Electronic Surface Level Positions of WO3 Thin Films for

Feb 5, 2008 - Table 1: Experimental Energetic Positions of VBM and CBM of WO3 Relative to EF, EVacuum, and on an NHE Energy Scalea ... indicate that a...
1 downloads 6 Views 168KB Size
3078

J. Phys. Chem. C 2008, 112, 3078-3082

Electronic Surface Level Positions of WO3 Thin Films for Photoelectrochemical Hydrogen Production L. Weinhardt,* M. Blum,† M. Ba1 r, and C. Heske Department of Chemistry, UniVersity of NeVada, Las Vegas, 4505 Maryland Parkway, Las Vegas, NeVada 89154-4003

B. Cole, B. Marsen, and E. L. Miller Hawaii Natural Energy Institute, School for Ocean, Earth Science, Technology, UniVersity of Hawaii, Manoa, Hawaii ReceiVed: October 15, 2007; In Final Form: December 2, 2007

Polycrystalline WO3 thin films for photoelectrochemical hydrogen production were investigated using photoelectron spectroscopy and inverse photoemission. First, we report on a careful study to minimize X-ray and electron beam-induced degradation. Second, we combined ultraviolet photoelectron spectroscopy and inverse photoemission to determine the surface positions of the valence and conduction band edges, respectively, and the work function (i.e., the position of the vacuum level). This allows us to paint a completely experimentbased picture of the WO3 surface level positions, which are of central relevance for the photoelectrochemical activity of such surfaces. We find the WO3 surface to be wide gap [3.28 ((0.14) eV] and n-type, with the conduction band minimum 0.39 ((0.10) eV above the Fermi level and 0.31 ((0.11) eV above the H+/H2 reduction potential. The valence band maximum is 2.89 ((0.10) eV below the Fermi level and 1.74 ((0.11) eV below the H2O/O2 oxidation potential.

Introduction WO3 has been widely studied as a material due to its relevance for various applications. Its excellent electrochromic properties make it a suitable material for displays (e.g., refs 1, 2), mirrors with variable reflectance (e.g., refs 3, 4), and smart windows (e.g., refs 5, 6). Furthermore, WO3 is commonly used as a material for gas sensors (e.g., refs 7, 8). Finally, WO3 has also been discussed as a photoanode material for photoelectrochemical (PEC) hydrogen production (e.g., refs 9-13). In such a photoelectrochemical cell, sunlight is absorbed in the photoanode, creating an electron-hole pair, which is then separated and used to split water (see, e.g., refs 14-18). Various deposition techniques have been used to deposit WO3 thin films. Among other techniques, RF-4,19 and DC-sputtering,4,20 screen-printing,21 and thermal evaporation22 were used. It was found that the optical and electronic properties of the films strongly depend on the actual stoichiometry. For example, slightly O-poor films exhibit a strong change in color and electrical properties as compared to stoichiometric films. The mechanisms of this stoichiometry dependence are discussed in detail in literature (e.g., refs 23-25). For the application as a photoanode in photoelectrochemical hydrogen production, one of the most crucial required properties of the material are suitable positions of the conduction band minimum (CBM) at the surface of the hydrogen electrode and of the valence band maximum (VBM) at the surface of the oxygen electrode, respectively. In most discussions of materials for PEC devices, one of the energy levels (the VBM for p-type systems or the CBM for n-type systems) is determined by * Corresponding author. E-mail: [email protected]. † Present address: Experimentelle Physik II, Universita ¨ t Wu¨rzburg, Am Hubland, 97074 Wu¨rzburg, Germany.

