Bilayer Hybrid Solar Cells Based on Triphenylamine

Jun 11, 2010 - polymer/TiO2 are typically in the order of 0.2%13-16 but can be higher in devices .... Photoelectron spectroscopy (PES) measurements we...
1 downloads 0 Views 1MB Size
Download PDF
J. Phys. Chem. C 2010, 114, 11659–11664

11659

Bilayer Hybrid Solar Cells Based on Triphenylamine-Thienylenevinylene Dye and TiO2 Eva L. Unger,† Emilie Ripaud,‡ Philippe Leriche,‡ Antonio Cravino,‡ Jean Roncali,‡ Erik M. J. Johansson,† Anders Hagfeldt,† and Gerrit Boschloo*,† Department of Physical and Analytical Chemistry, Uppsala UniVersity, Sweden, and UniVersity of Angers, Moltech Anjou CNRS UMR 6200, Group SCL, France ReceiVed: April 1, 2010; ReVised Manuscript ReceiVed: May 20, 2010

Photoinduced energy conversion from multilayers of organic dye on dense TiO2 films was investigated in bilayer hybrid solar cells. Dye layers of varying thicknesses were prepared by spin-casting the star-shaped dye [tris(dicyano-vinyl-2-thienyl)phenyl]amine (1) from solutions onto dense TiO2 on conducting glass substrates. A spin-cast layer of poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) and graphite powder was used for contacting the devices. Excitons generated in the dye multilayer contribute to the power conversion efficiency, reaching a maximum of ca. 0.3% at a dye layer thickness of ca. 8 nm for the devices described herein. For dye layers exceeding 5 nm, the cell performance becomes limited by the exciton diffusion length LED and the hole mobility in the organic layer. Using dye multilayers is a viable way to increase light harvesting in solid-state dye-sensitized solar cells. 1. Introduction In the search for low-cost photovoltaic devices, significant progress has been made in solar cells that use organic compounds as their photoactive component. The dye-sensitized solar cell (DSC) is a photoelectrochemical system, where a monolayer of dye is chemisorbed to a wide-band-gap mesoporous metal oxide film.1 Photon absorption leads to excitation of the dye, followed by electron injection from the dye to the conduction band of the metal oxide, leaving the dye in the oxidized state. Efficiencies of up to 11% have been obtained for these devices using the iodide/triiodide redox couple as the regenerating medium for the oxidized dye.2,3 Solid-state dye-sensitized solar cell devices have been realized utilizing 2,2′,7,7′-tetrakis(N,N-dipmethoxypheny-amine)-9,9′spirobifluorene (spiro-MeOTAD) as an organic hole conductor for dye regeneration.4 For these devices, efficiencies around 5% have been achieved.5 Because of a higher interfacial recombination, the optimized film thickness of the mesoporous titanium dioxide films is limited to 2 µm. This results in an insufficient optical thickness and thus incomplete absorption of the incident light, which limits the possible power conversion.5 Interfacial recombination can be decreased by dye design6 and blocking the metal oxide surface with coadsorbents and coatings.7,8 Another approach to increasing the light-harvesting efficiency ηA of these devices is to employ a dye multilayer layer instead of a monolayer. This would also increase the distance between the metal oxide and the hole conductor, which decreases interfacial recombination. Conventional dye-sensitized solar cells are not excitonic devices as light absorption and charge separation occur at the metal oxide/dye interface.9,10 In hybrid devices with dye multilayers or polymers, light absorption leads to the generation of excitons in the organic layer. These have to diffuse to the dye/metal oxide interface where they can be dissociated. Because of the limited lifetime of the excitons, only those generated within the exciton diffusion length LED can * To whom correspondence should be addressed. Tel: +46-(0)18-471 3303. Fax: +46-(0)18-471 3633. E-mail: [email protected]. † Uppsala University. ‡ University of Angers.

