Optical determination of charge transfer times from indoline dyes to ZnO in solid state dye-sensitized solar cells

The functionalization of inorganic semiconductors with organic materials is a highly topical field of research in semiconductor physics. Organic molecules and layers on ZnO are promising functional hybrids for advanced optoelectronic devices. For example, dye-sensitized solar cells of the Grätzel type¹ are candidates for commercial photovoltaic applications. Despite the relatively low quantum efficiencies compared to silicon or perovskite based solar cells,² there is continuous interest in Grätzel type solar cells, as they are low cost devices and thus provide short energy payback times. Especially, solid state dye-sensitized solar cells are a candidate for stable and sustainable photovoltaic devices since they avoid liquid components in the cell. Recently, their efficiency was shown to be greatly improved by starting from a liquid based solar cell and letting the solvent evaporate very slowly.^3,4 An important process in high efficient dye-sensitized solar cells is an optimized electron transfer from the photoexcited dye into the conduction band of the semiconductor at the organic-inorganic interface.^5,6 Efficient electron injection requires fast injection relative to the recombination time of the dye. Kinetic competition between these two processes strongly influences the quantum efficiency. Usually, techniques such as femtosecond transient absorption spectroscopy are used to determine the kinetics of the injection process by measuring the absorption changes. In a recent paper⁷ we presented a way to study the electron transfer from the lowest excited exciton state by photoemission. We revealed a correlation between the transfer time and three parameters: (i) the number of anchoring groups, (ii) the distance between the dye and the organic-inorganic interface, which was varied by the length of an alkyl-chain between the carboxylate anchoring group and the dye, (iii) the thickness of the adsorbed dye layer. Another important parameter is the level alignment between the excited dye (π*-level) and the conduction band minimum or the Fermi-level of the inorganic semiconductor. It is known, that due to different ionization energies of both hybrid materials substantial interface dipoles can occur.^8–10 Furthermore, the Fermi-level of the mesoporous semiconductor depends on the specific electrolyte used in the solar cell.^11–13 In the present paper we report about the influence of external applied bias on the charge transfer using fully operational solid state solar cells. By studying the photoluminescence (PL) and particularly time resolved photoluminescence (TRPL) of the dyes we are able to reveal differences in the level alignment of the indoline dyes D149 and D131.

Unlike noted otherwise, all chemicals were purchased from Roth, Aldrich or Merck in ACS grade or higher and used without further purification. We studied two equally prepared solar cells either sensitized by D131 or D149. The schematic structure of the indoline derivatives is depicted in Fig. 1. Both dyes have one carboxyl anchor group but differ in the size of the π electron system.

The porous ZnO substrates were prepared by electrodeposition as described in detail elsewhere.¹⁴ In short, precleaned FTO coated glass substrates (Asahi glass, 10 Ohm/sq) were put in a three-electrode setup with Ag/AgCl as reference and a Pt wire as counter electrode in an aqueous solution of 0.1 M KCl and saturated with oxygen. At first, samples were pre-electrolysed by applying -1.05 V for 30 min. Afterwards, 5 mM ZnCl2 were added to the solution and a potential of -1.05 V was applied for 10 min to deposit a compact layer of ZnO. For the deposition of the porous ZnO 300 μM of EosinY were added and a potential of -0.75 V was applied for 20 min. The samples were stored overnight in an aqueous solution of KOH (pH 10.5) to remove the EosinY from the film. For the sensitization, the porous ZnO films were cleaned with water, dried at 100°C for 1 h and put in an UV/ozone cleaner for 30 min. Then the samples were immersed into a 0.5 mM solution of D131 (Chemicrea) or D149 (Chemicrea) in a 50:50 mixture of acetonitrile:tert-Butanol for 2 h and subsequently rinsed with ethanol and dried in air. Solar cells were prepared by attaching a pre-drilled ATO coated glass (Geomantec, 5 Ohm/sq) with sputtered Pt onto the sensitized electrode using a hot-melt Surlyn sealant with thickness of 30 μm. An electrolyte with 1 M tetrapropylammonium iodide and 0.1 M iodine dissolved in 1:4 acetonitrile:ethylen carbonate was poured into the cell. To obtain solid state solar cells, the electrolyte was allowed to slowly evaporate.

We performed photoluminescence (PL) measurements by means of a standard setup with a high-resolution grating spectrometer. For the steady-state PL, a laser diode of 442 nm (2.8 eV) was used. For the electrooptical spectroscopy a computer-controllable power-supply was used with integrated measuring devices for voltage and current. This setup allowed recording of detailed voltage-dependent cw PL spectra in reflection geometry with excitation at 2.1 eV (514 nm) combined with a notch filter to suppress the reflected laser beam. The time-resolved PL experiments were performed using a standard streak-camera setup. Details are given for example in Ref. 15. The 100-fs pulses of the tunable Ti:Sapphire laser were frequency doubled to 2.8 eV (442 nm) for most experiments; a 475 nm color-glass edge filter was used in the detection to suppress scattered laser light. Voltage dependent time-resolved PL measurements on processed dye sensitized solar cells have been performed with excitation at 3.06 eV (405 nm) in combination with an edge filter at 2.76 eV (450 nm).