electrochemical methods that require specific sets of assumptions about the possibility to achieve flat band conditions. Moreover, the position of the other band edge (i.e., the CBM for p-type systems or the VBM for n-type systems) is generally inferred from optical bulk band gaps. However, in general, electronic band gaps and band edge positions at the surface (which are relevant for the interface formation with the electrolyte) are different from optical bulk gaps and bulk band edge positions (see, e.g., refs 26, 27 for detailed studies on the model system Cu(In,Ga)(S,Se)2). For a correct description, it is thus necessary to measure band edge positions and gaps directly with surfacesensitive techniques. In particular, WO3 is known to exhibit large changes of its properties as a function of small changes in composition, as discussed above. In this case, a direct experimental surface-sensitive determination of the valence and conduction band is indispensable. While various UV photoelectron spectroscopy (UPS) investigations of the valence band of WO3 can be found in literature (e.g., refs 22, 23, 28-30), no investigation of the conduction band is reported. In the present Article, we thus report the first direct measurement of conduction and valence band edges of polycrystalline WO3 thin film surfaces using UPS and inverse photoemission (IPES), allowing us to derive a completely experiment-based picture of the electronic surface structure. Experimental Section WO3 films were deposited onto both glass and TEC15 substrates (SnO2:F/glass) by reactive sputtering from a W target under an argon and oxygen ambient. The partial pressure of oxygen was kept at 2.1 mTorr to achieve fully stoichiometric films. The substrates were heated to 290 °C to ensure a high crystallinity for the films. The size of the crystallites was

10.1021/jp7100286 CCC: $40.75 © 2008 American Chemical Society Published on Web 02/05/2008

Electronic Surface Level Positions of WO2

J. Phys. Chem. C, Vol. 112, No. 8, 2008 3079

Figure 1. XPS survey spectrum of a WO3 surface after an exposure to Mg KR radiation for 5 min.

approximately 50 nm (a detailed discussion of crystallinity and morphology of the films can be found in ref 11). After deposition, great care was taken to minimize surface contaminations due to exposure to air. The samples were sealed under dry nitrogen conditions after film preparation at the University of Hawaii and immediately shipped to Las Vegas in an inert sample box equipped with desiccant. At UNLV the samples were unpacked in a glovebox under nitrogen atmosphere and directly introduced into the ultrahigh vacuum system used for the spectroscopic experiments. UPS measurements were performed using He I and He II excitation, and for the XPS measurements a Mg KR X-ray source (15 kV with 20 mA anode current) was used. The spectra were recorded with a PHOIBOS 150 electron analyzer with multichanneltron detector. For the IPES experiments, a low-energy electron gun (STAIB) and a Dose-type detector with SrF2 window and Ar:I2 filling were used. All experiments were performed in ultrahigh vacuum with a base pressure below 1 × 10-10 mbar. Results and Discussion Our XPS spectra indicate a very low amount of C adsorbates at the WO3 surface, as exemplarily shown by the survey spectrum in Figure 1. This can be attributed to the optimized shipping procedures as described above. Such an initially low surface contamination level is of critical importance for the study of WO3 surfaces, because conventional surface cleaning approaches such as Ar+ ion sputtering lead to strong changes in the spectra, as reported before by Dixon et al.24 Note that this is not only true for sputtering with high-energy ions; we also observed significant ion damage at energies as low as 50 eV. As will be discussed in detail below, the surface of the WO3 films is very sensitive to various different kinds of irradiation (ions, X-rays, UV light, and electrons). We have carefully monitored the changes induced by the excitation sources needed for our measurements and were thus able to minimize the experiment duration such that it was possible to measure spectra of an unaltered WO3 surface. In the next paragraph, we will first describe the beam-induced changes, followed by a discussion of the results obtained for the undamaged surface. Influence of the Excitation Sources on the WO3 Surface. In Figure 2 (left), the W 4f spectrum after increasing exposure time to non-monochromatized Mg KR radiation is shown. The first of the recorded spectra (exposure time about 1 min) shows a doublet of two symmetric peaks (W 4f5/2 at 37.92 ((0.02) eV and W 4f7/2 at 35.80 ((0.02) eV), which is representative

Figure 2. W 4f XPS spectra after increasing exposure time to Mg KR radiation (left), fit of the W 4f line after very short exposure (top right), and fit of the W 4f line after prolonged exposure (bottom right).

