Report

Nanoporous BiVO4 Photoanodes with Dual-Layer Oxygen Evolution Catalysts for Solar Water Splitting

See allHide authors and affiliations

Science  28 Feb 2014:
Vol. 343, Issue 6174, pp. 990-994
DOI: 10.1126/science.1246913

A Boost for Bismuth Vanadate

In theory, given its light-absorption spectrum, bismuth vanadate should be an effective photoanode for solar water-splitting. However, in prior studies, few of the “holes” generated upon photoexcitation have persisted long enough to strip electrons from water. Kim and Choi (p. 990, published online 13 February) now show that the use of a hydrophobic vanadium source in the semiconductor's synthesis results in a high-surface-area morphology with substantially enhanced hole lifetimes. Deposition of two successive catalyst layers enhanced the proportion of holes that reacted with water at the surface, thereby raising the efficiency of the oxygen evolution reaction.

Abstract

Bismuth vanadate (BiVO4) has a band structure that is well-suited for potential use as a photoanode in solar water splitting, but it suffers from poor electron-hole separation. Here, we demonstrate that a nanoporous morphology (specific surface area of 31.8 square meters per gram) effectively suppresses bulk carrier recombination without additional doping, manifesting an electron-hole separation yield of 0.90 at 1.23 volts (V) versus the reversible hydrogen electrode (RHE). We enhanced the propensity for surface-reaching holes to instigate water-splitting chemistry by serially applying two different oxygen evolution catalyst (OEC) layers, FeOOH and NiOOH, which reduces interface recombination at the BiVO4/OEC junction while creating a more favorable Helmholtz layer potential drop at the OEC/electrolyte junction. The resulting BiVO4/FeOOH/NiOOH photoanode achieves a photocurrent density of 2.73 milliamps per square centimenter at a potential as low as 0.6 V versus RHE.

N-type bismuth vanadate (BiVO4) has recently emerged as a promising photoanode for use in water-splitting photoelectrochemical cells because it absorbs a substantial portion of the visible spectrum (bandgap energy, ~2.4 eV) and has a favorable conduction band (CB) edge position very near the thermodynamic H2 evolution potential (1, 2). However, the solar-to-hydrogen (STH) conversion efficiency achieved with BiVO4 to date has been far below what is expected because the material suffers from poor electron-hole separation yield (ϕsep) (26). Previous efforts to improve the ϕsep of BiVO4 mainly focused on doping studies, which were intended to improve its poor electron transport properties (2, 612).

Here, we demonstrate that a high-surface-area, nanoporous BiVO4 electrode composed of particles smaller than its hole diffusion length can effectively increase ϕsep without additional doping. Furthermore, we investigated the effect of an oxygen evolution catalyst (OEC) layer on the interfacial recombination at the BiVO4/OEC junction, water oxidation kinetics, and the Helmholtz layer potential drop at the OEC/electrolyte junction using two different OECs, FeOOH and NiOOH. Our understanding of the BiVO4/OEC/electrolyte junction resulted in the development of a new strategy to serially apply dual layers of OECs that can optimize both the BiVO4/OEC and the OEC/electrolyte junctions simultaneously, enabling efficient utilization of surface-reaching holes for solar water oxidation.

Nanoporous BiVO4 electrodes were prepared by first electrochemically depositing BiOI electrodes and then applying a dimethyl sulfoxide (DMSO) solution of vanadyl acetylacetonate [VO(acac)2] onto their surface and heating in air at 450°C for 2 hours (experimental deatails are available in the supplementary materials, materials and methods). A schematic overview of the synthesis procedure is shown in fig. S1. The specific advantage of using BiOI is that its two-dimensional (2D) crystal structure enables electrodeposition of extremely thin plates (~20 nm) with sufficient voids between them (Fig. 1A). These voids inhibit grain growth of BiVO4 during the conversion process, resulting in nanoporous BiVO4 electrodes.

Fig. 1 Morphologies of nanoporous BiVO4 electrodes.

