Comparison of sandwich and fingers-crossing type WO₃/BiVO₄ multilayer heterojunctions for photoelectrochemical water oxidation

Abstract

WO₃ and BiVO₄ thin films were spin-coated on FTO in a intersecting way with the WO₃ and BiVO₄ overlapping only in the center part, resulting in multilayer fingers-crossing thin films (fingers-crossing heterojunction). Samples with only an overlapped area (sandwich heterojunction) were obtained by cutting off the non-overlapped area of both the WO₃ and BiVO₄ layers (the nodes) as a counterpart for comparison. The influence of layer number, layer thickness and junction type on the photoelectrochemical (PEC) water splitting performance under simulated solar light for the obtained heterojunctions was investigated. XRD, SEM, XPS and UV-vis absorption measurements were conducted to investigate the structural properties and morphology. The photocurrent density of the sandwich heterojunction is higher than that of the fingers-crossing heterojunction at a low applied bias. The IPCE shows that both the fingers-crossing heterojunction and sandwich heterojunction can effectively utilize light up to 510 nm. Mott–Schottky plots show that the sandwich heterojunction has a more negative flat band potential, which is desirable for water splitting, than the fingers-crossing heterojunction. In addition, the mechanism of photogenerated carrier separation for these two heterojunctions was discussed.

---

1 Introduction

Photoelectrochemical (PEC) water splitting for hydrogen production has been considered as one of the most promising artificial photosynthesis methods¹ to utilize solar energy and address critical energy security and environmental sustainability issues.^2–5 Although many semiconductor materials such as TiO₂, ZnO, Fe₂O₃, CdS, and SrTiO₃ (ref. 6–9) have been reported to be effective for PEC hydrogen production, most of them are limited in their utilization due to serious photocorrosion, high charge carrier recombination, or a large band gap.¹⁰ In order to take full advantage of each material’s merits and avoid its weaknesses, many composite semiconductor materials such as Si/Fe₂O₃,¹¹ TiO₂/CdS,¹² ZnO/CdS,¹³ BiVO₄/TiO₂,¹⁴ BiVO₄/ZnO,¹⁵ and WO₃/Fe₂O₃ (ref. 16) have been designed and tested to address those issues. In a typical semiconductor heterojunction with type II band alignment, photogenerated electron–hole pairs tend to be spatially separated on different sides of the heterojunction, achieving efficient photogenerated electron–hole pair separation.^17,18

Recently, BiVO₄ photoelectrodes have emerged as a popular semiconductor for photoelectrochemical application and have attracted considerable attention.^19–24 BiVO₄ owing to its band gap of 2.4 eV (monoclinic) is capable of sufficiently absorbing light of the solar spectrum and is stable against photocorrosion in an aqueous electrolyte with a mild pH (between 3 and 11),^25–27 but unfortunately its charge carrier recombination rate is relatively high.^4,21 In order to improve the charge carrier separation in BiVO₄, the method of combining it with other semiconductors with a suitable band gap energy level and good carrier mobility has been adopted. For example, WO₃ has a conduction band (CB) level that is more positive than that of BiVO₄ and possesses a good carrier mobility.^18,28,29 Many types of WO₃/BiVO₄ heterojunctions have been fabricated in the past five years. For example, Jae Sung Lee et al.³⁰ fabricated flat WO₃/BiVO₄ heterojunction photoelectrodes on FTO with the BiVO₄ layers optimized. Hyeok Park et al.³¹ fabricated WO₃/Mo–BiVO₄ heterojunction photoelectrodes by atomic doping. Ligang Xia et al.³² fabricated a BiVO₄/WO₃/W heterojunction photoanode on a nanoporous WO₃ film. Core/shell WO₃/BiVO₄ nanowire and nanorod photoanodes were fabricated by Zheng et al.³³ and Pihosh et al.³⁴ respectively. To further improve the PEC performance of the WO₃/BiVO₄ heterojunctions, cobalt phosphate (Co-Pi),^29,34,35 FeOOH³⁶ and NiOOH^36,37 oxygen evolution catalysts (OECs) were deposited on the WO₃/BiVO₄ heterojunctions using different technologies. The dynamics of the photogenerated charge carriers in WO₃/BiVO₄ heterojunction photoanodes was also studied by Ivan Grigioni and his co-workers.³⁸ In our previous study,^10,39 a nanostructural WO₃/BiVO₄ heterojunction with efficient carrier separation was realized. It is desirable to further study this efficient heterojunction to maximize the ability of charge separation in the heterojunction.

However, most reported WO₃/BiVO₄ heterojunction materials contain only a single heterojunction and don’t yet have efficient light absorption or charge transfer. Fatwa F. Abdi et al.⁴⁰ reported a charge carrier diffusion length of 70 nm for BiVO₄. Reducing the thickness of the BiVO₄ layers could be a solution, but its light absorbance is insufficient. Using multilayer heterojunctions may enable these deficiencies to be avoided via controlling the layer thickness and layer number.

