Multilayer WO 3 / BiVO 4 Photoanodes for Solar-Driven Water Splitting Prepared by RF-Plasma Sputtering

: A series of WO 3 , BiVO 4 and WO 3 / BiVO 4 heterojunction coatings were deposited on ﬂuorine-doped tin oxide (FTO), by means of reactive radio frequency (RF) plasma (co)sputtering, and tested as photoanodes for water splitting under simulated AM 1.5 G solar light in a three-electrode photoelectrochemical (PEC) cell in a 0.5 M NaSO 4 electrolyte solution. The PEC performance and time stability of the heterojunction increases with an increase of the WO 3 innermost layer up to 1000 nm. A two-step calcination treatment (600 ◦ C after WO 3 deposition followed by 400 ◦ C after BiVO 4 deposition) led to a most performing photoanode under back-side irradiation, generating a photocurrent density of 1.7 mA cm − 2 at 1.4 V vs. SCE (i.e., two-fold and ﬁve-fold higher than that generated by individual WO 3 and BiVO 4 photoanodes, respectively). The incident photon to current e ﬃ ciency (IPCE) measurements reveal the presence of two activity regions over the heterojunction with respect to WO 3 alone: The PEC e ﬃ ciency increases due to improved charge carrier separation above 450 nm (i.e., below the WO 3 excitation energy), while it decreases below 450 nm (i.e., when both semiconductors are excited) due to electron–hole recombination at the interface of the two semiconductors. of the best performing one reported in our previous work [24] prepared under similar sputtering conditions, due the Bi-rich atomic ratio of the here investigated material, revealed by EDX analysis. Indeed, the performance of BiVO 4 is strongly affected by the Bi:V atomic ratio, stoichiometric 1:1 ratio materials being best performing. The time stability test (CA at 1.0 V vs. SCE, Figure 3c), though confirming the higher performance of the sample calcined at 600 °C, also reveals a partial loss of activity along the irradiation time. Interestingly, when the light is turned on the two samples generate a similar photocurrent density, but the BiVO 4 coating post calcined at 600 °C shows a marked increase during the first 20 min of irradiation. between electrons photo-promoted in the WO 3 conduction band (CB) and holes photoproduced in the BiVO 4 valence band (VB), hindering the overall PEC performance.


Introduction
With the rapid increase in the global energy demand and the tremendous climate concerns due to combustion of the limited fossil fuels reserves, there is an urgent need for clean and renewable energy sources [1,2]. The exploitation of hydrogen as an energy vector of the future is thought as an ideal solution for these problems [3,4]. However, to date, most of its production is based on steam reforming of natural gas which still involves massive amounts of CO 2 emission, resulting environmentally unsustainable [5]. In this context, photoelectrochemical (PEC) water splitting is considered one of the most promising and green routes to harvest and convert solar energy through the production of H 2 [6].
Among plenty of semiconductor materials suitable for the PEC water splitting reaction, BiVO 4 has received significant attention in the last years due to its remarkable theoretical efficiency under visible light irradiation [7,8]. This is due to its narrow band gap of 2.4 eV, that extends its photoactivity up to 510 nm, and favorable band edges position. However, BiVO 4 usually suffers low photocatalytic efficiency due to poor charge carriers mobility [9]. A possible solution to overcome this limitation is to couple BiVO 4 with another semiconductor in order to form a heterojunction at the interface, achieving a spatial charge separation. Several semiconductors can be combined with BiVO 4 , including WO 3 [10], SnO 2 [11], Fe 2 O 3 [12] and CdS [13]. Among these, WO 3 /BiVO 4 is one of the most promising heterostructure, because both the valence and conduction bands of BiVO 4 are located at a higher energy than those of WO 3 . Hence, a type II heterojunction is generated at the interface between the two semiconductor materials that allows efficient photo-promoted electron-hole separation [14]. Moreover, WO 3 possesses a good electron transport ability of 12 cm 2 V −1 s −1 with a moderate length of hole diffusion of ca. 150 nm [15]. Thus, in WO 3 /BiVO 4 heterojunction, BiVO 4 primarily acts as an excellent visible-light harvester up to 510 nm (i.e., ca. 30% of sunlight), while WO 3 acts as efficient electron conductor. Several morphologies and preparation methods influencing the PEC activity of the WO 3 /BiVO 4 heterojunction have been widely investigated and reported in literature so far. These include dense [10,14,16] and porous films [17,18], nanorods [19,20], helix [21] and inverse opal [22] structures. In particular, the highest photocurrent density reported so far (ca. 6.72 mA cm −2 at 1.23 V vs. RHE under 1 sun illumination) has been achieved over core-shell WO 3 /BiVO 4 nanorods arrays decorated with a Co-Pi water oxidation co-catalyst system [23]. The WO 3 nanorods arrays were prepared by RF magnetron sputtering under glancing angle deposition.
