Surface Photovoltage Spectroscopy of Dye-Sensitized Solar Cells with

clear dependence of these onset wavelengths on the conduction band edge energies (ECB) of these oxides. This is manifested in a blue-shift for cells w...
1 downloads 0 Views 142KB Size
J. Phys. Chem. B 2001, 105, 6347-6352

6347

Surface Photovoltage Spectroscopy of Dye-Sensitized Solar Cells with TiO2, Nb2O5, and SrTiO3 Nanocrystalline Photoanodes: Indication for Electron Injection from Higher Excited Dye States Frank Lenzmann,† Jessica Krueger,† Shelly Burnside,† Keith Brooks,† Michael Gra1 tzel,*,† Doron Gal,‡ Sven Ru1 hle,‡ and David Cahen*,‡ Laboratory of Photonics and Interfaces, Swiss Federal Institute of Technology, 1015 Lausanne, Switzerland, and Department of Materials and Interfaces, Weizmann Institute of Science, RehoVoth, 76100, Israel ReceiVed: January 30, 2001; In Final Form: April 14, 2001

The onset wavelengths of the surface photovoltage (SPV) in dye-sensitized solar cells (DSSCs) with different mesoporous, wide-band gap electron conductor anode materials, viz., TiO2 (anatase), Nb2O5 (amorphous and crystalline), and SrTiO3, using the same Ru bis-bipyridyl dye for all experiments, are different. We find a clear dependence of these onset wavelengths on the conduction band edge energies (ECB) of these oxides. This is manifested in a blue-shift for cells with Nb2O5 and SrTiO3 compared to those with TiO2. The ECB levels of Nb2O5 and SrTiO3 are known to be some 200-250 meV closer to the vacuum level than that of our anatase films, while there is no significant difference between the optical absorption spectra of the dye on the various films. We, therefore, suggest that the blue shift is due to electron injection from excited-state dye levels above the LUMO into Nb2O5 and SrTiO3. Such injection comes about because, in contrast to what is the case for anatase, the LUMO of the adsorbed dye in the solution is below the ECB of these semiconductors, necessitating the involvement of higher vibrational and/or electronic levels of the dye, with the former being more likely than the latter. While for Nb2O5 hot electron injection has been proposed earlier, on the basis of flash photolysis experiments, this is the first evidence for such ballistic electron-transfer involving SrTiO3, a material very similar to anatase but with a significantly smaller electron affinity. Additional features in the SPV spectra of SrTiO3 and amorphous Nb2O5 (but not in those of crystalline Nb2O5) can be understood in terms of hole injection from the dye into the oxide via intraband gap surface states.

Introduction The mechanism of charge injection in dye-sensitized solar cells (DSSC’s) is an issue of considerable interest, both for improving the performance of this type of solar cell and because of basic issues of electron transfer across interfaces1 in semiconductors. Recently, Moser et al. reported data on the basis of which they suggested that hot electron injection occurs in cells with Nb2O5, rather than the generally used TiO2 (anatase), as the mesoporous electron conductor, into which the dye injects electrons.2 Earlier Ferrere and Gregg had found steady-state photophysical evidence for involvement of electron injection into TiO2 from higher lying excited states, using an Fe bipyridine complex.3 This followed earlier work of Damrauer et al.,4 who concluded, from femtosecond spectroscopy results on RuII(bipyridyl)3 complexes, that nonequilibrated excited states are of fundamental importance in the relaxation dynamics of such complexes. Building in part on earlier work of ours, where we compared the photovoltages of SrTiO3- with anatase-based cells;5 (see also below), we present here evidence for electron injection from higher excited dye states in such cells, using the voltage rather than the current spectral response. This has the advantage of extreme sensitivity to very weak absorbed light intensity, something that is particularly important when searching for effects in the absorption tail, as is the case here. The reason is that while the photocurrent, Iph, depends linearly on the absorbed † ‡

Swiss Federal Institute of Technology. Weizmann Institute of Science.

