The Versatile SALSAC Approach to Heteroleptic Copper ( I ) Dye Assembly in Dye-Sensitized Solar Cells

Surface-bound heteroleptic copper(I) dyes [Cu(Lanchor)(Lancillary)] are assembled using the “surfaces-as-ligands, surfaces as complexes” (SALSAC) approach by three different procedures. The anchoring and ancillary ligands chosen are ((6,6′-dimethyl-[2,2′-bipyridine]-4,4′-diyl)-bis(4,1-phenylene))bis(phosphonic acid) (3) and 4,4′-bis(4-iodophenyl)-6,6′-diphenyl-2,2′-bipyridine (4), respectively. In the first SALSAC procedure, the FTO/TiO2 electrode is functionalized with 3 in the first dye bath, and then undergoes ligand exchange with the homoleptic complex [Cu(4)2][PF6] to give surface-bound [Cu(3)(4)]+. In the second method, the FTO/TiO2 electrode functionalized with 3 is immersed in a solution containing a 1:1 mixture of [Cu(MeCN)4][PF6] and 4 to give surface-anchored [Cu(3)(4)]+. In the third procedure, the anchor 3, copper(I) ion and ancillary ligand 4 are introduced in a sequential manner. The performances of the DSSCs show a dependence on the dye assembly procedure. The sequential method leads to the best-performing DSSCs with the highest values of JSC (7.85 and 7.73 mA cm−2 for fully masked cells) and overall efficiencies (η = 2.81 and 2.71%, representing 41.1 and 39.6% relative to an N719 reference DSSC). Use of the 1:1 mixture of [Cu(MeCN)4][PF6] and 4 yields DSSCs with higher VOC values but lower JSC values compared to those assembled using the sequential approach; values of η are 2.27 and 2.29% versus 6.84% for the N719 reference DSSC. The ligand exchange procedure leads to DSSCs that perform relatively poorly. The investigation demonstrates the versatile and powerful nature of SALSAC in preparing dyes for copper-based DSSCs, allowing the photoconversion efficiency of dye to be optimized for a given dye. The SALSAC strategy provides alternative hierarchical strategies where the isolation of the homoleptic [Cu(Lancillary)2] is difficult or time-consuming; stepwise strategies are more atom-economic than ligand exchange involving the homoleptic [Cu(Lancillary)2].


