Martin
Rohloff
abc,
Björn
Anke
a,
Siyuan
Zhang
d,
Ulrich
Gernert
e,
Christina
Scheu
d,
Martin
Lerch
a and
Anna
Fischer
*bc
aInstitut für Chemie, Technische Universität Berlin, Straße des 17. Juni 135, 10623 Berlin, Germany
bInstitut für Anorganische und Analytische Chemie, Albert-Ludwigs-Universität Freiburg, Albertstraße 21, 79104 Freiburg, Germany. E-mail: anna.fischer@ac.uni-freiburg.de
cFMF – Freiburger Material Forschungszentrum, Stefan-Meier-Straße 21, 79104 Freiburg, Germany
dMax-Planck-Institut für Eisenforschung GmbH, Max-Planck-Straße 1, 40237 Düsseldorf, Germany
eZentraleinrichtung Elektronenmikroskopie (ZELMI), Technische Universität Berlin, Straße des 17. Juni 135, 10623 Berlin, Germany
First published on 7th September 2017
The n-type semiconductor bismuth vanadate (BiVO4) is one of the most promising ternary oxide materials for visible light-induced water oxidation, offering a theoretical solar-to-hydrogen efficiency of 9.1%. However, several factors strongly limit its actual efficiency and among these, poor charge transport has been identified as one major limitation to be overcome. Many efforts have been made to improve charge transport and charge separation in BiVO4 including various doping strategies as well as nanostructuring the absorber to the dimension of its charge carrier diffusion length. In this article, we present a new wet chemical synthesis approach to fabricate pristine and Mo-doped BiVO4 thin film photoanodes. Starting from a solution containing metalorganic Bi, V and Mo precursors, homogeneous and reproducible thin films can be fabricated by a simple procedure of dip-coating and subsequent calcination in air. Structural and morphological characterization reveals that the polycrystalline BiVO4 thin films crystallize in the monoclinic scheelite structure in micrometer-sized randomly oriented, porous domains. The small band-gap scheelite structure is maintained upon Mo doping into the BiVO4 lattice, yielding optimized light harvesting and improved charge carrier transport properties. As a result, the photoelectrochemical performance regarding water oxidation of the as-synthesized Mo-doped BiVO4 thin film photoanodes is highly improved with respect to their pristine counterparts. Photocurrent densities of 1.9 mA cm−2 and 4.6 mA cm−2 at 1.23 V vs. RHE under visible light illumination (100 mW cm−2) are measured for unmodified and CoPi-modified Mo-doped BiVO4 photoanodes respectively, both of which are amongst the highest values reported for modified BiVO4 single-layer design photoanodes so far.
Photoelectrochemical (PEC) water splitting using sunlight as the energy source and suitable semiconductors as photocathode and photoanode materials in a tandem system is considered as the Holy Grail in sustainable hydrogen production.4 In the context of developing a high-performance photoanode for the water oxidation half reaction under visible light, a semiconductor with a suitable band gap (1.9 eV < Eg < 3.1 eV) and a suitable position of the valence band (i.e. more positive than the OER redox potential) decorated with a thin layer of water oxidation catalyst is required.5–8
Since Kudo et al. discovered the ability of BiVO4 to oxidize water under visible light,9 it has become one of the most investigated metal oxide photoanode materials for water oxidation because of its manifold beneficial properties, including suitable valence band position for water oxidation, small band gap of 2.4 eV for efficient visible light absorption, long hole diffusion length (i.e. photo-generated charge carriers are separated efficiently within this material), high stability in aqueous media under oxidizing conditions, low toxicity, earth-abundance of its building elements and reasonable price.10
Many synthesis strategies including precipitation reactions,11–16 hydrothermal syntheses,17–23 sol–gel-methods,24–28 and sonochemical methods29–33 have been reported for the preparation of BiVO4 powders for photocatalytic water oxidation. However, for photoelectrocatalytic water oxidation, the construction of BiVO4 photoanodes in the form of electrodes is required. Hence, development of facile and scalable deposition methods of high quality BiVO4 thin films on transparent conductive substrates with tunable structure, morphology and thickness is essential. The thickness of the BiVO4 photoanode should thereby match the charge carrier diffusion length of BiVO4 (Ld ≈ 70–100 nm (ref. 34–36)) to ensure an optimized charge transport in the system.