light intensity (at least up to light intensities approaching normal insolation, for most solar cells), the photovoltage depends on ln(Iph/Io), for Iph > Io. Here Io is the reverse saturation current or, more generally, the preexponential factor in the simple diode equation.6 Io is often in the nA range, which implies that Iph > Io already at very low light intensities. Because direct measurement of Iph is well below the signal-to-noise level of most photovoltaic systems, at those intensities the voltage spectrum is a much more sensitive tool than the current one. Except where noted, we use cells, made with films of mesoporous, nanocrystalline Nb2O5, SrTiO3, and TiO2. Comparing cells with Nb2O5 to TiO2 is complicated by the rather different crystal (and surface) structures of these two oxides and by the complicated crystal chemistry of Nb2O5 per se. SrTiO3 has a structure that is much closer to that of TiO2 and on the average half of its surface exposes the truncated TiO6 octahedra that make up all of the various anatase surfaces.7 In any photovoltaic cell the photovoltage is determined by the light-induced difference in Fermi level (EF) between two electrodes. In a DSSC these Fermi levels can be identified with the redox potential of the solution and the conduction band edge (ECB) of the electron-conducting semiconductor;8 (cf. Figure 1). According to this understanding, increasing the energy difference between these two levels will increase the photovoltage. Because for liquid-based DSSCs the redox electrolyte appears to be optimized fairly well, different electron conducting semiconductors have been tried to increase the photovoltage. Indeed, a rough correlation has been reported between flatband potential and open circuit voltage (Voc) for ZnO, anatase, SnO2, and WO3.9

10.1021/jp010380q CCC: $20.00 © 2001 American Chemical Society Published on Web 06/16/2001

6348 J. Phys. Chem. B, Vol. 105, No. 27, 2001

Lenzmann et al.

Figure 1. Schematic energy level diagram of the illuminated electrolyte/ nanocrystalline semiconductor system, illustrating the source of the surface photovoltage (SPV). The vertical axis is the one-electron energy, with the arrow directed toward the vacuum level. The horizontal arrow is a real space coordinate. EF,n and EF,p are the electron and hole quasiFermi levels, respectively, the latter of which corresponds to the redox potential of the electrolyte solution. CB/VB (ECB/EVB in the text) are the conduction and valence band edges, respectively. For the case illustrated the dye LUMO is above the semiconductor conduction band edge, i.e., the normal situation encountered in anatase TiO2.

Also when we used, in an otherwise identical DSSC, films of SrTiO3 instead of anatase TiO2 where SrTiO3 has a lower electron affinity (more negative ECB on the NHE scale) than anatase, we could show several 100 mV increase in Voc.5 Earlier5,8 we have argued that the conduction band edge (ECB) will give a good estimate for the (quasi)Fermi level position under illumination. If, by changing the semiconductor, the ECB will be closer to the vacuum level, Evac, than the energy of the dye’s lowest unoccupied molecular orbital, LUMO, (ELUMO), direct electron injection from the excited dye into the conduction band of the semiconductor will not be possible. Therefore, under those conditions no photovoltaic effect will be seen. If higher energy photons could excite electrons into dye energy levels from where injection is possible, this should lead to a photovoltaic effect, but with a blue shift in the spectral response. The reason for this is that now only photons with energy

hν g (ELUMO - EHOMO) + (ECB - ELUMO) ) (ECB - EHOMO) (1) (where HOMO stands for highest occupied molecular orbital), rather than photons with energy

hν g (ELUMO - EHOMO)

(2)

will be able to create electrons that can be injected into the CB. This mechanism is possible only if the electrons can be injected while in an excited vibronic state, above the LUMO. Most likely such states are vibrationally excited ones, in which case “hot” electrons are injected. As noted above such a mechanism should manifest itself as a blue shift in the spectrum. Unfortunately, it is difficult to use the commonly recorded spectral response (generally presented as incident {or absorbed} photon to electron conversion efficiency, IPCE {APCE}) to determine the expected blue shift. The reason is that in those spectra the signals at long wavelengths are very weak and often in the noise. Therefore, it is problematic to use such spectra to ascertain the absence or presence of electron injection from higher excited dye states. Experimental Section 1. Materials. (a) Nanocrystalline Membranes. Details on the preparations and crystallographic and microstructural charac-

