Heterostructured WO3@CoWO4 bilayer nanosheets for enhanced visible-light photo, electro and photoelectro-chemical oxidation of water

Herein, a facile interface-induced synthesis method is first established to newly fabricate two-dimensional (2D) bilayer nanosheets of WO3@CoWO4 as highly efficient catalysts for enhanced photo, electro and photoelectro-chemical oxygen evolution reactions (OERs). The heterostructure and the interfacial oxygen vacancy of WO3@CoWO4 reduce the energy barriers in OER. Density functional theory (DFT) calculations and material characterizations reveal that WO3@CoWO4 p-n heterojunction endows the composite with a narrowed band gap for better visible-light harvesting, rapid charge transfer across the interface and a lower recombination rate of the photo-excited carriers. The interface O-vacancy vests the active Co site with an enhanced density of state (DOS) at valence band maximum (VBM), which can increase the concentration of the photogenerated holes to improve photocatalytic and photoelectrochemical (PEC) activity. This study presents a proofof-concept design towards low cost and multi-metal 2D/2D nanosheets for water oxidation applications.


Introduction
As an emerging technology for solar energy conversion and storage, photocatalytic and photoelectrochemical (PEC) splitting of water into H2 and O2 has attracted tremendous attention for sustainable environment and energy development. [1][2][3] In this course, decomposition of water to dissociative oxygen (known as water oxidation or OER) is kinetically sluggish, due to the multi-step, four-electron and multi-proton transfer processes. Rational design of efficient photocatalytic or PEC water oxidation catalysts (WOCs) is essential for advancing the technologies toward efficient water-splitting into hydrogen. 4,5 Enormous efforts have been devoted to pursuing suitable semiconductor materials that can achieve efficient solar-energyconversion and propel the complex water oxidation reactions by photo-generated holes. 6,7 Among them, low-cost tungsten oxide (WO3) has emerged as a promising n-type and visible-lightactive semiconductor material, which possesses up to 12% solar spectrum absorption with a bandgap energy of 2.7 eV. 8,9 Nonetheless, the low photon energy conversion efficiency, instability caused by photo-corrosion, and poor kinetics of pristine WO3 usually result in an unsatisfactory activity in OER. 10,11 It was revealed that WO3 in a 2D nanosheet configuration has planar conduction channels, promoting the exposure of catalytically active facets to accelerate fast transport of the photoexcited charge carriers. 12,13 However, the long migration route in WO3 makes it easy for electron-hole recombination. To solve this problem, construction of a heterojunction composite using two semiconductors is an excellent strategy. 14,15 As a ptype semiconductor with low cost and high stability, cobalt tungstate (CoWO4) has drawn our attention. We project that coupling WO3 nanosheet with CoWO4 nanosheet in a p-n heterostructure would be a promising WOC candidate. However, specific challenges have to be addressed in the synthesis of WO3 nanosheet because it is a nonlayered compound and lacks the driving force for 2D anisotropic growth. 16 Therefore, the design and integration of CoWO4-WO3 bilayered nanosheets are much more difficult. To the best of our knowledge, no such a material has been attempted and reported.
In this work, we elaborately propose a hydrothermal method to generate WO3 nanosheets. Then, a scalable interface-induced strategy for bilayer formation was established by coating CoWO4 onto WO3 to obtain 2D WO3@CoWO4 bilayer hybrids. In this format, CoWO4 produces benefits to photocatalytic or PEC OER with triple functions: i) construction of WO3@CoWO4 p-n heterojunction; ii) prevention of WO3 from corrosion; iii) CoWO4 as an active OER electrocatalyst [17][18][19] to serve as a cocatalyst to promote photocatalytic OER. For the first time,

Synthesis of WO3 nanosheets.
WO3 nanosheets were synthesized via a hydrothermal method. In detail, 0.38 g of tungstic acid was dispersed in 27 mL deionized (DI) water, followed by dissolving 0.5 g thiourea in the solution. The suspension was then transferred into a 50 ml Teflon liner and sealed in an autoclave. The autoclave was heated at 180 ºC for 24 h. The precipitates were separated using a centrifuge and washed with DI water and ethanol for several times. Finally, WO3 powders were obtained after drying at 60 ºC.

