3/CuWO4 composite thin films. The photocurrent was generated from water splitting as confirmed by hydrogen and oxygen gas evolution, and Faradic efficiency was calculated based on the amount of H2 produced. This work provides a low-cost and controllable method to prepare WO3-metal tungstate composite thin films, and also helps to deepen the understanding of charge transfer in WO3/CuWO4 heterojunction.


Introduction
In recent years, photoelectrochemical (PEC) water splitting has aroused tremendous research interest due to its sustainable and carbon neutral attributes for producing high energy density fuels [1][2][3][4]. A visible light active photoanode which utilizes more of the solar spectrum is highly desirable, and thus the earth-abundant α-Fe 2 O 3 (band gap 2.1 eV) and WO 3 (band gap 2.6 eV) have become popular candidates for water oxidation. Due to the intrinsic long-range antiferromagnetic order in α-Fe 2 O 3 , it has a short hole diffusion length (2-20 nm) [5,6] and low charge-carrier mobility (10´2-10´1 cm 2¨V´1¨S´1 ) [7]. In comparison, WO 3 has a much longer hole diffusion length (150 nm) and higher mobility (10 cm 2¨V´1¨S´1 ) [8,9]. Nevertheless, its limited utilization of the visible light

Synthesis and Characterization of Pristine WO 3 and WO 3 /CuWO 4 Composite Thin Films
In this report, three representative samples with 0, 4 and 8 runs of programmed dip coating of Cu 2+ solution are discussed, which gave rise to WO 3 , WO 3 /CuWO 4 composite and CuWO 4 materials respectively. XRD patterns in Figure 1 indicate that the WO 3 thin films obtained from magnetron sputtered metallic W thin film are of the monoclinic phase (JCPDS No. 43-1035), which is the most photocatalytically active phase for water oxidation [10]. For comparison, we also sputtered metallic W onto a normal glass slide, which produced the orthorhombic phase of WO 3 hydrate (JCPDS No.   [10], to the contrary, as shown in XRD patterns in Figure S1. It is thus believed that the crystalline layer of F:SnO 2 helped with the formation of monoclinic phase of WO 3 . Thin films of WO 3 /CuWO 4 and CuWO 4 showed the characteristic peaks of CuWO 4 (JCPDF 72-0616) [9]. UV-Vis/DRS results of the three samples are shown in Figure 1b. Pristine WO 3 absorbs light up to 464 nm, which is in accordance with the band gap of about 2.7 eV of monoclinic WO 3 . The light absorption of CuWO 4 extends to 540 nm, corresponding to the smaller band gap of CuWO 4

