Dye Sensitized Solar Cells (DSCs) are attracting much attention because of their relatively high conversion efficiencies, low cost production processes and short energy payback time1. In these devices, sunlight is absorbed by photoactive molecules that are attached to the surface of a wide band gap semiconductor oxide (typically TiO2 or ZnO) forming a dense monolayer. The system is completed using a hole transporting material which is usually a liquid electrolyte containing the iodide/tri-iodide (I/I3) red/ox couple2. The molecules act as sensitizers and upon photo-excitation inject an electron into the conduction band of the metal oxide. While the electrons are conducted through the nanostructured metal oxide to reach the external circuit, the oxidized dye is regenerated by the red/ox couple, which is itself regenerated at the counter electrode3,4. The complete description of the basic working principles of a DSC can be found elsewhere5,6.

Several different classes of materials are employed to fabricate a DSC, but the sensitizer is probably the key element since it governs the photon harvesting and the creation of free charges after injection of electrons into the nanostructured semi-conducting oxide. Moreover, the sensitizer structure has also been shown to control key electron transfer processes at the TiO2/dye/electrolyte interface such as the recombination of TiO2 electrons with electrolyte species or with dye cations themselves. For this reason many efforts have recently focused on the development of new efficient sensitizers that could help improve device performance and allow for practical and real use of this technology beyond the laboratory.

For many years, ruthenium based complexes were the “champion” dyes of DSCs and some of them were distinguished by achieving more than 11% efficiency7. However, despite the fact that high power conversion efficiencies and relatively stable DSCs have been fabricated using ruthenium sensitizers, Ruthenium cannot be considered an “earth abundant material” and, thus, it is desirable to look for alternative dyes. Besides, even though some ruthenium complexes show broad absorption spectra, they generally have modest molar extinction coefficients8,9 whichs limit their performances when employed to sensitize thin electrodes. As a consequence, this class of sensitizer is not ideal for DSCs based on ionic-liquids as thin electrodes are preferable due to the high inherent viscosity of these electrolytes10.

To overcome this problem, the use of organic sensitizers has been demonstrated to be a useful strategy. Organic sensitizers have attracted great interest because of their potentially lower production costs, easier and more versatile synthesis and much larger molar absorption coefficients. In addition, some of them have shown satisfactory stability11,12. Many different organic dyes with conversion efficiencies in the range of 6–8% have been reported in the last years but only a few examples have overcome efficiencies of 10%13,14,15,16. To date, the highest efficiency for a DSC employing a purely organic sensitizer has been achieved by Wang and co-workers using an electrolyte based on iodide/tri-iodide in a volatile electrolyte. Using C219, a high absorbing dye prepared in ten steps, they reported 10.1% power conversion efficiency under 1 Sun irradiation17.

In addition to purely organic dyes, metallated porphyrins showing strong absorption in the visible region, have been synthesized and tested in DSCs by Grätzel and co-workers18. After appropriate tailoring of the chemical structure and employment of a cobalt electrolyte, the porphyrin dye YD2-o-C8 showed power conversion efficiencies of up to 11.9% under standard conditions (AM 1.5 G, 100 mWcm−2 intensity) which was further improved to 12.3% when co-sensitized with an organic dye19. However, porphyrins usually require complicated synthetic strategies with relatively low yields, especially when an anchoring function and a redox group have to be introduced at specific positions20. For this reason, our motivation in this work was to design a simple, easy to prepare and efficient organic sensitizer. Our strategy was principally driven by the possibility to reduce the number of chemical steps involved in its preparation, the availability of the organic precursors and the possibility to confer to the dye tailored optical and electrochemical properties. In the literature, most of the new organic sensitizers are prepared using the very well-known donor-π-acceptor design approach, where the donor bears alkyls chains and the π conjugated bridge is mainly based on symmetrical chromophoric units. Our approach, on the other hand, is based on the use of a dissymmetric π-conjugated bridge that embeds an alkyls chain and an electro deficient unit localized close to the electron-withdrawing anchoring function and an electron-rich unit close to the aryl amine donating group21,22. The introduction of a phenyl ring between acceptor benzothiadiazole (BTD) and the cyanoacrylic acid group should stabilize the dye radical cation and decrease recombination rate as showed by Haid et al.23.