In the present work, a BiVO 4 thin film was deposited by radio-frequency (RF) plasma co-sputtering [24] over a bilayer WO 3 n-n heterojunction already optimized in our previous work [25]. The latter was obtained by plasma reactive sputtering, by depositing two consecutive layers of WO 3 at different pressures. In fact, it was demonstrated that the deposition pressure affects the position of the conduction band (CB) energy of the material, resulting in a built-in electric field at the interface between the two WO 3 layers. This obviously enhances the photo-promoted electron transfer.
The obtained WO 3 /BiVO 4 system can, in principle, exploit the visible-light photoactivity of BiVO 4 together with the efficient charge transport properties of the previously optimized WO 3 .
The effect that the WO 3 bilayer thickness and the post calcination temperature have on the performance of WO 3 /BiVO 4 photoanodes for water splitting was investigated under simulated AM 1.5 G solar light irradiation. The individual materials were also tested for comparison.

Photoanodes Preparation
The WO 3 and BiVO 4 individual photoanodes were prepared by RF diode plasma sputtering [26] and RF magnetron plasma co-sputtering [24], respectively. Both semiconductors were deposited on 10 × 10 mm 2 , 2 mm thick fluorine-doped tin oxide (FTO) conductive glass. For WO 3 photoanodes preparation, the substrates were placed at 4.5 cm from the W (99.9%) metal target in a reactive environment (40% O 2 /Ar). The reactor was evacuated to less than 10 −4 Pa before deposition. The sputtering power was set to 140 W and 1500 V of DC self-bias voltage. WO 3 was grown with a bilayer deposition strategy using two different gas pressures. The first layer was deposited at 3 Pa, followed by the deposition of the second layer at 1.7 Pa. The deposition time was set in order to obtain three samples with a thickness of 200, 500 and 1000 nm. Each photoanode was formed by two layers with the same thickness. The photoelectrodes were then calcined in air at 600 • C for 2 h.
The BiVO 4 films were deposited with a co-sputtering approach using separate bismuth oxide (Bi 2 O 3 , 99.9%) and metal vanadium (V, 99.99%) targets, 3.0 in. In diameter (Testbourne Ltd., Basingstoke, UK), with two separate power supplies in order to tune their power independently. The RF power of the metal vanadium target was fixed at 300 W (ca. 6 W/cm 2 ), while that of Bi 2 O 3 was fixed at 20 W (ca. 0.44 W/cm 2 ). The FTO substrate was placed at 10 cm distance from the targets on a rotating substrate holder (1.5 rev/min) to guarantee a uniform deposition. The chamber was evacuated to less than 10 −4 Pa before deposition. The deposition was performed in a 10% O 2 /Ar reactive mixture at a constant pressure of 1.7 Pa and for a deposition time necessary to achieve a thickness of 200 nm. The obtained photoelectrodes were then annealed in air at 400 or 600 • C for 2 h.