Synthesis and Characterization of Pristine WO3 and WO3/CuWO4 Composite Thin Films
In this report, three representative samples with 0, 4 and 8 runs of programmed dip coating of Cu 2+ solution are discussed, which gave rise to WO3, WO3/CuWO4 composite and CuWO4 materials respectively. XRD patterns in Figure 1 indicate that the WO3 thin films obtained from magnetron sputtered metallic W thin film are of the monoclinic phase (JCPDS No. 43-1035), which is the most photocatalytically active phase for water oxidation [10]. For comparison, we also sputtered metallic W onto a normal glass slide, which produced the orthorhombic phase of WO3 hydrate (JCPDS No.   [10], to the contrary, as shown in XRD patterns in Figure S1. It is thus believed that the crystalline layer of F:SnO2 helped with the formation of monoclinic phase of WO3. Thin films of WO3/CuWO4 and CuWO4 showed the characteristic peaks of CuWO4 (JCPDF 72-0616) [9]. UV-Vis/DRS results of the three samples are shown in Figure 1b. Pristine WO3 absorbs light up to 464 nm, which is in accordance with the band gap of about 2.7 eV of monoclinic WO3. The light absorption of CuWO4 extends to 540 nm, corresponding to the smaller band gap of CuWO4 (2.3 eV). The top and cross-section views of all the films observed by field emission scanning electron microscope (FESEM) are shown in Figure 2. A summary of the average thickness of WO3 and CuWO4 of each sample is provided in Table 1. Figure 2a,b shows the images of pristine WO3. The WO3 thin film was composed of densely assembled WO3 nanoparticles of irregular shape around 50-90 nm in diameter, and the average thickness of film was around 415 nm. Figure 2c,d are the images of WO3/CuWO4 composite thin film, with a porous layer of CuWO4 particles uniformly grown on the WO3 substrate. The thickness of WO3 layer was greatly reduced to 150 nm due to the in situ reaction with Cu 2+ ions to form CuWO4. The transmission electron microscope (TEM) image of the scraped particles from the composite film in Figure S2 displays a network-like structure in the thin film, corresponding to the porous CuWO4 nanostructure in FESEM observation. In Figure 2e,f, with further increased loading of Cu 2+ ions, the WO3 film was completely consumed, forming a porous layer of CuWO4 1 μm thick. The top and cross-section views of all the films observed by field emission scanning electron microscope (FESEM) are shown in Figure 2. A summary of the average thickness of WO 3 and CuWO 4 of each sample is provided in Table 1. Figure 2a,b shows the images of pristine WO 3 . The WO 3 thin film was composed of densely assembled WO 3 nanoparticles of irregular shape around 50-90 nm in diameter, and the average thickness of film was around 415 nm. Figure 2c,d are the images of WO 3 /CuWO 4 composite thin film, with a porous layer of CuWO 4 particles uniformly grown on the WO 3 substrate. The thickness of WO 3 layer was greatly reduced to 150 nm due to the in situ reaction with Cu 2+ ions to form CuWO 4 . The transmission electron microscope (TEM) image of the scraped particles from the composite film in Figure S2 displays a network-like structure in the thin film, corresponding to the porous CuWO 4 nanostructure in FESEM observation. In Figure 2e,f, with further increased loading of Cu 2+ ions, the WO 3 film was completely consumed, forming a porous layer of CuWO 4 1 µm thick. Ma teria ls 2016, 9, 348 4 of 12

Photoelectrochemical Performance of Thin Films
The pristine WO3, WO3/CuWO4, and CuWO4 thin films were used as photoanodes in a conventional three-electrode setup. Their PEC performance was investigated by measuring the photocurrent with back-illumination as shown in Figure 3. In the linear sweep voltammetry (LSV) results in Figure 3a, all the samples exhibited negligible photocurrent under dark condition. Composite thin films obtained from different runs are provided in Figure S3. At 1.20 VRHE, bare WO3 and CuWO4 were of very similar photocurrent density, and WO3/CuWO4 (0.45 mA/cm 2 ) showed a current density more than two times higher than WO3 and CuWO4 electrodes. As shown in the supporting information ( Figure S3), all the composite thin films demonstrate higher photocurrent compared to single component thin films. The relatively low photocurrent of the WO3 underlayer in our report could be due to the preparation condition and insufficient thickness. As we annealed the tungsten film in air to let it oxidize into WO3, the insertion of oxygen atoms will make the thin film expand. As a result, the back contact with fluorine doped tin oxide (FTO) could be adversely affected. In addition, the annealing temperature and time can also affect the oxygen vacancies and carrier density, which will subsequently influence the photocurrent of WO3 films. However, when a thin layer of WO3 was coupled with CuWO4, the photocurrent of the composite film was remarkably improved compared with the single phase. Figure 3b shows the on-off photocurrent profile of the composite WO3/CuWO4 under a constant bias of 1.20 VRHE, recorded over a duration of 600 s with interval of 5 s. The composite thin film showed a constant photocurrent though transient spikes were spotted for both samples, which could be caused by the surface-trapped photo-generated minority carriers which recombine with the photo-generated major carriers [32]. Figure 3c shows the IPCE