Introduction
Dye-sensitized solar cells (DSSCs) [1] harvest solar photons and convert them to electrical energy using a wide-bandgap semiconductor on which a dye is adsorbed [2][3][4].The n-type semiconductor is usually nanoparticulate anatase (TiO 2 ) and, ideally, the dye should be panchromatic, absorbing from the UV to near infrared.Photoconversion efficiencies (η) of up to ca. 14% have been achieved using organic, ruthenium(II) or zinc(II) porphyrin dyes [5][6][7][8][9][10].However, ruthenium is not abundant in the Earth's crust and this has encouraged us and others to explore the use of inorganic sensitizers containing more sustainable first-row metals such as copper(I) [11][12][13][14] and iron(II) [15,16].Values of η in the range 3-5% [17][18][19] have been achieved for DSSCs containing copper(I) dyes.Although these values are lower than those for state-of-the-art ruthenium(II) dyes, we must appreciate that it has taken over 20 years to optimize ruthenium dye structures and their combination with I 3 − /I − electrolytes.
In contrast, DSSCs sensitized with copper(I) dyes have only emerged in the last few years [11][12][13][14] and enhanced performances are gradually being achieved as a result of systematic modifications of dye and electrolyte components [17,[20][21][22][23][24] and the use of co-sensitization [19].The data points for 2017 in Figure 1 demonstrate that copper(I)-based DSSCs have achieved a relative photoconversion efficiency of 65.6% with respect to a reference DSSC sensitized with the ruthenium dye N719 (Scheme 1) [19].
Inorganics 2018, 6, x FOR PEER REVIEW 2 of 17 and iron(II) [15,16].Values of η in the range 3-5% [17][18][19] have been achieved for DSSCs containing copper(I) dyes.Although these values are lower than those for state-of-the-art ruthenium(II) dyes, we must appreciate that it has taken over 20 years to optimize ruthenium dye structures and their combination with I3 − /I − electrolytes.In contrast, DSSCs sensitized with copper(I) dyes have only emerged in the last few years [11][12][13][14] and enhanced performances are gradually being achieved as a result of systematic modifications of dye and electrolyte components [17,[20][21][22][23][24] and the use of co-sensitization [19].The data points for 2017 in Figure 1 demonstrate that copper(I)-based DSSCs have achieved a relative photoconversion efficiency of 65.6% with respect to a reference DSSC sensitized with the ruthenium dye N719 (Scheme 1) [19].Figure 1 summarizes the improvement in photoconversion efficiencies in copper-containing DSSCs relative to N719 reference cells.We include only our own results for sake of meaningful comparison, and data points are chosen for the best performing cells reported each year.Since 2013 [25], all DSSCs in Figure 1 have been fully masked [26].Use of unmasked DSSCs leads to significant overestimation of performance; for example, Dragonetti et al. report a 25−30% decrease in η when a mask is applied to DSSCs containing copper dyes coupled with Cu 2+ /Cu + electrolytes [23].Figure 1 demonstrates that 2011 saw a landmark increase in η and this corresponded to a change from and iron(II) [15,16].Values of η in the range 3-5% [17][18][19] have been achieved for DSSCs containing copper(I) dyes.Although these values are lower than those for state-of-the-art ruthenium(II) dyes, we must appreciate that it has taken over 20 years to optimize ruthenium dye structures and their combination with I3 − /I − electrolytes.In contrast, DSSCs sensitized with copper(I) dyes have only emerged in the last few years [11][12][13][14] and enhanced performances are gradually being achieved as a result of systematic modifications of dye and electrolyte components [17,[20][21][22][23][24] and the use of co-sensitization [19].The data points for 2017 in Figure 1 demonstrate that copper(I)-based DSSCs have achieved a relative photoconversion efficiency of 65.6% with respect to a reference DSSC sensitized with the ruthenium dye N719 (Scheme 1) [19].Figure 1 summarizes the improvement in photoconversion efficiencies in copper-containing DSSCs relative to N719 reference cells.We include only our own results for sake of meaningful comparison, and data points are chosen for the best performing cells reported each year.Since 2013 [25], all DSSCs in Figure 1 have been fully masked [26].Use of unmasked DSSCs leads to significant overestimation of performance; for example, Dragonetti et al. report a 25−30% decrease in η when a mask is applied to DSSCs containing copper dyes coupled with Cu 2+ /Cu + electrolytes [23].Figure 1 demonstrates that 2011 saw a landmark increase in η and this corresponded to a change from Figure 1 summarizes the improvement in photoconversion efficiencies in copper-containing DSSCs relative to N719 reference cells.We include only our own results for sake of meaningful comparison, and data points are chosen for the best performing cells reported each year.Since 2013 [25], all DSSCs in Figure 1 have been fully masked [26].Use of unmasked DSSCs leads to significant overestimation of performance; for example, Dragonetti et al. report a 25-30% decrease in η when a mask is applied to DSSCs containing copper dyes coupled with Cu 2+ /Cu + electrolytes [23].Figure 1 demonstrates that 2011 saw a landmark increase in η and this corresponded to a change from surface-bound homoleptic to surface-bound heteroleptic bis(diimine)copper(I) dyes.This approach predicates the use of heteroleptic complexes containing an anchoring ligand to bind to and facilitate electron-transfer to the semiconductor, and an ancillary ligand which influences the direction of electron flow from the excited state and facilitates hole-transfer to the electrolyte.Unfortunately, copper(I) complexes are labile and attempts to prepare heteroleptic [Cu(L 1 )(L 2 )] + complexes in solution typically result in equilibrium mixtures of [Cu(L 1 )(L 2 )] + and the homoleptic species [Cu(L 1 ) 2 ] + and [Cu(L 2 ) 2 ] + (Equation ( 1)) as shown by NMR spectroscopy [27].
To facilitate the preparation of heteroleptic complexes, Odobel and coworkers capitalized [18,28] upon the HETPHEN (HETeroleptic PHENanthroline) approach [29] in which bulky groups in the 6,6 -positions of a 2,2 -bipyridine (bpy) ligand stabilize the heteroleptic [Cu(L 1 )(L 2 )] + with respect to ligand redistribution.Similar approaches have been introduced previously by, for example, Sauvage [30], Rehahn [31] and Swager [32].Our analysis addressed the problem of ligand redistribution in a different manner.Instead of prioritizing the isolation of discrete heteroleptic complexes from solution, we recognized that the device relevant species was the surface-bound heteroleptic compound.This is encompassed in our "surfaces-as-ligands, surfaces as complexes" (SALSAC) approach in which the heteroleptic dye is assembled on the FTO/TiO 2 electrode is a stepwise manner [11].In this approach, the heteroleptic complex is essentially removed from equilibrium as a solid-state species with a solution activity of zero.In Figure 1, the first significant rise in photoconversion efficiency for DSSCs with copper(I) dyes was achieved [33] using SALSAC.For performance enhancement, heteroleptic dyes are essential.The SALSAC strategy is extremely adaptable, and has the added benefit of permitting dye regeneration [34,35].
Starting from a [Cu(bpy) 2 ] + core, the first requirements for the dye are substituents in the 6,6 -positions to stabilize the tetrahedral coordination geometry preferred by copper(I), and militate against the formation of square-planar copper(II) in both the ground and excited states [36].Methyl groups are often selected, but use of 6,6 -diphenyl-2,2 -bipyridine has the advantage of extending the absorption range of the copper(I) complex towards the red as observed in the absorption spectrum of [Cu(dpp) 2 ] + (dpp = 2,9-diphenyl-1,10-phenanthroline) [37,38].The second design feature to consider is the anchoring domain.Of the different anchors that we have investigated in copper(I) dyes, phosphonic acids (or phosphonates because the protonation state of the adsorbed anchor is not clear [39]) with a phenyl spacer between the PO(OH) 2 and bpy units are superior to carboxylic and cyanoacrylic acids if one considers both synthetic accessibility and DSSC performance [40][41][42].Finally, we come to the ancillary ligand where there is near-infinite scope for functional group modification.Despite investigating a wide range of functionalities for realizing "push-pull" dyes with improved DSSC performances, we have found that simple ancillaries often prove to be the most promising.Halo-substituents in L ancillary are not obvious contenders for inclusion in dyes, but DSSCs containing dyes with peripheral halo-substituents [17] or CF 3 groups in the 6-or 6,6 -positions of the bpy unit [43] function surprisingly well.
We have previously demonstrated the beneficial effects of using ligands 1 and 2 (Scheme 2) as ancillary ligands with 3 as the anchoring ligand [17,40].We now present an investigation of the effects on DSSC performance of combining 4,4 -bis(iodophenyl) and 6,6 -diphenyl substitution patterns in a bpy ancillary ligand 4, and also demonstrate the superiority of the SALSAC approach to the hierarchical assembly of copper(I) sensitizers on mesoporous TiO 2 to optimize DSSC performance.

Ligand Synthesis and Characterization
Anchoring ligand 3 was synthesized as previously described [40].Ancillary ligand 4 was prepared using Kröhnke [44] methodology.The 1 H NMR spectrum of 4 is shown in Figure 2a and this and the 13 C NMR spectrum were assigned by COSY, NOESY, HMQC and HMBC methods; the atom labeling for assignments is given in Scheme 2. Signals for protons H A3 and H A5 were distinguished by the appearance of NOESY cross peaks between H C2 and H A5 , and H B2 and H A3 (Figure S1).The electrospray mass spectrum of 4 exhibited a base peak at m/z = 713.3arising from the [M + H] + ion.
atom labeling for assignments is given in Scheme 2. Signals for protons H A3 and H A5 were distinguished by the appearance of NOESY cross peaks between H C2 and H A5 , and H B2 and H A3 (Figure S1).The electrospray mass spectrum of 4 exhibited a base peak at m/z = 713.3arising from the [M + H] + ion.Scheme 2. Structures of ligands and atom labeling for NMR assignments of 4.

