Photoelectrochemical Decomposition of Water into ... - ACS Publications

Science UniVersity of Tokyo, Yamazaki 2641, Noda, Chiba 278-8514, Japan, and Nissan Research Center,. Nissan Motor Co., Ltd., 1 Natsushima, Yokosuka, ...
0 downloads 0 Views 432KB Size
11352

J. Phys. Chem. B 2006, 110, 11352-11360

Photoelectrochemical Decomposition of Water into H2 and O2 on Porous BiVO4 Thin-Film Electrodes under Visible Light and Significant Effect of Ag Ion Treatment Kazuhiro Sayama,*,† Atsushi Nomura,‡ Takeo Arai,† Tsuyoshi Sugita,‡ Ryu Abe,† Masatoshi Yanagida,† Takashi Oi,§ Yasukazu Iwasaki,§ Yoshimoto Abe,‡ and Hideki Sugihara† Energy Technology Research Institute, National Institute of AdVanced Industrial Science and Technology (AIST), Central 5, 1-1-1 Higashi, Tsukuba, Ibaraki 305-8565, Japan, Faculty of Science and Technology, Science UniVersity of Tokyo, Yamazaki 2641, Noda, Chiba 278-8514, Japan, and Nissan Research Center, Nissan Motor Co., Ltd., 1 Natsushima, Yokosuka, Kanagawa 237-8523, Japan ReceiVed: December 28, 2005; In Final Form: April 1, 2006

The photoelectrochemical properties of porous BiVO4 thin-film electrodes on conducting glass for H2 production from water under visible light were investigated. BiVO4 films were prepared by the metal-organic decomposition method, and particles were 90-150 nm in diameter. Under visible-light irradiation, H2 and O2 evolved in a stoichiometric ratio (H2/O2 ) 2) from an aqueous solution of Na2SO4 with an external bias. The photocurrent increased with addition of methanol. The band structure of BiVO4 was investigated by opencircuit potential, flat-band potential, X-ray photoelectron spectroscopy, and calculations based on density functional theory. The top of the valence-band potential of BiVO4 was shifted negatively compared to the potentials of the conventional oxide semiconductors without Bi. We surmise that hybridization between the O-2p and Bi-6s orbitals might contribute to the negative shift of the BiVO4 valence band. Treatment with an aqueous solution of AgNO3 improved the photocurrent of the BiVO4 electrode significantly. The maximum incident photon-to-current conversion efficiency at 420 nm was 44%. This value was the highest among mixed-oxide semiconductor electrodes under visible light irradiation. AgNO3 treatment also improved the stability of the photocurrent. The Ag+ ion in/on the BiVO4 catalyzed the intrinsic photogeneration of oxygen with the holes.

1. Introduction Since the publication of the Honda-Fujishima effect on TiO2 electrodes under UV light,1 photoelectrochemical water splitting with semiconductor photoelectrodes has been regarded as an ideal means of converting solar energy directly into clean fuel, namely, hydrogen energy.2-22 The energy density of solar light is relatively low; therefore, many featuresssuch as high energyconversion efficiency, high stability, low-cost production, and simple structure for large-area utilizationsare required for semiconductor photoelectrodes. To increase the solar energy conversion efficiency, effective utilization of visible light over a wide wavelength region is indispensable. It should be noted that oxide semiconductors have many advantages, for example, they are generally stable, are suitable for O2 evolution, and are produced easily compared to non-oxide semiconductors. However, the efficiencies of various oxide semiconductor electrodes with conventional types, such as single crystal or sintered pellet, are low in the visible-light region, even when a high external bias was applied. It was recently reported that a nanocrystalline WO3 thin film on a conducting glass electrode showed an excellent incident photon-to-current conversion efficiency (IPCE > 75%) for water decomposition under visible light.5,6 The nanocrystalline WO3 film was prepared by spreading a colloidal solution of WO3 on * Address correspondence to this author. Phone: +81-(0)298-61-6234. Fax: +81-(0)298-61-4760. E-mail: [email protected]. † National Institute of Advanced Industrial Science and Technology (AIST). ‡ Science University of Tokyo. § Nissan Motor Co., Ltd.

