DOI: 10.1002/cssc.201402918
Full Papers
High-Throughput Screening of Thin-Film Semiconductor Material Libraries II: Characterization of Fe¢W¢O Libraries Robert Meyer,[a] Kirill Sliozberg,[b] Chinmay Khare,[a] Wolfgang Schuhmann,*[b, c] and Alfred Ludwig*[a, c] Metal oxides are promising materials for solar water splitting. To identify suitable materials within the ternary system Fe¢W¢ O, thin-film material libraries with combined thickness and compositional gradients were synthesized by combinatorial reactive magnetron sputtering. These libraries (> 1000 different samples) were investigated by means of structural and functional high-throughput characterization techniques to establish correlations between composition, crystallinity, morphology,
thickness, and photocurrent density in the compositional range between (Fe6W94)Ox and (Fe61W39)Ox. In addition to the well-known phase WO3, the binary phase W5O14 and the ternary phase Fe2O6W show enhanced photoelectrochemical activity. The highest photocurrent density of 65 mA cm¢2 was achieved for the composition (Fe15W85)Ox, which contains the W5O14 phase and has a thickness of 1060 nm.
Introduction The conversion of renewable energy sources into technically usable forms of energy is a key factor for a sustainable economy in the future. A promising technology is solar water splitting (SWS) in a photoelectrochemical cell to convert solar energy directly to chemical energy in form of molecular hydrogen (and oxygen).[1] Possible configurations of a photoelectrochemical cell are the combination of n-type semiconductor photoanodes with p-type semiconductor photocathodes or metal counter electrodes immersed in an aqueous electrolyte. Many candidates for anode materials have been identified, most of them being metal-oxide semiconductors such as TiO2, WO3, and Fe2O3.[2–6] These materials are matching several necessary criteria, such as high abundance and thereby a relatively low price, high stability in aqueous media, and non-toxicity. However, these single-metal-oxide semiconductors do not exhibit the necessary performance for application in SWS devices. Recent research, therefore, focused on multinary-metal-oxide semiconductors, many of which are based on TiO2, WO3, or Fe2O3, either mixed with each other or doped with single elements such as Ta.[7–9]
[a] R. Meyer, Dr. C. Khare, Prof. Dr. A. Ludwig Institute for Materials Ruhr-University Bochum Universittsstraße 150, 44801 Bochum (Germany) E-mail: [email protected] [b] K. Sliozberg, Prof. Dr. W. Schuhmann Analytical Chemistry—Center for Electrochemical Sciences (CES) Ruhr-University Bochum Universittsstraße 150, 44801 Bochum (Germany) E-mail: [email protected] [c] Prof. Dr. W. Schuhmann, Prof. Dr. A. Ludwig Materials Research Department Ruhr-University Bochum Universittsstr. 150, 44801 Bochum
ChemSusChem 2015, 8, 1279 – 1285
An effective method to identify suitable materials for SWS is seen in the thin-film combinatorial material science approach.[10–13] Instead of analyzing only a single point in a huge parameter space, the combinatorial approach provides a large number of different but very well defined and well comparable samples in a material library (ML). Using specially designed high-throughput characterization tools, this method has the potential to accelerate the development of thin-film-based SWS systems as compared to the “one by one” approach.[14] Herein, the thin-film combinatorial approach was applied to investigate a large fraction of the Fe¢W¢O system. The binary systems Fe¢O and W¢O are well studied materials for SWS photoanodes, but both have disadvantages such as a high recombination rate (a-Fe2O3) or a limited width of the absorption spectrum (WO3).[15, 16] By combining Fe¢O and W¢O in the Fe¢ W¢O system, these disadvantages might be overcome. By changing the short- and long-range order as a function of composition, the electronic structure might improve towards a higher incident photon-to-current conversion efficiency (IPCE). The altered electronic structure is hard to predict theoretically by means of density functional theory.[8, 17] However, analyzing well-comparable samples from MLs over a wide range of compositions by means of photoelectrochemical screening may provide both qualitative and quantitative trends that provide precise information for further in-depth analysis of hits obtained during screening. Furthermore, new ternary oxide phases exhibiting interesting photoelectrochemical properties could be discovered. To identify correlations between composition, crystallinity, morphology, thickness, and photocurrent density, ternary Fe¢W¢O thin-film MLs were deposited by reactive magnetron sputtering. The MLs were designed such that they consisted of well-defined compositions and thickness gradients (Figure 1).
1279
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers
Figure 1. Schematic of the combinatorial sputter setup to achieve combined composition and thickness gradients on a material library in top view (a) and side view (b).
Results and Discussion To screen for potential SWS materials within the W-rich part of the Fe¢W¢O system, three MLs (ML1, ML2, and ML3) were fabricated. On each ML 342 measurement areas were defined, adding up to 1026 samples ranging from 6 to 61 at % Fe (binary) with thicknesses ranging from 220 to 1160 nm. The photocurrent density, as a generally accepted screening parameter, was used to evaluate the properties of each measurement area on the MLs. To screen for interesting materials, an in-house-developed, fully automated optical scanning droplet cell (OSDC) setup was used, which is described elsewhere.[18] XRD mapping revealed the existence of three different crystalline phases and an amorphous region in the investigated Fe¢ W¢O composition range. These regions show also a thickness dependence, which is not unusual for sputtered thin films.[19] Due to the large amount of obtained data, XRD data are presented in complementary ways. Figure 2 shows discrete q/ 2q scans of ten measurement areas for each ML recorded along the compositional gradient through the center of each ML. In Figure 3, color-coded intensity maps of characteristic major XRD peaks are presented.[20] These can be directly compared to the color-coded chemical composition, photocurrent density, and the thickness maps of the MLs. For samples with Fe contents < 11 at % only the monoclinic WO3 phase is present, showing reflections of (002), (020), (111), (200), and (220) orientation (Figure 2 a). From about 11 at % Fe onwards, these peaks disappear and the main phase changes to W5O14, which has a distinct (001) reflection. This Magn¦li phase was reported by McColm et al. to be a tetragonal (tp152) structure with an oxygen-to-metal ratio of 2.8, which is stabilized by the exisChemSusChem 2015, 8, 1279 – 1285
www.chemsuschem.org
Figure 2. XRD patterns of selected measurement areas of the thin-film MLS a) ML1, b) ML2, and c) ML3, with the corresponding Fe content (binary) and thickness. Originating from the W-filament in the XRD Cu cathode, the W-La peaks appear for the high-intensity peaks of Pt (111) (from the Pt back electrode) and W5O14 (001) and are visible because of the use of the very sensitive PIXcel detector. The same applies to the Kb peaks. The chemical compositions were measured by means of EDX in the center of each measurement area and, therefore, present an average for the area measured by XRD.