Synthesis and Characterization of [Cu(4)2][PF6]
The syntheses of homoleptic bis(bpy)copper(I) complexes with various bpy derivatives are typically carried out in a mixture of MeCN and CH2Cl2.However, when The electrospray mass spectrum of [Cu(4)2][PF6] in the positive mode exhibited a base peak at m/z 1487.0 corresponding to the [Cu(4)2] + ion.The 1 H NMR spectrum of the compound is shown in Figure 2b.Assignments of the 1 H and 13 C NMR spectra were made using 2D techniques.A comparison of Figure 2a and 2b reveals significant shifts in proton signals on going from the free ligand 4 to [Cu(4)2][PF6].The presence of the 6,6′-diphenyl substituents should lead to considerable flattening of the tetrahedral coordination sphere caused by π-stacking interactions between the phenyl substituents of one ligand with arene rings of the adjacent ligand, as has been crystallographically demonstrated for the related complex [Cu(1)2][PF6] (Scheme 1 shows ligand 1) [32].The absorption spectrum of a CH2Cl2 solution of [Cu(4)2][PF6] (Figure 3) consists of intense bands in the UV region arising from ligand-centred π*←π transitions.As expected, the presence of the 6,6′-diphenyl substituents extends the absorption range giving two bands in the visible region (435 and 585 nm) as observed for the bromo-analogue [Cu(2)2][PF6] [40].Similar dual bands in the visible have been described for [Cu(dpp)2] + (dpp = 2,9-diphenyl-1,10-phenanthroline) and arise from the highly distorted and flattened structure caused by intramolecular π-stacking involving the phenyl substituents [37,38].Interestingly, when a CH2Cl2 solution of [Cu(4)2][PF6] was left standing Scheme 2. Structures of ligands and atom labeling for NMR assignments of 4.

Synthesis and Characterization of [Cu(4) 2 ][PF 6 ]
The syntheses of homoleptic bis(bpy)copper(I) complexes with various bpy derivatives are typically carried out in a mixture of MeCN  The electrospray mass spectrum of [Cu(4) 2 ][PF 6 ] in the positive mode exhibited a base peak at m/z 1487.0 corresponding to the [Cu(4) 2 ] + ion.The 1 H NMR spectrum of the compound is shown in Figure 2b.Assignments of the 1 H and 13 C NMR spectra were made using 2D techniques.A comparison of Figure 2a and 2b reveals significant shifts in proton signals on going from the free ligand 4 to [Cu(4) 2 ][PF 6 ].The presence of the 6,6 -diphenyl substituents should lead to considerable flattening of the tetrahedral coordination sphere caused by π-stacking interactions between the phenyl substituents of one ligand with arene rings of the adjacent ligand, as has been crystallographically demonstrated for the related complex [Cu(1) 2 ][PF 6 ] (Scheme 1 shows ligand 1) [32].The absorption spectrum of a CH 2 Cl 2 solution of [Cu(4) 2 ][PF 6 ] (Figure 3) consists of intense bands in the UV region arising from ligand-centred π*←π transitions.As expected, the presence of the 6,6 -diphenyl substituents extends the absorption range giving two bands in the visible region (435 and 585 nm) as observed for the bromo-analogue [Cu(2) 2 ][PF 6 ] [40].Similar dual bands in the visible have been described for [Cu(dpp) 2 ] + (dpp = 2,9-diphenyl-1,10-phenanthroline) and arise from the highly distorted and flattened structure caused by intramolecular π-stacking involving the phenyl substituents [37,38].Interestingly, when a CH 2 Cl 2 solution of [Cu(4) 2 ][PF 6 ] was left standing at room temperature, the colour changed from dark red to yellow over a 10 day period, and the dual absorptions in the visible were replaced with one, lower intensity maximum at 452 nm (Figure 3).This is consistent with the loss of the flattened structure and loss of one ligand 4 from the [Cu(4) 2 ] + ion.After 14 days, the solution was colourless with a small amount of greenish precipitate, and the absorption spectrum (Figure 3) showed complete loss of the MLCT band.The shift in the ligand-centred band to 270 nm (Figure 3) was consistent with the dominant solution species being compound 4 (Figure S5).An ESI mass spectrum of the 14 day old solution was dominated by a peak at m/z 713 corresponding to 4 (calculated [M + H] + = 713).
at room temperature, the colour changed from dark red to yellow over a 10 day period, and the dual absorptions in the visible were replaced with one, lower intensity maximum at 452 nm (Figure 3).This is consistent with the loss of the flattened structure and loss of one ligand 4 from the [Cu(4)2] + ion.After 14 days, the solution was colourless with a small amount of greenish precipitate, and the absorption spectrum (Figure 3) showed complete loss of the MLCT band.The shift in the ligand-centred band to 270 nm (Figure 3) was consistent with the dominant solution species being compound 4 (Figure S5).An ESI mass spectrum of the 14 day old solution was dominated by a peak at m/z 713 corresponding to 4 (calculated [M + H] + = 713).] is given in Table 1.A reversible process at +0.36 V is assigned to the copper(I) oxidation in both cases.The oxidation potential is the same as for [Cu(2) 2 ][PF 6 ], consistent with the negligible effect of the remote iodo (in 4) versus bromo (in 2) substituents.An irreversible ligand-centred reduction is observed at −1.96 V.
at room temperature, the colour changed from dark red to yellow over a 10 day period, and the dual absorptions in the visible were replaced with one, lower intensity maximum at 452 nm (Figure 3).This is consistent with the loss of the flattened structure and loss of one ligand 4 from the [Cu(4)2] + ion.After 14 days, the solution was colourless with a small amount of greenish precipitate, and the absorption spectrum (Figure 3) showed complete loss of the MLCT band.The shift in the ligand-centred band to 270 nm (Figure 3) was consistent with the dominant solution species being compound 4 (Figure S5).An ESI mass spectrum of the 14 day old solution was dominated by a peak at m/z 713 corresponding to 4 (calculated [M + H] + = 713).