Herein we report the preparation of a new purely organic dye, RK1, whose synthesis involves just five steps which starts from low cost precursors, resulting in a solar to electrical power conversion efficiency of 10.2% under standard conditions (AM 1.5 G, 100 mWcm−2 intensity). To the best of our knowledge, this represents one of the best power conversion efficiencies ever reported for a DSC based on a purely organic dye. Moreover, this high efficiency is coupled with outstanding long-term stability as demonstrated by devices based on an ionic liquid electrolyte.


RK1 synthesis

The chemical structure of RK1 is shown in Scheme 1. The thiophene BTD phenyl chromophore constitutes the π bridge in this D–π–A structure. RK1 has a very simple chemical structure and its preparation does not require complicated synthetic procedures or purification steps allowing for its preparation in the tens of grams scale.

The convergent synthetic strategy to access RK1 relies on the preparation of two key building blocks (1 and 2) that are subsequently coupled together via a Suzuki or a Stille cross coupling reaction (see Figure 1). The preparation of these two intermediates is based on the use of low cost and commercially available precursors and a classical palladium cross coupling reaction. With the idea in mind of potential future synthesis of this dye at an industrial level, toxic reaction solvents such as DMF were purposely not used in our synthetic scheme. In addition, the replacement of DMF by a mixture of toluene, THF and water helped us to strongly improve the yields of the Suzuki coupling reaction. The overall yield of preparation for RK1 is 50% after just five steps.

Figure 1
figure 1

Synthetic route and reaction conditions for the preparation of RK1.

Optical and electrochemical data

The optical properties of the dye were investigated by UV-Visible spectroscopy in ethanol solution and compared to those of N719 which was used as a reference compound in this study (Figure 2 and Table 1). The UV-Vis spectrum of RK1 shows, as expected, two absorption bands between 300 and 580 nm. The first one which is located in the UV region (λ = 366 nm, ε = 45000 M−−1) is assigned to the π-π* transition of the aromatic rings, whereas the second absorption band in the visible region (λ = 470 nm, ε = 26600 M−−1) is attributed to the internal charge transfer (ICT) transition that occurs between the electron-withdrawing and electron-donating segments of the molecule. This band is at shorter wavelength with respect to that of N719 but the measured molar extinction coefficient at 470 nm for RK1 is almost two times higher than the corresponding values for N719. As a consequence, RK1 shows an aesthetically pleasing orange colour both in solution and on TiO2 films.

Table 1 Optical and electrochemical data of RK1
Figure 2
figure 2

UV-Visible absorption spectra of RK1 and N719 in ethanol solution.

The electrochemical properties of RK1 were investigated by cyclic voltammetry (CV) with the goal to determine the energy level positions of frontier orbitals. Upon the application of negative potentials, the dye undergoes a quasi-reversible reduction/re-oxidation process and upon the application of positive potentials it showed two reversible oxidation waves corresponding to an oxidation/re-reduction process. The first oxidation peak (Eox1) was found at 0.98 V versus NHE which can be assigned to the oxidation of the triphenylamine donor moiety. The second oxidation potential (Eox2) was observed at 1.41 V and is attributed to the oxidation of the π-conjugated backbone. The HOMO and LUMO energy level positions determined from the onset of the first oxidation and reduction potential were 0.93 and −0.72 V vs. NHE. RK1 therefore showed a bandgap (EgCV) of 1.65 eV. The experimentally determined LUMO energy level of RK1 lies above the conduction band edge of TiO2 (0.40 V versus NHE), which ensures the necessary driving force for electron injection from the dye into the conduction band of the semiconductor. Moreover, the HOMO energy level positioned at 0.93 V versus NHE also indicates that there is a sufficient energy offset for regeneration by the I/I3 redox couple at −0.35 V versus NHE24.

