Enhancing photoelectrochemical water splitting by combining work function tuning and heterojunction engineering

We herein demonstrate the unusual effectiveness of two strategies in combination to enhance photoelectrochemical water splitting. First, the work function adjustment via molybdenum (Mo) doping significantly reduces the interfacial energy loss and increases the open-circuit photovoltage of bismuth vanadate (BiVO4) photoelectrochemical cells. Second, the creation and optimization of the heterojunction of boron (B) doping carbon nitride (C3N4) and Mo doping BiVO4 to enforce directional charge transfer, accomplished by work function adjustment via B doping for C3N4, substantially boost the charge separation of photo-generated electron-hole pairs at the B-C3N4 and Mo-BiVO4 interface. The synergy between the above efforts have significantly reduced the onset potential, and enhanced charge separation and optical properties of the BiVO4-based photoanode, culminating in achieving a record applied bias photon-to-current efficiency of 2.67% at 0.54 V vs. the reversible hydrogen electrode. This work sheds light on designing and fabricating the semiconductor structures for the next-generation photoelectrodes.

P hotoelectrochemical cell (PEC) for water splitting is a key technology of the future for hydrogen production [1][2][3] . Despite the widespread attention that has been received, this technology still has many hurdles to overcome and uncharted territories to explore. Ultimately, the photon to hydrogen conversion efficiency has yet to be increased to such a level that commercial applications could become viable 4 .
In the PEC water splitting process, photons are first absorbed by the photoelectrode producing electrons and holes, which are then separated and participated in the hydrogen evolution reaction (HER) on cathode and the oxygen evolution reaction (OER) on anode, possibly with the assistance of a bias voltage 1 . Therefore, one way to enhance the PEC efficiency is to increase the quantum efficiency of photons in a PEC system by improving the efficiencies of light harvesting, charge separation and surface charge transfer [3][4][5][6][7][8][9][10] . Another strategy to enhance the PEC efficiency is to minimize the overpotential by reducing the voltage loss related to charge recombination, sluggish surface kinetics, etc. BiVO 4 has received great attention in recent years because it is a promising sustainability-inspired photoanode material for PEC with a suitable band gap for visible light absorption 11 and favorable conduction band edge position (0.1-0.2 V vs. NHE) for H 2 evolution 7 . However, the PEC efficiency at low bias voltages of BiVO 4 photoanode still has much room for improvement [11][12][13] . In particular, due to the presence of numerous trap states and surface defects as well as the associated surface Fermi-level pinning effect, the BiVO 4 films variously prepared so far are still plagued by the quite low open-circuit photo-voltage when used as photoanodes 11,14,15 . To address this issue, doped photoanodes, such as W-BiVO 4 , Mo-BiVO 4 , have been fabricated aiming to enhance charge transport and to reduce the charge recombination 11,12,16,17 . Meanwhile, W-BiVO 4 /BiVO 4 , Co 2 O 3 / BiVO 4 , and BiOI/BiVO 4 photoanodes have been developed in the form of so-called homojunctions and heterojunctions to enhance the charge separation in PEC systems 6,13,18 . Other problems of the BiVO 4 -based photoanode include the still low coverage of the solar spectrum which it is able to harvest as well as the low charge separation efficiency. To address these problems, carbon quantum dots/BiVO 4 and nitrogen doped BiVO 4 photoanodes have been reported showing broadened light absorption range, enhanced light harvesting efficiency, and boosted interfacial charge transfer for PEC water splitting 9,18 . As for improving the utilization efficiency of surface charge for oxygen evolution, the combined catalyst/photoelectrode systems, such as FeOOH/ BiVO 4 , NiFeO x /BiVO 4 , Co-Pi/BiVO 4 , and NiOOH/FeOOH/ BiVO 4 , have been commonly used 7,[19][20][21] .