Synthesis of WO3@CoWO4 nanosheet composites.
The obtained WO3 (0.1 g) was dispersed into 10 mL DI water under stirring for 10 min at room temperature. After that, certain amount of Co(NO3)2·6H2O was added into the suspension solution and stirred for another 10 min. Then, 2.5 mL ammonium hydroxide solution was added dropwise into the above solution and stirred for 1 h at room temperature before evaporating at 80 ºC. Finally, the residual powders were collected and heated at 300 ºC for 2 h under air with a heating rate of 5 ºC min −1 . The synthesized catalysts were designated as WO3@CoWO4-1, WO3@CoWO4-2, WO3@CoWO4-3, and WO3@CoWO4-4, and WO3@CoWO4-5 according to the different additive amount of Co(NO3)2·6H2O at 3.2 mg, 6.4 mg, 9.6 mg, 64 mg and 128 mg, respectively.

Electrochemical measurements.
Electrocatalytic tests were conducted in N2-saturated 0.1 M KOH in a three-electrode electrochemical system using a rotating disk electrode (RDE) configuration (Pine Instrument Company, USA), which is controlled by a Gamry electrochemical workstation (Reference 3000). Ag/AgCl (KCl sat.) and Pt wire were adopted as the reference electrode and the counter electrode, respectively.
Preparation of working electrode is described as follows: 7 mg catalyst was added into a solution containing 25 µL Nafion ® 117 solution and 500 µL ethanol to generate a suspension by sonication. Then, 10 µL of the catalyst ink was dripped onto a glassy carbon electrode (5.0 mm in diameter) and dried in air. All potentials were converted into reversible hydrogen electrode (RHE) values based on equation (1): The electrodes were activated by running cyclic voltammetry (CV) cycles from 1.2 to 1.8 V (vs RHE) for at least 10 times till stable and reproducible curves were obtained. Then, polarization curves using linear sweep voltammetry (LSV) were recorded with a rotation speed of 1600 rpm at a scan rate of 5 mV s -1 . The overpotential (η) was calculated according to the following formula: All polarization plots were recorded after iR-correction.

Photocatalytic oxygen evolution reaction tests.
For each reaction, 50 mg catalyst was dispersed in 50 mL phosphate buffer solution and the pH value was adjusted to around 6.8, followed by adding 0.49 g Na2SO4, 0.18 g Na2S2O8 and 0.03 g [Ru(bpy)3]Cl2· 6H2O. After that, the solution was transferred to a sealed double jacketed reactor (800 mL) with a quartz window, which was connected to an on-line gas chromatograph (Agilent 490 Micro GC) with a thermal conductivity detector. To remove air in the reactor completely, N2 was pumped in for at least 30 min. Then, the solution was stabilized in the dark for 10 min, which was probed as the baseline. The solution was irradiated via a 300 W Xeon lamp (Newport) through a light filter (λ＞420 nm) and aligned to 200 mW/cm 2 (2 suns) to start the reaction. The reaction temperature was maintained at 25 ºC by a flow of cooling water, controlled by a thermostatic water bath.

Photoelectrochemical Measurement.
Photoelectrochemical tests were carried out on a Zennium workstation (Zahner, Germany) in a three-electrode framework, with Ag/AgCl electrode as the reference electrode and Pt plate (1.5×1.5 cm 2 ) as the counter electrode. F-doped tin oxide (FTO) glasses were adopted as the photoanode substrate, which were cleaned before use under sonication by acetone, ethanol, and distilled water successively. Samples were loaded onto the FTO as below: 8 mg catalyst, 25 μL Nafion 117 solutions and 500 μL ethanol were mixed by ultrasonication and 40 μL of the resulted suspension was loaded onto the 1×1 cm 2

