Combining phosphonic acid-functionalized anchoring ligands with asymmetric ancillary ligands in bis(diimine)copper(i) dyes for dye-sensitized solar cells

A ‘surfaces-as-ligands’ strategy is used to assemble heteroleptic copper(I) dyes [Cu(Lanchor)(Lancillary)] + on FTO/ TiO2 electrodes for dye-sensitized solar cells (DSCs). The anchoring domain, Lanchor, is either ((6,60-diphenyl[2,20-bipyridine]-4,40-diyl)bis(4,1-phenylene))bis(phosphonic acid) (1) or ((6,60-dimethyl-[2,20-bipyridine]4,40-diyl)bis(4,1-phenylene))bis(phosphonic acid) (2). Asymmetric ancillary ligands with a 2,20-bipyridine metal-binding domain are used to counter the sterically-demanding 6,60-diphenyl-substitution pattern in anchor 1; Lancillary 1⁄4 6-methyl-4-phenyl-2,20-bipyridine (3), 6-methyl-4-(4-bromophenyl)-2,20-bipyridine (4), 6-methyl-4-(4-methoxyphenyl)-2,20-bipyridine (5) or 6-methyl-4-(3,4,5-trimethoxyphenyl)-2,20bipyridine (6). Solid-state absorption spectra of adsorbed [Cu(1)(Lancillary)] + and [Cu(2)(Lancillary)] , and external quantum efficiency (EQE) spectra of DSCs containing these dyes confirm that incorporation of 6,60-diphenyl-substituted 1 leads to a broadened spectral response at lower energies compared to dyes with anchor 2; dye-loading is higher for [Cu(2)(Lancillary)] + than for [Cu(1)(Lancillary)] , and EQEmax is >41% for [Cu(2)(Lancillary)] + compared to <12% for [Cu(1)(Lancillary)] . Enhanced values of the short-circuit current density (JSC) are observed on going from anchor 1 to 2, independent of Lancillary. For the series of [Cu(2)(Lancillary)] + dyes, photoconversion efficiencies (confirmed using four DSCs per dye) vary with Lancillary in the order 3 5 > 6 > 4 on the day of DSC assembly, and 5 > 3 6 > 4 after a week. The best performing DSCs achieve efficiencies of 37% relative to N719 set at 100%.