for W atoms with an oxidation state of +6 (as expected for the WO3 environment).21-25,28,31 A decomposition of this spectrum is shown in Figure 2 (top right) using a fit with linear background and two Voigt profiles with equal width, a (fitderived) spin-orbit splitting of 2.12 eV, and a peak ratio of 4:3. Upon further exposure to Mg KR radiation, we find strong changes of the spectra on the time scale of a few minutes. The two symmetric spin-orbit split lines decrease in intensity, and, in parallel, new spectral features emerge on both sides and between the W 4f doublet, indicating at least two additional W 4f components. A corresponding fit of the data is shown in Figure 2 (bottom right). The spin-orbit splitting and peak ratio was fixed at 2.12 eV and 4:3, respectively, and all peak widths were kept equal. A similar splitting of the W 4f line of WO3 has been reported, caused by various different treatments/ exposures such as sample heating in UHV at T > 50 °C,22,25 UV irradiation23,28 (low-pressure Hg discharge lamp), Ar+ ion treatment24 (1.5-2.0 keV), or exposure to X-rays31 (synchrotron radiation with hν ) 200 eV). It is generally agreed that the low binding energy component ((2) in Figure 2) is due to a loss of O at the surface and corresponds to W atoms in a +5 oxidation state.22-25,28,31 The high binding energy component (3) has been attributed either to surface defects21,31 or to a plasmon loss peak.28 In the case of the UPS measurements, we did not observe any changes in our spectra, independent of the used excitation line (He I or He II). This is illustrated in Figure 3a and b. Control W 4f spectra recorded after the first and last of the UPS spectra (not shown) corroborate the result that the irradiation of our UV source does not affect the WO3 surface. This seeming contrast to the observations of Fleisch et al.23 and Hollinger et al.,28 who found strong changes after UV-irradiation (so-called “UV-coloration”), can be explained by the much lower photon flux and the considerably higher photon energies used for our measurements in comparison to the UV-coloration experiments (He I at 21.2 eV and He II at 40.8 eV vs a low-pressure Hg discharge lamp in ref 28). For our inverse photoemission (IPES) experiments, the sample is irradiated with low-energy electrons (between 7 and 15 eV) at a current density of approximately 10 µA/cm2. As seen in Figure 3c, these electrons induce changes in the IPES spectra on the time scale of a few tens of minutes. By using the W 4f

3080 J. Phys. Chem. C, Vol. 112, No. 8, 2008

Weinhardt et al.

Figure 3. (a) He II and (b) He I UPS spectra as a function of exposure time to He II and He I, respectively. (c) IPES spectra after increasing exposure to low-energy electrons.

Figure 4. UPS and IPES spectra of an undamaged WO3 surface after minimized exposure. The linear extrapolations of the edges give the positions of the VBM and CBM. The surface is determined to be n-type with an electronic band gap of 3.28 ((0.14) eV.

line as a probe, the induced changes are found to be similar to the case of irradiation with X-rays (note that the probing XPS measurements were short enough not to induce considerable changes themselves). Electronic Surface Properties of the WO3 Surface. Taking the findings of the previous paragraphs into account, it was possible to derive an undisturbed electronic picture of the WO3 surface. This could be achieved by using a fresh film (prepared under the same conditions) for each of the experiments (XPS, UPS, and IPES) and keeping the total exposure (measurement) time short enough for the above-described changes to be negligible. Figure 4 shows the result of these efforts, where the UPS and IPES spectra of the WO3 surface are plotted on a common energy scale. We can now derive the positions of the valence band maximum (VBM) and conduction band minimum (CBM) relative to the Fermi level (EF) by a linear extrapolation of the leading edges in the respective spectra (for a justification of this procedure, see ref 32). We find the VBM at -2.89 ((0.10) eV and the CBM at 0.39 ((0.10) eV relative to EF, showing that the surface of the investigated WO3 films is n-type. Both numbers together give us a value of 3.28 ((0.14) eV for the surface band gap of the film. This value is significantly larger than the bulk value of 2.5-2.6 eV determined by optical

measurements of the investigated samples. Note that the observation of an increased band gap at the surface of a polycrystalline semiconductor film is not unusual; in the case of CuInSe2, for example, the observed electronic surface band gap of (solar cell) device-grade films (1.4 eV26) is significantly larger than the corresponding optical bulk band gap (1.0 eV). In the present case, both structural as well as compositional differences between bulk and surface could be responsible for the observed enhancement of the surface band gap. This will be the subject of further investigations. To utilize the derived positions of VBM and CBM for the design of a PEC device, we have to find their values relative to the vacuum level EVacuum or, as commonly used, relative to the normal hydrogen electrode (NHE). For this purpose, we use the absolute potential of the normal hydrogen electrode of -4.44 eV and the following expression (as recommended by the International Union of Pure and Applied Chemistry33):