(A) SEM image of BiOI. (B and C) Top-view and side-view SEM images of BiVO4 electrode prepared using NH4OH/V2O5. (D to F) Top-view and side-view SEM images of nanoporous BiVO4 prepared using DMSO/VO(acac)2.

In a previous attempt to prepare nanoporous BiVO4 (13), we used an NH4OH solution of V2O5 as the vanadium source, which could not easily wet the BiOI surface because air in the voids between the BiOI plates renders the surface highly hydrophobic (fig. S2). Thus, the distribution of V2O5 was uneven, few BiVO4 nucleation processes were induced within a single 2D BiOI sheet, and the resulting electrodes manifested limited porosity (Fig. 1, B and C). The use of comparatively hydrophobic VO(acac)2/DMSO solution overcame this problem and resulted in a remarkable increase in surface area. The top-view and side-view scanning electron microscopy (SEM) images show the formation of much smaller BiVO4 nanoparticles (mean particle size = 76 ± 5 nm) (fig. S3) creating a 3D nanoporous network (Fig. 1, D to F).

N2 adsorption-desorption-isotherm measurements show that the nanoporous BiVO4 electrode contains micropores within BiVO4 particles as well as mesopores and macropores between BiVO4 nanoparticles (fig. S4 and table S1) (14). The specific surface area of the nanoporous BiVO4 electrode was estimated to be 31.8 ± 2.3 m2/g based on a fitting analysis using the Brunauer-Emmett-Teller equation (14). BiVO4 electrodes prepared by using other synthesis methods (such as metal organic decomposition, spray deposition, or direct electrodeposition of BiVO4) possess limited surface areas, and no attempts to measure surface areas of these samples were reported (2).

The purity and crystal structure of the nanoporous BiVO4 electrode (monoclinic scheelite structure) were confirmed with x-ray diffraction (fig. S5). The bandgap of the nanoporous BiVO4 electrode was estimated to be ~2.50 to 2.55 eV, using ultraviolet-visible absorption spectra (fig. S6), which is slightly larger than the bandgap of BiVO4 samples prepared by other methods that result in larger grain composition (~2.4 eV) (13, 15).

The photoelectrochemical properties were first examined in the presence of 1 M sodium sulfite (Na2SO3), which served as the hole scavenger. The oxidation of sulfite is thermodynamically and kinetically more facile than oxidation of water (11, 13, 1518), and therefore, measuring photocurrent for sulfite oxidation enables investigation of the photoelectrochemical properties of BiVO4 independently of its poor water oxidation kinetics. A typical photocurrent-potential (J-V) curve of the sulfite oxidation with nanoporous BiVO4 is shown in Fig. 2A. A very early photocurrent onset potential, 0.1 V versus reversible hydrogen electrode (RHE), and a rapid increase in photocurrent in the 0.2 V < E versus RHE < 0.6 V region, representing an excellent fill factor, resulted in a photocurrent density of 3.3 ± 0.3 mA/cm2 at a potential as low as 0.6 V versus RHE. The incident photon-to-current conversion efficiency (IPCE) and the absorbed photon-to-current conversion efficiency (APCE) of the nanoporous BiVO4 at 0.6 V versus RHE are 60 and 72%, respectively, at 420 nm (Fig. 2B).

Fig. 2 Photoelectrochemical properties of nanoporous BiVO4 electrode for sulfite oxidation.

(A) J-V curve of nanoporous BiVO4 electrode measured in a 0.5 M phosphate buffer (pH 7) containing 1 M Na2SO3 as hole scavenger under AM 1.5 G, 100 mW/cm2 illumination (scan rate, 10 mV/s). Dark current is shown as a dashed line. (Inset) ϕsep calculated from the J-V curve after dark current is subtracted. (B) IPCE (red circles) and APCE (blue triangles) measured in the same solution at 0.6 V versus RHE.