contribute to the photocurrent. Thus, the exciton diffusion efficiency ηED presents a limitation to the overall device performance and diminishes the benefit of an increased ηA with increasing dye layer thickness d. Additionally, the chargetransfer efficiency ηCT at the interface and collection efficiency ηCC of the separated charges have to be taken into account. ηCT should be predominantly determined by the interfacial properties between the dye and the metal oxide. Although the electron mobility in TiO2 is comparatively high,11 the hole mobility in the dye layer can become a limiting factor to the device performance at thicker dye layers.12 Thus, dyes with good hole mobility are needed to realize the suggested devices.13 There are several examples for hybrid solar cells in the literature using organic polymers as a sensitizer and holetransporting medium in combination with metal oxides as an electron acceptor. Solar conversion efficiencies of bilayer hybrid polymer/TiO2 are typically in the order of 0.2%13-16 but can be higher in devices with efficient interfacial charge transfer and high exciton diffusion length.17 There are few examples using dyes as molecular semiconductors in hybrid solar cells because the hole mobility in these materials is usually low. Additionally, molecular compounds, such as porphyrins, often exhibit an anisotropy in the exciton diffusion and hole transport.18,19 For efficient devices, the film morphology needs to be controlled, which is not easy to achieve in nanostructured optoelectronic devices.20 Compounds based on triphenylamines (TPAs) have shown isotropic optical and electronic properties and are thus prominent building blocks in organic hole conductors.20,21 Despite forming disordered materials, hole mobilities up to 0.015 cm2 V-1 s-1 have been measured.22 The pseudo three-dimensionality in these conjugated systems makes these compounds reluctant to crystallization, prevents photoluminescence (PL) quenching in the material, and thus gives rise to a high exciton diffusion length.23 TPAs typically absorb visible light in the blue region of the electromagnetic spectrum, but by attaching electron-accepting groups, these compounds can be made more sensitive in the visible region. This approach led to the development of very efficient organic sensitizers for dye-sensitized solar cells.24,25

10.1021/jp102946z  2010 American Chemical Society Published on Web 06/11/2010

11660

J. Phys. Chem. C, Vol. 114, No. 26, 2010

Figure 1. Schematic picture of the hybrid solar cell test devices and the chemical structure of [tris(dicyano-vinyl-2-thienyl)phenyl]amine (1).

A number of star-shaped donor-acceptor compounds based on TPA gave efficient organic solar cells in combination with C60.26,27 In previous work, the dye [tris(dicyano-vinyl-2-thienyl)phenyl]amine (1) was tested in devices using C60 as an electron acceptor and power conversion efficiencies of 1.85% were achieved in optimized heterojunction devices. Moreover, this compound gave a power conversion efficiency of 0.4% in a single layer device with an Al contact. The open-circuit voltage VOC reached 0.76 V, and the short-circuit current JSC was 1.7 mA cm-2.28 We report here a first step in employing dye multilayers as sensitizers and hole-transporting materials in solid-state dyesensitized solar cells. Bilayer devices using smooth TiO2 films as an electron acceptor on transparent conducting electrodes and varying dye layer thicknesses of compound 1 were investigated. By fitting the experimental EQE dependency on the dye layer thickness d to a theoretical model,18,29 the exciton diffusion length LED and charge-transfer efficiency ηCT for 1 on TiO2 as an electron acceptor were estimated. Alternatively, the ddependent photoluminescence quenching can be used to estimate the LED.13,30 2. Experimental Section The synthesis of the dye [tris(dicyano-vinyl-2-thienyl)phenyl]amine 1 (structure shown in Figure 1) has been reported elsewhere.26,27 Bilayer devices of spin-casted layers of 1 on dense TiO2 were built to characterize the photovoltaic performance of 1 in hybrid solar cell devices. The back contact was established by a layer of (3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS, Aldrich), graphite powder, and a second conducting glass slide (schematically shown in Figure 1). Solid films of 1 were prepared by spin-casting with a Chemat Technology KW-4A spin-coater. The dye layer thickness d was varied by applying 10 µL/cm2 of solution with concentrations varying between 0.3 and 20 mM (applied at 1000 rpm, followed by 30 s at 2000 rpm) in dichloromethane (Aldrich). The solid films of 1 were characterized by UV-vis and photoluminescence (PL) spectroscopy. UV-vis measurements were carried out on an Ocean Optics HR2000 fiberoptic spectrometer using the blank substrate samples as a reference. PL measurements were carried out on a Horiba Fluorolog (Jobin Yvon) fluorimeter equipped with double monochromators for excitation and emission. The emission upon excitation at 530 nm was detected at a right angle with a R2658P detector. The sample was placed at a 30° angle relative to the excitation beam. The relative PL quenching of 1 by the two interfaces present in the solar cell

