Copper Complexes with Tetradentate Ligands for Enhanced Charge Transport in Dye-Sensitized Solar Cells

In dye-sensitized solar cells (DSCs), the redox mediator is responsible for the regeneration of the oxidized dye and for the hole transport towards the cathode. Here, we introduce new copper complexes with tetradentate 6,6′-bis(4-(S)-isopropyl-2-oxazolinyl)-2,2′-bipyridine ligands, Cu(oxabpy), as redox mediators. Copper coordination complexes with a square-planar geometry show low reorganization energies and thus introduce smaller losses in photovoltage. Slow recombination kinetics of excited electrons between the TiO2 and CuII(oxabpy) species lead to an exceptionally long electron lifetime, a high Fermi level in the TiO2, and a high photovoltage of 920 mV with photocurrents of 10 mA·cm−2 and 6.2% power conversion efficiency. Meanwhile, a large driving force remains for the dye regeneration of the Y123 dye with high efficiencies. The square-planar Cu(oxabpy) complexes yield viscous gel-like solutions. The unique charge transport characteristics are attributed to a superposition of diffusion and electronic conduction. An enhancement in charge transport performance of 70% despite the higher viscosity is observed upon comparison of Cu(oxabpy) to the previously reported Cu(tmby)2 redox electrolyte.


Introduction
A steep incline in energy demands, paired with an increased awareness of climate change, has driven the development of renewable energy sources [1].Technologies for harvesting solar energy have evolved and photovoltaics has become a major focus in today's politics and research.Dye-sensitized solar cells (DSCs), as first described by Gerischer [2] and implemented by Grätzel [3], provide a low-cost alternative for large-scale solar energy conversion based on inexpensive and Earth-abundant materials [4][5][6].Since the beginning of DSC development in 1991, impressive power conversion efficiencies above 14% have been demonstrated [7].Light absorption in a DSC is based on dye molecules anchored to a porous metal oxide semiconductor (typically TiO 2 , Figure 1a).Photon absorption leads to injection of electrons from the excited dye molecules (D*) into the conduction band of the semiconductor and subsequently an external circuit, while the dye Inorganics 2018, 6, 53; doi:10.3390/inorganics6020053www.mdpi.com/journal/inorganics is regenerated by a redox electrolyte.Crucial photovoltaic parameters of solar cells include the photocurrent, determined by light absorption and carrier injection by the dye, and the photovoltage, largely dependent on the molecular engineering of the redox mediators.Initially, the iodide-triiodide redox couple played an important role in the electrolyte development [8].The I − /I 3 − electrolyte allows high photocurrents by reducing recombination losses [9].However, it also carries drawbacks such as potential drops and competitive light absorption [10,11].In 2010, Feldt et al. demonstrated high-efficiency DSCs with Co II/III (bpy) 3 as redox mediators [12].Upon further investigation, however, deficient diffusion through the mesoporous TiO 2 layer limited the photovoltaic performance of these bulky octahedral tris-bipyridine-based cobalt mediators [13][14][15].Furthermore, cobalt complexes showed limited stability especially under irradiative conditions due to e.g., ligand exchanges, impeding their potential for large-scale applications [16].Cu I/II -based redox shuttles emerged as premiere high-performing, tunable redox mediators and stand out through straightforward synthetic paths and high photovoltages based on low-cost materials.Hattori et al. introduced for the first time bis(2,9-dimethyl-1,10-phenanthroline)copper(I/II) complexes as redox mediators with 1.4% power conversion efficiency (PCE) [17].Consecutively, PCEs of 7% with the C218 dye were achieved by Bai et al.And a 8.3% PCE was presented by Freitag et al. with the organic D-π-A dye LEG4 [18,19].The mark of 10% PCE was surpassed for copper-based DSCs when Saygili et al. introduced copper bipyridyl redox mediators in 2016 [20].Recent results present improved PCEs up to 11.7% [21,22].DSCs, however, carry a disadvantage in the usage of liquid volatile electrolytes.Difficulties concerning encapsulation have thus far impeded the commercial production of DSCs.Therefore, the development of electrolytes with higher viscosity shows higher potential for commercialization [23].Furthermore, solid-state hole transport materials (HTMs) evolved as solvent-free hole transport materials based on organic hole conductors as well as metal complexes [24,25].
Reorganization energies upon changes of oxidation state are one reason for performance losses in DSCs mediated by metal complexes [26][27][28].In previously reported approaches, this reorganization was prevented by ligands with sterically hindering groups to suppress coordination changes upon charging/discharging [18,20].
