Photocharged BiVO4 photoanodes for improved solar water splitting

Bismuth vanadate (BiVO 4 ) is a promising semiconductor material for the production of solar fuels via photoelectrochemical water splitting, however, it su ﬀ ers from substantial recombination losses that limit its performance to well below its theoretical maximum. Here we demonstrate for the ﬁ rst time that the photoelectrochemical (PEC) performance of BiVO 4 photoanodes can be dramatically improved by prolonged exposure to AM 1.5 illumination in the open circuit (OC) con ﬁ guration. Photoanodes subjected to such light treatment achieve a record photocurrent for undoped and uncatalysed BiVO 4 of 3.3 mA cm (cid:1) 2 at 1.23 V RHE . Moreover, photoelectrochemical tests with a sacri ﬁ cial agent yield signi ﬁ cantly enhanced catalytic e ﬃ ciency over the whole operating potential range, suggesting elimination of major losses at the semiconductor – electrolyte interface. Finally, we demonstrate that this so-called ‘ photocharging ’ technique induces a considerable cathodic shift in the photocurrent onset potential and increases the photovoltage extracted from BiVO 4 photoanodes.


Introduction
Photoelectrochemical (PEC) water splitting offers an efficient and sustainable way to store solar energy in the form of chemical bonds. [1][2][3] One of the most attractive ways to achieve this solar driven water splitting is by using light-absorbing semiconductors to capture and convert sunlight into chemical energy by driving the corresponding water reduction and oxidation reactions. Bismuth vanadate (BiVO 4 ) is a promising n-type semiconducting material for photoelectrochemical water splitting devices due to its ability to absorb visible light (band-gap energy $ 2.4 eV), a favorable conduction band edge position (just below the water reduction potential), and its relative stability in near-neutral aqueous environments. Moreover, it is made of cheap, non-toxic, earth-abundant elements and can be easily produced on a large scale. In order to make the production of solar hydrogen with BiVO 4 viable, its performance bottlenecks and limitations have to be identied and addressed accordingly.
Based on its band gap energy, BiVO 4 should be able to produce upwards of $7.5 mA cm À2 of photocurrent under AM 1.5 irradiation. 4,5 However, to date, this photoanode has achieved far less than its full potential. It has been suggested that BiVO 4 suffers from poor surface catalytic activity, and thus many strategies have focused on passivating or putting dedicated oxygen evolution catalysts (OECs) on its surface. The results of both passivation and co-catalysis have achieved signicantly enhanced photoactivity at reduced overpotentials, but the mechanism for how these methods work are limited. In fact, a new performance benchmark for BiVO 4 photoanodes has recently been set by Kim and Choi,6 who observed photocurrent densities of 1.7 mA cm À2 (at 1.23 V RHE ) for undoped nanostructured BiVO 4 , and 4.5 mA cm À2 (at 1.23 V RHE ) for BiVO 4 catalysed with a dual layer FeOOH/NiOOH OEC. Many other attempts have been made to improve the surface catalytic reactivity of BiVO 4 by the addition of other OEC's such as Co-Pi, Co-Bi and Ni-Bi. 7-10 However, strong parasitic light absorption by OECs (up to 20% for layers as thin as 2 nm) 11 can make them prohibitive to use in tandem water splitting devices that require illumination through the electrolyte/OEC interface. In addition, the catalytic activity at low bias potentials remains low for photoanodes even when modied with co-catalysts. It is therefore of paramount importance to understand the physical and chemical nature of the semiconductor/electrolyte interface in BiVO 4 photoanodes, specically the role of electronic surface states that may constitute the favorable pathways for unwanted recombination. Therefore new strategies to tackle surface catalytic reactivity are necessary.