In a recent paper we have shown, that the electron transfer times from the lowest excited dye state can be determined with high accuracy by studying the dye PL. A detailed discussion of the kinetic model is given in Ref. 7. In what follows, we will apply this model to the solar cells under consideration.

The PL spectra of both indoline derivatives on mesoporous ZnO in air at room temperature are depicted in the inset of Fig. 2. The PL maximum of D149 is seen at about 650 nm, whereas the PL peak of D131 at about 540nm is clearly blue shifted compared to D149. This is typical for the smaller π electron system. In Fig. 2 the transients of the PL are depicted for D131 and D149 on mesoporous ZnO. Both curves exhibit the typical dependence with an initially fast decay and a slow tail at later times. The experimental curves can be well fitted by the biexponential function Eq. (1) (see full black lines in Fig. 2).

Accordingly to our kinetic model the long time τ_exc is the excitonic lifetime comprising the radiative and nonradiative recombination probability of the dye. The short time τ_eff is caused by the fast electron transfer to ZnO in competition with the exciton recombination. Describing the electron transfer probability to the ZnO by w_trans = 1/τ_trans the resulting lifetime for the dye excitons is given by Eq. (2).

As discussed in detail in Ref. 7 only those excitons close enough to the interface can contribute to the charge transfer. Excitons which are too far away from the interface cannot dissociate. The annihilation of those excitons exhibit the time constant τ_exc yielding the weak but long living tail of the decay curve. The exciton lifetimes, the transfer times and the respective weighting factors are given in Table I. The weighting factor n_a gives the amount of excitons which are created by absorption close enough to the interface and can perform either a charge transfer or recombine radiatively. The weighting factor n_b gives the number of excitons which are too far away from the interface and will therefore recombine radiatively or nonradiatively, respectively. In case of D131 about 95% of the excited molecules are able to transfer the electron to ZnO and only 5% of the dye molecules are too far away. Though, the solar cells have been prepared equally, the amount of transferred electrons is substantially smaller in case of D149. Note, the electron transfer time to ZnO of D131 is substantially shorter compared to D149, i.e. the transfer probability of D131 is higher. This is a first indication for a higher lying excited-state of D131 yielding a higher electron injection rate due to the higher number of unoccupied states for electron injection above the ZnO Fermi-level. We come back to this point later again.

It is known for a long time that an increase in PL intensity can be observed upon negative electrical bias to the semiconductor electrode.^16–18 Such an increase in PL intensity is typically attributed to the reduced electron injection process from the dye. In Fig. 3 the PL intensities of solid-state solar cells with D131 and D149 on ZnO are depicted as function of the applied bias voltage. A total suppression of the PL was achieved in reversed bias at about -600 mV, i.e. the positive potential at the ZnO electrode. It can be seen, that in forward bias, i.e. the Fermi-level in ZnO is now shifted upwards, the PL intensity increases. It is important to note, that the shape of the PL is independent of the applied voltage as can be seen in the inset of Fig. 3 for D131 as example. The intensity increase is caused by a decreasing probability of the electron transfer. The maximum PL intensity, in other words the total suppression of electron transfer, was found at U_s=+750 mV and U_s=+850 mV for D149 and D131, respectively.

To reveal the influence of the applied bias on the electron transfer in more detail, we now apply our kinetic model on the PL transients of the D131 solar cell. In Fig. 4 the PL decay transients of D131 are depicted for various applied voltages and compared with calculated curves. Between 0 mV and the open circuit voltage V_oc=508 mV no significant changes of the decay curves could be found, although the integrated PL intensity is already somewhat higher at 508 mV compared to 0 mV. A significant change of the transients was found however between 508 mV and 750 mV. The overall decay becomes significantly slower. Between 750 mV and 800 mV the remaining changes are weak. Above 800 mV we could not detect further changes within the experimental errors. At a first glance the changes can be explained by a reduced electron transfer and respective increasing amount of recombining excitons. Actually, the transfer times determined by means of Eq. (1) increase with increasing voltage and the respective weighting factors decline (see upper part of Table II).

It should be noted, however, that the transients for bias voltages above 600 mV could no longer be fitted well by means of Eq. (1). A third exponential with increasing weighting factors was necessary at higher applied voltages. To make this clear, for U=600 mV and U=700 mV the best fits using Eq. (1) are depicted as dotted black lines in Fig. 4. It is obvious that for late decay a new slow decay component comes into play. Almost perfect fits could be achieved using a third exponential function using Eq. (3). All the respective parameters are given in the lower part of Table II.