The SALSAC Approach to On-Surface Dye Assembly
The first step in the SALSAC strategy for on-surface heteroleptic copper(I) dye assembly is the functionalization of the semiconductor surface with an anchoring ligand (Figure 4, top left).We have typically followed this with exposure of the ligand-functionalized surface to a solution of a homoleptic complex [Cu(Lancillary)2][PF6] which results in ligand exchange and the formation of a surface-bound heteroleptic species (Figure 4).This is a simple procedure that allows rapid screening of surface-bound heteroleptic dyes.A disadvantage, however, is that one equivalent of ancillary ligand is lost (Figure 5) and this is particularly detrimental if ligand synthesis is time-consuming, for example for ligands incorporating multiple-generation hole-transporting dendrons [25].
We have previously demonstrated that a mixture of [Cu(1)2][PF6] and [Cu(1)(MeCN)2][PF6] (see Scheme 1 for 1) can be used in place of pristine [Cu(1)2][PF6] in the assembly step shown at the top of Figure 4 using 3 as the anchoring ligand, Lanchor [17].Furthermore, heteroleptic dyes may also be assembled by treating the anchor-functionalized TiO2 surface with a 1:1 mixture of [Cu(MeCN)4][PF6] and ancillary ligand (Figure 4, left) [17].This procedure has the advantage of optimizing atom efficiency because no ancillary ligand is lost during surface-bound heteroleptic dye assembly.We confirmed that values of the current density (JSC) were similar for DSSCs with dyes assembled using a mixture of [Cu(1)2][PF6] and [Cu(1)(MeCN)2][PF6] or a 1:1 mixture of [Cu(MeCN)4][PF6] and 1 in the second dye bath (Figure 4).Values of the open-circuit voltage (VOC) were not significantly affected (range 551 to 582 mV).Overall, the photoconversion efficiencies did not show a significant dependence on the dye-assembly method [17].The data are reproduced in Table 2 for convenience.While the data in Table 2 illustrate two variations of the SALSAC approach, a third variant is also possible involving the sequential exposure of the FTO/TiO2 electrode to

The SALSAC Approach to On-Surface Dye Assembly
The first step in the SALSAC strategy for on-surface heteroleptic copper(I) dye assembly is the functionalization of the semiconductor surface with an anchoring ligand (Figure 4, top left).We have typically followed this with exposure of the ligand-functionalized surface to a solution of a homoleptic complex [Cu(L ancillary ) 2 ][PF 6 ] which results in ligand exchange and the formation of a surface-bound heteroleptic species (Figure 4).This is a simple procedure that allows rapid screening of surface-bound heteroleptic dyes.A disadvantage, however, is that one equivalent of ancillary ligand is lost (Figure 5) and this is particularly detrimental if ligand synthesis is time-consuming, for example for ligands incorporating multiple-generation hole-transporting dendrons [25].
We have previously demonstrated that a mixture of [Cu(1) 2 ][PF 6 ] and [Cu(1)(MeCN) 2 ][PF 6 ] (see Scheme 1 for 1) can be used in place of pristine [Cu(1) 2 ][PF 6 ] in the assembly step shown at the top of Figure 4 using 3 as the anchoring ligand, L anchor [17].Furthermore, heteroleptic dyes may also be assembled by treating the anchor-functionalized TiO 2 surface with a 1:1 mixture of [Cu(MeCN) 4 ][PF 6 ] and ancillary ligand (Figure 4, left) [17].This procedure has the advantage of optimizing atom efficiency because no ancillary ligand is lost during surface-bound heteroleptic dye assembly.We confirmed that values of the current density (J SC ) were similar for DSSCs with dyes assembled using a mixture of [Cu(1) 2 ][PF 6 ] and [Cu(1)(MeCN) 2 ][PF 6 ] or a 1:1 mixture of [Cu(MeCN) 4 ][PF 6 ] and 1 in the second dye bath (Figure 4).Values of the open-circuit voltage (V OC ) were not significantly affected (range 551 to 582 mV).Overall, the photoconversion efficiencies did not show a significant dependence on the dye-assembly method [17].The data are reproduced in Table 2 for convenience.While the data in Table 2 illustrate two variations of the SALSAC approach, a third variant is also possible involving the sequential exposure of the FTO/TiO 2 electrode to anchoring ligand, followed by [Cu(MeCN) 4 ][PF 6 ], and finally the ancillary ligand (Figure 3, lower right).We were therefore motivated to compare the performances of DSSCs with dyes assembled using all three dipping processes shown in Figure 4, and chose to investigate the combination of anchor 3 with ancillary ligand 4. As a control, we also looked at the effects of treating the FTO/TiO   The SALSAC approach to a TiO2-bound heteroleptic copper(I) dye using ligand exchange.The SALSAC approach to a TiO2-bound heteroleptic copper(I) dye using ligand exchange.4) leads to maxima in the absorption spectra at 480 nm arising from MLCT transitions of the anchored dye [Cu(3)(4)] + .Although the electrode functionalized using ligand exchange appeared orange to the eye, the corresponding absorption band at 480 nm is barely visible in Figure 6.The single absorption maximum is consistent with the heteroleptic complex [Cu(3)(4)] + rather than [Cu(4) 2 ] + (see Section 2.2) because of the different 6,6 -substituents in ligands 3 and 4. As detailed in Section 2.2, the four phenyl substituents around the copper centre in [Cu(4) 2 ] + lead to a flattened structure and two bands in the visible region of the absorption spectrum.Ligand 3 has 6,6 -dimethyl substituents and the coordination environment around the copper(I) centre in [Cu(3)(4)] + will be close to tetrahedral.Given the mesoporous nature of the semiconductor surface and different orientations of anchored ligands 3 on the functionalized surface after the first dipping step (Figure 2), it is possible that some of the surface-anchored ligand 3 initially binds copper(I) to give anchored [Cu(3) 2 ] + in which both ligands 3 are bound to TiO 2 .Control experiments were carried out in which FTO/TiO 2 electrodes without a scattering layer were immersed in a DMSO solution of anchor 3, followed by one day in a MeCN solution of [Cu(MeCN) 4 ][PF 6 ].The appearance of an MLCT band at ca. 480 nm (Figure 6) indicates the formation of [Cu(3) 2 ] + , presumably arising from the closure of two surface-bound ligands 3 around a Cu + ion.