In this study, we fabricated fingers-crossing and sandwich type multilayer heterojunction photoelectrodes and compared their PEC performance to investigate the junction-type effect on the PEC water oxidation. Although the maximum photocurrent of the films is much less than the recently recorded value for core–shell WO₃/BiVO₄ nanorods (6.72 mA cm⁻² at 1.23V_RHE) reported by Pihosh and his co-workers,³⁴ the effect of the heterojunction structure including the layer number and the way of connecting the components on its performance is rarely investigated. So we have discussed the carrier separation in these two types of multilayer heterojunction, and investigated the influence of layer number and layer thickness on the PEC performance. At last, a possible mechanism for the influence of the junction nodes on PEC water splitting was proposed.

2 Experimental

2.1 Film preparation

For the WO₃ precursor solution (0.125 M), 0.01 mol tungsten acid (H₂WO₄) and 25 ml H₂O₂ (30 wt%) were dissolved in 15 ml de-ionized water (H₂O), then mixed with 40 ml of a PVA (poly vinyl alcohol) precursor solution (0.5 g PVA dissolved in 10 ml H₂O). For the BiVO₄ precursor solution (0.125 M), 0.01 mol Bi(NO₃)₃·5H₂O, 0.01 mol NH₄VO₃ and 10 ml concentrated nitric acid (HNO₃) were dissolved in 30 ml de-ionized water, then mixed with 40 ml PVA solution. Fluorine-doped SnO₂ (F:SnO₂, FTO)-coated glass substrates (TEC-15, 15 Ω sq⁻¹) were cleaned through successive ultrasonic cleaning with acetone, de-ionized water and ethanol for 30 min, and then dried using a nitrogen flow. An intersecting and overlapping WO₃ and BiVO₄ structure was synthesized on FTO substrates by the spin-coating of WO₃ and BiVO₄ layer by layer, the spin rate was controlled at 3000 rpm for 30 s. For the fingers-crossing heterojunction, each layer (any parts undesired for coating were covered with scotch tape) was coated on top of the previous layers after calcination (500 °C/2 h) between the coatings. For the sandwich heterojunction, both the WO₃ node and BiVO₄ node of the fingers-crossing heterojunction were cut off.

2.2 Film characterization

An X’pert PRO diffractormeter (PANalytical, using Cu Ka irradiation, λ = 15.4184 nm) was employed for performing X-ray diffraction (XRD) analysis and determining the structure and phase of the samples. A JSM-7800FE instrument (JEOL) was employed for performing field emission scanning electron microscopy (FE-SEM). X-ray photoelectron spectroscopy (XPS, Axis UltraDLD, Kratos) was used with monochromatic aluminum Kα radiation to obtain the atomic ratio of the samples by etching. Hitachi U-4100 UV-vis-NIR spectrophotometer equipment was used for recording the optical UV-vis absorption spectra of the samples.

2.3 Photoelectrochemical measurements

The PEC measurements were performed with a scanning potentiostat (CH Instruments, model CHI 760D) using a standard three-electrode system in 0.5 M Na₂SO₄ as the electrolyte with Ag/AgCl and a Pt plate as the reference electrode and counter electrode, respectively. The applied bias was converted to a value against a reversible hydrogen electrode (RHE) using eqn (1) below.

The simulated solar spectrum was generated with a fiber optic spectrometer (AvaSpec-2048) using 350 W xenon lamps set 100 mW cm⁻² as the light source. The electrochemical impedance spectroscopy (EIS) was performed at 1.1V_RHE with a frequency of 100 kHz–1 Hz under illumination. Apart from the above, a Newport monochromator 74126 and burnt glass made of quartz were used to measure the incident photon to current conversion efficiency (IPCE).

3 Results and discussion

To compare the PEC performances of the sandwich heterojunction and fingers-crossing heterojunction, we prepared bare WO₃, fingers-crossing/sandwich WO₃–BiVO₄-n (n = 1, 2, 3, 4, 5) heterojunctions and bare BiVO₄ photoelectrodes via a spin-coating method. The number n denotes the repeat number of the WO₃/BiVO₄ layers in the fingers-crossing or sandwich heterojunctions. Sandwich heterojunction photoelectrodes were made by cutting off the WO₃ node and BiVO₄ node areas of the fingers-crossing heterojunction with only the multijunction area left to provide clean cross sections. All the WO₃ and BiVO₄ layers were prepared using the same conditions. The configurations of the samples are presented in Scheme 1.

Scheme 1 Geometric illustration of the fingers-crossing heterojunction and sandwich heterojunction. The yellow layer represents BiVO₄ and the light blue layer represents WO₃.