Unger et al. devices (TiO2 and PEDOT:PSS) was compared. Samples with varying d were prepared by spin-casting 1 on glass substrates that were half covered with a thin layer of TiO2 (for details in preparation, see the Supporting Information) or PEDOT:PSS. Before spin-casting (4000 rpm, 30 s), the aqueous PEDOT:PSS suspension was ultrasonicated for 15 min and filtered through a 0.45 µm syringe filter (RC Minisart). The absorption coefficient R(λ) of 1 was determined by correlating the absorbance A and dye layer thickness d for a set of four samples prepared by thermal evaporation (10-5 mbar). Photoelectron spectroscopy (PES) measurements were performed to determine the d of the samples with an in-house ESCA 300 spectrometer, using monochromated Al KR radiation (1486.7 eV; take-off angle, 90°). Charging and radiation effects were checked for by measuring the specific core level repetitively and were found negligible. The decrease in the intensity of the substrate (Ti 2p) signal depends on d of the overlying organic layer (for more information, see the Supporting Information).31 Conducting fluorine-doped tin oxide (FTO) coated glass (TEC15, Pilkington, substrate thickness ) 2.3 mm) was used as transparent electrodes in the solar cell devices. Compact layers of TiO2 were prepared onto the FTO substrates by spray pyrolysis deposition.32 A 1.2 M solution of diisopropoxy titanium(IV) bis(acetoacetonate) was manually sprayed with a commercial air brush (distance ∼6 cm, speed ∼5 cm/sec) using nitrogen as a carrier gas (1.5 bar) onto the FTO substrates (T ) 450 °C) on a glass ceramic hot plate (Harry Gestigkeit GmbH). The TiO2 substrates for the devices described herein were prepared in 12 spray cycles (for more information, see the Supporting Information). The thickness of the TiO2 layers on FTO was determined from the cross-sectional scanning electron microscope picture (SEM, LEO 1550). Compound 1 was spincast onto the TiO2-covered FTO substrates from dichloromethane with varying solution concentrations. The absorbance A of each sample was measured using the TiO2-coated substrates without the dye as reference. PEDOT:PSS was applied as described earlier. To remove solvent residues, the samples were first heated to 120 °C and then evacuated (10-5 mbar, 30 min). A piece of double-sided self-adhesive foam tape (tesa, mounting tape) with a round hole (standard hole punch, hole diameter ) 5.5 mm) was attached to the sample, graphite powder was poured into the hole, and electric contact was established with a second piece of conducting glass.33 During characterization, the solar cell devices were pressed together with clamps. The active area of the test devices was the area of the graphite contact defined by the diameters of the punch hole (0.24 cm2). For measuring solar cell performance, the cells were masked using the same dimensions as the active area. The current-voltage J-V characteristics of the devices were recorded using a solar simulator (Newport model 91160) and a computer-controlled Keithley 2400 source meter. External quantum efficiencies (EQE) were recorded using a computer-controlled setup consisting of a xenon light source (ASB-XE-175), a monochromator (Spectral Products CM110), and a potentiostat (PAR 273). Additionally, EQE measurements illuminating from the PEDOT:PSS side (further referred to as back-side illumination) were performed on a sample with d ) 48 nm by clamping flexible indium tin oxide coated PET substrates (60 Ω/square, IST) directly onto the PEDOT:PSS layer. The experimental results were compared to the theoretical EQE for front-side (from the TiO2 side) and back-side illumination derived in the literature18,29,34 (for details, see the Supporting Information). A reflective loss of 10% at the sample substrates was included in

Bilayer Hybrid Solar Cells

J. Phys. Chem. C, Vol. 114, No. 26, 2010 11661

Figure 2. Normalized absorbance of the dye 1 as a solid film on glass (a). Photoluminescence PL of a 1.9 nm solid sample on glass (b) and PEDOT:PSS (b′). PL of a 1.3 nm solid sample on glass (c) and TiO2 (c′). PEDOT:PSS quenches the PL by 83% (red arrow), whereas TiO2 quenches the PL less efficiently (-44%, blue arrow).

Figure 4. (a) Relation between the absorbance of dye films at 530 nm and the thickness determined by photoelectron spectroscopy (PES) measurements; linear fit giving a proportionality constant of 0.024 nm-1. (b) Dependency of the dye layer thickness d on the concentration of dye in the dichloromethane solution used for spin-casting.

Figure 3. Energy level diagram of the bilayer hybrid solar cell devices derived from electrochemical and spectroscopic data for all the components. The driving force for electron injection from the dye LUMO to the TiO2 conduction band is comparatively low.

the model. The exciton diffusion length LED and charge-transfer efficiency ηCT at the dye/metal oxide interface were estimated by fitting the d-dependent experimentally determined EQE to the theoretical EQE. 3. Results A schematic picture of the solar cell devices and the chemical structure of 1 is shown in Figure 1. The absorption maximum of 1 lies at 530 nm and the maximum photoluminescence at 700 nm (Figure 2). From the intercept between the absorbance and photoluminescence spectrum, the E0-0 transition of 1 was determined to be 2.0 eV. The electrochemical oxidation potential Eox of 1 was determined to be 1.35 V versus NHE,27 which was taken as the molecular HOMO. The excited-state potential was estimated to be -0.65 V versus NHE by adding the E0-0 to Eox. In Figure 3, the energy level diagram for the devices reported herein is shown. The flat-band potential of anatase TiO2 is approximately -0.5 V versus NHE35 and the work function of PEDOT:PSS is ca. 0.4 V versus NHE.36 In Figure 4a, the measured absorbance A is correlated to the dye layer thickness d determined from PES measurements. The linear regression of the data resulted in a proportionality factor of R′(530 nm) ) 0.024 nm-1, which corresponds to an absorption coefficient R(530 nm) of 5.53 × 107 m-1. Spray pyrolysis deposition of the TiO2 precursor onto conducting FTO glass gave smooth films. TiO2 films prepared