DSSC Performances
DSSCs sensitized with [Cu(3)(4)] + were fabricated with working electrodes assembled using the three dipping procedures detailed in Figure 4. FTO/TiO 2 electrodes with a scattering layer were subjected to the same dye-bath sequences detailed in Section 2.4 for the solid-state absorption spectroscopic measurements.For device measurements, the DSSCs were fully masked to prevent overestimation of their performance [26].
DSSC performance parameters for duplicate cells containing [Cu(3)(4)] + are given in Table 3, and J-V curves measured on the day of device fabrication (day 0) are displayed in Figure 7. Figure S1 shows the J-V curves compared to that of an N719 reference cell.All fill factors are around 70%, providing approximately equal contributions to the global efficiencies (Equation ( 2)).Comparisons with a reference cell containing N719 are best made by setting the photoconversion efficiency of this cell to 100% (Table 3, right column).The trends in dependence of DSSC performance on method of dye assembly are confirmed by the use of multiple cells.The best-performing devices are those made using the three-step assembly process, with subsequent steps introducing sequentially the anchor 3, copper(I) centre and ancillary ligand 4.These DSSCs achieve values of J SC = 7.73 and 7.85 mA cm −2 , respectively, and V OC = 524 and 517 mV, leading to overall conversion efficiencies of η = 2.81 and 2.71%, respectively.These values convert to 41.1 and 39.6% of η for the N719 reference.Significantly lower values of J SC are obtained by using a 1:1 mixture of [Cu(MeCN) 4 ][PF 6 ] and 4 in CH 2 Cl 2 for the second dye-bath step, although a gain of ca.35 mV in V OC is observed (Table 3).Overall, these DSSCs show photoconversion efficiencies relative to N719 of 33.2 and 33.4% (Table 3).Use of the ligand exchange procedure (Figure 4) leads to the lowest performing DSSCs.Although values of V OC are close to those of the DSSCs assembled by the sequential protocol, values of J SC are significantly lower (4.36 and 4.88 mA cm −2 , Table 3 and Figure 7).The trends in J SC are consistent with the external efficiency (EQE) spectra (Figure 8 and Table 4).The observation that DSSC performance is significantly affected by the method used to assemble the dye on the semiconductor surface underlines a key advantage of the SALSAC approach in that the strategy can be adjusted to optimize DSSC performance on the day of cell fabrication.3 and Figure 7).The trends in JSC are consistent with the external efficiency (EQE) spectra (Figure 8 and Table 4).The observation that DSSC performance is significantly affected by the method used to assemble the dye on the semiconductor surface underlines a key advantage of the SALSAC approach in that the strategy can be adjusted to optimize DSSC performance on the day of cell fabrication.Performance parameters and EQE spectra for DSSCs sensitized by [Cu(3)(4)] + were also measured after 3 and 7 days (Table 3 and Table S1).The devices were stored in the dark between measurements.DSSCs with dyes assembled using the sequential procedure or with the 1:1 mixture of [Cu(MeCN)4][PF6] and 4 are reasonably stable over a period of 7 days (Figure 9) whereas DSSCs containing dyes assembled using the homoleptic complex [Cu(4)2] + exhibit reduced values of JSC after 3 days leading to lower photoconversion efficiencies.The reason for this is not clear.The fill factors remained around 68-70% indicating that cell integrity remained constant.The EQE spectra shown in Figure 10 are consistent with the trends in JSC values observed in the J-V curves in Figure 9. Table S1 gives values of EQEmax and λmax for the spectra in Figure 10.Performance parameters and EQE spectra for DSSCs sensitized by [Cu(3)(4)] + were also measured after 3 and 7 days (Table 3 and Table S1).The devices were stored in the dark between measurements.DSSCs with dyes assembled using the sequential procedure or with the 1:1 mixture of [Cu(MeCN) 4 ][PF 6 ] and 4 are reasonably stable over a period of 7 days (Figure 9) whereas DSSCs containing dyes assembled using the homoleptic complex [Cu(4) 2 ] + exhibit reduced values of J SC after 3 days leading to lower photoconversion efficiencies.The reason for this is not clear.The fill factors remained around 68-70% indicating that cell integrity remained constant.The EQE spectra shown in Figure 10 are consistent with the trends in J SC values observed in the J-V curves in Figure 9. Table S1 gives values of EQE max and λ max for the spectra in Figure 10.  3 and Figure 7).The trends in JSC are consistent with the external efficiency (EQE) spectra (Figure 8 and Table 4).The observation that DSSC performance is significantly affected by the method used to assemble the dye on the semiconductor surface underlines a key advantage of the SALSAC approach in that the strategy can be adjusted to optimize DSSC performance on the day of cell fabrication.Performance parameters and EQE spectra for DSSCs sensitized by [Cu(3)(4)] + were also measured after 3 and 7 days (Table 3 and Table S1).The devices were stored in the dark between measurements.DSSCs with dyes assembled using the sequential procedure or with the 1:1 mixture of [Cu(MeCN)4][PF6] and 4 are reasonably stable over a period of 7 days (Figure 9) whereas DSSCs containing dyes assembled using the homoleptic complex [Cu(4)2] + exhibit reduced values of JSC after 3 days leading to lower photoconversion efficiencies.The reason for this is not clear.The fill factors remained around 68-70% indicating that cell integrity remained constant.The EQE spectra shown in Figure 10 are consistent with the trends in JSC values observed in the J-V curves in Figure 9. Table S1 gives values of EQEmax and λmax for the spectra in Figure 10.    3) with photoanodes assembled using the three dipping methods (ligand exchange, 1:1 mixture and sequential) described in the Materials and Methods section.
As discussed in Section 2.3, solid-state absorption spectra indicated that [Cu(3)2] + was present on the semiconductor surface after FTO/TiO2/3 electrodes were immersed in solutions of [Cu(MeCN)4][PF6].A working electrode with a scattering layer was therefore prepared using this assembly procedure and incorporated into a DSSC.This control assesses the contribution of any [Cu(3)2] + present in DSSCs.In principle, all the dye-assembly methods shown in Figure 4 3) with photoanodes assembled using the three dipping methods described in the Materials and Methods section.3) with photoanodes assembled using the three dipping methods (ligand exchange, 1:1 mixture and sequential) described in the Materials and Methods section.
As discussed in Section 2.3, solid-state absorption spectra indicated that [Cu(3) 2 ] + was present on the semiconductor surface after FTO/TiO 2 /3 electrodes were immersed in solutions of [Cu(MeCN) 4 ][PF 6 ].A working electrode with a scattering layer was therefore prepared using this assembly procedure and incorporated into a DSSC.This control assesses the contribution of any [Cu(3) 2 ] + present in DSSCs.In principle, all the dye-assembly methods shown in Figure 4 3) with photoanodes assembled using the three dipping methods (ligand exchange, 1:1 mixture and sequential) described in the Materials and Methods section.
As discussed in Section 2.3, solid-state absorption spectra indicated that [Cu(3)2] + was present on the semiconductor surface after FTO/TiO2/3 electrodes were immersed in solutions of [Cu(MeCN)4][PF6].A working electrode with a scattering layer was therefore prepared using this assembly procedure and incorporated into a DSSC.This control assesses the contribution of any [Cu(3)2] + present in DSSCs.In principle, all the dye-assembly methods shown in Figure 4 3) with photoanodes assembled using the three dipping methods described in the Materials and Methods section.3) with photoanodes assembled using the three dipping methods described in the Materials and Methods section.3) with photoanodes assembled using the three assembly protocols described in the Materials and Methods section.