Mullins et al. showed in 2011 that nanostructured BiVO4 thin films can be synthesized on F:SnO2-coated glass slides (FTO) by reactive ballistic deposition, involving co-evaporation of bismuth and vanadium in an oxygen atmosphere.37 After deposition of a cobalt based water oxidation electrocatalyst, Mullins et al. achieved water oxidation photocurrents of ca. 1 mA cm−2 at 1.23 V vs. RHE under simulated sunlight (AM 1.5 filter, 100 mW cm−2). By applying spray pyrolysis for the direct deposition of BiVO4 thin films and an optimized deposition of CoPi (a cobalt based water oxidation catalyst developed by Kanan and Nocera38), van de Krol et al. showed in 2013 that W-doping allows enhancing the charge transport efficiency in BiVO4 thin film photoanodes, reaching water oxidation photocurrents of 2.3 mA cm−2 at 1.23 V vs. RHE under simulated sunlight (AM 1.5 filter, 100 mW cm−2).39 Introducing the concept of gradient doping in W-doped BiVO4, van de Krol et al. further improved the charge separation in W-doped BiVO4 photoanodes, yielding photocurrents of 4 mA cm−2 at 1.23 V vs. RHE for CoPi functionalized photoanodes under simulated sunlight.40 Nanostructuring of the BiVO4 absorber provides another approach to improve the photoelectrochemical performance. For example, nanostructuring and deposition of a multilayer water oxidation catalyst allowed Choi and Kim to further improve the photocurrent density to 4.5 mA cm−2 at 1.23 V under simulated sunlight.41
Despite the large progress made in the performance of BiVO4 photoanodes over the last few years, the employed synthesis procedures are often technically advanced and their scalability and reproducibility is often an issue. A simple, wet chemical synthesis method for the reproducible synthesis of high-performance BiVO4 thin film photoanodes for visible light induced water oxidation, i.e. with optimized light absorption, charge separation and water oxidation catalysis, is still missing.
In the present work, we present a new and facile synthesis method for high-performance single-layer Mo-doped BiVO4 photoanodes for water oxidation under visible light. Our approach is based on non-aqueous sol–gel chemistry starting from metalorganic bismuth, vanadium and molybdenum precursors and allows the facile and direct deposition of Mo-doped BiVO4 thin films on various substrates via simple dip-coating and subsequent calcination. Structural, morphological and photoelectrochemical characterizations reveal the as-synthesized BiVO4 thin films to crystallize in the photoelectrochemically favorable monoclinic scheelite structure in large, micrometer sized, single crystalline, flake-like domains with optimized charge transport properties. The as-synthesized Mo-doped BiVO4 photoanodes revealed remarkable water oxidation activity in a neutral electrolyte (potassium phosphate buffer at pH 7.3), yielding photocurrents as high as 1.9 mA cm−2 and 4.6 mA cm−2 at 1.23 V vs. RHE (100 mW cm−2; 400–700 nm) prior to and after CoPi functionalization respectively, both values being amongst the highest values reported so far for unmodified and water oxidation catalyst modified metal-doped single-layer BiVO4-based photoanodes.
Bi(OHex)3 | VO(OEt)3 | MoO2(acac)2 | |
---|---|---|---|
0% Mo:BiVO4 | 445 mg/0.70 mmol | 140 mg/123 μl/0.70 mmol | — |
5% Mo:BiVO4 | 445 mg/0.70 mmol | 140 mg/123 μl/0.70 mmol | 11.3 mg/0.036 mmol |
10% Mo:BiVO4 | 445 mg/0.70 mmol | 140 mg/123 μl/0.70 mmol | 22.6 mg/0.072 mmol |
15% Mo:BiVO4 | 445 mg/0.70 mmol | 140 mg/123 μl/0.70 mmol | 33.9 mg/0.104 mmol |
20% Mo:BiVO4 | 445 mg/0.70 mmol | 140 mg/123 μl/0.70 mmol | 45.2 mg/0.138 mmol |
For determination of the faradaic efficiency, the amount of dissolved oxygen in the electrolyte (0.1 M phosphate buffer at pH 7.3) was monitored using a Clark electrode (Strathkelvin Instruments oxygen meter model 782). A potential of 0.8 V vs. RHE was applied under illumination of white light (400–700 nm, 100 mW cm−2, backside illumination) over a time period of 50 minutes. Prior to measurements the electrolyte was purged with N2 for 12 h. The theoretical O2 amount was calculated according to the number of transferred electrons (assuming 100% faradaic efficiency). The faradaic efficiency was calculated by comparing the experimentally observed O2 amount with that obtained by the theoretical results.
For chronoamperometric measurements of the photocurrent transients a potential of 1.23 V vs. RHE was applied and blue light (440 nm) at an intensity of 2 mW cm−2 was used.