tence of small amounts of Fe atoms.[21] Although the amount of Fe is much lower for the reported bulk diffusion samples than for the sputter-deposited samples, the XRD data clearly point to the existence of the W5O14 phase, especially for sam-
1280
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers ples from ML3 (Figure 2 c). Nevertheless the WO3 and the W5O14 phases are likely to co-exist over a large compositional range, where the more pronounced XRD signals are covering the less pronounced signals of the other phase, especially in the range between 11 and 15 at % Fe. Starting at about 20 at % Fe, the W5O14 peak intensity is decreasing again and disappears at around 27 –29 at % Fe. As the measurement areas do not show any phase separation, it is most likely that W atoms are substituted by Fe atoms in the crystal lattice and that the number of substituted W atoms is increasing with increasing Fe content. Yet it is possible that the non-stoichiometry is compensated by amorphous fractions in the film and no significant substitution takes place. At about 44 at % Fe, a ternary crystal phase appears, which was identified as Fe2O6W, showing a (200) reflection peak. The intensity of this peak is increasing with increasing Fe content up to the maximum Fe content of 61 at % and accompanied by the Fe2O6W (061) peak starting from around 49 at % Fe. Unfortunately, the (200) peak is located in the left shoulder of the high-intensity Pt (111) peak of the back electrode. For high intensities, the Fe2O6W (200) signal is clearly distinguishable, but the Pt peak might cover low-intensity peaks at lower Fe contents (Figure 2 b and c). In contrast to the binary phases, for the ternary phase it is most likely that Fe atoms are substituted by W atoms as the Fe-to-W ratio is lower than the stoichiometric ratio of 2:1 in Fe2O6W. Hence, the number of substituted W atoms is decreasing with increasing Fe content. But, as for the W-rich measurement areas, it is possible that the crystalline parts of the film are indeed stoichiometric whereas excess atoms are present in an amorphous phase. Measurement areas with Fe contents between 29 and 44 at % Fe are at least X-ray amorphous for all MLs (thickness depending). As aforementioned for identical compositions, the appearance, intensity, and fading of the observed peaks varies slightly as a function of film thickness for all crystalline and amorphous phases. To understand these variations, three effects have to be considered: 1) With increasing thickness the volume increases, which leads to a higher XRD signal intensity. Thus, crystal phases with low-intensity peaks are easier to detect and to identify in thicker films. 2) The composition is measured over an area of 400 mm Õ 600 mm [energy-dispersive X-ray analysis (EDX)], whereas the phase information (XRD) is averaged over an area of up to 3 mm Õ 3 mm. 3) The longer deposition time for ML3 (60 % longer than for ML1 or ML2) leads to about the same amount of additional thermal power applied to the growing film. This introduces more energy and gives more time for an improved short- and long-range order in the atomic structure and, therefore, higher peak intensities.
Figure 3. Logarithmic color maps of the relative intensities of the main XRD peaks for a) ML1, b) ML2, and c) ML3 (top row). For a direct comparison, the logarithmic intensity maps of the photocurrent density (center), the absolute thickness (right bottom), and the Fe content (left bottom) are shown side by side. The photocurrent density color map is according to Figure 4. Each color map shows a ML on a 100 mm diameter substrate.
ChemSusChem 2015, 8, 1279 – 1285
www.chemsuschem.org
1281
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers The correlations of the measured photocurrent values with thickness, composition and crystal structure are discernible from Figure 4. Taking into account the crystal structure information, the diagram is divided into five sections by dashed black lines and labeled with roman numbers (I–V). The photocurrent density data is presented in two complementary ways:
thicknesses between 700 and 800 nm on ML1. The local maximum for the ternary Fe2O6W phase (V) is located at about 51 at % Fe and a thickness of 345 nm. The absolute value of 15 mA cm¢2 is low compared to the photocurrents of the binary phase. Within the Fe2O6W region, ML2 and ML3 show a deviation in photocurrent density. Although the trend of increasing photocurrent density for increasing crystallinity (starting at 44 at % Fe) is consistent, for ML3 the absolute photocurrent densities are lower than for ML2 for measurement areas with similar composition and thickness. Furthermore, the measurement areas on ML3 show an increase in photocurrent density for increasing thickness, whereas this is not apparent in the same ternary region for ML2. These differences are at least partially related to the morphology of the MLs along the compositional gradient, which is shown in Figure 5. The SEM images are obtained from measurement areas in accordance with the positions of the XRD results in Figure 2, whereas the XRD patterns are only displayed for every other measurement area. Distinct changes in morphology may be related to a change in the crystal phase (c.f. Figure 3). For most of the identical compositions, the morphology of all MLs is identical independent of thickness except for the ternary Fe2O6W phase > 44 at % Fe (ML2 and ML3). ML3 Figure 4. Color-coded map showing the correlations between thickness, Fe-content, and measured photocurrent density acquired for 1026 measurement areas from three MLs. shows a relatively smooth and homogeneous surface The map is divided in five phase regions (I–V) by black dashed lines, indicating the Fe with only few inclusions of bigger grains. The avercontent and thickness for which the measurement areas exhibit the according phase. age grain size, as determined from the SEM microGrey dotted lines trace the measurement areas for which SEM micrographs were recordgraphs, is about 90 nm for 44 at % Fe and increases ed (Figure 5 ), with the start/end points encircled. to 100 nm at 52 at %. ML2, however, shows a morphology consisting of two grain types for composiIn Figure 3 it is plotted as a function of the position on the ML, tions equal or higher than 49 at % Fe. The film consists of large whereas in Figure 4 the photocurrent density is shown colorcrystallites (180–250 nm) embedded in a matrix of smaller coded as a function of composition and thickness for all meagrains (50–70 nm). The number of larger grains decreases with surement areas. The color map scales are identical for Figures 3 increasing Fe content. An EDX intensity map reveals no differand 4. In general, photocurrent densities of measurement ence in composition for the large grains and the surrounding areas with similar composition and thickness exhibit comparamatrix. ble values and show the same trends for all MLs. In a recent To account for the difference in photocurrent density, two paper by Kollender et al., data on thermally evaporated and main effects have to be considered. The first is the difference annealed Fe¢W¢O thin films were published.[22] Contrary to in thickness. Measurement areas of ML3 are about 200 nm thicker for identical compositions (SEM micrographs in our MLs obtained by reactive co-sputtering, which are mostly Figure 5). The second is that the deposition time for ML3 was single phase (except for the WO3 + W5O14 part), the evaporated 60 % longer, giving an equivalent of 60 % more thermal power and annealed films consisted of multiple phases over the applied during growth. Consequently, the morphological differwhole compositional range. The photocurrents are in the same ences might lead to higher photocurrent densities due to an order of magnitude. However, the influence of film thickness increased specific surface area for ML3, although the difference on photocurrents could not be taken into account, precluding in photocurrent seems to be larger than the increase in surface a more detailed comparison of the data. area. According to Figure 4, two compositional regions with relaFor the binary W¢O phases, the influence of the microstructively high photocurrents were identified: the binary W¢O ture on the photocurrent density is minor compared to the phases < 30 at % Fe and the ternary phase > 44 at % Fe. The “degree of crystallinity”, as shown by means of peak intensity amorphous region in between shows minimal photocurrents in XRD measurements. However, it has to be taken into considfor all MLs. Measurement areas within the W5O14 phase (III) diseration that the morphology and texture of sputtered thin play the highest photocurrent densities, with an absolute maxfilms are not independent although it is possible to influence imum of about 65 mA cm¢2 at around 15–17 at % Fe and thickmorphology and texture without changing the crystal phase, nesses ranging from 900 to 1060 nm (ML3). A local maximum for example, by altering deposition pressure or rate or the of about 49 mA cm¢2 is detected at 5–6 at % Fe (WO3 ; I) and ChemSusChem 2015, 8, 1279 – 1285
www.chemsuschem.org
1282
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers
Figure 5. SEM micrographs of selected measurement areas of a) ML1, b) ML2, and c) ML3. The binary Fe content is labeled top left, the thickness top right, the phase region (c.f. Figure 3) bottom left, and the photocurrent density (mA cm¢2) bottom right for the presented measurement areas. Each micrograph has a size of 2 mm Õ 2 mm and was recorded at a magnification of 30 kx.