3.1 Structure and morphology

The crystal structures of the bare film or heterojunctions were examined using XRD analysis. For comparison, the XRD pattern obtained from FTO is also shown in Fig. 1. The XRD patterns of all the photoelectrodes exhibited strong peaks corresponding to FTO of the substrate. The crystal phase of the bare WO₃ photoelectrode was identified to be monoclinic (JCPDS card no. 01-071-2141). The bare BiVO₄ photoelectrode exhibited a monoclinic scheelite type crystal structure (JCPDS card no. 00-014-0688). For the WO₃–BiVO₄-n (n = 1, 2, 3, 4, 5) photoelectrodes, the exhibited peaks corresponding to WO₃ and BiVO₄ indicate the absence of any new compound formation. The X-ray diffraction patterns of WO₃–BiVO₄-n (n = 2, 4) are shown in Fig. S1.†

Fig. 1 X-ray diffraction patterns of FTO, bare WO₃, WO₃–BiVO₄-n (n = 1, 3, 5) and bare BiVO₄ samples.

The surface morphologies of the multilayer heterojunctions are presented in Fig. 2. Bare WO₃, bare BiVO₄ and WO₃–BiVO₄-5 show a similar porous morphology but smaller grain sizes compared to that reported by Jae Sung Lee et al.³⁰ This could correspond to the different binder used for spin-coating (PVA in our work and polyethyleneimine (PEI) in Lee’s work). For the bare WO₃ film, the porous network with a 50 nm grain size provides a large number of reaction sites and a short distance for holes to travel to the surface.^41,42 For the bare BiVO₄ film, a larger grain size (∼150 nm) was observed. There is no significant grain size change for the BiVO₄ layers in the WO₃–BiVO₄-5 film as shown in Fig. 2c. The thickness of the WO₃–BiVO₄-5 film is ca. 324 nm as shown in Fig. 2d. As five cycles of the coating-anneal steps were conducted in the preparation of the WO₃–BiVO₄-5 film, the thickness of a single WO₃/BiVO₄ heterojunction was calculated to be ca. 65 nm by dividing 324 nm by 5. All the surface morphologies of the multilayer heterojunctions have little difference since the same synthetic conditions except for the layer number were used.

Fig. 2 SEM images of the different samples: (a) bare WO₃; (b) bare BiVO₄; (c) WO₃–BiVO₄-5; and (d) cross-sectional views of sample WO₃–BiVO₄-5.

The distribution of the Bi/W atomic ratio as a function of etching time for the multilayer heterojunctions (taking WO₃–BiVO₄-3(F) as a representative sample) is illustrated in Fig. 3. The etching rate was controlled at 0.1 nm s⁻¹. As the etching time increased, obvious changing of the Bi/W ratio was found along with the different layers, confirming successful fabrication of a multilayer heterojunction. The thickness of a single WO₃/BiVO₄ bi-layer was determined to be 66 nm by taking the average of 64 nm and 68 nm, which is consistent with the SEM result.

Fig. 3 XPS etching profile of the sample WO₃–BiVO₄-3(F).

3.2 Optical properties

The optical behavior of the heterojunction photoelectrodes was investigated using UV-vis absorbance spectra as shown in Fig. S2.† The onset of light absorption by the bare WO₃ photoelectrode started at 450 nm (Fig. S2a†) in correspondence with its band gap energy of 2.76 eV (Fig. S2b†). The bare BiVO₄ photoelectrode exhibited an absorption edge at 500 nm, commensurate with a band gap of 2.56 eV. For the fingers-crossing heterojunction photoelectrodes, the absorption edges were between at 450 nm and 500 nm. All the absorbance intensities increased with increasing layer number while the band absorption edges, which closely resemble the bandgap of BiVO₄, were identical. Similar absorbance spectra were observed for the sandwich heterojunction photoelectrodes.

3.3 PEC performance of multilayer heterojunction photoelectrodes

The photocurrent densities were measured using a three electrode system. It was found that the photocurrents of both the fingers-crossing heterojunctions and sandwich heterojunctions (Fig. 4) increased first and then decreased with the layer number increasing. Within the labels for the curves, the number 1 to 5 indicates the layer number of the WO₃–BiVO₄ bi-layers, F stands for fingers-crossing heterojunction and S stands for sandwich heterojunction. The maximum photocurrent densities for the samples of WO₃–BiVO₄-3(F) and WO₃–BiVO₄-3(S) were 1.08 mA cm⁻² and 1.05 mA cm⁻² at 1.60V_RHE, respectively. This can be explained as the number of photogenerated carriers gradually achieved saturation with the increasing layer number while internal recombination started to play a major role when the layer number was higher than 3. At the rising part of the curve, the increase of the layer number lead to higher light absorption and thus a higher photocurrent. While at the decreasing part of the curve, further increase of the layer number increased the distance for charge transport and thus gave a higher recombination rate. A trade-off between the light absorption and the charge transport distance was also reported by Kazuhiro Sayama et al.⁴³

Fig. 4 Current density–potential curves for start–stop simulated AM 1.5G light irradiation of (a) fingers-crossing heterojunctions and (b) sandwich heterojunctions with different numbers of bi-layers.

