Molecular engineering on semiconductor surfaces: design, synthesis and application of new efficient amphiphilic ruthenium photosensitizers for nanocrystalline TiO₂ solar cells

Abstract

A new series of amphiphilic heteroleptic ruthenium(II) sensitizers [Ru(H₂dcbpy)(dhbpy)(NCS)₂] (C1), [Ru(H₂dcbpy)(dchobpy)(NCS)₂] (C2), [Ru(H₂dcbpy)(mubpy)(NCS)₂] (C3), [Ru(H₂dcbpy)(dhabpy)(NCS)₂] (C4) have been developed and fully characterized by UV-visible, emission, NMR and cyclic voltammetric studies (where mubpy=4-methyl-4′-perfluoro 1H, 1H, 2H, 2H, 3H, 3H undecyl-2,2′-bipyridine, dhabpy=4-4′-dihexylamide-2,2′-bipyridine, dhbpy=4-4′-dihexyl-2,2′-bipyridine, dchobpy=4-4′-dicholesteryl-2,2′-bipyridine). The amphiphilic amide heteroleptic ruthenium(II) sensitizers, when anchored to nanocrystalline TiO₂ films from ethanol solutions, achieve very efficient sensitization in the visible range yielding ≈80% incident photon-to-current conversion efficiency (IPCE) and display a short circuit photocurrent density of 15 mA/cm² and an open circuit voltage of 0.75 V corresponding to an overall conversion efficiency of 7.4% under standard AM 1.5 sunlight.

Introduction

Photosynthesis converts only 0.02–0.05% of the incident solar energy of about 10²² kJ per year into biological material. This is 100 times more than the food needed for mankind [1]. In green plants and purple bacteria, photosynthetic units harvest light to induce electron transfer to finally initiate consecutive chemical conversion [2], [3]. The importance of light harvesting in such systems explains why currently there is a considerable interest in ruthenium polypyridyl [4], [5], [6] complexes used to sensitize nanocrystalline TiO₂ thin films based solar cells [7], [8], [9], [10]. The common homoleptic cis-dithiocyanatobis(2,2′-bipyridyl-4,4′-dicarboxylate)ruthenium(II) complex, also known as N3, shows a high incident photon-to-current conversion efficiency (IPCE) yielding close to 10% under AM 1.5 solar conditions [11]. The class of heteroleptic complexes opened the field of dye molecular engineering on TiO₂, because it became possible to mix in one dye anchoring groups like H₂dcbpy (H₂dcbpy=4,4′-dicarboxy-2,2′-bipyridine) as well as electron donating and/or protective groups held on the second bipyridine. Alkyl chains have been introduced and opened the generation of amphiphilic heteroleptic ruthenium polypyridyl photosensitizers [12]. The presence of hydrophobic chains afforded a strong improvement in the stability of solar cells performances emerging from the water insolubility of these dyes. In this paper we report a new series of heteroleptic amphiphilic ruthenium dyes [Ru(H₂dcbpy)(dhbpy)(NCS)₂] (C1), [Ru(H₂dcbpy)(dchobpy)(NCS)₂] (C2), [Ru(H₂dcbpy)(mubpy)(NCS)₂] (C3), [Ru(H₂dcbpy)(dhabpy)(NCS)₂] (C4) (Fig. 4) based on the idea to use weak forces (van der Waals contacts and hydrogen bonds) to induce self-organization on the surface of TiO₂ nanocrystals.

The rate of electron transport in dye-sensitized solar cells is a major element of the overall efficiency of the cells. The cell is composed of a nanocrystalline TiO₂ layer on which the dye is attached by carboxylic groups (Fig. 1). When exposed to light, the dye interacts with photons and goes to an excited state sufficiently energetic to inject an electron into the TiO₂ conduction band. This electron can be collected at the transparent conducting glass (ideal case) or can return to the Ru(III) center (charge recombination reaction) or can even be trapped by the redox mediator (dark currents generated).

One possible way of avoiding such dark currents is to cover the surface, after the adsorption of dye, with hydrophobic chains. These chains beeing previously attached on the Ru complex (Fig. 2). The strategy is in accordance with the idea expressed by Francis Garnier [13] in which “molecular materials formed by assemblies of molecules held together by weak forces implies that the properties of the solid are directly governed by those of the individual molecules which form the elemental”.

Physical techniques have been developed toward this aim such as Langmuir–Blodgett deposition [14]. A chemical route can also be used based on self recognition properties inducing a spatial ordering of the constituting molecules. This explains why we used increasing energy interactions emerging respectively from hydrogenated alkyl, cholesteryl and perfluorinated chains followed by hydrogen bonding induced by alkyl amide groups. In the particular case of fluorinated bipyridine, two parameters should be controlled. The fluorine load should be enough to increase the hydrophobic character (compares to hydrocarbon chains) without strongly increasing the fluorophilic affinity [15], [16], [17]. However, the fluorinated compound will only be soluble in perfluorinated solvents. Another crucial modification is the introduction of several methylene groups (spacers) between the fourth position of the 2,2′-bipyridine and the first CF₂ group of the fluoro tail, in order to mitigate the powerful electron withdrawing effect of the CF₂ groups that can decrease electron density on the ligand and reduce the IPCE expected value of the dye. Following these requirements we introduced only one perfluoro chain on fourth position with three CH₂ groups spacers.

In the case of bis amide bipyridine LA6 a six carbons chain length was chosen to avoid insolubility problems usually observed with amide structures.