in 12 spray cycles were found to be 150 nm thick from crosssectional scanning electron microscope analysis (see the Supporting Information). Spin-casting layers of compound 1 resulted in dye layers almost linearly dependent on the concentration of the dye solution, which is shown in Figure 4b. Samples prepared from high solution concentration appeared to be less homogeneous than the ones prepared from low solution concentration. The current-voltage (J vs V) characteristics of the hybrid bilayer devices were measured under AM 1.5 conditions (1 sun, 100 mW cm-2) with an active area of 0.24 cm2. The data for selected samples with d ranging from 1 to 23 nm (a-e) are shown in Figure 5a. The corresponding solar cell parameters, VOC, JSC, and FF, and the device efficiency η for all investigated devices are listed in Table S-1 (see the Supporting Information). Under simulated solar illumination, the JSC initially increases with d but reaches an asymptotic value of ca. 1.0 mA cm-2, which seems to be the upper limit (Figures 5a and S-4a, Supporting Information). From the plot of VOC versus d (Figure 5b) an upper limit for the open-circuit voltage can be estimated to 0.90 V. In the same figure, it is shown that FF decreases from 0.6 to 0.25 with increasing d. The maximum efficiency for these bilayer devices was ca. 0.3% for devices with d ) 7.5 nm. Figure 6a depicts the external quantum efficiency EQE response for selected devices with d varying between 1 and 23 nm (a-e). In the same graph, the light-harvesting efficiency ηA derived from the absorbance spectrum of 1 is shown for d equal to 1 (a′), 2.2 (b′), and 3.6 nm (c′). The latter corresponds to the determined exciton diffusion length LED discussed later. The EQE spectrum shows a broad band with a maximum at around 530 nm and a shoulder around 380 nm. For d e 5 nm, the EQE response increases homogeneously over the whole spectrum (a-c). For the thinnest dye layers, the EQE is proportional to the light-harvesting efficiency ηA. To illustrate this, the ηA for d ) 1 (a′), 2.2 (b′), and 3.6 nm (c′) were included in Figure 6a.

11662

J. Phys. Chem. C, Vol. 114, No. 26, 2010

Unger et al.

Figure 5. (a) Current density J vs voltage V measurements for hybrid solar cells with varying dye layer thicknesses: d ) 1 (a), 2.2 (b), 4.5 (c), 7.5 (d), and 23 nm (e) on dense TiO2. (b) Dependency of the opencircuit voltage VOC (triangles) and fill factor (circles) on the dye layer thickness d; superimposed lines are a guide to the eye and have no mathematical significance.

The maximum EQE of ca. 5% for d ) 4.5 nm coincides with the absorption maximum of 1. The calculated ηA is 6, 11.5, and 18% for the three different d (a′-c′). Below 400 nm, light absorption by the TiO2 starts to contribute to the photogenerated current. At d > 5 nm, a progressive blue shift of the EQE maximum from 530 to 470 nm is observed while the shoulder at 380 nm progressively evolves. At higher photon energies, the measured EQE is higher than ηA, which is peculiar when ηCT and ηCC are assumed to be independent of the photon energy. Integration of the EQE spectra with respect to the AM 1.5G photon flux yields a short-circuit current density of 0.5 mA cm-2, which is lower than the JSC determined from J-V measurements. This discrepancy is probably due to the absence of UV light and lower intensity of the light source for the monochromatic EQE determination, which can affect the conductivity in the TiO2 and the organic layer.37 In Figure 6b, the experimental EQE at 530 nm was plotted as function of d. In the theoretical model for the d dependency of the EQE, the two cases of a nonquenching (nq) and a perfectly quenching (q) PEDOT:PSS interface were considered. The efficiency ηCT was, in both cases, treated as an independent factor to give the best overall fit in magnitude to the experimental data. For the nq-PEDOT:PSS, the best fit was achieved with LED ) 3.6 nm and ηCT ) 36% (nq, black solid line). For the q-PEDOT:PSS case, the best fit was achieved with LED ) 1.7 nm and an ηCT of 67% (q, red dashed line). For thicker d, the experimental EQE decreases instead of reaching a steadystate value as predicted by the model. To rule out that this is caused by a filter effect, we performed front-side/back-side illumination EQE on a sample with d ) 48 nm. The results are

