Photoelectrochemical Oxidation of Water at Transparent Ferric Oxide

Jun 24, 2005 - “In rust we trust”. Hematite – the prospective inorganic backbone for artificial photosynthesis. Debajeet K. Bora , Artur Braun ,...
0 downloads 7 Views 284KB Size
Download PDF
J. Phys. Chem. B 2005, 109, 13685-13692

13685

Photoelectrochemical Oxidation of Water at Transparent Ferric Oxide Film Electrodes Chantal Jorand Sartoretti, Bruce D. Alexander, Renata Solarska, Iwona A. Rutkowska, and Jan Augustynski* De´ partement de Chimie Mine´ rale, Analytique et Applique´ e, UniVersite´ de Gene` Ve, 1211 Gene` Ve, Switzerland

Radovan Cerny Laboratoire de Cristallographie, UniVersite´ de Gene` Ve, 1211 Gene` Ve, Switzerland ReceiVed: March 25, 2005; In Final Form: May 23, 2005

The fabrication of thin-film Fe2O3 photoanodes from the spray pyrolysis of Fe(III)-containing solutions is reported along with their structural characterization and application to the photoelectrolysis of water. These films combine good performance, measured in terms of photocurrent density, with excellent mechanical stability. A full investigation into the effects that modifications of the spray-pyrolysis method, such as the addition of dopants or structure-directing agents and changes in precursor species or carrier solvent, have on the performance of the photoanodes has been realized. The largest photocurrents were obtained from photoanodes prepared from ferric chloride precursor solutions, simultaneously doped with Ti4+ (5%) and Al3+ (1%). Doping with Zn2+ also shows promise, cathodically shifting the onset potential by ∼0.22 V.

1. Introduction Apart from very few exceptions, semiconducting materials that are able to efficiently absorb visible light are unstable in aqueous solutions when polarized anodically where they most frequently undergo either photo-oxidation or photodecomposition. This is the reason recent reports describing the extension of the response of a chemically stable photocatalyst, titanium dioxide, into the visible region of the solar spectrum have provoked a great deal of interest. Although doping with carbon, nitrogen, or sulfur1-4 allows extension of the photocatalytic activity of TiO2 into visible wavelengths, attempts to use such materials as photoanodes for water splitting have demonstrated low conversion efficiencies up until now.3,5 In particular, the corresponding photocurrents recorded under simulated solar illumination remain an order of magnitude lower than those obtained using nanocrystalline tungsten trioxide photoanodes synthesized in this laboratory.6 The latter material, which has a band gap energy (Eg) of 2.5 eV, exhibits a large photoresponse to the blue part of the visible spectrum up to ∼500 nm. It is also important to note recent reports7 which describe the photooxidation of water at nanocrystalline BiVO4 (Eg of 2.4 eV) thinfilm electrodes, showing significant photon-to-current-conversion efficiencies in the 400-450 nm wavelength range. Another semiconductor which has been the subject of investigation for some time with regard to an application for the photoelectrolysis of water is iron(III) oxide. Apparently, R-Fe2O3, hematite (Eg ) 2.2 eV), is the smallest band gap n-type semiconductor with proven photoanodic stability in aqueous solutions. When used as a photoanode, it should, in principle, extend the accessibility to the amount of useful solar energy to ∼40%. Thus, the attractiveness of this material has prompted a large number of reports concerning the production and photoelectrochemical performance of thin films of hematite for quite some time.8-11 Interestingly, some of the older8,12 and quite * Corresponding author. Fax: [email protected].

+41

22

379

6830.

E-mail:

recent reports13,14 describe also the formation of p-type Fe2O3, obtained through Mg(II) or Zn(II) doping, and its characteristics as a photocathode.15 However, the long-term stability of the latter materials as photocathodes for hydrogen evolution remains to be proven. Although n-type R-Fe2O3 photoanodes exhibit apparently good chemical stability in alkaline solutions, they also notoriously suffer from poor conductivity and from a short holediffusion length, which results in a high electron-hole recombination rate. The problem of how best to fabricate efficient Fe(III) oxide-based photoanodes remains. With a variety of methods of thin-film production, numerous experimental parameters evidently exist which can influence the composition and performance of the resultant photoanodes. This article reports the preparation of thin-film hematite photoanodes that are effective for the photosplitting of water. The effects of various experimental modifications, such as different dopants, carrier solvents, deposition temperatures, and structure-directing additives, to a previously reported basic procedure9 are determined in order to identify the most efficient method for the production of Fe(III) oxide photoanodes and in an attempt to link their structural characteristics to their photoelectrochemical performance. 2. Experimental Section Photoelectrodes made from thin films of hematite were prepared by spray pyrolysis of Fe(III)-containing solutions. Precursor solutions were initially prepared in absolute ethanol (puriss.) from 0.01 to 0.05 M Fe(acetylacetonate)3 (giving type A electrodes), 0.1 M FeCl3‚6H2O (affording type C electrodes), and 0.1 M Fe(NO3)3‚9H2O (producing type N electrodes). Acetylacetonate, methanol, n-butanol, and 2-propanol (all puriss.) were also used as alternative solvents, as indicated in the text. The effect of a wide range of dopants on the performance of the photoelectrodes was investigated by the addition of different dopants to the precursor solutions. These consisted of lithium perchlorate, lithium acetylacetonate, alu-