Interestingly, the third component with long decay time τ_mol = 500 ps is the known lifetime of unbound dye molecules without any molecule-molecule interaction (see e.g. Ref. 7). It is known, that a certain portion of unbound molecules can be found in dye sensitized solar cells. It is interesting to note, however, that the amount of unbound molecules increases drastically from very small amounts 2% at 508mV to about 31% at 800mV. The molecules adsorb again by reducing the applied voltage. It is worth mentioning, that Lemon and Hupp¹⁹ reported a similar result on the basis of electrochemical quartz crystal microbalance experiments. They observed reversible potential-induced dye desorption from nanocrystalline TiO₂.

We now consider the variation of the transfer times by applying an increasing negative potential at the ZnO electrode, i.e. shifting the Fermi-level in ZnO upwards. It can be seen in Table II, the number of transferred electrons declines tremendously with increasing voltage whereas the emission of excitons without electron transfer increases respectively. The decay curve measured at 800 mV is the first curve, which can be fitted without any electron transfer (parameters see lowest line in Table II), i.e. the transfer would be completely suppressed. Unfortunately, the fit of the curve is not unique in this case. A complete suppression of the transfer is just a limiting case. An almost identical curve could be achieved assuming a weak remaining electron transfer (full black line in Fig. 4). The respective values are also given in Table II. The electron transfer time τ_trans = 140 ps approaches the radiative lifetime and the number of transferred electrons is reduced to 13%. Both limiting cases, no transfer on the one side and a remaining transfer of about 10% on the other side were applicable for all transients measured even at higher voltages. Actually, we can not exclude such remaining electron transfer at higher voltages. The reduction of the oxidized dye by electrons from ZnO or interface states is possible. This would provide a certain number of unoccupied states below the Fermi-level, which can be refilled either by electrons from π^⋆ or by refilling from upper lying but filled electron states of ZnO.

The shorter electron transfer times from D131 to ZnO compared to D149 are caused by a different level alignment of the π^⋆ to the quasi Fermi-level in ZnO as depicted in Fig. 5. Unfortunately, it is not possible to reveal the absolute offset directly from the electro-optical measurements. However, a relative consideration of the D131 and D149 solar cells is possible, as both cells have been prepared equally. We can assume therefore equal positions of the Fermi-energy in ZnO. In so far, the voltage difference needed to suppress the electron transfer can be used to estimate the difference of the ZnO conduction band and the π^⋆ energy of the dyes. The maximum of the dye PL is reached if the ZnO Fermi-energy reaches a level, where an electron transfer from the dye is no longer possible. It is important to note, that the binding energy due to the coulomb interaction must be taken into account to release electrons completely from ionized dye molecules. As discussed above, the intensity curves of D149 and D131 in Fig. 3 are shifted by 100 mV. We can conclude therefore, that the excited state π^⋆ of D131 lies about 100 meV above the respective state of D149. The respective band alignments of D149 or D131 and the ZnO conduction band edge is depicted in Fig. 5.

Care needs to be taken comparing energy levels from the literature for analyses of the actual energetic situation. The values are often taken from measurements of the individual components, excluding effects due to adsorption of the dye and electrolyte species to the oxide surface. For example, Matsui et al.²⁰ determined by means of cyclic voltammetry for the energy difference between the ZnO conduction band and HOMO of D131 and D149 $ECBZnO−Eox=1.42eV$ and $ECBZnO−Eox=1.35eV$⁠, respectively. However, in the here presented method a direct measure for the relative LUMO for two similar dyes position is presented when adsorbed to a porous semiconductor. Cyclic voltammetry measurements of dissolved D149 and D131 dyes reveal a LUMO difference of around 240 meV²⁰ which somewhat differs from the here obtained 100 meV highlighting the importance of in situ measured energy levels of adsorbed molecules.

In summary, we studied and compared the charge transfer from optically excited indoline dyes D131 and D149 into mesoporous ZnO by applying an external electrical bias. We were able to reveal the electron transfer times using time resolved photoluminescence in combination with electrical bias of solid state solar cells. The transfer probability of electrons from the excited dye to ZnO can be strongly reduced by shifting the Fermi-level of the ZnO to higher energies. In case of D131 the transfer time can be tuned from τ_trans = 18 ps to τ_trans = 140 ps. The longer times are caused by the reduced number of unoccupied states available for electron injection. At open circuit voltage, the transfer time of D131 is τ_trans = 23 ps and substantially faster than the transfer time of D149 with τ_trans = 61 ps. This is due to the different level alignment. The excited state π^⋆ of D131 is about 100 meV above the respective state of D149.

We are grateful to financial support by the DFG (Deutsche Forschungsgemeinschaft) in the framework of the SFB 1083 (IM and WH) and within the project SCHL340/19-1 and the GRK (Research Training Group) 2204 (RR, MR and DS). The authors would like to thank H. Miura (Chemicrea Co., Japan) for supplying the dyes D149 and D131.