Guglielmo Risia,
Mike Devereuxb,
Alessandro Prescimonea,
Catherine E. Housecrofta and
Edwin C. Constable*a
aDepartment of Chemistry, University of Basel, BPR 1096, Mattenstrasse 24a, 4058 Basel, Switzerland. E-mail: edwin.constable@unibas.ch
bDepartment of Chemistry, University of Basel, Klingelbergstrasse 80, CH-4056 Basel, Switzerland
First published on 31st January 2023
Metal complexes used as sensitisers in dye-sensitised solar cells (DSCs) are conventionally constructed using a push–pull strategy with electron-releasing and electron-withdrawing (anchoring) ligands. In a new paradigm we have designed new DπA ligands incorporating diarylaminophenyl donor substituents and phosphonic acid anchoring groups. These new ligands function as organic dyes. For two separate classes of DπA ligands with 2,2′-bipyridine metal-binding domains, the DSCs containing the copper(I) complexes [Cu(DπA)2]+ perform better than the push–pull analogues [Cu(DD)(AA)]+. Furthermore, we have shown for the first time that the complexes [Cu(DπA)2]+ perform better than the organic DπA dye in DSCs. The synthetic studies and the device performances are rationalised with the aid of density functional theory (DFT) and time-dependent DFT (TD-DFT) studies.
The injection of the electron into the conduction band is facilitated by the covalent binding of the photosensitiser to the surface of the semiconductor. This both optimizes the stability of the device and provides a low-energy pathway for the transport of the electron from photosensitiser to the semiconductor. Functionality used for the covalent binding of the photosensitiser to the surface (anchoring groups) includes carboxylic acids, phosphonic acids, phenols and polyphenols. In addition to the anchoring group, the molecular design of photosensitisers is predicated upon a dipolar species with the negative pole in the direction of the anchoring group. Using these design criteria, three main classes of photosensitisers have been investigated: organic molecules,5–7 metal-containing macrocycles such as porphyrins and phthalocyanines8 and metal complexes. This paper is concerned with investigating a new design paradigm in the latter class of sensitisers.
Photosensitisation of crystalline TiO2 and SnO2 by ruthenium(II) complexes was reported in 1979 (ref. 9) but the potential was first fully developed in 1991 when Grätzel described a dye-sensitised solar cell (DSC) in which semiconductor electrodes, prepared by the sintering of TiO2 (anatase) nanoparticles, were sensitised with a ruthenium(II) complex.10 The use of the sintered nanoparticles resulted in an electrode with a very large surface area which could bind a large amount of the photosensitiser. After the first modest results, there has been a steady improvement in the performance of DSCs containing ruthenium photosensitisers through tuning of the ligands and anchoring groups, the electrolyte and the chemical and electronic structure of the semiconductor. The prototypical ruthenium(II) photosensitisers are six-coordinate and have two 2,2′-bipyridine (bpy) or 1,10-phenanthroline (phen) ligands bearing anchoring groups bound to the metal centre and two ancillary donors which both define the dipole and tune the redox and photophysical properties of the complexes.11 The design principle has been more generally extended to other coordination12 and organometallic13,14 compounds of ruthenium. The incorporation of the dipole within the photosensitiser through ligand design is often described as the “push–pull” strategy in which one ligand is electron-releasing and the other is electron withdrawing. State-of-the-art ruthenium dyes have attained maximum photoconversion efficiencies (PCE) of up to 12% in DSCs.8,15–22 It is argued that metal complexes have inherent advantages over organic photosensitisers as they are likely to exhibit higher thermal and photochemical stability. Nevertheless, the use of ruthenium complexes as photosensitisers has a significant, potentially critical, disadvantage. Ruthenium is present in the Earth's crust in low abundance (ca. 0.001 ppm)23 and is expensive, raising questions about both the sustainability and the commercial viability of the technology. As a consequence, significant effort has been invested in the search for photosensitisers based on other metal centres which would be more sustainable and lower cost.
In 1994, Sauvage and coworkers reported a DSC with a TiO2 photoanode sensitised with a Cu(phen)2+ complex.31 Subsequently, we designed a series of homoleptic copper(I) complexes with bpy ligands containing carboxylic acids as anchoring groups and reported a PCE which corresponds to 23.7% relative to a device with a reference ruthenium(II) dye N719.32 Since then, homoleptic copper(I) complexes have attracted interest with improvements in PCEs by tailored modifications of the ligands.33–35 This type of complex is shown in Scheme 1a. The anchoring ligand is conveniently described as Lanchor and is typically also an electron-withdrawing substituent.