Figure 6. (a) EQE response of hybrid devices with the dye layer thickness d ) 1 (a), 2.2 (b), 4.5 (c), 7.5 (d), and 23 nm (e); lightharvesting efficiency ηA of compound 1 with d ) 1 (a′), 2.2 (b′), and 3.6 nm (c′). (b) Experimental (circles) and theoretical EQE (lines) as a function of d. The determined LED was 3.6 and 1.7 nm and the ηCT was 34% and 67% for the nonquenching PEDOT:PSS (nq) and quenching PEDOT:PSS (q) cases, respectively. (c) Experimental EQE (data points) for front-side (a) and back-side (b) illuminations for a device with d ) 48 nm; theoretical EQE for front-side (a′) and backside (b′) illuminations. The latter case exhibits a filter effect.

shown in Figure 6c together with the theoretical EQE for this d, assuming both the TiO2 and the PEDOT:PSS to be quenching and using the same values for LED ) 1.7 nm and ηCT ) 67%, as determined from the d-dependent EQE analysis (q-PEDOT: PSS). The relative photoluminescence (PL) quenching at the emission maximum (700 nm) of a 1.3 nm sample of 1 on TiO2 and glass is included in Figure 2. Relative to glass, the PL of 1 was quenched by 44% on a thin, smooth TiO2 film. In the same figure, it is shown that the PL of a 1.9 nm sample of 1 on PEDOT:PSS was quenched by 83% compared with the PL on glass (PL0), showing that PEDOT:PSS quenches the PL of 1 very efficiently. The analysis of the d-dependent relative PL

Bilayer Hybrid Solar Cells quenching (PL/PL0) gave an LED of 1.2 nm and η ) 55% for 1 on TiO2 (Figure S-5; see the Supporting Information). 4. Discussion Up to a dye layer thickness d around 3.6 nm, the EQE increases in accordance with an increasing light-harvesting efficiency ηA (Figure 6a). This shows that it is possible to increase the optical thickness of a solid-state dye-sensitized solar cell device by employing a dye multilayer as a sensitizer. Dye layers exceeding the exciton diffusion length LED are not contributing to the photocurrent. Although the photoluminescence PL of 1 is quenched quite efficiently by PEDOT:PSS (Figure 2), this interface does not contribute to the photogenerated current by charge separation. This was confirmed by performing EQE measurements under illumination from the PEDOT:PSS side of a solar cell device (back-side) with d ) 48 nm, in which case, a filter effect was observed (Figure 6c). As d . LED, excitons are mainly generated close to the PEDOT:PSS interface and cannot reach the charge separating TiO2 interface. The dye layer acts as a filter for the incident light in this case. Only a small part of the incident light generates excitons within the LED of the charge-separating interface. These results also confirm the suggested working principle of the devices. Light absorption leads to the generation of excitons in the dye multilayer. These have to diffuse to the dye/TiO2 interface where they are subsequently charge separated. The open-circuit voltage VOC approaches a constant value of 0.9 V for higher d, which should correspond to the difference in the quasi Fermi level for electrons in TiO2 and the quasi Fermi level for holes in the dye layer. Considering that the Fermi level in TiO2 should be close to the conduction band (Figure 1), the Fermi level for holes would have to be close to mid band gap in compound 1. On the other hand, the maximum VOC could be set by the conducting level in PEDOT:PSS. Upon exciton dissociation at the interface, a gradient in the respective electrochemical potentials of the charge carriers is established in both TiO2 and the dye layer, which drive the charge carriers away from the interface.9,10 This concentration gradient-assisted charge separation becomes more effective at large d, whereas at small d, holes and electrons remain close to the interface and are thus more likely to recombine. Additionally, the coverage of the compact TiO2 at low d might be incomplete, giving rise to additional charge recombination pathways. The observed trend in VOC could also be explained by an energy band bending effect of the molecular semiconductor on the TiO2 surface. Such an effect was observed in d-dependent PES measurements of an organic hole-conducting molecule on TiO2, and it was argued that this led to hole trapping and enhanced interfacial recombination.38 Compared with the organic devices prepared by subsequent evaporation of 1 and C60, the JSC and EQE are substantially lower in the hybrid devices. This is the main reason for the difference in device performance as the VOC and FF are comparable.26,28 Both the evaporated and the spin-cast films of 1 have an amorphous character, and hybrid devices with evaporated dye did not give higher efficiencies. It is, therefore, unlikely that the difference in device performance can be attributed to morphology differences in the dye layer. On the other hand, C60 could have diffused into the organic layer, giving rise to a bigger interface area,30 thus leading to a higher current. Also, photogenerated charge carriers in C60 could contribute to the JSC. Furthermore, a lower charge transfer efficiency ηCT at the heterointerface is the probable cause for the lower performance.