General
1 H and 13 C NMR spectra were recorded on a Bruker Avance III-500 NMR spectrometer (Bruker BioSpin AG, Fällanden, Switzerland); 1 H and 13 C chemical shifts were referenced to residual solvent peaks with respect to δ(TMS) = 0 ppm.Solution and solid-state absorption spectra were recorded using a Cary 5000 spectrophotometer (Agilent Technologies Inc., Santa Clara, CA, USA).Electrospray ionization (ESI) and high resolution ESI mass spectra were measured on a Shimadzu LCMS-2020 (Shimadzu Schweiz GmbH, Roemerstr, Switzerland), or Bruker maXis 4G QTOF instrument (Bruker Daltonics GmbH, Faellanden, Switzerland), respectively.Electrochemical measurements were made using a CH Instruments 900B potentiostat (CH Instruments, Inc., Austin, TX, USA) with glassy carbon, platinum wire and leakless AgCl/Ag + electrode (eDAQ ET069, eDAQ Europe, Warszawa, Poland) as the working, counter and reference electrodes, respectively.Samples were dissolved in dry CH2Cl2

DSSC Fabrication
TiO 2 electrodes (Solaronix Test Cell Titania Electrodes, Solaronix SA, Aubonne, Switzerland) were washed with milliQ water and HPLC grade EtOH, heated at 450 • C for 30 min, then cooled to ca. 80 • C. The electrodes were placed in a DMSO solution of 3 (1.0 mM) for 24 h at ambient temperature.The electrodes were taken out of the solution, washed with DMSO and EtOH, then dried in an N 2 stream.For the ligand exchange method, each electrode was then immersed in a CH 2 Cl 2 solution of [Cu(4) 2 ][PF 6 ] for 3 days at room temperature.For the 1:1 dipping procedure, each electrode was then immersed in a CH 2 Cl 2 solution of a 1:1 mixture (each 0.1 mM) of 4 and [Cu(MeCN) 4 ][PF 6 ] for 3 days at room temperature.For the sequential assembly, each electrode was immersed in a MeCN solution of [Cu(MeCN) 4 ][PF 6 ] (2.0 mM) for 1 day followed by immersion in a CH 2 Cl 2 solution of 4 (0.1 mM) for 3 days at room temperature.For the control, each FTO/TiO 2 /3 electrode was immersed for a day in a MeCN solution of [Cu(MeCN) 4 ][PF 6 ].After removal from the final dye-bath, the electrodes were washed with CH 2 Cl 2 and dried in an N 2 stream.For N719 (Solaronix SA, Aubonne, Switzerland), TiO 2 electrodes (Solaronix Test Cell Titania Electrodes, Solaronix SA, Aubonne, Switzerland) were soaked in a solution of N719 (EtOH, 0.3 mM) for 3 days.The electrodes were removed from the dye-bath, washed with EtOH and dried in a stream of N 2 .Counter electrodes (Solaronix Test Cell Platinum Electrodes, Solaronix SA, Aubonne, Switzerland) were washed with EtOH and then heated for 30 min at 450 • C to remove volatile organics.