Staircase potentio-electrochemical impedance spectroscopy (PEIS, Mott–Schottky) was carried out in 0.5 M KPi buffer at pH 7.3 scanning from anodic to cathodic and using a sinusoidal modulation of 10 mV at frequencies of 500 Hz and 1 kHz. Estimation of the amount of free charge carriers in the thin film and evaluation of the flat band potential (VFB) was performed by Mott–Schottky analysis (vide infra).
Fig. 1 Synthesis scheme: (a) precursor solution, (b) as-deposited amorphous thin film and (c) calcined, crystalline BiVO4 thin film. |
As demonstrated by the group of Vest in the early nineties, metal hexanoates are suitable compounds for solution-based synthesis of metal oxide thin films because they are quantitatively transformed into metal oxides when pyrolized.44 For a bimetallic compound like BiVO4, the exclusive use of separate bismuth and vanadium hexanoates is however not ideal because these metal hexanoates do not tend to react with each other, eventually leading to certain compositional inhomogeneity. According to Lange however, once mixed with metal alkoxides, metal hexanoates react to form so-called “double-alkoxide” precursors with defined cationic ratios between the two metals.45 As such, by using hexanoate and alkoxide precursors of different metals, metalorganic single-source precursors can be formed with a 1:1 stoichiometry as demonstrated for example for PbTiO3.45
Herein, we adapted this synthesis concept to the synthesis of BiVO4, i.e. by mixing bismuth 2-ethyl-hexanoate and vanadium oxytriethoxide to form in situ a bismuth–vanadium double-alkoxide precursor in the starting precursor solution. Regarding the incorporation of molybdenum, molybdenum dioxydiacetylacetonate was chosen to take advantage of the condensation inhibiting properties of the acetylacetonate ligand.46 All compounds were mixed to form a clear, deep red precursor solution after 4 hours of stirring (Fig. 1a). After stirring, the aged solution was used to deposit thin films on various substrates (silicon wafers, FTO) by dip coating. The dip-coating process led to the formation of soapy, amorphous thin films on the substrates (as can be seen in the TEM images and SAED diffraction pattern in Fig. 1b and 2a).
Calcination at 450 °C leads to grain/crystallite growth driven by the change in free energy due to elimination of grain boundaries by either coalescence or dissolution of smaller grains as well as decomposition of the organic residues,45,47 resulting in the formation of BiVO4 thin films comprised of large, porous, single-crystalline 2D-domains (vide infra). Homogeneous crystalline thin films with good substrate coverage and homogeneous film thicknesses (200 to 250 nm) were obtained for both silicon and FTO coated glass as substrates, as demonstrated by SEM top-view, SEM cross-section and HRTEM investigations (see Fig. 6 for FTO substrates and Fig. S1† for silicon substrates). In the case of Mo-doped thin films, similar results were obtained.
As demonstrated by ICP-OES and EDX, the amount of Mo incorporated within the BiVO4 thin films can be adjusted between 0 and 10.5% (at% with respect to Bi) by varying the Mo concentration in the precursor solution. However, one has to note (see Table 2) that the Mo-content incorporated in the thin films is always lower (by ca. 50%) than the actual Mo-concentration in the precursor solution. This discrepancy, especially for higher Mo-concentrations, can be explained by the limited solubility of the Mo-precursor in the precursor solution, as revealed by the increasing turbidity and hence higher absorbance of the precursor solution (see ESI, Fig. S2† for UV/Vis spectra of Mo-containing synthesis solutions). In the following discussion, the nominal Mo-content of the precursor solution is used to label the five samples, as X% Mo:BiVO4 (X = 0, 5, 10, 15 or 20).