J. Phys. Chem. C, Vol. 114, No. 26, 2010 11663 The latter is affected by the energetic band offset at the interface and the electronic coupling between the dye and the inorganic semiconductor. The polarizability of the interface, interfacial electronic states, and band-bending effects in the dye layer can also affect the exciton dissociation and electron-hole recombination rates at the interface.9,38,39 For efficient electron injection from the excited dye molecule to the acceptor levels in TiO2, the excited-state levels need to be higher in energy than the conduction band edge of TiO2. In the devices investigated here, this energy difference is only 0.15 eV (Figure 3), which is a comparatively low driving force for charge transfer.24 On the other hand, the conduction band in TiO2 and the LUMO in C6028 are quite similar. From DSC research, it is known that chemisorbtion of the dye onto the TiO2 surface by carboxylate groups significantly improves injection efficiency due to improved electronic coupling. Though giving moderate efficiency compared to the organic devices, the results reported here are comparable with other bilayer hybrid devices with TiO2 and organic semiconductors, such as P3HT or MEH-PPV.14,16,40 Hybrid solar cell devices give generally lower conversion efficiencies compared with their organic counterparts, which is probably a cause of the inherently different interfacial properties. Hybrid organic-inorganic junctions can be improved by modifying the heterointerface with amphiphilic molecules.15,41,42 This could also lead to an enhanced performance of the devices investigated here and will be further investigated. The exciton diffusion length LED should be an inherent property of the organic semiconductor but depends on how this value is determined experimentally. Optical interference effects can have a major impact on the experimentally determined LED,43,44 which we did not account for. We estimated the exciton diffusion length from both analysis of the d-dependent EQE and PL quenching experiments. When measuring complete solar cell devices, non-Ohmic electrode contacts and exciton deactivation at the cathode can affect the determined LED.18 From the EQE measurements, LED was determined to be 3.6 and 1.7 nm, assuming nonquenching and quenching back contacts, respectively. This is comparable to values for the LED in P3HT/ TiO2 bilayer solar cells reported by Kroeze et al.,18 who used the same exciton diffusion model. They determined the LED in P3HT to be 5.3 and 2.6 nm for nonquenching and quenching back contacts, respectively. We also determined the LED from d-dependent PL quenching experiments (see the Supporting Information) where LED for 1 was determined to be 1.2 nm with a ηCT of 55%. This is in agreement with our analysis of the d-dependent EQE, assuming a quenching PEDOT:PSS interface. Using a value of 3.6 nm for LED, it can be approximated that the EQE is at most 18% for a bilayer device with d ) LED and ηCT ) 1. For organic (1, C60) bilayer devices, however, EQE values of up to 28% have been reported.27 As mentioned earlier, this may be attributed to a larger interface area. Another explanation is that the determined LED can be affected by the quenching rate of excitons at the interface.46 This effect was not considered in this study, where complete quenching of excitons, but fractional charge transfer at the interface was assumed. We note that our assumption disagrees with the limited PL quenching on TiO2 substrates (Figure 2). The values reported here represent therefore of lower limit of LED in compound 1. Higher values can be expected when more efficient exciton dissociation takes place at the interface. The change of the shape of the EQE spectra in Figure 6a gives indications that ηCT is wavelength dependent. For higher photon energies, the EQE does not decrease but reaches a steady-state value