In order to compare these experimental values and to gain insight into structural properties, the electron density distribution of the frontier orbitals of RK1 was determined by density functional theory (DFT) using the B3LYP hybrid functional (see Experimental Section for details). We report in Figure 3 the calculated molecular orbital energy diagram for RK1, along with iso density plots of the HOMO and LUMO. The dyes show a directional electron distribution, with the delocalization of the HOMO on the triphenylamine group and the thiophene of the π-conjugated bridge. On the other hand, the LUMO is predominantly delocalized on the BTD-phenylvinylcyanoacetic units. This directional electron distribution is consistent with the dissymmetrisation of the core and should facilitate the reduction of the oxidized dyes by the I/I3 based electrolyte and the injection of electrons into the conduction band of TiO2.

Figure 3
figure 3

Frontier molecular orbitals of RK1 calculated at B3LYP level of theory.

Dye sensitized solar cells

Solar cells with liquid electrolyte

The photovoltaic parameters of the best RK1 DSCs using a standard acetonitrile based electrolyte are shown in Table 2 alongside those for a device made using the ruthenium sensitizer N719 for comparison. The photoactive electrodes of these devices were fabricated using standard double layer structure consisting of mesoporous TiO2 and a TiO2 scatter layer. The I-V curves of champion cells recorded under AM 1.5, 1 Sun illumination and in the dark and their photocurrent response (IPCE) are shown in Figure 4. It should be underlined that the results presented here were reproduced in two different laboratories (ICIQ and Solaronix) and that three different cells were fabricated for each entry of the table.

Table 2 Photovoltaic parameters of DSCs employing an acetonitrile electrolyte
Figure 4
figure 4

(a) I-V curves recorded under AM 1.5 illumination and in the dark and (b) IPCE spectra for RK1 and N719 devices.

In order to investigate and optimize the DSC performances of the new dye, the TiO2 film thickness was varied between 4 and 13 μm onto which a 3.5 to 4 μm scattering layer was deposited. For the dyeing of the TiO2 electrodes, 0.2 to 0.5 mM solutions of RK1 containing various amounts of 3α, 7α-dihydroxy-5β-cholic acid (chenodeoxycholic acid) were employed. Though the use of chenodeoxycholic acid with organic sensitizers is known to reduce the sensitizer's loading, it has also been shown to diminish the undesirable formation of dye aggregates on the TiO2 surface and decrease recombination processes between TiO2 electrons and electrolyte species25.

Interestingly for RK1 devices performances remain higher than 7% even for low TiO2 film thicknesses. When thicker electrodes (10 to 13 μm) are employed, very high current densities of up to 20.25 mA/cm2 can be attained. This results in a power conversion efficiency of 10.2% for the RK1 champion cell which is outstanding for an organic sensitizer.

The Jsc, FF and Voc for champion cells of N719 and RK1 with power conversion efficiencies of over 10%, are rather similar. The dark current in both devices is nearly identical. The photocurrent response (IPCE) shows that though the N719 device performs better at longer wavelengths, RK1 performs significantly better at shorter wavelengths, in agreement with the respective absorption spectra of these dyes (Figure 2).

The Voc of a DSC is determined by the quasi-Fermi-level of the metal oxide semiconductor, which is correlated with its conduction band edge (Ec) and electron density. Electron density is itself dependent on rate of recombination between semiconductor electrons and oxidized electrolyte species. To investigate the difference in voltage for these devices, charge extraction and transient photovoltage measurements were conducted. Charge extraction data (Figure 5(a)) show that the RK1 device has a slightly higher charge density with respect to the N719. This indicates that the TiO2 conduction band in RK1 devices is shifted downward somewhat compared to N719 devices. Electron lifetime measured using transient photovoltage measurements (Figure 5(b)) for the RK1 device is shorter than for the N719 device. This is routinely observed for DSCs employing organic sensitizers26. Therefore, the combined effect of the lower lying TiO2 conduction band and shorter electron lifetime explains the general trend of lower Voc for RK1 devices.