Inorganics 2018, 6, x FOR PEER REVIEW 2 of 17 band of the semiconductor and subsequently an external circuit, while the dye is regenerated by a redox electrolyte.Crucial photovoltaic parameters of solar cells include the photocurrent, determined by light absorption and carrier injection by the dye, and the photovoltage, largely dependent on the molecular engineering of the redox mediators.Initially, the iodide-triiodide redox couple played an important role in the electrolyte development [8].The I − /I3 − electrolyte allows high photocurrents by reducing recombination losses [9].However, it also carries drawbacks such as potential drops and competitive light absorption [10,11].In 2010, Feldt et al. demonstrated high-efficiency DSCs with Co II/III (bpy)3 as redox mediators [12].Upon further investigation, however, deficient diffusion through the mesoporous TiO2 layer limited the photovoltaic performance of these bulky octahedral tris-bipyridine-based cobalt mediators [13][14][15].Furthermore, cobalt complexes showed limited stability especially under irradiative conditions due to e.g., ligand exchanges, impeding their potential for large-scale applications [16].Cu I/II -based redox shuttles emerged as premiere highperforming, tunable redox mediators and stand out through straightforward synthetic paths and high photovoltages based on low-cost materials.[20].Recent results present improved PCEs up to 11.7% [21,22].DSCs, however, carry a disadvantage in the usage of liquid volatile electrolytes.Difficulties concerning encapsulation have thus far impeded the commercial production of DSCs.Therefore, the development of electrolytes with higher viscosity shows higher potential for commercialization [23].Furthermore, solid-state hole transport materials (HTMs) evolved as solvent-free hole transport materials based on organic hole conductors as well as metal complexes [24,25].
Reorganization energies upon changes of oxidation state are one reason for performance losses in DSCs mediated by metal complexes [26][27][28].In previously reported approaches, this reorganization was prevented by ligands with sterically hindering groups to suppress coordination changes upon charging/discharging [18,20].In the present work, we report a copper(I/II) redox couple with a tetradentate (6,6 -bis(4-(S)-isopropyl-2-oxazolinyl)-2,2 -bipyridine) ligand, Cu I/II (oxabpy) (Figure 1b).Linking all four coordination sites into one ligand molecule greatly reduces geometry re-arrangements between different copper oxidation states, compared to mono-or bidentate ligands [29,30].Therefore, charges can be mediated in the Cu(oxabpy) redox electrolyte more rapidly and with fewer potential losses.In addition to smaller geometry rearrangements, the tetradentate ligand also impedes ligand exchange or loss processes [31].Low absorption coefficients in the visible range render competitive light absorption negligible.Recombinative charge transfer of electrons from the TiO 2 to the Cu II species is largely suppressed for the Cu(oxabpy) electrolyte, leading to a high Fermi level in the TiO 2 .Therefore, open-circuit potentials of 920 mV, 9.75 mA•cm −2 photocurrent and 6.2% PCE were reached in Y123-sensitized DSCs despite the comparatively low redox potential (0.66 V vs. normal hydrogen electrode (NHE)) compared to previously reported Cu-based redox mediators [20].Furthermore, the square planar coordination geometry of Cu(oxabpy) complexes enables superior charge transport properties.Cu(oxabpy) electrolytes exhibit a more than 2.5 times higher viscosity relative to the previously reported Cu(tmby) 2 redox system.Nonetheless, the Cu(oxabpy) electrolyte shows a 1.6 times higher conductivity at similar bias potentials.

Characterization of Cu(oxabpy) Complexes
Absorption and photoluminescence spectra of Cu I/II (oxabpy) complexes are shown in Figure 2. Absorption and photoluminescence maxima as well as extinction coefficients are tabulated together with the formal redox potential in Table 1.The redox potential was obtained through the cyclic voltammogram of the Cu I (oxabpy) complex (see Figure S1).
In the present work, we report a copper(I/II) redox couple with a tetradentate (6,6′-bis(4-(S)isopropyl-2-oxazolinyl)-2,2′-bipyridine) ligand, Cu I/II (oxabpy) (Figure 1b).Linking all four coordination sites into one ligand molecule greatly reduces geometry re-arrangements between different copper oxidation states, compared to mono-or bidentate ligands [29,30].Therefore, charges can be mediated in the Cu(oxabpy) redox electrolyte more rapidly and with fewer potential losses.In addition to smaller geometry rearrangements, the tetradentate ligand also impedes ligand exchange or loss processes [31].Low absorption coefficients in the visible range render competitive light absorption negligible.Recombinative charge transfer of electrons from the TiO2 to the Cu II species is largely suppressed for the Cu(oxabpy) electrolyte, leading to a high Fermi level in the TiO2.Therefore, open-circuit potentials of 920 mV, 9.75 mA•cm −2 photocurrent and 6.2% PCE were reached in Y123-sensitized DSCs despite the comparatively low redox potential (0.66 V vs. normal hydrogen electrode (NHE)) compared to previously reported Cu-based redox mediators [20].Furthermore, the square planar coordination geometry of Cu(oxabpy) complexes enables superior charge transport properties.Cu(oxabpy) electrolytes exhibit a more than 2.5 times higher viscosity relative to the previously reported Cu(tmby)2 redox system.Nonetheless, the Cu(oxabpy) electrolyte shows a 1.6 times higher conductivity at similar bias potentials.