Photoelectrochemical Performance of Thin Films
The pristine WO 3 , WO 3 /CuWO 4 , and CuWO 4 thin films were used as photoanodes in a conventional three-electrode setup. Their PEC performance was investigated by measuring the photocurrent with back-illumination as shown in Figure 3. In the linear sweep voltammetry (LSV) results in Figure 3a, all the samples exhibited negligible photocurrent under dark condition. Composite thin films obtained from different runs are provided in Figure S3. At 1.20 V RHE , bare WO 3 and CuWO 4 were of very similar photocurrent density, and WO 3 /CuWO 4 (0.45 mA/cm 2 ) showed a current density more than two times higher than WO 3 and CuWO 4 electrodes. As shown in the supporting information ( Figure S3), all the composite thin films demonstrate higher photocurrent compared to single component thin films. The relatively low photocurrent of the WO 3 underlayer in our report could be due to the preparation condition and insufficient thickness. As we annealed the tungsten film in air to let it oxidize into WO 3 , the insertion of oxygen atoms will make the thin film expand. As a result, the back contact with fluorine doped tin oxide (FTO) could be adversely affected. In addition, the annealing temperature and time can also affect the oxygen vacancies and carrier density, which will subsequently influence the photocurrent of WO 3 films. However, when a thin layer of WO 3 was coupled with CuWO 4 , the photocurrent of the composite film was remarkably improved compared with the single phase. Figure 3b shows the on-off photocurrent profile of the composite WO 3  both samples, which could be caused by the surface-trapped photo-generated minority carriers which recombine with the photo-generated major carriers [32]. Figure 3c shows the IPCE curves for WO 3 , WO 3 /CuWO 4 and CuWO 4 electrodes measured at 1.20 V RHE . The WO 3 /CuWO 4 composite thin films had much higher IPCE values compared with WO 3 and CuWO 4 films. Though light absorption of CuWO 4 covered the wavelengths up to 540 nm, the identical wavelength ranges of the three thin films in the IPCE curves indicate that the charge carriers generated in CuWO 4 in the wavelength range 470-540 nm make little contribution to photocurrent generation [29,31]. The simulated photocurrents in Figure 3d for all samples were also calculated by integrating the IPCE spectra with a standard AM 1.5 G solar spectrum from Equation (1). The simulated photocurrent is independent of the light source and applied filters, and thus it is more accurate in evaluation of the PEC performance of the thin films. The values obtained were reasonable compared to measured photocurrents.
where λ is the wavelength of light in unit of nm and IPCE (λ) is measured and calculated as will be described in Experiment Section 3.5. Φ(λ) is the photon flux of sunlight in photons/m 2 /s. The photon flux can be measured from tabulated solar irradiance data, E(λ), via Φ(λ) = E(λ)/(1240/λ) [33].
Ma teria ls 2016, 9, 348 5 of 12 curves for WO3, WO3/CuWO4 and CuWO4 electrodes measured at 1.20 VRHE. The WO3/CuWO4 composite thin films had much higher IPCE values compared with WO3 and CuWO4 films. Though light absorption of CuWO4 covered the wavelengths up to 540 nm, the identical wavelength ranges of the three thin films in the IPCE curves indicate that the charge carriers generated in CuWO4 in the wavelength range 470-540 nm make little contribution to photocurrent generation [29,31]. The simulated photocurrents in Figure 3d for all samples were also calculated by integrating the IPCE spectra with a standard AM 1.5 G solar spectrum from Equation (1). The simulated photocurrent is independent of the light source and applied filters, and thus it is more accurate in evaluation of the PEC performance of the thin films. The values obtained were reasonable compared to measured photocurrents.
where λ is the wavelength of light in unit of nm and IPCE (λ) is measured and calculated as will be described in Experiment Section 3.5. Φ(λ) is the photon flux of sunlight in photons/m 2 /s. The photon flux can be measured from tabulated solar irradiance data, E(λ), via Φ(λ) = E(λ)/(1240/λ) [33]. In order to measure the band structures of WO3 and CuWO4, the Mott-Schottky measurement is shown in Figure 4a,b, and the result for WO3/CuWO4 composite thin film is shown in Figure S4. Given the band gap measured from UV-Vis absorption, the band structures of WO3 and CuWO4 are shown in Figure 4c. The conduction band of CuWO4 is located at +0.2 eV (vs. normal hydrogen electrode, NHE) and that of WO3 is located at +0.4 eV (vs. NHE), while valence band of CuWO4 is at +2.4 eV (vs. NHE), lower than that of WO3 at +3.0 eV (vs. NHE). Therefore, these two components can form a heterojunction pair, and photo-generated holes from the inner WO3 layer will be transferred to the outer layer of CuWO4 in WO3/CuWO4 composite electrode. In order to measure the band structures of WO 3 and CuWO 4 , the Mott-Schottky measurement is shown in Figure 4a,b, and the result for WO 3 /CuWO 4 composite thin film is shown in Figure S4. Given the band gap measured from UV-Vis absorption, the band structures of WO 3 and CuWO 4 are shown in Figure 4c. The conduction band of CuWO 4 is located at +0.2 eV (vs. normal hydrogen electrode, NHE) and that of WO 3 is located at +0.4 eV (vs. NHE), while valence band of CuWO 4 is at +2.4 eV (vs. NHE), lower than that of WO 3 at +3.0 eV (vs. NHE). Therefore, these two components can form a heterojunction pair, and photo-generated holes from the inner WO 3 layer will be transferred to the outer layer of CuWO 4 in WO 3 /CuWO 4 composite electrode.
Electrochemical impedance spectroscopy (EIS) measurements were carried out to evaluate the overall resistance of the three photoanodes, and is shown in Figure 4d under illumination and 1.2 VRHE bias. The semicircle in the medium-frequency region was attributed to the charge-transfer process. The diameter of the WO3/CuWO4 semicircle was the smallest among the three electrodes, which was in accordance with the LSV results [16,25]. CuWO4 had the largest semicircle diameter, indicating a very high charge transfer resistance in the thin film, which was possibly related to the poor intrinsic charge transfer property of CuWO4. However, when CuWO4 was coupled with WO3 to form a heterojunction composite anode, the photogenerated electrons in CuWO4 could be transferred to the WO3 underlayer, with good charge transport characteristics, and contribute to the reduced resistance. Thus, the composite electrode combined the excellent charge transfer characteristics of WO3 and good light absorption capability of CuWO4.  The carrier density can be calculated by where C SC , q, ε 0 , ε, N D , and E fb are capacitance, the electron charge, permittivity in vacuum, dielectric constant, donor carrier density and flat-band potential of the semiconductor, respectively. Using Equation (2), the carrier densities are 1.73ˆ10 19 for WO 3 and 3.36ˆ10 18 for CuWO 4 . Our low carrier density is possibly due to the synthesis method for the thin films. The literature has shown that annealing in air can affect both the oxygen vacancies and the types of other vacancies/defects within a nanostructured thin film [28]. Electrochemical impedance spectroscopy (EIS) measurements were carried out to evaluate the overall resistance of the three photoanodes, and is shown in Figure 4d under illumination and 1.2 V RHE bias. The semicircle in the medium-frequency region was attributed to the charge-transfer process. The diameter of the WO 3 /CuWO 4 semicircle was the smallest among the three electrodes, which was in accordance with the LSV results [16,25]. CuWO 4 had the largest semicircle diameter, indicating a very high charge transfer resistance in the thin film, which was possibly related to the poor intrinsic charge transfer property of CuWO 4 . However, when CuWO 4 was coupled with WO 3 to form a heterojunction composite anode, the photogenerated electrons in CuWO 4 could be transferred to the WO 3 underlayer, with good charge transport characteristics, and contribute to the reduced resistance. Thus, the composite electrode combined the excellent charge transfer characteristics of WO 3 and good light absorption capability of CuWO 4 .