the conducting glass and then firing it. Moreover, relatively high IPCEs (ca. 25%) were also reported for porous Fe2O3 thin-film electrodes on conducting glass prepared by spraying the glass with a solution containing iron salts.10,15,16 It is worth mentioning that high IPCEs were obtained with this simple wet process, although many defects in the porous film electrodes may have been present. To effectively absorb visible light from solar light, utilization of semiconductors having a narrower band gap is essential. In studies of multijunction hybrid photoelectrodes, which were designed to decompose water directly with no external bias,20-22 development of effective oxide semiconductors for O2 evolution has been important. However, oxide semiconductor materials that have efficient O2 evolution ability and a band gap that is narrower than that of WO3 are limited. Some porous thin-film electrodes of various oxide semiconductorsssuch as metal-iondoped oxides and mixed oxides on conducting glassshave been investigated;7-10,13-17 however, their IPCEs are not satisfactory. We have studied new oxide semiconductors for their watersplitting ability under visible light. In the field of photocatalysis, BiVO4 semiconductor powder was reported to show high O2 evolution when Ag+ or Fe3+ ions were used as electron acceptors under visible light, but this semiconductor had no H2 evolution ability.19,23,24 The band gap of BiVO4 (2.4 eV) was smaller than that of WO3 (2.7 eV). It is speculated that the top of the valence band consists of Bi-6s and O-2p orbitals;23 therefore, such physical properties as mobility of charge and band potentials of BiVO4 are probably different from those of simple oxide semiconductors. In an earlier work, we reported

10.1021/jp057539+ CCC: $33.50 © 2006 American Chemical Society Published on Web 05/19/2006

Photoelectrochemical Decomposition of Water

J. Phys. Chem. B, Vol. 110, No. 23, 2006 11353

briefly that a nanocrystalline BiVO4 film electrode on conducting glass showed an excellent IPCE for the decomposition of water.12 In this study, the physical and photoelectrochemical properties of BiVO4 films, the effect of sacrificial reagent addition, and the band structure of BiVO4 were investigated in detail. Moreover, surface modifications of the semiconductor photoelectrodes,18,25 as well as semiconductor photocatalysts, were important for improving efficiency and stability. We report herein that surface modification of a BiVO4 electrode by treatment with an aqueous solution of AgNO3 substantially improved both efficiency and stability under visible-light irradiation. 2. Experimental Section 2.1. Preparation of BiVO4 Electrode. The BiVO4 films were prepared by a modified metal-organic decomposition (MOD) method.12,26 Bi(NO3)3‚5H2O (Kanto Chemical Co.) in acetic acid (0.2 mol/L) and vanadium(IV) (oxy)acetylacetonate (Azmax Co.) in acetylacetone (0.03 mol/L) were mixed with a 1:1 stoichiometric ratio of Bi to V. The solution was coated by means of a spin coater (Mikasa Co., 1H-D7, 1000 rpm, 10 s) on a conducting glass (FTO glass, Nippon Sheet Glass Co., F-doped SnO2, 10 Ω/sq) and fired at 500 °C in air for 30 min. The spin coatings and the calcinations were repeated 6 times. The average thickness of the rough film, as measured by a step meter (Tencor, Alpha step), was ca. 0.5 µm for 6 layers. In the case of the surface modifications, the BiVO4/FTO electrode was immersed in various aqueous solutions (0.01 mol/ L) in the dark for 12 h and was then thoroughly washed with distilled water. 2.2. Photoelectrochemical Measurement and Characterization. The photoelectrochemical and capacitance measurements were conducted with a potentiostat (BAS, 612A) and a Pyrex glass cell. A Pt wire and a Ag/AgCl electrode (0.21 V vs NHE, pH 0) were used as the counter and reference electrodes, respectively. For stability testing (the photocurrent-time measurement), the photocurrent was measured by a two-electrode system without a reference electrode. An aqueous solution of 0.5 mol/L of Na2SO4 (pH 5.8) was used most often as the electrolyte solution because the stability of the BiVO4 film in Na2SO4 (aq sol) was good compared to that in other solutions. It should be noted, however, that because the Na2SO4 solution was unbuffered, the applied potential was probably overestimated owing to the pH change around the electrode surface during the photoreaction. The light source was a Xe lamp (500 W, Ushio Denki Co.) with/without UV cutoff (HOYA, L-42) or band-pass filters. The unfiltered light intensity was measured by a spectroradiometer (Yamashita Denso Co., YSR-1100M) between 350 and 500 nm, corresponding to the active wavelength of BiVO4, to be 450 W/m2. The light intensity of AM1.5G (Japan Industrial Standard, JIS-C8911) between 350 and 500 nm was 175 W/m2; therefore, the unfiltered Xe light intensity was nearly equivalent to a condition of 2.6 Sun. The semiconductor electrodes were irradiated through the FTO conducting glass. The masked-off irradiated area was 0.28 cm2. The light-harvesting efficiency (LHE) of the electrode was calculated from the transmittance (T) and reflectance (R) with use of an integrating sphere (Jasco, V-570):

LHE ) 1 - R - T

(1)

The IPCE was then calculated from the following equation:

IPCE )

1240 × photocurrent density [µA/cm2] wavelength [nm] × photon flux [W/m2]