overall film thickness. The influence of the thickness is not as distinct as the influence of the crystallinity but still apparent. Two main factors influencing the capability of a thin-film semiconductor absorber to convert radiation energy into electrical/chemical energy are the absorption (higher thickness favorable) and the charge-carrier transport, usually addressed as recombination rate or carrier lifetime (lower thickness favorable). Hence, an optimum thickness clearly exists and has been reported before, for example, for pure WO3.[23] Herein, Fedoped binary W¢O phases exhibit the highest photocurrent densities. For low Fe contents (6.4 at %) within the WO3 phase (ML1), the photocurrent density gradually increases until a maximum thickness is reached (765 nm). It cannot be excluded that the maximum photocurrent density for this phase is located at even higher thicknesses and lower Fe contents. The region of the globally highest photocurrents within the W5O14 (ML3) phase is located between 910 and 1060 nm thicknesses for Fe contents between 15 and 16 at %. As described above, no thickness-dependent maximum region of photocurrent density could be identified for the ternary Fe2O6W phase. For ChemSusChem 2015, 8, 1279 – 1285
www.chemsuschem.org
ML2, the higher photocurrent densities depend more on composition (45–55 at % Fe) than on thickness. For ML3, the photocurrent density steadily increases with increasing thickness but at low absolute values. Thus, the dependence on morphology might in this case be stronger than the thickness effect. Generally, it was found that the morphology has a weaker influence on the photoelectrochemical properties than crystal structure and thickness. The geometric constraints are of course a key factor, as the photocurrent scales linear with the surface area. To achieve a maximum efficiency, the semiconductor-to-electrolyte interface has to be as large as possible. This leads to the requirement of a rough or porous surface structure for the anode material. For the region where the global maximum in photocurrent density was identified, a SEM micrograph of higher magnification is presented in Figure 6. The morphology shows longish spiky grains with a size of 50–150 nm. Unlike the amorphous and partially the ternary phase with their flat and dense surfaces, the binary regions contain cavities, which can be seen on the higher magnification micrograph. Although the
1283
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers be interesting candidates for solar water-splitting photoanodes.
Experimental Section Synthesis of thin film Fe¢W¢O MLs
Figure 6. SEM micrograph taken in the region of highest photocurrent (ML3 at 15.2 at % Fe, thickness 1060 nm), after PEC measurement. The sample morphology is identical to samples with the same composition but different thickness as shown in Figure 5.
absolute or relative surface areas have not been quantified, it is clear to see that the W-rich part and with this the binary phases show a much higher specific surface area than the amorphous and ternary parts of the MLs. Despite the effect of an increased photocurrent for a larger semiconductor-to-electrolyte interface being undisputed, a wide range of compositions of all MLs have similar surface morphologies but significantly different photocurrent densities. This again points to the importance of the crystal structure, respectively the degree of crystallinity and the film thickness for the photocatalytic activity.
The Fe¢W¢O thin-film MLs were fabricated by co-deposition on 100 mm (4 in; 1 in = 25.4 mm) diameter Si/SiO2 (525 mm/1.5 mm) wafers using reactive magnetron sputtering from 4 in elemental metallic targets in an AJA ATC2200V sputter system. The sputter configuration was chosen to obtain MLs with composition- and thickness gradients almost perpendicular to each other (Figure 1). Thus, regions of similar chemical composition but different thickness and vice versa were obtained. The MLs were segmented in measurement areas of 4.5 mm Õ 4.5 mm, thus defining 342 measurement areas per ML. Fe (purity 99.99 %) was sputtered using pulsed direct current (DCp) magnetron sputtering with a frequency f = 35 kHz and a reverse time r = 1.2 ms. W (purity 99.99 %) was deposited by radio frequency (RF) magnetron sputtering. The MLs were deposited at a substrate temperature of 400 8C. Due to the use of a SiC-Inconel static heater, the temperature variation of the substrate was negligible. The base pressure was < 1.3 Õ 10¢4 Pa for all depositions. The gas pressure was 1.33 Pa with an Ar/O2 ratio of 1:3. The deposition times and powers as well as the composition and thickness ranges for the three MLs are listed in Table 1. The back electrode was sputtered prior to the MLs and consisted of a 10 nm Ti adhesion layer and a 100 nm Pt film sputtered at room temperature [250 W RF (Pt), 200 W RF (Ti)] and an Ar pressure of 0.67 Pa.