Figure 5
figure 5

(a) Charge extraction data showing electron density as a function of induced voltage and (b) transient photovoltage data showing electron lifetimes versus electron density for N719 and RK1 DSC devices.

Solar cells with ionic liquid electrolyte

For the practical application of dye-sensitized solar cells (DSCs), in addition to high performances, long device lifetime is a major requirement. Because highly volatile solvents such as acetonitrile are inappropriate for stability measurements, a solvent-free ionic liquid electrolyte was used instead in order to evaluate the lifetime of the RK1 sensitizer under thermal stress and light soaking. The photovoltaic parameters of these solar cells are given in Table 3. Two different types of electrodes were used to fabricate the devices. In addition to regular mesoporous TiO2, to further increase the thickness of the electrode, a macro channel TiO2 paste was employed. This strategy was found to be particularly efficient to improve the light-harvesting capacity of these devices without alteration of the capability of the viscous ionic liquid electrolyte to penetrate deeply into the nanostructured electrode. With 8 μm thick TiO2 film coated with a 3.5 μm thick reflecting layer, devices showing a Jsc of 15.40 mA/cm2, a Voc of 0.665 V, a FF of 69% and a power conversion efficiency of 7.36% at 1 Sun were obtained. This performance ranks among the best reported for organic sensitizers using ionic liquid electrolyte and is considerably better than most of the performances reported for Ruthenium based dyes27,28. The fabricated DSCs were irradiated continuously in a solar simulator with a light intensity of 1000 W/cm−2 at 65°C. The evolution of the photovoltaic parameters of a cell measured under the irradiance of AM 1.5 G sunlight during successive full-sun visible-light soaking is presented in Figure 6. After light soaking, the power conversion efficiency of the solar cell significantly increases; this behaviour is related to the improved penetration of the ionic liquid electrolyte through the complete thickness of the electrode and the “activation” of the entire electrode29. After this period we observed a stabilisation of the performances and no degradation; this cell exhibits outstanding stability with no measurable degradation of the initial performances after 2200 h of irradiation. After this period the cells start to undergo a linear degradation of their performances but it is noteworthy that after 5000 h continuous irradiation at 65°C, the degraded solar cells still show 75% of their initial power conversion efficiency. From the best of our knowledge this stability performance is the best ever achieved for an organic sensitizer.

Table 3 Photovoltaic parameters of DSCs with an ionic liquid electrolyte
Figure 6
figure 6

Detailed photovoltaic parameters of a cell measured under the irradiance of AM 1.5 G sunlight during successive full-sun visible-light soaking (1 sun 1000 W/m2) at 65°C.

Jsc: short-circuit photocurrent density; Voc: open-circuit voltage; FF: fill factor; η: power conversion efficiency.


We have designed and synthesized a novel donor-acceptor organic dye (RK1) based on a dissymmetric central core embedding a benzothiadiazole chromophoric unit. This dye has a simple chemical structure and can be prepared very easily in five steps with an overall yield of 50%. When employed in DSCs with liquid electrolytes containing the I/I3 redox couple, under standard conditions a high short circuit current of 18.26 mA/cm2, an open circuit voltage of 0.76 V and a FF of 74% were achieved leading to an overall power conversion efficiency of 10.2%. This represents one of the highest efficiencies ever recorded for a DSC device employing a fully organic dye and iodine/iodide electrolytes. In addition, when a solvent-free ionic liquid electrolyte was used a short circuit current of 15.40 mA/cm2, an open circuit voltage of 0.665 V and a FF of 69% leading to a power conversion efficiency of 7.36% under standard conditions was achieved. These DSC devices showed outstanding stability when irradiated continuously with 1000 W/cm−2 light intensity at 65°C with no measurable degradation of the initial performances after 2200 h and only a loss of circa 25% after 5000 h. We note that these results are obtained with iodide/tri-iodide that has a strong absorption band located at 360 nm30 and that the photocurrent generated by the sensitizer could be further improved in the blue part of the spectrum with the replacement of this electrolyte by a cobalt redox shuttle. In addition, such redox shuttles can be used to increase Voc further improving device performance. To conclude, we believe that RK1 is a most promising organic dye for future low cost, stable and efficient dye sensitized solar cells.


