The Role of Underlayers and Overlayers in Thin Film BiVO4 Photoanodes for Solar Water Splitting

Light‐driven water splitting with metal oxide semiconductor materials to produce H2 constitutes one of the most promising energy conversion technologies built on solar power. BiVO4 stands out as one of the most attractive metal oxides with reported photocurrents close to its theoretical maximum of 7.5 mA cm−2 at 1 sun illumination. The present work addresses the state‐of‐the‐art strategies to enhance the performance of this material for water oxidation by heterostructuring with different underlayer (SnO2 and WO3) and overlayer (NiOOH/FeOOH, Co–Pi, Co–Fe Prussian Blue derivative) materials, with particular emphasis on the physico‐chemical mechanisms responsible for the reported enhancements.


Introduction
Monoclinic bismuth vanadate (BiVO 4 ) stands out as the most promising metal oxide n-type semiconductor material for photoelectrochemical water oxidation. After some initial controversy, it has been demonstrated that the material has an indirect band gap energy of 2.4 eV (≈516 nm band edge), [1] which leads to a maximum theoretical photocurrent of ≈7.5 mA cm −2 under 1 sun AM 1.5G illumination and a solar to hydrogen (STH) efficiency of 9.2%, [2] which is close to the technological target for solar driven water-splitting (STH efficiency of 10%). [3] Its conduction band edge is more positive than the hydrogen evolution potential (0 V RHE , vs reversible hydrogen electrode) and consequently, photogenerated holes lie below 2.4 V RHE with a large driving force to oxidize water. Compared to other polymorphs of BiVO 4 , the higher distortion in the local environments of Bi and V ions within the monoclinic phase is believed to be at the origin of its superior photocatalytic activity. [4] Electron transport has been recognized as the main limiting factor for performance. [5,6] This poor electron transport is attributed to the crystalline structure of BiVO 4 where the VO 4 tetrahedral units are not interconnected. [7] A more recent study claims space charge-limited current in the presence of trap states with with Sn 2+ species, which can introduce deep levels inside the bandgap of FTO. [13] The presence of oxygen vacancies related defects, usually detected by luminescence, inducing deep states inside the bandgap of SnO 2 has been widely reported. [14] Electron trapping by DFS leads to a negative charging of the FTO/ BiVO 4 interface, repelling other electrons and flattening the bands. This band flattening effect allows photogenerated holes in BiVO 4 to recombine with trapped electrons at DFS, which act as recombination centers at the FTO/BiVO 4 interface. The introduction of the SnO 2 underlayer between FTO and BiVO 4 , block the path of the photogenerated holes to the DFS (hole mirror effect), enhancing charge extraction by reducing the recombination, as depicted in Figure 1a. A more detailed analysis of the role of the SnO 2 underlayer involved a thickness dependent study showing that the thicker the SnO 2 layer (from 20 nm up to 65 nm), the more effectively passivated the DFS at FTO in the FTO/SnO 2 /BiVO 4 heterostructure. [12a] This same effect has been already identified for other photoanode material like Fe 2 O 3 . [15]