Photocurrent density obtained for sulfite oxidation was used to calculate ϕsep by using Eq. 1, where JPEC is the measured photocurrent density and Jabs is the photon absorption rate expressed as current density, which is calculated assuming 100% APCE (calculation details are available in the supplementary materials, materials and methods) (4, 6, 1921). Jabs of the nanoporous BiVO4 electrode was calculated to be 4.45 mA/cm2. ϕsep is the yield of the photogenerated holes that reach the surface, and ϕox is the yield of the surface reaching holes that are injected into the solution species (21). JPEC = Jabs × ϕsep × ϕox (1)For sulfite oxidation with extremely fast oxidation kinetics, surface recombination is negligible, and ϕox is ~1. Therefore, ϕsep is obtained by dividing JPEC by Jabs (Fig. 2A, inset). The result shows that the nanoporous BiVO4 electrode achieves ϕsep = 0.70 ± 0.03 and 0.90 ± 0.03 at 0.6 V and 1.23 V versus RHE, respectively, which is remarkable because a typical ϕsep value for a BiVO4 photoanode is below 0.3 at 1.23 V versus RHE (4, 6). The highest ϕsep achieved recently by Abdi and coworkers using gradient doping was ~0.6 at 1.23 V versus RHE (12). The hole diffusion length of BiVO4 was recently reported to be ~100 nm when using single-crystal BiVO4 (22). The mean particle size of BiVO4 composing the nanoporous BiVO4 electrode shown in Fig. 1D is 76 ± 5 nm (fig. S3), and the particle size obtained from the XRD peaks (fig. S5) when using the Scherrer equation is 27 ± 2 nm. Therefore, the nanoporosity incorporated into BiVO4 electrodes in this study appears to be ideal for effectively suppressing bulk carrier recombination, resulting in a record high ϕsep.

Photocurrent from the nanoporous BiVO4 for water oxidation shown in Fig. 3A (black line) is considerably lower than the photocurrent for sulfite oxidation (Fig. 2A), indicating that the majority of the surface-reaching holes were lost to surface recombination because of the poor catalytic nature of the BiVO4 surface for water oxidation (2). To improve water oxidation kinetics, we photodeposited a thin FeOOH or NiOOH layer on the nanoporous BiVO4 surface as an OEC layer in order to assemble BiVO4/FeOOH and BiVO4/NiOOH electrodes. Their thicknesses were optimized so as to maximize photocurrent generation (fig. S7). It has been previously demonstrated that FeOOH interfaces well with BiVO4 (13, 15), whereas NiOOH is known to be a more active OEC than FeOOH (less overpotential required to achieve the same current density) as a dark electrocatalyst on a conducting substrate (fig. S8) (2325).

Fig. 3 Effect of OECs on photocurrents for water oxidation and sulfite oxidation.

(A) J-V curves of BiVO4 (black solid line), BiVO4/FeOOH (blue), BiVO4/NiOOH (green), BiVO4/FeOOH/NiOOH (red), and BiVO4/NiOOH/FeOOH (pink) for water oxidation measured in a 0.5 M phosphate buffer (pH 7) under AM 1.5 G illumination. Dark current is shown as a dashed line. (B to E) J-V curves of (B) BiVO4/FeOOH, (C) BiVO4/NiOOH, (D) BiVO4/FeOOH/NiOOH, and (E) BiVO4/NiOOH/FeOOH comparing photocurrent for sulfite oxidation (dashed) and water oxidation (solid) measured with and without the presence of 1.0 M Na2SO3 as hole scavenger. Photocurrent for sulfite oxidation by BiVO4 is shown as the black dashed line for comparison. The mean values and SDs of photocurrent onset potentials and photocurrent densities are summarized in tables S2 and S3.