Heteroleptic complexes of the type [Cu(Lanchor)(Lancillary)]+ in which the second, ancillary, ligand, Lancillary is electron-releasing (Scheme 1b) possess the push–pull architecture expected to improve the electron injunction into the conduction band of the semiconductor.24 Copper(I) complexes are labile and attempts to isolate homoleptic [Cu(Lanchor)(Lancillary)]+ complexes typically lead to a statistical mixture of [Cu(Lanchor)(Lancillary)]+, [Cu(Lanchor)2]+ and [Cu(Lancillary)2]+. Elegant strategies have been developed to isolate heteroleptic complexes including the HETPHEN36,37 approach developed by Schmittel in which bulky substituents direct the preferential formation of a heteroleptic complex.38–40 We have introduced the SALSAC approach,24,41 based on the stepwise assembly of heteroleptic copper(I) complexes by immobilising Lanchor on the semiconductor, followed by formation of the heteroleptic complex as a surface-bound species. These approaches have allowed the evaluation of a wide range of Lanchor and Lancillary combinations to be evaluated.42–54
Chemisorption on the photoanode of a DSC does not completely prevent the dissociation of Lancillary and this phenomenon contributes to deterioration of device performance following prolonged contact with the electrolyte, a phenomenon described as ‘bleaching’.55–58
In this paper we describe a new strategy for the design of push–pull complexes to address these issues. Instead of using heteroleptic complexes with two different ligands containing electron-withdrawing and electron-accepting (anchoring) ligands respectively, we considered a single asymmetrical ligand type bearing both electron-withdrawing (donor) and electron-accepting (acceptor) substituents (Scheme 1c). We describe this as a donor–π–bridge–acceptor (DπA) architecture. This strategy has been little investigated and has not delivered convincing results to date.59–61 We also note that the DπA ligands themselves possess the characteristics of a purely organic photosensitiser and it is instructive to compare the performance of the ligand as an organic dye sensitiser with that of the homoleptic [Cu(DπA)2]+ complexes.
The core of our investigation was the comparison of [Cu(Lanchor)(Lancillary)]+ complexes with their isotopic [Cu(DπA)2]+ congeners. Thus, the ligands 3 and 6 were designed as the DπA partners for the Lanchor and Lancillary pairs 1 and 2, or 4 and 5, respectively (top panel, Scheme 2). The complexes [Cu(3)(3–H)] and [Cu(6)(6–H)] were compared with the heteroleptic push–pull [Cu(1)(2)]+ and [Cu(4)(5)]+, respectively (bottom panel, Scheme 2). The photophysical and electrochemical properties of the compounds and the DSC performances of devices were studied, complemented by DFT and TD-DFT calculations.
The working and counter-electrode for each DSC were joined together using thermoplast hot-melt sealing foil (Solaronix Test Cell Gaskets, 60 μm) and the gap between them was filled with electrolyte (LiI (0.1 M), I2 (0.05 M), 1-methylbenzimidazole (0.5 M), 1-butyl-3-methylimidazolinium iodide (0.6 M) in 3-methoxypropionitrile) by vacuum backfilling through a hole in the counter-electrode. Finally, the hole was sealed (Solaronix Test Cell Sealings and Solaronix Test Cell Caps).
C62H44Br2N4, Mr = 1004.83, yellow plate, monoclinic, space group P21/c, a = 6.5334(6), b = 25.031(2), c = 15.1033(13) Å, β = 102.382(3)°, V = 2412.5(4) Å3, Dc = 1.383 g cm−3, T = 150 K, Z = 2, μ(CuKα) = 2.478 mm−1. Total 14983 reflections, 4367 unique (Rint = 0.0246). Refinement of 3963 reflections (307 parameters) with I > 2σ(I) converged at final R1 = 0.0280 (R1 all data = 0.0313), wR2 = 0.0751 (wR2 all data = 0.0776), gof = 1.063. CCDC 2203243.
Single crystals of 10 were grown by slow evaporation of a CH2Cl2 solution. Yellow plates of 10 crystallise in the monoclinic space group P21/c with half of the molecule present in the asymmetric unit; the second half is generated by inversion (Fig. 1a) and thus, as the pyridine ring is near-planar, the bpy unit is necessarily planar. Selected bond lengths and angles are given in the caption to Fig. 1. The substituents at the alkene unit describe the (E)-stereoisomer, and the arene ring containing C10 is essentially coplanar with the pyridine ring (angle between the least squares planes = 14.0°). However, the conjugated substituents are somewhat ‘bowed’ as illustrated in Fig. 1b. The angle between the planes of the pyridine ring and arene ring containing C4 is 36.0°, consistent with alleviating repulsive H⋯H contacts between the rings.
Most compounds were obtained in yields of around 60%, with the exception of 10 (31%) and 5 (90%). The steric hindrance of the 6,6′-substituted 2,2′-bipyridine systems reduces the N–HOSi bonding interactions between the pyridine rings and silica particles, making the purification by column chromatography a viable option in the case of ligands 4e, 5 and 6e. Note that the presence of a phosphonate ester is the cause of loss of material because of the bonding interactions (following hydrolysis of the ester) between the phosphonic acid groups and the silica. The phosphonic acid derivatives were used after precipitation without further purification. However, for the phosphonic acids and their copper(I) complexes, it was not possible to obtain satisfactory elemental analysis. The 31P NMR spectra of the phosphonate esters 3e and 6e (δ = 17.9 and 18.3 ppm, respectively) exhibited an upfield shift upon deprotection (δ = 14.9 and 11.9 ppm, respectively). Similar shifts are seen on comparing the complexes of the phosphonate esters to those of the phosphonic acids (see the Experimental section in ESI†). In the copper(I) complexes of ligands 3 and 6, the absence of the PF6− anion was verified, thus indicating that at least one proton has been lost from the phosphonic acid substituents, negating the need for an anion. With no further evidence, we assumed one of the two phosphonic acid to be deprotonated (with the ligands this referred to as [3–H]− and [6–H]−) and the copper(I) complexes to be zwitterions (i.e. [Cu(3)(3–H)] and [Cu(6)(6–H)]). Furthermore, a low stability of the complexes in solution was observed after their isolation, and after one week, the 1H NMR spectra showed broadening and loss of resolution of the signals. We did not conduct time stability studies in this work. Instead, we based the photophysical and electrochemical characterisation of the compounds on their ester analogues given their long-term stability in solution. The 1H and 13C{1H} NMR spectra of all compounds were assigned using two-dimensional (2D) methods. It is interesting to analyse the nature of 6e from the 1H NMR spectrum as representative of the asymmetrical ligands (Fig. 2): the presence of a donor group and an acceptor group breaks the symmetry of the molecule, generating a larger number of signals. The comparison of peaks with spectra of ligands 4e and 5, reveals the presence of two patterns compatible with those of these ancillary and anchoring ligands.