terizations of the different mesoscopic, nm-sized thin films (membranes) used as electron conductors are given in the literature.5,10,11 Films with 5-6 µm geometric thickness (measured by profilometry, using an alpha step apparatus) were prepared on glass, covered with 300-400 nm of transparent conducting oxide, SnO2:F with a sheet resistance of 15 Ω/0 (TEC 15 from Libby-Owens-Ford). The electrodes consist of interconnected nm-sized particles with ∼10-20 nm diameter empty channels between them. The particle sizes are unimodal ∼15 nm for anatase, bimodal 10-15 nm and 40-60 nm diameter for SrTiO3, and have a relatively broad distribution centered around 30 nm for Nb2O5 (600 °C; cf. below) based on BET (inert gas adsorption) and scanning electron microscopy measurements. The SrTiO3 particles have the perovskite structure. The films of those particles have been described in detail in ref 5. The nature of the Nb2O5 films is strongly influenced by the annealing temperature. Unless otherwise noted we report on films annealed at 600 °C, which showed excellent (nano)crystallinity, corresponding to the orthorhombic T-structure. Up to 450 °C, the Nb2O5 films appeared amorphous.10 (b) Electrolyte. The electrolyte is a standard one for anatasebased DSSC’s. It had the following composition: 600 mM dimethylpropyl-imidazolium iodide, 50 mM I2, 500 mM tertbutylpyridine, and 100 mM LiI in methoxyacetonitrile. Similar electrolytes are described in ref 12. (c) Dye. The dye used in this work is the bis(tetrabutylammonium) salt of cis-di(thiocyanato)-N,N-bis(2,2′-bipyridyl-4,4′dicarboxylic acid)-ruthenium(II), called N719 hereafter. The synthesis and photo- and electrochemical properties of this compound are described in the literature.13 For staining the electrodes, the films were soaked in the dye solution for 15 h at room temperature (concentration: 3 × 10-4 M, solvent mixture: acetonitrile and tert-butyl alcohol in a volume ratio of 1:1). 2. Technique. Surface photovoltage (SPV) spectroscopy of photovoltaic cells essentially yields their open circuit spectral response (cf.14 for a review). We measure the SPV, using the one-contact Kelvin probe method, rather than the two contact configuration, known for characterizing photovoltaic cells.15,16 In the Kelvin probe method, a vibrating Au reference electrode, situated above the sample, serves as second electrode, and the contact potential difference (CPD) between it and the sample surface is measured. Changes in CPD reflect differences in work function. This allows measurement between the differences in Fermi level (EF) in equilibrium and the electron quasi-Fermi level (EFn) that reflects the electron distribution upon photoexcitation. This difference gives the photovoltage of the photoexcited surface. When this is done as a function of wavelength, the SPV spectrum (SPS) results. Hitherto it was only possible to use dry, solid/gas interface configurations, because of severe experimental problems encountered with liquid/solid interfaces (described, for example, in ref 14). However, recently Bastide et al. have shown how it is possible to measure samples in contact with a liquid.17 They did so by separating the sample and liquid from the vibrating reference electrode with a thin glass slide, making it into a container, using wax to seal it. Sample illumination is through the transparent container wall. Thanks to this extension we could measure the photovoltage of the liquid electrolyte/(dye+ semiconductor) junction that lies at the heart of a DSSC. Thus, we measured SPS of the different oxide membranes in a setup closely resembling a typical dye-sensitized solar cell. In all measurements the cells were illuminated from the front side, through the solution, as explained in ref 17. The accuracy for

Surface PhotoVoltage Spectroscopy

J. Phys. Chem. B, Vol. 105, No. 27, 2001 6349

Figure 2. Approximate one-electron energy schemes of different interfaces between polyiodide (I-/I3-) electrolyte and semiconductors onto which the dye N719 (for TiO2 anatase and ZrO2) or erythrosin B (for NiO) is adsorbed. All numbers are in eV. I-/I3- indicates the polyiodide redox potentials in organic electrolyte (for anatase and ZrO2, as given in the Experimental Section) and in aqueous solution for the NiO system. These and all the other numerical values are taken from refs 2, 8, 13, 20. Evac: energy level of electron in a vacuum; CB/VB (ECB/EVB in the text): bottom of the conduction band/top of valence band; HOMO/LUMO (EHOMO/ELUMO in the text): highest occupied/lowest unoccupied molecular orbital. Note that while the absolute values given here are subject to uncertainties due to the use of -5.0 and -4.5 eV as values for the work function of Au and the NHE on the solid-state scale, respectively, the relative values are not affected by those uncertainties.