A series of heterojunction photoelectrodes was prepared by depositing a 200 nm-thick BiVO 4 layer above the WO 3 bilayer pre-calcinated at 600 • C for 2 h, with different thicknesses. After BiVO 4 deposition, the photoelectrodes were calcinated at 400 • C for 2 h. Finally, one more heterojunction photoelectrode was prepared by depositing a 200 nm-thick BiVO 4 layer above a pristine (i.e., not pre-annealed at 600 • C) 1000 nm-thick WO 3 double layer and subsequently calcined at 600 • C for 2 h.

Photoanodes Characterization
In order to measure the film thickness, a portion of sample was covered with a silicon mask during deposition. After taking it away, the height difference between the deposited and non-deposited parts was measured with a P15 surface profilometer (KLA Tencor, San Jose, CA, USA).
The morphology of the photoelectrodes was investigated by scanning electron microscopy (SEM) using a LEO 1430 microscope (Zeiss, Jena, Germany) equipped with an INCA Energy Dispersive X-Ray Spectroscopy (EDX) Detector (Oxford Instruments, Buckinghamshire, UK) and by atomic force microscopy (AFM) in air by means of a Nano-RTM System Pacific Nanotechnology (Santa Clara, CA, USA). The AFM images were acquired in contact mode and the measurements were recorded at a scan speed of 0.5 µm/s for a scanned area of 5 × 5 µm 2 .
X-ray diffraction (XRD) patterns of the deposited coatings were recorded with a PW3020 powder diffractometer (Philips, Amsterdam, the Netherlands), using the Cu K radiation (λ = 1.5418 Å) .
UV-Vis diffuse reflectance spectra (DRS) were recorded in the 220-800 nm range with a UV3600 Plus spectrophotometer (Shimadzu, Kyoto, Japan) equipped with an ISR-603 integrating sphere, using barium sulfate as blank.

Photoelectrocatalytic Water Splitting Tests
The PEC activity was tested using a three-electrode homemade PEC cell, with the film deposited on FTO as working electrode, a Pt wire as counter electrode and a saturated calomel electrode (SCE) as reference. They were all connected to an Amel (Milan, Italy), mod. 2549 potentiostat/galvanostat and immersed in a 0.5 M Na 2 SO 4 electrolyte solution. All measurements were performed under N 2 bubbling into the electrolyte solution; N 2 bubbling always started 20 min prior to the beginning of the tests. The samples were tested by means of linear sweep voltammetry (LSV, scan speed 5 mV s −1 ) and chronoamperometry (CA), both under simulated AM 1.5 G solar light (300 W LOT-qd Xe lamp, equipped with an AM 1.5 G filter). The photoanodes were placed at ca. 50 cm from the light source in order to have an incident power of 100 mWcm −2 on the photoactive surface. The incident photon to current efficiency (IPCE) curves were measured using the set-up described elsewhere [25]. The incident power was measured with a calibrated Thorlabs S130VC photodiode connected to a Thorlabs PM200 power meter. The electrodes were tested under both front-and back-side (i.e., through the FTO support) irradiation.

Photoanodes Characterization
The WO 3 and BiVO 4 coatings prepared as above described display the characteristic XRD peaks of the orthorhombic and monoclinic crystal structures, respectively, as already reported in our previous works [24,25]. The SEM micrograph of the cross section of the multilayer WO 3 /BiVO 4 /FTO photoanode ( Figure 1a) confirms an overall film thickness of 1.28 µm due to a ca. 1.05 µm thick WO 3 double layer and a ca. 238 nm-thick BiVO 4 layer on top of it.
The EDX spectrum of the BiVO 4 coatings reveals a V:Bi atomic ratio of 0.8, i.e., the coatings obtained under the adopted experimental conditions are Bi-rich. This result reflects the difficult tuning of the V:Bi atomic ratio during the co-sputtering process. The average surface roughness of the coating is 20 nm according to AFM measurements ( Figure 1b). The UV-vis DRS spectra reported in Figure 1c show that the absorption edge of the individual WO 3 coating is located at ca. 450 nm while that of BiVO 4 , calcined at either 400 or 600 • C, extends up to ca. 500 nm. The two heterojunction samples are dominated by the absorption properties of BiVO 4 . The direct band gap energy determined by the Tauc plot ( Figure 1d) confirms that the two heterojunction samples show a band gap value similar to that of the single BiVO 4 coating (i.e., ca. 2.5 eV; corresponding to a photon wavelength of 496 nm). The individual WO 3 coating possesses a larger band gap energy of 2.8 eV, corresponding to a 443 nm wavelength.