Mo% in the synthesis solution | Mo% in the BiVO4 thin films | Mo% in the BiVO4 powders |
---|---|---|
0 | 0 | 0 |
5 | 3 | 4 |
10 | 6 | 9 |
15 | 8.5 | 15 |
20 | 10.5 | 18 |
Fig. 3a and c give an overview of the diffraction patterns of the X% Mo:BiVO4 thin films and powders. Details about the powder diffraction measurements as well as the outcome of the Rietveld refinements can be found in the ESI (S4, Fig. S5, and Table S1†). From these data (see Fig. 3e), one can see that a continuous phase transition from the monoclinic scheelite to the tetragonal scheelite phase takes place with increasing molybdenum concentration. As can be seen in Fig. 3d, the monoclinic scheelite structure forms for the BiVO4, 5% Mo:BiVO4, 10% Mo:BiVO4 and 15% Mo:BiVO4 samples. Only the 20% Mo:BiVO4 sample crystallizes in the tetragonal scheelite structure, as revealed e.g. by the merging of the (200) and (020) reflections of the monoclinic scheelite structure into a single reflection corresponding to the (200) plane of the tetragonal scheelite structure. This molybdenum induced phase transition agrees with previous literature reports and is related to Mo6+ cations with a larger ionic radius (0.41 Å, coordination number 4) substituting V5+ cations (0.35 Å, coordination number 4).50 For a better understanding of the Mo-doped materials, a closer look was given at the possible defects arising from substituting V5+ by Mo6+. In the ESI (S4, and Table S4†), all reasonable defect models are summarized. The observed under-occupation of the bismuth position (see site occupation factor, s.o.f. in S4, Table S2† determined by Rietveld refinement) clearly points to the presence of bismuth vacancies as the main defects in the Mo-doped BiVO4 structure, which accordingly should be written as Bi1−x/3MoxV1−xO4. This is in good agreement with single crystal diffraction results on such phases, as reported for example by Cesari et al.51
Starting from the structural information gained from the powder samples, one would expect the thin films to exhibit the same crystal structure as the corresponding BiVO4 powders given a similar Mo composition. Considering that the Mo contents within the Mo-doped BiVO4 thin films are below 11 at% (see ICP-OES results of the powders and thin film samples summarized in Table 2), all synthesized X% Mo:BiVO4 thin films are expected to crystallize in the monoclinic scheelite structure.
Fig. 5 Illustration of discussed thin film features, highlighted in (a) an SEM top view image and (b) a cross-section image of a 5% Mo:BiVO4 thin film on an Si substrate. |
Top view and cross-sectional images with corresponding EDX spectra of pristine BiVO4 and Mo-doped BiVO4 thin films (deposited onto FTO substrates) are shown in Fig. 6. All films have similar thicknesses ranging from 215 to 270 nm, are porous and homogeneously cover the rough FTO surface. EDX of all X% Mo:BiVO4 samples proves the thin films to consist of Bi, V, O and Mo. Interestingly, the molybdenum content strongly influences the morphology of the thin films, as can be seen from Fig. 6. Both the average domain size and the size of the inner domain material streaks depend on the Mo content. As summarized in Table 3, undoped BiVO4 has an average domain size of 7.2 ± 3.3 μm2 and an average inner domain material streak size of 68 ± 7 nm. With increasing Mo content, the average domain size decreases (down to 1 μm2 for 20% Mo–BiVO4) while the inner domain material streak size increases (up to 139 ± 8 nm for 20% Mo–BiVO4, see Table 3). Interestingly, the inner domain material streak sizes of the 5%, 10% and 15% Mo:BiVO4 samples, which are 80 ± 7 nm, 84 ± 9 nm and 105 ± 9 nm, respectively, match the BiVO4 diffusion length of 70–100 nm.34–36 The specific geometrical surface area, as determined by krypton physisorption, decreases with increasing Mo content from 8.2 cmfilm2 cmsubstrate−2 for the pristine BiVO4 thin films down to 2.9 cmfilm2 cmsubstrate−2 for the 20% Mo:BiVO4 sample, which correlates with the coarsening of the inner domain material streaks. Similar morphology was observed for films deposited onto Si (see S1, and Fig. S1†), suggesting that there is no influence of the substrate on the microstructure and morphology of the BiVO4 films.
Nominal Mo content (%, atomic ratio) | Molybdenum content determined by ICP-OES (at% with respect to Bi) | Average geometrical domain size by SEM (μm2) | Average inner domain streak size (nm) | Average film thickness (nm) | Specific geom. surface area by Kr-adsorption (cmfilm2 cmsubstrate−2) |
---|---|---|---|---|---|
0 | 0 | 7.2 ± 3.3 | 68 ± 7 | 269 ± 5 | 8.2 |
5 | 3 | 6.4 ± 2.4 | 80 ± 7 | 245 ± 12 | 8.0 |
10 | 6 | 2.8 ± 1.7 | 84 ± 9 | 238 ± 10 | 7.0 |
15 | 8.5 | 1.7 ± 1.1 | 105 ± 9 | 223 ± 12 | 5.5 |
20 | 10.5 | <1 | 139 ± 8 | 214 ± 14 | 2.9 |
The thin film crystallinity and orientation was further investigated by selected area electron diffraction (SAED) and electron backscatter diffraction (EBSD). TEM/SAED of a 5% Mo:BiVO4 sample are exemplarily shown in Fig. 7a. SAED analysis performed at three different spots in one thin film domain reveals identical single-crystalline diffraction pattern close to the [10] zone axis. This result shows that the whole domain exhibits one preferred orientation. EBSD analysis shown in Fig. 7b further confirms that each domain has a preferred orientation. Furthermore, the misorientation between domains is random, so that the thin film is overall polycrystalline, in agreement with our GI-XRD result, which shows no preferred orientation/texture in the film.