The photocurrents for water oxidation from the resulting BiVO4/FeOOH and BiVO4/NiOOH photoanodes were markedly higher than those from the bare BiVO4 electrode (Fig. 3A and table S2), but their photocurrents were still lower than that generated for sulfite oxidation at the bare BiVO4 electrode. This comparison suggested that neither BiVO4/FeOOH nor BiVO4/NiOOH engages all surface-reaching holes in the oxygen evolution reaction, instead losing a portion to surface recombination at the BiVO4/OEC junction. The interface states formed at the BiVO4/OEC junction can serve as recombination centers and cause surface recombination. The fact that BiVO4/FeOOH generated higher photocurrent than did BiVO4/NiOOH, although NiOOH shows faster water oxidation kinetics as an electrocatalyst, suggests that the interface recombination at the BiVO4/NiOOH junction is more substantial than that at the BiVO4/FeOOH junction. This can be easily confirmed by comparing photocurrents for sulfite oxidation by BiVO4, BiVO4/FeOOH, and BiVO4/NiOOH (Fig. 3, B and C, and table S3). Because the interfacial hole transfer rates for sulfite oxidation on the BiVO4, FeOOH, and NiOOH surfaces should be equally fast, any difference observed in photocurrents for sulfite oxidation by BiVO4, BiVO4/FeOOH, and BiVO4/NiOOH should be mainly due to the recombination at the BiVO4/OEC junction. The comparison shows that photocurrent for BiVO4/FeOOH is very close to that for BiVO4, whereas the photocurrent for BiVO4/NiOOH is considerably lower, which indicates that the interface recombination at the BiVO4/NiOOH junction is indeed more substantial.

In addition to the interface recombination at the BiVO4/OEC junction, slow water oxidation kinetics at the OEC/solution junction can cause additional surface recombination during water oxidation (26, 27). This additional surface recombination can be shown as the difference in photocurrent for sulfite oxidation and water oxidation (Fig. 3, B and C). When the rate of interfacial hole transfer for water oxidation is slower than the rate of holes entering the OEC layer, holes are accumulated in the OEC layer and at the BiVO4/OEC junction, which in turn increases the electron current from the CB of BiVO4 to the OEC layer for surface recombination (27). Because FeOOH has slower water oxidation kinetics than that of NiOOH, the difference in photocurrent for water oxidation and sulfite oxidation is more pronounced for BiVO4/FeOOH than BiVO4/NiOOH (Fig. 3, B and C). However, when the effects of interface recombination at the BiVO4/OEC junction and water oxidation kinetics are combined, BiVO4/NiOOH loses more surface-reaching holes to surface recombination and generates lower photocurrent than does BiVO4/FeOOH for water oxidation (Fig. 3A).

Regardless of more substantial surface recombination, BiVO4/NiOOH shows an earlier photocurrent onset and generates higher photocurrent than does BiVO4/FeOOH in the low bias region (E < 0.44 V versus RHE) for water oxidation (Fig. 3A and table S2). This means that BiVO4/NiOOH has a more negative flatband potential (EFB) than that of BiVO4/FeOOH. The photocurrent onset potential for sulfite oxidation with fast oxidation kinetics should be very close to EFB. The results show that BiVO4 has the most negative onset potential (0.11 ± 0.02 V versus RHE), followed by BiVO4/NiOOH (0.12 ± 0.02 V versus RHE), and then BiVO4/FeOOH (0.18 ± 0.02 V versus RHE) (table S3 and fig. S9A). The EFBs obtained by Mott-Schottky plots of BiVO4, BiVO4/FeOOH, and BiVO4/NiOOH show the same trend; BiVO4 has the most negative EFB (0.10 ± 0.03 V versus RHE), followed by BiVO4/NiOOH (0.11 ± 0.03 V versus RHE), and then BiVO4/FeOOH (0.15 ± 0.02 V versus RHE) (table S4 and fig. S10).