Characterization of Cu(oxabpy) Complexes
Absorption and photoluminescence spectra of Cu I/II (oxabpy) complexes are shown in Figure 2. Absorption and photoluminescence maxima as well as extinction coefficients are tabulated together with the formal redox potential in Table 1.The redox potential was obtained through the cyclic voltammogram of the Cu I (oxabpy) complex (see Figure S1).
The optical absorption spectrum of Cu I (oxabpy) shows a first optical transition associated with the HOMO-LUMO transition around 520 nm.This is confirmed by the calculated weak oscillator strength for the first vertical transition computed with time-dependent DFT (TD-DFT) (see below for computational details, corresponding data are listed in Table S1).The absorption maximum at 415 nm is attributed to excitation of electrons from the HOMO − 2 to the LUMO (Figure S5).The oscillator  697 93 * The formal potential E 0 of the redox system was determined from anodic and cathodic current peaks in the cyclic voltammogramm of the Cu I species (see Figure S1).
The optical absorption spectrum of Cu I (oxabpy) shows a first optical transition associated with the HOMO-LUMO transition around 520 nm.This is confirmed by the calculated weak oscillator strength for the first vertical transition computed with time-dependent DFT (TD-DFT) (see below for computational details, corresponding data are listed in Table S1).The absorption maximum at 415 nm is attributed to excitation of electrons from the HOMO − 2 to the LUMO (Figure S5).The oscillator strength for this transition is calculated to be two orders of magnitude higher than for the first transition.An extinction coefficient of 1587 M −1 cm −1 was measured for Cu I (oxabpy) and negligible absorption was observed around 700 nm for the Cu II (oxabpy).Therefore, the Cu(oxabpy) electrolyte competes for light absorption significantly less than previously introduced Cu-based electrolytes (e.g., 5300 M −1 cm −1 for Cu I (tmby) 2 ) [20].
The formal potential of the Cu(oxabpy) redox system is found at 0.66 V vs. NHE.It is thus situated significantly lower compared to previously reported Cu-based redox electrolytes (e.g., 0.87 V for Cu(tmby) 2 ) [20].
Electrolyte solutions with the Cu I/II (oxabpy) redox couple were gel-like in appearance.To study the gelation of Cu(oxabpy) complexes, viscosity measurements were conducted for Cu(oxabpy) with respect to the Cu(tmby) 2 redox system in Cu I 5:1 Cu II composition (as in the electrolyte).The flow speed through a capillary measured with a ViscoClock showed that the addition of Cu(oxabpy) complexes to chloroform lead to a 2.6 times greater increase of the solution viscosity compared to the addition of an equal amount of the Cu(tmby) 2 couple (see Figure S2).
Raman spectra of Cu I/II (oxabpy) complexes are shown in Figure 3.
Inorganics 2018, 6, x FOR PEER REVIEW 4 of 17 strength for this transition is calculated to be two orders of magnitude higher than for the first transition.An extinction coefficient of 1587 M −1 cm −1 was measured for Cu I (oxabpy) and negligible absorption was observed around 700 nm for the Cu II (oxabpy).Therefore, the Cu(oxabpy) electrolyte competes for light absorption significantly less than previously introduced Cu-based electrolytes (e.g., 5300 M −1 cm −1 for Cu I (tmby)2) [20].
The formal potential of the Cu(oxabpy) redox system is found at 0.66 V vs. NHE.It is thus situated significantly lower compared to previously reported Cu-based redox electrolytes (e.g., 0.87 V for Cu(tmby)2) [20].
Electrolyte solutions with the Cu I/II (oxabpy) redox couple were gel-like in appearance.To study the gelation of Cu(oxabpy) complexes, viscosity measurements were conducted for Cu(oxabpy) with respect to the Cu(tmby)2 redox system in Cu I 5:1 Cu II composition (as in the electrolyte).The flow speed through a capillary measured with a ViscoClock showed that the addition of Cu(oxabpy) complexes to chloroform lead to a 2.6 times greater increase of the solution viscosity compared to the addition of an equal amount of the Cu(tmby)2 couple (see Figure S2).