WO 3 /BiVO 4
Among the different heterojunctions with BiVO 4 underlayers engineered in the last years (WO 3 /BiVO 4 , [16] SnO 2 /BiVO 4 , [13] Bi 2 WO 6 /BiVO 4 , [17] Co 3 O 4 /BiVO 4 , [18] CuO/BiVO 4 , [19] etc.), WO 3 /BiVO 4 has attracted significant attention, due to the highest water oxidation photocurrents obtained (6.72 mA cm −2 at 1.23 V vs RHE), close to the theoretical maximum (7.5 mA cm −2 ). [2] This heterojunction synergistically combines the excellent conductivity of WO 3 with the good absorption properties of BiVO 4 , in a favorable type II band alignment to promote charge separation leading to a significant reduction of charge recombination. [20] An excellent recent review has extensively described the different synthetic approaches leading to different WO 3 /BiVO 4 nanostructures. [21] On the other hand, different mechanistic studies have shed light on the carrier dynamics at different timescales in this system. At the ultrafast timescale (fs-ps), Kamat and Selli studied with transient absorption spectroscopy (TAS) the response of the heterojunction and individual components, showing that under selective BiVO 4 excitation, a favorable electron transfer from photoexcited BiVO 4 to WO 3 occurs immediately after excitation and leading to an enhanced lifetime of the trapped holes at BiVO 4 . However, upon simultaneous excitation of both oxides, a new recombination channel is activated. This is reflected on a shorter lifetime of the trapped holes in BiVO 4 . [22] Furthermore, Hammarström and Selli confirmed through nanosecond mid-IR TA experiments that charge carrier separation occurs in WO 3 /BiVO 4 electrodes under visible-light excitation, persisting up to the microsecond timescale. [23] Additionally, photogenerated electrons live much longer in WO 3 compared to BiVO 4 , in line with the far better electron conductivity of the former oxide. At more relevant timescale for water oxidation (µs-s), impedance spectroscopy measurements together with physical modeling were employed to understand the role of WO 3 on the enhanced performance of the WO 3 /BiVO 4 heterostructure. [24] It was hypothesized that the enhancement of the photoelectrochemical performance was due to either a reduction of Her main topic of interest is on the synthesis and characterization of semiconductor metal oxide thin films, for the development of efficient photoelectrochemical devices, especially those for solar fuels production from water splitting.
Sixto Giménez is currently an Associate Professor at Universitat Jaume I de Castelló (Spain). He received his B.S. in Physics at University of Granada (Spain) and his Ph.D. in Physics at University of Navarra (Spain) studying the relationship between processing of metallic and ceramic powders, their sintering behavior and mechanical properties. Currently, he is actively involved in the development of new concepts for photovoltaic and photoelectrochemical devices based on nanoscaled materials, particularly studying the optoelectronic and electrochemical responses of the devices by different spectroscopic techniques. www.advmatinterfaces.de bulk recombination or to a more favorable electron extraction kinetics at the back contact. The relative contribution of both processes could be easily evaluated by the behavior of the minimum value in the total resistance, evaluated from the j-V curve as A decreased value of the minimum was ascribed to decreased bulk recombination and a cathodic shift was related to more favorable charge extraction at the electron contact. Figure 1c schematically illustrates the dc resistance in two materials with different bulk recombination and electron extraction properties. The material (i) exhibits lower bulk recombination and enhanced electron extraction compared to (ii), as reflected by the lower and cathodically shifted value of the resistance valley. In line with this analysis, WO 3 was found to control the transport properties in the heterojunction, significantly reducing bulk recombination by boosting the charge extraction, while BiVO 4 was responsible of the enhanced light harvesting properties, which explained the synergistic effect in WO 3 /BiVO 4 heterostructure. On the other hand, Kafizas and co-workers [25] have studied the dynamics of photogenerated electron and holes in the WO 3 /BiVO 4 heterojunction at relevant timescales for water oxidation by using TAS and transient photocurrent (TPC) spectroscopy. The origin of the enhanced performance of the WO 3 /BiVO 4 heterojunction was based on the faster electron transfer from BiVO 4 to WO 3 (<µs) compared to that from BiVO 4 to FTO (≈100 µs). Since the photogenerated electrons in BiVO 4 are transferred faster to WO 3 , bulk recombination at BiVO 4 (which dominates at early time scales, <µs-ms, and explains the poor performance of bare BiVO 4 ) is significantly reduced, consequently increasing the population of photogenerated holes accumulated at the semiconductor/liquid junction leading to more favorable water oxidation conditions. It is worth noting, that even if electron extraction is slower from WO 3 to FTO (≈ms) than from BiVO 4 to FTO (≈20-100 µs), this factor does not limit the enhanced performance of the heterojunction. This was explained on the basis of fast charge transfer from BiVO 4 to WO 3 , concomitantly reducing the bulk recombination losses at BiVO 4 , as illustrated in Figure 1b.
An additional beneficial effect of heterostructuring with WO 3 underlayers is the suppresion of the "dead-layer" effect observed in the bare BiVO 4 thin layers (≤125 nm), also reported for other metal oxide photoanodes as hematite. [26] This is probably due to the reduction in lattice mismatch between the back contact (FTO) and the absorber upon deposition of the underlayer.
A more detailed theoretical-experimental study of the WO 3 /BiVO 4 interface based on density functional theory (DFT) calculations could satisfactorily explain the different photoelectrochemical performance with front and back illumination. [27] The DFT model revealed a nonfavorable alignment (nonstaggered) between the valence bands of WO 3 and BiVO 4 attributed to the hybridization of interfacial states (most likely oxygen Adapted with permission. [13] Copyright 2011, American Chemical Society. c) dc resistance illustrating two materials with different bulk recombination and surface recombination velocity at the selective contact for electrons. Adapted with permission. [24] Copyright 2016, Royal Society of Chemistry.