Fig. 5a shows the photocurrent densities of the three bi-layers heterojunction photoelectrodes of WO₃–BiVO₄-3(F) and WO₃–BiVO₄-3(S). As the applied bias increased, a faster photocurrent density increase was observed for sample WO₃–BiVO₄-3(S). The photocurrent density of WO₃–BiVO₄-3(S) was greater than that of WO₃–BiVO₄-3(F), by approximately 1.1 times at 1.4V_RHE. This is ascribed to the presence of nodes which reduced the performance for separation of the photogenerated electron–hole pairs. While with an increased applied bias, currents close to the saturation photocurrent were obtained for both samples. This could be ascribed to the larger applied bias promoting charge carrier separation²⁰ and the influence of WO₃ and BiVO₄ nodes on the photocurrent was reduced.

Fig. 5 (a) Current density–potential curves for start–stop simulated AM 1.5G light irradiation, (b) IPCE performance, (c) charge separation efficiency and (d) charge injection efficiency of the samples WO₃–BiVO₄-3(F) and WO₃–BiVO₄-3(S).

In order to compare the photoactive wavelength region for the various photoelectrodes and make a quantitative correlation between the fingers-crossing heterojunction and sandwich heterojunction, incident photon conversion efficiency measurements were conducted under 1.4V_RHE in a 0.5 M Na₂SO₄ aqueous solution. The IPCE can be expressed as:³⁷

where I is the photocurrent density, λ is the incident light wavelength, and J_light is the measured irradiance. Fig. 5b shows the comparison of the IPCE for WO₃–BiVO₄-3(F) and WO₃–BiVO₄-3(S). The IPCE characteristics show a similar dependence on wavelength with optical band onsets at 510 nm. The highest efficiency of 31% at 400 nm for WO₃–BiVO₄-3(S) was approximately 1.41 times that of the 22% at 400 nm for WO₃–BiVO₄-3(F). This improvement is much higher than that under white light illumination, which is possibly due to the low light intensity.

To gain better insight into the efficient water oxidation, charge separation efficiency (η_separation) and injection efficiency (η_injection) tests were conducted in a 0.5 M Na₂SO₄ electrolyte containing 0.5 M H₂O₂. η_separation or η_injection is the fraction of photogenerated holes that does not recombine with electrons in the bulk or at surface traps, respectively. The relationship between the water splitting photocurrent (J_H2O), photon absorption photocurrent density (J_absorbed), η_separation and η_injection can be expressed as follows:^44,45

J_H2O = J_absorbed × η_separation × η_injection (3)

Having H₂O₂ as a hole scavenger in the electrolyte means that the η_injection becomes 100% (η_injection = 1). So, the photocurrent measured with H₂O₂ in the electrolyte (J_H2O2) can be expressed as follows:

J_H2O2 = J_absorbed × η_separation (4)

Hence, the η_separation and η_injection are obtained by dividing J_H2O2 by J_absorbed and dividing J_H2O by J_H2O2, respectively. The details can be expressed as follows:

η_separation = J_H2O2/J_absorbed (5)

η_injection = J_H2O/J_H2O2 (6)

By measuring the light absorption of the sample WO₃–BiVO₄-3 and integrating it with respect to the AM 1.5G solar spectrum, J_absorbed was calculated to be 6.17 mA cm⁻². Fig. 5c and d present the charge separation efficiency and injection efficiency, respectively. WO₃–BiVO₄-3(F) and WO₃–BiVO₄-3(S) show the same η_separation as the applied bias increases, which indicates that the nodes show no effect on the recombination between electrons and holes in the bulk. While the obtained different η_injection values for WO₃–BiVO₄-3(F) and WO₃–BiVO₄-3(S) suggests that the nodes influences the recombination behaviour at the surface. At a low applied bias of 0.6V_RHE, values of 7.6% and 45.7% were obtained for the WO₃–BiVO₄-3(F) and WO₃–BiVO₄-3(S) respectively. However, at a high bias of 1.85V_RHE, the same η_injection (97.2%) was achieved for both the films. These results can be explained as the sandwich type structure of the WO₃/BiVO₄ multi-layer heterojunction possesses a better surface reaction activity, while for the fingers-crossing structure, a higher applied bias is required to achieve an efficient surface reaction.