The chemical shifts of protons in pyridine rings B and C are well separated (HB3 and HC3, HB5 and HC5), with the ring C signals shifted downfield. The same is true for 13C NMR assignments in 6e, where the signals of ring C are shifted downfield with respect to ring B (see the Experimental section in ESI†). The same is found in the case of ligand 3e and similar conclusions are drawn for all asymmetric ligands and their complexes described in this work. The differences in the chemical environments between the moieties in the 2,2′-bipyridine scaffold are consistent with the asymmetry, leading to a chemical shift distribution oriented from the donor (downfield) through the acceptor unit (upfield). The full assignment for the 6,6′-substituents of the asymmetric 2,2′-bipyridine ligands and corresponding copper(I) complexes could not be made unambiguously; whereas the alkenyl protons could be assigned in most cases, the TPA units found in 3e, 6e and derivatives experience a similar chemical environment, resulting in not unequivocal signal assignment. The deprotection of the esters was easily confirmed by the disappearance of the HEt-CH2 and HEt-CH3 signals.
The NMR assignments for the ligands can be compared to their homoleptic copper(I) complexes [Cu(3)(3–H)] and [Cu(6)(6–H)] and their ester counterparts. The 1H and 13C{1H} NMR assignments indicate that the electron density distribution within the 2,2′-bipyridine is nonuniform and oriented from the donor through the acceptor unit. Additionally, L and L–H cannot be distinguished due to fast proton exchange. In turn, it implies that the new copper(I) complexes preserve the chemical environment of the ligands when engaged in coordination to the metal, regardless of the conformational change of the bpy domain from transoid to cisoid.
During the isolation of [Cu(6)(6–H)], we observed the presence of another species present in solution (Fig. 3). Some signals (δ = 7.87 ppm, δ = 7.35 ppm, and δ = 7.1 ppm) can be ascribed to a species compatible with the free ligand in solution. We were, therefore, interested in investigating the ligand-complex binding dynamics in solution in order to assess how the bulky chelating site found in the phosphonic esters and acids affects the binding equilibria during complexation.
Fig. 3 Comparison of 1H NMR spectra 6 (top) and [Cu(6)(6–H)] (bottom). Solvent: DMSO-d6. The spectrum of [Cu(6)(6–H)] reveals a second pattern of peaks of low intensity compatible with that of 6. |
For this purpose, we conducted DOSY experiments in CD2Cl2 for solutions containing ancillary ligand 5 and [Cu(CH3CN)4][PF6] as copper(I) source (Table S1†). Diffusion coefficients of 5.037 × 10−10 m2 s−1 and 4.466 × 10−10 m2 s−1 were found for 5 and [Cu(5)2]+ (1:0.5 ligand to copper ratio), respectively, resulting in a slower diffusion for the homoleptic complex. As anticipated, the nature of CuL2 was clarified by comparing the hydrodynamic radii of the two species (≈42% larger volume for [Cu(5)2]+ as compared to 5 assuming ideal spheres without a solvation shell). We screened two additional combinations of ligand and [Cu(CH3CN)4][PF6], namely 1:1 and 1:2 ratios. In both cases, [Cu(5)2]+ was the predominant species in solution with an essentially constant diffusion coefficient (D = 4.455 × 10−10 m2 s−1 and 4.525 × 10−10 m2 s−1, respectively). Additionally, we could identify a species with D = 4.929 × 10−10 m2 s−1 compatible with [CuL]+ and with only ≈7% volume excess with respect to 5. This not only confirms the solution stability of [Cu(5)2]+ at high molar ratios of copper(I), but also confirms that, in spite of the bulky substituents of 2,2′-bipyridine ligands, the formation of the homoleptic complexes occurs.