www.advmatinterfaces.de
anions at the WO 3 /BiVO 4 ) pinning the valence band. Upon back illumination, the photogenerated holes in WO 3 must cross the junction to reach the BiVO 4 /electrolyte interface. Since the valence band energies of both materials are pinned, there is not effective driving force to facilitate the travelling of the photogenerated holes, which can more easily recombine with the electrons at WO 3 , contrary to the favorable band alignment for the extraction of photogenerated electrons at BiVO 4 under front illumination. When the heterojunction is frontilluminated, the generation of electron-hole pairs is higher compared to the recombination at BiVO 4 , conversely to the situation in WO 3 . This has been depicted with the thick and thin arrows in Figure 1b.
In summary, the WO 3 /BiVO 4 heterojunction improves the performance for water oxidation compared to bare BiVO 4 , due to a significant reduction of bulk recombination at BiVO 4 caused by the fast transfer of the photogenerated electrons to the back contact. This faster electron extraction takes place as a result of the staggered alignement between WO 3 and BiVO 4, characteristic of a type II heterojunction. Finally, the double heterojunction SnO 2 /WO 3 /BiVO 4 has also been recently studied, showing enhanced performance with respect to the single heterostrucutured systems described above. [28] This system allows a cascade band alignment facilitating the charge transport across the interface. Moreover, the introduction of a WO 3 layer between the SnO 2 and the BiVO 4 also provides slightly enhanced optical absorption under visible light.

Other Examples
Other relevant examples of underlayers include TiO 2 , [29] and ZnO, [30] although most of the published reports are based on heterostructures, where TiO 2 or ZnO act as charge transport/ collection layers and BiVO 4 as an inorganic sensitizer, overcoming the poor charge transport properties of BiVO 4 . Lu 2 O 3 has also been reported as an efficient hole blocking underlayer, when combined with BiVO 4 , due to the significant reduction of interfacial defects at the BiVO 4 /Lu 2 O 3 and Lu 2 O 3 /ITO interfaces. [31]

Overlayers
The deposition of protective, passivating, or catalytic layers on top of BiVO 4 has also demonstrated to lead to significant functional improvements for water oxidation. The photoanode/ overlayer/electrolyte interface can be more favorable compared to the photoanode/electrolyte, since: (i) The built-in potential generated when the photoanode and the overlayer are brought together is independent of the redox potential of the electrolyte. This built-in potentail can be tuned by employing overlayers with appropiated band alignements with BiVO 4 , improving the photogenerated hole injection through the overlayer into the solution. and (ii), the surface of the overlayer can be designed to minimize the overpotential needed to start extracting photogenerated carriers. [32] On the other hand, the employed overlayers for water splitting applications are often made of noble metal catalysts. [33] However, a real alternative for cost-effective renewable energy sources cannot be based on noble metals. [34] Below, we focus on different overlayer materials based on Earth-abundant elements, like nickel oxide, iron oxide, cobalt phosphate (Co-Pi) and Prussian Blue derivatives.