Recent efforts have improved the photocurrent density of BiVO 4 based photoanodes for water splitting to nearly 90% of its theoretical value at 1.23 V vs. RHE 22,23 . However, the photon to hydrogen conversion efficiency is still far from its theoretical value mainly due to the stagnant carrier transport. Especially when a PEC cell works at low bias, the carrier transport is more susceptible to blockage by any potential barriers in the energy landscape along the carrier passage. Specifically, poor performance at low bias of BiVO 4 based photoanodes led to poor applied bias photon-to-current efficiencies (ABPEs) as reported in some PEC water splitting systems, such as Bi-NiFeO x /BiVO 4 (2.25%) 20 , NiOOH/FeOOH/BiVO 4 (1.75%) 7 , NiO/CoO x /BiVO 4 (1.5%) 11 , NiOOH/FeOOH/N-BiVO 4 (2.2%) 9 . Previously, nanostructures and cocatalysts have been used to promote photocurrents and to minimize onset potentials, respectively 7,19-21 .
Meanwhile, because C 3 N 4 has a favorable conduction band edge position relative to that of BiVO 4 , a heterojunction between the two could increase the charge separation. Prompted by this expectation, the conjugation of C 3 N 4 with BiVO 4 has received great attention in recent years [24][25][26][27][28] . However, before the C 3 N 4 / BiVO 4 junction could efficiently drive the PEC water splitting, new strategies must be developed to elaborate the band structure at the junction to optimize charge separation by minimizing interfacial kinetic barriers and energy losses.
In this work, we endeavored to explore such ways to further improve the PEC performance of the BiVO 4 photoanode. First, we systematically studied the effect of Mo doping on the electron band structure of BiVO 4 , and discovered that a moderate Mo doping of BiVO 4 , a low end doping regime that has not been explored before, can increase the photo-voltage of BiVO 4 photoanodes from 0.24 V to~1 V in 10 s irradiation. Second, to further improve the charge separation efficiency at low bias, we elaborated a cliff like junction between B-C 3 N 4 and Mo-BiVO 4 , for which the band structure of C 3 N 4 was judiciously tuned as well by B doping. With such an elaborated junction, interfacial charge transfer was remarkably enhanced. As the main thread running through this work, we make special efforts to advance our ability to modulate the work functions with a view to toning up the NiFeO x /B-C 3 N 4 /Mo-BiVO 4 photoanodes for PEC water splitting. We have significantly increased the light harvesting efficiency (LHE) of the B-C 3 N 4 /Mo-BiVO 4 photoanode, achieving photocurrent densities of 4.7 mA cm −2 at 0.6 V vs. RHE (Φ Sep = 79%) and 6 mA cm −2 at 1.23 V vs. RHE (Φ Sep = 98%) in potassium phosphate buffer (PPB) solution with 0.5 M Na 2 SO 3 hole scavenger (pH 7). When the NiFeO x was anchored on the B-C 3 N 4 /Mo-BiVO 4 photoanode as an OER catalyst layer forming the NiFeO x /B-C 3 N 4 /Mo-BiVO 4 photoanode, we obtained photocurrent densities of 3.85 mA cm −2 at 0.54 V vs. RHE (71% IPCE) and 5.93 mA cm −2 at 1.23 V vs. RHE (92% IPCE) in PPB solution without any hole scavengers such as Na 2 SO 3 (pH 7). Significantly, the NiFeO x /B-C 3 N 4 /Mo-BiVO 4 photoanode has achieved an ABPE up to 2.67% at 0.54 V vs. RHE, which is the highest reported to date and yet, with the lowest biased-voltage, for BiVO 4 -based PEC devices.
Work function tuning. The first strategy we used to optimize the PEC performance of BiVO 4 -based photoanodes, more precisely, to lower the onset potential, was to systematically adjust the work function by Mo doping below 1%, which is a previously uncharted doping regime (the atomic ratio is shown in Supplementary Table 1). Figure  RHE. From our extensive measurements, the photocurrent densities assuming 100% absorbed photon-to-current efficiency (J abs ) of the BiVO 4 -based and the Mo-BiVO 4 -based photoanodes were consistently~4.7 mA cm −2 and~5.01 mA cm −2 , respectively ( Supplementary Fig. 3). The increase in J abs could be ascribed to the enhanced LHE, carrier concentration ( Supplementary Fig. 27) and mobility ( Supplementary Fig. 28) resulting from the Mo doping [33][34][35][36][37] . More interestingly, the onset potential of Mo doped BiVO 4 (0.05, 0.1, and 0.5%) photoanodes became significantly more negative than that of pure BiVO 4 photoanode, and the rapid photocurrent increase region against bias for the 0.05% and 0.1% Mo doped BiVO 4 is also more negative than that of the BiVO 4 photoanode. However, the rapid photocurrent increase region against bias for the 0.5% Mo doped BiVO 4 is more positive than that of BiVO 4 . This phenomenon is caused by the changed opencircuit photo-voltage (OCP) of BiVO 4 due to the work function adjustment by the Mo doping.