Introduction
Since the discovery by Sauvage and coworkers 1 that copper(I) sensitizers combined with wide band-gap semiconducting metal oxides such as TiO 2 or ZnO could be used for photoconversion, signicant progress has been made in the development of homoleptic [Cu(N^N) 2 ] + and heteroleptic [Cu(N^N)(N^N) 0 ] + sensitizers (N^N ¼ diimine ligand) in dyesensitized solar cells (DSCs). [2][3][4] The advantages of copper dyes over conventional ruthenium dyes lie in the Earth abundance and lower cost of copper over ruthenium. However, while photon-to-electrical current conversion efficiencies (h) for ruthenium dyes reach z12%, 5-7 the efficiencies of DSCs containing copper(I) dyes have only recently surpassed 3% (relative to z7.5% for the dye N719 used as a reference). 8,9 In an n-type DSC, heteroleptic [Cu(N^N)(N^N) 0 ] + dyes anchored to TiO 2 are preferred to homoleptic dyes because their electronic properties are readily tuned to a 'push-pull' design to encourage electron transfer from electrolyte, through the dye to the semiconductor. One ligand (L anchor ) in the heteroleptic dye is designed to anchor the dye to the TiO 2 surface, and the second (L ancillary ) can be variously functionalized. The exceptional global efficiencies of 4.42% and 4.66% reported by Odobel and coworkers 9 were achieved using the heteroleptic [Cu(L anchor )(L ancillary )] + dyes shown in Scheme 1 in which both anchoring and ancillary ligands contain 6,6 0 -substituted 2,2 0 -bipyridine (bpy) metalbinding domains. The use of bulky mesityl groups in one ligand stabilizes the copper(I) complex with respect to ligand dissociation, permitting isolation of the dye 9,10 before TiO 2surface functionalization. In contrast, we have applied the lability of bis(diimine) copper(I) complexes to assemble sensitizers on TiO 2 in our 'surfaces-as-ligands' approach. 11 As part of our continued efforts to enhance the performance of copper(I) sensitizers, we recently described the use of the phosphonic acid anchor 1 (Scheme 2) which contains phenyl substituents in the 6,6 0 -positions of the bpy unit. 12 We have previously shown that phosphonic acid anchoring groups are preferred over carboxylic acids in copper(I) dyes, 13,14 and the 6,6 0diphenyl substituents in 1 not only help to shield the copper(I) centre but also improve light absorption towards the red-end of the visible spectrum. 15 One of the major hurdles to overcome with copper(I) dyes is broadening their spectral responses. Comparisons of DSC performances of dyes with anchor 1 or 2 (the 6,6 0 -dimethyl analogue of 1, Scheme 2) and common ancillary ligands which contained 6,6 0 -disubstituted bpy units conrmed 2 to be the preferred anchor. 12 We observed 12 that dyes with anchor 1 bleached in the presence of I À /I 3 À electrolyte or iodide ion. We have now investigated these observations further using asymmetrical ancillary ligands 3-6 (Scheme 3) to eliminate possible effects of steric crowding in the copper(I) bis(diimine) coordination sphere when both ligands are 6,6 0disubstituted.  13 C NMR (126 MHz, CD 2 Cl 2 ) d/ppm 158.2 (C B6 ), 153.0 (C A2/B2 ), 152.4 (C A2/B2 ), 151.1 (C B4 ), 149.2 (C A6 ), 138.6 (C A4 ), 137.6 (C C1 ), 130.5 (C C4 ), 129.9 (C C3 ), 127.7 (C C2 ), 126.8 (C A5 ), 124.3 (C B5 ), 122.6 (C A3 ), 117.7 (C B3 ), 25 13 13

DSC fabrication
Commercial TiO 2 electrodes (Solaronix Test Cell Titania Electrodes) were washed with miliQ H 2 O and HPLC grade EtOH, then heated at 450 C for 30 min; aer cooling to z80 C, they were immersed in a DMSO solution of 1 (1.0 mM) and le for 24 h at ambient temperature. The electrodes were then taken out of the solution, washed with DMSO and EtOH and dried in a stream of N 2 . Each electrode was then soaked in a CH 2 Cl 2 solution of [Cu(L ancillary ) 2 ][PF 6 ] (L ancillary ¼ 3-6, 0.1 mM) for 3 days at room temperature. Aer the electrodes had been taken out of the dye-bath, they were washed with CH 2 Cl 2 and then dried in a stream of N 2 . N719 (Solaronix) reference electrodes were made by dipping TiO 2 electrodes (Solaronix Test Cell Titania Electrodes) in an EtOH solution of N719 (0.3 mM) for 3 days. Aer removal of the electrodes from the solution, they were washed with EtOH and dried in a stream of N 2 . Commercial counter electrodes (Solaronix Test Cell Platinum Electrodes) were washed with EtOH, and then heated on a hot plate at 450 C for 30 min to remove volatile organic impurities.
The working and counter-electrodes were combined with thermoplast hot-melt sealing foil (Solaronix Test Cell Gaskets, 60 mm) by heating while pressing them together. The electrolyte (LiI (0.1 M), I 2 (0.05 M), 1-methylbenzimidazole (0.5 M), 1-butyl-3-methylimidazolinium iodide (0.6 M) in 3-methoxypropionitrile) was added to the device by vacuum backlling through a hole in the counter electrode that was then sealed (Solaronix Test Cell Sealings) and covered (Solaronix Test Cell Caps).