-E(NHE) - 4.44 eV ) EVacuum ) EF + Φ

(1)

where Φ is the work function of the investigated material and EVacuum is the energy position of the vacuum level at the surface of the sample. Φ can be derived by measuring the secondary electron cutoff in the UPS spectra, as shown in Figure 5. We find a value of 4.49 ((0.05) eV. In literature, values between 4.3 eV (colorized films) and 4.9 eV are reported.34 To compare the experimentally determined edge positions with the H2O/O2 oxidation and H+/H2 reduction potentials, it is important to note that the latter are given for the standard state (i.e., pH ) 0). In contrast, the here-derived data correspond to the levels at the WO3 surface in vacuum. In a recent publication, Chun et al. assume that theses “vacuum values” correspond to the isoelectric point; that is, that they are equal to those in a solution with a pH for which the surface charge of the sample is zero.35 In most cases, however, the surface of a semiconductor in vacuum does exhibit a nonzero charge, inducing a surface band bending. In the present case of n-type WO3, the (upward) band bending has to be smaller than 0.39 ((0.10) eV (i.e., the distance between CBM and EF at the surface). An exact determination of the magnitude of the band bending could be achieved by determining the bulk Fermi level position within the band gap using Hall measurements. Alternatively, one could measure the surface photovoltage under illumination with a sufficiently high intensity to achieve flat-

Electronic Surface Level Positions of WO2

J. Phys. Chem. C, Vol. 112, No. 8, 2008 3081

Figure 5. Secondary electron cutoff measured by UPS. The work function (4.49 ((0.05) eV) can be derived by a linear extrapolation of the edge.

TABLE 1: Experimental Energetic Positions of VBM and CBM of WO3 Relative to EF, EVacuum, and on an NHE Energy Scalea VBM CBM H2O/O2 oxidation potential H+/H2 reduction potential

E rel. EF (eV)

E rel. EVacuum (eV)

E rel. NHE (eV)

-2.89 ((0.10) 0.39 ((0.10)

-7.38 ((0.11) -4.10 ((0.11) -5.67

2.97 ((0.11) -0.31 ((0.11) 1.23

-4.44

0

a The positions of the H2O/O2 and H+/H2 redox potentials are given for comparison. The values in the “E rel. NHE” column do not include a potential surface band bending in the vacuum-based experiments.

band conditions. In the framework of this Article, we will first correlate the vacuum measurements with the electrochemical energy scale by assuming a zero surface charge in vacuum (i.e., following the approach in ref 35). We then will discuss the numerical results in view of the surface-induced band bending discussed above. For this approach, we have corrected for the energetic difference of the electronic levels present between the isoelectric point and standard state (i.e., pH ) 0). For several metal oxides, it has been found that a change of the pH value will shift the levels by approximately 60 mV/pH36 due to surface interactions with the electrolyte. Because, for WO3, it was found that the isoelectric point corresponds to a pH value of 0.5,37 the correction for a direct comparison with the redox potentials is very small (i.e., 0.5 × 60 meV ) 30 meV). A further deliberate modification of the pH value of the electrolyte will shift the redox potentials by the same amount as the electronic levels (according to the Nernst equation). This effectively “couples” the relative electronic levels throughout the full pH range, and thus no further corrections due to pH variation are necessary. By using eq 1, we can now derive the positions of VBM and CBM rel. to the NHE, as summarized in Table 1 and Figure 6 including the 30 meV correction. For comparison, the relevant levels for the splitting of water are also included. As discussed, we need to take into account that the sample in vacuum might exhibit a possible surface band bending (less than 0.39 ((0.10) eV in our case). This correction would shift the CBM downward, that is, closer to the H+/H2 reduction potential. Semiconductor band-edge positions are critical parameters in determining the energetics of PEC water splitting devices.18

Figure 6. Positions of VBM and CBM of WO3 relative to EF, EVacuum, and the NHE. For comparison, the H2O/O2 oxidation and the H+/H2 reduction potentials are given. Gray bars for VBM and CBM indicate the error bars along the energy axis. The alignment of the “E rel. NHE (eV)” axis does not include a potential surface band bending in the vacuum-based experiments.