10.1021/jp051546g CCC: $30.25 © 2005 American Chemical Society Published on Web 06/24/2005

13686 J. Phys. Chem. B, Vol. 109, No. 28, 2005 minium(III) chloride hexahydrate, titanium(IV) ethoxide, chrome(III) chloride hexahydrate, nickel(II) chloride hexahydrate, indium(III) chloride hydrate, indium(III) acetylacetonate, tin(II) chloride dihydrate, tantalum(V) ethoxide, zinc(II) chloride, ammonium tetrachloroplatinate(II), and hydrogen hexachloroplatinate(IV) hydrate. In certain cases, poly(ethylene glycol) 300 (PEG) or diethylene glycol monobutyl ether was added to the precursor solution. All chemicals were purchased from Fluka Chemie AG (Buchs, Switzerland). Glass plates with a 0.5-µm-thick F-doped SnO2 overlayer (Nihon Sheet Glass Co., 12 Ω/sq) were sprayed with the precursor solution at a temperature between 370 and 450 °C for type C electrodes and 370 and 440 °C for type A electrodes. A layer of iron oxide was deposited by spraying the precursor solution onto the substrate from a carrier gas of nitrogen (flow rate ∼7.5 L min-1) at a distance of 12 cm from the plate. A spray of 10 s was followed by a pause of 5 min to allow for the thermal hydrolysis of the freshly sprayed layer and to limit temperature losses in the substrate. This procedure was repeated until the desired number of layers had been deposited. Typically, the type A photoelectrodes consisted of 6-12 layers, whereas the type C photoelectrodes consisted of six layers. Following the spray pyrolysis, the freshly prepared photoelectrodes were rapidly cooled under a flow of nitrogen during a period of 5-10 min. The photoelectrodes were then annealed in air at 550 °C for 60 min before being rapidly cooled under a flow of nitrogen for 5-10 min. The samples have been routinely analyzed by Raman microscopy. Spectra were recorded using a LabRam I Raman microscope (Jobin Yvon), fitted with a CCD camera, operating in confocal mode, that is, by closing the confocal pinhole to 200 µm, with a 100× sampling objective (N.A. 0.95). The irradiation source was a frequency doubled Nd:YAG laser (532.0 nm); the power at the sample was kept below 100 µW in order to minimize sample degradation or structural alteration due to laser heating. Spectra were recorded from a number of positions on the surface and were found to be highly reproducible. Further X-ray structural analysis of the films was performed on a Bruker D8 diffractometer with Bragg-Brentano geometry using Cu KR1 radiation monochromatized with a curved Ge-crystal primary monochromator. Patterns were measured in the range 5-100° 2θ with a step of 0.007 24 using a Braun positionsensitive detector (PSD-50M) with a window opening of 2.5° 2θ. The total measuring time was 3.7 h per pattern. Diffraction patterns were analyzed by Rietveld refinement using the FullProf.2k16 program. The pseudo-Voigt function was used to model the reflection profile, and the preferred orientation of the crystallites was modeled with the March-Dollase function.17 One scale factor, two lattice parameters, one overall isotropic displacement parameter, four profile parameters, and one parameter of the preferred orientation were varied in the refinement. The background was interpolated between fixed points. UV-vis transmission spectra were recorded on a PerkinElmer Lambda 900 spectrophotometer. The thicknesses of the thin films were measured using a Tencor Alpha Step 200 profilometer. For the photoelectrochemical measurements, the Fe(III) oxide electrode was illuminated from the side of the electrolyte|film interface in a 0.1 M NaOH(aq) solution in a Teflon cell equipped with a quartz window. The counter and reference electrodes were a large-area Pt foil and a mercuric oxide electrode, respectively. All potentials are quoted versus a normal hydrogen electrode (NHE), and unless otherwise stated, all photocurrents mentioned in the text are those recorded at 0.45 V versus NHE.