the determination of the onset wavelength is estimated to be 15 nm or better, from experiments repeated several times with different electrodes (cf. Figure 4b). 3. Cell Assembly. The stained electrodes were assembled into liquid sandwich cells by covering the nanocrystalline films with a very thin glass sheet (Thomas Scientific, micro-cover glass slides #1, 0.16 mm thick), leaving a part of the conductive glass uncovered for contact. The cell was sealed with candlewax after allowing the electrolyte to wet the space between the film and the glass cover by capillary action. When carefully assembled, the wax provides a leak-free seal, well beyond the time of the measurement (cf.17 for details). In this setup the vibrating Au mesh is outside the liquid container. It acts as the counter-electrode of this DSSC-like arrangement. Results and Discussion 1. Relative Energy Level Configurations for Semiconductor/Dye Interfaces. There are several mechanisms by which the dye molecules, adsorbed on the surface of a nanocrystalline film, can lose the energy that they gained by photon absorption. Not considering destructive processes, the following processes are possible: i: thermal de-excitation ii: radiative de-excitation iii: electron or hole injection into the semiconductor, followed by regeneration of the ground state by hole or electron transfer from the electrolyte. Which of these mechanisms applies to a given (semiconductor+dye)/redox couple interface depends largely on the relative positions of the energy levels involved (see Figure 2). Electron injection is the well-established foundation of the dye-sensitized solar cell. The dye N719 acts as a sensitizer and after photoexcitation injects electrons into the conduction band of TiO2 (anatase), over its whole optical absorption spectrum with a quantum yield close to unity (Figure 2a).18 We verified experimentally (using also SPV) that if the same dye is adsorbed on ZrO2, a material with ECB at least 1 eV closer to Evac than that of fresh TiO2 anatase,2,19 the dye cannot inject electrons (Figure 2b). In that case the excited state decays by thermal or radiative de-excitation. A further possibility, i.e., hole injection from the dye (Figure 2c), was recently suggested as an explanation for results obtained with the p-NiO/erythrosin B system.20

As discussed in the Introduction there are intermediate cases, where ECB of the semiconductor is above the LUMO of the dye, but still has some overlap with higher lying vibrational (hot) states of the LUMO or higher electronically excited levels. The former possibility was suggested as an explanation for the excitation wavelength dependence of the injection quantum yield of the Nb2O5/N719 interface, obtained from laser flash photolysis experiments. The interpretation was as follows:2 Electron injection at this interface is possible only if the excitation wavelength is such, that excitation of the sensitizer to a state above the vibrational ground state (of the LUMO) can take place. In that case, the electron reaches a level above ECB of the solid. To explain the results obtained in ref 2, it was suggested there that ECB of Nb2O5 is 0.2-0.3 eV above (closer to Evac) that that of anatase. Such an energy level configuration is depicted schematically in Figure 3 for the case of SrTiO3, as compared to that of anatase, where we rely on our contact potential difference measurements in the light and in the dark for fresh anatase TiO2 and SrTiO3 surfaces. On the basis of the results from these, we find the following ECB values: ∼-4.4 and -4.2 eV (∼-3.9 and -3.7 eV) vs Evac, respectively. The numbers in parentheses are the estimated values in solution.8 The differences are based on electrochemical data and confirmed qualitatively by comparative CPD measurements (Ru¨hle et al., unpublished) and by comparison between gas-phase and liquid electrochemical data. The reason for the shift is polarization by solvent molecules.21 The work function values for anatase and SrTiO3, as measured by us by Kelvin Probe, are ∼-5.2 and -5.05 eV (∼-4.5 and -4.6 eV) vs Evac, respectively. The polyiodide redox potential in the organic solution used is at -4.85 eV, and the HOMO and LUMO levels of N719 are estimated to be at -5.45 and -3.85 eV below Evac.13 These values are calculated by taking the normal hydrogen electrode (NHE) to be at -4.5 eV vs Evac8 and using 1.6 eV as the HOMO-LUMO energy separation. This latter value, which is less than the 1.7 eV value found for the free dye, agrees with the ∼780 nm onset wavelength for SPV that we find for the anatase-based cell (cf. Figure 4). It is consistent with reported red-shifts (up to 0.2 eV) of the charge-transfer transitions of these dyes upon adsorption.13a 2. Surface Photovoltage (SPV) Spectroscopic Evidence for Hot Electron Injection. We measured the SPV spectra of