Comparison of Absorption Efficiency (η abs ), and Charge Separation Efficiency (η sep ) of WO 3 , CuWO 4 and WO 3 /CuWO 4 Thin Films
Since the WO 3 /CuWO 4 electrode demonstrated remarkable improvement in photoelectrochemical and EIS measurement compared with pristine WO 3 substrate and CuWO 4 films, we further extracted the efficiency values of light absorption (η abs ) and charge separation (η sep ) based on Equations (3) and (4) to quantify the contribution of each factor [34,35].
where J abs is the photo current density calculated by multiplying η abs by the standard AM 1.5 G (100 mW/cm 2 ) solar spectrum, Φ(λ) is the photon flux of sunlight in photons/m 2 /s, e is the charge of an electron (C). J sca is the photocurrent of the photoelectrode in the presence of scavengers as a function of applied bias. By measuring the light transmittance and reflectance in an integrated sphere (see Figure S5), we were able to obtain the η abs of pristine WO 3 , CuWO 4 and WO 3 /CuWO 4 in Figure 5a. It shows that more photons can be absorbed in the presence of CuWO 4 . With the measured values of η abs and Equation (1), the integrated J abs over the AM 1.5 spectrum of pristine WO 3 , CuWO 4 and WO 3 /CuWO 4 films are 1.8, 4.7 and 3.2 mA/cm 2 , respectively. The charge separation efficiency (η sep ) can be determined by adding the hole scavenger H 2 O 2 to the electrolyte. The presence of H 2 O 2 increased photocurrent density of all the three thin films ( Figure S6), which was due to the much faster charge transfer rate promoted by the hole scavenger. According to Equation (4), charge separation (η sep ) efficiency of the three films was obtained and shown in Figure 5b. The composite WO 3 /CuWO 4 thin film showed significant improvement of charge separation efficiency compared with CuWO 4 film. This indicates that with a thin underlayer of WO 3 , the charge separation characteristics of CuWO 4 are greatly enhanced, which leads to much higher photocurrent.