Materials synthesis

2-bromo-3-octylthiophene, 4-(N,N-diphenylamine) phenylboronic acid, 4-formylphenyl-boronic acid, 2-cyanoacetic acid, piperidine, n-BuLi [2.5 M solution in Tetrahydrofuran (THF)], trimethyl tin chloride solution [1 M solution in n-hexane] were purchased from Aldrich or TCI chemicals and used as received. N-Bromo-Succinimide (NBS) was purchased from Fisher Chemicals and 4,7-dibromo-2,1,3-benzothiadiazole from Orgalight or provided by Kaïronkem. The solvents, such as anhydrous toluene, chloroform and acetonitrile from Aldrich were used as received. THF was used after distillation under sodium and benzophenone. Spectroscopic grade solvents from Aldrich were used for spectral measurements. 4-bromo-7-(4-formylbenzyl)-2,1,3-benzothiadiazole was synthesised according to literature31.

Synthesis of 4-(3-Octylthien-2-yl)-N,N-dipphenylamine (2)

Under argon, 4-(diphenylamino)phenylboronic acid (1.65 g, 4.63 mmol), 2-bromo-3-octylthiophene (1.16 g, 4.21 mmol), K2CO3 (700 mg, 5.05 mmol) and Pd(PPh3)4 (200 mg, 0.17 mmol) were dissolved in a degazed toluene/THF/water solution (30 mL, 20/7.5/2.5). The mixture was stirred overnight at 75°C before being poured into water. The organic phase was recovered with Et2O, washed with brine, dried on Na2SO4 and concentrated under reduced pressure. The crude oil was purified by chromatography on silica gel using n-hexane/CH2Cl2 2:1 to afford pale yellow oil 2 (1,67 g, 3.62 mmol, 86,0%). 1H NMR (CD2Cl2, 200 MHz): δ = 7.33–7.96 (m, 16H), 2.66 (t, 2H, J = 7.8 Hz), 1.64–1.53 (m, 2H), 1.41–1.20 (m, 10H), 0.88 (t, 3H, J = 3 Hz). 13C NMR (CD2Cl2, 50 MHz): 148.10, 147.52, 138.76, 138.09, 130.48, 130.09, 129.83, 129.17, 125.06, 124.63, 123.69, 123.60, 32.45, 31.60, 30.030, 29.95, 29.83, 29.20, 23.25, 14.55, HRMS (ESI): [M + Na]+ = 439.23 (calcd. for C30H33NS: 439.23); Anal. Calcd for C30H33NS: C,81.96; H,7.57; N,3.19; S,7.29,. Found: C,81.66; H,7.67; N,3.22; S,7.47.

Synthesis of 4-(7-(5-(4-(diphenylamino)phenyl-4-octylthien-2-yl)benzo[c][1,2,5]thiadia-zol-4-yl)benzaldehyde (4)