6350 J. Phys. Chem. B, Vol. 105, No. 27, 2001

Lenzmann et al.

Figure 3. Approximate one-electron energy schemes of electrolyte/ (dye + semiconductor) interface of DSSCs (symbols as in Figure 2; cf. also note on uncertainties in numerical values in legend to Figure 2). All numbers are in eV. Probable positions of surface states are indicated (grey squares), together with (im)possibility for electron transfer from them to the HOMO, if the latter is emptied upon photoexcitation. Left: Interface that allows electron injection from higher excited states. The numbers refer to SrTiO3 (cf. ref 5). The situation for Nb2O5 is similar. Its CB edge has been reported to be some 250 meV closer to Evac than that of anatase,2,25 in qualitative agreement with our own data (which suggest approximately 100 meV difference between Nb2O5 and TiO2 anatase in aqueous medium). Right: For comparison the anatase TiO2 system, shown in Figure 2, that allows normal electron injection, is indicated in this figure, too.

mesoporous, nanocrystalline TiO2, SrTiO3, and Nb2O5 semiconductor membranes as changes in the contact potential difference (CPD), in the liquid cell setup, described in the experimental part, using a Kelvin probe. Representative spectra are shown in Figure 4. The negative signal observed for TiO2 starting at around 780 nm (Figure 4) is due to electron injection from the dye into the conduction band of the TiO2. This electron injection leads to a photodoping effect, bringing about a change in EF which, under the nonequilibrium (illumination) conditions becomes EF,n, the quasi-Fermi level of the electrons (cf. Figure 1). The result is that under illumination (ECB - EF,n) < (ECB - EF) in the dark. In the present case this is experimentally measurable as a decrease of the work function and detected as a negative change in CPD.22,23 While for SrTiO3 and Nb2O5 we observe such signals as well, they are, however, significantly blue-shifted with respect to the TiO2 signal (35 and 50 nm, i.e., 0.08 and 0.11 eV, for Nb2O5 and SrTiO3, respectively). The optical absorbance spectra of N719 on the different membranes in the cells are shown in Figure 5. A priori, we might expect for Nb2O5 a red-shift for the long wavelength onset with respect to the anatase one, because of the higher acidity of the Nb2O5 surface and the known red-shift of N719 with decreasing pH.13 According to this reasoning, for SrTiO3 the opposite would be expected (with respect to anatase) because the SrO parts of the surface will be more basic than the TiO parts. However, the spectra, which were corrected for reflectance losses, do not show any significant differences, and definitely do not give any evidence for blue shifts comparable to the ones seen by SPV (Figure 5a). Therefore, we conclude that the observed blue shifts are not due to changes in the optical absorption of the dye, adsorbed on the different oxide surfaces. Rather, we propose that they are most likely due to electron injection from levels above the LUMO level, in accordance with the relative energy level configurations, shown in Figure 3.24 If we now take a fresh look at earlier reported IPCE data for SrTiO3- and anatase-based DSSCs5 (Figure 5b), we note a blue

Figure 4. (a) Comparison of the surface photovoltage (SPV ) ∆CPD; CPD ) contact potential difference) spectra of nanocrystalline TiO2, Nb2O5, and SrTiO3 electrodes in contact with the polyiodide electrolyte (see Experimental section). The decrease in voltage signal at short wavelengths is due mainly to the decreasing light intensity from the lamp.27 (The optical band gaps of the three oxides, 3.15-3.3 eV, should show up only e400 nm). (b) Demonstration of the accuracy of the determination of the onset wavelength of the SPV, showing the results for measurements on three different TiO2 and three different Nb2O5 (600 °C annealed) samples, each in their own liquid SPV cell.