(2)

The monochromatic photon flux through the band-pass filter (Nihon Shinku Kogaku Co.) was measured by an optical photodiode power meter (Advantest Co., TQ8210), and it was corrected for the reflection loss at the Pyrex glass cell and the FTO glass. The electrodes were characterized by X-ray diffractometry (XRD, Mac Science, MX-Labo), using Cu KR radiation at 40 mA, 40 kV, and by Raman spectroscopy (Jasco, NRS1000, 532 nm). Scanning electron microscopy (SEM) was performed on a Hitachi S-800 instrument. The composition of the semiconductors was measured with an X-ray fluorescence analyzer (XRF, Rigaku Co., ZSX-mini, 40 kV, 1.2 mA). In the case of X-ray photoelectron spectroscopy (XPS) (Ulvac-Phi, XPS-1800), an Al anode with a monochromator was used to significantly reduce the background signal. To completely prevent the charge-up effect, many techniques were used. Both a low-energy electron ( Bi2O3. Therefore, we believe that Bi-containing semiconduc-

11356 J. Phys. Chem. B, Vol. 110, No. 23, 2006

Sayama et al.

Figure 10. Mott-Schottky (1/C2 versus applied potential, where C is the capacitance) plot of the BiVO4 electrode in the dark at 600 Hz. Figure 7. Time course of H2 and O2 evolution under visible-light irradiation (>420 nm). The applied bias was 1.6 V to the Pt counter electrode.

Figure 8. Density of state (DOS) of monoclinic sheelite-type BiVO4 calculated by the DFT method. Note that the x axis is inverted compared to the Figure 9 XPS and that “zero energy” refers to the occupied level.

Figure 9. XPS spectra of various semiconductors on FTO conducting glass. The binding energy was referenced to the C-1s at 284.8 eV: (a) BiVO4, (b) Bi2O3, (c) V2O5, (d) WO3, and (e) TiO2.

tors easily discharged the electron and that the top of Evb of BiVO4 and Bi2O3 was shifted negatively toward those of conventional Bi-free oxide semiconductors. The flat-band potential (Efb) of semiconductor electrodes is usually measured by Mott-Schottky analysis.10,17 Figure 10 shows the Mott-Schottky plot of the BiVO4 electrode. The Efb of the BiVO4 was -0.6 ( 0.1 V (vs Ag/AgCl). Values of Efb for the WO3 and TiO2 electrodes in the same electrolyte solution were ca. -0.6 and -0.8 V, respectively. Efb is strongly related

to the bottom of the Ecb and is considered to be located just under the Ecb for n-type semiconductors. According to the small band gap of BiVO4, the top of Evb of BiVO4 was calculated to be more negative than those of TiO2 and WO3. Absolute values of Ecb and Evb of semiconductors are very important but are controversial in each method mentioned above. The comparison method for Voc was simple; however, Voc was sensitive to light intensity, electrolyte solution, and quality of film preparation. In the case of XPS measurement, complete neutralization for the charge-up effect and precise correction of the binding energy are essential. The Efb from the MottSchottky analysis was affected by the Helmholtz capacitance for dielectric semiconductors.32 It is considered that the dielectric property of BiVO4 might cause the upward deviation of the plots below -0.5 V in Figure 10.32,33 However, it should be emphasized that all results from these three methodss comparison of Voc, XPS measurement, and Mott-Schottky analysissindicated that the top of Evb of BiVO4 was shifted to less positive potentials in comparison with the values for the conventional semiconductors without Bi. The negative shift of the valence band is beneficial because it decreases the band gap of oxide semiconductors without the positive shift of the conduction band, leading to the decrease of applied bias. According to the DFT calculation in Figure 8, the Bi-6s orbital was mainly located -10 eV apart from the valence band made mainly of O-2p. However, we also found that the top of the valence band contained a slight amount of the Bi-6s component (O-2p, 97%; Bi-6s, 2%; V-3d, 1%) and the Bi-6s peak at -10 eV also contained an O-2p component (O-2p, 15%; Bi-6s, 84%; V-3d, 1%), suggesting hybridization between the O-2p and Bi6s orbitals.34 This hybridization may have contributed to the upper shift of the BiVO4 valence band. Similar behavior about the shift of valence band was also suggested in the Bi-6s, the Pb-6s, or Ag-4d hybrid orbitals of PbBi2Nb2O9 and (AgBi)0.5MO4 (M ) W and Mo).35,36 We suggest that the transfer of holes on the valence band becomes easy with the formation of a hybrid Bi-6s-O-2p orbital. We also investigated other MVO4 (M ) Y, In) electrodes; however, high efficiencies were not obtained. These results suggest that the Bi orbital probably played an important role in the high efficiency of the BiVO4 electrode. 3.3. Stability of the BiVO4 Electrode and the Ag Treatment Effect. Figure 11a shows the time course of the photocurrent for the BiVO4 electrode. The photocurrent decreased gradually, but it remained constant at ca. 0.5 mA/cm2 for more than 10 h. The turnover number of observed photoelectrons to the BiVO4 unit was ca. 60 for 1 h. No change in the BiVO4 electrode after a 10-h photoreaction was observed in the XRD, XRF, and UV-vis spectra, suggesting that the bulk of BiVO4 was intrinsically stable for the photoelectrochemical reaction.

