Ionic liquid (IL)-based electrolytes were prepared by adding 0.8 M of 1-methyl-3-propylimidazolium iodide (PMII, Fluka) in propylene carbonate (PC, Fluka). PC solvent is used in DSCs for its high dielectric constant 14 15 . Keeping the PMII composition stable, we examined the effect of the I₂ concentration by adding different quantities (0.02-0.08 M) of iodine (resublimed I₂, Aldrich) in the above mixture. Further addition of NMBI or TBP (Aldrich) was performed. When the optimum NMBI or TBP-based electrolyte compositions (0.45 M) were determined (respectively), 0.05 M of GuNCS (Aldrich) was added in each system 16 17 . To avoid performance instability and assure longer lifetime of the cells, no Li⁺ cations were added in the above mixture 15 .

To construct the DSCs, opaque TiO₂ films (about 12-15 μm thick, measured with a profilometer) were deposited by doctor-blading a paste of Degussa P25 powder on transparent conductive glass substrates (Pilkington, Active glass, 15 Ohm/square) and sintered at 450°C for 60 min in air 18 . For optimum photovoltaic performance studies, a second layer of large scattering particles (Ti-Nanoxide 300, Solaronix) was deposited on top of the first one and then the double-layered films were post-treated with TiCl₄. After final calcination at 550°C for 1 h, sensitization was achieved by immersing the photoelectrodes in a solution of standard N719 dye (Dyesol Ltd.). Open (non-sealed) DSCs were fabricated by putting a small drop of the novel liquid electrolytes onto the sensitized photoelectrode and simply sandwiching the Pt counter electrode against the first electrode. The techniques used for the photoelectrochemical investigation of the cells (J-V characteristics, Intensity Modulated Voltage Spectroscopy and Intensity Modulated Photocurrent Spectroscopy) are described in detail in 18 . It should be noted that a good batch of DSCs was tested and a mean value for the obtained results has been taken into account. Low statistical errors (see tables) validate the repeatability of the measurements and justify the representative outcome for our analysis.

After incorporation of the electrolytes inside the cells, DSCs were assembled and their J-V characteristics were determined. In a first attempt, the effect of the iodine concentration in the electrolyte (without any additive) was investigated. Table S1 in Additional file 1 presents the photovoltaic parameters obtained from the J-V characteristic curves (Figure S1 in Additional file 1) of the PMII-I₂-PC electrolytes, while Figure 1 presents analytically the dependence of the cell parameters (V _oc, ff, J _sc and η) on the iodine concentration. As we can see from Figure 1b, the short-circuit photocurrent density (J _sc) decreases systematically as the [I₂] increases, in agreement with literature results 19 . The above behaviour can be understood by assuming that whereas a specific critical level of I₃ ^- is necessary for cell functioning, further increase of [I₂] (or of the produced [I₃ ^-]), increases recombination at short-circuit conditions, thus reducing J _sc 20 .

When TBP was added in the PMII redox electrolyte (the J-V characteristic curves are shown in Figure S2 (Additional file 2), while Table S2 in Additional file 2 summarizes the photovoltaic parameters in detail), the V _oc of the corresponding cells was increased under all circumstances (Figure 1a). Most importantly, the tendency that was previously observed regarding the systematic reduction of J _sc with increasing [I₂] is generally preserved (Figure 1b). Again, the J _sc is significantly decreased (at least until 0.04 M), while a further increase of the iodine content does not significantly modify the observed J _sc.

To make a comparison between electrolytes containing different additives which play an identical effective role in DSCs, NMBI was now added in the pristine (no TBP) PMII electrolytes. The J-V characteristics of cells with NMBI are depicted in Figure S3 (Additional file 3), while Table S3 in Additional file 3 presents the photovoltaic parameters derived from the current-potential curves. By comparing the V _oc values in Figure 1a, it is thus evident that, like in the case of TBP, V _oc is increased for all [I₂] by 45-115 mV, in line with literature 11 21 . On the other hand, concerning the J _sc variation with [I₂], the behaviour (previously met with the no additives and TBP-based systems) changes now radically. The results reveal that for electrolytes containing [I₂] < 0.04 M, the addition of NMBI results in a remarkable decrease of J _sc (Figure 1b), in agreement with relevant literature results 22 . However, for [I₂] higher than 0.04 M, the J _sc was substantially increased and finally the highest photocurrent (9.57 mA/cm²) was obtained using the electrolyte with the highest concentration of iodine (0.08 M).

All the above results are quantified through Figure 1d that presents the overall power conversion efficiencies of the DSCs using electrolytes with I₂ in varying concentrations, with and without additives. By comparing Figure 1b, d, it is evident, then, that J _sc is the main factor determining the photovoltaic efficiency as we can see that both J _sc and η curves follow very similar trends. However, from Figure 1a, c it is clear that the other two parameters, V _oc and ff, are varied in a non-systematic way, when changing [I₂]. Another significant feature is that the highest efficiencies are obtained at [I₂] = 0.02 M when no additive is used in the electrolyte (η = 3.64%) or when TBP is added (η = 4.14%), Figure 1d. On the contrary, the highest efficiency for NMBI is obtained at very high iodine concentrations ([I₂] = 0.08 M). This value is very similar to the one measured with the TBP additive (η = 4.12%), but the relative efficiency increase is now spectacular (since η of the DSC at 0.08 M without NMBI is only 2.61%), leading us to the conclusion that the presence of NMBI additive smooths the iodine concentration effects.