To provide holes for the oxidation of H2O to O2 and H+, the VBM of the photoanode (e.g., WO3) at the semiconductor surface has to be sufficiently lower than the H2O/O2 oxidation potential in the electrolyte solution. In the case of the WO3 films characterized in this study, the surface VBM on the vacuum energy scale lies 1.74 ((0.11) eV below the H2O/O2 oxidation potential at standard conditions, indicating favorable energetics for the water oxidation reaction. As an additional requirement for PEC water-splitting, the Fermi level of the counter-electrode surface, in this case the cathode, has to be higher than the H+/ H2 reduction potential to provide electrons for the reduction of H+ to H2. The energy levels of this backside cathode are defined by the Fermi level position (more correctly, the quasi-Fermi levels under illumination) in the bulk of the anode material. If we assume that (a) flat-band conditions can be achieved and (b) the bulk Fermi level lies close to the CBM, then for the device to work without additional bias, it is necessary (but not sufficient) for the CBM of the photoanode surface to be higher than the H+/H2 reduction potential. Under these assumptions, the measured level positions in Figure 6 at first sight suggest that a WO3 device might be able to work without additional bias. However, the CBM is at most 0.31 ((0.11) eV above the H+/H2 reduction potential. As discussed above, if a band bending is present at the surface in vacuum, this value can be smaller; in the most extreme case, it would be 0.08 eV below the reduction potential. This makes it very improbable to expect direct water splitting without bias. Furthermore, typical sunlight is not sufficiently intense to achieve flat-band conditions, and the need to draw a reasonable current from the device will inevitably incur overpotential and other system losses, further prohibiting unassisted PEC water splitting. Summary We have determined an entirely experiment-based description of the electronic valence and conduction band edge positions at the WO3 surface using UV and inverse photoemission. Changes of the surface upon irradiation with X-ray photons and low-energy electrons were found, monitored, and minimized for the determination of these electronic levels. By measuring the work function at the WO3 surface, the band edge positions could be related to an electrochemical energy scale and discussed on the basis of the redox potentials necessary for the splitting

3082 J. Phys. Chem. C, Vol. 112, No. 8, 2008 of water. We find the valence band maximum to be low enough to initiate the oxidation of H2O to O2 and H+. In contrast, the conduction band minimum is only high enough for the reduction of H+ to H2 at a backside metal cathode if assuming ideal conditions. Thus, the measurements indicate that an additional bias will be needed for an effective splitting of water using a WO3 photoanode, for example, by an additional solar cell in a tandem configuration. Acknowledgment. We acknowledge funding by the U.S. Department of Energy through subcontract #RF-05-SHGR-004 under grant no. DE-FG36-03GO13062. M.Ba¨. acknowledges support of the Deutsche Forschungsgemeinschaft through the Emmy Noether program. References and Notes (1) Masetti, E.; Grilli, M. L.; Dautzenberg, G.; Macrelli, G.; Adamik, M. Sol. Energy Mater. Sol. Cells 1999, 56, 259. (2) Wang, X. G.; Jang, Y. S.; Yang, N. H.; Wang, Y. M.; Yuan, L.; Pang, S. J. Sol. Energy Mater. Sol. Cells 2000, 63, 197. (3) Baucke, F. G. K. Sol. Energy Mater. 1987, 16, 67. (4) Franke, E. B.; Trimble, C. L.; Schubert, M.; Woollam, J. A.; Hale, J. S. Appl. Phys. Lett. 2000, 77, 930. (5) Lampert, C. M. Sol. Energy Mater. 1984, 11, 1. (6) Niklasson, G. A.; Granqvist, C. G. J. Mater. Chem. 2007, 17, 127. (7) Gyorgy, E.; Socol, G.; Mihailesciu, I. N.; Ducu, C.; Ciuca, S. J. Appl. Phys. 2005, 97, 093527. (8) Thiele, J. A.; da Cunha, M. P. Sens. Actuators, B: Chem. 2006, 113, 816. (9) Santato, C.; Ulmann, M.; Augustynski, J. J. Phys. Chem. B 2001, 105, 936. (10) Miller, E. L.; Marsen, B.; Cole, B.; Lum, M. Electrochem. SolidState Lett. 2006, 9, G248-250. (11) Marsen, B.; Miller, E.; Paluselli, D.; Rocheleau, R. Int. J. Hydrogen Energy 2007, 32, 3110-3115. (12) Marsen, B.; Cole, B.; Miller, E. L. Sol. Energy Mater. Sol. Cells 2007, 91, 1954-1958. (13) Park, J. H.; Park, O. O.; Kim, S. Appl. Phys. Lett. 2006, 89, 163106. (14) Fujishima, A.; Honda, K. Nature 1972, 238, 37. (15) Nozik, A. J.; Memming, R. J. Phys. Chem 1996, 100, 13061.