Comparison of Absorption Efficiency (ηabs), and Charge Separation Efficiency (ηsep) of WO3, CuWO4 and WO3/CuWO4 Thin Films
Since the WO3/CuWO4 electrode demonstrated remarkable improvement in photoelectrochemical and EIS measurement compared with pristine WO3 substrate and CuWO4 films, we further extracted the efficiency values of light absorption (ηabs) and charge separation (ηsep) based on Equations (3) and (4) to quantify the contribution of each factor [34,35].
where Jabs is the photo current density calculated by multiplying ηabs by the standard AM 1.5 G (100 mW/cm 2 ) solar spectrum, Φ(λ) is the photon flux of sunlight in photons/m 2 /s, e is the charge of an electron (C). Jsca is the photocurrent of the photoelectrode in the presence of scavengers as a function of applied bias. By measuring the light transmittance and reflectance in an integrated sphere (see Figure S5), we were able to obtain the ηabs of pristine WO3, CuWO4 and WO3/CuWO4 in Figure 5a. It shows that more photons can be absorbed in the presence of CuWO4. With the measured values of ηabs and Equation (1), the integrated Jabs over the AM 1.5 spectrum of pristine WO3, CuWO4 and WO3/CuWO4 films are 1.8, 4.7 and 3.2 mA/cm 2 , respectively. The charge separation efficiency (ηsep) can be determined by adding the hole scavenger H2O2 to the electrolyte. The presence of H2O2 increased photocurrent density of all the three thin films ( Figure S6), which was due to the much faster charge transfer rate promoted by the hole scavenger. According to Equation (4), charge separation (ηsep) efficiency of the three films was obtained and shown in Figure 5b. The composite WO3/CuWO4 thin film showed significant improvement of charge separation efficiency compared with CuWO4 film. This indicates that with a thin underlayer of WO3, the charge separation characteristics of CuWO4 are greatly enhanced, which leads to much higher photocurrent.