BiVO 4 /NiOOH and BiVO 4 /FeOOH
At present, Earth-abundant Ni-based oxides constitute one of the most attractive water oxidation electrocatalysts to be integrated in photoelectrochemical devices for the production of solar fuels. Kim and Choi improved the electron-hole separation at BiVO 4 through nanostructuring and minimized the recombination losses at the BiVO 4 /electrolyte interface by the deposition of two different catalytic overlayers (FeOOH and NiOOH), boosting the hole injection efficiency into the solution. [35] More specifically, a record charge separation efficiency was obtained by reducing the particle size of the nanoporous BiVO 4 (30-75 nm) below the hole diffusion length (≈100 nm). However, it was observed that the majority of the photogenerated holes recombined before being injected into the solution. Consequently, two different catalytic overlayers (NiOOH and FeOOH) were deposited on top of the nanoporous BiVO 4 . Although the obtained photocurrents were significantly higher in the presence of the catalytic overlayers, they were lower compared to that of bare BiVO 4 in the hole scavenger. This suggested that the recombination pathways were not totally suppressed by these overlayers. BiVO 4 /NiOOH showed higher surface recombination and a more cathodic onset and higher photocurrents compared to BiVO 4 /FeOOH, due to a more favorable potential drop at the Helmholtz layer. As a next step, the complex systems BiVO 4 / NiOOH/FeOOH and BiVO 4 /FeOOH/NiOOH were also investigated, and a photocurrent density of 2.73 mA cm −2 at 0.6 V RHE was obtained for the latter case. This was attributed to the combination of two beneficial factors: (i) the FeOOH layer reduced the recombination at the BiVO 4 /FeOOH interface, while (ii) the NiOOH catalyst reduced the potential drop at the Helmholtz layer at the NiOOH/solution interface shifting cathodically the BiVO 4 flat band potential (which also shifted the photocurrent onset to more negative potentials). More specifically, FeOOH/ NiOOH has been reported as one of the best oxygen evolution catalysts for BiVO 4 due to its excellent surface kinetics for water oxidation. [36] Remarkable improvement of the long-term stability was achieved by saturating the electrolyte with V 5+ ions in order to prevent the photocorrosion. [37] BiVO 4 is usually unstable under anodic bias and long-term illumination. The degradation process mainly involves the dissolution of V 5+ species, caused by photo-oxidation, which segregates from the BiVO 4 lattice leading to a concomitant decrease in performance. Some of the photogenerated holes which accumulate at the BiVO 4 / electrolyte interface take part in this photocorrosion process. By saturating the employed borate buffer with V 5+ ions, the photocorrosion was totally suppressed for 60 h, without any change in the water oxidation kinetics. In addition, the presence of V 5+ ions in the electrolyte can form an interfacial Fe/Ni-V-O layer between the BiVO 4 and the NiOOH/FeOOH catalysts, inducing high stability (≈450 h) under illumination and anodic bias, enhancing water oxidation kinetics. www.advmatinterfaces.de