Table 1. Overview of the composition- and thickness ranges, and the deposition parameters for sputter deposition of the Fe¢W¢O material libraries. Material Binary Fe Thickness Deposition power Deposition time library content [at %] [nm] W (RF)/Fe (DC-p) [W] [min]
Conclusions The combinatorial synthesis and high-throughput screening of a large part of the Fe¢W¢O system, using more than 1000 different measurement areas located on three material libraries (MLs), is presented. Each measurement area is characterized in terms of chemical composition, thickness, crystal structure, photocurrent density at a constant bias potential [1.475 V vs. reversible hydrogen electrode (RHE)], and for selected measurement areas also in terms of morphology. The screening revealed correlations between increased photocurrent density and crystallinity, especially for the phases WO3 and W5O14 and less distinct but still apparent for the Fe2O6W phase. The photocurrent density increases with an increase in the degree of crystallinity (here: increased peak intensity in XRD). The thickness was confirmed to be an important factor. Measurement areas of identical crystal structure and composition show thickness-dependent maxima in photocurrent density. The morphology was also important, but its influence was not as strong as that of crystallinity and thickness. Only for measurement areas of similar thickness and crystal phase but different morphology a strong impact of the latter could be confirmed. The phases W5O14 and Fe2O6W should be studied in detail, as they might ChemSusChem 2015, 8, 1279 – 1285
www.chemsuschem.org
ML1 ML2 ML3
6–34 17–61 15–56
220–800 220–730 360–1160
460/70 330/140 330/140
214 250 400
High-throughput characterization of thin-film MLs Chemical compositions were determined by automated energy dispersive X-ray analysis (EDX) using an INCA X-act detector (Oxford Instruments) attached to a JEOL 5800 SEM operating at 20 kV. The compositional data was measured in the center (approx. 400 mm Õ 600 mm) of each measurement area. When using a Si-drift-detector (SDD), quantification of very light (O) and very heavy (W) elements at the same time is not accurate. Therefore, the presented compositional data are always provided for binary Fe/W ratios with an accuracy of about 0.5 at %. Due to the deposition conditions, with high O2 excess at elevated temperature, the investigated materials were expected to be stoichiometric. EDX elemental mappings in selected measurement areas were performed using an EDAX system (APOLLO XV SDD) attached to a Carl Zeiss LEO Supra 55 FEG SEM. Using the latter, SEM micrographs were recorded. The film thicknesses were measured using a second set of samples fabricated at identical parameters except for the deposition temperature. The control samples were deposited at RT to allow the use of photoresist patterning. The film thickness was subsequently
1284
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers measured by tactile profilometry using an Ambios XP2 profilometer. X-ray diffraction (XRD) patterns were recorded in Bragg–Brentano geometry by using a PANalytical X‘Pert PRO system equipped with a PIXcel detector and CuKa radiation. The area illuminated by the X-ray beam (2q dependent) was of elliptic shape with a size of about 3 mm Õ 3 mm. The identification of the phases was based on data from the “Pauling File” database (WO3—S1250995, W5O14— S1251922) and “Pearson’s Crystal Data” database (Fe2O6W— 1402895), respectively. Measurement of the photoelectrochemical (PEC) properties was carried out using a custom-designed automated OSDC.[18] The measurements were performed in a three-electrode configuration using a Pt wire as counter electrode and a Ag/AgCl 3 m KCl reference electrode. The selected measurement area was connected as working electrode. The surface area of the working electrode was 0.785 mm2. Measurement areas screened for PEC performance had the same locations as those analyzed by EDX and profilometry to allow for correlation of composition, thickness, and PEC performance. The electrolyte was a 0.5 m NaClO4 aqueous solution (pH 4.5). An optical-fiber-guided 150 W Xe lamp was used as light source. By using an aperture, the effective light intensity illuminating the sample surface was set to 0.785 mW. This corresponded to a normalized intensity of 100 mW cm¢2. A bias potential of 1 V versus Ag/AgCl 3 m KCl (1.475 V vs. RHE) was applied, and the steady-state current was measured in the dark and under illumination. The dark currents were < 100 nA cm¢2 for all measurement areas. The photocurrents were subsequently calculated as the difference between the steady-state current under illumination and in the dark.
Acknowledgements
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]
[13]
[14] [15]
[16] [17] [18]
This work was in part funded by DFG within the Priority Program SPP 1613, “Fuels Produced Regeneratively Through Light-Driven Water Splitting: Clarification of the Elemental Processes Involved and Prospects for Implementation in Technological Concepts” (LU1175/10-1, Schu929/12-1). K.S. acknowledges a PhD fellowship by the International Max Planck Research School for Surface and Interface Engineering (IMPRS-SurMat). We thank C. Long and I. Takeuchi for providing the XRDsuite software.
ChemSusChem 2015, 8, 1279 – 1285
Keywords: combinatorial chemistry · energy conversion · high-throughput screening · photoelectrochemistry · thin films
www.chemsuschem.org
[19] [20] [21] [22] [23]
M. Grtzel, Nature 2001, 414, 338. A. Fujishima, K. Honda, Nature 1972, 238, 37. M. A. Butler, J. Appl. Phys. 1977, 48, 1914. R. K. Quinn, R. D. Nasby, R. J. Baughman, Mater. Res. Bull. 1976, 11, 1011. H. P. Maruska, A. K. Ghosh, Sol. Energy 1978, 20, 443. M. A. Gondal, A. Hameed, Z. H. Yamani, A. Suwaiyan, Chem. Phys. Lett. 2004, 385, 111. J. Georgieva, E. Valova, S. Armyanov, N. Philippidis, I. Poulios, S. Sotiropoulos, J. Hazard. Mater. 2012, 211, 30. E. Thimsen, S. Biswas, C. S. Lo, R. Biswas, J. Phys. Chem. C 2009, 113, 2014. V. M. Aroutiounian, V. M. Arakelyan, G. E. Shahnazaryan, G. M. Stepanyan, J. A. Turner, O. Khaselev, Int. J. Hydrogen Energy 2002, 27, 33. X. D. Xiang, X. Sun, G. Briceno, Y. Lou, K. A. Wang, H. Chang, W. G. Wallace-Freedman, S. W. Chen, P. G. Schulz, Science 1995, 268, 1738. E. Danielson, J. H. Golden, E. McFarland, C. M. Reaves, W. H. Weinberg, X. D. Wu, Nature 1997, 389, 944. J. Cui, Y. Chu, O. O. Famodu, Y. Furuya, J. Hattrick-Simpers, R. D. James, A. Ludwig, S. Thienhaus, M. Wuttig, Z. Zhang, I. Takeuchi, Nat. Mater. 2006, 5, 286. S. Hamann, M. E. Gruner, S. Irsen, J. Buschbeck, C. Bechtold, I. Kock, S. G. Mayr, A. Savan, S. Thienhaus, E. Quandt, S. Fhler, P. Entel, A. Ludwig, Acta Mater. 2010, 58, 5949. M. Woodhouse, B. A. Parkinson, Chem. Soc. Rev. 2009, 38, 197. Z. Huang, Y. Lin, X. Xiang, W. Rodriguez-Cordoba, K. J. McDonald, K. S. Hagen, K.-S. Choi, B. S. Brunschwig, D. G. Musaev, C. L. Hill, D. Wang, T. Lian, Energy Environ. Sci. 2012, 5, 8923. M. Barroso, S. R. Pendlebury, A. J. Cowan, J. R. Durrant, Chem. Sci. 2013, 4, 2724. S. Curtarolo, G. L. W. Hart, M. B. Nardelli, N. Mingo, S. Sanvito, O. Levy, Nat. Mater. 2013, 12, 191. K. Sliozberg, D. Schfer, T. Erichsen, R. Meyer, C. Khare, A. Ludwig, W. Schuhmann, ChemSusChem 2015, DOI: 10.1002/cssc.201402917. C. V. Thompson, Annu. Rev. Mater. Sci. 2000, 30, 159. R. Zarnetta, P. J. S. Buenconsejo, A. Savan, S. Thienhaus, A. Ludwig, Intermetallics 2012, 26, 98. I. J. McColm, R. Steadman, S. J. Wilson, J. Solid State Chem. 1978, 23, 33. J. P. Kollender, A. I. Mardare, A. W. Hassel, ACS Comb. Sci. 2013, 15, 601. V. S. Vidyarthi, M. Hoffmann, A. Savan, K. Sliozberg, D. Kçnig, R. Beranek, W. Schuhmann, A. Ludwig, Int. J. Hydrogen Energy 2011, 36, 4724.