Photoelectrochemical Water Splitting
In order to confirm that photocurrent was generated by water splitting, we conducted hydrogen and oxygen evolution under AM 1.5 illumination with 1.20 VRHE in 0.5 M Na2SO4 electrolyte. Figure 6a illustrates how the charge carriers are transported in the WO3/CuWO4 composite photoanode. Both WO3 and CuWO4 are excited by back illumination and generate charge carriers. Holes from WO3 are transferred to CuWO4 due to the formation of heterojunction, and are injected into the electrolyte from porous CuWO4 surface to oxidize water into O2. Electrons are directed to the Pt electrode where water molecules are reduced to hydrogen gas. As shown in Figure 6b, the total amount of oxygen

Photoelectrochemical Water Splitting
In order to confirm that photocurrent was generated by water splitting, we conducted hydrogen and oxygen evolution under AM 1.5 illumination with 1.20 V RHE in 0.5 M Na 2 SO 4 electrolyte. Figure 6a illustrates how the charge carriers are transported in the WO 3 /CuWO 4 composite photoanode. Both WO 3 and CuWO 4 are excited by back illumination and generate charge carriers. Holes from WO 3 are transferred to CuWO 4 due to the formation of heterojunction, and are injected into the electrolyte from porous CuWO 4 surface to oxidize water into O 2 . Electrons are directed to the Pt electrode where water molecules are reduced to hydrogen gas. As shown in Figure 6b, the total amount of oxygen and hydrogen evolved in three hours is about 5.0 and 14.0 µmol, respectively. The ratio of H 2 to O 2 produced is greater than the stoichiometric ratio [36]: Faradaic Efficiency pFEq % " actual hydrogen evolution rate{calculated amount from photocurrent generationˆ100 (5) which was about 79% using the hydrogen quantity calculated according to Equation (5). The loss of faradaic efficiency was possibly due to the slow kinetics of water oxidation and back reaction of H 2 and O 2 . The 3 h time course photocurrent density is presented in the inset of Figure 6b. The photocurrent dropped to around 63% of the initial current density within the first hour, but we believe that the stability of WO 3 /CuWO 4 photoanode could be improved by loading oxygen evolution reaction (OER) co-catalyst on the electrode surface.
Ma teria ls 2016, 9, 348 8 of 12 Faradaic Efficiency (FE) % = actual hydrogen evolution rate/calculated amount from photocurrent generation × 100 (5) which was about 79% using the hydrogen quantity calculated according to Equation (5). The loss of faradaic efficiency was possibly due to the slow kinetics of water oxidation and back reaction of H2 and O2. The 3 h time course photocurrent density is presented in the inset of Figure 6b. The photocurrent dropped to around 63% of the initial current density within the first hour, but we believe that the stability of WO3/CuWO4 photoanode could be improved by loading oxygen evolution reaction (OER) co-catalyst on the electrode surface.

Preparation of W Thin Film from Magnetron Sputtering
The synthesis route of the WO3/CuWO4 film is shown in Scheme 1. Tungsten film was deposited onto F-doped tin oxide (FTO) glass (sheet resistance ≤ 15 Ω/square, size: 10 mm × 25 mm and thickness: 2.2 mm) using direct current (DC) magnetron sputtering. The FTO glass was cleaned using acetone, ethanol and DI water prior to the sputtering. A metallic tungsten target (W, 3.00′′ diameter × 0.250′′ thick, 99.95% purity, Kurt J. Lesker, Jefferson Hills, PA, USA) was used as the sputtering target. The distance between the target and the substrate was set at around 10 cm. The sputtering chamber was evacuated to 8.0 × 10 −6 Torr or lower using a rotary pump and a turbo pump before introducing argon gas. The argon flow rate was kept constant at 20 sccm. A manual gate valve was used to fix the pressure inside the sputtering chamber at 20.0 mTorr during film deposition. The surface of the target was cleaned by sputtering the W target for 10 min before deposition onto FTO glass. The whole sputtering process lasted for 5.0 min at a constant working power of 230 W to obtain the black metallic W thin film.