In addition to its poor surface catalytic reactivity, BiVO 4 also suffers from poor charge separation efficiency due to bulk recombination losses. It has been reported that spray deposited BiVO 4 photoanodes have 35% charge separation efficiency at 1.23 V RHE , 12 showing that bulk processes are severely limiting its performance. Previous research has aimed at tackling this problem by introducing elemental doping, most notably with W and Mo. [12][13][14] Of particular note, the most signicant improvement in the charge separation efficiency was reported by utilizing a novel W-gradient doping prole in the BiVO 4 , which signicantly increased the charge separation efficiency in the bulk of the photoanode. 15 Herein we report a new photoanode treatment technique made possible by so-called 'photocharging'. We demonstrate that the PEC performance of BiVO 4 photoanodes can be dramatically improved by prolonged exposure to AM 1.5 illumination in the open circuit (OC) conguration. Such a light treatment results in double the record benchmark photocurrent density for undoped and uncatalyzed BiVO 4 at 1.23 V RHE , a 0.3 V cathodic shi of the photocurrent onset potential, and near unity catalytic efficiency in the operational potential range. Preliminary analysis shows that this overall performance enhancement is largely due to improvements in the surface of the BiVO 4 photoanodes, though it was also observed that bulk modications may also play a role in the enhanced activity as well. Ultimately, we demonstrate a novel approach to investigate and improve the nature of the BiVO 4 /electrolyte interface leading to a new benchmark performance for un-doped and un-catalyzed BiVO 4 photoanodes.

Preparation of BiVO 4 thin lm photoanodes
Thin lms of 200 nm thick BiVO 4 were prepared by spray pyrolysis on FTO coated glass substrates. Details of the deposition procedure are described elsewhere 5,12 and are also available in the ESI. † Prior to deposition of the BiVO 4 , a SnO 2 interfacial layer ($80 nm) was deposited onto the FTO substrate at 425 C to prevent recombination at the FTO/BiVO 4 interface. 16 The substrate temperature during spraying of BiVO 4 was maintained at 450 C. Aer deposition, the samples were further annealed for 2 h in a tube furnace at 450 C in air.

Photoelectrochemical (PEC) measurements
Photoelectrochemical characterization of BiVO 4 photoanodes was carried out in an electrochemical cell using a three-electrode conguration. A 0.1 M phosphate buffer (0.1 K-Pi) was used as electrolyte, prepared by dissolving K 2 HPO 4 (98%, Sigma) and KH 2 PO 4 (99%, Sigma) in Milli-Q water (18.2 MU cm) to obtain pH 7.2. The potential of the working electrode was controlled by a multi-channel potentiostat (Parstat MC, Princeton Applied Research). An Ag/AgCl electrode (XR300, saturated KCl + AgCl solution (KS120), Radiometer Analytical) and a coiled Pt wire were used as the reference and counter electrodes, respectively. Measurements under illumination were performed with a Newport Sol3A Class AAA solar simulator (type 94023A-SR3) producing simulated AM 1.5 solar illumination (100 mW cm À2 ). In all the experiments involving illumination, BiVO 4 samples were illuminated from the back-side, i.e. the light came through the substrate side rst.
The photovoltage of BiVO 4 photoanodes was investigated with a series of open circuit (OC) measurements; potential of the working electrode (V WE ) was monitored in the dark and under AM 1.5 illumination until an equilibrium plateau in the V WE (t) was reached.
The photocharging of BiVO 4 photoanodes was performed in OC conditions under illumination. BiVO 4 samples were placed in the electrochemical cell and exposed to AM 1.5 light until the plateau of V WE (t) was reached (10 hours on average).
Cyclic voltammetry scans were taken at a scan speed of 50 mV s À1 unless stated otherwise. To examine the catalytic efficiency, 1.5 ml of H 2 O 2 (30%, Merck) was added to the electrolyte solution.
X-ray photoelectron spectroscopy (XPS) XPS experiment was performed using the Thermo Scientic K-alpha apparatus equipped with an Al K-alpha X-ray Source and a Flood Gun. Parameters used for the measurements were: spot size of 400 mm, pass energy of 50 eV, energy step size of 0.1 eV, dwell time of 50 ms, 10 scans in the vicinity of V2p orbital binding energy. XPS spectra were corrected for the C peak position.