Fig. 4 Solution absorption spectra of investigated compounds (CH2Cl2, 10−5 mol dm−3, 6 × 10−6 mol dm−3 for [Cu(2)2][PF6]). |
vs. NHE | Epa/V | Epc/V | λmax/nm (ε/dm3 mol−1 cm−1) | λonset/nm | EHOMO/eV | ELUMOb/eV | Eoptga/eV |
---|---|---|---|---|---|---|---|
a Obtained using formula Eoptg = 1240/λonset.b Calculated subtracting Eoptg to EHOMO. | |||||||
3e | +1.23 | +1.13 | 246 (45650); 298 (27840); 353 (23740) | 393 | 1.23 | −1.93 | 3.16 |
4 | +0.96 | +0.90 | 267 (71150); 401 (59850) | 446 | 0.96 | −1.82 | 2.78 |
5 | +0.95 | — | 298 (69600); 385 (91570) | 439 | 0.95 | −1.87 | 2.82 |
6e | +0.92 | — | 265 (57240); 296 (59080); 391 (71880) | 440 | 0.92 | −1.90 | 2.82 |
[Cu(3e)2][PF6] | +1.04 | +0.92 | 259 (62130); 321 (47890); 494 (33630) | 545 | 1.04 | −1.23 | 2.27 |
[Cu(2)2][PF6] | +1.03 | — | 310 (57223); 360 (49511); 492 (16801) | 546 | 1.03 | −1.24 | 2.27 |
[Cu(5)2][PF6] | +0.87 | — | 299 (141040); 413 (141040) | 487 | 0.87 | −1.68 | 2.55 |
[Cu(6e)2][PF6] | +1.03 | +0.98 | 297 (124240); 414 (134030) | 495 | 1.03 | −1.48 | 2.51 |
Fig. 5 Energy level diagram of ligands and complexes according to EHOMO and ELUMO found in Table 1. |
The phosphonate esters and N,N-bis(4-methoxyphenyl)phenylamine (MeOTPA) substituents in 4e, 5 and 6e do not appear to have a significant impact on the low energy band, with a distribution of λmax values within a range of 15 nm. Instead, the second lowest energy band is more affected: ligand 4e (with phosphonate esters) presents a λmax of 267 nm compared to 298 nm in ligand 5 (with MeOTPA). Ligand 6e features a band with two peaks centred at 265 and 296 nm (combination of phosphonate ester and MeOTPA substituents, respectively). In addition, an increase in ε is noticed in ligand 5 as the number of MeOTPA groups is doubled. The high energy bands are attributed to the π* ← π transitions, while the most red-shifted are ascribed to intra-ligand charge transfer (ILCT) between the TPA groups and the bpy domain.
Upon coordination of ligands 3e, 5 and 6e to copper(I), the absorption maxima of [Cu(3e)2][PF6], [Cu(5)2][PF6] and [Cu(6e)2][PF6] undergo a significant red-shift. This is rationalised with the interaction between electrons of the metal d-manifold and the ligand π-system, resulting in a more electron-rich system, in turn reducing the HOMO–LUMO gap. The complexes have the following λmax in decreasing order: 494 ≥ 492 > 414 ≥ 413 nm for [Cu(3e)2][PF6], [Cu(2)2][PF6], [Cu(6e)2][PF6] and [Cu(5)2][PF6], respectively. In the spectra of [Cu(3e)2][PF6], [Cu(2)2][PF6], these absorptions are attributed to the MLCT transitions. Whereas for [Cu(6e)2][PF6] and [Cu(5)2][PF6], the lowest energy band appears as an extension towards higher wavelengths of the ILCT bands found in spectra of 6e and 5. It is not possible to unambiguously ascribe the lowest energy transitions to either ILCT or MLCT, thus a contribution of both is likely; this was followed by an investigation with TD-DFT calculations (see later subsection).
The values of λonset present a similar trend to that seen with λmax, with 546, 545, 495 and 487 nm for [Cu(2)2][PF6], [Cu(3e)2][PF6], [Cu(6e)2][PF6] and [Cu(5)2][PF6], respectively. However, in comparison to ligands 3e, 5 and 6e, the complexes exhibit an opposite trend. In fact, with a red-shift of λmax ≈ 141 nm, ligand 3e dominates over the increase in λmax when ligands 5 and 6e are coordinated to copper(I) (Δλ ≈ 28 and 23 nm for [Cu(5)2][PF6] and [Cu(6e)2][PF6], respectively). This can be rationalised in terms of the steric hindrance induced by the ligands around the metal centre: the methyl groups in 3e allow a partial distortion of the complex which is closer to that typical of copper(II), decreasing the MLCT energy and causing a red-shift, whereas the alkenyl-TPA groups in ligands 5 and 6e entangle the metal in a constrained geometry.
The electrochemical behaviour of ligands and complexes was investigated by cyclic voltammetry in CH2Cl2 with 0.1 mol dm−3 [nBu4N][PF6] as supporting electrolyte. The cyclic voltammograms are summarised in Fig. S76.† All were referenced to the Fc/Fc+ redox couple, and then all the values transposed against NHE.71 Ligands 4e, 5 and 6e show multiple forward irreversible oxidations. The complex pattern of the voltammograms follows from the structure of the ligands. The presence of aromatic amines (easily subjected to oxidation) may be source of subsequent chemical transformations, modifying the electrochemistry of the compound and hampering the its reversibility (ECE processes). The copper(I) complexes also present a pattern of multiple irreversible oxidation processes. This is consistent with that observed for the ligands, which are an integral part of the complexes. For these reasons, we based our analysis exclusively on Epa of the first forward oxidations. Ligands 4e, 5 and 6e present similar Epa values of +0.96, +0.95 and +0.92 V, while 3e shows a more positive Epa of +1.23 V. This difference is attributable to the alkenyl-TPA-substituted bpys bearing multiple electrondonating amine. Overall, alkenyl-TPA groups play an important role in the electrochemical properties of ligands 4e, 5 and 6e, whereas the 4,4′-phenylene substituents do not seem to have a significant effect.
The complexes show positive Epa values of +1.04, +1.03, +1.03, +0.87 V for [Cu(3e)2][PF6], [Cu(2)2][PF6], [Cu(6e)2][PF6] and [Cu(5)2][PF6] respectively.