Synthesis and structure analysis.
X-ray diffraction (XRD) patterns ( Fig. 1 Transmission electron microscopy (TEM) images in Fig. 2a and Fig. S1a reveal the nanosheet morphology of WO3. The highresolution TEM (HRTEM) analysis indicates that (001) planes are exposed planes of WO3 (Fig. 2b). WO3@CoWO4 composites display very similar nanosheet morphologies ( Fig. 2c and S1-S5). Representative three-dimensional (3D) atomic force microscopy (AFM) measurements ( Fig. 2e and f) on WO3@CoWO4-3 show that the nanosheets could be thinner than 10 nm. High angle annular dark field scanning TEM (HAADF-STEM) and corresponding energy-dispersive X-ray spectroscopy (EDX) elemental mapping images of the composites (Fig. 2g, S2c, S3c, S4c and S5c) implicates the different coverage levels of CoWO4 on WO3. The full-scan X-ray photoelectron spectroscopy (XPS) spectra of WO3 and WO3@CoWO4-3 are provided in Fig. S7 samples. 21,22 In the O1s spectra (Fig. S7c), the peak at 530.4 eV can be associated with oxygen bonded to metal species, whereas the one centred at 531.6 eV is typical of the low coordination oxygen ions on the surface. 23 (Fig. S7d). 22,23 The formation mechanism of the bilayer nanosheet composite is briefly proposed in Fig. 3

Characterization of the electrochemical OER performance.
The electrocatalytic OER activity of the samples was evaluated by polarization curves (Fig. 4a). The overpotential at a current density (J) of 10 mA cm -2 is a criterion for assessing OER properties. WO3 was almost inactive in the electrocatalytic OER.
With higher CoWO4 loading in the composites, the electrocatalytic activity increased dramatically and then declined. WO3@CoWO4-4 gave the smallest overpotential of 0.38 V, which is lower than commercial RuO2 (0.40 V).
As the best OER catalyst, WO3@CoWO4-4 displayed excellent durability, as provided in Fig. S8 and Supporting Information. In addition, WO3@CoWO4 composites all displayed much higher electrochemically active surface area (EASA) than WO3, with WO3@CoWO4-4 to be the highest. (Fig. S9). This result indicates that abundant active sites were introduced to WO3 by loading of CoWO4.