Electrodes for Solid-State Absorption Spectroscopy
Dye-functionalized electrodes were assembled using the immersion procedures above (the same solution concentrations and times), but using Solaronix Test Cell Titania Electrodes Transparent (Solaronix SA, Aubonne, Switzerland).

DSSC and External Quantum Efficiency (EQE) Measurements
The DSSCs were masked; the mask was made from a black-coloured copper sheet with an aperture (ca.0.06 cm 2 , each mask accurately calibrated) smaller than the surface area of TiO 2 .Black tape was used to complete the top and side masking of each DSSC.Performance measurements were made by irradiating the DSSC from behind with a LOT Quantum Design LS0811 instrument (LOT-QuantumDesign GmbH, Darmstadt, Germany, 100 mW cm −2 = 1 sun, AM1.5 G conditions) and the simulated light power was calibrated with a silicon reference cell.
EQE measurements were made using a Spe Quest quantum efficiency setup (Rera Systems, Nijmegen, The Netherlands) operating with a 100 W halogen lamp (QTH) and a lambda 300 grating monochromator (LOT-Oriel GmbH & Co. KG, Darmstadt, Germany).The monochromatic light was modulated to 3 Hz using a chopper wheel (ThorLabs Inc., Newton, NJ, USA), and the cell response was amplified with a large dynamic range IV converter (Melles Griot B.V., Didam, The Netherlands) and measured with a SR830 DSP Lock-In amplifier (Stanford Research Systems Inc., Sunnyvale, CA, USA).

Conclusions
We have reported the synthesis and characterization of ligand 4 and its homoleptic copper(I) complex.The solution absorption spectrum of [Cu(4) 2 ][PF 6 ] exhibits two MLCT bands at 435 and 585 nm, consistent with a flattened, tetrahedral structure in which there is intramolecular π-stacking involving the 6-and 6 -phenyl substituents.We have fabricated DSSCs in which the photoanode is sensitized with the heteroleptic copper(I) dye [Cu(3)(4)] + ; the adsorbed dye is characterized by an MLCT band at ca. 480 nm.The FTO/TiO 2 electrode surface was first functionalized with the phosphonic acid anchor 3, and then the dyes were assembled by ligand exchange using the homoleptic complex [Cu(4) 2 ][PF 6 ] or in a hierarchical manner using a 1:1 mixture of [Cu(MeCN) 4 ][PF 6 ] and 4, or sequential exposure to [Cu(MeCN) 4 ][PF 6 ] and then 4, thereby demonstrating the versatility of the SALSAC approach to the formation of surface-bound copper(I) dyes.The performances of the DSSCs show a marked dependence on the dye assembly procedure.Use of the sequential method leads to the best-performing DSSCs with the highest values of J SC (7.85 and 7.73 mA cm −2 ) and overall efficiencies (2.81 and 2.71% for masked cells versus 6.84% for a reference N719 DSSC).Using a 1:1 mixture of [Cu(MeCN) 4 ][PF 6 ] and 4 in the second dye bath (Figure 4) leads to DSSCs that have higher V OC values but lower J SC values compared to those assembled using the sequential dye baths.The net result is slightly reduced global efficiencies (2.27 and 2.29% versus 6.84% for the reference cell with N719).Both the sequential and 1:1 mixture approaches yield DSSCs that are stable over a 7 days period.In contrast, the ligand exchange procedure leads to devices that perform relatively poorly; values of J SC fall after 3 days from 4.36 and 4.88 mA cm −2 to 2.63 and 3.45 mA cm −2 .Although the reasons why the ligand exchange protocol leads to inferior DSSCs are unclear, a contributing factor must be the presence of the 6,6 -diphenyl substitution pattern because, when phenyl groups are replaced by methyls (ligand 4 to 1, both the ligand exchange and use of the 1:1 mixture results in comparable DSSC performances (Table 3) [17].
This comparison of the three dye-assembly methods for surface-bound [Cu(3)(4)] + demonstrates the powerful nature of SALSAC in preparing dyes for copper-based DSSCs.The approach allows us to optimize the conditions for the surface assembly of a particular dye and gives alternative hierarchical strategies where the isolation of the homoleptic [Cu(L ancillary ) 2 ] + is difficult.The sequential approach has the advantage of being more atom-economic than ligand exchange involving the homoleptic [Cu(L ancillary ) 2 ] + .

Figure 1 .
Figure 1.Progress of our own work in dye-sensitized solar cells (DSSCs) with copper(I) dyes presented in terms of photoconversion efficiencies relative to N719 set at 100%.Data points are chosen for the best performing cells reported each year.

Figure 1 .
Figure 1.Progress of our own work in dye-sensitized solar cells (DSSCs) with copper(I) dyes presented in terms of photoconversion efficiencies relative to N719 set at 100%.Data points are chosen for the best performing cells reported each year.

Figure 1 .
Figure 1.Progress of our own work in dye-sensitized solar cells (DSSCs) with copper(I) dyes presented in terms of photoconversion efficiencies relative to N719 set at 100%.Data points are chosen for the best performing cells reported each year.

a
No reduction process observed within the solvent accessible window.
2 working electrodes only with only [Cu(MeCN) 4 ][PF 6 ].Inorganics 2018, 6, x FOR PEER REVIEW 7 of 17 anchoring ligand, followed by [Cu(MeCN)4][PF6], and finally the ancillary ligand (Figure 3, lower right).We were therefore motivated to compare the performances of DSSCs with dyes assembled using all three dipping processes shown in Figure 4, and chose to investigate the combination of anchor 3 with ancillary ligand 4. As a control, we also looked at the effects of treating the FTO/TiO2 working electrodes only with only [Cu(MeCN)4][PF6].

Figure 4 .
Figure 4. Three approaches to heteroleptic copper(I) dye assembly using the "surfaces-as-ligands, surfaces as complexes" (SALSAC) strategy.See text for discussion.