Under argon, 2 (200 mg, 0.45 mmol) was dissolved in distilled THF (8 mL) then n-BuLi (0.32 mL, 0.58 mmol) was added at −78°C. The solution was stirred for an hour at −50°C before adding a n-hexane solution of Me3SnCl (0.73 mL, 0.72 μmol) at −78°C. The solution was allowed to reach room temperature and stirred for 2 hours. The reaction was quenched with water and the organic phase was extracted with n-hexane, dried with Na2SO4, filtered and concentrated under reduced pressure. The resulting oil 3 was engaged without any further purification in a Stille coupling with 1. Under argon, solids 3, 1 (174 mg, 0.55 mmol), Pd2dba3 (16 mg, 18 μmol) and P(o-tolyl)3 (44 mg, 0.14 mmol) were dissolved in anhydrous toluene (8 mL) and refluxed for 24 hours. The mixture was then poured into HCl (1 M). The organic phase was extracted with Et2O, washed with HCl (1 M), dried over Na2SO4 and concentrated under reduced pressure. The crude solid was purified by chromatography on silica gel using DCM/n-hexane 6:4 first then DCM as eluent to afford red solid 4(270 mg, 0.40 mmol, 87%). 1H NMR (CDCl3, 200 MHz): δ = 10.11 (s, 1H), 8.19 (d, 2H, J = 8.4 Hz), 8.08 (s, 1H), 8.07 (d, 2H, J = 8.2 Hz), 7.96 (d, 1H, J = 7.6 Hz), 7.82 (d, 1H, J = 7.6 Hz), 7.42–7.26 (m, 6H), 7.19–7.03 (m, 8H), 2.77 (t, 2H, J = 7.8 Hz), 1.80–1.66 (m, 2H), 1.45–1.19 (m, 10H), 0.87 (t, 3H, J = 6.6 Hz); 13C NMR (CDCl3, 50 MHz): δ = 191.83, 153.68, 152.64, 147.42, 147.36, 143.23, 140.30, 139.46, 136.11, 135.632, 131.03, 130.50, 129.93, 129.81, 129.66, 129.33, 128.94, 127.84, 127.71, 124.74, 123.24, 122.87, 31.880, 31.04, 29.56, 29.42, 29.27, 28.97, 22.67, 14.13,; HRMS (ESI): [M + Na]+ = 677.25 (calcd. for C43H39N3OS2: 677.25); Anal. Calcd for C43H39N3OS2: C, 76.18; H, 5.80; N, 6.20; S, 9.46. Found C, 76.07; H, 5.73; N, 6.17; S, 9.65.

Synthesis of RK1 dye

Under argon, 4 (134 mg, 0.18 mmol), cyanoacetic acid (78 mg, 0.90 mmol), were dissolved in a mixture of acetonitrile (6 mL) and chloroform (4 mL). A catalytic amount of piperidine was added and the solution was refluxed for 3 hours. Solvent was removed under reduced pressure and the solid dissolved in chloroform. The organic phase was washed with HCl solution (1.5 M), dried on Na2SO4 and concentrated. The crude solid was purified by chromatography on silica gel using DCM/MeOH/Acetic acid 90:5:5 as eluent to afford RK1 as red solid 5 (128 mg, 0.16 mmol, 87%). 1H NMR (CDCl3, 200 MHz): δ = 8.36 (s, 1H), 16 (s, 4H), 8.06 (s, 1H), 7,91 (d, 1H, J = 7.4 Hz), 7.79 (d, 1H, J = 7.6 Hz), 7.40–7.26 (m, 6H), 7.18–7.03 (m, 8H), 2.74 (t, 2H, J = 7.8 Hz), 1.78–1.66 (m, 2H), 1.42–1.21 (m, 10H), 0.87 (t, 3H, J = 6.6 Hz); 13C NMR (THF-d8, 50 MHz): δ = 163.57, 154.40, 153.76, 153.38, 148.32, 148.29, 142.22, 140.94, 139.85, 137.13, 132.27, 131.70, 131.65, 130.84, 130.39, 130.28, 129.99, 129.53, 128.78, 128.15, 125.44, 125.36, 123.95, 123.47, 116.17, 104.35, 32.681, 31.75, 30.31, 30.22, 30.09, 29.61, 23.38, 14.29,; HRMS (Q-TOF): [M − H] = 743.25145 (calcd. for C46H39N4O2S2: 743.2519), [M-CO2H] = 699.26162 (calcd. for C45H39N4S2: 699.2628); Anal. Calcd for C46H40N4O2S2: C, 74.16; H, 5.41; N, 7.52; S, 8.61. Found: CC, 73.87; H, 5.51; N, 7.41; S, 8.89.