Assuming the Epa for the first oxidation is representative of the HOMO levels of the compounds, we can derive the LUMO energies by means of the optical bandgap (Eoptg, by extrapolating the intersecting abscissa from the linear section of the lowest energy band in UV-visible spectra) and assemble the energy diagram displayed in Fig. 5. It is possible to take the [Cu(Lancillary)2]+ homoleptic complex as representative for the properties of corresponding [Cu(Lanchor)(Lancillary)]+ heteroleptic complex; therefore, we considered [Cu(1)(2)]+ and [Cu(4)(5)]+ photophysical and electrochemical properties represented by compounds [Cu(2)2][PF6] and [Cu(5)2][PF6] as extensively seen in the literature.72
The EHOMO of compounds 3e, 4, 6e, [Cu(2)2][PF6], [Cu(3e)2][PF6], [Cu(6e)2][PF6] and [Cu(5)2][PF6] show values between 0.87–1.23 eV, below the redox potential of the I2/I3− (≈+0.40 eV vs. NHE73), thus, indicating a good driving force for the regeneration of the dyes. The ELUMO are more negative than the redox potential of TiO2 (≈−0.50 eV vs. NHE73), being beneficial for the electron injection. These observations suggest that these new dyes are good candidates for applications in DSC devices.
The Eoptg values of the ligands range from 2.78 to 3.16 eV, whereas the complexes display values between 2.27 and 2.55 eV. In general, this can be explained as the result of both destabilisation of the HOMO and stabilisation of the LUMO occurring upon coordination of the ligands to copper(I). Exception is made for ligand 6e, where the complexation stabilises both the HOMO and LUMO energies (Table 1, from +0.92 and −1.90 eV to +1.03 and −1.48 eV for EHOMO and ELUMO of 6e and [Cu(6e)2][PF6], respectively).
The structures were minimised at MM2 level and this geometry was used as the input for a single point DFT calculation (B3LYP 6-31G* basis set level with polar solvent) to determine the orbital distributions. Ligands 3 and 6 exhibit analogous HOMO and LUMO characteristics (Fig. S77†): the HOMO is localised on the TPA and the MeOTPA units, reaching the vicinal pyridine ring. Both LUMOs extend through the bpy domain to the phosphonic acid, which is beneficial for the anchor character of the dyes. This MO distribution is in line with the chemical environment found in NMR analysis of ligands 3 and 6 and consistent with the architecture of DπA dyes. An inspection of the HOMO-manifold of 6 reveals orbitals distributed over the alkenyl-TPA units. The symmetry in [Cu(3)2]+ and [Cu(6)2]+ was found to affect the energies of the MOs, delivering pairs of degenerate orbitals at the beginning of both manifolds (Table S2†). The HOMO/HOMO−1 are centred on the metal in both complexes. The degenerate HOMO/HOMO−1 of [Cu(6)2]+ show equal contributions from the MeOTPA groups from both ligands besides that of the metal (Fig. 6c and d). The HOMO/HOMO−1 of [Cu(4)(5)]+, likewise degenerate, with both MeOTPA from 5 taking part to the delocalisation (Fig. 6g and h). The donor contribution of alkenyl-TPA units is found in [Cu(4)(5)]+ and [Cu(6)2]+ in the lower HOMO-manifold (Fig. S78†). In LUMO/LUMO+1 orbitals of [Cu(4)(5)]+ and [Cu(6)2]+ a neat difference between the two complex architectures is found. The LUMO/LUMO+1 in [Cu(4)(5)]+ are non-degenerate: whereas the LUMO is fully localised over the bpy domain and the acceptor groups of Lanchor 4, the LUMO+1 is limited to the bpy domain of Lancillary 5 (Fig. 6e). The [Cu(6)2]+ LUMO/LUMO+1 are degenerate orbitals, instead, each of which is delocalised between the bpy domain and the phosphonic acid (Fig. 6a and b). The same conclusions can be made assessing both [Cu(3)2]+ and [Cu(1)(2)]+ MOs. This may be detrimental for transitions to LUMO+1 (devoid of anchor character) in ‘push–pull’ architecture, in contrast with [Cu(DπA)2]+ architecture where the LUMO/LUMO+1 degeneracy makes it possible to deliver electron injection from both orbitals.
