Influence of a co-adsorbent on the performance of bis(diimine) copper(i)-based dye-sensitized solar cells

In the bis(diimine) copper(I) dyes, [Cu(1)(3)] and [Cu(1)(4)] (1 1⁄4 ((6,60-dimethyl-[2,20-bipyridine]-4,40-diyl)bis(4,1-phenylene))bis(phosphonic acid), the ancillary ligands 3 and 4 contain sterically demanding secondgeneration hole-transporting dendrons with methyl or phenyl substituents adjacent to the N-donor atoms of the 2,20-bipyridine metal-binding domain. The performances of DSCs containing [Cu(1)(3)] and [Cu(1)(4)] depend on both the solvent (acetone or CH2Cl2) used in the dye-bath and on the presence of a co-adsorbent. Irrespective of solvent, the dye [Cu(1)(4)] (6,60-Ph2-substituted) only performs well if chenodeoxycholic acid (cheno) is added as a co-adsorbent; for [Cu(1)(3)], cheno has a noticeable effect when the dye assembly is carried out in CH2Cl2. Overall, the results indicate that a combination of small 6,60-substituents in the ancillary ligand and acetone in the dye-dipping cycle lead to the best performing dyes, and for the second-generation dyes, the addition of cheno is essential. Conditions to form TiO2bound [Cu(1)(5)] (5 1⁄4 4,40-bis(4-iodophenyl)-6,60-dimethyl-2,20-bipyridine) in a stepwise manner have been optimized and the effects of introducing cheno at different points during the dye-assembly process have been investigated. When cheno is added to the [Cu(MeCN)4][PF6]/5 dye-bath, the DSCs exhibit values of JSC, VOC and h values that are similar to those with no co-adsorbent. However, competitive binding of 1 and cheno in the first dipping-cycle leads to lower values of JSC and lower photoconversion efficiencies.


Introduction
Dye-sensitized solar cells (DSCs) with bis(diimine) copper(I) complexes 1 as sensitizers have recently been reported with photoconversion efficiencies (PCE) exceeding 3%. 2,3 Several approaches have been adopted for the anchoring of copper(I) dyes to the nanoparticulate TiO 2 semiconductor surface. Initial studies focused on the use of homoleptic [CuL 2 ] + complexes as dyes with anchoring domains in both ligands. [4][5][6][7][8][9] An improved method of dye assembly utilizes the lability of bis(diimine) copper(I) complexes and permits stepwise assembly of surfacebound heteroleptic dyes. The metal oxide is rst treated with an anchoring ligand such as the bis(phosphonic acid) 1 10 (Scheme 1), and the functionalized material is then soaked in a solution containing a homoleptic complex [Cu(L ancillary ) 2 ] + . Equilibration to generate the surface-bound heteroleptic complex {Cu(1)(L ancillary )} is rapid, and the strategy permits effective screening of a wide range of dyes comprising different anchoring and ancillary diimine ligands. 2,[10][11][12][13][14][15][16] To optimize atom efficiency, surface-bound heteroleptic dyes can also be assembled by treating the anchor-functionalized TiO 2 surface with a 1 : 1 mixture of [Cu (MeCN) 4 ] + and the ancillary ligand. 2 The DSC performance of these cells is comparable to those of cells comprising the corresponding dye made by the on-surface ligand exchange procedure.