Optical and electrochemical characterizations

UV-vis absorption spectra were recorded in solution on a Perkin-Elmer Lambda 2 spectrometer (wavelength range: 180–820 nm; resolution: 2 nm). Electrochemical studies of the synthesized molecules were carried out in a one compartment, three-electrode electrochemical cell equipped with a flat platinum working electrode (7 mm2), a Pt wire counter electrode and a Ag wire pseudo-reference electrode, whose potential was checked using the Fc/Fc+ couple as an internal standard. The electrolyte consisted of 0.1 M tetrabutylammonium tetrafluoroborate (Bu4NBF4) solution in dichloromethane containing 2 × 10−3 M of the dye. The experiments were carried out in a glove box.


The density functional theory (DFT) calculations were performed on the two organic sensitizers by using the Amsterdam Density Functional package (ADF 2010.01) [ADF2010, SCM, Theoretical Chemistry, Vrije Universiteit, The Netherlands Amsterdam, (scientific computing and modelling accessed 21-01-2014)]32,33. The structures of the dyes have been fully optimized at a GGA level using the Perdew-Burke-Ernzerhof (PBE) functional and triple zeta plus 2 polarization Slater functions (TZ2P set in ADF) basis sets for all atoms, with 1 s orbitals frozen for C, N and O atoms. On these optimized geometries, B3LYP hybrid functional calculations with TZ2P all electron basis sets were then performed to yield more reliable frontier orbital energies. Acetonitrile solvent was taken into account using a polarizable continuum model (COSMO in ADF)34 for all geometry optimizations and B3LYP single point calculations. Orbitals were drawn using the graphical interface of ADF (ADF-GUI).

Solar cell fabrication and characterization

For optimized DSC devices employing acetonitrile electrolyte, photoanodes consisted of a 13 μm layer of mesoporous TiO2 and a 4 μm scatter layer were prepared. Prior to the deposition of the TiO2 paste the conducting glass substrates were immersed in a solution of TiCl4 (40 mM) at 70°C for 30 min, washed with water and ethanol and then dried. The TiO2 nanoparticle paste was deposited onto a conducting glass substrate (NSG glass with 8 Ω cm−2 resistance) using the screen printing technique. The TiO2 electrodes were gradually heated under airflow at 325°C for 5 min, 375°C for 5 min, 450°C for 15 min and 500°C for 15 min. The heated TiO2 electrodes were immersed again in a solution of TiCl4 (40 mM) at 70°C for 30 min and then washed with water and ethanol. The electrodes were heated again at 500°C for 30 min and cooled before sensitization. In order to reduce scattered light from the edge of the glass electrodes of the dyed TiO2 layer, a light shading mask was used on the DSCs, so the active area of DSCs was fixed to 0.16 cm2. The counter electrode was made by spreading a 5 mM solution of H2PtCl6 in isopropyl alcohol onto a conducting glass substrate (TEC15, Pilkington) with a small hole to allow the introduction of the liquid electrolyte using vacuum, followed by heating at 390°C for 15 min. TiO2 films were dipped overnight in the corresponding dye solution. The sensitized electrodes were washed with ethanol and dried under air. Finally, the working and counter electrodes were sandwiched together using a thin thermoplastic (Surlyn) frame that melts at 100°C. The volatile electrolyte used consisted of 0.5 M 1-butyl-3- methylimidazolium iodide (BMII), 0.1 M lithium iodide, 0.05 M iodine and 0.5 M tert-butylpyridine in acetonitrile. The non-volatile electrolyte used consisted of (molar ratio) DMII (12), EMII (12), EMITCB (16), I2 (1.67), n Butyl benzimidazole (3.33), Guanidinium thiocyanate (0.67).

Charge extraction and transient photovoltage measurements

For charge extraction white light from a series of LEDs was used as the light source. When the LEDs are turned off, the cell is immediately short circuited and the charge is extracted allowing the electron density in the cells to be calculated. By changing the intensity of the LEDs, the electron density can be estimated as a function of cell voltage. In transient photovoltage measurements, in addition to the white light applied by the LEDS, a diode pulse (660 nm, 10 mW) is applied to the sample inducing a change of 2–3 mV within the cell. The resulting photovoltage decay transients are collected and the τ values are determined by fitting the data to the equation exp(−t/τ).