Raman spectra of Cu I/II (oxabpy) complexes are shown in Figure 3. Modes between 250 cm −1 and 410 cm −1 as well as at 749 cm −1 in the spectra of Cu(oxabpy)TFSI and Cu(oxabpy)TFSI2 were identified to origin from the TFSI counterion (black boxes).Furthermore, features at 1131 cm −1 and 1247 cm −1 were associated with the TFSI counterion.The TFSI modes are generally found in higher relative intensity for the Cu II (oxabpy) compared to the Cu I (oxabpy), confirming the presence of two TFSI counterions for the Cu II (oxabpy) complex.The Cu II (oxabpy) complex shows strong modes at 1003 cm −1 and 1040 cm −1 , which can be affiliated with a C-N stretch and a pyridine breathing mode as previously reported for similar compounds [32].The breathing mode, however, is not visible in the spectrum of Cu I (oxabpy).This is attributed to the longer bond to the copper metal center (see Figure 4), which impedes this vibration and serves as a strong signature to distinguish the coordination of Cu I or Cu II to the pyridine nitrogen.Additional modes at 630 cm −1 and 668/680 cm −1 in the Cu I/II (oxabpy) complexes are attributed to the oxazoline units.A pair of bands arises at 668/680 cm −1 upon coordination in Cu II (oxabpy), while the vibration at 680 cm −1 is again not visible in Cu I (oxabpy).

Density Functional Theory Calculations
In order to estimate the reaction-free energies and internal reorganization energies for the Cu I/II (oxabpy) complexes, DFT calculations with the Gaussian 16 suite of programs [33] were performed at the B3LYP-D3B [34] level of theory [34][35][36], with SDD as basis set for Cu and 6-31++G(d,p) basis set for other atoms.Structural optimizations, molecular frequencies, and thermochemistry data have been computed in acetonitrile by applying the polarizable continuum Modes between 250 cm −1 and 410 cm −1 as well as at 749 cm −1 in the spectra of Cu(oxabpy)TFSI and Cu(oxabpy)TFSI 2 were identified to origin from the TFSI counterion (black boxes).Furthermore, features at 1131 cm −1 and 1247 cm −1 were associated with the TFSI counterion.The TFSI modes are generally found in higher relative intensity for the Cu II (oxabpy) compared to the Cu I (oxabpy), confirming the presence of two TFSI counterions for the Cu II (oxabpy) complex.The Cu II (oxabpy) complex shows strong modes at 1003 cm −1 and 1040 cm −1 , which can be affiliated with a C-N stretch and a pyridine breathing mode as previously reported for similar compounds [32].The breathing mode, however, is not visible in the spectrum of Cu I (oxabpy).This is attributed to the longer bond to the copper metal center (see Figure 4), which impedes this vibration and serves as a strong signature to distinguish the coordination of Cu I or Cu II to the pyridine nitrogen.Additional modes at 630 cm −1 and 668/680 cm −1 in the Cu I/II (oxabpy) complexes are attributed to the oxazoline units.A pair of bands arises at 668/680 cm −1 upon coordination in Cu II (oxabpy), while the vibration at 680 cm −1 is again not visible in Cu I (oxabpy).

Density Functional Theory Calculations
In order to estimate the reaction-free energies and internal reorganization energies for the Cu I/II (oxabpy) complexes, DFT calculations with the Gaussian 16 suite of programs [33] were performed at the B3LYP-D3B [34] level of theory [34][35][36], with SDD as basis set for Cu and 6-31++G(d,p) basis set for other atoms.Structural optimizations, molecular frequencies, and thermochemistry data have been computed in acetonitrile by applying the polarizable continuum model (PCM) for the solvent degrees of freedom [37].Vertical excitation energies for the Cu I/II (oxabpy) complexes have been computed on the minimum energy structures with time-dependent DFT (TD-DFT) by applying the same level of theory and solvent model [38].
The minimum-energy structures of Cu I/II (oxabpy) complexes are shown in Figure 4.
Inorganics 2018, 6, x FOR PEER REVIEW 5 of 17 model (PCM) for the solvent degrees of freedom [37].Vertical excitation energies for the Cu I/II (oxabpy) complexes have been computed on the minimum energy structures with timedependent DFT (TD-DFT) by applying the same level of theory and solvent model [38].