BiVO 4 /Co-Pi
Cobalt phosphate (Co-Pi) [38] is a well-known, efficient and earth-abundant electro-catalyst and consequently, its coupling with photoabsorbers for photoelectrochemical water splitting, including BiVO 4 , has been widely studied. Several authors have reported increased charge injection efficiency and consequently, enhanced water oxidation kinetics when using Co-Pi modified BiVO 4 photoanodes. [5,39,40] However, the origin of such improvements is currently under debate, since two different mechanisms can explain the observed enhanced photoelectrochemical behavior: (i) suppression of surface recombination at the BiVO 4 /solution interface (Figure 2a) and (ii) "true" catalytic activity via Co-Pi enhancing charge transfer kinetics (Figure 2b). Mechanism (i) was claimed first by Gamelin and co-workers, who observed a large (≈440 mV) cathodic shift of the onset potential for OER after photoassisted electrodeposition of Co-Pi onto W:BiVO 4 photoanodes. [39] Experiments with and without a sacrificial hole scavenger ( j-V curves and chronoamperometric tests) concluded that the modification of BiVO 4 photoanodes with Co-Pi, yielded to nearly complete suppression of recombination losses, together with enhanced charge injection efficiency.
A more detailed mechanistic study was carried out by Durrant and co-workers using photo induced absorption spectroscopy (PIAS) and spectroelectrochemistry (SEC). [41] A signi ficantly larger PIAS signal was observed on Co-Pi modified BiVO 4 , attributed to an additional photoinduced species, rather than to photogenerated holes in BiVO 4 . By comparing with SEC data for FTO/Co-Pi, the large PIAS signal observed for Co-Pi-modified BiVO 4 was assigned to the oxidation of Co 2+ to Co 3+ . Furthermore, the analysis of steady-state photocurrent during SEC showed that electrochemical water oxidation by Co-Pi takes place with a density of Co 3+ sites of 1 × 10 17 cm −2 , three times higher compared to that for the Co-Pi-modified BiVO 4 under water oxidation condition. Since the hole transfer kinetics from BiVO 4 was still faster compared to that via Co-Pi oxidation states, it was concluded that Co-Pi did not significantly contribute to the overall water oxidation photocurrent, which was instead enhanced by the capability of Co-Pi to retard electron/hole recombination at the BiVO 4 /solution interface. Similar conclusions were obtained by Abdi and co-workers, through incident modulated photocurrent spectroscopy (IMPS) measurements. [42] In these experiments, light intensity is used to modulate the surface concentration of the photogenerated carriers, and consequently both the potential across the space charge region and the reaction rate constants remain unaltered. Therefore, the hole injection into the electrolyte and surface recombination at the semiconductor/electrolyte interface can be clearly differentiated. At low applied potentials, surface recombination rate constant (k rec ) was reduced by the introduction of electrodeposited Co-Pi, and this was attributed to the passivation of surface recombination sites. Then, k rec decreased with increasing potential, which was attributed to the formation of oxidized species, with higher intrinsic catalytic activity compared to BiVO 4 , due to the oxidation of Co-Pi by the increasing fraction of photogenerated holes. On the other hand, the charge transfer rate constant (k tr ) was not affected in the presence of Co-Pi, suggesting that charge transfer still occurs via the BiVO 4 surface. This conclusion was also supported by in situ UV-Vis absorption measurements, also suggesting the gradual shift of water oxidation from the BiVO 4 surface to the Co-Pi, at higher applied potentials. Despite all these mechanistic studies, the chemical nature of the surface defects and the accurate passivation mechanism via Co-Pi modification are still elusive. Nonetheless, these studies clearly concluded that electrocatalysis is not the main function of Co-Pi in these systems.
A different mechanistic picture of the role of Co-Pi on BiVO 4 was provided by Boettcher and collaborators. [43,44] They could directly measure the charging of the Co-Pi overlayer by the photogenerated holes in BiVO 4 by using an electrochemical atomic force microscopy (AFM) as a second working electrode. A rise in the Co-Pi potential detected at the onset of the photocurrent revealed how the photogenerated holes in the absorber layer were collected by the CoPi catalyst. By comparing both FTO/Co-Pi and BiVO 4 /Co-Pi, it was possible to demonstrate the accumulation of photogenerated holes at BiVO 4 within the Co-Pi overlayer until enough anodic potential was reached for water oxidation. Consequently, Co-Pi was found to act as a hole reservoir of photogenerated charges at BiVO 4 , behaving as a "true" oxygen evolution catalyst. It was concluded that the holes involved in the oxidation of water were firstly transferred to the Co-Pi catalyst, and OER did not take place on the BiVO 4 surface. A systematic analysis of identical photoelectrodes with different spectroscopic and electrochemical tools would help to reconcile both interpretations.

BiVO 4 /Co-Fe Prussian Blue derivatives
As an attractive alternative to metal oxides-based water oxidation catalysts like NiOOH and FeOOH or to Co-Pi, which cracks upon drying of the electrode, the Prussian Blue derivatives (metal hexacyanometallate structures) have demonstrated exceptional activity and stability in neutral an acid media [45] and the possibility to be easily prepared and processed by soft chemistry methods, both as nanoparticles and thin film, from Earth abundant materials. Their unique electronic and structural Adv. Mater. Interfaces 2019, 6, 1900299 Figure 2. Proposed band diagrams for acting mechanisms in Co-Pi-modified BiVO 4 photoanodes: a) Co-Pi acts as suppressor of recombination loses at the BiVO 4 /solution interface, meanwhile water oxidation is primarily driven by direct BiVO 4 holes. Adapted with permission. [42] Copyright 2017, The Royal Society of Chemistry. b) Co-Pi acts as "true catalyst", where water oxidation takes place primarily at the Co-Pi sites. Adapted with permission. [44] Copyright 2019, American Chemical Society.