Fig. 6 Character of MOs from LUMO+1 to HOMO−1 for [Cu(6)2]+ (left column, (a–d) and [Cu(4)(5)]+ (right column, (e–h) calculated at a DFT level 6-31G* basis set in polar solvent. |
Fig. 10 Day 0 J–V curves measured for cells with dyes 3, 4 and 6. The inset shows curves referred to that of N719. |
Dye | JSC/mA cm−2 | VOC/mV | FF/% | η/% | ηrel./% |
---|---|---|---|---|---|
N719 | 15.02 | 615 | 59 | 5.42 | 100.0 |
3 c1 | 2.47 | 565 | 65 | 0.91 | 15.7 |
3 c2 | 2.51 | 578 | 65 | 0.94 | 16.2 |
3 c3 | 2.52 | 570 | 65 | 0.94 | 16.2 |
3 c4 | 2.32 | 567 | 66 | 0.86 | 14.9 |
3 Average | 2.45 ± 0.09 | 570 ± 6 | 65 | 0.91 ± 0.04 | 16.8 ± 0.7 |
4 c1 | 4.74 | 527 | 72 | 1.79 | 33.0 |
4 c2 | 5.19 | 516 | 72 | 1.92 | 35.5 |
4 c3 | 5.05 | 514 | 71 | 1.85 | 34.2 |
4 c4 | 4.68 | 512 | 72 | 1.72 | 31.8 |
4 Average | 4.91 ± 0.24 | 517 ± 7 | 72 | 1.82 ± 0.09 | 33.6 ± 1.6 |
6 c1 | 5.28 | 551 | 66 | 1.93 | 35.7 |
6 c2 | 5.33 | 563 | 67 | 2.01 | 37.1 |
6 c3 | 5.54 | 546 | 67 | 2.01 | 37.1 |
6 c4 | 4.96 | 563 | 69 | 1.92 | 35.5 |
6 Average | 5.28 ± 0.24 | 556 ± 8 | 67 ± 1 | 1.97 ± 0.05 | 36.4 ± 0.9 |
Fig. 11 Day 0 J–V curves measured for cells with dyes [Cu(3)(3–H)], [Cu(1)(2)]+, [Cu(6)(6–H)] and [Cu(4)(5)]+. The inset shows curves referred to that of N719. |
Dye | JSC/mA cm−2 | VOC/mV | FF/% | η/% | ηrel./% |
---|---|---|---|---|---|
a From electrodes functionalised with method b, see Fig. 7.b Set and parameters from our previous work.29c From electrodes functionalised with method a. | |||||
N719 | 15.02 | 615 | 59 | 5.42 | 100.0 |
[Cu(3)(3–H)]a c1 | 5.11 | 600 | 61 | 1.88 | 32.5 |
[Cu(3)(3–H)]a c2 | 4.35 | 598 | 65 | 1.68 | 29.1 |
[Cu(3)(3–H)]a c3 | 4.53 | 607 | 65 | 1.79 | 30.9 |
[Cu(3)(3–H)]a c4 | 5.07 | 593 | 64 | 1.94 | 33.5 |
[Cu(3)(3–H)] average | 4.77 ± 0.38 | 599 ± 6 | 64 ± 2 | 1.82 ± 0.11 | 33.6 ± 2.1 |
[Cu(1)(2)]+b,c c1 | 4.82 | 532 | 68 | 1.74 | 32.2 |
[Cu(1)(2)]+b,c c2 | 5.25 | 523 | 70 | 1.93 | 35.7 |
[Cu(1)(2)]+b,c c3 | 4.64 | 536 | 69 | 1.71 | 31.5 |
[Cu(1)(2)]+b,c c4 | 4.89 | 533 | 67 | 1.74 | 32.1 |
[Cu(1)(2)]+ average | 4.90 ± 0.26 | 531 ± 6 | 68 ± 1 | 1.78 ± 0.10 | 32.9 ± 3.2 |
[Cu(6)(6–H)]a c1 | 6.77 | 564 | 65 | 2.48 | 45.9 |
[Cu(6)(6–H)]a c2 | 6.84 | 565 | 67 | 2.59 | 47.9 |
[Cu(6)(6–H)] average | 6.81 ± 0.05 | 564 ± 1 | 66 ± 1 | 2.54 ± 0.08 | 46.9 ± 0.24 |
[Cu(4)(5)]+c c1 | 4.92 | 507 | 70 | 1.76 | 32.5 |
[Cu(4)(5)]+c c2 | 4.92 | 501 | 71 | 1.76 | 32.5 |
[Cu(4)(5)]+c c3 | 4.79 | 501 | 71 | 1.70 | 31.4 |
[Cu(4)(5)]+c c4 | 4.97 | 504 | 72 | 1.80 | 33.3 |
[Cu(4)(5)]+ average | 4.90 ± 0.08 | 503 ± 3 | 71 ± 1 | 1.76 ± 0.04 | 32.4 ± 0.8 |
The heteroleptic dye [Cu(1)(2)]+ exhibits a slightly higher JSC of 4.90 mA cm−2 than [Cu(3)(3–H)] (4.77 mA cm−2). However, the major contribution to the PCE of [Cu(3)(3–H)] is found in the VOC with 599 mV (against 531 mV for [Cu(1)(2)]+, Table 3, visible in Fig. 11). It should be noted that whereas Lanchor 1 has two phosphonic acid substituents and ligand 3 has one, the complex bearing ligand 3 has a higher VOC. A slightly higher fill factor (FF) is reported for [Cu(1)(2)]+ with respect to that of [Cu(3)(3–H)] (68%, against 64%, respectively). The performances were monitored for a week after day 0 (Tables S3 and S4†). The measurements revealed a higher stability of [Cu(3)(3–H)] versus [Cu(1)(2)]+; especially on going from day 0 to day 7, a decrease in η of about 5% was observed in the case of [Cu(3)(3–H)] compared to 17% observed for [Cu(1)(2)]+. As seen with these two complex congeners, the [Cu(DπA)2]+ design can perform as well as the classic [Cu(Lanchor)(Lancillary)]+ design.
However, the comparison of complexes [Cu(6)(6–H)] and [Cu(4)(5)]+ reveals the potential of the [Cu(DπA)2]+ dye design. The gap in terms of PCE is large: while cells functionalised with [Cu(4)(5)]+ have an average η of 1.76%, those functionalised with [Cu(6)(6–H)] deliver an average of 2.54%, corresponding a relative PCE of 46.9% (ηrel, relative to an N719 reference), delivering the highest PCE for a homoleptic bis-diimine copper(I) complex and one of the highest PCEs attained with a copper(I)-based dye (without the use of a co-absorbent or co-sensitisation43,50). The complex [Cu(6)(6–H)] exhibits about 60 mV more than [Cu(4)(5)]+ cells (red and yellow curves, Fig. 11). Again, it is the case that when the anchoring unit belongs to an asymmetrical DπA ligand, the VOC of the corresponding complex is higher than that of its heteroleptic congener. The largest contribution to the PCE comes from the JSC, with values as high as 6.81 mA cm−2 for [Cu(6)(6–H)] (average of two duplicates, Table 3), versus 4.90 mA cm−2 measured with [Cu(4)(5)]+, having a determining impact on the PCEs.