Photoelectrochemical (PEC) measurements
The PEC performance of the untreated BiVO 4 photoanodes was rst measured under AM 1.5 irradiation to obtain the photocurrent density as a function of the applied potential, as shown in Fig. 1. The untreated samples show a photocurrent onset potential (V on ) of $0.7 V RHE , and a photocurrent density of 0.8 mA cm À2 at 1.23 V RHE (solid black curve). Within the range of potentials applied, the current density vs. voltage (J-V) curve has a very poor slope, a concave shape, and doesn't obtain a saturated current density, suggesting a substantial amount of surface recombination. It has been previously shown, 12 that H 2 O 2 is an excellent hole scavenger for BiVO 4 photoanodes in these conditions, and thus it is proposed that when using H 2 O 2 in the electrolyte solution, the performance can be estimated to yield near 100% catalytic efficiency. Therefore, using an electrolyte with H 2 O 2 provides the benchmark for the highest expected activity that these photoelectrodes can achieve if there are no surface recombination losses. When the un-treated BiVO 4 photoanodes were measured in an electrolyte containing H 2 O 2 , the JV curves show a remarkable, but expected, improvement in both the on-set potential (V on ¼ 0.6 V RHE ) and photocurrent density at 1.23 V vs. RHE (J ¼ 3.3 mA cm À2 ), as shown in the dashed black curve in Fig. 1. The PEC performance of the BiVO 4 photoanodes in H 2 O and in an H 2 O 2 containing electrolyte are used to act as a reference for the activity of our untreated samples in an aqueous solution, and with no surface catalytic losses.
Aer obtaining the JV characteristics of the pure BiVO 4 photoanodes, the samples were subjected to a simple pretreatment technique. The BiVO 4 photoanodes were placed in the PEC cell and held in OC conditions (i.e. no applied bias) and exposed to AM 1.5 illumination until an equilibrium potential was observed (10 hours on average). We subsequently refer to this process as photocharging. Immediately following this photocharging treatment, the JV characteristics under AM 1.5 illumination were measured for the treated samples (shown in Aer the photocharging treatment, the shape of the JV characteristic becomes convex and a record photocurrent of 3.3 mA cm À2 at 1.23 V RHE for undoped and uncatalyzed BiVO 4 was achieved. In addition, the treated photoanodes show a substantial cathodic shi of the V on by $0.3 V. Remarkably, the J-V of the photocharged photoanode follows almost exactly the photocurrent of the un-treated BiVO 4 measured in the presence of H 2 O 2 . The shape of the J-V curve suggests 17 major suppression of the surface recombination due to the very steep photocurrent slope, relatively low applied potential required to achieve a saturated current density, and convex shape of the curve. The PEC performance of the photocharged BiVO 4 samples was similarly tested aer adding H 2 O 2 to the electrolyte, as shown in the dashed red curve in Fig. 1. Surprisingly, the JV characteristics under these conditions showed a further decreased onset potential, V on ¼ 0.43 V RHE , and a larger photocurrent density at 1.23 V RHE , J ¼ 4.0 mA cm À2 . To further quantify the change in the surface reactivity for each of the BiVO 4 photoanodes (before and aer photocharging), the catalytic efficiency, h cat , was calculated using the method developed by Dotan et al., 18 where the current density in the normal aqueous solution is directly compared to that obtained in the presence of H 2 O 2 , as shown in Fig. 2. It is conrmed that the untreated BiVO 4 photoanode has very poor catalytic activity, shown by a very low h cat (maximum of 25% at 1.23 V RHE ). In contrast, the h cat of the photocharged sample showed signicant improvements as shown in the dashed and solid red curves. If the photocharged BiVO 4 photoanode was compared to the initial untreated sample measured in H 2 O 2 , the dashed red curve in Fig. 2 is obtained, showing unity h cat across the entire potential range. This is expected since the JV curves overlap for the two samples. However, if the photocharged sample was compared to itself with and without H 2 O 2 , the solid red curve in Fig. 2 is obtained. The h cat for this newly photocharged sample now reaches a maximum of 80% at 1.23 V RHE , and even