Visible-light photocatalytic OER performance.
Fig. 5a displays photocatalytic OER activities of the samples. Elevated performance was observed on WO3@CoWO4 composites compared with WO3. With rising CoWO4/WO3 ratios, the OER activity increased first and then decreased. WO3@CoWO4-3 presented the highest O2 evolution rate (1.6 mmol g -1 O2 in 1 h), which is over 9 times higher than WO3. This rate is among the highest values reported for nonprecious metallic OER catalysts, as compared in Table S2.
A series of photo-dependent tests were carried out on the synthesized samples. As shown in UV-visible diffuse reflectance spectra (UV-Vis DRS, Fig. 5b), the visible light absorption intensity of the composites gradually improved with increasing CoWO4 loadings. In addition, a red shift occurred in the absorption band-edge of the composites, reflecting better absorption at longer wavelengths. WO3 presented a positive slope, typical for n-type semiconductors in the Mott-Schottky (M-S) plot (Fig. 5c). Since CoWO4 is a p-type semiconductor with a negative slope, inverted "V-shapes" were then observed on WO3@CoWO4 composites, reflecting a well-matched p-n heterostructure. [28][29][30] Fig. 5d shows a band structure diagram for WO3@CoWO4 system. The conduction band minimum (CBM) of WO3 (0.73 eV) was evaluated from the flat band potential of the M-S plot (Fig. 5c). The band gap of WO3 (2.40 eV) was obtained from the Tauc's plot (Fig. S10). VBM was acquired by the sum of CBM and band gap. Since we did not prepare pure CoWO4, the band structure data of CoWO4 were obtained by DFT calculations using the monoclinic CoWO4 slab (M-1, Fig.  S11a). The band gap (2.84 eV), CBM (-0.11 eV) and VBM (2.73 eV) of CoWO4 were obtained via the corresponding DOS (Fig.  S11b), which are close to those values reported in the literature. 31 The band gap of the WO3@CoWO4 system can be narrowed with CoWO4 as the VBM, and WO3 as the CBM, which helps explain their better light-harvesting ability than WO3. Since the CBM of CoWO4 is more negative than WO3, it is thermodynamically favorable for the photo-excited electrons to move from CoWO4 to WO3. Meanwhile, the holes generated in the valence bands of the two semiconductors can transfer from WO3 to CoWO4 due to their potential difference in VBM. Due to effective electron-hole separation by WO3@CoWO4 heterojunction, their To explore the charge redistribution across the WO3@CoWO4 interface, molecular models including a hexagonal WO3 slab (M-2) and a WO3@CoWO4 slab (M-3) were constructed (as given in Fig. S12 and Supporting Information). The charge redistribution across the interface (Fig. 5e) was explored by subtracting the electronic charge of M-3 from those of the M-1 and M-2. Specifically, charge accumulation mainly occurs at the side of WO3 while charge depletion focuses on CoWO4 near the interface. Therefore, interface electric dipole forms at the interface, which enables electrons to transfer from CoWO4 to WO3 while the holes transfer from WO3 to CoWO4. 33 This makes it easier for electron-hole transport and separation across the interface under light irradiation, which can also be experimentally demonstrated by photoluminescence (PL) spectra (Fig. S13). As shown, the lower luminescence intensity of the composites than WO3 suggests the reduced charge recombination.
Room temperature in-situ electron paramagnetic resonance (EPR) was performed in this water oxidation system with 5, 5dimethyl-1-pyrroline N-oxide (DMPO) as the trapping agent (Fig. 5f). No visible signal was obtained under dark. Interestingly, active seven-line paramagnetic signals were captured as • DMPO-X under irradiation after adding WO3 or WO3@CoWO4-3, which might arise from the excessive oxidation of DMPO by peroxide generated during OER. [34][35][36] In general, water oxidation reaction proceeds via four-electrontransfer steps based on the following mechanism: 27 *+H 2  WO3@CoWO4-3 displayed much stronger peak signals of • DMPO-X than that with WO3, indicating the higher concentration of peroxo species and the more active OER performance.

Photoelectrochemical (PEC) OER performance.
Based on the enhanced electro and photo-catalytic OER performance of WO3@CoWO4 composites, their PEC properties were further evaluated. Current-voltage curves were first recorded under dark (Fig. S14). By comparison, all the photoanodes presented deeply enhanced anodic photocurrent densities upon illumination (Fig. 6a). Compared to WO3, the photocurrent density of the composites firstly improved significantly along with the increased loading of CoWO4 on WO3 with WO3@CoWO4-2 reaching a maximum. The photocurrent density of WO3@CoWO4-2 is about 2 times larger than that of WO3 at 1.3 V. However, the anodic photocurrent densities decreased dramatically in WO3@CoWO4-3, -4 and -5 with further higher ratios of CoWO4. Similar to the above discussion in photocatalytic OER test, this phenomenon can be explained by the destruction of the optimum synergistic function of heterostructure for PEC activity. The transient photoresponse of the composites were assessed by measuring i-t curves at 1.0 V (Fig. 6b). Prompt and steady photocurrent responses can be captured on the photoanodes during on and off cycles of illumination, which had the same trend with Fig. 6a. It is noted that an applied bias is imposed on the photoanode during PEC OER, which promotes the output of the photogenerated electrons in WO3 through FTO-glass. The electron-hole recombination rate is reduced and thus the OER activity of WO3 is largely elevated in PEC test compared with that in photocatalysis. Due to the different mechanisms, WO3 exhibited the worst performance in photocatalysis but it was not the worst in PEC OER and the optimal CoWO4/WO3 ratio was also various in the two systems.
EIS measurements were conducted on the photoanodes under dark and irradiation, respectively (Fig. S15). Compared to those collected under dark, all the semicircles in EIS were largely minished under irradiation, proving the lowered charge transfer resistance by photo-induced charge carries. Especially, WO3@CoWO4-2 exhibited the smallest resistance diameter, which helps explain its highest photocurrent response in PEC test. Therefore, an appropriate construction of WO3/CoWO4 heterostructure could effectively boost the conductivity and PEC activity of WO3.
Besides, the instability of WO3 caused by photo-corrosion was largely improved. As provided in Fig. S16, WO3@CoWO4-2 (decayed by 4 %) exhibited much better long-term PEC stability than WO3 (decayed by 57 %) tested by the potentiostatic method.