Electrodes for solid-state absorption spectroscopy
Dye-functionalized electrodes were assembled as described above replacing the Solaronix Test Cell Titania Electrodes by Solaronix Test Cell Titania Electrodes Transparent.

DSC and external quantum efficiency (EQE) measurements
The DSCs were masked for measurements; the mask was made from a black-coloured copper sheet with an aperture of average area 0.06012 cm 2 (standard deviation ¼ 1%) placed over the active area of the cell. The area of the aperture in the mask was smaller than the surface area of TiO 2 (0.36 cm 2 ). Black tape was used to complete the masking of the cell. Performance measurements were made by irradiating the DSC from behind with a LOT Quantum Design LS0811 instrument (100 mW cm À2 ¼ 1 sun), and the simulated light power was calibrated with a silicon reference cell.
The external quantum efficiency (EQE) measurements were made using a Spe-Quest quantum efficiency setup (Rera Systems, Netherlands) operating with a 100 W halogen lamp (QTH) and a lambda 300 grating monochromator (Lot Oriel). The monochromatic light was modulated to 3 Hz using a chopper wheel (ThorLabs), and the cell response was amplied with a large dynamic range IV converter (CVI Melles Griot) and then measured with a SR830 DSP Lock-In amplier (Stanford Research).

Ligand syntheses and characterizations
We have previously described the synthesis of the anchoring ligand 1 from the corresponding tetraethyl ester. 12 Deprotection of tetraethyl ((6,6 0 -dimethyl-[2,2 0 -bipyridine]-4,4 0 -diyl)bis(4,1phenylene))bis(phosphonate) is carried out using concentrated aqueous HCl followed by treatment with glacial acetic acid. This reaction takes 3 days, and we present here an alternative route involving treatment of the ester with Me 3 SiBr which gives acid 1 in 58.6% yield within 24 hours.
The Kröhnke strategy 21 was used to prepare compounds 3-6, and the synthetic route is summarized in Scheme 4; a stoichiometric amount of 1-(2-oxopropyl)pyridinium chloride was combined with intermediate 3a, 4a, 5a or 6a. Ligand 3 has previously been prepared in 25% yield, 22 but by using the Kröhnke method, we were able to isolate 3 in 79.5% yield. The base peak in the electrospray mass spectrum of each of 3-6 arose from the [M + H] + and the isotope pattern matched that simulated. The 1 H and 13 C NMR spectra of the ligands are consistent with the asymmetric structures shown in Scheme 2 and were assigned using COSY, NOESY, HMQC and HMBC methods; the spectrum of ligand 5 is shown in Fig. 1a as a representative example. In the NOESY spectrum for 5, cross peaks between H C2 and H B3 , and between H C2 and H B5 allowed proton H C2 to be distinguished from H C3 ; this was conrmed by a NOESY peak between H C3 and H OMe ; an H B5 -H Me(ring B) NOESY cross peak distinguished H B3 from H B5 . Signals in the 1 H NMR spectra of 3, 4 and 6 were assigned in a similar manner.