The minimum-energy structures of Cu I/II (oxabpy) complexes are shown in Figure 4. From analysis of DFT minimum energy geometries for Cu I and Cu II complexes, the most evident feature is the significant shift of the copper position upon oxidation from Cu I to Cu II .While both Cu I and Cu II complexes have a planar configuration of the N4 open ring, the coordination of Cu I tends to involve only the two nitrogen atoms in the oxazoline groups in a quasi linear geometry (Cu-N distances are reported in Figure 4).In the Cu II complex the copper atom is accommodated in a planar square configuration with the four nitrogen atoms with about 2.0 Å of distance.Moreover, the N4 pseudo-ring appears to be more open by ~10% in the case of Cu I than in the Cu II one.These coordination differences have a direct consequence on the computed spectroscopic properties of the complexes, in agreement with the observed experimental differences between the Raman spectra of Cu I /Cu II complexes (see Figure 3).Quenching of the pyridine breathing vibration at 1247 cm −1 for the Cu I (oxabpy) complex proves the alteration of the metal-ligand bonding with respect to Cu II (oxabpy).Therefore, the change in bond lengths as suggested by DFT calculations is confirmed.
I II From analysis of DFT minimum energy geometries for Cu I and Cu II complexes, the most evident feature is the significant shift of the copper position upon oxidation from Cu I to Cu II .While both Cu I and Cu II complexes have a planar configuration of the N4 open ring, the coordination of Cu I tends to involve only the two nitrogen atoms in the oxazoline groups in a quasi linear geometry (Cu-N distances are reported in Figure 4).In the Cu II complex the copper atom is accommodated in a planar square configuration with the four nitrogen atoms with about 2.0 Å of distance.Moreover, the N4 pseudo-ring appears to be more open by ~10% in the case of Cu I than in the Cu II one.These coordination differences have a direct consequence on the computed spectroscopic properties of the complexes, in agreement with the observed experimental differences between the Raman spectra of Cu I /Cu II complexes (see Figure 3).Quenching of the pyridine breathing vibration at 1247 cm −1 for the Cu I (oxabpy) complex proves the alteration of the metal-ligand bonding with respect to Cu II (oxabpy).Therefore, the change in bond lengths as suggested by DFT calculations is confirmed.
To study the stability of Cu I /Cu II (oxabpy) complexes and the possibility of an additional coordination of a TFSI counterion, we first considered the binding between Cu I /Cu II species and the (oxabpy) ligand.Total binding energies have been computed as total energy differences between the complexes and the isolated species (including zero-point vibrational energies).For the Cu I /Cu II (oxabpy) complexes the computed values are −3.27 and −8.20 eV for Cu I and Cu II , respectively.The negative values mean that the formation of these complexes is favorable.When adding a TFSI anion to these systems, we considered the interaction between its N atom and the copper, forming a pseudo-square pyramid coordination with the Cu I/II (see Figure S3-S4).In these new complexes the computed binding energies become −3.54 and −8.77 eV for Cu I and Cu II , respectively.The TFSI alone provides an extra stabilization of −0.27 and −0.57eV for Cu I and Cu II complexes, respectively.
From these data we obtain that the Cu II interacts more strongly with the ligand, probably due to the more efficient coordination of all the four N atoms of the oxabpy ligand.The favorable interaction with TFSI is mostly due to classic electrostatics, as a matter of facts the TFSI does not alter the nature and the shape of the frontier orbitals (as shown for the HOMO in Figure S4).These results highlight the importance of considering also the possibility of other co-ligands for the Cu I/II (oxabpy) complexes, paving the route to further studies in this direction.

Photovoltaic Performance in Dye-Sensitized Solar Cells
Photovoltaic data of Y123-sensitized solar cells fabricated with the Cu(oxabpy) and Cu(tmby) 2 redox couples are shown in Figure 5a and listed in Table 2. Electrolyte solution for the solar cells (as well as for ensuing studies in this article) was comprised of 0.2 M Cu I and 0.04 M Cu II species of the respective complex.The reader is referred to Figures S6-S9 for measurement statistics, solar cell characterization at lower light intensity and studies concerning electrolyte compositions.
complexes and the isolated species (including zero-point vibrational energies).For the Cu I /Cu II (oxabpy) complexes the computed values are −3.27 and −8.20 eV for Cu I and Cu II , respectively.The negative values mean that the formation of these complexes is favorable.When adding a TFSI anion to these systems, we considered the interaction between its N atom and the copper, forming a pseudo-square pyramid coordination with the Cu I/II (see Figure S3-S4).In these new complexes the computed binding energies become −3.54 and −8.77 eV for Cu I and Cu II , respectively.The TFSI alone provides an extra stabilization of −0.27 and −0.57eV for Cu I and Cu II complexes, respectively.
From these data we obtain that the Cu II interacts more strongly with the ligand, probably due to the more efficient coordination of all the four N atoms of the oxabpy ligand.The favorable interaction with TFSI is mostly due to classic electrostatics, as a matter of facts the TFSI does not alter the nature and the shape of the frontier orbitals (as shown for the HOMO in Figure S4).These results highlight the importance of considering also the possibility of other co-ligands for the Cu I/II (oxabpy) complexes, paving the route to further studies in this direction.