PEC Performance of Individual WO3 and BiVO4 Photoanodes
The effect of the WO3 thickness on the LSV chopped curves is reported in Figure 2. The photocurrent density as a function of the applied potential on the as-prepared samples was recorded under both front ( Figure 2a) and back (Figure 2b) irradiation. Similar PEC performances were obtained for the two modes of irradiation, proving an excellent photo-promoted electron transfer through the WO3 bilayer. The film thickness apparently does not significantly affect the PEC performance. However, during the subsequent chronoamperometry (CA) test at 1.0 V vs. SCE ( Figure  2c), the photocurrent density significantly increases along the 3 h irradiation time, revealing a clear effect of the film thickness at the end of the run, the WO3 1000 nm-thick photoanode being the most performing one. This is reasonably due to the increasing light absorption capability with an increase of film thickness. The increasing PEC performance with increasing WO3 film thickness is further proved by the LSV curves recorded after CA tests ( Figure 2d).
The BiVO4 photoanode shows a clear difference in photocurrent density between the two illumination conditions (Figure 3a,3b), generating higher photocurrent when irradiated through FTO (back-side irradiation). This confirms the poor mobility of photo-promoted electrons through the BiVO4 material [27]. Moreover, the sample post-calcined at 600 °C shows a slightly higher performance than that calcined at 400 °C. The photoactivity of this BiVO4 coating is lower with respect to that of the best performing one reported in our previous work [24] prepared under similar sputtering conditions, due the Bi-rich atomic ratio of the here investigated material, revealed by EDX analysis. Indeed, the performance of BiVO4 is strongly affected by the Bi:V atomic ratio, stoichiometric 1:1 ratio materials being best performing. The time stability test (CA at 1.0 V vs. SCE, Figure 3c), though confirming the higher performance of the sample calcined at 600 °C, also reveals a partial loss of activity along the irradiation time. Interestingly, when the light is turned on the two samples generate a similar photocurrent density, but the BiVO4 coating post calcined at 600 °C shows a marked increase during the first 20 min of irradiation.

PEC Performance of Individual WO 3 and BiVO 4 Photoanodes
The effect of the WO 3 thickness on the LSV chopped curves is reported in Figure 2. The photocurrent density as a function of the applied potential on the as-prepared samples was recorded under both front ( Figure 2a) and back (Figure 2b) irradiation. Similar PEC performances were obtained for the two modes of irradiation, proving an excellent photo-promoted electron transfer through the WO 3 bilayer. The film thickness apparently does not significantly affect the PEC performance. However, during the subsequent chronoamperometry (CA) test at 1.0 V vs. SCE (Figure 2c), the photocurrent density significantly increases along the 3 h irradiation time, revealing a clear effect of the film thickness at the end of the run, the WO 3 1000 nm-thick photoanode being the most performing one. This is reasonably due to the increasing light absorption capability with an increase of film thickness. The increasing PEC performance with increasing WO 3 film thickness is further proved by the LSV curves recorded after CA tests (Figure 2d).