Preparation of W Thin Film from Magnetron Sputtering
The synthesis route of the WO 3 /CuWO 4 film is shown in Scheme 1. Tungsten film was deposited onto F-doped tin oxide (FTO) glass (sheet resistance ď 15 Ω/square, size: 10 mmˆ25 mm and thickness: 2.2 mm) using direct current (DC) magnetron sputtering. The FTO glass was cleaned using acetone, ethanol and DI water prior to the sputtering. A metallic tungsten target (W, 3.00 11 diameter0 .250 11 thick, 99.95% purity, Kurt J. Lesker, Jefferson Hills, PA, USA) was used as the sputtering target.
The distance between the target and the substrate was set at around 10 cm. The sputtering chamber was evacuated to 8.0ˆ10´6 Torr or lower using a rotary pump and a turbo pump before introducing argon gas. The argon flow rate was kept constant at 20 sccm. A manual gate valve was used to fix the pressure inside the sputtering chamber at 20.0 mTorr during film deposition. The surface of the acetone, ethanol and DI water prior to the sputtering. A metallic tungsten target (W, 3.00′′ diameter × 0.250′′ thick, 99.95% purity, Kurt J. Lesker, Jefferson Hills, PA, USA) was used as the sputtering target. The distance between the target and the substrate was set at around 10 cm. The sputtering chamber was evacuated to 8.0 × 10 −6 Torr or lower using a rotary pump and a turbo pump before introducing argon gas. The argon flow rate was kept constant at 20 sccm. A manual gate valve was used to fix the pressure inside the sputtering chamber at 20.0 mTorr during film deposition. The surface of the target was cleaned by sputtering the W target for 10 min before deposition onto FTO glass. The whole sputtering process lasted for 5.0 min at a constant working power of 230 W to obtain the black metallic W thin film.

Fabrication of WO 3 Thin Film
The as-prepared metallic W thin film was placed in a clean porcelain crucible, and was transferred to a muffle furnace (CWF 12/5, Carbolite, Derbyshire, UK). The thin film was calcined at 500˝C for 2 h with a ramping rate of 2˝C/min for heating step and was cooled down naturally. Monoclinic WO 3 thin film with a light yellow colour was obtained. The process was repeated for 4 and 8 runs to load different amount of Cu 2+ ions. The films were subsequently heated in air at 550 C for 4 h in a muffle furnace, with a ramping rate of 2˝C/min for both heating and cooling steps. After annealing, the colour of the thin film changed to bright yellow, indicating the formation of a layer of CuWO 4 during heat treatment. CuO that formed from decomposition of excess Cu(NO 3 ) 2 along the edges of WO 3 thin film and FTO glass during heating was dissolved away by soaking in a 0.5 M HCl solution for 10 min.

Photoelectrochemical (PEC) Measurement
To investigate the photoelectrochemical properties of WO 3 , WO 3 /CuWO 4 , and CuWO 4 photoanodes, a conventional three-electrode system was used. All the PEC measurements were carried out in a home-made Teflon PEC cell with an illumination window of 5 mm inner diameter for back-illumination. The surface area of electrode exposed under illumination was about 0.2 cm 2 . 0.5 M of sodium sulphate (Na 2 SO 4 , Sigma-Aldrich, St. Louis, MO, USA) in deionized water solution with a pH value of 6 was used as the electrolyte. Platinum (Pt) coil and silver/silver chloride (Ag/AgCl) were used as counter and reference electrodes, respectively. The prepared thin films were used as working electrodes. The light source was simulated sunlight from a 150 W xenon solar simulator (67005, Newport Corp., Irvine, CA, USA) through an Air Mass filter (AM 1.5, Global, 81094, Newport Corp., Irvine, CA, USA) with a constant light intensity to standard AM1.5 sunlight (100 mW/cm 2 ) at the photoanode surface. Linear sweep voltammetry (LSV) was carried out by an electrochemistry workstation (CHI 852C, CH Instruments, Shanghai, China) both in dark and under AM1.5 sunlight simulator. Stability of WO 3 /CuWO 4 sample was carried out at a bias potential of 1.20 V RHE for 600 s with illumination on-off interval of 5 s. Charge separation efficiency was obtained by measuring light absorption of thin films and photocurrent in 0.5 M Na 2 SO 4 + 0.5 M H 2 O 2 aqueous solution. Incident photon to electron conversion efficiency (IPCE) was measured with a xenon light source (66983, Newport Corp., Irvine, CA, USA) coupled with a monochromator (74125, Newport Corp., Irvine, CA, USA) at a bias of 1.20 V RHE from back illumination. A Si photodiode (DH-Si, Bentham, Reading, Berkshire, UK) with known IPCE was used to calculate the IPCE of prepared thin films. A source meter (Keithley Instruments Inc., Solon, OH, USA, Model: 2400) was used to record the photocurrent of Si diode. CHI 852C electrochemistry workstation was used to record the photocurrent of each photoanode. IPCE calculation is given in the following formula: IPCE pλq " 100ˆ1240ˆpJpλq´J dark q{λ{Ipλq r38s (6) where λ is the wavelength of light in unit of nm; J(λ) is the photocurrent density in mA/cm 2 under illumination at λ; J dark is the photocurrent density measured at dark; and I(λ) is the incident light intensity in mW/cm 2 at λ [38]. Electrochemical impedance spectroscopy (EIS) was performed using an Autolab PGSTAT 302N system (Metrohm Autolab, Utrecht, The Netherlands) equipped with the FAR2 Faraday impedance module (Metrohm Autolab, Utrecht, The Netherlands). The flat band potential of CuWO 4 was determined using the Mott-Schottky equation on a CuWO 4 sample at frequencies of 5 k and 10 k Hz. The Nyquist plot was measured at 1.20 V RHE with a frequency ranging from 0.01 Hz to 100 kHz at 10 mV amplitude potential under AM1.5 illumination.