Received: August 29, 2014 Published online on February 27, 2015
1285
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers
High-Throughput Screening of Thin-Film Semiconductor Material Libraries II: Characterization of Fe¢W¢O Libraries Robert Meyer,[a] Kirill Sliozberg,[b] Chinmay Khare,[a] Wolfgang Schuhmann,*[b, c] and Alfred Ludwig*[a, c] Metal oxides are promising materials for solar water splitting. To identify suitable materials within the ternary system Fe¢W¢ O, thin-film material libraries with combined thickness and compositional gradients were synthesized by combinatorial reactive magnetron sputtering. These libraries (> 1000 different samples) were investigated by means of structural and functional high-throughput characterization techniques to establish correlations between composition, crystallinity, morphology,
thickness, and photocurrent density in the compositional range between (Fe6W94)Ox and (Fe61W39)Ox. In addition to the well-known phase WO3, the binary phase W5O14 and the ternary phase Fe2O6W show enhanced photoelectrochemical activity. The highest photocurrent density of 65 mA cm¢2 was achieved for the composition (Fe15W85)Ox, which contains the W5O14 phase and has a thickness of 1060 nm.
Introduction The conversion of renewable energy sources into technically usable forms of energy is a key factor for a sustainable economy in the future. A promising technology is solar water splitting (SWS) in a photoelectrochemical cell to convert solar energy directly to chemical energy in form of molecular hydrogen (and oxygen).[1] Possible configurations of a photoelectrochemical cell are the combination of n-type semiconductor photoanodes with p-type semiconductor photocathodes or metal counter electrodes immersed in an aqueous electrolyte. Many candidates for anode materials have been identified, most of them being metal-oxide semiconductors such as TiO2, WO3, and Fe2O3.[2–6] These materials are matching several necessary criteria, such as high abundance and thereby a relatively low price, high stability in aqueous media, and non-toxicity. However, these single-metal-oxide semiconductors do not exhibit the necessary performance for application in SWS devices. Recent research, therefore, focused on multinary-metal-oxide semiconductors, many of which are based on TiO2, WO3, or Fe2O3, either mixed with each other or doped with single elements such as Ta.[7–9]
[a] R. Meyer, Dr. C. Khare, Prof. Dr. A. Ludwig Institute for Materials Ruhr-University Bochum Universittsstraße 150, 44801 Bochum (Germany) E-mail: [email protected] [b] K. Sliozberg, Prof. Dr. W. Schuhmann Analytical Chemistry—Center for Electrochemical Sciences (CES) Ruhr-University Bochum Universittsstraße 150, 44801 Bochum (Germany) E-mail: [email protected] [c] Prof. Dr. W. Schuhmann, Prof. Dr. A. Ludwig Materials Research Department Ruhr-University Bochum Universittsstr. 150, 44801 Bochum
ChemSusChem 2015, 8, 1279 – 1285
An effective method to identify suitable materials for SWS is seen in the thin-film combinatorial material science approach.[10–13] Instead of analyzing only a single point in a huge parameter space, the combinatorial approach provides a large number of different but very well defined and well comparable samples in a material library (ML). Using specially designed high-throughput characterization tools, this method has the potential to accelerate the development of thin-film-based SWS systems as compared to the “one by one” approach.[14] Herein, the thin-film combinatorial approach was applied to investigate a large fraction of the Fe¢W¢O system. The binary systems Fe¢O and W¢O are well studied materials for SWS photoanodes, but both have disadvantages such as a high recombination rate (a-Fe2O3) or a limited width of the absorption spectrum (WO3).[15, 16] By combining Fe¢O and W¢O in the Fe¢ W¢O system, these disadvantages might be overcome. By changing the short- and long-range order as a function of composition, the electronic structure might improve towards a higher incident photon-to-current conversion efficiency (IPCE). The altered electronic structure is hard to predict theoretically by means of density functional theory.[8, 17] However, analyzing well-comparable samples from MLs over a wide range of compositions by means of photoelectrochemical screening may provide both qualitative and quantitative trends that provide precise information for further in-depth analysis of hits obtained during screening. Furthermore, new ternary oxide phases exhibiting interesting photoelectrochemical properties could be discovered. To identify correlations between composition, crystallinity, morphology, thickness, and photocurrent density, ternary Fe¢W¢O thin-film MLs were deposited by reactive magnetron sputtering. The MLs were designed such that they consisted of well-defined compositions and thickness gradients (Figure 1).
1279
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers
Figure 1. Schematic of the combinatorial sputter setup to achieve combined composition and thickness gradients on a material library in top view (a) and side view (b).
Results and Discussion To screen for potential SWS materials within the W-rich part of the Fe¢W¢O system, three MLs (ML1, ML2, and ML3) were fabricated. On each ML 342 measurement areas were defined, adding up to 1026 samples ranging from 6 to 61 at % Fe (binary) with thicknesses ranging from 220 to 1160 nm. The photocurrent density, as a generally accepted screening parameter, was used to evaluate the properties of each measurement area on the MLs. To screen for interesting materials, an in-house-developed, fully automated optical scanning droplet cell (OSDC) setup was used, which is described elsewhere.[18] XRD mapping revealed the existence of three different crystalline phases and an amorphous region in the investigated Fe¢ W¢O composition range. These regions show also a thickness dependence, which is not unusual for sputtered thin films.[19] Due to the large amount of obtained data, XRD data are presented in complementary ways. Figure 2 shows discrete q/ 2q scans of ten measurement areas for each ML recorded along the compositional gradient through the center of each ML. In Figure 3, color-coded intensity maps of characteristic major XRD peaks are presented.[20] These can be directly compared to the color-coded chemical composition, photocurrent density, and the thickness maps of the MLs. For samples with Fe contents < 11 at % only the monoclinic WO3 phase is present, showing reflections of (002), (020), (111), (200), and (220) orientation (Figure 2 a). From about 11 at % Fe onwards, these peaks disappear and the main phase changes to W5O14, which has a distinct (001) reflection. This Magn¦li phase was reported by McColm et al. to be a tetragonal (tp152) structure with an oxygen-to-metal ratio of 2.8, which is stabilized by the exisChemSusChem 2015, 8, 1279 – 1285
www.chemsuschem.org
Figure 2. XRD patterns of selected measurement areas of the thin-film MLS a) ML1, b) ML2, and c) ML3, with the corresponding Fe content (binary) and thickness. Originating from the W-filament in the XRD Cu cathode, the W-La peaks appear for the high-intensity peaks of Pt (111) (from the Pt back electrode) and W5O14 (001) and are visible because of the use of the very sensitive PIXcel detector. The same applies to the Kb peaks. The chemical compositions were measured by means of EDX in the center of each measurement area and, therefore, present an average for the area measured by XRD.