The BiVO 4 photoanode shows a clear difference in photocurrent density between the two illumination conditions (Figure 3a,b), generating higher photocurrent when irradiated through FTO (back-side irradiation). This confirms the poor mobility of photo-promoted electrons through the BiVO 4 material [27]. Moreover, the sample post-calcined at 600 • C shows a slightly higher performance than that calcined at 400 • C. The photoactivity of this BiVO 4 coating is lower with respect to that of the best performing one reported in our previous work [24] prepared under similar sputtering conditions, due the Bi-rich atomic ratio of the here investigated material, revealed by EDX analysis. Indeed, the performance of BiVO 4 is strongly affected by the Bi:V atomic ratio, stoichiometric 1:1 ratio materials being best performing. The time stability test (CA at 1.0 V vs. SCE, Figure 3c), though confirming the higher performance of the sample calcined at 600 • C, also reveals a partial loss of activity along the irradiation time. Interestingly, when the light is turned on the two samples generate a similar photocurrent density, but the BiVO 4 coating post calcined at 600 • C shows a marked increase during the first 20 min of irradiation.   However, contrarily to WO 3 , when irradiation is chopped off (5 min in the dark), the gain of activity is progressively lost and stabilized after the fourth cycle of 30 min long irradiation. The LSV recorded under back irradiation after CA tests (Figure 3d) shows a decreased performance of the aged samples, as expected by the observed CA trend. Moreover, being the LSV a transient experiment, the performance of the two samples is similar, as expected by the similar photocurrent density generated during the CA experiment as the light is turned on. Figure 4 shows the PEC performance of the WO 3 /BiVO 4 heterojunction as a function of the WO 3 film thickness. All samples were annealed at 600 • C after WO 3 deposition and at 400 • C after BiVO 4 deposition. The PEC performance of the heterojunction electrodes shows a behavior partly common to those of individual WO 3 and BiVO 4 electrodes, namely: (i) Backside irradiation leads to higher generated photocurrent, as in the case of BiVO 4 (Figure 4a However, contrarily to WO3, when irradiation is chopped off (5 min in the dark), the gain of activity is progressively lost and stabilized after the fourth cycle of 30 min long irradiation. The LSV recorded under back irradiation after CA tests (Figure 3d) shows a decreased performance of the aged samples, as expected by the observed CA trend. Moreover, being the LSV a transient experiment, the performance of the two samples is similar, as expected by the similar photocurrent density generated during the CA experiment as the light is turned on. Figure 4 shows the PEC performance of the WO3/BiVO4 heterojunction as a function of the WO3 film thickness. All samples were annealed at 600 °C after WO3 deposition and at 400 °C after BiVO4 deposition. The PEC performance of the heterojunction electrodes shows a behavior partly common to those of individual WO3 and BiVO4 electrodes, namely: i) Backside irradiation leads to higher generated photocurrent, as in the case of BiVO4 (Figure 4a,b); ii) the PEC performance increases during CA tests, as in the case of WO3 (Figure 4c); and iii) the performance increases with an increase of the WO3 double layer thickness. The highest photocurrent density (i.e., 1.05 mA cm −2 at 1.4 V vs. SCE) was attained with the 1000 nm-thick WO3 film illuminated from the backside.   The effect of the WO 3 layer thickness becomes more evident after their stability test (Figure 4c,d). In fact, the WO 3 1000 nm/BiVO 4 sample benefits most from the "photo-assisted activation", reaching a maximum photocurrent of 1.7 mA cm −2 at 1.4 V vs. SCE (Figure 4c), with an 80% increase with respect to the as-prepared heterojunction electrode. This corresponds to ca. two-fold and to ca. five-fold increase compared to the photocurrent produced by individual 1000 nm-thick WO 3 (0.8 mA cm −2 ) and BiVO 4 (0.35 mA cm −2 ), respectively. Moreover, the overall photocurrent density generated by the heterojunction electrode significantly exceeds the sum of the photocurrent densities generated by the individual semiconductors ( Figure 5). Thus, it confirms the capability of the heterojunction structure to attain an efficient space separation of the charge carriers due to the optimal band alignment of these two semiconductor materials.