Fig. 6 shows the relationship of the layer number with the photocurrent difference value (DV) between samples WO₃–BiVO₄-n(F) and samples WO₃–BiVO₄-n(S) at 1.1V_RHE. The photocurrent difference is the value obtained by deducting from the photocurrent of the sandwich samples that of the fingers-crossing samples at 1.1V_RHE. With the increasing layer number, the photocurrent difference increased first then decreased and reached its maximum value of 0.322 mA cm⁻² at three bi-layers. When the layer number is below 3, the photogenerated holes are easily transferred to the electrode/electrolyte interface to oxidize water as the carrier diffusion length of BiVO₄ is ca. 70 nm.⁴⁰ When the layer number is over 3, the large thickness of the films inhibits the hole transfer and reduces the photocurrent.

Fig. 6 The relationship of the layer number and the photocurrent difference value (DV) between samples WO₃–BiVO₄-n(F) and WO₃–BiVO₄-n(S) at 1.1V_RHE.

In order to investigate the influence of the thickness of each layer on the photocurrent density, we kept the bi-layer number of the heterojunctions at two and adjusted the thickness of each layer by changing the spin-coating circle number for each layer. The thickness of all the layers was controlled simultaneously by using the same spin-coating circle number. The samples were labelled as double-n (n = 1, 2, 3, 4, 5, n indicates the spin-coating circle number). Fig. 7a shows the photocurrent density of the fingers-crossing heterojunction (double-n (F)) and sandwich heterojunction (double-n (S)) at an applied bias of 1.2V_RHE. As the layer thickness increased, the photocurrent densities for both structures decreased. This can be interpreted as due to enhancement of the recombination with the increasing thickness of each layer. The photocurrent density decay with the increase of the thickness of each layer for the sandwich heterojunction is slower when compared to that of the fingers-crossing heterojunction. 76% of the photocurrent (at 1.2V_RHE) remained for the sandwich heterojunction while only 31% of photocurrent was left for the fingers-crossing heterojunction. This can be due to a higher recombination rate of the photogenerated electron–hole pairs with the presence of the nodes in the fingers-crossing heterojunction. These results mean that a smaller thickness of each layer could reduce the effects of the nodes. When the concentration of the precursors was reduced to half (0.0625 M), the BiVO₄ layer could not cover all of the FTO substrate (Fig. S3†). Thus, the optimum thickness for each layer was controlled by using one spin-coating cycle and a thickness of a single WO₃/BiVO₄ heterojunction of ca. 65 nm was obtained. It was found that the photocurrent performance of all the fingers-crossing heterojunctions is worse than that of a sandwich heterojunction with the same layer number and thickness. Fig. 7b shows the photocurrents of the fingers-crossing heterojunction and sandwich heterojunction of double-5. The photocurrent density of the double-5 (S) is 2.8 times that of double-5 (F) at 1.6V_RHE.

Fig. 7 The relationship between layer thickness and photocurrent performance. (a) Photocurrent densities of the heterojunctions at an applied bias of 1.2V_RHE; (b) comparison of the photocurrent performances of double-5 (F) and double-5 (S).

In order to study the effects of WO₃ nodes and BiVO₄ nodes on the flat band potential, impedance measurements were conducted at different frequencies (1, 2, 5, 10 kHz) on prepared samples in a 0.5 M Na₂SO₄ electrolyte in the dark and this allowed the construction of Mott–Schottky plots based on the following eqn (7):³⁰

where C_sc is the space charge capacitance in F; e is the electronic charge in C; ε is the dielectric constant of the semiconductor; ε₀ is the permittivity of free space; N_d is the carrier density in cm⁻³; E is the applied bias in V; E_fb is the flat band potential in V; k is the Boltzmann constant; T is the temperature in K.

The x intercept and the slope of the linear region in a plot of C_sc⁻² versus E are usually used to determine the flat band potential and the density of carriers. A plot of impedance versus frequency (Fig. S4†) was plotted to check the validity of the Mott–Schottky results. As shown in Fig. 8, the flat band potential values of WO₃–BiVO₄-5(F) and WO₃–BiVO₄-5(S) were determined to be 0.55V_RHE and 0.08V_RHE, respectively, and both values are between those of pure WO₃ and BiVO₄ (1.14V_RHE and 0.01V_RHE, respectively) and are similar to reported results.^30,46 It can be found that the existence of nodes has effects on the flat band potential and the density of carriers. The sandwich heterojunction possesses a more negative flat band potential than the fingers-crossing heterojunction. From the slope, it was found that all the deposited films possess the nature of an n-type semiconductor. For n-type semiconductors, a smaller slope value of WO₃–BiVO₄-5(S) than that of WO₃–BiVO₄-5(F) indicates that WO₃–BiVO₄-5(S) has a larger density of carriers compared to the latter.

Fig. 8 Mott–Schottky plots of (a) WO₃–BiVO₄-5(F) and (b) WO₃–BiVO₄-5(S) at different frequencies.