www.advmatinterfaces.de
features, [46][47][48] as well as their versatile redox properties, [49] have made possible their use as catalyst for oxidation of H 2 O 2 [50] and organic compounds. [51] The electrocatalytic activity for water oxidation of Prussian Blue analogues was firstly investigated by Galán-Mascarós and co-workers in 2013 with the cobalt iron analog [46] (cobalt hexacyanoferrate; from now CoFe-PB). Since then, several authors have also reported effective water oxidation catalysis with Prussian Blue coordinated polymers. [45,47,52] A recent report described a heterostructured BiVO 4 /CoFe-PB photoanode with a 10-fold enhancement of the photocurrent with respect to bare BiVO 4 , a shift of the onset potential of 0.8 V versus RHE (Figure 3a) and excellent stability through ≈55 h chronoamperometric test. [53] As a BiVO 4 overlayer, CoFe-PB outperforms Co-Pi and FeOOH. Mechanistic studies with impedance spectroscopy and linear voltammetry in the presence of a sacrificial hole scavenger, showed that the origin of the functional improvement was related to the more efficient hole transfer to the solution via CoFe-PB layer, indicating that CoFe-PB could effectively act as a "true" catalyst. This was also supported by hybrid density functional theory (DFT) calculations, which predicted the existence of a strong energetic offset (thermodynamic driving force) for hole transfer from the valence band of BiVO 4 to the Co states of CoFe-PB and then to water (Figure 3b) In good agreement with these findings, a more detailed mechanistic investigation with TAS demonstrated that the holes from BiVO 4 were quickly and efficiently transferred to CoFe-PB, leading to persistent oxidized CoFe-PB states. [54] TAS showed the appearance of a long-lived oxidized CoFe-PB + on a sub-microsecond (µs) time-scale, even under very low applied bias. This indicated that photogenerated holes in BiVO 4 rapidly transferred to CoFe-PB. This observation was also supported by PIAS, which emulates in operando water oxidation conditions. In this case, the appearance of CoFe-PB + was also observed, and also a doubly oxidized CoFe-PB 2+ species, upon further oxidation of CoFe-PB + . On the other hand, the typical BiVO 4 hole signal vanished, and consequently the CoFe-PB 2+ state was assumed to be the catalytically active species for water oxidation. Consequently, the BiVO 4 /CoFe-PB heterostructure led to fast (µs) interfacial hole transfer, with concomitant suppression of electron-hole recombination at BiVO 4 and, consequently to enhanced water oxidation performance (Figure 3c).

Other Examples
Some other relevant overlayers on BiVO 4 include TiO 2 (or defective TiO x ). On discontinuous BiVO 4 layers, coating the FTO/BiVO 4 structures with amorphous TiO 2 leads to the effective blocking of surface recombination and to solution-mediated recombination at surface defects and at exposed regions of the conductive substrate. [55] Similar mechanistic enhancement has been suggested for the heterostructured WO 3 /BiVO 4 system. [56] Furthermore, TiO 2 [57] and TiO 2 /Ni [58] dual layers have been reported to enable water oxidation with BiVO 4 in basic media providing effective protection against photocorrosion. Another interesting example involves mixed CeO 2 overlayers (Fe 0. 26   bands. c) Pathways of charge transfer thought water oxidation, elucidated by time-resolved absorption spectroscopies investigation. a,b) Adapted with permission. [53] Copyright 2017, American Chemical Society. c) Adapted with permission. [54] Copyright 2019, American Chemical Society.

www.advmatinterfaces.de
from BiVO 4 while deactivating recombination. [59] On the other hand, amorphous Co−La mixed double hydroxides (MDH) on BiVO 4 /FTO have showed to reduce charge recombination and enhance transport by controlling the grain size. [60] In 2 O 3 has also been used as overlayer on BiVO 4 , enhancing both separation and injection efficiencies. [61] Finally, ultrathin Al 2 O 3 coatings also showed enhanced water oxidation kinetics, which was not due to improved reaction kinetics, but rather, inhibited bimolecular recombination and increased the yield of long-lived holes for water oxidation. [62]

Conclusions
We have showed that interfacing metal oxide semiconductors like BiVO 4 with different underlayers and overlayers provides a convenient materials design platform to exploit synergistic interaction between different constituents. The functional enhancements for solar-driven water oxidation observed upon heterostructuring BiVO 4 can be ascribed to different processes like suppression of bulk and surface recombination, passivation of defect states, or boosting the catalytic activity. Consequently, the detailed understanding of operating mechanisms is essential to unravel the limitations of these heterostructured devices. Therefore, an accurate mechanistic description is key to wisely select the best combination of underlayer/overlayers for each material under study. Furthermore, synthetic modifications and post-synthetic treatments combined to the deposition of underlayers and overlayers constitute a novel paradigm to target theoretical efficiencies on metal oxide photoelectrodes.