Sartoretti et al.

Figure 1. Dependence of photocurrent density recorded in 0.1 M NaOH(aq) at 0.6 V vs NHE on substrate deposition temperature for type C electrodes doped with 5% Ti and 1% Al.

The wavelength photoresponse (i.e., the incident-photon-tocurrent-conversion efficiency vs excitation wavelength) of the Fe(III) oxide electrodes was determined using a 500 W xenon lamp (Ushio UXL-502HSO) set in an Oriel model 66021 housing and a Multispec 257 monochromator (Oriel). The absolute intensity of the incident light from the monochromator was measured with a model 730 A radiometer/photometer (Optronic Lab). The typical bandwidth was 4 nm. The Fe(III) oxide electrodes were also illuminated with the full output of the 150 W xenon lamp in order to perform an effective screening procedure of the photoanodes. All solutions were made from analytical-grade reagents and Milli-Q water, and photoelectrochemical measurements were performed under potentialcontrolled conditions. 3. Results and Discussion Standard Films. Along with the spray time, the temperature of deposition had been previously found to be an important factor in thin-film formation.13 Figure 1 shows the dependence of the photocurrent densities obtained from our type C electrodes upon the deposition temperature of the conducting glass substrate, which was measured directly by a K-type thermocouple. No such dependence was obtained for type A electrodes; the annealing temperature was found to be the determining factor in terms of photocurrents. This was due to the fact that iron(III) acetylacetonate decomposes completely below 370 °C and therefore a higher temperature of deposition becomes unnecessary.18 At the outset, it had to be verified at what level the choice of precursor solution would be important in determining the performance of the photoelectrodes. Electrodes were initially prepared from standard inorganic and organic ferric salts. A clear dependence upon the nature of the precursor was quickly found. Undoped type A and C electrodes were found to consistently give good to satisfactory photocurrents (∼0.9 mA cm-2) and, equally crucially, good mechanical stability (no instances of corrosion were observed, despite repeated photocurrent-voltage or incident-photon-to-current-conversion efficiency (IPCE) measurements). As can be seen in Figure 2, the IPCEs of undoped type A photoanodes are already promising, reaching 15% for photoanodes prepared from our standard procedure described above (Figure 2c and d). Interestingly, it seems that increasing the thickness of type A electrodes does not appreciably enhance the IPCEs, reflecting the high holeelectron recombination rate of R-Fe2O3, and it is only after an extended thermal treatment that an increase in the IPCE is observed. Analysis of the composition of these electrodes by Raman spectroscopy showed (Figure 3) that both types of electrodes contained mainly R-Fe2O3, although a small amount

Photoelectrochemical Oxidation of Water

J. Phys. Chem. B, Vol. 109, No. 28, 2005 13687

Figure 2. IPCEs recorded in 0.1 M NaOH(aq) at 0.7 V vs NHE for (a) a 6-layer type C electrode doped with 5% Ti and 1% Al, (b) a 15-layer undoped type A electrode, (c) a 12-layer undoped type A electrode, and (d) a 6-layer undoped type A electrode. All electrodes were subjected to a 1 h annealing step at 600 °C, except the 12-layer type A electrode (c) which was treated thermally for 3 h. All type A electrodes were prepared from a 0.05 M solution of Fe(acetylacetonate)3.

Figure 4. (a) Scanning electron micrographs of undoped type A electrodes prepared from a 0.05 M solution of Fe(acetylacetonate)3 (i) before and (ii) after annealing. (b) Raman spectra recorded from the same electrodes.

Figure 3. Typical Raman spectra of iron oxide photoanodes: (a) type C; (b) type A; (c) type N.