Electrochemical impedance spectroscopy (EIS) was also carried out under front illumination at 1.1V_RHE to further compare the charge transfer properties of the fingers-crossing and sandwich heterojunctions. The Nyquist plots of the EIS results are shown in Fig. 9. The circular dots in the plots represent the experimental data, and the solid lines represent the curve fitting using the equivalent circuit model shown inside the plots. In the equivalent Randle circuit, R_s is the solution resistance, Q₁ and Q₂ are the constant phase elements (CPE) which stand for the electrolyte–photoelectrode interface and the WO₃–BiVO₄ interface. The R_ct1 and R_ct2 are the charge transfer resistances across the interfaces of electrolyte–photoelectrode and WO₃–BiVO₄. The parameters extracted from the equivalent circuit fitting are shown in Table 1. The semicircle at the high frequency end in the Nyquist plot corresponds to the charge transfer resistance (R_ct).^4,47 As can be seen in Table 1, the sandwich heterojunction WO₃–BiVO₄-3(S) shows a smaller R_ct2 and Q₂ than WO₃–BiVO₄-3(F), presumably resulting from the absence of WO₃ nodes and BiVO₄ nodes. The smaller R_ct2 and Q₂ of WO₃–BiVO₄-3(S) than for WO₃–BiVO₄-3(F) indicate that the sandwich heterojunction is a structure more favourable for charge transfer than the fingers-crossing heterojunction. While the extracted parameters of R_ct1 and Q₁ for the fingers-crossing and sandwich heterojunctions are close, indicating that both the WO₃ node and BiVO₄ node have little effect on the electron transport at the interface of the semiconductor/electrolyte.

Fig. 9 Nyquist plots of the samples WO₃–BiVO₄-3(F) and WO₃–BiVO₄-3(S).

Table 1 Simulated parameters of an equivalent circuit model

Samples	Rs (Ω)	Rct1 (kΩ)	Q1		Rct2 (kΩ)	Q2
			Y01 (10^−5)	n1		Y02 (10^−5)	n2
WO3–BiVO4-3(F)	78.19	1.01	6.68	0.80	3.03	2.68	0.89
WO3–BiVO4-3(S)	80.12	1.32	5.36	0.74	0.19	1.89	1.00

---

For a WO₃–BiVO₄ heterojunction in Na₂SO₄ solution the energy band positions of BiVO₄ (V_CB = 0.02 eV, V_VB = 2.53 eV, V vs. NHE) and WO₃ (V_CB = 0.41 eV, V_VB = 2.77 eV, V vs. NHE) have been reported by Lee.³⁰ It is widely accepted that the bottoms of the conduction bands in several n-type semiconductors are more negative by 0.1 V–0.3 V than the flat band potential.^37,48 Based on the band positions, the optical band gap energies and the experimental results discussed above, we proposed a band structure model for our two different types of multilayer heterojunctions as shown in Scheme 2. We take BiVO₄/WO₃/BiVO₄ as one cycle junction and discuss the band structures of each outermost cycle junction (shown in Scheme 2a and b). Both the conduction band and valence band edge of BiVO₄ are more negative than the corresponding band edges of WO₃. This band structure facilitates the injection of photogenerated electrodes from the conduction band of BiVO₄ to that of WO₃. For heterojunctions under light irradiation, symmetrical space charge fields E_n1 and E_n2 should generate two voltages (ΔV_n1 and ΔV_n2) in opposite directions.⁴⁹ For the fingers-crossing heterojunction, the WO₃ node connects all the WO₃ fingers resulting in an equipotential within all the WO₃ layers. The same situation should be expected for all the BiVO₄ layers. These equipotential connections contribute to the same amplitude of ΔV_n1 and ΔV_n2. This indicates that the total voltage difference (ΔV_n) in one cycle junction should be zero (ΔV_n1 + ΔV_n2 = 0) at the adjacent junctions. However, underlying junctions will generate a smaller photovoltage (ΔV_n) as less light is absorbed. The case for the sandwich heterojunctions without WO₃ or BiVO₄ nodes is different. With the absence of an equipotential connection, there is an additional Fermi level drop in the outer BiVO₄ layer induced by electrons flowing to the double layer across the BiVO₄/electrolyte. Thus, the space charge region at BiVO₄/WO₃ was strengthened while that at WO₃/BiVO₄ was weakened, resulting in ΔV_n1 < ΔV_n2. As for the next BiVO₄/WO₃/BiVO₄ cycle, the space charge region of WO₃/BiVO₄ was weakened because that of BiVO₄/WO₃ in the previous cycle was strengthened. With repetition of the BiVO₄/WO₃/BiVO₄ cycles, the conduction band edges of the semiconductors for each layer in the sandwich heterojunction gradually shifted to negative values, as shown in Scheme 2b. That is why the flat-band potential of the sandwich heterojunction is more negative compared to that of the fingers-crossing heterojunction, and this flat-band potential shift is beneficial for charge separation and thus the photoelectrochemical activity.