Supplementary Fig. 5 shows OCP changes of the Mo doped BiVO 4 (0.05, 0.1, and 0.5%) relative to the pure BiVO 4 photoanode due to the truncation of Fermi-level pinning and surface trap states of BiVO 4 . OCP is essentially the difference between open-circuit voltage in dark (OCV dark ) and light (OCV light ). The work function is tuned in such a way that the Fermi level (E f ) is far from CBM when light is off resulting in a small OCV dark , but when light is on (AM 1.5 G), the quasi-Fermilevel (E fn ) becomes as close as possible to CBM giving rise to a large OCV light . Consequently, a large OCP (V ph ) can be obtained. Shown in Supplementary Fig. 5 are the OCP timing profiles of the Mo doped BiVO 4 (0.05, 0.1, and 0.5%) and the pure BiVO 4 photoanodes in PPB solution with the Na 2 SO 3 hole scavenger (pH 7) over a testing interval of 30 s ( Supplementary Fig. 5a) and 2000 s ( Supplementary Fig. 5b). The most important observation is that the 0.1% Mo-BiVO 4 photoanode achieved the highest OCP (~1 V in the first irradiation on/off cycle, and~0.55 V in the cycles after testing for 1600 s). These OCP values are much higher than those of pure-BiVO 4 (~0.35 V and~0.15 V), 0.05% Mo-BiVO 4 (~0.8 V and~0.4 V), and 0.5% Mo-BiVO 4 photoanode (~0.23 V and~0.16 V) when tested under otherwise the same conditions. A higher OCP value means a more favorable driving force for water oxidation since it determines the difference between the hole quasi-Fermi-level of the semiconductor heterojunction and the redox potential of the electrolyte.
The above presented OCP result can be captured by the picture illustrated in Fig. 2b. For the pure BiVO 4 , the Fermi-level position  [38][39][40] . When the pure BiVO 4 photoanode was immersed in the solution, the E f of pure BiVO 4 became more negative than the redox potential due to Fermi-level pinning by surface trapped electrons, making the OCV dark of BiVO 4 relatively high 11 . Meanwhile, under AM 1.5 G illumination, the Fermi-level pinning effect prevented the E fn moving very close to the conduction band minimum (CBM) of BiVO 4 11,38 , leading to a moderate OCV light and thus a low OCP (V ph ) of pure BiVO 4 very low. Importantly, the Mo doping in BiVO 4 could reduce the surface trap states and at the same time introduced new states, thus moderating the Fermi-level pinning effect ( Supplementary Fig. 4) 41 . Due to the reduced Fermi-level pinning effect by Mo doping, for the 0.05% and 0.1% Mo-BiVO 4 photoanodes, the E f and E fn became more positive and negative than E f of the pure BiVO 4 photoanode, respectively, thereby enhancing the OCPs. However, for the 0.5% Mo-BiVO 4 , new states were introduced due to the excess Mo doping, and the E f became much closer to CBM, leading to high OCV dark ( Supplementary Fig. 6) and thus a low OCP. As such, the OCP of 0.1% Mo-BiVO 4 is the best of all the samples we studied, in agreement with the corresponding PEC performance as will be presented below. As can be seen from Fig. 2a, the photocurrent density of 0.1% Mo-BiVO 4 reached 5.0 (±0.2) mA cm −2 at 1.23 V vs. RHE in solution with hole scavenger, which represents~73% the theoretical water oxidation photocurrent density (J max ) of BiVO 4 (6.8 mA cm −2 ). Thus the moderate Mo doping of BiVO 4 in the low end doping regime can increase the photo-voltage carrier concentration and mobility of the BiVO 4 photoanodes, and improve their onset potential and photocurrent density.