The 10% Mo-doped BiVO4 thin film sample, which features the best PEC performance for water oxidation (vide infra), was further examined by high resolution scanning transmission electron microscopy (HR-STEM). The cross-sectional TEM lamella was prepared by focused ion beam (FIB) cutting. The interface between the FTO back-contact and the Mo:BiVO4 is shown in Fig. 8a and FFT analysis was performed at several areas across the interface. The FFT pattern of the Mo:BiVO4 crystal (red square) exhibits only one set of reflections corresponding to a d-spacing of 2.9 Å and the (004) plane. In the adjacent area (green square), the (110) plane with a corresponding d-spacing of 3.4 Å of the FTO is shown. At the interface region with a projected width of approximately 5 nm, spatial frequencies of both crystals were found within the FFT. No amorphous interlayer between the FTO substrate and the Mo-doped BiVO4 thin film was found. The crystalline character of the interface can facilitate an efficient interfacial electron transfer by reducing the ohmic resistivity. The distribution of Mo within the 10% Mo:BiVO4 thin film was investigated by STEM-EDX. As shown in Fig. 8b, the STEM-EDX elemental mapping proves the homogeneous distribution of Mo all over the thin film, indicating the homogeneous incorporation of Mo within the BiVO4 lattice and excluding any phase segregation, in agreement with the XRD results. Furthermore, the determined amount of Mo (≈5 at%) is in agreement with the results obtained by ICP-OES (6 at%) for the 10% Mo:BiVO4 sample. From the STEM-EDX mapping one can further see that the Mo-surface concentration (i.e. the Mo-concentration in the first 10 nm of the photoanode surface) is quite similar to the bulk concentration in the film underneath (see the inset, no shadowing). According to recent reports,52 the superficial dopant content in BiVO4 can be significantly different from the dopant content in the bulk, which in some cases can significantly influence the photocatalytic performance regarding water oxidation. According to our STEM/EDX investigations, no increased Mo concentration at the electrode surface could be detected, so that this effect can most likely be excluded in the present study.
Fig. 9a shows j–V curves of undoped and Mo-doped BiVO4 thin films supported on FTO. For the undoped BiVO4 samples, low photocurrents were observed, accounting for 0.2 mA cm−2 at 1.23 V vs. RHE. With increasing Mo content, the photocurrents dramatically increase by a factor of 10, with maximum photocurrent densities of 1.7 mA cm−2 and 1.9 mA cm−2 at 1.23 V vs. RHE being reached for the 5% Mo:BiVO4 and 10% Mo:BiVO4 samples, respectively. Monitoring the amount of dissolved evolved oxygen in the electrolyte using a Clark electrode revealed a faradaic efficiency of 93% in the case of the 10% Mo:BiVO4 photoanode, a result which confirms that the obtained photocurrents can almost completely be assigned to the water oxidation reaction (see ESI, Fig. S9†). This result is in line with the faradaic efficiency determined for other BiVO4 photoanode systems (see Table S4†).
Fig. 9 j–V curves for X% Mo:BiVO4 thin film photoanodes: (a) without a water oxidation catalyst, (b) with a thin deposited CoPi layer. Forward scan of the second cycle is shown. |
To rationalize the photocurrent increase by Mo-doping, frontside/backside illumination tests were performed. In the case of frontside illumination (i.e. illumination through the absorber film), photo-generated electrons generated at the top surface of the absorber film need to travel longer distances towards the FTO back-contact than in the case of backside illumination through the FTO. In the case of a poor electron conducting material, recombination is hence more likely to occur under frontside than under backside illumination, leading to a large discrepancy between frontside and backside currents. As such, frontside/backside illumination tests are commonly used to assess limitations due to insufficient electron/hole transport properties. As summarized by the ratios of photocurrents obtained in frontside and backside illumination in Fig. 10 (see also ESI, Fig. S7† for j–V curves in frontside and backside illumination), the photocurrents obtained for pristine BiVO4 in frontside-illumination are about 5 times lower than the photocurrent yielded in backside-illumination. This clearly indicates that bulk recombination due to poor electron transport through the BiVO4 thin film is a limiting factor in the pristine system. In contrast, the photocurrents under frontside and backside illumination are highly comparable for the Mo-doped samples. This clearly points to an improved electronic conductivity and decreased bulk recombination in the Mo-doped materials, results which will be further corroborated by Mott–Schottky and photocurrent transient analysis (vide infra).