Scheme 2 Schematic illustration of the interfacial band edge structure within (a) two cycle junctions of the fingers-crossing heterojunction, and (b) two cycle junctions of the sandwich heterojunction.

4 Conclusions

Fingers-crossing heterojunction and sandwich heterojunction WO₃/BiVO₄ photoelectrodes were fabricated on FTO substrates. Both the as-prepared WO₃ layer and BiVO₄ layer possess a monoclinic phase. The influence of layer number and layer thickness on the PEC performance was investigated, and the optimal layer number was found to be 3 layers and the best thickness for each layer was the thinnest thickness used. It was also found that the flat band potential for the sandwich heterojunction shifted to be more negative compared to the fingers-crossing one and an enhancement of the PEC performance for hydrogen production was observed especially when applied with a low bias. This work compared two different types of heterojunctions and showed that the PEC properties are sensitive to the junction type. With regard to further improvement of the PEC performance of heterojunctions, optimizing the structure of the electrode remains a promising approach, and it can also inspire ideas for investigating other heterostructures.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (No. 51202186).

Notes and references

1. H. W. Jeong, T. H. Jeon, J. S. Jang, W. Choi and H. Park, J. Phys. Chem. C, 2013, 117, 9104–9112.
2. K. Zhang and L. Guo, Catal. Sci. Technol., 2013, 3, 1672.
3. D. Eisenberg, H. S. Ahn and A. J. Bard, J. Am. Chem. Soc., 2014, 136, 14011–14014.
4. X. Fu, M. Xie, P. Luan and L. Jing, ACS Appl. Mater. Interfaces, 2014, 6, 18550–18557.
5. P. Chatchai, A. Y. Nosaka and Y. Nosaka, Electrochim. Acta, 2013, 94, 314–319.
6. F. E. Osterloh, Chem. Soc. Rev., 2013, 42, 2294–2320.
7. R. Daghrir, P. Drogui and D. Robert, Ind. Eng. Chem. Res., 2013, 52, 3581–3599.
8. K. Sivula, F. L. Formal and M. Grätzel, ChemSusChem, 2011, 4, 432–449.
9. T. Hisatomi, J. Kubota and K. Domen, Chem. Soc. Rev., 2014, 43, 7520–7535.
10. J. Su, L. Guo, N. Bao and C. A. Grimes, Nano Lett., 2011, 11, 1928–1933.
11. M. T. Mayer, C. Du and D. Wang, J. Am. Chem. Soc., 2012, 134, 12406–12409.
12. S. Liu, N. Zhang, Z. Tang and Y. Xu, ACS Appl. Mater. Interfaces, 2012, 4, 6378–6385.
13. Y. Tak, S. J. Hong, J. S. Lee and K. Yong, J. Mater. Chem., 2009, 19, 5945.
14. S. Ho-Kimura, S. J. A. Moniz, A. D. Handoko and J. Tang, J. Mater. Chem. A, 2014, 2, 3948.
15. S. J. A. Moniz, J. Zhu and J. Tang, Adv. Energy Mater., 2014, 4, 1301590.
16. K. Sivula, F. L. Formal and M. Grätzel, Chem. Mater., 2009, 21, 2862–2867.
17. S. Shenawi-Khalil, V. Uvarov, S. Fronton, I. Popov and Y. Sasson, J. Phys. Chem. C, 2012, 116, 11004–11012.
18. R. Saito, Y. Miseki and K. Sayama, Chem. Commun., 2012, 48, 3833–3835.
19. H. Yoon, M. G. Mali, J. Y. Choi, M.-w. Kim, S. K. Choi, H. Park, S. S. Al-Deyab, M. T. Swihart, A. L. Yarin and S. S. Yoon, Langmuir, 2015, 31, 3727–3737.
20. M. Long, W. Cai and H. Kisch, J. Phys. Chem. C, 2008, 112, 548–554.
21. Y. H. Ng, A. Iwase, A. Kudo and R. Amal, J. Phys. Chem. Lett., 2010, 1, 2607.
22. R. Saito, Y. Miseki and K. Sayama, J. Photochem. Photobiol., A, 2013, 258, 51–60.
23. Y. Park, K. J. McDonald and K. S. Choi, Chem. Soc. Rev., 2013, 42, 2321–2337.