tence of small amounts of Fe atoms.[21] Although the amount of Fe is much lower for the reported bulk diffusion samples than for the sputter-deposited samples, the XRD data clearly point to the existence of the W5O14 phase, especially for sam-
1280
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers ples from ML3 (Figure 2 c). Nevertheless the WO3 and the W5O14 phases are likely to co-exist over a large compositional range, where the more pronounced XRD signals are covering the less pronounced signals of the other phase, especially in the range between 11 and 15 at % Fe. Starting at about 20 at % Fe, the W5O14 peak intensity is decreasing again and disappears at around 27 –29 at % Fe. As the measurement areas do not show any phase separation, it is most likely that W atoms are substituted by Fe atoms in the crystal lattice and that the number of substituted W atoms is increasing with increasing Fe content. Yet it is possible that the non-stoichiometry is compensated by amorphous fractions in the film and no significant substitution takes place. At about 44 at % Fe, a ternary crystal phase appears, which was identified as Fe2O6W, showing a (200) reflection peak. The intensity of this peak is increasing with increasing Fe content up to the maximum Fe content of 61 at % and accompanied by the Fe2O6W (061) peak starting from around 49 at % Fe. Unfortunately, the (200) peak is located in the left shoulder of the high-intensity Pt (111) peak of the back electrode. For high intensities, the Fe2O6W (200) signal is clearly distinguishable, but the Pt peak might cover low-intensity peaks at lower Fe contents (Figure 2 b and c). In contrast to the binary phases, for the ternary phase it is most likely that Fe atoms are substituted by W atoms as the Fe-to-W ratio is lower than the stoichiometric ratio of 2:1 in Fe2O6W. Hence, the number of substituted W atoms is decreasing with increasing Fe content. But, as for the W-rich measurement areas, it is possible that the crystalline parts of the film are indeed stoichiometric whereas excess atoms are present in an amorphous phase. Measurement areas with Fe contents between 29 and 44 at % Fe are at least X-ray amorphous for all MLs (thickness depending). As aforementioned for identical compositions, the appearance, intensity, and fading of the observed peaks varies slightly as a function of film thickness for all crystalline and amorphous phases. To understand these variations, three effects have to be considered: 1) With increasing thickness the volume increases, which leads to a higher XRD signal intensity. Thus, crystal phases with low-intensity peaks are easier to detect and to identify in thicker films. 2) The composition is measured over an area of 400 mm Õ 600 mm [energy-dispersive X-ray analysis (EDX)], whereas the phase information (XRD) is averaged over an area of up to 3 mm Õ 3 mm. 3) The longer deposition time for ML3 (60 % longer than for ML1 or ML2) leads to about the same amount of additional thermal power applied to the growing film. This introduces more energy and gives more time for an improved short- and long-range order in the atomic structure and, therefore, higher peak intensities.
Figure 3. Logarithmic color maps of the relative intensities of the main XRD peaks for a) ML1, b) ML2, and c) ML3 (top row). For a direct comparison, the logarithmic intensity maps of the photocurrent density (center), the absolute thickness (right bottom), and the Fe content (left bottom) are shown side by side. The photocurrent density color map is according to Figure 4. Each color map shows a ML on a 100 mm diameter substrate.
ChemSusChem 2015, 8, 1279 – 1285
www.chemsuschem.org
1281
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers The correlations of the measured photocurrent values with thickness, composition and crystal structure are discernible from Figure 4. Taking into account the crystal structure information, the diagram is divided into five sections by dashed black lines and labeled with roman numbers (I–V). The photocurrent density data is presented in two complementary ways:
thicknesses between 700 and 800 nm on ML1. The local maximum for the ternary Fe2O6W phase (V) is located at about 51 at % Fe and a thickness of 345 nm. The absolute value of 15 mA cm¢2 is low compared to the photocurrents of the binary phase. Within the Fe2O6W region, ML2 and ML3 show a deviation in photocurrent density. Although the trend of increasing photocurrent density for increasing crystallinity (starting at 44 at % Fe) is consistent, for ML3 the absolute photocurrent densities are lower than for ML2 for measurement areas with similar composition and thickness. Furthermore, the measurement areas on ML3 show an increase in photocurrent density for increasing thickness, whereas this is not apparent in the same ternary region for ML2. These differences are at least partially related to the morphology of the MLs along the compositional gradient, which is shown in Figure 5. The SEM images are obtained from measurement areas in accordance with the positions of the XRD results in Figure 2, whereas the XRD patterns are only displayed for every other measurement area. Distinct changes in morphology may be related to a change in the crystal phase (c.f. Figure 3). For most of the identical compositions, the morphology of all MLs is identical independent of thickness except for the ternary Fe2O6W phase > 44 at % Fe (ML2 and ML3). ML3 Figure 4. Color-coded map showing the correlations between thickness, Fe-content, and measured photocurrent density acquired for 1026 measurement areas from three MLs. shows a relatively smooth and homogeneous surface The map is divided in five phase regions (I–V) by black dashed lines, indicating the Fe with only few inclusions of bigger grains. The avercontent and thickness for which the measurement areas exhibit the according phase. age grain size, as determined from the SEM microGrey dotted lines trace the measurement areas for which SEM micrographs were recordgraphs, is about 90 nm for 44 at % Fe and increases ed (Figure 5 ), with the start/end points encircled. to 100 nm at 52 at %. ML2, however, shows a morphology consisting of two grain types for composiIn Figure 3 it is plotted as a function of the position on the ML, tions equal or higher than 49 at % Fe. The film consists of large whereas in Figure 4 the photocurrent density is shown colorcrystallites (180–250 nm) embedded in a matrix of smaller coded as a function of composition and thickness for all meagrains (50–70 nm). The number of larger grains decreases with surement areas. The color map scales are identical for Figures 3 increasing Fe content. An EDX intensity map reveals no differand 4. In general, photocurrent densities of measurement ence in composition for the large grains and the surrounding areas with similar composition and thickness exhibit comparamatrix. ble values and show the same trends for all MLs. In a recent To account for the difference in photocurrent density, two paper by Kollender et al., data on thermally evaporated and main effects have to be considered. The first is the difference annealed Fe¢W¢O thin films were published.[22] Contrary to in thickness. Measurement areas of ML3 are about 200 nm thicker for identical compositions (SEM micrographs in our MLs obtained by reactive co-sputtering, which are mostly Figure 5). The second is that the deposition time for ML3 was single phase (except for the WO3 + W5O14 part), the evaporated 60 % longer, giving an equivalent of 60 % more thermal power and annealed films consisted of multiple phases over the applied during growth. Consequently, the morphological differwhole compositional range. The photocurrents are in the same ences might lead to higher photocurrent densities due to an order of magnitude. However, the influence of film thickness increased specific surface area for ML3, although the difference on photocurrents could not be taken into account, precluding in photocurrent seems to be larger than the increase in surface a more detailed comparison of the data. area. According to Figure 4, two compositional regions with relaFor the binary W¢O phases, the influence of the microstructively high photocurrents were identified: the binary W¢O ture on the photocurrent density is minor compared to the phases < 30 at % Fe and the ternary phase > 44 at % Fe. The “degree of crystallinity”, as shown by means of peak intensity amorphous region in between shows minimal photocurrents in XRD measurements. However, it has to be taken into considfor all MLs. Measurement areas within the W5O14 phase (III) diseration that the morphology and texture of sputtered thin play the highest photocurrent densities, with an absolute maxfilms are not independent although it is possible to influence imum of about 65 mA cm¢2 at around 15–17 at % Fe and thickmorphology and texture without changing the crystal phase, nesses ranging from 900 to 1060 nm (ML3). A local maximum for example, by altering deposition pressure or rate or the of about 49 mA cm¢2 is detected at 5–6 at % Fe (WO3 ; I) and ChemSusChem 2015, 8, 1279 – 1285
www.chemsuschem.org
1282
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers
Figure 5. SEM micrographs of selected measurement areas of a) ML1, b) ML2, and c) ML3. The binary Fe content is labeled top left, the thickness top right, the phase region (c.f. Figure 3) bottom left, and the photocurrent density (mA cm¢2) bottom right for the presented measurement areas. Each micrograph has a size of 2 mm Õ 2 mm and was recorded at a magnification of 30 kx.