11664

J. Phys. Chem. C, Vol. 114, No. 26, 2010

or increases further for the investigated range of d. When the diffusion model is applied to model the EQE at higher photon energies, both higher ηCT and slightly longer LED result from the fit to the experimental data. This is another indication that ηCT might have an impact on the determined LED and cannot be treated as an independent parameter in the exciton diffusion model.45 As discussed earlier, the decreasing EQE at the absorption maximum is not caused by a filter effect. We interpret the deviation from the predicted d-dependent EQE to be caused by a limited hole transport and higher series resistance in the organic layer for thicker d.46 Upon charge separation at the interface, the holes have to be transported through the entire dye layer to the back contact. This is a field-dependent process, and there is an increasing competition between charge carrier transport and recombination.47 ηCC is thus a d-dependent parameter and cannot be assumed to be unity. This limited hole transport is also evident from the decreasing FF for higher d. 5. Conclusion and Outlook Bilayer hybrid solar cells were realized using flat TiO2 layers as an acceptor and a dye multilayer of 1 as donor material. The device efficiency increased gradually up to ca. 0.3% for a dye layer thickness d of ca. 8 nm. Ultimately, the devices become limited by the exciton diffusion length LED in the organic layer. These results suggest that dye multilayers can be employed to increase the optical thickness of the absorber layer in solidstate dye-sensitized solar cells. Similar to other hybrid solar cells, the investigated devices exhibited lower efficiencies compared with equivalent devices using organic compounds as electron acceptors. Different interfacial properties are the probable cause for the lower charge-transfer efficiency ηCT. The exciton diffusion length LED was estimated by fitting the experimental data to a theoretical model. The determined value of LED ) 1.7 nm is a lower limit for the LED in compound 1. For thicker d, the EQE and fill factors of the solar cell devices decrease, indicating a limited hole mobility in the organic layer. To further enhance the device performance, modification of the heterointerface15,42 will be investigated. In addition, device performance can be enhanced by increasing the interfacial area in nanostructured hybrid solar cell devices. Acknowledgment. We thank the Swedish Research Council (Vetenskapsrådet) for funding this project. We thank James Gardner and Hans-Christian Becker from the Department of Photochemistry and Molecular Science, Uppsala University, for assistance in performing photoluminescence measurements on the Horiba Fluorolog and UV-vis spectroscopy measurements with an integrating sphere. Supporting Information Available: Optimization and determination of the thickness of the dense TiO2 layers used in this study (S.1.), thickness determination of the dye layers (S.2.), table with solar cell parameters (Table S-1) investigated in this study and JSC and efficiency η dependency on d (S.3.), and details about the employed exciton diffusion model and PL quenching experiments (S.4.). This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) O’Regan, B.; Gra¨tzel, M. Nature 1991, 353, 737. (2) Gra¨tzel, M. J. Photochem. Photobiol., A 2004, 164, 3. (3) Nazeeruddin, M. K.; De Angelis, F.; Fantacci, S.; Selloni, A.; Viscardi, G.; Liska, P.; Ito, S.; Takeru, B.; Gra¨tzel, M. J. Am. Chem. Soc. 2005, 127, 16835.