Copper(I) complexes
The homoleptic complexes [Cu(L ancillary ) 2 ][PF 6 ] with L ancillary ¼ 3-6 were prepared by reaction of [Cu(NCMe) 4 ][PF 6 ] with two equivalents of ligand and were isolated as dark red-orange solids in 71.2-90.0% yield. The highest mass peak in the electrospray mass spectrum of each complex corresponded to the [M À PF 6 ] + ion and showed a characteristic isotope pattern corresponding to the simulated pattern. The 1 H and 13 C NMR spectra of the complexes were recorded in CD 2 Cl 2 , and assigned using 2D methods; NOESY spectra were used to distinguish between pairs of protons H B3 /H B5 and H C2 /H C3 as described earlier. Fig. 1 shows a comparison of the 1 H NMR spectrum of [Cu(5) 2 ][PF 6 ] with that of the free ligand (in CD 3 CN), and similar changes (e.g. a shi to lower frequency for H A6 and H Me ) occur upon coordination of each ligand. Proton signals for the peripheral phenyl substituent are little affected by coordination to copper.
The solution absorption spectra of the complexes are compared in Fig. 2. The absorption maxima are summarized in Table 1 and were veried by recording the spectra at different concentrations. The spectra are dominated by high-energy bands assigned to p* ) p transitions and a broad metal-toligand charge transfer (MLCT) band with l max z 470 nm ( Table 1). The extinction coefficient of 9300 dm 3 mol À1 cm À1 for the MLCT band in [Cu (6)    drying, the electrodes were placed in a tailor-made holder in the spectrophotometer and the absorption spectra recorded in transmission mode. Duplicate electrodes were used to validate the data, and Fig. 3a shows the spectra of one set of electrodes.
The four dyes with anchoring ligand 2 exhibit MLCT absorptions with l max ¼ 466-468 nm. Replacing the 6,6 0 -methyl groups by 6,6 0 -phenyl substituents on going from anchor 2 to 1 leads to a reduction in the intensity of the MLCT bands while gaining slightly at longer wavelengths ( Fig. 3a and b). The data are consistent with a lower dye-loading on increasing the steric requirements of the anchoring ligand. By eye, the electrodes with adsorbed dyes [Cu(2)(3)] + to [Cu(2)(6)] + appear a more intense orange colour than those with [Cu(1)(3)] + to [Cu(1)(6)] + ; this is seen in the photographs of the electrodes shown in Fig. 4a. An electrode with N719 is shown in Fig. 4a for comparison. In Fig. 4b, the absorption spectra of TiO 2 functionalized with [Cu(1)(3)] + and [Cu(2)(3)] + are compared with that of an electrode with adsorbed N719 and underline the fact that light-harvesting by the copper(I) dyes lacks contributions from the red-end of the visible region in particular.     Tables 2  and 3 gives the photoconversion efficiency relative to a reference DSC containing N719 as the sensitizer; the value of h for this cell was set at 100%. This is a procedure that we use routinely to permit comparisons of data between different light sources (see later). 26 Fig. 6 shows current density (J) versus potential (V) plots for the DSCs showing the best performances for each ligand combination in a [Cu(L anchor )(L ancillary )] + dye.