PEC Performance of the WO3/BiVO4 Heterojunction
The effect of the WO3 layer thickness becomes more evident after their stability test (Figure 4c,d). In fact, the WO3 1000 nm/BiVO4 sample benefits most from the "photo-assisted activation", reaching a maximum photocurrent of 1.7 mA cm −2 at 1.4 V vs. SCE (Figure 4c), with an 80% increase with respect to the as-prepared heterojunction electrode. This corresponds to ca. two-fold and to ca. fivefold increase compared to the photocurrent produced by individual 1000 nm-thick WO3 (0.8 mA cm −2 ) and BiVO4 (0.35 mA cm −2 ), respectively. Moreover, the overall photocurrent density generated by the heterojunction electrode significantly exceeds the sum of the photocurrent densities generated by the individual semiconductors ( Figure 5). Thus, it confirms the capability of the heterojunction structure to attain an efficient space separation of the charge carriers due to the optimal band alignment of these two semiconductor materials.
Finally, we explored the effect of the post calcination temperature on the PEC performance of the heterojunction photoanodes ( Figure 5). Two samples were prepared under the same WO3 and BiVO4 sputtering conditions. One sample was post annealed at 600 °C for 2 h after WO3 deposition over FTO and then at 400 °C for 2 h after BiVO4 deposition (as the above investigated series of samples). This sample is labelled as WO3/BiVO4_400C. The other sample was subjected to a single-step post calcination treatment at 600 °C for 2 h after BiVO4 deposition above WO3 (WO3/BiVO4_600C). Figure 5 shows that the post calcination procedure strongly affects the PEC performance of the heterojunction. Surprisingly, WO3/BiVO4_600C displays similar efficiency under both front and backside irradiation, reaching ca. 1 mA cm −2 photocurrent density at 1.4 V vs. SCE over the as-prepared sample. Moreover, the sample Finally, we explored the effect of the post calcination temperature on the PEC performance of the heterojunction photoanodes ( Figure 5).
Two samples were prepared under the same WO 3 and BiVO 4 sputtering conditions. One sample was post annealed at 600 • C for 2 h after WO 3 deposition over FTO and then at 400 • C for 2 h after BiVO 4 deposition (as the above investigated series of samples). This sample is labelled as WO 3 /BiVO 4 _400C. The other sample was subjected to a single-step post calcination treatment at 600 • C for 2 h after BiVO 4 deposition above WO 3 (WO 3 /BiVO 4 _600C). Figure 5 shows that the post calcination procedure strongly affects the PEC performance of the heterojunction. Surprisingly, WO 3 /BiVO 4 _600C displays similar efficiency under both front and backside irradiation, reaching ca. 1 mA cm −2 photocurrent density at 1.4 V vs. SCE over the as-prepared sample. Moreover, the sample annealed at 600 • C generates ca. two-fold photocurrent density with respect to that annealed at 400 • C under front irradiation, while the two samples show similar activity under backside irradiation. The post calcination protocol affects the time stability of the samples as well. WO 3 /BiVO 4 _600C suffers from partial deactivation along the irradiation time similar to that of BiVO 4 _600C (Figure 5c). Hence, the sample calcined at 600 • C shows a similar activity to WO 3 after CA tests, which is half of the photocurrent generated by WO 3 /BiVO 4 _400C (Figure 5d).
The IPCE curves recorded after CA tests ( Figure 6) help to get a deeper insight into the different PEC performances of these two heterojunction samples.