Fig. 10 j(Frontside)/j(backside) for X% Mo:BiVO4 photoanodes (a) without a water oxidation catalyst, (b) with CoPi deposited. |
Further improvement of the PEC performance of the as-synthesized BiVO4 and Mo-doped BiVO4 photoanodes could be achieved by CoPi deposition at the semiconductor–electrolyte interface. j–V curves of BiVO4 and Mo-doped BiVO4 photoanodes with a deposited layer of CoPi are shown in Fig. 9b. As can be seen, CoPi deposition enhances the photocurrents in all cases, yielding a maximum value of 4.6 mA cm−2 at 1.23 V vs. RHE for the 10% Mo:BiVO4 samples. Quantification of the amount of evolved oxygen revealed a faradaic efficiency of 88% in the case of the CoPi-modified 10% Mo:BiVO4 photoanode, a result which confirms that the obtained photocurrents can be almost completely assigned to the water oxidation reaction (see ESI, Fig. S9†).
Table 4 gives a brief overview on well-performing BiVO4-based photoanodes with respect to water oxidation under illumination of 100 mW cm−2 at neutral pH. To make a fair comparison, we compare our single-layered BiVO4 photoanode with comparable single-layered BiVO4-based systems. Despite the comparatively simple synthesis method applied in our work, our 10% Mo:BiVO4/CoPi photoanode easily keeps up with most of the listed photoanodes consisting of a single BiVO4 thin film layer modified with a water oxidation co-catalyst (see Table 4). For a more detailed summary of well-performing BiVO4 photoanode systems the reader is referred to a recent review by Tolod et al.53 Future work will concentrate on the application of our simple synthesis route to more complex heterojunction photoanode architectures like, for instance, those given in Table 4.
Anode/catalyst | j at 1.23 V vs. RHE without a co-catalyst (mA cm−2) | j at 1.23 V vs. RHE with a co-catalyst (mA cm−2) | Buffer/pH | Author/reference | Publication year |
---|---|---|---|---|---|
Single-layer BiVO 4 /co-catalyst photoanodes | |||||
Mo:BiVO4/CoPi | 0.2 | 1.0 | 0.5 M Na2SO4, pH = 7 | Pilli et al.54 | 2011 |
Mo:BiVO4/FeOOH | — | 2.1 | 0.1 M KH2PO4, pH = 6.8 | Chen et al.55 | 2015 |
Mo:BiVO4/RhO2 | n.a. | 2.9 | Natural seawater | Luo et al.56 | 2011 |
Mo:BiVO4/FeOOH | 1.1 | 3.0 | 0.1 M KH2PO4, pH = 7 | Park et al.57 | 2014 |
BiVO4/FeOOH | 0.2 | 2.3 | 0.1 M KH2PO4, pH = 7 | Seabold et al.58 | 2012 |
BiVO4/CoOx/NiO | 1.1 | 3.5 | 0.1 M KH2PO4, pH = 7 | Zhong et al.59 | 2015 |
W:BiVO4/CoPi | 1.1 | 4.0 | 0.1 M KH2PO4, pH = 7.3 | Abdi et al.40 | 2013 |
BiVO4/FeOOH/NiOOH | 1.9 | 4.5 | 0.5 M KH2PO4, pH = 7 | Kim et al.41 | 2014 |
Mo:BiVO 4 /CoPi | 1.9 | 4.6 | 0.1 M KH 2 PO 4 , pH = 7.3 | This work | 2017 |
H2-treated Mo:BiVO4/CoPi | 2.5 | 4.9 | 0.1 M KH2PO4, pH = 7 | Kim et al.60 | 2015 |
H2-treated Mo:BiVO4/CoPi before H2-treatment | 1.5 | 3.0 | 0.1 M KH2PO4, pH = 7 | Kim et al.60 | 2015 |
N:BiVO4/FeOOH | 3.0 | 5.0 | 0.5 KH2PO4, pH = 7 | Kim et al.61 | 2015 |
More complex BiVO 4 -based heterojunction photoanodes | |||||
WO3 helices/BiVO4/FeOOH/NiOOH | 3.8 | 5.4 | 0.5 KH2PO4, pH = 7 | Shi et al.62 | 2014 |
SiOx/Pt/SnO2 nanocones/Mo:BiVO4/Fe(Ni)OOH | 4.1 | 5.8 | 0.5 KH2PO4, pH = 7 | Qiu et al.63 | 2016 |
WO3 nanorods/BiVO4/FeOOH/NiOOH | — | 6.7 | KH2PO4, pH = 7 | Pihosh et al.64 | 2015 |
To gather more information about the origin of the high photoelectrochemical performance, Mott–Schottky-type electrochemical impedance spectroscopy was applied. This method relies on measuring the capacitance of the space charge region CSC at the semiconductor/electrolyte junction. According to the Mott–Schottky-equation (eqn (1)) the flat band potential VFB and the carrier concentration N can be estimated by the linear region of the plot of 1/CSC2 as a function of the applied potential V, as seen in Fig. 11.