The behaviour exhibited by the TBP-based versus the NMBI-based electrolyte (both are electron donating nitrogen heterocycles) can be easily understood if someone takes into account the differences in chemical affinity of the two organic additives with respect to the iodine 23 24 25 , fact that significantly affects their adsorption on the TiO₂ surface.

The DSCs presenting the optimum efficiency have been then used as reference cells for further experiments. To this end, GuNCS was added in the electrolytes already containing the TBP or NMBI additives to significantly increase the J _sc (and slightly the V _oc) of the cells. The J-V curves of the cells were drawn in Figure 2 while a comparative evaluation of the results is summarized on Tables 1 and 2. Additionally, Intensity Modulated Voltage Spectroscopy (IMVS) and Intensity Modulated Photocurrent Spectroscopy (IMPS) were used to extract valuable information about recombination kinetics under open-circuit and charge transport properties at short-circuit conditions, respectively 26 . Figure 3a, c presents the plots of the electron lifetime (τ_n) versus the photovoltage while Figure 3b, d displays the plots of the electron diffusion coefficients (D _e-) versus the photocurrent.

Taking a close look at the results of Table 1, it appears that the addition of the TBP leads to V _oc increase due to the synergistic effects of the negative E _c shift along with the increase of electrons lifetimes (Figure 3a), in perfect agreement with the literature 10 . Simultaneously, TBP decreases J _sc, due to the E _c shift (which also entails a smaller driving force for electron injection from the excited dye 10 ) as well as due to the reduced electron diffusion coefficients (Figure 3b). The extra addition of GuNCS in the TBP-based electrolytes increased further the gained V _oc (by only 17 mV in perfect agreement with the 20 mV increase observed by Kopidakis et al. in 8 ), whereas the recombination kinetics at open-circuit are not affected (no variation of the electron lifetimes, Figure 3a). However, and most importantly, the addition of GuNCS restored the values of J _sc by positively shifting the E _c of TiO₂ (that consequently increases the electrons injection efficiency 12 ), and this happens despite the fact that the D _e- were further decreased (Figure 3b). The V _oc invariance in conjunction with the large J _sc increase can be only explained by a large E _c shift towards positive potentials giving rise to a large injection efficiency under both open and short-circuit conditions or, most unlikely, by a Fermi level pinning.

On the contrary, the addition of NMBI in the electrolytes causes different effects. Figure 2b (and more clearly Table 2) proves that the addition of NMBI increases both V _oc and J _sc; the increase of J _sc is due to the enhanced electron diffusion coefficients (Figure 3d) and despite the fact that the conduction band edge of TiO₂ moves to more negative potentials 27 . On the other hand, V _oc also increases due to the reduction of the electrons back-reaction and this is expressed in increased values of electron lifetime (Figure 3c), in agreement with 11 . Furthermore, the addition of GuNCS in the NMBI-based electrolytes increased further the gained V _oc (by 19 mV in line with the V _oc increase observed in the TBP/GuNCS electrolytes), due to the additional increase of the electron lifetimes; the reduction of recombination seems to be strong enough to determine the V _oc of the cell (which is slightly increased, instead of an anticipated large decrease due to E _c shift towards positive potentials). The above results are in great accordance with Zhang et al. 12 . Most considerably, GuNCS increased the J _sc values (from 9.5 up to 11.2 mA/cm²) despite the decreased D _e- (Figure 3d); it seems that the E _c shift is the main parameter (and not electron transport kinetics) determining the overall photocurrent delivered by the cell.

It was thus concluded that although some additives (TBP and NMBI) have similar chemical properties, both shift the E _c towards negative values and reduce recombination but, finally, they act in a total different way in IL-based liquid electrolytes dissolved in PC solvent, affecting also electron transport kinetics and thus influencing J _sc. GuNCS was added in the electrolyte to compensate the E _c shifts and increase the photocurrent (while also slightly increase V _oc by reducing recombination). Even if this was indeed the result, GuNCS again acts in a different manner in each studied system (affecting recombination at open-circuit only in the NMBI system). The overall efficiencies with TBP-GuNCS and NMBI-GuNCS electrolytes was very similar reaching values up to 4.59 and 4.71%, respectively.

The above efficiencies were determined using non-optimized DSCs because we decided to stay in a system as simple as possible, in order to clearly discriminate the additive effects. Thus in a last step, we have used optimized TiO₂ films electrodes (with a three layer stratification of 22 μm in total thickness, incorporating Degussa P25, scattering layer and titania nanoparticles from TiCl₄ treatment) with enhanced electrooptical properties to increase the above efficiencies and find out, at the end, what is the maximum gain for a solar cell when using the two additives together in the electrolyte. From Figure 4, depicting the two I-V curves with optimum DSCs, one can infer that the electrolyte with NMBI-GuNCS gives a maximum efficiency of 5.78% (with J _sc = 13.7 mA/cm², V _oc = 762 mV and ff = 0.55). However, a much higher efficiency of 8.03%, resulting from much higher J _sc of the order of 17.5 mA/cm² (V _oc = 763 mV and ff = 0.60), is achieved when TBP and GuNCS are simultaneously incorporated inside the electrolyte. The above results are in great accordance with the optimum composition of electrolytes found in recent literature dealing with state-of-the-art cells 14 . In the case of the optimized DSCs, the much better performance of the TBP-based electrolyte (with respect to the cells using NMBI-based electrolyte) is mainly due to the particular stratification of the composite electrode. In fact, the adsorption behaviour of the TBP and NMBI additives on TiO₂ could be differentiated by the small-sized titania nanoparticles (produced following the TiCl₄ treatment), thus affecting in a different way the electron injection/recombination dynamics.