Surfaces 2020, 3 FOR PEER REVIEW 8 annealed at 600 °C generates ca. two-fold photocurrent density with respect to that annealed at 400 °C under front irradiation, while the two samples show similar activity under backside irradiation. The post calcination protocol affects the time stability of the samples as well. WO3/BiVO4_600C suffers from partial deactivation along the irradiation time similar to that of BiVO4_600C (Figure 5c). Hence, the sample calcined at 600 °C shows a similar activity to WO3 after CA tests, which is half of the photocurrent generated by WO3/BiVO4_400C (Figure 5d). The IPCE curves recorded after CA tests ( Figure 6) help to get a deeper insight into the different PEC performances of these two heterojunction samples. The onset photoactivity of WO3 and of BiVO4 are below 450 and 500 nm, respectively, in agreement with their band gap energy determined by the Tauc plot. BiVO4_600C shows higher IPCE under backside irradiation with a maximum of 6% at 350 nm. On the other hand, the WO3 double layer confirms to be a good performing photoanode, with a maximum IPCE of 90% at 350 nm under backside irradiation. Both heterojunction samples post-calcined at different temperatures display a significantly improved photoactivity in the 450-500 nm range (i.e., in the photon energy range below the WO3 absorption onset). This result highlights the role of the WO3 double layer in improving the photoactivity of BiVO4 by providing a downhill pathway in the CB levels that allows an efficient electron-hole separation, as shown in Scheme 1a. Interestingly, the IPCE of the WO3/BiVO4_600C sample in this spectral region is higher than the IPCE of the WO3/BiVO4_400C sample, especially under front-side irradiation. However, at higher photon energy (i.e., below the WO3 absorption threshold), both heterojunction photoanodes exhibit an IPCE lower than the WO3 coating alone. This phenomenon has been already reported in the literature [14,28] and attributed [29,30] to an electronhole recombination path occurring at the interface when both semiconductors are excited, as depicted in Scheme 1b. Interestingly, the two heterojunction samples display very different behaviors in the high photon energy region (λ < 450 nm) under front-or back-side irradiation. Indeed, the WO3/BiVO4_600C sample keeps a relatively high IPCE under front-side irradiation, reaching a maximum of 23% at 370 nm, with a minor loss with respect to single WO3. In contrast, under backside irradiation the same photoelectrode shows a maximum IPCE of 27% at 440 nm (i.e., in correspondence of the WO3 absorption onset), followed by a dramatic loss of efficiency at lower wavelengths, while WO3/BiVO4_400C shows an increasing IPCE up to 60% at 350 nm (though lower than that of WO3 alone). This dramatic loss of efficiency of the WO3/BiVO4_600C sample below 450 nm is responsible for its relative low PEC performance under simulated solar light, as shown in Figure 5c,d. This result suggests that the single step post annealing treatment at 600 °C very likely produces a better contact at the interface between the two semiconductor materials that favors a more efficient electron transfer at λ > 450 nm (Scheme 1a), but also a higher probability of charge carriers The onset photoactivity of WO 3 and of BiVO 4 are below 450 and 500 nm, respectively, in agreement with their band gap energy determined by the Tauc plot. BiVO 4 _600C shows higher IPCE under backside irradiation with a maximum of 6% at 350 nm. On the other hand, the WO 3 double layer confirms to be a good performing photoanode, with a maximum IPCE of 90% at 350 nm under backside irradiation. Both heterojunction samples post-calcined at different temperatures display a significantly improved photoactivity in the 450-500 nm range (i.e., in the photon energy range below the WO 3 absorption onset). This result highlights the role of the WO 3 double layer in improving the photoactivity of BiVO 4 by providing a downhill pathway in the CB levels that allows an efficient electron-hole separation, as shown in Scheme 1a. Interestingly, the IPCE of the WO 3 /BiVO 4 _600C sample in this spectral region is higher than the IPCE of the WO 3 /BiVO 4 _400C sample, especially under front-side irradiation. However, at higher photon energy (i.e., below the WO 3 absorption threshold), both heterojunction photoanodes exhibit an IPCE lower than the WO 3 coating alone. This phenomenon has been already reported in the literature [14,28] and attributed [29,30] to an electron-hole recombination path occurring at the interface when both semiconductors are excited, as depicted in Scheme 1b. Interestingly, the two heterojunction samples display very different behaviors in the high photon energy region (λ < 450 nm) under front-or back-side irradiation. Indeed, the WO 3 /BiVO 4 _600C sample keeps a relatively high IPCE under front-side irradiation, reaching a maximum of 23% at 370 nm, with a minor loss with respect to single WO 3 . In contrast, under back-side irradiation the same photoelectrode shows a maximum IPCE of 27% at 440 nm (i.e., in correspondence of the WO 3 absorption onset), followed by a dramatic loss of efficiency at lower wavelengths, while WO 3 /BiVO 4 _400C shows an increasing IPCE up to 60% at 350 nm (though lower than that of WO 3 alone). This dramatic loss of efficiency of the WO 3 /BiVO 4 _600C sample below 450 nm is responsible for its relative low PEC performance under simulated solar light, as shown in Figure 5c,d. This result suggests that the single step post annealing treatment at 600 • C very likely produces a better contact at the interface between the two semiconductor materials that favors a more efficient electron transfer at λ > 450 nm (Scheme 1a), but also a higher probability of charge carriers recombination (according to Scheme 1b) when both semiconductors are excited (i.e., λ < 450 nm). The higher IPCE attained with this sample under front-side irradiation reflects its intrinsic higher efficiency, because under this irradiation mode, even at λ < 450 nm, the largest part of the photons are absorbed by the outermost BiVO 4 layer, thus reducing the probability of electron-hole recombination at the interface.