Photovoltaic Performance in Dye-Sensitized Solar Cells
Photovoltaic data of Y123-sensitized solar cells fabricated with the Cu(oxabpy) and Cu(tmby)2 redox couples are shown in Figure 5a and listed in Table 2. Electrolyte solution for the solar cells (as well as for ensuing studies in this article) was comprised of 0.2 M Cu I and 0.04 M Cu II species of the respective complex.The reader is referred to Figures S6-S9 for measurement statistics, solar cell characterization at lower light intensity and studies concerning electrolyte compositions.The open-circuit voltage (Voc) of a dye-sensitized solar cell is given by the difference in energy  The open-circuit voltage (V oc ) of a dye-sensitized solar cell is given by the difference in energy between the Fermi level (E F ) in the mesoporous TiO 2 and the effective redox potential of the electrolyte [11].Solar cells with the Cu(tmby) 2 redox couple show a photovoltage of 1040 mV, while those with Cu(oxabpy) exhibit a maximum photovoltage of 920 mV.Intriguingly, the Cu(oxabpy) redox couple (0.66 V vs. NHE) shows a 210 mV lower formal redox potential compared to Cu(tmby) 2 (0.87 V vs. NHE) [20], while solar cells generate a photovoltage which is only 120 mV lower.Slow recombination kinetics of excited electrons with Cu II (oxabpy) lead to a higher Fermi level in the TiO 2 and thus to a high photovoltage.This effect will be further discussed in the lifetime measurements in Figure 6.
A slightly lower photocurrent of 9.75 mA•cm −2 is measured for the Cu(oxabpy) cells with respect to 10.5 mA•cm −2 for the Cu(tmby) 2 system.At 10% light intensity, similar photocurrents are recorded with both electrolytes.Figure 5b shows IPCE spectra for abovementioned Y123-sensitized solar cells recorded at 10% illumination intensity.IPCE spectra of DSCs with Cu(tmby) 2 and Cu(oxabpy) electrolytes match well across the entire spectral range, yielding a similar integrated photocurrent for both electrolytes.Therefore, the lower photocurrent at full illumination for Cu(oxabpy) cells has to be attributed to mass transport deficiencies through the porous TiO 2 network that originate from the high viscosity of the Cu(oxabpy) redox electrolyte.Investigations at 10% light intensity (Figures S8 and S9) highlight that photocurrents with the Cu(oxabpy) redox couple can be enhanced up to 1.58 mA/cm 2 , beyond the 1.45 mA/cm 2 obtained for Cu(tmby) 2 , upon further addition of Cu II (oxabpy) up to a 1:1 mixture of Cu I/II .This indicates that the presence of the Cu II species facilitates charge transport in the Cu(oxabpy) redox electrolyte.The increased presence of recombination sites in Cu II (oxabpy), however, leads to a decrease in photovoltage.

Electron Lifetime Measurements
Recombination kinetics between excited electrons in the TiO 2 and the redox electrolyte were further investigated with electron lifetime measurements (Figure 6).redox couple (0.66 V vs. NHE) shows a 210 mV lower formal redox potential compared to Cu(tmby)2 (0.87 V vs. NHE) [20], while solar cells generate a photovoltage which is only 120 mV lower.Slow recombination kinetics of excited electrons with Cu II (oxabpy) lead to a higher Fermi level in the TiO2 and thus to a high photovoltage.This effect will be further discussed in the lifetime measurements in Figure 6.
A slightly lower photocurrent of 9.75 mA•cm −2 is measured for the Cu(oxabpy) cells with respect to 10.5 mA•cm −2 for the Cu(tmby)2 system.At 10% light intensity, similar photocurrents are recorded with both electrolytes.Figure 5b shows IPCE spectra for abovementioned Y123-sensitized solar cells recorded at 10% illumination intensity.IPCE spectra of DSCs with Cu(tmby)2 and Cu(oxabpy) electrolytes match well across the entire spectral range, yielding a similar integrated photocurrent for both electrolytes.Therefore, the lower photocurrent at full illumination for Cu(oxabpy) cells has to be attributed to mass transport deficiencies through the porous TiO2 network that originate from the high viscosity of the Cu(oxabpy) redox electrolyte.Investigations at 10% light intensity (Figures S8  and S9) highlight that photocurrents with the Cu(oxabpy) redox couple can be enhanced up to 1.58 mA/cm 2 , beyond the 1.45 mA/cm 2 obtained for Cu(tmby)2, upon further addition of Cu II (oxabpy) up to a 1:1 mixture of Cu I/II .This indicates that the presence of the Cu II species facilitates charge transport in the Cu(oxabpy) redox electrolyte.The increased presence of recombination sites in Cu II (oxabpy), however, leads to a decrease in photovoltage.