It is unlikely that the shift in EFB is due to the change in charge carrier density of BiVO4 because the addition of an extremely thin OEC layer should not affect the doping level or carrier density within the BiVO4 electrode. This assumption is also supported by the comparable slopes in the Mott-Schottky plots for these three electrodes at each frequency (table S4). Then, the difference in EFB should be caused by the change in the Helmholtz layer potential drop (VH), which is the only other factor that can affect the EFB, as shown in Eq. 2, where ϕSC is the work function of the semiconductor versus vacuum, and 4.5 is the scale factor relating the H+/H2 redox level to vacuum (28). At the BiVO4/electrolyte junction, the dominant charges that affect the solid side of the Helmholtz double layer come from the adsorption of H+ and OH ions on the BiVO4 surface, which depends on the solution pH and the point of zero ζ potential (pHPZZP) of BiVO4 (Eq. 3) (28, 29). EFB (NHE) = ϕSC + VH – 4.5 (2)VH = 0.059 (pHPZZP – pH) (3)The pHPZZP of BiVO4 is reported to be between 2.5 and 3.5, and therefore, VH should be negative in a pH 7 solution (30, 31). However, when a thin layer of FeOOH or NiOOH is deposited on BiVO4, because the Helmholtz double layer is now formed at the OEC/solution junction the VH at the solid/solution interface is no longer determined by the pHPZZP of BiVO4 but by the pHPZZP of the OEC layer. This means that the pHPZZP of OEC is an important factor to consider in optimizing the photoanode/OEC junction because it can affect the EFB of the photoanode.

The pHPZZP of FeOOH is reported to be between 7 and 9 (32, 33). Thus, the resulting more positive VH at the FeOOH/solution junction will shift the EFB of BiVO4/FeOOH positively, which is in agreement with the observed shift direction of the EFB. The pHPZZP of NiOOH cannot be straightforwardly determined because chemical composition of NiOOH varies with pH. However, it is known that NiOOH has a negative ζ potential (~–20 mV) in a pH 7 solution (34), meaning that pHPZZP of NiOOH is lower than 7, and the EFB of BiVO4/NiOOH is expected to be more negative than that of BiVO4/FeOOH, which again agrees with the observed EFB shift.

The pHPZZP of a material depends on its specific surface termination and solution composition. Therefore, the most reliable estimations of the VHs of electrodes discussed in this study can be obtained when the ζ potential measurement is performed by using the electrodes and the solution that were used in this study. The ζ potentials measured for our nanoporous BiVO4, BiVO4/FeOOH, and BiVO4/NiOOH in a 0.5 M phosphate buffer (pH 7) were –36 ± 3, –8 ± 3, and –36 ± 4 mV, respectively. These results indicate that the VHs at the BiVO4/solution and the NiOOH/solution junctions should indeed be more negative than that at the FeOOH/solution junction.

On the basis of our new understanding of the BiVO4/OEC and the OEC/electrolyte interfaces, we deposited consecutive layers of FeOOH and NiOOH, simultaneously optimizing the BiVO4/OEC and the OEC/electrolyte junctions. The FeOOH at the BiVO4/OEC junction will reduce the interface recombination, whereas the NiOOH at the OEC/electrolyte junction will decrease VH to achieve a more negative EFB for BiVO4 while realizing faster water oxidation kinetics than if FeOOH was used as the outermost layer.

The photocurrent onset for sulfite oxidation (fig. S9A and table S3) as well as the Mott-Schottky plot (fig. S10 and table S4) of the resulting BiVO4/FeOOH/NiOOH photoanode shows that its EFB is comparable with that of BiVO4/NiOOH, indicating that the EFB of the BiVO4 photoanode is indeed affected by the pHPZZP of the outermost OEC layer. In addition, BiVO4/FeOOH/NiOOH and BiVO4/FeOOH show comparable J-V curves for sulfite oxidation, confirming that the BiVO4/FeOOH junction effectively reduces the interface recombination at the BiVO4/OEC junction (Fig. 3, B and D). As a result, BiVO4/FeOOH/NiOOH shows impressive overall performance for water oxidation, reaching a photocurrent density of 2.8 ± 0.2 mA/cm2 at 0.6 V versus RHE (Fig. 3A and table S2), which is markedly better than those of BiVO4/FeOOH and BiVO4/NiOOH and is almost comparable with the performance of bare BiVO4 for sulfite oxidation.