shift for the SrTiO3-based cell as compared to the anatase-based one. Because of the sensitivity problem for the photocurrent spectrum at very low light intensities, noted in the Introduction, the data of Figure 5b, taken alone, could not be presented as evidence for a blue shift, because it could be argued that the signal/noise of the IPCE signal of the SrTiO3-based DSSC in the 800-740 nm wavelength range is too poor. However, taken together with the SPV spectrum with its superior signal/noise, we can see that the ∼50 nm shift, indicated by the IPCE spectra, is a real one. 3. Intraband Gap States. The SrTiO3-based DSSC yields a small SPV signal with sign opposite to that of the one for electron injection. This SPV signal appears around 780 nm, just before the onset of the electron injection signal (scanning from long to short wavelengths) and is absent for anatase TiO2- as well as for Nb2O5-based DSSCs. The sign of the signal suggests that it is caused by electrons leaVing the oxide membrane rather than entering it. Further investigations of cells made with Nb2O5 membranes, that were annealed at various temperatures, showed that as the crystallinity of the oxide film decreased, a similar positive SPV signal started to emerge. Figure 6 shows the SPV response of a DSSC using a near-amorphous Nb2O5 film (annealed at 500 °C10) and compares it with that of the cell made with a crystalline Nb2O5 film (i.e., annealed at 600 °C), shown also in Figure 4. An explanation for these observations may be found in the possible role of intraband gap states. Such states, which are due

Surface PhotoVoltage Spectroscopy

J. Phys. Chem. B, Vol. 105, No. 27, 2001 6351 measurements on dye-free membrane surfaces. These show evidence for two types of surface states, as deduced from VB f surface state (increased negative charge on surface, more negative work function) and surface state f CB (decreased negative charge on surface, less negative work function) transitions at roughly (ECB - 2.2) and (EVB + 1.8) eV, where EVB stands for the energy of an electron at the top of the valence band. In Figure 3 we show the approximate positions of the state that is closest to the CB (deduced from the VB f surface state transition). Especially for SrTiO3 electron injection from this state into the N719 HOMO appears possible, while from anatase it appears less plausible. Further work is needed to confirm this mechanism and refine it. Another way of describing this mechanism is in terms of hole injection from the dye into the oxide upon excitation and subsequent regeneration by the redox electrolyte (thereby underlining the inverse analogy to the electron injection mechanism). The general feasibility of such type of a process was indicated above for the interface p-NiO/erythrosin B.20 Although the hole is injected into the valence band in the NiO case, while it is injected into intraband gap states in our case, the general analogy seems apparent.27

Figure 5. (a) Optical absorption spectra of the dye N719, adsorbed on mesoporous, nanocrystalline TiO2, SrTiO3, and Nb2O5 membranes, corrected for reflectance and diffuse transmittance (absorbance, i.e., in optical density units). (b) IPCE (incident photon-to-electron conversion efficiency) of DSSCs using N719 as the dye and SrTiO3 or anatase TiO2 membranes. IPCE values were calculated using IPCE ) 1250 (Jph)/(λ‚Φinc), where Jph is the photocurrent density in µA/cm2, λ is the wavelength in nm and Φinc is the total incident photon flux in W/m2 (from ref 5).

Figure 6. Comparison of the SPV spectra of a crystalline and an amorphous film of Nb2O5. The temperatures indicated in the figure are the annealing temperatures, used to prepare the films. The decrease in voltage signal at short wavelengths is due mainly to the decreasing light intensity from the lamp (as the optical band gap of the oxide is e400 nm.27

to the presence of defects, are expected to occur especially in poorly crystallized materials. They are well-known to play an important role in the generation of the SPV signal.14,26 The coincidence of the onset of this signal with the absorption onset of the dye suggests that the dye’s HOMO and LUMO play a role in the appearance of this signal. Thus, we propose the existence of filled intraband gap states, which are energetically close to the HOMO of the dye. When the dye is excited, electrons can be transferred into its HOMO, forming the dye anion (because the LUMO is occupied). This excited-state dye anion may then relax back to the ground state by electron transfer to I3- in the solution. The net result of such a process is electron loss from the oxide membrane, resulting in a decrease of its Fermi level, which translates into a positive SPV signal. Some indication for the validity of this idea is found in SPV