The isolation of heteroleptic bis(diimine) copper(I) complexes is usually difficult because of rapid equilibration between homo-and heteroleptic species in solution. 17 One way to overcome this is the HETPHEN approach 18 which employs sterically demanding substituents adjacent to the N-donors in one diimine ligand. 3,19 Using this strategy, a record PCE of 4.66% has been reached with the [Cu(L anchor )(L ancillary )] + dye in which L anchor is 6,6 0 -dimesityl-2,2 0 -bipyridine-4,4 0 -dicarboxylic acid and L ancillary is a 2,2 0 -bipyridine ancillary ligand bearing peripheral triphenylamino domains. The use of the coadsorbant chenodeoxycholic acid (cheno, Scheme 2) is critical to the attainment of this high PCE. 3 Surprisingly few studies have probed the effects of co-adsorbants on copper(I) dyes, although it is well known that cheno enhances the open-circuit voltage (V OC ) of ruthenium(II)-containing dyes such as N719, 20 and zinc(II) porphyrin dyes. 21,22 We have reported that the addition of cheno to the homoleptic dyes shown in Scheme 3a does not enhance the PCE, 6 whereas Robertson and coworkers report improved photocurrents when cheno is added to the sensitizer [Cu(4,4 0 -(HO 2 C) 2 bpy)(POP)] + (Scheme 3b). 23 The reduction in dye loading resulting from the presence of co-adsorbants should counter the ripening effects that DSCs utilizing copper(I)-containing dyes experience with an I À /I 3 À electrolyte. 2,10,12,13 This aging effect has also been observed for ruthenium(II) dyes and is explained in terms of disaggregation and reorganization of the sensitizer molecules on the TiO 2 surface. [24][25][26] On the other hand, we have also noted that if the ancillary ligand is small (e.g. 6,6 0 -dimethyl-2,2 0 -bipyridine, dmbpy), optimum PCE is achieved on the day of DSC fabrication, indicating that molecular dyes such as [Cu(1)(dmbpy)] + undergo little or no time-dependent reorganization on the surface. 27 Controlled dye loading is critical to achieving optimal DSC performance immediately aer DSC fabrication as shown by varying the concentration of the homoleptic [Cu(L ancillary ) 2 ] + during the soaking step in the stepwise assembly of [Cu(L anchor )(L ancillary )] + on a TiO 2 surface. 28 In this latter investigation, L anchor was phosphonic acid 1 and the ancillary ligand was the rst-generation dendron 2 (Scheme 4).
Here we report a study of the effects on DSC performance of combining cheno with three dyes. The latter were selected because, for copper(I) containing dyes, they give moderate to good photoconversion efficiencies. 2

DSC fabrication and measurements
DSCs were made by adapting a method described by Grätzel and coworkers. 30,31 For the photoelectrode, Solaronix Test Cell Titania Electrodes were used. The electrodes were washed with EtOH and sintered at 450 C for 30 min, then cooled to z80 C and immersed in a 1 mM DMSO solution of 1 for 24 h, or in a mixture of 1 (0.1 mM) with cheno in molar ratios (1 : 1, 1 : 3 and 1 : 6) in DMSO for 24 h. The electrode was removed from the solution, washed with DMSO and EtOH and dried with a heat gun at z60 C. For dyes containing 3 or 4, the functionalized electrode was placed in either a 0.1 mM CH 2 Cl 2 or a 0. Each reference electrode was prepared by soaking a commercial electrode in a 0.3 mM EtOH solution of dye N719 (Solaronix) for 3 days. Aer soaking in the dye-baths, the electrodes were removed, washed with the same solvent as used in the dye-bath and dried with a heat gun (z60 C).
For the counter electrode Solaronix Test Cell Platinum Electrodes were used, and organic impurities were removed by heating on a heating plate at 450 C for 30 min.
The dye-covered TiO 2 electrode and Pt counter-electrode were combined using thermoplast hot-melt sealing foil (Solaronix Test Cell Gaskets) by heating while pressing them together. The electrolyte comprised LiI (0.1 M), I 2 (0.05 M), 1-methylbenzimidazole (0.5 M) and 1-butyl-3-methylimidazolinium iodide (0.6 M) in 3-methoxypropionitrile and was introduced into the DSC by vacuum backlling. The hole in the counter electrode was sealed using hot-melt sealing foil (Solaronix Test Cell Sealings) and a cover glass (Solaronix Test Cell Caps).
Measurements were made by irradiating from behind using a light source SolarSim 150 (100 mW cm À2 ¼ 1 sun). The power of the simulated light was calibrated using a reference Si cell. All DSCs were completely masked 32 before measurements were made.

External quantum efficiencies
The external quantum efficiency (EQE) measurements were made using a Spe-Quest quantum efficiency instrument from Rera Systems (Netherlands) 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). The cell response was amplied with a large dynamic range IV converter (CVI Melles Griot) and measured with a SR830 DSP Lock-In amplier (Stanford Research).

Co-adsorbant cheno with sterically demanding dyes [Cu(1)(3)] + and [Cu(1)(4)] +
We have previously reported that extending the holetransporting dendron in ancillary ligand 2 to give 3 leads to inferior performances of the second-generation dye [Cu(1)(3)] + compared to rst-generation [Cu(1)(2)] + . 16 We were interested in seeing whether the performance of DSCs incorporating the very sterically hindered second-generation ancillary ligand could be improved by adding cheno. Ancillary ligands 3 and 4 both contain second-generation dendrons, but differ in having methyl or phenyl substituents in the 6,6 0 -positions, respectively. All data reported for cells without cheno were obtained in a parallel study, where we showed that a combination of small 6,6 0 -substituents and acetone in the dye-dipping cycle yielded the best performing dendron-containing dyes. 16 The solvent used in the dye bath is known to inuence the efficiency of DSCs, 16,33 and in the present study, we have used different solvents (acetone and CH 2 Cl 2 ) in the copper(I) dye baths.