of substoichiometric iron oxide, either Fe3O4 or FeO, was always present.9,19 Measurement of the thickness of a six-layer type A film by profilometry was found to be 0.35 µm on average. Sixlayer and nine-layer type C films were 0.5 and 0.8 µm thick, respectively. Type C electrodes are thicker than the corresponding type A electrodes due to the increased concentration of the precursor solution, which is itself a function of the increased solubility of ferric chloride hexahydrate in ethanol. Type N electrodes gave moderately poor photocurrents in comparison. Furthermore, analysis of the Raman spectra obtained from type N electrodes revealed a lack of order, shown in Figure 3. The very weak spectra were dominated by a broad peak at 656 cm-1, which is attributable to Fe3O4. Small amounts of R-Fe2O3 were also present. Given the poor photocurrents (e.g., 0.38 mA cm-2) obtained from type N electrodes in comparison to those of types A or C, coupled with the extremely poor adhesion of the type N films to the conducting glass substrate, often detaching from the substrate after two photocurrent measurements, the development of type N electrodes was abandoned. It quickly became apparent that the postdeposition thermal treatment was also an important factor in obtaining electrodes of good performance. For example, a large increase (up to 7fold) in the photocurrents obtained from undoped type A electrodes was observed upon annealing in air at 550 °C for 1 h.

SEM micrographs for undoped type A electrodes are shown in parts i (before annealing) and ii (after annealing) of Figure 4a. Here, the micrographs look remarkably similar: in both cases, a weak porosity is observed with particle diameters in the 70300 nm range. While the annealing process seems to have little influence on the surface morphology, the composition of the films is altered, as evidenced by Raman microscopy (Figure 4b). All of the peaks present in the Raman spectra before (i) and after (ii) annealing at 550 °C for 1 h are due to R-Fe2O3, except the band present at ∼660 cm-1, which was previously assigned to magnetite (Fe3O4) or disorder (such as that induced by the presence of wu¨stite, FeO).9,19 It is notable that this peak decreases in intensity upon annealing, indicating an increase in the crystalline content of the film, which is predominantly R-Fe2O3. It is further notable that the photocurrents are lower when there is an increase in disorder visible in the Raman spectra, as shall be discussed for other cases such as the effect of dopants and carrier solvent (vide infra). Further confirmation of the composition of the type A and C films came from X-ray diffraction. Hematite was identified in the diffraction patterns of both type A and type C electrodes. Rietveld plots are given in Figure 5 for type C electrodes. The powder patterns are of mediocre quality due to the very small diffracting volume of the thin layers; however, χ2 ) 1.18 and 1.12 for the as-deposited and annealed samples, respectively. The refined lattice parameters were found to be a ) 5.0321(3) Å and c ) 13.7377(6) Å and a ) 5.0316(3) Å and c ) 13.7328(6) Å for the as-deposited and annealed samples, respectively, showing a decrease of the parameter c upon annealing. Both types of samples show strong preferred orientation, with the basal plane (001) preferentially oriented in the plane of the substrate. The diffraction pattern of the as-deposited layer shows an impurity peak at d ) 2.714, which is difficult to assign with

13688 J. Phys. Chem. B, Vol. 109, No. 28, 2005

Figure 5. X-ray diffractogram for a type C electrode doped with 2.5% Ti (a) before and (b) after annealing. Part c shows an expanded view of the impurity peak. Raw diffractograms are shown in all cases. Expected peak positions for R-Fe2O3 are indicated.

confidence but may be related to a bixbyite-type structure, β-Fe2O3,20,21 or to Ti2O3.22 The impurity peak is not observed in the diffractogram of the annealed sample (see Figure 5). Dopants. The addition of dopants, either in small or in large quantities, to iron oxides has been known in some cases to improve the photocurrents, either by increasing the conductivity, by favoring the formation of mixed oxides, or by acting as a structure-directing agent.8,23,24 To this end, a full investigation