DSC performances
Firstly, we consider the effect of changing the 6,6 0 -substituents in the anchoring ligand. The data in Tables 2 and 3, and the J-V curves in Fig. 6 reveal that the dyes containing the 6,6 0diphenyl-substituted anchoring ligand 1 perform poorly with    respect to those with the 6,6 0 -dimethyl-substituted anchor 2, both in terms of values of J SC and V OC . The trend in values of J SC upon changing the anchoring ligand are corroborated by the external quantum efficiency (EQE) spectra shown in Fig. 7. Although use of the 6,6 0 -diphenyl-substituted 1 leads to a broad spectral response between 420 and 630 nm (Fig. 7) 12 On the day of cell fabrication, DSCs containing anchor 2 with ancillary ligands 3-6 achieve overall efficiencies, h, of between 27.9 and 38.4% relative to that of the reference dye N719 set at 100% (Table 3); the efficiencies decrease somewhat over a period of a week, but remain $30% that of N719 set at 100%. In contrast, use of the more sterically demanding anchor 1 results in relative efficiencies of #11.0%. Secondly, we focus on the dyes containing the 6,6 0 -dimethylsubstituted anchor 2. The performance parameters for DSCs containing [Cu(2)(4)] + may be compared with those with the dye [Cu (2)(8) The incorporation of the electron-releasing methoxy substituent in the 4-position of the bpy unit in 5 is benecial, as expected in terms of the desirable 'push-pull' characteristics of a dye for an n-type DSC. Use of [Cu(2)(5)] + leads to enhanced values of V OC with respect to [Cu(2)(4)] + (Fig. S1 † and 6). Overall, the efficiencies of DSCs containing [Cu(2)(5)] + are higher than those with [Cu(2)(6)] + and this can be understood in terms of the dominant +M effect of the 4-methoxy substituent being partially countered by the ÀI effects of the 3,5-dimethoxy groups. Data in Table 3 show that this trend is maintained over a week aer the cells were assembled and that the gain in performance originates from enhanced J SC . Although values of EQE max for DSCs with [Cu(2)(5)] + and [Cu(2)(6)] + are the same (Fig. 7), the EQE spectrum for the DSC with [Cu(2)(5)] + is appreciably extended towards the red-end of the spectrum, consistent with higher J SC values for cells sensitized with [Cu(2)(5)] + . An unexpected observation is that DSCs containing [Cu(2)(3)] + perform well, especially on the day of cell fabrication. The relatively high values of J SC (Table 3, Fig. S1 † and 7) are an important contributing factor, and effective values of V OC are also achieved ( Table 3, Fig. S1 † and 6). Simple ancillary ligands such as 3 (Scheme 3) are oen overlooked in a desire to design 'pushpull' dyes for the n-type semiconductor surface. 2 However, the present results suggest that such ancillary ligands deserve further exploration. Table 3 shows that over a 7 day period aer the DSCs were made, the cells containing [Cu(2)(L ancillary )] + (L ancillary ¼ 3-6) exhibited stable photoconversion performances, whereas Table  2 reveals that the performances of cells with [Cu(1)(L ancillary )] + generally decayed. By eye, the red-orange colour of the photoanodes in the DSCs containing anchor 1 bleached. To investigate the process, solid-state absorption spectra of transparent FTO/TiO 2 electrodes functionalized with [Cu(1)(4)] + were recorded and then the electrodes were soaked 15 min in either a 3methoxypropionitrile solution of LiI or in a solution of the standard I À /I 3 À electrolyte used for the DSCs. In both cases, the orange colour of the dye bleached (Fig. 8, le) and the MLCT band arising from the adsorbed dye decreased in intensity ( Fig. 9), consistent with our previous proposal that bleaching is caused by attack at the copper(I) centre by I À ion. 12 The electrodes were then soaked in either MeCN solutions of    6 ]. Aer removal from the dye-baths, the electrodes were dried and washed with CH 2 Cl 2 and dried. This dye-bath step regenerated the characteristic orange colour (Fig. 8, right) of a bis(diimine) copper(I) complex. The spectra in Fig. 9 conrm that the anchor 1 remains on the surface during the bleaching process and is able to bind copper(I) and ancillary ligand (Fig. 9a) or undergo ligand exchange with the homoleptic complex (Fig. 9b) to regenerate the surface-bound [Cu(1)(4)] + dye. These observations indicate that our previous report of copper(I) dye-regeneration on bleached photoanodes 12 may be applied more generally.

Conclusions
The homoleptic complexes [Cu (3) 6 ] containing asymmetric diimine ligands have been prepared and characterized. By use of a 'surfaces-as-ligands' strategy, the heteroleptic dyes [Cu(1)(L ancillary )] + and [Cu(2)(L ancillary )] + with L ancillary ¼ 3, 4, 5 or 6 were assembled on FTO/TiO 2 electrodes. The phosphonic acid anchoring ligands 1 and 2 differ in having phenyl or methyl substituents in the 6-and 6 0 -positions of the bpy metal-binding domain. The reproducibility of DSC performances was conrmed by using multiple devices for each dye. Solid-state absorption spectra of adsorbed [Cu(1)(L ancillary )] + and [Cu(2)(L ancillary )] + , and external quantum efficiency (EQE) spectra of DSCs containing these dyes show that use of the diphenyl-substituted 1 results in a broadened spectral response at lower energies compared to dyes with anchor 2. However, dye-loading is higher for [Cu (2) 6 ] followed by L ancillary . We are currently investigating dyes containing symmetrical analogues of 3, 5 and 6, as well as the effects of isomerization within the methoxyphenyl rings.