Fig. 11 (a) Mott–Schottky plot of Mo-doped BiVO4 thin films, (b) Mott–Schottky plots with undoped BiVO4 sample included, (c) Mott–Schottky-plot of a 10% Mo:BiVO4 sample at two different frequencies. |
Mott–Schottky equation.
(1) |
In eqn (1), εr represents the relative permittivity of the semiconductor, ε0 the permittivity in vacuum, A the surface area of the semiconductor electrode, e the charge of an electron, kb the Boltzmann constant, and T the temperature. Considering the BiVO4/FTO junction structure, it was uncertain whether accurate results would be provided by the Mott–Schottky analysis. The obtained positive slopes of the linear region of the Mott–Schottky plot, typical of n-type semiconductors, and the VFB values obtained (summarized in Table 5) are (i) frequency independent as exemplarily shown for the 10% Mo:BiVO4 sample (Fig. 11c), and (ii) in line with previously reported VFB values for BiVO4 electrodes.36,41 As can be seen in Fig. 11a and b, the slope of the Mott–Schottky plot for undoped BiVO4 is much steeper than for the Mo-doped BiVO4 samples. This result reveals a drastic increase of free charge carriers in the Mo-doped BiVO4 electrodes.
Sample | V FB vs. RHE (V) measured at 500 Hz | V FB vs. RHE (V) measured at 1 kHz |
---|---|---|
BiVO4 | 0.11 | 0.07 |
5% Mo:BiVO4 | 0.17 | 0.16 |
10% Mo:BiVO4 | 0.15 | 0.15 |
15% Mo:BiVO4 | 0.13 | 0.13 |
20% Mo:BiVO4 | 0.11 | 0.12 |
To further clarify the origin of the different photoelectrochemical performances of the X% Mo:BiVO4 photoanodes, chronoamperometric investigations of the photocurrent transients of the pristine BiVO4 and the 10% Mo:BiVO4 samples were performed. The results are shown in Fig. 12a and b. The transients were acquired at an applied potential of 1.23 V vs. RHE using blue light (440 nm) with an intensity of 2 mW cm−2. To analyze the transients a phenomenological approach commonly used in the literature was used.65–68 As can be seen in Fig. 12a and b, the BiVO4 photoanodes respond to light on/light off illumination with a characteristic “spike and overshoot” photocurrent transient. This transient response is very typical of photoanode systems with pronounced surface electron–hole recombination. When the illumination is turned on, holes generated in the space charge region are rapidly displaced towards the semiconductor–electrolyte interface. Due to the sluggish kinetics of the water oxidation reaction, holes initially accumulate at the semiconductor–electrolyte interface. In the steady state, the rate of hole arrival is then balanced by the rates of hole transfer towards the electrolyte and the rate of electron–hole surface recombination. As such, the resulting photocurrent transient is the sum of the hole and electron contributions. The instantaneous photocurrent (jini) measured when the illumination is switched on corresponds to a charging or displacement current due to the initial movement of photo-generated holes towards the surface. By contrast, the steady-state photocurrent (jss) corresponds to the flux of holes that are transferred successfully to the electrolyte without undergoing recombination with electrons at the surface. As a consequence, the ratio between jss and jini is a measure of the hole transfer efficiency ηtransfer (see eqn (2)). Using a simple approach and assuming that both hole transfer and surface recombination follow a pseudo-first order law in terms of surface hole concentration, the hole transfer efficiency can be expressed by the corresponding first order rate constants of hole transfer, ktrans, and hole recombination, krec.
Fig. 12 Photocurrent transients, exemplarily shown for an undoped and a 10% Mo:BiVO4 sample. (a) Undoped BiVO4, (b) 10% Mo:BiVO4, (c) undoped BiVO4 with CoPi deposited, (d) 10% Mo:BiVO4 with CoPi deposited. Blue light (440 nm, 2 mW cm−2) switched on and off at an applied potential of 1.23 V vs. RHE was used to record the transients. Transients for other X% Mo:BiVO4 samples are given in the ESI.† |
Hole transfer efficiency expressed as a function of jss and jini as well as a function of hole transfer and hole surface recombination rate constants.