achieves reasonable efficiencies at lower potentials, with a h cat of 55% at 0.8 V RHE . These results have three major implications: (1) the photocharging treatment signicantly improves the catalytic efficiency of the BiVO 4 photoanodes, (2) there are still some surface catalytic losses for photocharged electrodes and (3) a bulk/non-surface related improvement was also achieved. Comparing the solid red curve and the solid black curve in Fig. 2 clearly shows that there is an enhancement in the catalytic efficiency due to the photocharging treatment. However, if the improvement was achieved only due to the modication of the surface of BiVO 4 , then it would be expected that the performance in a solution containing H 2 O 2 would be identical for the untreated and photocharged samples (dashed black and dashed red in Fig. 1), which is not the case. Therefore, since the JV curves of the photocharged BiVO 4 sample and the untreated sample in H 2 O 2 overlap, and there is still a further improvement in current when a photocharged sample is measured in presence of H 2 O 2 , there must be a non-catalytic enhancement, which will be discussed later.
The PEC performance of the photocharged photoanodes was found to be stable within 20 voltammetric cycles in the potential range of 0.5-1.3 V RHE , with no observable decrease in the photocurrent density (ESI, Fig. S1 †).
However, a photocharged sample le in the electrochemical cell over-night and re-measured on the next day exhibits very similar J-V characteristics to what is measured initially, before photocharging. This shows that dark OC conditions facilitate the deactivation of the improved PEC performance. Nonetheless, the process of photocharging repeated later on is equally effective and results in recovery of the improved performance (ESI, Fig. S2 †). This nding may limit this technique to have a lasting effect in a practical device for PEC water splitting, but a prospective challenge for further development would be to 'lock-in' the photocharging effect on the photocharged BiVO 4 surface.

Photovoltage of BiVO 4
Numerous reports have shown that the performance of metal oxide photoanodes can be greatly hindered by the presence of electronic surface states (SuS). 19,20 It has been proposed that the electronic nature of the SuS results in the pinning of the Fermi level (E F ), which has a detrimental effect on the obtainable photovoltage (V ph ). Likewise, it has been demonstrated that these SuS can be passivated, leading to an increased V ph , a cathodically shied V on and higher photocurrent densities. [21][22][23][24][25] Therefore, to better understand whether the observed change in PEC properties of our BiVO 4 photoanodes is related to the presence of SuS, an investigation of the change in V ph for our samples before and aer the photocharging treatment has been conducted.
The equilibration conditions of n-type semiconductors in aqueous-based electrolytes have been described in detail in previously reported literature. [26][27][28] It has been proposed that in the dark and in an OC conguration and ideal conditions, i.e. with a pristine electrode surface with no surface defect states, the Fermi level (E F ) of a photoanode equilibrates with the water oxidation potential (E o OH À =O2 ¼ 1.23 V RHE ) ( Fig. 3(a)). Whereas under illumination, the quasi-Fermi level of the electrons (E F,n ) shis towards the conduction band edge (E S C ) and quasi-Fermi level of the holes (E F,p ) equilibrates with E o OH À =O2 (Fig. 3(b)). However, when surface states with sufficient density (DOS $ 10 13 to 10 14 ) 29 exist at the semiconductor/electrolyte interface, a signicant potential drop in the Helmholtz layer is likely to develop. 30 Consequently the E F doesn't equilibrate with E o OH À =O2 and instead aligns with the energy level of the SuS (E SuS ) 21,28 (Fig. 3(c)). The potential drop in the Helmholtz layer results in a so-called Fermi level pinning effect. 31 Under Fermi level pinning conditions, any change in pH or applied potential is accommodated by a change in the occupation of the SuS and a corresponding change in the potential drop across the Helmholtz layer. 28 As a result, the Fermi level stays virtually xed and the degree of band bending remains constant, and a signicantly lower photovoltage is obtained.