DFT calculations for OER.
To study the effect of interface oxygen vacancies or defects (as verified by Fig. 2d, S3b, S4b Fig.  7b. This is because OER prefers to occur on the side of CoWO4 in the composites due to migration of holes as analyzed in Fig.  5d and e. Moreover, Co sites are believed to be more active centres for OER in CoWO4. 37,38 The specific Gibbs free energy changes during the four elementary steps are shown in Fig. 7b and Table S3. All adsorption scenarios in the four models shared uphill/endothermic energy profiles from *OH (step 1), *O (step 2) to *OOH (step 3), suggesting that an external driving force (light irradiation or electrical potential) is required to initiate the OER reaction. Once the reaction got to *OOH, the diagrams became downhill/exothermic in M-1, M-2 and M-3, indicating that step 4 is apt to happen and *OOH will convert to O2 (step 4) automatically. The step with the highest free energy barrier is referred as the overpotential-determining step. 39 It was noted that steps 1 and 2 were the potential-determining steps of WO3, while these energy barriers were lower in CoWO4 with step 2 being the hardest one. CoWO4 was catalytically more active than WO3 in OER. Compared with CoWO4, the energy barriers in steps 1, 2 and 3 were reduced in M-3, indicating the lowered OER overpotential on WO3@CoWO4 interface. This can be associated with the hole accumulation at the side of CoWO4 across the interface (Fig. 5e). Interestingly, after an O was removed in the interface, the barriers in steps 1 and 2 were significantly lowered a) b) c) c a (M-4), implying a simpler adsorption of water molecules onto the active Co site (step 1) and easier formation of OH* (step 2). As these two initial steps were the hardest in M-1, M-2 and M-3, their easier proceedings promoted by interface-O-vacancy were considered to contribute considerably to the overall OER activity in WO3@CoWO4 composites. Therefore, the WO3@CoWO4 interface and especially interface-O-vacancies can serve as active sites for both electro, photo-catalytic and PEC oxidation of water.
DOS of the designated Co site in M-3 and M-4 were projected in Fig. 7c. Especially, the interface-O-vacancy induced dramatically increased DOS of Co site at both VBM and CBM, which can accelerate the transport of photon-generated carriers under light irradiation. 16,40 Faster diffusion kinetics, higher photoconversion efficiency and higher concentration of the photogenerated holes to react with H2O can be achieved. Thus, interface-O-vacancies can not only reduce OER energy barriers but also induce enhanced photo-responsive behavior.

Conclusions
In summary, a versatile method was proposed for the synthesis of WO3@CoWO4 bilayer nanosheets as excellent WOCs for enhanced visible-light-driven photo, electro-catalytic and PEC OER processes. Because of the theoretically reduced OER barriers by WO3@CoWO4 interface and the interface-Ovacancies, WO3@CoWO4-4 displayed a low overpotential of 0.38 V in 0.1 M KOH for electrocatalysis. The creation of p-n heterojunctions and interface-O-vacancies can increase the photo-energy conversion efficiency and the water oxidation ability, enabling WO3@CoWO4-3 to present over 9-time-higher O2 evolution rate than WO3. A larger photocurrent with high stability was also observed in WO3@CoWO4-2 for PEC OER. The paradigm we presented in this work could provide a refreshing perspective for pursuing and designing more efficient low-dimensional photocatalytic, electrocatalytic and PEC OER catalysts.

Conflicts of interest
There are no conflicts to declare.