Experimentally, syntheses of ligands LF8 and L6 were done using classical coupling reactions involving lithium diisopropylamide (LDA) at −78 °C (Fig. 3). The mono- and di-anion obtained from the reaction of 4,4′-dimethyl-2,2′-bipyridine with LDA¹ at −78 °C were treated in situ with commercially available I(CH₂)₂(CF₂)₇CF₃, hexylbromide. After the reaction was quenched with water the desired bipyridine was extracted with dichloromethane and purified on silica column chromatography to afford respectively LF8 and L6 in 50 and 60% yield. The 4-4′-dihexylamide-2,2′-bipyridine (dhabpy=LA6) was obtained following a described procedure [18] involving 4,4′-dicarbonyl chloride-2,2′-bipyridine [19] and an excess of a freshly purified amine at room temperature for 24 h in dry toluene. The 4,4′-dicholesteryl-2,2′-bipyridine LC was obtained from the reaction of cholesterol (from lanolin: Fluka) with 4-4′-dicarbonyl chloride-2,2′-bipyridine. Respective yields were 80 and 60%.

Fig. 4 shows the structure of complexes C1, C2, C3 and C4 which were obtained by refluxing an excess of NCS⁻ ligand with the corresponding heteroleptic dichloro complexes using a published procedure [20] (see Section 3 for synthetic details). Table 1 shows UV-visible, emission and electrochemical data of complexes C1–C6, which were measured in ethanol. The two distinct absorption bands in the UV region at 296 and 314 nm are due to the second bipyridine (L6, LC, LF8, LA6, mhdbpy) and H₂dcbpy ligands centered π–π^∗ transitions, respectively. Bands between 390 to 530 nm region are due to metal-to-ligand charge transfer (MLCT) transitions. MLCT band of C2 reach a maximum at 551 nm. The presence of ester groups lower the energy of the LUMO. Presence of amide groups on the bipy also generate a reduction of the MLCT energy. We can observe that N3 and C4 complexes present almost the same UV properties with, however, a lower extinction coefficient for C4. For the complex C3, UV values are similar to C1, C5 and C6. Moreover emission values for C3 and C5 complexes, obtained in ethanol by excitation at 520 nm, are similar. We can conclude that the covalently attached fluorinated chain does not affect the orbitals of pyridines rings. Our hypothesis to block the electron withdrawing effect of fluorinated chain by intercaling three CH₂ groups is valid and confirmed.

Cyclic voltammograms of complexes C1–C6 in degazed dimethylformamide (DMF) show respectively reversible waves at 0.72, 0.79, 0.78, 0.80, 0.70, 0.74 V vs. Ag/AgCl, at room temperature, on a glassy carbon electrode at a scan rate of 200 mV s⁻¹, which can be readily assigned to the Ru(II)/(III) couple. The ratio between the oxidation peak current and the reduction peak current is close to one. In the case of amide and ester complexes the potentials are around 100 mV more positive than aliphatiques complexes C1 or C6. Not only ester, even amide groups tend to lower electron density on the bipyridine. Nevertheless, they remain 50–60 mV below the N3 value [21]. On the cathodic side, there are quasireversible peaks around −1.5 V corresponding to the reduction of H₂dcbpy ligands.

Photovoltaic measurements were done using 12 μm thick TiO₂ electrodes. In a previous study [12] the IR data showed unambiguously that C5 is adsorbed on the surface using the two carboxylate groups as a bidentate chelation or bridging mode rather than an ester type linkage [22]. The nanocrystalline TiO₂ (anatase) films were prepared on conducting glass using a previously described procedure [23]. The electrodes were heated up to 450 °C for 10 min and then allowed to cool to ≈50 °C before dipping into the dye solution (3×10⁻⁴ M in ethanol) for 20 h. The dark red colored films were tested in photovoltaic cells in conjunction with a redox electrolyte composed by 0.6 M butylmethylimidazolium, 50 mM iodine, 500 mM t-butylpyridine and 100 mM lithium iodide in a 1 to 1 solvent mixture of acetonitrile–valeronitrile.

On the four new complexes synthesized the bis amide complex C4 presents the best photovoltaic results with a photocurrent density of 15.5 mA/cm² and an open circuit potential of 706 mV. With another electrolyte² we obtained a better potential Voc=750 mV and Jsc=14.9 mA. Fig. 5 shows the photocurrent action spectrum of the cell containing sensitizer C4 where the incident photon to current conversion efficiency is plotted as a function of wavelength and reaches 75%.

For C2 we observed a low Jsc value equal to 8.5 mA/cm² and Voc equal to 570 mV. Low values induced by only 40% of IPCE. We have seen in UV data that ester groups directly attached to the bipyridine lower the LUMO and, consequently, after light excitation the electron went on the carboxy H₂dcbpy and also on the ester dchobpy bipyridine. It is not surprising to see this effect on the LUMO in the case of ester but it is surprising not to see it in the amide C4 complexes. Particularly, when we also observe a 8–10 nm red shift of the MLCT in the case of C4 with respect to aliphatic complexes C1, C5, C6. Finally, the fluorinated complex C3 gave a better value compared to ester but a worse compared to his hydrogenated equivalent C5. Indeed the value is decreased by 20–25% and IPCE is equal to 54% at 540 nm. We can not argue the same reason as for ester C2 complex because absorption bands energy of C3 and C5 are identical and not red shifted. One explanation may be that the C3 complex once adsorbed on the surface generates a network of electropositive carbon tails. The reaction of conduction band electrons with electropositive carbon tails significantly reduces the charge-collection efficiency and there-by decreases the total efficiency of the cell [24].