Sartoretti et al. into the effect of a wide range of dopants on the prepared thin films of R-Fe2O3 has been carried out. Figure 6 details a schematic representation of the effect of some dopant levels on typical photocurrents. Photocurrents are reported at two potentials, 0.45 and 0.59 V versus NHE. The former potential may be considered to be safely below the thermodynamic oxidation potential and facilitates comparison with most published results. In certain cases in the literature, photocurrents have been reported at more anodic potentials,10,25 and thus, the 0.59 V values are included to assist comparison. As can be seen from the current-voltage curves (vide infra), the dark current recorded from the photoanodes reported herein can be considered to be negligible at both potentials. It should be stressed that, in the following discussion, only photocurrent values at 0.45 V are discussed. Doping type C electrodes with a substantial amount (5%) of Ti(IV) improved photocurrents markedly from 0.78 mA cm-2, in the undoped case, to 4.05 mA cm-2, most likely owing to the increased conductivity of the films and the stabilization of oxygen vacancies by Ti4+ cations. Similar attempts were made using Pt4+, which has been reported to cause an extension of the photoresponse of hightemperature sintered R-Fe2O3 photoanodes. However, the addition of chloroplatinic acid (1%) caused an almost total suppression of the photocurrent, which was at best 0.05 mA cm-2. Interestingly, multiple doping of up to 5% Pt2+ in the presence of 2.5% titanium gives an improvement in the photocurrents with respect to Pt4+-doped electrodes of around 0.4 mA cm-2. For similar reasons, electrodes were prepared from precursor solutions doped with 1% Ta5+, although no real improvements in the photocurrents were observed. Another strategy for improving the performance of the electrodes was to add small quantities of dopants which should, in principle, form oxides that are isostructural with R-Fe2O3, such as Al3+, Cr3+, or In3+, thereby directing the crystallinity of the thin films. These were added along with 5% titanium, which had previously been found to be the most effective dopant.9 It was thought that, as indium oxide is also an n-type semiconductor with a band gap similar to that of R-Fe2O3, a clear effect on the performance of the electrodes could be observed. However, the addition of indium in large quantities (up to the solubility limit of a solid solution, i.e., 30%)26 caused an almost complete suppression of the photocurrents, even after an additional heat treatment. In fact, the Raman spectra of the electrodes prepared with the addition of such large quantities of indium were much weaker than those without indium and displayed a substantial increase (4-5-fold) in the broad band at ∼660 cm-1, previously assigned to magnetite or disorder,9,19 therefore indicating that indium is a rather poor additive. Better results were evident upon the substitution of indium for 5% Cr3+, although photocurrents remained weaker than those of the electrodes containing only 5% titanium (see Figure 6). The coaddition of aluminium, on the other hand, seems to be more efficient than that of indium or chromium. Simultaneous doping of the precursor solutions with 5% aluminium and 5% titanium led to a slight reduction in photocurrents. However, an optimum amount of aluminium doping (1% introduced as AlCl3‚6H2O) was found to improve the performance of the photoanodes, increasing photocurrents from 4.05 mA cm-2, in the case of 5% Ti alone, to 4.38 mA cm-2, in the case of 1% Al and 5% Ti for type C electrodes. Indeed, incident-photonto-current efficiencies (IPCEs) are very promising, as shown in Figure 2. With multiple doping of 1% Al and 5% Ti, an IPCE of 25% is attained at 400 nm. Furthermore, the spectral photoresponse spectrum extends well into the visible part of

Photoelectrochemical Oxidation of Water

J. Phys. Chem. B, Vol. 109, No. 28, 2005 13689

Figure 6. Comparison of the effect of dopants and annealing (labeled “TT”) for type C electrodes.

the spectrum: at 580 nm, the IPCE remains greater than 2%. When aluminium was employed in isolation, that is, monodoping, the measured photocurrents were low. Furthermore, when aluminium was introduced in the form of R-aluminium(III) oxide, that is, a preformed species isostructural with R-Fe2O3, it lead to poor performance (at best, 1.56 mA cm-2) and thus AlCl3‚6H2O was used as the source of Al doping throughout this work. Simultaneous doping with dopants that should favor the formation of n- and p-type ferric oxide was also attempted. Lithium was added in small quantities to the precursor solutions along with 5% Ti as a direct attempt to improve the holediffusion length of the resulting material and therefore the performance. As can be seen in Figure 7a and b, the photocurrent responses of the electrodes are improved slightly (by ∼0.15 mA cm-2) upon augmenting the quantity of lithium added from 0.1 to 1%; however, photocurrents remain comparatively low (see Figures 6 and 7). In this regard, much more convincing results were obtained when Li+ was replaced with a Zn2+ dopant recently reported to induce p-type behavior of ferric oxide.14 The type C photoelectrodes doped with 4% Zn2+ along with 5% Ti (Figure 7d) showed a clear (by ∼0.22 V) cathodic shift of the photocurrent onset potential and a steeper rise of the photocurrent with the applied voltage. Further increase in the amount of Zn2+ to 8%, found by Ingler et al. to be the optimum doping level for the p-type ferric oxide films,14 did not produce any further improvement in the behavior of the present dual Ti4+/Zn2+-doped n-type photoelectrodes. It is also to be noted that even at the highest Zn2+ doping level our ferric oxide films still exhibited only anodic photocurrents consistent with the n-type behavior. Finally, the effect of doping with nickel on the performance of the photoanodes was examined. The addition of 5% nickel caused a dramatic suppression of the photocurrents.