In our experiments, we performed a set of PEC measurements in the OC conguration to investigate the behaviour of our BiVO 4 photoanodes in this respect. The potential of the working electrode (V WE ) was monitored in the dark and under illumination until a steady state equilibrium was observed, i.e. the potential of the working electrode remained constant over time (ESI, Fig. S3 and S4 †). The photovoltage (V ph ) was calculated as V WE,dark À V WE,light . Fig. 4 contains a summary of the measured and estimated OC potentials in the dark and under illumination for BiVO 4

photoanodes in water (H 2 O), in the presence of hydrogen peroxide (H 2 O 2 ) and aer photocharging (H 2 O photo ).
The equilibrium potential of the untreated BiVO 4 photoanode was found to be more negative than the water oxidation potential by ca. 0.4 V. A likely reason for the mismatch between the observed value of E F and E o OH À =O2 is the presence of SuS's on the BiVO 4 surface. These results are in agreement with studies on a-Fe 2 O 3 and TiO 2 electrodes, 20 where equilibration potentials of ca. 0.8 V RHE have been reported (OC dark conditions), suggesting that this may be an effect typical for metal oxide photoanodes. This consistency implies that SuS are independent of the exact composition of a metal oxide and originate from some surface adsorbate species and/or water oxidation intermediates that bind preferentially to metal oxide surfaces. Furthermore, when a hole scavenger is present in the electrolyte, we observed values of V WE (H 2 O 2 dark) much closer to 1.23 V RHE , i.e. 1.0 V RHE . This can be explained by the fact that H 2 O 2 is not only an efficient acceptor of photogenerated holes, but can also easily donate electrons to the SuS's of BiVO 4 , as the oxidation potential of H 2 O 2 (0.68 V RHE ) is more negative than E SuS .
The dark OC potential of the photocharged anodes could not be directly determined using the standard approach. That's because dark OC conditions facilitate the deactivation of the photocharging effect, therefore the measured dark OC potential corresponds rather to the untreated material. Consequently, two different approaches were used to estimate the photovoltage of the photocharged photoanodes. It's been proposed 21 that there is direct correlation between the onset potential and the position of the dark OC potential. Therefore in the rst approach the photovoltage was estimated based on the cathodic shi of the onset potential observed for the photocharged anodes. In the second approach the dark OC potential was  assumed to be the same as the one of the untreated photoanodes.

Photocharged BiVO 4
To further explore the PEC properties of the photocharged BiVO 4 photoanodes, dark electrochemical experiments were performed before and aer photocharging. Fig. 5 shows three consecutive dark JV scans taken immediately aer the photoinduced equilibrium of E F was reached. An oxidation peak can be observed at potentials of about 0.85 V RHE . As such a peak is not present in dark JV curves taken for untreated BiVO 4 , we conclude that upon exposure to illumination in OC conditions, the electrochemical behaviour of our BiVO 4 photoanodes are altered, i.e. they attain some additional negative charge. Once the photoanodes are swept anodically with potential greater than 0.85 V RHE , the electrons are removed, which results in the presence of an oxidation peak in the JV scan. This peak appears to be the largest aer the rst cycle, and becomes successively smaller aer each cycle until aer >3 cycles the peak is gone, suggesting that within the range of potentials applied (0.2-2.1 V RHE ) and within the timescale of a CV scan, this effect is irreversible without the addition of light. The position of the observed oxidation peak (0.85 V RHE ) is consistent with the recorded V WE dark H 2 O (0.85 V RHE ), which directly corresponds to the energy level of the proposed SuS.