Surfaces 2020, 3 FOR PEER REVIEW 9 recombination (according to Scheme 1b) when both semiconductors are excited (i.e., λ < 450 nm). The higher IPCE attained with this sample under front-side irradiation reflects its intrinsic higher efficiency, because under this irradiation mode, even at λ < 450 nm, the largest part of the photons are absorbed by the outermost BiVO4 layer, thus reducing the probability of electron-hole recombination at the interface.

Conclusions
Reactive RF plasma (co)sputtering demonstrates to be an effective technique for the deposition of multilayered WO3/BiVO4 heterojunction photoanodes for solar-driven water splitting. This can open the way to a future scale-up of the PEC technology by deposition of large surface area photoanodes. The thickness of the WO3 double layer, the post annealing treatment and the irradiation mode (back-or front-side) strongly affect the efficiency of the WO3/BiVO4 heterojunction. In this work, the best performing heterojunction was obtained by a two-step annealing treatment (i.e., over a 1000 nm-thick WO3 double layer calcined at 600 °C followed by deposition of a 200 nm-thick BiVO4 layer further calcined at 400 °C ). The electron-hole recombination at the interface between the two semiconductors occurring at photon energy above the WO3 band gap was confirmed to be the main drawback of this system, especially for heterojunction photoanodes prepared in a single-step calcination treatment at 600 °C. Nonetheless, coupling two semiconductor materials with appropriate bands alignment confirms to be a very promising approach to improve charge carrier separation and, hence, significantly increase the PEC performance under solar light irradiation.

Conclusions
Reactive RF plasma (co)sputtering demonstrates to be an effective technique for the deposition of multilayered WO 3 /BiVO 4 heterojunction photoanodes for solar-driven water splitting. This can open the way to a future scale-up of the PEC technology by deposition of large surface area photoanodes. The thickness of the WO 3 double layer, the post annealing treatment and the irradiation mode (back-or front-side) strongly affect the efficiency of the WO 3 /BiVO 4 heterojunction. In this work, the best performing heterojunction was obtained by a two-step annealing treatment (i.e., over a 1000 nm-thick WO 3 double layer calcined at 600 • C followed by deposition of a 200 nm-thick BiVO 4 layer further calcined at 400 • C). The electron-hole recombination at the interface between the two semiconductors occurring at photon energy above the WO 3 band gap was confirmed to be the main drawback of this system, especially for heterojunction photoanodes prepared in a single-step calcination treatment at 600 • C. Nonetheless, coupling two semiconductor materials with appropriate bands alignment confirms to be a very promising approach to improve charge carrier separation and, hence, significantly increase the PEC performance under solar light irradiation. Funding: This research was funded by "Ministero dell'istruzione, dell'università e della ricerca, MIUR", PRIN2017, grant number 20179337R7 (MULTI-e project)" and PRIN2015, grant number 2015K7FZLH (Solar-driven chemistry: new materials for photo-and electro-catalysis). It was also supported by the University of Milano through the project PSR2017, linea 2, Azione A" and within the CNR-Regione Lombardia Agreement No. 19366/RCC, January 10th, 2017, Decr. Reg. No. 7784-05/08/2016.