Surface Treatment. The photo-oxidation of water occurring at the electrode surface may be further facilitated by the addition of surface catalysts. Electrodes were modified with an additional thin layer of ruthenium oxide (RuO2), obtained after spraying the R-Fe2O3 thin films with a solution of RuCl3 (0.5 mM) in HClO4 (0.1 M). Only one layer was added before annealing at 550 °C in oxygen for 30 min. As shown in Figure 8, covering the surface with a layer of RuO2 electrocatalyst did not significantly modify the measured photocurrents, nor the onset potential, as has previously been claimed.25 Precursor Solution Effects. It has already been found that the addition of an organic structure-directing agent to the precursor solution may affect both the crystallinity and porosity of the resulting thin films.6,27 In the present case, poly(ethylene glycol) (PEG) was added in either 1:1 or 1:4 Fe/PEG (v/v) ratios to both type A and type C electrodes. In almost all cases, the photocurrents of electrodes measured before the final annealing process were negligible and were typically of the order of a few microamperes per square centimeter. This fact is clearly linked to the delay of crystallization induced by the PEG. Raman spectra of the PEG-modified electrodes are much weaker than those of the unmodified electrodes and are dominated by broad features centered at 665 cm-1 due to Fe3O4, as shown in Figure 9. It may also be seen from Figure 9 that the thermal treatment affects a conversion between the substoichiometric iron oxide and R-Fe2O3, evidenced by an increase in the relative intensities of the bands between 200 and 600 cm-1. Indeed, the introduction of an additional coordinating component into the ethanol solvent suppressed the photocurrents in general. For example, the addition of diethylene glycol monobutyl ether had a clear detrimental effect, with the largest photocurrents being at best half of those obtained from a glycol-free electrode. Here, the quantity of Fe3O4 found in the Raman spectra was found to be higher than that of electrodes prepared from pure ethanolic

13690 J. Phys. Chem. B, Vol. 109, No. 28, 2005

Figure 7. Photocurrent-voltage curves for type C electrodes doped with (a) 0.2% Li and 5% Ti, (b) 1.0% Li and 5% Ti, (c) 5% Ti and 5% Sn, and (d) 4% Zn and 5% Ti. Parts a and b were before thermal treatment, whereas parts c and d were after annealing. (e) Typical photocurrent-voltage characteristics for a 5% Ti-doped electrode. (f) Dark current from the same electrode.

solutions. Furthermore, the amount of Fe3O4 seemed to increase with the glycol fraction of the precursor solution. The spray-pyrolysis method employed herein will clearly be sensitive to the choice of solvent, which dictates, among others things, the size of the particles in the spray due to a tradeoff between solvent viscosity and boiling point. A range of different solvents has been tested for both type A and type C electrodes. A summary of these is displayed in Figure 10. For type C electrodes, changing the solvent from ethanol to n-butanol decreases the photocurrent, while the use of a coordinating solvent such as acetylacetone has an even more dramatic effect. In fact, it appears that the electrodes prepared from acetylacetonic solutions require a longer heat treatment in order to crystallize the iron oxide. Furthermore, these electrodes have a substantially thinner iron oxide layer than those prepared from other solvents, evidenced by the very pale appearance of the electrodes and the difficulty in obtaining Raman spectra or XRD diffractograms. For one electrode, the photocurrent at 0.45 V is 0.012 mA cm-2 before thermal treatment. This value rises to only 0.074 mA cm-2 after annealing at 600 °C for 16 h. Raman analysis shows that the former photoanode is largely disordered, containing only trace crystalline material, whereas the latter becomes much more ordered, being predominantly Fe3O4 with a sufficient quantity of R-Fe2O3. Similar drops in photocurrent when acetylacetone was used as the solvent were observed for type A electrodes.

Sartoretti et al.

Figure 8. Effect of the application of a RuO2 catalyst overlayer on the photocurrent-voltage curves for undoped type A electrodes prepared from a 0.05 M solution of Fe(acetylacetonate)3: (a) before; (b) after the application of the overlayer; (c) dark current.

Figure 9. Effect on the Raman spectra of 5% Ti-doped type C electrodes modified by an organic additive shown in parts b-d. (a) No organic additive after annealing for 1 h at 600 °C; (b) following the addition of 1:4 (v/v) PEG before thermal treatment and (c) after annealing for 1 h at 600 °C; and (d) 1:4 (v/v) diethylene glycol monobutyl ether before thermal treatment.