Figure 6
Figure 6 compares the solid-state absorption spectra of FTO/TiO2 electrodes (without a scattering layer) functionalized with anchor 3 followed by a solution of homoleptic complex [Cu(4)2][PF6] (ligand exchange strategy), or a 1:1 mixture of [Cu(MeCN)4][PF6] and 4, or sequential dye baths with [Cu(MeCN)4][PF6] and then 4. Use of the 1:1 dye bath or the sequential approach (Figure 4) leads to maxima in the absorption spectra at 480 nm arising from MLCT transitions of the anchored dye [Cu(3)(4)] + .Although the electrode functionalized using ligand exchange appeared orange to the eye, the corresponding absorption band at 480 nm is barely visible in Figure 6.The single absorption maximum is consistent with the heteroleptic complex [Cu(3)(4)] + rather than [Cu(4)2] + (see Section 2.2) because of the different 6,6′-substituents in ligands 3 and 4. As detailed in Section 2.2, the four phenyl substituents around the copper centre in [Cu(4)2] + lead to a flattened

Figure 4 .
Figure 4. Three approaches to heteroleptic copper(I) dye assembly using the "surfaces-as-ligands, surfaces as complexes" (SALSAC) strategy.See text for discussion.

Figure 4 .
Figure 4. Three approaches to heteroleptic copper(I) dye assembly using the "surfaces-as-ligands, surfaces as complexes" (SALSAC) strategy.See text for discussion.

Figure 6 Figure 5 .
Figure 6 compares the solid-state absorption spectra of FTO/TiO2 electrodes (without a scattering layer) functionalized with anchor 3 followed by a solution of homoleptic complex [Cu(4)2][PF6] (ligand exchange strategy), or a 1:1 mixture of [Cu(MeCN)4][PF6] and 4, or sequential dye baths with [Cu(MeCN)4][PF6] and then 4. Use of the 1:1 dye bath or the sequential approach (Figure 4) leads to maxima in the absorption spectra at 480 nm arising from MLCT transitions of the anchored dye [Cu(3)(4)] + .Although the electrode functionalized using ligand exchange appeared orange to the eye, the corresponding absorption band at 480 nm is barely visible in Figure 6.The single absorption maximum is consistent with the heteroleptic complex [Cu(3)(4)] + rather than [Cu(4)2] + (see Section 2.2) because of the different 6,6′-substituents in ligands 3 and 4. As detailed in Section 2.2, the four phenyl substituents around the copper centre in [Cu(4)2] + lead to a flattened

Inorganics 2018, 6 ,
x FOR PEER REVIEW 9 of 17 close to those of the DSSCs assembled by the sequential protocol, values of JSC are significantly lower (4.36 and 4.88 mA cm −2 , Table

Figure 7 .
Figure 7. J-V curves for duplicate DSSCs with photoanodes assembled using the three dipping methods (ligand exchange, 1:1 mixture and sequential) described in the Materials and Methods section.

Inorganics 2018, 6 ,
x FOR PEER REVIEW 9 of 17 close to those of the DSSCs assembled by the sequential protocol, values of JSC are significantly lower (4.36 and 4.88 mA cm −2 , Table

Figure 7 .
Figure 7. J-V curves for duplicate DSSCs with photoanodes assembled using the three dipping methods (ligand exchange, 1:1 mixture and sequential) described in the Materials and Methods section.

Figure 7 .
Figure 7. J-V curves for duplicate DSSCs with photoanodes assembled using the three dipping methods (ligand exchange, 1:1 mixture and sequential) described in the Materials and Methods section.

Figure 8 .
Figure 8. External efficiency (EQE) spectra for DSSCs (cell 2 in each case, see Table3) with photoanodes assembled using the three dipping methods (ligand exchange, 1:1 mixture and sequential) described in the Materials and Methods section.

Figure 9 .
Figure 9. J-V curves for DSSCs (cell 2 in each case, see Table3) with photoanodes assembled using the three dipping methods described in the Materials and Methods section.

Figure 8 .
Figure 8. External efficiency (EQE) spectra for DSSCs (cell 2 in each case, see Table3) with photoanodes assembled using the three dipping methods (ligand exchange, 1:1 mixture and sequential) described in the Materials and Methods section.

17 Figure 8 .
Figure 8. External efficiency (EQE) spectra for DSSCs (cell 2 in each case, see Table3) with photoanodes assembled using the three dipping methods (ligand exchange, 1:1 mixture and sequential) described in the Materials and Methods section.

Figure 9 .
Figure 9. J-V curves for DSSCs (cell 2 in each case, see Table3) with photoanodes assembled using the three dipping methods described in the Materials and Methods section.

Figure 9 .
Figure 9. J-V curves for DSSCs (cell 2 in each case, see Table3) with photoanodes assembled using the three dipping methods described in the Materials and Methods section.

Figure 10 .
Figure 10.EQE spectra for DSSCs (cell 2 in each case, see Table3) with photoanodes assembled using the three assembly protocols described in the Materials and Methods section.

Figure 10 .
Figure 10.EQE spectra for DSSCs (cell 2 in each case, see Table3) with photoanodes assembled using the three assembly protocols described in the Materials and Methods section.
Figure S2: J-V curves for duplicate DSSCs with photoanodes assembled using the three dipping methods (ligand exchange, 1:1 mixture and sequential).Figures S3 and S4: J-V curves and EQE spectra of the control DSSC.
Figure S5: Solution absorption spectrum of 4. Table
a 40 a No reduction process observed within the solvent accessible window.

Table 2 .
[17]ormance parameters (under 1 sun illumination) of duplicate, masked DSSCs with [Cu(3)(1)] + on the day of sealing the cells.The data are compared to a DSSC containing N719 and relative η values are with respect to N719 set at 100%.Data are from ref.[17].

Table 3 .
Performance parameters (under 1 sun illumination) of duplicate, masked DSSCs with [Cu(3)(4)] + on the day of sealing the cells, and after 3 and 7 days.The data are compared to a DSSC containing N719 and relative η values are with respect to N719 set at 100%.