Conclusions We have compared the behavior of three different mesoporous, nanocrystalline membranes of semiconducting oxides as electrodes in dye-sensitized solar cell structures. For this we used a well-characterized dye13 and a novel adaptation of the Kelvin probe method to measure the surface photovoltage of such electrodes in contact with the actual electrolyte used in the cells. The results give strong indication for involvement of electron injection from higher excited-state energy levels of dye molecules into the oxide conductors onto which the dye is attached by a chemical bond. In principle the levels involved can be either higher electronically or vibrationally excited ones.4 If the first effect were at work here we would not expect differences between the SPV spectra of the Nb2O5 and SrTiO3 cells. Experimentally, though, we find that the two systems give different blue shifts for the same dye (cf. Figure 4a). Such a situation is compatible with the existence and involvement of a manifold of closely spaced energy levels. Thus, using Occam’s razor28 we argue that Vibronic excitations are involved. Taken together with earlier indications this leads to the following conclusions: •DSSCs can yield photovoltages that are even higher than those predicted by the simple (Eredox - ECB) model.8 •Knowledge of the relevant energy levels of the various DSSC components will allow predictions about the cells’ behavior. •Injection from higher excited states, most probably vibrational ones (i.e., of hot electrons) from the dye into the semiconductor implies that the contact between the two is a highly intimate one with no significant density of scattering defects/kinetic barriers (at kT) between them. This justifies treating the dye plus the semiconductor as one effective medium.8,29 •Under the right conditions the HOMO of the dye can also be involved in charge exchange with the semiconductor. Acknowledgment. D.C. thanks the EPFL for a visiting professorship. Student visits to Rehovoth and Lausanne were made possible by the European Science Foundation, in the framework of its NANO program.

6352 J. Phys. Chem. B, Vol. 105, No. 27, 2001 References and Notes (1) Miller, R. J. D.; McLendon, G. L.; Nozik, A. J.; Schmickler, W.; Willig, F.; Surface Electron-Transfer Processes; VCH: New York, 1995; Ch. 1. (2) Moser, J.-E.; Wolf, M.; Lenzmann, F.; Gra¨tzel, M. Zeit. Phys. Chem. 1999, 212, 85. (3) Ferrere, S.; Gregg, B. J. Am. Chem. Soc. 1998, 120, 843. (4) Damrauer, N. H.; Cerullo, G.; Yeh, A.; Boussie, T. R.; Shank, C. V.; McCusker, J. K.; Science 1997, 275, 54. (5) Burnside, S.; Moser, J. E.; Brooks, K.; Gra¨tzel, M.; Cahen, D. J. Phys. Chem. 1999, 103, 9328. (6) The simple (illuminated) diode equation is I ) Io{exp(q[V - IRS]/ nkT) - Iph, where n ) diode factor (g1), k ) Boltzmann’s constant, T ) the absolute temperature in degrees K, q ) the electron charge and RS ) the series resistance). cf. Fahrenbruch, A. L.; Bube, R. H. Fundamentals of Solar Cells; Academic Press: New York, 1983. (7) Wells, A. F. Structural Inorganic Chemistry, 5th ed.: Oxford University Press: Oxford, 1986. (8) Cahen, D.; Hodes, G.; Gra¨tzel, M.; Guillemoles, J. F.; Riess, I. J. Phys. Chem B 2000, 104, 2053. (9) Sayama, K.; Sugihara, H.; Arakawa, H. Chem. Mater. 1998, 10, 3825. (10) Lenzmann, F.; Shklover, V.; Brooks, K.; Gra¨tzel, M. J. Sol-Gel Sci. Technol. 2000, 19, 175. (11) Barbe, C. J.; Arendse, F.; Comte, P.; Jirousek, M.; Lenzmann, F.; Shklover, V.; Gra¨tzel, M. J. Am. Ceram. Soc. 1997, 80, 3157. (12) Hinsch, A.; Kroon, J. M.; Spa¨th, M.; et al. Proceedings of the 16th European PhotoVoltaic Solar Energy Conference and Exhibition; Glasgow, May 1-5, 2000; paper OB9.2; James and James: London, in press. (13) (a) Nazeeruddin, M. K.; Kay, A.; Rodicio, I.; Humphry-Baker, R.; Mueller, E.; Liska, P.; Vlachopoulos, N.; Gra¨tzel, M. J. Am. Chem. Soc. 1993, 115, 6382. (b) Nazeeruddin, M. K.; Zakeeruddin, S. M.; HumphryBaker, R.; Jirousek, M.; Liska, P.; Vlachopoulos, N.; Shklover, V.; Fischer, C.-H.; Gra¨tzel, M. Inorg. Chem. 1999, 38, 6298. (14) Kronik, L.; Shapira, Y. Surface Sci. Rep. 1999, 37, 1. (15) (a) Goldstein, B.; Redfield, D.; Szostak, D. J.; Carr, L. A. Appl. Phys. Lett. 1981, 39, 258. (b) Szostak, D. J.; Goldstein, B. J. Appl. Phys. 1984, 56, 522. (16) Kronik, L.; Mishori, B.; Fefer, E.; Shapira, Y.; Riedl, W. Sol. Ener. Mater. Sol. Cells 1998, 51, 21. (17) Bastide, S.; Gal, D.; Cahen, D.; Kronik, L. ReV. Sci. Instr. 1999, 70, 4032.