Interestingly, initial results indicate that small modifications in the solvent viscosity, introduced through using binary mixtures of ethanol with methanol or 2-propanol, indeed affect the photocurrents, if only slightly. For example, a type C electrode of six layers doped with 5% Ti and 1% Al prepared from a pure ethanol solution gives a photocurrent of 3.92 mA cm-2, whereas similar type C electrodes prepared from 1:1 ethanol/methanol and ethanol/2-propanol give photocurrents of 3.05 and 4.08 mA cm-2, respectively. These differences are

Photoelectrochemical Oxidation of Water

J. Phys. Chem. B, Vol. 109, No. 28, 2005 13691 addition of tin to the precursor solution does not represent an improvement upon the electrodes doped with 5% titanium; in fact, measured photocurrents are reduced in comparison. To further emphasize the fact that there is little positive influence on the performance of the electrodes due to the migration of atoms from the substrate to the R-Fe2O3 layer during the annealing process, other substrates were employed: metallic titanium sheets (pre-etched in boiling oxalic acid for 15 min), Sb-doped SnO2 conducting glass, and ITO electrodes. No improvement in the performance of the electrodes was observed, and in the case of the titanium sheets sprayed with ferric chloride precursor solution containing 5% of Ti(IV), the photocurrents obtained reached only two-thirds of those normally observed for 5% Ti-doped type C electrodes deposited on conducting glass substrates, although the addition of titanium into the precursor solution has beneficial effects. Significantly, the dark currents remained very low (less than 20 µA cm-2 up to a voltage of 0.45 V). 4. Conclusions

Figure 10. Photocurrent-voltage curves for 5% Ti-doped type C electrodes prepared from (a and b) n-BuOH and (c and d) acetylacetonate solutions. Part b shows the photocurrent-voltage curve after annealing for 1 h at 550 °C, whereas part d shows the photocurrentvoltage curve after annealing for 16 h at 600 °C. (e) Dark current for the 5% Ti-doped type C electrode prepared from n-BuOH.

most probably due to the changing size of the particles in the spray on going from a less viscous, more volatile solvent mixture to a more viscous, less volatile one. Finally, attempts were made to initiate the formation of colloidal particles in the precursor solution in order to act as a seed for the directed formation of R-Fe2O3. To this extent, solutions of Fe(acac)3 in ethanol or acetone were subjected to ultrasonic pretreatment prior to spray deposition, which has previously been shown to influence the formation of microcrystallites. The Raman spectra of electrodes prepared with and without ultrasonic pretreatment were, in certain cases, almost superimposable, indicating a negligible change in the crystallinity of the films. Nevertheless, measured photocurrents indicate that an improvement may be possible in the ultrasonicated samples. Effect of Substrate. Given that the electrodes are subjected to extended heat-treatment processes, it seemed logical to examine the possible effects of the transfer of tin from the conducting glass overlayer into the R-Fe2O3 thin film during the prolonged heating process. To this extent, electrodes were prepared from precursor solutions with the addition of 1-5% tin. Figure 7c shows that, with the addition of 5% Ti4+, the electrodes exhibit poor photoelectrochemical behavior and photocurrents are often much less than 1 mA cm-2. Upon thermal treatment, a substantial improvement of the photocurrent is observed: in the case of the electrodes doped with both tin and titanium, annealing at 550 °C for 6 h increases the photocurrent from 0.06 to 1.29 mA cm-2. Nevertheless, the