Heterojunction engineering. We now turn to our second strategy to optimize the PEC performance of BiVO 4 -based photoanodes by further increasing the utilization of J max . To accomplish it, we started with the basic C 3 N 4 /BiVO 4 junction, and then work up for optimization by B-doping C 3 N 4 and the Mo-doping BiVO 4 . Both LHE and charge separation have been enhanced, leading to the increase of J abs and Φ Sep of the photoanode. Supplementary Fig. 7 shows the UV-vis absorption spectra from the diffuse reflectance measurements and photographs of B-C 3 N 4 and C 3 N 4 . From visual inspection, the yellow color of B-C 3 N 4 is clearly deeper than C 3 N 4 , and correspondingly, the absorption of B-C 3 N 4 is also stronger than C 3 N 4 . From the plots in Supplementary Fig. 8a of (αhν) 2 vs. the photon energy (hν), the band-gap energy of C 3 N 4 and B-C 3 N 4 are 2.53 eV and 2.41 eV, respectively. Supplementary Fig. 8b is the LHE of B-C 3 N 4 /Mo-BiVO 4 , which exhibits stronger absorption in the range between 300 nm and 500 nm than Mo-BiVO 4 . Supplementary  Fig. 8c shows the spectra of the solar irradiance of AM 1.  4 . This result shows that when C 3 N 4 is compositing the Mo-BiVO 4 , it has no effect on the separation of the photo-generated charges. These results mean that although the thermodynamic potential of pure C 3 N 4 and Mo-BiVO 4 are match, the heterojunction of pure C 3 N 4 and Mo-BiVO 4 became a compound center of photo-generated charge.
In detail, when Mo-BiVO 4 and C 3 N 4 come into contact to form a heterojunction, the bands on the two sides bend oppositely into the spike-like structure (Fig. 3e), and thus the electrons from the C 3 N 4 side can hardly transfer to the Mo-BiVO 4 side. The holes inside Mo-BiVO 4 will hardly transfer to C 3 N 4 , as the existence of energy barrier at the interface. Therefore, the contact interface of pure C 3 N 4 and Mo-BiVO 4 will reduce the separation of photogenerated charges. On the other hand, due to the B element doping, the Fermi level of B-C 3 N 4 is getting closer to the VBM ( Supplementary Fig. 9b). As shown in Fig. 3d, when the Mo-BiVO 4 and B-C 3 N 4 are in contact, a cliff like junction is formed with the correct charge transfer direction, which will increase the separation of photo-generated charges. Supplementary Fig. 15     , which shows that the photocurrent density of samples measured in solution with Na 2 SO 3 will decline after the NiFeO x layer deposition ( Supplementary Fig. 20). Figure 4c shows the half-cell applied bias photo-to-current efficiency (ABPE) of the NiFeO x /B-C 3 N 4 /Mo-BiVO 4 photoanode. The ABPE is calculated to be 2.67% at 0.54 V vs. RHE, which is the highest recorded for BiVO 4 -based photoanodes (Fig. 6c). Evidently, the highest efficiency has been achieved for the NiFeO These results confirm that the NiFeO x is a stable co-catalyst for the BiVO 4 -based photoanodes, essentially consistent with the recent reports. The generation rates of H 2 and O 2 by our half-cell system are measured to be 77.5 μM/h and 336 μm 33.6 μM/h, respectively, with Faradic efficiency of 98% (Fig. 5f).