(2) |
The exponential decay of the initial photocurrent jini towards the steady state photocurrent jss is characterized by the time constants (krec + ktrans)−1. The decay can be analyzed and ktrans and krec can be separated using eqn (2). After switching the light off, a cathodic (photo-)current overshoot is observed. This results from electrons flowing back to the electrode surface to recombine with holes (rapid) and with accumulated holes at surface states (slow), a phenomenon which is called back-electron–hole recombination.69–71 Detailed results of the analysis carried out in this work for all samples are given in the ESI (S8, and Table S5†) and the essential outcome is described in the following.
As can be seen in Fig. 12a, the 0% Mo:BiVO4 sample shows a typical “spike and overshoot” behavior during light-on/off experiments, accounting for an initial photocurrent jini of 0.055 mA cm−2 which is rapidly decreasing towards the steady state photocurrent jss accounting for 0.002 mA cm−2. This response is typical for a system with pronounced surface electron–hole recombination. Consequently, the hole transfer efficiency of the pristine BiVO4 sample was determined to be very low accounting for 0.03 (see ESI, Table S5†).
In the case of the Mo-doped BiVO4 samples the situation is different (Fig. 12b). Only a moderate photocurrent spike is observed for the Mo-doped samples while any overshoot is suppressed. This result is in good agreement with previous reports about charge carrier dynamics in undoped and doped BiVO4 photoanodes.71,72 The absence of any overshoot can be explained by two effects related to n-type doping in BiVO4: (a) decrease of the space charge width due to increased donor concentration reducing back electron–hole recombination with surface accumulated holes and (b) removal of surface trapping states upon doping, therefore reducing the number of surface accumulated holes which are prone to back electron–hole recombination when the light is switched off. Consequently, Mo-doping also substantially increases the hole transfer efficiencies ηtransfer from 0.45 for the 5% Mo:BiVO4 sample up to 0.53 for the 20% Mo:BiVO4 sample (see Table S5†).
Deposition of a layer of CoPi onto the undoped BiVO4 anode increases the steady state photocurrent by a factor of 4, as a result of increased water oxidation kinetics (Fig. 12c). However, the CoPi deposition does not affect the spike and overshoot behavior positively, indicating that the undoped CoPi-modified BiVO4 sample still suffers from a high degree of electron–hole recombination, which most probably occurs because of its limited electron transport.35
For all Mo:BiVO4 samples the observed photocurrents are largely increased after CoPi deposition indicating a drastic reduction of surface recombination due to CoPi-enhanced heterogeneous electron transfer rates at the semiconductor–electrolyte interface. In line, the recombination rates krec become very small compared to the hole transfer rates ktrans for the 5% Mo:BiVO4 sample and the 10% Mo:BiVO4 sample (factor of 6 and 7.5, respectively). A maximum hole transfer efficiency ηtransfer of 0.88% was yielded for the 10% Mo:BiVO4 sample, which is in good agreement with the obtained j–V-curves and photocurrent densities, respectively (see Fig. 9 and ESI, S8†).
As such, in line with other literature reports,35 the main limiting factor for the PEC performance of the undoped BiVO4 photoanodes was figured out to be the poor electron transport properties through the BiVO4 bulk. According to our results, this problem can be overcome by Mo-doping, which increases the charge carrier density in the material. Depositing a layer of CoPi enhances water oxidation kinetics and preeminently reduces possible surface recombination.
Fig. 13 Morphological features (a) and photoelectrochemical properties without (b) and with (c) CoPi of the X% Mo:BiVO4 thin film photoanodes illustrated as a function of Mo content. |
The applied bias photon to current efficiency (ABPE) curves for the X% Mo:BiVO4 samples calculated from the j–V curves are shown in Fig. 14. ABPE corresponds to the solar-to-hydrogen efficiency (STH) if an external bias is needed to enable photoelectrolysis, i.e. the electrical power output is corrected by the electrical power input provided by the external bias according to eqn (3):
Applied bias to photon to current efficiency.
(3) |
As can be seen in Fig. 14, the 10% Mo:BiVO4 yields the best ABPE without and with CoPi deposited accounting for 0.22% and 1.09%, respectively. Aiming at higher efficiencies, the 10% Mo:BiVO4 thin film photoanodes in principle can easily be coupled to a suitable photocathode or be biased externally by a coupled solar cell. However, in consideration of a reasonable target STH value for light-driven water splitting of 16%, as for instance proposed by Laurence Peter,73 there is still much room for improvement.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c7se00301c |
This journal is © The Royal Society of Chemistry 2017 |