Photoelectrochemical Water Splitting
The photoelectrochemical water splitting was carried out in an air-tight reactor using back illumination. The light source (AM 1.5 sunlight, 100 mW/cm 2 ) and applied bias 1.20 V RHE were kept the same during the measurement. The amount of hydrogen and oxygen was analyzed by a gas chromatographer (GC-7890A, Agilent, Agilent, Santa Clara, CA, USA) equipped with thermal conductivity detector (TCD) detector.

Conclusions
In conclusion, we have converted monoclinic WO 3 thin film into WO 3 /CuWO 4 composite and CuWO 4 films through a facile dip-coating step followed by heat treatment. The PEC and EIS measurements showed that the presence of a thin layer of WO 3 beneath CuWO 4 can enhance the photocurrent density and reduce the charge transfer resistance compared with pure CuWO 4 film. By separately studying the photo absorption and charge transfer efficiencies, it was demonstrated that the WO 3 /CuWO 4 composite film exhibited enhancements in each process compared with single-phase WO 3 and CuWO 4 . Hydrogen and oxygen evolution was conducted to confirm that the photocurrent was generated from water splitting. Our result has provided a low-cost and controllable method to prepare WO 3 -metal tungstate heterojunction thin films, and helped to provide a reference for designing CuWO 4 -based photoanodes with greater efficiency.
Supplementary Materials: The following are available online at www.mdpi.com/1996-1944/9/5/348/s1. Figure S1: XRD patterns of different WO 3 phases obtained from magnetron sputtering on FTO (red) and normal glass slide (blue) substrates, which showed the FTO layer helped to induce the crystal growth of monoclinic WO 3 . Figure S2: TEM images of particles scraped from WO 3 /CuWO 4 , indicating network morphology of the CuWO 4 layer. Figure S3: Photocurrent comparison of thin film obtained from different runs of dip coating. Figure S4: Mott-Schottky plots of WO 3 /CuWO 4 thin film at 10 k and 5 k Hz under dark conditions. Figure S5: Absorption efficiency of the WO 3 , CuWO 4 and WO 3 /CuWO 4 thin film by measuring the transmission and reflection spectra using an integrating sphere (Absorbance (η abs ) = 1´Transmittance´Reflectance). Figure S6: Linear sweep voltammetry of all samples with (solid lines) and without the illumination of AM 1.5 (dashed lines), measured in 0.5 M Na 2 SO 4 + 0.5 M H 2 O 2 aqueous solution.