Photoelectrochemical Decomposition of Water

J. Phys. Chem. B, Vol. 110, No. 23, 2006 11357

Figure 13. Dependence of IPCE on wavelength in Na2SO4 (aq sol): (a) no AgNO3 treatment and (b) after AgNO3 treatment. The applied potential was +1.0 V versus Ag/AgCl.

TABLE 1: Dependence of Treatment Time of AgNO3 Aqueous Solution on Photocurrent and Stabilitya Figure 11. Time course of the photocurrent of BiVO4 electrodes: (a) no AgNO3 treatment and (b) after AgNO3 treatment. The applied bias was 1.2 V to the Pt counter electrode.

treatment no treatment AgNO3 treatment in the dark AgNO3 treatment with light

treatment time

initial photocurrent (mA/cm2)

photocurrent after 30 min (mA/cm2)

1 min 1h 3h 12 h 1 min

1.2 1.9 2.2 2.1 2.0 2.1

0.8 0.9 2.3 2.2 2.1 1.3

a Two-electrode system; applied bias, 1.2 V; AgNO3 concentration, 0.01 mol/L.

Figure 12. Potential-current curve of the BiVO4 electrodes. The applied bias was 1.2 V to the Pt counter electrode. Unfiltered whole light from a Xe lamp was used for irradiation. (Upper panel) In Na2SO4 aqueous solution: (a) no AgNO3 treatment and (b) after AgNO3 treatment. (Lower panel) In Na2SO4 aqueous solution with methanol addition (10 vol %): (c) no AgNO3 treatment and (d) after AgNO3 treatment.

The photocurrent was restored to 70% of the initial efficiency by keeping it in the dark for 1 day or by cyclic voltammetry. It is speculated that slight changes, for example, in the valence state in BiVO4 or in the surface structure, may influence the photocurrent stability. In the study of photocatalysts, surface modifications and loading of cocatalyst have often been investigated for the promotion of the charge separation and the formation of active

reaction site. On the other hand, reports on surface modifications of photoelectrodes are limited.14,15,17,18,25 Ohno et al. reported an improvement in the IPCE of a TiO2 single-crystal photoelectrode by adsorption of Fe3+ ion.25 We investigated surface modifications of the BiVO4 electrode by aqueous solutions (0.01 mol/L, 12 h) containing various metal nitrates and chlorides, such as AgNO3, Pd(NO3)2, Cr(NO3)3, Cu(NO3)2, Ni(NO3)2, Fe(NO3)3, RuCl3, RhCl3, H2PtCl6, and HAuCl4. Among them, treatment with AgNO3 aqueous solution improved the photocurrent significantly. Figure 12 (upper panel) shows the potential-current curve of BiVO4 without (a) and after (b) the AgNO3 treatment. The photocurrent increased sharply with a positive potential sweep in the case of the Ag-treated electrode. The IPCE of the BiVO4 electrode in the visible-light region nearly doubled with AgNO3 treatment (see Figure 13). The maximum IPCE at 420 nm was 44%. This value was the second highest to that of the nanocrystalline WO3 electrode among oxide semiconductor electrodes under the visible-light region and was highest among mixed-oxide semiconductor electrodes, suggesting the great potential for porous thin-film electrodes with use of various mixed-metal oxide semiconductors. The photocurrent and stability were not improved by treatment with HNO3 and many other nitrates, except AgNO3, suggesting that H+ and NO3- ions did not contribute to the promotion of photocurrent but that Ag+ ions did contribute to the promotion effect. The promotion effect was observed even at low concentrations of AgNO3 aqueous solution (1 mmol/L). AgNO3 treatment improved not only the IPCE but also the stability of the photocurrent (see Figure 11b), whereas the photocurrent without AgNO3 treatment decreased gradually with irradiation time (Figure 11a). The photocurrent with AgNO3 treatment was maintained at around 4 mA/cm2. The turnover number of total electrons per BiVO4 unit was found to be more than 200 for 1 h. The Ag in the BiVO4 electrode was not detected by XRF measurement, suggesting that the total amount of Ag was less than the measurement’s limit of detection (