We found the similar performances of complexes [Cu(3)(3–H)] and [Cu(1)(2)]+ encouraging, proving that dyes with the [Cu(DπA)2]+ design can deliver PCEs in the same range as those dyes with [Cu(Lanchor)(Lancillary)]+ design. Furthermore, the PCE behaviour from day 0 through day 7 of [Cu(3)(3–H)] may be indicative of a stabilising influence brought by the phosphonic acid located on each asymmetrical DπA ligand. Of greater significance is the comparison of complexes [Cu(6)(6–H)] and [Cu(4)(5)]+ which shows that shifting the dye design from [Cu(Lanchor)(Lancillary)]+ to [Cu(DπA)2]+ achieves a higher PCE than the ‘push–pull’ architecture.
The PCE of ligand 3 is 0.91% whereas that of complex [Cu(3)(3–H)] is 1.82%. The JSC of ligand 3 is 2.45 mA cm−2 (Table 2) whereas that delivered by the homoleptic complex [Cu(3)(3–H)] is 4.77 mA cm−2 (Table 3). As seen in the EQE spectra (Fig. 12), complex [Cu(3)(3–H)] is redshifted (λmax 470 nm) and more intense than that of ligand 3 (450 nm). An increase in VOC is observed for [Cu(3)(3–H)] (Table 3, 599 mV) compared to ligand 3 (Table 3, 570 mV).
This proves that an organic dye with a metal-binding domain can be coordinated to a metal centre resulting in an improvement in the DSC performances. We believe that this is the first time that this phenomenon has been observed.
Cells sensitised with ligand 6 exhibit a PCE of 1.97%, whereas those with [Cu(6)(6–H)] show values about 30% higher (Table 3, 2.54%). As with ligand 3, coordination of ligand 6 to copper(I) improves the DSC performance, having a greater impact on the JSC with values of 5.28 mA cm−2 (Table 2) for 6 and 6.81 mA cm−2 (Table 3) for the corresponding complex [Cu(6)(6–H)]. This is also visible in the EQE spectra in Fig. 12, where complex [Cu(6)(6–H)] exhibits higher EQEs than 6 over the whole spectral region. It is worth noting that both dyes show a broad spectrum with λonset at approximately 660 nm. The spectra feature similar shapes, with maximum EQE around 370 nm gradually decreasing towards a shoulder at about 550 nm in 6 and 560 nm in [Cu(6)(6–H)] (EQE values ranging from 71 to 40% in [Cu(6)(6–H)], from 60 to 28% in 6). Again, we observe a beneficial effect on the performances when the organic dye is coordinated.
Finally, it is worth spending a word on the comparison of [Cu(3)(3–H)] and [Cu(6)(6–H)]. It was shown how the complexation to copper(I) of both ligands 3 and 6 to the respective complexes exhibited enhanced the performances. However, copper(I) has a greater impact on 3 than 6, with PCE improvement rates of 100% and 29% in [Cu(3)(3–H)] and [Cu(6)(6–H)], respectively. We attribute the discrepancy to the overlap between the absorption region of the two ligands with respect to that of the corresponding complexes (Fig. 4). The alkenyl-TPA substituents in 6e (λmax = 391, Table 1) extend the spectral absorption of the ligand towards the visible region of copper(I) (ranging between 400 and 500 nm with diimine ligands11), as opposed to the case of 3e (λmax = 353, Table 1) and its complex. Hence, it is desirable to limit the absorption of the ligand away from that of copper(I) and benefit from the absorption of both the ligand and the metal.
Dye and cell number | [Cu(CH3CN)4][PF6]/mM | JSC/mA cm−2 | VOC/mV | FF/% | η/% | ηrel./% |
---|---|---|---|---|---|---|
N719 | — | 15.02 | 615 | 59 | 5.42 | 100.0 |
3 c1 | 0.01 | 2.73 | 591 | 68 | 1.09 | 20.2 |
3 c2 | 0.01 | 1.64 | 561 | 61 | 0.56 | 10.4 |
3 c3 | 0.01 | 3.04 | 579 | 67 | 1.18 | 21.8 |
3 c4 | 0.01 | 2.18 | 582 | 63 | 0.81 | 14.9 |
Average | — | 2.40 ± 0.62 | 578 ± 13 | 65 ± 3 | 0.91 ± 0.28 | 16.8 ± 5.2 |
3 c1 | 0.1 | 4.20 | 588 | 68.0 | 1.68 | 31.0 |
3 c2 | 0.1 | 4.25 | 578 | 69.6 | 1.71 | 31.6 |
3 c3 | 0.1 | 4.20 | 578 | 70.6 | 1.72 | 31.7 |
3 c4 | 0.1 | 4.39 | 597 | 68.8 | 1.80 | 33.3 |
Average | — | 4.26 ± 0.09 | 585 ± 9 | 69 ± 1 | 1.73 ± 0.05 | 31.9 ± 1.0 |
3 c1 | 1.0 | 3.12 | 543 | 65.3 | 1.11 | 20.4 |
3 c2 | 1.0 | 3.40 | 554 | 64.4 | 1.21 | 22.4 |
3 c3 | 1.0 | 3.30 | 554 | 63.5 | 1.16 | 21.4 |
Average | — | 3.27 ± 0.14 | 550 ± 6 | 64 ± 1 | 1.16 ± 0.05 | 21.4 ± 1.0 |
6 c1 | 0.1 | 7.31 | 537 | 66 | 2.60 | 48.1 |
6 c2 | 0.1 | 7.14 | 535 | 65 | 2.50 | 46.1 |
6 c3 | 0.1 | 7.18 | 538 | 64 | 2.46 | 45.3 |
6 c4 | 0.1 | 6.92 | 539 | 66 | 2.45 | 45.2 |
Average | — | 7.14 ± 0.16 | 537 ± 1 | 65 ± 1 | 2.50 ± 0.07 | 46.2 ± 1.3 |