Lenzmann et al. (18) Hagfeldt, A.; Gra¨tzel, M. Acc. Chem. Res. 2000, 33, 269. (19) Morrison, S. R. Electrochemistry at Semiconductor and Oxidized Metal Electrodes; Plenum Press: New York, 1980; p 183. (20) He, J.; Lindstrom, H.; Hagfeld, A.; Lindquist, S.-E. J. Phys. Chem. B 1999, 103, 8940. (21) Zaban, A.; Ferrere, S.; Gregg, B. J. Phys. Chem. B 1998, 102, 452. (22) This is so because, while normally the change in CPD upon illumination will be due to a change in band bending, here any band bending in the nm-sized particles and in the film, contacted by the electrolyte, is negligible [cf. Cahen et al., loc. cit.; ref 8 and references therein]. In addition the space charge capacity inside the film is much lower than that of the Helmholtz layer, under all conditions except strong forward bias (accumulation) (cf. Rothenberger et al., loc. cit.; ref 23). Hence shifts in work function under illumination will be due to EFn shifts unless extremely high light intensities are used. (23) Rothenberger, G.; Fitzmaurice, D.; Graetzel, M. J. Phys. Chem. 1992, 96, 5983. (24) To establish the process proposed here requires further work, especially use of well-characterized dyes with LUMO levels, different from that of the dye used here, when adsorbed on these semiconducting oxides and under the conditions of the DSSC. If the ELUMO is confirmed above the ECB of all three the oxides investigated here no blue shift should be observed. Work along this direction is in progress. (25) Vogel, R.; Weller, H. SPIE Proc. 1992, 1729, 82. (26) (a) Cohen, R.; Kronik, L.; Liu, A.; Rosenwaks, Y.; Shanzer, A.; Lorenz, J.; Ellis, A. B.; Cahen, D. J. Am. Chem. Soc. 1999, 121, 10545. (b) Cohen, R.; Kronik, L.; Vilan, A.; Shanzer, A.; Cahen, D. AdV. Mater. 2000, 12, 33. (27) We note that positive SPV signals are seen also at short wavelengths for the SrTiO3 and Nb2O5 500 °C samples, very close to their band gap energies (3.2-3.3 eV). Possibly these may result from holes in the VB as a result of VB to LUMO excitation, viewing the LUMO of the adsorbed dye as an oxide surface state. However, they can also be artifacts because of the low light intensities at those wavelengths and optical absorption of dye and triiodide electrolyte at these wavelengths. (28) “If two competing theories lead to the same predictions, the one that is simpler is the better.” See http://www.weburbia.com/physics/ occam.html. (29) Ferber, J.; Stangl, R.; Luther, J. Sol. Energy Mater. Sol. Cells 1998, 53, 29.