Unger et al. (4) Bach, U.; Lupo, D.; Comte, P.; Moser, J. E.; Weissortel, F.; Salbeck, J.; Spreitzer, H.; Gra¨tzel, M. Nature 1998, 395, 583. (5) Snaith, H. J.; Moule, A. J.; Klein, C.; Meerholz, K.; Friend, R. H.; Gra¨tzel, M. Nano Lett. 2007, 7, 3372. (6) Karthikeyan, C. S.; Wietasch, H.; Thelakkat, M. AdV. Mater. 2007, 19, 1091–1095. (7) Karthikeyan, C. S.; Thelakkat, M. Inorg. Chim. Acta 2008, 361, 635. (8) Snaith, H. J.; Schmidt-Mende, L. AdV. Mater. 2007, 19, 3187– 3200. (9) Gregg, B. A. MRS Bull. 2005, 30, 20–22. (10) Gregg, B. A. J. Phys. Chem. B 2003, 107, 4688–4698. (11) Solbrand, A.; Lindstrom, H.; Rensmo, H.; Hagfeldt, A.; Lindquist, S. J. Phys. Chem. B 1997, 101, 2514. (12) Savenije, T. J.; Goossens, A. Phys. ReV. B 2001, 64, 115323. (13) Savenije, T. J.; Warman, J. M.; Goossens, A. Chem. Phys. Lett. 1998, 287, 148. (14) Daoud, W. A.; Turner, M. L. React. Funct. Polym. 2006, 66, 13. (15) Goh, C.; Scully, S. R.; McGehee, M. D. J. Appl. Phys. 2007, 101, 114503. (16) Lira-Cantu, M.; Krebs, F. C. Sol. Energy Mater. Sol. Cells 2006, 90, 2076. (17) Arango, A. C.; Johnson, L. R.; Bliznyuk, V. N.; Schlesinger, Z.; Carter, S. A.; Ho¨rhold, H. H. AdV. Mater. 2000, 12, 1689–1692. (18) Kroeze, J. E.; Savenije, T. J.; Vermeulen, M. J. W.; Warman, J. M. J. Phys. Chem. B 2003, 107, 7696–7705. (19) Savenije, T. J.; Moons, E.; Boschloo, G. K.; Goossens, A.; Schaafsma, T. J. Phys. ReV. B 1997, 55, 9685. (20) Shirota, Y. J. J. Mater. Chem. 2000, 10, 1–25. (21) Thelakkat, M. Macromol. Mater. Eng. 2002, 287, 442–461. (22) Ohishi, H.; Tanaka, M.; Kageyama, H.; Shirota, Y. Chem. Lett. 2004, 33, 1266. (23) Roncali, J.; Leriche, P.; Cravino, A. AdV. Mater. 2007, 19, 2045– 2060. (24) Hagberg, D. P.; Marinado, T.; Karlsson, K. M.; Nonomura, K.; Qin, P.; Boschloo, G.; Brinck, T.; Hagfeldt, A.; Sun, L. J. Org. Chem. 2007, 72, 9550–9556. (25) Hagberg, D. P.; Yum, J.-H.; Lee, H.; De Angelis, F.; Marinado, T.; Karlsson, K. M.; Humphry-Baker, R.; Sun, L.; Hagfeldt, A.; Gra¨tzel, M.; Nazeeruddin, M. K. J. Am. Chem. Soc. 2008, 130, 6259. (26) Cravino, A.; Roquet, S.; Leriche, P.; Ale´veˆque, O.; Fre`re, P.; Roncali, J. Chem. Commun. 2006, 1416. (27) Roquet, S.; Cravino, A.; Leriche, P.; Ale´veˆque, O.; Fre`re, P.; Roncali, J. J. Am. Chem. Soc. 2006, 128, 3459–3466. (28) Cravino, A.; Leriche, P.; Ale´veˆque, O.; Roquet, S.; Roncali, J. AdV. Mater. 2006, 18, 3033–3037. (29) So¨dergren, S.; Hagfeldt, A.; Olsson, J.; Lindquist, S.-E. J. Phys. Chem. 1994, 98, 5552–5556. (30) Markov, D. E.; Amsterdam, E.; Blom, P. W. M.; Sieval, A. B.; Hummelen, J. C. J. Phys. Chem. A 2005, 109, 5266. (31) Briggs, D., Seah, M. P., Eds. Practical Surface Analysis by Auger and X-ray Photoelectron Spectroscopy; Wiley: Chichester, U.K., 1983; p 211. (32) Kavan, L.; Gra¨tzel, M. Electrochim. Acta 1995, 40, 643. (33) Smestad, G. P.; Spiekermann, S.; Kowalik, J.; Grant, C. D.; Schwartzberg, A. M.; Zhang, J.; Tolbert, L. M.; Moons, E. Sol. Energy Mater. Sol. Cells 2003, 76, 85. (34) Peumanns, P.; Aharon, Y.; Forrest, S. R. J. Appl. Phys. 2003, 93, 3693–3723. (35) Boschloo, G. K.; Goossens, A.; Schoonman, J. J. Electrochem. Soc. 1997, 144, 1311–1317. (36) Greczynski, G.; Kugler, T.; Salaneck, W. R. Thin Solid Films 1999, 354, 129. (37) Snaith, H. J.; Gra¨tzel, M. Phys. ReV. Lett. 2007, 98, 177402. (38) Johansson, E. M. J.; Odelius, M.; Karlsson, P. G.; Siegbahn, H.; Sandell, A.; Rensmo, H. J. Chem. Phys. 2008, 128, 184709. (39) Gledhill, S. E.; Scott, B.; Gregg, B. A. J. Mater. Res. 2005, 20, 3167–3179. (40) Liu, Y.; Summers, M. A.; Edder, C.; Fre´chet, J. M. J.; McGehee, M. D. AdV. Mater. 2005, 17, 2960–2964. (41) Boucle´, J.; Ravirajan, P.; Nelson, J. J. Mater. Chem. 2007, 17, 3141– 3153. (42) Zhu, R.; Jiang, C.-Y.; Liu, B.; Ramakrishna, S. AdV. Mater. 2008, 20, 1–20. (43) Scully, S. R.; McGehee, M. D. J. Appl. Phys. 2006, 100, 034907. (44) Theander, M.; Yartsev, A.; Zigmantas, D.; Sundstrom, V.; Mammo, W.; Andersson, M.; Inganas, O. Phys. ReV. B. 2000, 61, 12957. (45) Gregg, B. A.; Sprague, J.; Peterson, M. W. J. Phys. Chem. B 1997, 101, 5362. (46) Kerp, H. R.; Donker, H.; Koehorst, R. B. M.; Schaafsma, T. J.; van Faassen, E. E. Chem. Phys. Lett. 1998, 298, 302. (47) Hoppe, H.; Sariciftci, N. S. J. Mater. Res. 2004, 19, 1924–1945.

JP102946Z