Discussion
In conclusion, the NiFeO x /B-C 3 N 4 /Mo-BiVO 4 photoanode has provided an archetype to exploit the potential of boosting the photoelectrochemical performance by the synergistic combination of work function tuning and heterojunction construction. The bespoke photoanode achieved a remarkable photocurrent density of 3.85 mA cm −2 , ABPE of 2.67% and IPCE of 71% at 0.54 V vs. RHE, which are the highest yet reported with the lowest biased-voltage for BiVO 4 -based PEC materials. The NiFeO x /B-C 3 N/Mo-BiVO 4 photoanode exhibited significantly enhanced PEC activity for water splitting by systematically work function adjustment (Fig. 6a). We have demonstrated the work function adjustment via Mo doping could reduce the interfacial energy loss and increase the opencircuit photo-voltage of BiVO 4 PEC cells. In addition, the creation and optimization of the heterojunction (p-n) of B-C 3 N 4 and Mo-BiVO 4 with correct charge transfer direction were accomplished by work function adjustment via B doping for C 3 N 4 , thereby increasing the separation of photo-generated electron-hole pairs at the B-C 3 N 4 and Mo-BiVO 4 interface (Fig. 6a, b).  4,7,9,11,13,18,20,[43][44][45] . The demonstration of the NiFeO x / B-C 3 N 4 /Mo-BiVO 4 photoanode with excellent PEC water splitting capability achieved by the synergistic combination of work function tuning and heterojunction deliberation will inform the design and development of the next-generation PEC materials and devices.

Methods
Preparation of BiVO 4 and Mo-BiVO 4 electrode. BiVO 4 photoanodes were fabricated by a two-step process via a modified method which was originally developed by Kim and Choi 7 . At first, a template-free electrochemical deposition was applied to prepare the BiOI nanosheets using a conventional three-electrode glass cell, where a piece of F-doped SnO 2 coated glass (FTO, Nippon Sheet Glass, 1 × 2 cm) served as the working electrode, a Pt electrode served as the counter electrode and an Ag/AgCl electrode served as the reference electrode. Generally, 50 mL of solution containing 0. 4   The corresponding concentration ratio of Mo/Bi was 0.05%, 0.1%, and 0.5%, respectively.
Preparation of C 3 N 4 and B-C 3 N 4 . The bulk graphite-C 3 N 4 (C 3 N 4 ) was fabricated by directly heating low-cost melamine (99%, Aladdin). In detail, 5 g melamine powder was placed in an alumina crucible with a cover, then heated to 500°C for 2 h in a muffle furnace with a heating rate of 2°C min −1 . The obtained bulk C 3 N 4 was grind into small powder, and 100 mg C 3 N 4 powder was dispersed in 100 mL isopropyl alcohol (AR, Tianjin Damao Reagent) and exfoliated by ultrasonication for 24 h to obtain C 3 N 4 nanosheets (C 3 N 4 -NS). The resultant dispersion was centrifuged at 3000 rpm for 10 min, and the supernatant containing exfoliated C 3 N 4 -NS was collected by pipette. Boron doped C 3 N 4 nanosheets (B-C 3 N 4 -NS) supernatants were prepared in the same way but heating the mixture of 0.5 g boric acid (GR, Aladdin) and 5 g melamine. Analysis. The as-synthesized products were characterized by a scanning electron microscope (SEM, Zeiss G-500), transmission electron microscopy (TEM, JEOL 2100 F, FEI Tecnai G 2 F30), X-Ray Diffractometer (XRD, D8 ADVANCE), X-ray Photoelectron Spectroscopy (XPS) and Ultroviolet Photoelectron Spectroscopy (UPS, Thermo Fisher Scientific ESCALab250) and Raman spectroscopy (Renishaw inVia). The optical properties of the products were measured with an UV-vis-NIR Spectrophotometer (UV-vis-NIR, Shimadzu UV-2450).
Photoelectrochemical and electrochemical measurements. All the PEC and electrochemical measurements were carried out in a three-electrode cell with a flat quartz window to facilitate illumination of the photoelectrode surface. The working electrode is the product fabricated in this work, while Pt electrode was used as a counter electrode and Ag/AgCl electrode was used as a reference electrode, respectively. The illumination source was AM 1.5 G solar simulator (Newport, LCS 100 94011 A (class A, Supplementary Fig. 29) directed at the quartz PEC cell (100 mW cm −2 ). Incident-photon-to-current conversion efficiency (IPCE) were collected by a Solartron 1280B electrochemical station with a solar simulator (Newport 69920, 1000 W xenon lamp), coupled with an infrared water filter (Oriel 6127) and aligned monochromator (Oriel Corner-  7), c specific photocurrent density at 1.23 V vs. RHE and applied bias photo-tocurrent efficiency (ABPE) of BiVO 4 based photoanode 4,7,9,11,13,18,20,[41][42][43] .