Electron Lifetime Measurements
Recombination kinetics between excited electrons in the TiO2 and the redox electrolyte were further investigated with electron lifetime measurements (Figure 6).To demonstrate the influence of the Cu II content in the electrolyte on recombination kinetics, measurements are presented for electrolyte compositions from the standard 5:1 electrolyte (orange) stepwise to a 1:1 mixture of Cu I/II (blue).Cu I concentration was kept constant in all compositions, the respective entry indicates the relative concentrations of Cu I and Cu II .The observation of a Fermi level shifts up to 70 mV for the Cu(oxabpy) standard electrolyte compared to the Cu(tmby) 2 reference supports the hypothesis of slow recombination kinetics for the Cu(oxabpy) redox couple.The shift in the Fermi level follows a consistent trend towards lower potentials upon increase of Cu II concentration in the electrolyte, as a larger number of recombination sites becomes available for excited electrons.
A number of processes are to be taken into account for recombination in DSCs [11].Ultimately, electron lifetimes in the Cu(oxabpy) electrolyte (foremost for electrolytes with little Cu II ) exceed electron lifetimes for solar cells with the Cu(tmby) 2 redox couple.In agreement with aforementioned observations, lifetimes of excited electrons in electrolytes with higher Cu II concentrations decrease due to additional recombination sites in the electrolyte.Therefore, an increased Cu II concentration leads, while it does facilitate charge transport (see Figure S10), to a lower open-circuit voltage.

Dye Regeneration
Photoinduced absorption spectroscopy (PIA) was employed to assess dye regeneration by the Cu(oxabpy) redox system.Figure 7a shows PIA spectra of Y123-sensitized photoanodes in inert and in Cu(oxabpy) electrolyte.electron lifetimes for solar cells with the Cu(tmby)2 redox couple.In agreement with aforementioned observations, lifetimes of excited electrons in electrolytes with higher Cu II concentrations decrease due to additional recombination sites in the electrolyte.Therefore, an increased Cu II concentration leads, while it does facilitate charge transport (see Figure S10), to a lower open-circuit voltage.

Dye Regeneration
Photoinduced absorption spectroscopy (PIA) was employed to assess dye regeneration by the Cu(oxabpy) redox system.Figure 7a shows PIA spectra of Y123-sensitized photoanodes in inert and in Cu(oxabpy) electrolyte.Scans were performed with a constant Cu I concentration of 0.2 M in the electrolyte and increasing Cu II content according to the color code up to a 1:1 mixture (blue).In the case of inert electrolyte (0.1 M LiTFSI, 0.6 M tBP), the Y123-dye shows the characteristic Stark-shifted ground state bleach (500-600 nm) and the absorption of the oxidized dye (680 nm, 800 nm) [39].For all constitutions of the Cu(oxabpy) electrolyte, the absorption of the dye in the oxidized state is largely suppressed.Therefore, it can be deduced that electrons are rapidly transferred from the redox electrolyte to the oxidized dye molecules.A bleach around 550-600 nm persists and can be attributed to local electric fields.Minor remaining absorption at near-infrared wavelengths are associated with electrons in the TiO2 [40].Quenching of the oxidized dye decreases with increasing Cu II content in the electrolyte due to a reduced presence of the Cu I species in the immediate proximity of the oxidized dye.This effect was further studied with transient absorption spectroscopy (TAS, Figure 7b).Results are summarized in Table 3.  Scans were performed with a constant Cu I concentration of 0.2 M in the electrolyte and increasing Cu II content according to the color code up to a 1:1 mixture (blue).In the case of inert electrolyte (0.1 M LiTFSI, 0.6 M tBP), the Y123-dye shows the characteristic Stark-shifted ground state bleach (500-600 nm) and the absorption of the oxidized dye (680 nm, 800 nm) [39].For all constitutions of the Cu(oxabpy) electrolyte, the absorption of the dye in the oxidized state is largely suppressed.Therefore, it can be deduced that electrons are rapidly transferred from the redox electrolyte to the oxidized dye molecules.A bleach around 550-600 nm persists and can be attributed to local electric fields.Minor remaining absorption at near-infrared wavelengths are associated with electrons in the TiO 2 [40].Quenching of the oxidized dye decreases with increasing Cu II content in the electrolyte due to a reduced presence of the Cu I species in the immediate proximity of the oxidized dye.This effect was further studied with transient absorption spectroscopy (TAS, Figure 7b).Results are summarized in Table 3.In inert electrolyte (black), excited electrons in the TiO 2 recombine with oxidized dye molecules with a half-time (τ 1/2,inert ) of 104 µs.In the case of Cu(oxabpy) electrolyte, a significantly faster absorption decay (τ 1/2,reg ) is observed for all compositions.Correspondingly, dye regeneration efficiency φ reg can be calculated.