Spray pyrolysis of Fe(III)-containing solutions onto a conducting glass substrate affords electrodes predominantly consisting of R-Fe2O3, although a small amount of disorder in the dense submicron thick films is observed by Raman spectroscopy. Good mechanical stability was observed for type A and type C electrodes, whereas poor adherence characterized electrodes prepared from a ferric nitrate precursor, which separated from the substrate after two photocurrent measurements. An extensive optimization of the current spray-pyrolysis technique has been carried out. For the present system, the most efficient electrodes are those prepared from ethanolic solutions of ferric chloride or ferric acetylacetonate. Other solvent choices resulted in electrodes that exhibited a poorer performance, usually due to an observed increase in disorder in the Fe2O3 films. In most cases, the addition of dopants had a detrimental effect on the performance of the electrodes. However, large photocurrents were observed following the addition of either Al3+ (1%) or Zn2+ (4%) in conjunction with Ti4+ (5%). In particular, the Zn2+/Ti4+ doping of ferric oxide seems to affect the holediffusion length, which leads to a favorable negative shift of the photocurrent onset potential. A clear link could be established between the extent of crystallinity of the films, monitored by Raman spectroscopy, and their photoactivity. The most active films predominantly consist of well-crystallized R-Fe2O3 with the presence of only small amounts of substoichiometric iron oxide, most probably Fe3O4. Acknowledgment. This work was supported by the Swiss Federal Office of Energy. The authors also wish to thank Prof. Milena Koudelka-Hep and Dr Olivier Guenat at the Institute of Microtechnology, University of Neuchaˆtel, for their assistance with the alpha-step profilometric measurements and Dr Hubert Cachet, Universite´ Pierre et Marie Curie, Paris, for the generous donation of the Sb-doped SnO2 conducting glass. References and Notes (1) Khan, S. U. M.; Al-Shahry, M.; Ingler, W. B., Jr. Science (Washington, D.C.) 2002, 297, 2243-2245. (2) Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science (Washington, D.C.) 2001, 293, 269-271. (3) Torres, G. R.; Lindgren, T.; Lu, J.; Granqvist, C.-G.; Lindquist, S.-E. J. Phys. Chem. B 2004, 108, 5995-6003. (4) Diwald, O.; Thompson, T. L.; Zubkov, T.; Goralski, E. G.; Walck, S. D.; Yates, J. T., Jr. J. Phys. Chem. B 2004, 108, 6004-6008. (5) Noworyta, K.; Augustynski, J. Electrochem. Solid-State Lett. 2004, 7, E31-E33.

13692 J. Phys. Chem. B, Vol. 109, No. 28, 2005 (6) Santato, C.; Ulmann, M.; Augustynski, J. J. Phys. Chem. B 2001, 105, 936-940. (7) Sayama, K.; Nomura, A.; Zou, Z.; Abe, R.; Abe, Y.; Arakawa, H. Chem. Commun. (Cambridge) 2003, 2908-2909. (8) Anderman, M.; Kennedy, J. H. In Studies in Physical and Theoretical Chemistry; Finklea, H. O., Ed.; Elsevier: Amsterdam, The Netherlands, 1988; Vol. 55. (9) Jorand Sartoretti, C.; Ulmann, M.; Alexander, B. D.; Augustynski, J.; Weidenkaff, A. Chem. Phys. Lett. 2003, 376, 194-200. (10) Khan, S. U. M.; Akikusa, J. J. Phys. Chem. B 1999, 103, 71847189. (11) Lindgren, T.; Vayssieres, L.; Wang, H.; Lindquist, S.-E. Chem. Phys. Nanostruct. Semicond. 2003, 83-110. (12) Leygraf, C.; Hendewerk, M.; Somorjai, G. J. Solid State Chem. 1983, 48, 357-367. (13) Ingler, W. B., Jr.; Khan, S. U. M. Thin Solid Films 2004, 461, 301-308. (14) Ingler, W. B., Jr.; Baltrus, J. P.; Khan, S. U. M. J. Am. Chem. Soc. 2004, 126, 10238-10239. (15) Leygraf, C.; Hendewerk, M.; Somorjai, G. A. J. Phys. Chem. 1982, 86, 4484-4485.

Sartoretti et al. (16) Frontera, C.; Rodriguez-Carvajal, J. Physica B 2003, 335, 219222. (17) Dollase, W. A. J. Appl. Crystallogr. 1986, 19, 267-272. (18) Pal, B.; Sharon, M. Thin Solid Films 2000, 379, 83-88. (19) De Faria, D. L. A.; Silva, S. V.; De Oliveira, M. T. J. Raman Spectrosc. 1997, 28, 873-878. (20) Ikeda, Y.; Takano, M.; Bando, Y. Bull. Inst. Chem. Res., Kyoto UniV. 1986, 64, 249-258. (21) Geller, S. Acta Crystallogr., Sect. B 1970, 27, 821-828. (22) Abrahams, S. C. Phys. ReV. 1963, 130, 2230-2237. (23) Kennedy, J. H.; Frese, K. W., Jr. J. Electrochem. Soc. 1978, 125, 709-714. (24) Morin, F. J. Phys. ReV. 1951, 83, 1005-1010. (25) Majumder, S. A.; Khan, S. U. M. Int. J. Hydrogen Energy 1994, 19, 881-887. (26) Ristic, M.; Popovic, S.; Tonkovic, M.; Music, S. J. Mater. Sci. 1991, 26, 4225-4233. (27) Santato, C.; Odziemkowski, M.; Ulmann, M.; Augustynski, J. J. Am. Chem. Soc. 2001, 123, 10639-10649.