Fig. 13 Day 0 average EQE spectra of cells built with ‘method c’ with varying concentrations of copper(I). |
It is not surprising that when ligand-functionalised electrodes are exposed to different concentrations of copper(I) salt, the performances vary according to the amount of copper in solution; we have previously reported similar effects in which different concentrations of [Cu(Lancillary)2]+ impacted the performances of devices assembled using ‘method a’ (Fig. 9).68
Normalised solid-state absorption spectra are presented in Fig. 14. An electrode functionalised with 3 (showing no absorption in the visible region) was immersed into a solution with [Cu(I)] = 0.1 mM (red, Fig. 14); the corresponding band overlays almost perfectly with that of an electrode dipped into a solution of complex [Cu(3)(3–H)] (Fig. 14, dotted line in blue). Thus, the two bear extremely similar (if not equal) photophysical properties. This is consistent with the concept that the desired species is formed on surface by means of a stepwise approach, similar to the assembly of heteroleptic copper(I) complexes (SALSAC, method a). Additionally, we demonstrate that a surface-bound organic photosensitiser can coordinate to copper(I) and form the corresponding complex on surface.
Fig. 14 Normalised solid state UV-visible spectra of electrodes functionalised with dyes 3 + [Cu(I)] = 0.1 mM and [Cu(3)(3–H)]. |
The same methodology was applied to electrodes functionalised with ligand 6. Cells with electrodes immersed in a 0.1 mM solution of [Cu(CH3CN)4][PF6] exhibited an average PCE of 2.50%, essentially the same as that of cells sensitised with [Cu(6)(6–H)] (Table 4, 2.54%). The JSC of 7.14 mA cm−2 is remarkably high when compared to that measured for devices with ligand 6 alone (Table 2, 5.28 mA cm−2), slightly higher than that of cells sensitised with [Cu(6)(6–H)] (6.81 mA cm−2). Instead, the VOC of 537 mV is lower than that of 6 and [Cu(6)(6–H)] (Tables 2 and 3, 556 and 564 mV, respectively). The EQE spectrum matches almost perfectly that of [Cu(6)(6–H)] (dashed and bold line in red, Fig. 13). With this, not only we see again that the assembly on surface with method c is possible, but also that in this case the performances are well representative of a dye with [Cu(DπA)2]+ design (namely [Cu(6)(6–H)]), whose cells are built with method b.
The TD-DFT study emphasises that the lowest energy transition is MLCT-based in [Cu(3e)2]+, but it is predominantly ILCT in [Cu(6e)2]+. From the DSC performances, we observe that: (i) a comparison of [Cu(3)(3–H)] and [Cu(1)(2)]+ shows that the [Cu(DπA)2]+ design can deliver as good efficiencies as the traditional [Cu(Lanchor)(Lancillary)]+ design, and (ii) a comparison of [Cu(6)(6–H)] and [Cu(4)(5)]+ confirms that the [Cu(DπA)2]+ design surpasses the performances of the [Cu(Lanchor)(Lancillary)]+ dye. [Cu(6)(6–H)] displays a PCE of 2.54% (46.9% relative to N719), which is the highest PCE for a homoleptic bis(diimine) copper(I) complex and among the highest PCEs attained by copper(I)-based dyes in DSCs.
We have also compared the asymmetrical DπA ligands as organic dyes with their homoleptic copper(I) complexes. The DSC performances demonstrate that coordination to copper(I) increases the PCEs on going from 3 and 6 (0.91 and 1.97%) to [Cu(3)(3–H)] and [Cu(6)(6–H)] (1.82 and 2.54%), respectively.
In conclusion, we have demonstrated the appealing potential of asymmetrical DπA 2,2′-bpy derivatives for applications in DSCs both as organic dyes and as sensitisers in the form of their copper(I) complexes. The [Cu(DπA)2]+ dyes perform better than the corresponding organic DπA dyes in DSCs. Although the DπA design is well-established for organic dyes, its implementation in ligands and subsequently their coordination compounds is, to the best of our knowledge, new. We hope that these results with encourage further investigations of asymmetrical DπA ligands and their copper(I) complexes.
Footnote |
† Electronic supplementary information (ESI) available: Experimental section. Fig. S1–S45; NMR spectra; Fig. S46–S60: FT-IR spectra; HR-MS spectra: S61–S75; Table S1: DOSY data; Fig. S76: cyclic voltammograms; Fig. S77, S78 and Table S2: DFT calculations; J–V parameters: Tables S3–S8. CCDC 2203243. For ESI and crystallographic data in CIF or other electronic format see DOI: https://doi.org/10.1039/d3ra00437f |
This journal is © The Royal Society of Chemistry 2023 |