Weinhardt et al. (16) Graetzel, M. Nature 2001, 414, 338. (17) Khaselev, O.; Bansal, A.; Turner, J. A. Int. J. Hydrogen Energy 2001, 26, 127. (18) Bak, T.; Nowotny, J.; Rekas, M.; Sorrell, C. C. Int. J. Hydrogen Energy 2002, 27, 991. (19) Yamada, Y.; Tabataa, K.; Yashima, T. Sol. Energy Mater. Sol. Cells 2007, 91, 29. (20) DeVries, M. J.; Trimble, C.; Tiwald, T. E.; Thompson, D. W.; Woollam, J. A.; Hale, J. S. J. Vac. Sci. Technol., A 1999, 17, 2906. (21) Bittencourt, C.; Felten, A.; Mirabella, F.; Ivanov, P.; Llobet, E.; Silva, M. A. P.; Nunes, L. A. O.; Pireaux, J. J. J. Phys.: Condens. Matter 2005, 17, 6813. (22) Santucci, S.; Cantalini, C.; Crivellari, M.; Lozzi, L.; Ottaviano, L.; Passacantando, M. J. Vac. Sci. Technol., A 2000, 18, 1077. (23) Fleisch, T. H.; Mains, G. J. J. Chem. Phys. 1982, 76, 780. (24) Dixon, R. A.; Williams, J. J.; Morris, D.; Rebane, J.; Jones, F. H.; Egdell, R. G.; Downes, S. W. Surf. Sci. 1998, 399, 199. (25) Romanyuk, A.; Oelhafen, P. Sol. Energy Mater. Sol. Cells 2006, 90, 1945. (26) Morkel, M.; Weinhardt, L.; Lohmu¨ller, B.; Heske, C.; Umbach, E.; Riedl, W.; Zweigart, S.; Karg, F. Appl. Phys. Lett. 2001, 79, 4482. (27) Weinhardt, L.; Ba¨r, M.; Muffler, H.-J.; Fischer, Ch.-H.; Lux-Steiner, M. C.; Niesen, T. P.; Karg, F.; Gleim, Th.; Heske, C.; Umbach, E. Thin Solid Films 2003, 431-432, 272. (28) Hollinger, G.; Duc, T. M.; Deneuville, A. Phys. ReV. Lett. 1976, 37, 1564. (29) Pertosa, P.; Hollinger, G.; Michel-Calendini, F. M. Phys. ReV. B 1978, 18, 5177. (30) Bringans, R. D.; Ho¨chst, H.; Shanks, H. R. Phys. ReV. B 1981, 24, 3481. (31) Ottaviano, L.; Bussolotti, F.; Lozzi, L.; Passacantando, M.; La Rosa, S.; Santucci, S. Thin Solid Films 2003, 436, 9. (32) Gleim, Th.; Heske, C.; Umbach, E.; Schumacher, C.; Gundel, S.; Faschinger, W.; Fleszar, A.; Ammon, Ch.; Probst, M.; Steinru¨ck, H.-P. Surf. Sci. 2003, 531, 77. (33) Trasatti, S. Pure Appl. Chem. 1986, 58, 955. (34) Subrahmanyam, A.; Karuppasamy, A.; Kumar, C. S. Electrochem. Solid-State Lett. 2006, 9, H111. (35) Chun, W.-J.; Ishikawa, A.; Fujisawa, H.; Takata, T.; Kondo, J. N.; Hara, M.; Kawai, M.; Matsumoto, Y.; Domen, K. J. Phys. Chem. B 2003, 107, 1798. (36) Matsumoto, Y.; Yoshikawa, T.; Sato, E. J. Electrochem. Soc. 1989, 136, 1389. (37) Parks, G. A. Chem. ReV. 1965, 65, 177.