Additionally, current densities up to 0.043 mA cm À2 at 2.1 V RHE in the dark can be observed for photocharged BiVO 4 , suggesting some dark electrocatalytic activity (Fig. 5). Untreated BiVO 4 is a very poor electrocatalyst, with virtually no current at 2.1 V RHE in the dark (ESI, Fig. S5 †). Even with a thin layer of OEC such as Ni-Bi only 0.011 mA cm À2 can be observed at 2.1 V RHE (ESI, Fig. S5 †). It further conrms that photocharging can successfully modify the surface of BiVO 4 photoanodes leading to enhanced catalytic activity. However, it is also possible that the photocharging of BiVO 4 photoanodes is a non-surface related phenomenon. For untreated BiVO 4 photoanodes, the rst voltammetric cycle under illumination always yields the highest photocurrent density, and subsequent cycles show a gradually decreasing performance ( Fig. 6(a) inset). Usually from the 5th cycle onwards, the obtained CVs start to overlap as shown in Fig. 6(a), where the blue curve is obtained for all CVs recorded aer 5 cycles. Interestingly, this decrease in photocurrent over cycling for untreated BiVO 4 is accompanied with a change in the redox waves intensities and positions at low potentials. Initially a small oxidation wave at $0.65 V RHE and a reduction wave at $0.55 V RHE can be observed. Both waves disappear over cycling and a reduction wave at $0.75 V RHE arises instead. Conversely, the photocharging technique leads to the diminishing of this reduction wave at $0.75 V RHE , as shown in Fig. 6(b). The observed redox waves have previously been shown to correspond to the reduction and re-oxidation of vanadium V 4+ /V 5+ , which may be a bulk process. 32 The fact that photocurrent density is related to the intensity of the redox waves, and hence to the concentration of V 4+ and V 5+ species, suggests that photocharging leads to alteration of the V 4+ to V 5+ ratio.
To further investigate the possible change in vanadium oxidation state, XPS measurements were performed on the BiVO 4 photoanodes before and aer photocharging. Prior to the XPS experiment two samples were photocharged in K-Pi buffer pH 7.2; one for 8 h, and another one for 16 h. One other sample, an untreated one, was tested as a reference. Measurements were focused in the energy range expected for vanadium, and the results are presented in Fig. 7.  A clear shi of the V2p peak towards lower binding energies can be observed for photocharged samples (Fig. 7a). The position of V2p peak as a function of photocharging time is plotted in Fig. 7b. It shows quantitatively that the negative shi of the binding energy for V2p 3/2 orbital scales up linearly with photocharging time. Moreover, it has been previously reported 33 that the binding energy of vanadium in the 5+ oxidation state is found at 517.1 eV, and is correlated to the V 2 O 5 species, while the binding energy of vanadium in the 4+ oxidation state is 516.4 eV. Comparison of our data with these V2p 3/2 references for different oxidation states (dotted lines on Fig. 7b) suggests that photocharging treatment results in gradual reduction of V in our BiVO 4 lms, i.e. the untreated samples are rich in V in 5+ state, while photocharging results in higher and higher fraction of V 4+ .
It is important to note that the XPS technique only gives quantitative information about the oxidation state of the surface of the sample, and therefore we cannot conclude that the oxidation state of vanadium changes in the bulk. Ion beam etching was attempted to probe the oxidation state of vanadium at a depth several nanometers below the surface for our samples, but due to the complex morphology and composition of our lms, reliable data was not obtainable. Therefore we can conclude that at least the surface vanadium species reduce during the photocharging process, but further investigations are required to determine if the bulk vanadium also reduces.

Material properties of photocharged BiVO 4
To probe whether the photocharging treatment leads to any changes in the microstructural, structural and/or chemical properties of our BiVO 4 lms, a detailed material characterization was performed before and aer photocharging. Specically, the aim was to clarify if the photocharging treatment leads to any corrosion process of BiVO 4 , enhancement/change in the surface area, or any other surface altering phenomena. The faradaic efficiency is also included to test if the enhancement in photocurrent generation is related to water oxidation and not any other reaction.