When NiOOH was first deposited on the BiVO4 surface and FeOOH was added as the outermost layer to form BiVO4/NiOOH/FeOOH (reversed OEC junction), the resulting EFBs determined by sulfite photocurrent onset (fig. S9A and table S3) and Mott-Schottky plot (fig. S10 and table S4) are comparable with those of BiVO4/FeOOH, again confirming that the EFB of the BiVO4 photoanode is affected by the pHPZZP of the outermost OEC. Also, the J-V curve for sulfite oxidation by BiVO4/NiOOH/FeOOH was comparable with that by BiVO4/NiOOH, confirming that a BiVO4/NiOOH junction is not favorable for interface recombination (Fig. 3, C and E). As a result, BiVO4/NiOOH/FeOOH shows the lowest photocurrent for water oxidation. These results prove that the photocurrent enhancement achieved by the BiVO4/FeOOH/NiOOH photoanode for photoelectrolysis of water is truly due to the simultaneous optimization of the BiVO4/OEC and OEC/electrolyte junctions, using an optimum dual OEC structure.

The applied bias photon-to-current efficiency (ABPE) of the BiVO4/FeOOH/NiOOH electrode calculated by using its J-V curve, assuming 100% Faradaic efficiency, is plotted in Fig. 4A (35). The maximum ABPE of 1.75% achieved by the system is impressive because it is obtained by using unmodified BiVO4 as a single photon absorber. Moreover, this efficiency is achieved at a potential as low as 0.6 V versus RHE, which is a highly favorable feature for assembling a tandem cell or a photoelectrochemical diode (12, 36, 37). The ABPE obtained by using a two-electrode system (working electrode and a Pt counter electrode), which achieves the maximum ABPE of 1.72%, is also shown in fig. S11 (35). The long-term stability of BiVO4/FeOOH/NiOOH was tested by obtaining a J-t curve. A photocurrent density of 2.73 mA/cm2, obtained by applying 0.6 V between the working and counter electrodes, was maintained for 48 hours without showing any sign of decay, proving its long-term stability (Fig. 4B). The O2 measurement made by using a fluorescence O2 sensor confirmed that the photocurrent generated at 0.6 V versus counter-electrode was mainly associated with O2 production (> 90% photocurrent-to-O2 conversion efficiency) (Fig. 4C). The same results were obtained when the measurement was performed at 0.6 V versus RHE. H2 production at the Pt counter electrode was also detected with gas chromatography (GC) (Fig. 4C). The molar ratio of the produced H2:O2 was 1.85:1. The slight deviation from the stoichiometric ratio of 2:1 is due to our imperfect manual sampling method of H2 for GC analysis.

Fig. 4 Photoelectrolysis of water by BiVO4/FeOOH/NiOOH photoanode.

(A) ABPE obtained using a three-electrode system. (B) J-t curve measured at 0.6 V versus counter electrode in a phosphate buffer (pH 7) under AM 1.5 G illumination. (C) Detection of H2 and O2 at 0.6 V versus counter electrode.

Because this outstanding performance was achieved by using simple, unmodified BiVO4 (no extrinsic doping and no composition tuning) as the only photon absorber, further improvement of the cell efficiency is expected when various strategies of tuning compositions or forming heterojunctions and tandem cells are used to enhance photon absorption and electron-hole separation.

Supplementary Materials

www.sciencemag.org/content/343/6174/990/suppl/DC1

Materials and Methods

Figs. S1 to S11

Tables S1 to S4

References (3841)

References and Notes

  1. Acknowledgments: We acknowledge support from the Center for Chemical Innovation of the National Science Foundation (POWERING THE PLANET: grant CHE-1305124). We thank M. A. Woo for electrodepositing FeOOH and NiOOH on FTO substrates and testing their electrochemical water oxidation performances.
View Abstract

Stay Connected to Science

Navigate This Article