Dye-functionalized photoanodes were made by soaking electrodes in a solution of anchoring ligand 1 followed by a dipping cycle in CH 2 Cl 2 or acetone solutions containing a 1 : 1 mixture of equimolar amounts of cheno and either [Cu (3) (1)(4)] + when acetone is the solvent for the dye dipping cycle. The efficiency increases from 1.29 to 2.45% on adding cheno Table 1). In contrast, the improvement in J SC and V OC for [Cu (1)(3)] + is small when the dye soaking-cycle involves acetone.
We have previously observed that (in the absence of cheno), a ripening effect is observed for CH 2 Cl 2 -fabricated dyes. 16 The data in Table 1 show that this time-dependent improvement in efficiency is less (h increases for [Cu(1)(3)] + from 1.49 to 2.01% over 22 days) than that produced on the rst day by the addition of cheno (h ¼ 2.23%); a small ripening effect is observed when cheno is present (h increases for [Cu(1)(3)] + from to 2.23-2.47% over 22 days).
The ancillary ligand in [Cu(1)(4)] + possesses a 6,6 0 -Ph 2 bpy core (Scheme 4), and the presence of the co-adsorbant has a dramatic effect. The J-V curves shown in Fig. 1 illustrate that the addition of cheno is important for immediately optimizing the DSC performance of the second-generation dye [Cu(1)(4)] + derived from the acetone dye-soaking cycle (grey versus blue J-V curves in Fig. 1). Most importantly, the enhanced values of short-circuit current density and open-circuit voltage which result from the addition of cheno are maintained over a 3 week period (Fig. 1).
The EQE spectra shown in Fig. 2 for DSCs containing [Cu(1)(4)] + summarize the combined effects of solvent used in the dye bath and cheno additive. The maximum EQE of 49.1% (l max ¼ 470 nm) aer 3 weeks is essentially the same as on the day of sealing the cell (48.3% at l max ¼ 480 nm) for acetonederived dyes with cheno. This compares to only 18.0% rising to 21.3% (l max ¼ 480 nm) for [Cu(1)(4)] + /CH 2 Cl 2 with no coadsorbant. Of note is that the [Cu(1)(4)] + /acetone/cheno combination shows improved electron injection to lower energy (blue curves, Fig. 2).  Compared to dye assembly using the exchange reaction shown in eqn (1), the stepwise method has the advantage of atom efficiency (eqn (2)).  (2) The typical soaking time of an anchoring ligandfunctionalized electrode in a solution of a homoleptic complex (eqn (1)) is 3 days. 2,10-16 However, we argued that a shorter dipping cycle might be sufficient with the stepwise method shown in eqn (2). Table 2 summarizes the results of experiments in which DSCs were assembled by rst functionalizing the TiO 2 -coated electrode with anchoring ligand 1, and then immersing the electrode in a CH 2 Cl 2 solution containing a mixture of [Cu(MeCN) 4 ][PF 6 ] and 5 (each 0.1 mM) for either 1 or 3 days. Duplicate cells were also tested and showed the same trend as observed in Table 2. A longer dipping time results in improved J SC and slightly enhanced V OC , leading to a greater overall conversion efficiency. The results also indicate that the efficiency, h, increases as the DSC matures when a short dyebath cycle is used; this is not the case for the longer dipping time (Table 1). Fig. 3 shows the external quantum efficiency (EQE) spectra for the 7 day old DSCs with EQE maxima of 47.8% (l max ¼ 480 nm, 1 day in the dye-bath) and 44.7% (l max ¼ 470 nm, 3 days in the dye bath). The increased spectral response at both  higher and lower energies in the EQE spectrum for the 3 day soaking time corresponds to the enhanced J SC for the DSC prepared under these conditions. A control experiment was also carried out; a photoanode functionalized with anchoring ligand 1 was immersed in a CH 2 Cl 2 solution of [Cu(MeCN) 4 ][PF 6 ] in the absence of ligand 5 for a day. Aer drying, the electrode appeared pale yellow in colour and the EQE spectrum of the DSC made with this electrode and I 3 À /I À electrolyte gave a maximum of only 4.36% (l max ¼ 460 nm, Fig. 3) in sharp contrast to the value of 47.8% (l max ¼ 480 nm) obtained using a mixture of [Cu(MeCN) 4 ][PF 6 ] and 5 (eqn (2)) under the same assembly conditions. To quantify the dye loading aer different dye bath times, electrodes without a scattering layer were prepared by immersing TiO 2 /FTO glass slides functionalized with anchor 1 in a CH 2 Cl 2 solution containing a 1 : 1 mixture of [Cu(MeCN) 4 ] + and 5. Aer dipping times of 1 or 3 days, the solid-state absorption spectra of the electrodes were recorded and each spectrum was corrected for the background spectrum of a blank TiO 2 /FTO electrode. The difference in absorption is signicant (Fig. 4a) and is also seen by eye (Fig. 4b). In subsequent experiments with the co-adsorbant cheno, a period of 3 days was used for the dye dipping-cycle.  Table 3 summarizes the performance data. The J-V curves for the DSCs are shown in Fig. 5 and are compared with that for a DSC in which no coadsorbant was present (red curve in Fig. 5); this corresponds to the DSC in Table 2 with 3 days in the dye bath. Within experimental error (see Table S2 † for data with duplicate cells), the addition of cheno makes no signicant difference to the overall DSC performance. The PCE of 2.71% without   (1)(5)] + sensitizer can be understood in terms of the dye possessing little steric crowding, and suggests that near-optimal dye coverage is obtained immediately the cells are assembled. When using the ligand exchange protocol (eqn (1)), we observed ripening effects, 2 but these are not as dramatic as with more sterically encumbered ancillary ligands, 16 nor do we see significant time-dependent performance using the stepwise assembly of [Cu(1)(5)] + (eqn (2)). We were therefore interested to see the effects of introducing the co-adsorbant with the anchoring ligand (specically whether there is competitive binding), prior to complexation with copper(I) and ancillary ligand. The dyefunctionalized photoanodes for the DSCs were prepared by immersing the electrode for a day in a DMSO solution containing 1 and cheno in molar ratios of 1 : 1, 1 : 3 and 1 : 6; stock solutions of each component were 0.1 mM. Aer drying, the electrode was then soaked for 3 days in a CH 2 Cl 2 solution containing [Cu(MeCN) 4 ][PF 6 ] and ligand 5 in a 1 : 1 molar ratio. A reference electrode with N719 was also prepared. Performance parameters for the DSCs containing these photoanodes are shown in Table 4. Measurements were made at intervals up to a week aer cell fabrication, and J-V curves corresponding to 7 day old cells are depicted in Fig. 6.
Measurements made with a duplicate set of DSCs conrmed that the general trends in Table 4 and Fig. 6 were reproducible (Table S3 †). The results indicate that adding cheno to the anchoring ligand during the rst stage of dye assembly is detrimental. The open-circuit voltage achieved aer 7 days without cheno (578 m, Table 2) is higher than the V OC values observed with cheno (547 to 574 mV), and values of the shortcurrent current density are signicantly lowered when cheno is added (from 6.96 mA cm À2 in the absence of cheno to 6.15 mA cm À2 and 5.75 mA cm À2 with 1 : 1 and 1 : 6 ratios of 1 : cheno). The consequence of the lower J SC values (Fig. 6) is a drop in the cell PCE from 2.71% (no cheno, Table 2) to 2.49% (1 : cheno ¼ 1 : 1) to 2.27% (1 : cheno ¼ 1 : 6). The EQE spectra of 7 day old DSCs are shown in Fig. 7. Although there is no unambiguous trend in the shape of the spectra with the amount of cheno added, it is apparent that the highest amount of cheno results in a loss of quantum efficiency at both lower and higher wavelengths.
Although both experiment and theory show that phosphonic acid anchors bind more strongly to TiO 2 than carboxylic acid domains, 34 our results are consistent with competitive binding of the bis(phosphonic acid) 1 with cheno (CO 2 H anchor) which ultimately leads to a poorer surface coverage of the copper(I) dye.   Table 3) in which the dye was assembled by dipping the electrode functionalized with 1 into a dye bath containing [Cu(MeCN) 4 ][PF 6 ], 5 and cheno. The red curve corresponds to a DSC without cheno. The red curve corresponds to a DSC without cheno.