We investigated the surface morphology before and aer photocharging by taking AFM scans of 200 nm thick BiVO 4 samples, as shown in ESI Fig. S6 and S7 † respectively. From these images, there are no clearly noticeable changes in the surface morphology before and aer the treatment. Our AFM investigation suggests that the photocharging treatment does not alter the microstructure of our lms, and therefore the observed enhancement in photocurrent appears not to be related to an increased surface roughness or etching induced nano-structuring. Our SEM study further conrms these conclusions. The surface morphology of BiVO 4 samples before and aer photocharging was investigated by SEM, as shown in ESI Fig. S8 and S9 † respectively. Similar to the AFM observations, any qualitative difference in the surface morphology before and aer the treatment is not clearly visible.
The crystal structure of BiVO 4 before and aer photocharging was also investigated by collecting XRD scans (ESI, Fig. S10 †). Our spray deposited lms show a structure typical for monoclinic scheelite-type BiVO 4 . Our diffractograms demonstrate that there is no clear difference in the crystal structure between the untreated and photocharged lms. This implies that no additional lattice stress or strain inducing defects are responsible for the enhanced performance of the photocharged BiVO 4 lms.
The chronoamperometric (J-t) tests of photocharged BiVO 4 thin lms were performed for three different potentials over the course of an hour for each potential step. We choose moderately anodic potentials to avoid any stability issues related to BiVO 4 itself, which are more likely to occur at highly oxidative potentials. We also choose these potentials for the sake of applicability of the photocharging treatment. Potentials in the 0.6-1.0 V range are close to the intersection points for a BiVO 4 /a-Si PV cell tandem device, 15 therefore they represent practical operating potentials for a potential bias free solar water splitting device.
The results of the photocurrent stability measurements are presented in ESI Fig. S11. † The photocurrent within the timeframe of our experiment seems to be stable, i.e. for all 3 potentials tested, the J-t curves look fairly at as a function of time. Larger uctuations in the photocurrent density at 1.0 V RHE are related to bubble formation. Our results suggest that the enhancement in current due to the photocharging treatment can be sustained for at least several hours. Any decrease in photocurrent of the photocharged BiVO 4 held at xed potential is therefore likely to originate from the instability of the material itself, and not as a result of the photocharging treatment.
To show that the observed enhancement in photocurrent is indeed related to O 2 evolution, a gas chromatography (GC) measurement was performed. Given the time resolution of our GC apparatus (18 min) and any potential intrinsic stability issues related to the BiVO 4 itself, we performed the GC measurements under a xed current density (0.65 mA cm À2 ) rather than at a xed potential. The outcome of the GC investigation is presented in Fig. 8. All the experimental data points are in the close proximity of the oxygen evolution trend calculated based on the applied current and the GC calibration data. Our results suggest that the faradaic efficiency of the water oxidation process on photocharged BiVO 4 photoanodes is close to unity, meaning the enhancement in photocurrent is not related to any other electron pathway or side reaction.

Conclusions
In summary, we showed that prolonged exposure of BiVO 4 photoanodes to AM 1.5 illumination in the OC conguration, a technique that we call 'photocharging', greatly improves its PEC properties. We found that such light-induced treatment results in a major suppression of surface recombination, and thus a record photocurrent for bare, undoped, uncatalysed BiVO 4 photoanodes was achieved. Furthermore, we showed that photocharging induces a signicant cathodic shi in the photocurrent onset potential, and an increased photovoltage.
According to our results, the photocharging treatment improves both bulk and surface properties of BiVO 4 . Therefore we hypothesize about the mechanisms related to these two effects. Surface related enhancement could possibly be explained by the fact that photocharging leads to a photoinduced passivation of the surface states at the solid-liquid interface. Such passivation would result in a diminished Fermi level pinning effect and consequently in an increase of the photovoltage. Bulk related effect of photocharging could possibly be explained by reduction of V 5+ to V 4+ in the bulk of the lm. Work is in progress to reveal what is the effect of photocharging on physio-chemical properties of BiVO 4 , and to further determine the exact mechanism of this process and how it leads to the increased photocatalytic activity.