overall film thickness. The influence of the thickness is not as distinct as the influence of the crystallinity but still apparent. Two main factors influencing the capability of a thin-film semiconductor absorber to convert radiation energy into electrical/chemical energy are the absorption (higher thickness favorable) and the charge-carrier transport, usually addressed as recombination rate or carrier lifetime (lower thickness favorable). Hence, an optimum thickness clearly exists and has been reported before, for example, for pure WO3.[23] Herein, Fedoped binary W¢O phases exhibit the highest photocurrent densities. For low Fe contents (6.4 at %) within the WO3 phase (ML1), the photocurrent density gradually increases until a maximum thickness is reached (765 nm). It cannot be excluded that the maximum photocurrent density for this phase is located at even higher thicknesses and lower Fe contents. The region of the globally highest photocurrents within the W5O14 (ML3) phase is located between 910 and 1060 nm thicknesses for Fe contents between 15 and 16 at %. As described above, no thickness-dependent maximum region of photocurrent density could be identified for the ternary Fe2O6W phase. For ChemSusChem 2015, 8, 1279 – 1285
www.chemsuschem.org
ML2, the higher photocurrent densities depend more on composition (45–55 at % Fe) than on thickness. For ML3, the photocurrent density steadily increases with increasing thickness but at low absolute values. Thus, the dependence on morphology might in this case be stronger than the thickness effect. Generally, it was found that the morphology has a weaker influence on the photoelectrochemical properties than crystal structure and thickness. The geometric constraints are of course a key factor, as the photocurrent scales linear with the surface area. To achieve a maximum efficiency, the semiconductor-to-electrolyte interface has to be as large as possible. This leads to the requirement of a rough or porous surface structure for the anode material. For the region where the global maximum in photocurrent density was identified, a SEM micrograph of higher magnification is presented in Figure 6. The morphology shows longish spiky grains with a size of 50–150 nm. Unlike the amorphous and partially the ternary phase with their flat and dense surfaces, the binary regions contain cavities, which can be seen on the higher magnification micrograph. Although the
1283
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers be interesting candidates for solar water-splitting photoanodes.
Experimental Section Synthesis of thin film Fe¢W¢O MLs
Figure 6. SEM micrograph taken in the region of highest photocurrent (ML3 at 15.2 at % Fe, thickness 1060 nm), after PEC measurement. The sample morphology is identical to samples with the same composition but different thickness as shown in Figure 5.
absolute or relative surface areas have not been quantified, it is clear to see that the W-rich part and with this the binary phases show a much higher specific surface area than the amorphous and ternary parts of the MLs. Despite the effect of an increased photocurrent for a larger semiconductor-to-electrolyte interface being undisputed, a wide range of compositions of all MLs have similar surface morphologies but significantly different photocurrent densities. This again points to the importance of the crystal structure, respectively the degree of crystallinity and the film thickness for the photocatalytic activity.
The Fe¢W¢O thin-film MLs were fabricated by co-deposition on 100 mm (4 in; 1 in = 25.4 mm) diameter Si/SiO2 (525 mm/1.5 mm) wafers using reactive magnetron sputtering from 4 in elemental metallic targets in an AJA ATC2200V sputter system. The sputter configuration was chosen to obtain MLs with composition- and thickness gradients almost perpendicular to each other (Figure 1). Thus, regions of similar chemical composition but different thickness and vice versa were obtained. The MLs were segmented in measurement areas of 4.5 mm Õ 4.5 mm, thus defining 342 measurement areas per ML. Fe (purity 99.99 %) was sputtered using pulsed direct current (DCp) magnetron sputtering with a frequency f = 35 kHz and a reverse time r = 1.2 ms. W (purity 99.99 %) was deposited by radio frequency (RF) magnetron sputtering. The MLs were deposited at a substrate temperature of 400 8C. Due to the use of a SiC-Inconel static heater, the temperature variation of the substrate was negligible. The base pressure was < 1.3 Õ 10¢4 Pa for all depositions. The gas pressure was 1.33 Pa with an Ar/O2 ratio of 1:3. The deposition times and powers as well as the composition and thickness ranges for the three MLs are listed in Table 1. The back electrode was sputtered prior to the MLs and consisted of a 10 nm Ti adhesion layer and a 100 nm Pt film sputtered at room temperature [250 W RF (Pt), 200 W RF (Ti)] and an Ar pressure of 0.67 Pa.