24. M. Zhong, T. Hisatomi, Y. Kuang, J. Zhao, M. Liu, A. Iwase, Q. Jia, H. Nishiyama, T. Minegishi, M. Nakabayashi, N. Shibata, R. Niishiro, C. Katayama, H. Shibano, M. Katayama, A. Kudo, T. Yamada and K. Domen, J. Am. Chem. Soc., 2015, 137, 5053–5060.
25. S. K. Cho, H. S. Park, H. C. Lee, K. M. Nam and A. J. Bard, J. Phys. Chem. C, 2013, 117, 23048–23056.
26. N. F. Fatwa, F. Abdi, A. Dabirian and R. van de Krol, Mater. Res. Soc. Symp. Proc., 2012, 1446, 7–12.
27. M. T. McDowell, M. F. Lichterman, J. M. Spurgeon, S. Hu, I. D. Sharp, B. S. Brunschwig and N. S. Lewis, J. Phys. Chem. C, 2014, 118, 19618–19624.
28. C. A. Bignozzi, S. Caramori, V. Cristino, R. Argazzi, L. Meda and A. Tacca, Chem. Soc. Rev., 2013, 42, 2228–2246.
29. S. K. Pilli, R. Janarthanan, T. G. Deutsch, T. E. Furtak, L. D. Brown, J. A. Turner and A. M. Herring, Phys. Chem. Chem. Phys., 2013, 15, 14723–14728.
30. S. J. Hong, S. Lee, J. S. Jang and J. S. Lee, Energy Environ. Sci., 2011, 4, 1781–1787.
31. K. Zhang, X. J. Shi, J. K. Kim and J. H. Park, Phys. Chem. Chem. Phys., 2012, 14, 11119–11124.
32. L. Xia, J. Bai, J. Li, Q. Zeng, X. Li and B. Zhou, Appl. Catal., B, 2016, 183, 224–230.
33. P. M. Rao, L. Cai, C. Liu, I. S. Cho, C. H. Lee, J. M. Weisse, P. Yang and X. Zheng, Nano Lett., 2014, 14, 1099–1105.
34. Y. Pihosh, I. Turkevych, K. Mawatari, J. Uemura, Y. Kazoe, S. Kosar, K. Makita, T. Sugaya, T. Matsui, D. Fujita, M. Tosa, M. Kondo and T. Kitamori, Sci. Rep., 2015, 5, 11141.
35. Y. Pihosh, I. Turkevych, K. Mawatari, T. Asai, T. Hisatomi, J. Uemura, M. Tosa, K. Shimamura, J. Kubota, K. Domen and T. Kitamori, Small, 2014, 10, 3692–3699.
36. X. Shi, I. Y. Choi, K. Zhang, J. Kwon, D. Y. Kim, J. K. Lee, S. H. Oh, J. K. Kim and J. H. Park, Nat. Commun., 2014, 5, 4775.
37. T. W. Kim and K. S. Choi, Science, 2014, 343, 990–994.
38. I. Grigioni, K. G. Stamplecoskie, E. Selli and P. V. Kamat, J. Phys. Chem. C, 2015, 119, 20792–20800.
39. J. Su, X. Feng, J. D. Sloppy, L. Guo and C. A. Grimes, Nano Lett., 2011, 11, 203–208.
40. F. F. Abdi, T. J. Savenije, M. M. May, B. Dam and R. van de Krol, J. Phys. Chem. Lett., 2013, 4, 2752–2757.
41. S. J. Hong, H. C. Jun, P. H. Borse and J. S. Lee, Int. J. Hydrogen Energy, 2009, 34, 3234.
42. M. U. Clara Santato and J. Augustynski, J. Phys. Chem. B, 2001, 105, 936–940.
43. I. Fujimoto, N. Wang, R. Saito, Y. Miseki, T. Gunji and K. Sayama, Int. J. Hydrogen Energy, 2014, 39, 2454–2461.
44. H. Dotan, K. Sivula, M. Grätzel, A. Rothschild and S. C. Warren, Energy Environ. Sci., 2011, 4, 958–964.
45. J. H. Kim, Y. J. Jang, J. H. Kim, J. W. Jang, S. H. Choi and J. S. Lee, Nanoscale, 2015, 7, 19144–19151.
46. M. Li, L. Zhao and L. Guo, Int. J. Hydrogen Energy, 2010, 35, 7127–7133.
47. L. Zhang, E. Reisner and J. J. Baumberg, Energy Environ. Sci., 2014, 7, 1402–1408.
48. P. Bornoz, F. F. Abdi, S. D. Tilley, B. Dam, R. van de Krol, M. Graetzel and K. Sivula, J. Phys. Chem. C, 2014, 118, 16959–16966.
49. J. Cao, J. Xing, Y. Zhang, H. Tong, Y. Bi, T. Kako, M. Takeguchi and J. Ye, Langmuir, 2013, 29, 3116–3124.

---

Footnote

† Electronic supplementary information (ESI) available: XRD of multilayer heterojunctions (Fig. S1), optical behavior of heterojunctions (Fig. S2), SEM image of BiVO₄ (Fig. S3), and the dependence of impedance on frequency (Fig. S4). See DOI: 10.1039/c5ra25601a

---