Table 1. Overview of the composition- and thickness ranges, and the deposition parameters for sputter deposition of the Fe¢W¢O material libraries. Material Binary Fe Thickness Deposition power Deposition time library content [at %] [nm] W (RF)/Fe (DC-p) [W] [min]
Conclusions The combinatorial synthesis and high-throughput screening of a large part of the Fe¢W¢O system, using more than 1000 different measurement areas located on three material libraries (MLs), is presented. Each measurement area is characterized in terms of chemical composition, thickness, crystal structure, photocurrent density at a constant bias potential [1.475 V vs. reversible hydrogen electrode (RHE)], and for selected measurement areas also in terms of morphology. The screening revealed correlations between increased photocurrent density and crystallinity, especially for the phases WO3 and W5O14 and less distinct but still apparent for the Fe2O6W phase. The photocurrent density increases with an increase in the degree of crystallinity (here: increased peak intensity in XRD). The thickness was confirmed to be an important factor. Measurement areas of identical crystal structure and composition show thickness-dependent maxima in photocurrent density. The morphology was also important, but its influence was not as strong as that of crystallinity and thickness. Only for measurement areas of similar thickness and crystal phase but different morphology a strong impact of the latter could be confirmed. The phases W5O14 and Fe2O6W should be studied in detail, as they might ChemSusChem 2015, 8, 1279 – 1285
www.chemsuschem.org
ML1 ML2 ML3
6–34 17–61 15–56
220–800 220–730 360–1160
460/70 330/140 330/140
214 250 400
High-throughput characterization of thin-film MLs Chemical compositions were determined by automated energy dispersive X-ray analysis (EDX) using an INCA X-act detector (Oxford Instruments) attached to a JEOL 5800 SEM operating at 20 kV. The compositional data was measured in the center (approx. 400 mm Õ 600 mm) of each measurement area. When using a Si-drift-detector (SDD), quantification of very light (O) and very heavy (W) elements at the same time is not accurate. Therefore, the presented compositional data are always provided for binary Fe/W ratios with an accuracy of about 0.5 at %. Due to the deposition conditions, with high O2 excess at elevated temperature, the investigated materials were expected to be stoichiometric. EDX elemental mappings in selected measurement areas were performed using an EDAX system (APOLLO XV SDD) attached to a Carl Zeiss LEO Supra 55 FEG SEM. Using the latter, SEM micrographs were recorded. The film thicknesses were measured using a second set of samples fabricated at identical parameters except for the deposition temperature. The control samples were deposited at RT to allow the use of photoresist patterning. The film thickness was subsequently
1284
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Papers measured by tactile profilometry using an Ambios XP2 profilometer. X-ray diffraction (XRD) patterns were recorded in Bragg–Brentano geometry by using a PANalytical X‘Pert PRO system equipped with a PIXcel detector and CuKa radiation. The area illuminated by the X-ray beam (2q dependent) was of elliptic shape with a size of about 3 mm Õ 3 mm. The identification of the phases was based on data from the “Pauling File” database (WO3—S1250995, W5O14— S1251922) and “Pearson’s Crystal Data” database (Fe2O6W— 1402895), respectively. Measurement of the photoelectrochemical (PEC) properties was carried out using a custom-designed automated OSDC.[18] The measurements were performed in a three-electrode configuration using a Pt wire as counter electrode and a Ag/AgCl 3 m KCl reference electrode. The selected measurement area was connected as working electrode. The surface area of the working electrode was 0.785 mm2. Measurement areas screened for PEC performance had the same locations as those analyzed by EDX and profilometry to allow for correlation of composition, thickness, and PEC performance. The electrolyte was a 0.5 m NaClO4 aqueous solution (pH 4.5). An optical-fiber-guided 150 W Xe lamp was used as light source. By using an aperture, the effective light intensity illuminating the sample surface was set to 0.785 mW. This corresponded to a normalized intensity of 100 mW cm¢2. A bias potential of 1 V versus Ag/AgCl 3 m KCl (1.475 V vs. RHE) was applied, and the steady-state current was measured in the dark and under illumination. The dark currents were < 100 nA cm¢2 for all measurement areas. The photocurrents were subsequently calculated as the difference between the steady-state current under illumination and in the dark.
Acknowledgements
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]
[13]
[14] [15]
[16] [17] [18]
This work was in part funded by DFG within the Priority Program SPP 1613, “Fuels Produced Regeneratively Through Light-Driven Water Splitting: Clarification of the Elemental Processes Involved and Prospects for Implementation in Technological Concepts” (LU1175/10-1, Schu929/12-1). K.S. acknowledges a PhD fellowship by the International Max Planck Research School for Surface and Interface Engineering (IMPRS-SurMat). We thank C. Long and I. Takeuchi for providing the XRDsuite software.
ChemSusChem 2015, 8, 1279 – 1285
Keywords: combinatorial chemistry · energy conversion · high-throughput screening · photoelectrochemistry · thin films
www.chemsuschem.org
[19] [20] [21] [22] [23]
M. Grtzel, Nature 2001, 414, 338. A. Fujishima, K. Honda, Nature 1972, 238, 37. M. A. Butler, J. Appl. Phys. 1977, 48, 1914. R. K. Quinn, R. D. Nasby, R. J. Baughman, Mater. Res. Bull. 1976, 11, 1011. H. P. Maruska, A. K. Ghosh, Sol. Energy 1978, 20, 443. M. A. Gondal, A. Hameed, Z. H. Yamani, A. Suwaiyan, Chem. Phys. Lett. 2004, 385, 111. J. Georgieva, E. Valova, S. Armyanov, N. Philippidis, I. Poulios, S. Sotiropoulos, J. Hazard. Mater. 2012, 211, 30. E. Thimsen, S. Biswas, C. S. Lo, R. Biswas, J. Phys. Chem. C 2009, 113, 2014. V. M. Aroutiounian, V. M. Arakelyan, G. E. Shahnazaryan, G. M. Stepanyan, J. A. Turner, O. Khaselev, Int. J. Hydrogen Energy 2002, 27, 33. X. D. Xiang, X. Sun, G. Briceno, Y. Lou, K. A. Wang, H. Chang, W. G. Wallace-Freedman, S. W. Chen, P. G. Schulz, Science 1995, 268, 1738. E. Danielson, J. H. Golden, E. McFarland, C. M. Reaves, W. H. Weinberg, X. D. Wu, Nature 1997, 389, 944. J. Cui, Y. Chu, O. O. Famodu, Y. Furuya, J. Hattrick-Simpers, R. D. James, A. Ludwig, S. Thienhaus, M. Wuttig, Z. Zhang, I. Takeuchi, Nat. Mater. 2006, 5, 286. S. Hamann, M. E. Gruner, S. Irsen, J. Buschbeck, C. Bechtold, I. Kock, S. G. Mayr, A. Savan, S. Thienhaus, E. Quandt, S. Fhler, P. Entel, A. Ludwig, Acta Mater. 2010, 58, 5949. M. Woodhouse, B. A. Parkinson, Chem. Soc. Rev. 2009, 38, 197. Z. Huang, Y. Lin, X. Xiang, W. Rodriguez-Cordoba, K. J. McDonald, K. S. Hagen, K.-S. Choi, B. S. Brunschwig, D. G. Musaev, C. L. Hill, D. Wang, T. Lian, Energy Environ. Sci. 2012, 5, 8923. M. Barroso, S. R. Pendlebury, A. J. Cowan, J. R. Durrant, Chem. Sci. 2013, 4, 2724. S. Curtarolo, G. L. W. Hart, M. B. Nardelli, N. Mingo, S. Sanvito, O. Levy, Nat. Mater. 2013, 12, 191. K. Sliozberg, D. Schfer, T. Erichsen, R. Meyer, C. Khare, A. Ludwig, W. Schuhmann, ChemSusChem 2015, DOI: 10.1002/cssc.201402917. C. V. Thompson, Annu. Rev. Mater. Sci. 2000, 30, 159. R. Zarnetta, P. J. S. Buenconsejo, A. Savan, S. Thienhaus, A. Ludwig, Intermetallics 2012, 26, 98. I. J. McColm, R. Steadman, S. J. Wilson, J. Solid State Chem. 1978, 23, 33. J. P. Kollender, A. I. Mardare, A. W. Hassel, ACS Comb. Sci. 2013, 15, 601. V. S. Vidyarthi, M. Hoffmann, A. Savan, K. Sliozberg, D. Kçnig, R. Beranek, W. Schuhmann, A. Ludwig, Int. J. Hydrogen Energy 2011, 36, 4724.
Received: August 29, 2014 Published online on February 27, 2015
1285
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