The fastest regeneration of 6.19 µs is observed for the electrolyte with the Cu I species only, corresponding to a regeneration efficiency above 94%.A coherent decrease in dye regeneration is subsequently observed upon addition of Cu II to the electrolyte as it hinders Cu I species from transferring charges to the dye molecules.The high regeneration efficiencies of the Y123 dye by the Cu(oxabpy) redox system prove efficient charge compensation in DSCs.

Charge Transport in the Cu(oxabpy) Redox Electrolyte
The charge transport in the Cu(oxabpy) electrolyte was investigated with symmetrical cells with poly(3,4-ethylenedioxythiophene) (PEDOT) contacts on either electrode.Figure 8a shows cyclic voltammograms of these symmetrical cells with the Cu(tmby) 2 and Cu(oxabpy) redox couples.Results of these investigations are summarized in Table 4.
Inorganics 2018, 6, x FOR PEER REVIEW 9 of 17 In inert electrolyte (black), excited electrons in the TiO2 recombine with oxidized dye molecules with a half-time (τ1/2,inert) of 104 μs.In the case of Cu(oxabpy) electrolyte, a significantly faster absorption decay (τ1/2,reg) is observed for all compositions.Correspondingly, dye regeneration efficiency Фreg can be calculated.
The fastest regeneration of 6.19 μs is observed for the electrolyte with the Cu I species only, corresponding to a regeneration efficiency above 94%.A coherent decrease in dye regeneration is subsequently observed upon addition of Cu II to the electrolyte as it hinders Cu I species from transferring charges to the dye molecules.The high regeneration efficiencies of the Y123 dye by the Cu(oxabpy) redox system prove efficient charge compensation in DSCs.

Charge Transport in the Cu(oxabpy) Redox Electrolyte
The charge transport in the Cu(oxabpy) electrolyte was investigated with symmetrical cells with poly(3,4-ethylenedioxythiophene) (PEDOT) contacts on either electrode.Figure 8a shows cyclic voltammograms of these symmetrical cells with the Cu(tmby)2 and Cu(oxabpy) redox couples.Results of these investigations are summarized in Table 4.The shape of the cyclic voltammograms at low potentials follows that of a characteristic ohmic resistance.From potentials around 1 V, however, mass transport deficiencies impede the charge transport.In this diffusion-limited region at high potentials, cells with the Cu(oxabpy) redox electrolyte exhibit a 1.7 times higher current at similar potential compared to the Cu(tmby)2 reference cells.The limiting current can be directly associated with the diffusion of the copper complexes in the electrolyte, The shape of the cyclic voltammograms at low potentials follows that of a characteristic ohmic resistance.From potentials around 1 V, however, mass transport deficiencies impede the charge transport.In this diffusion-limited region at high potentials, cells with the Cu(oxabpy) redox electrolyte exhibit a 1.7 times higher current at similar potential compared to the Cu(tmby) 2 reference cells.The limiting current can be directly associated with the diffusion of the copper complexes in the electrolyte, where n is the number of exchanged electrons (n = 1 for common copper-based redox systems), F represents the Faraday constant, c and D are the concentration and the diffusion coefficient of the limiting species (i.e., the minority species, in these cases the Cu II ), and δ is the spacing between the electrodes.The diffusion coefficients for Cu II (tmby) 2 and Cu II (oxabpy) obtained from the cyclic voltammograms are calculated in Table 4.The diffusion coefficient for Cu(tmby) 2 is estimated around 8.6 cm 2 /s, slightly lower but within reasonable agreement with previously reported values from similar experimental setups [41,42].The higher limiting current for the Cu(oxabpy) system yields a correspondingly 1.7 times higher diffusion coefficient of 14.3 cm 2 /s.A faster transport in the Cu(oxabpy) gel-like electrolyte appears intriguing when related to the fact that its viscosity was estimated more than 2.5 times higher than that of Cu(tmby) 2 .The enhanced charge transport properties are hence to be attributed to superposition of diffusion with an additional transport mechanism such as charge hopping.The square-planar coordination geometry of Cu(oxabpy) complexes appears to enable direct electronic interactions between the complexes, leading to charge-hopping transitions along Cu(oxabpy) complexes and thus a more effective charge transport in the electrolyte.Such charge transport based on intermolecular transfers had previously been successfully demonstrated in copper complexes as solid hole transport materials [25,43].
The advanced charge transport in the Cu(oxabpy) redox electrolyte was also observed in electrochemical impedance spectroscopy on symmetrical PEDOT-PEDOT cells (Figure 8b).Experimental results (circles/stars) were fitted to the alternative circuit shown in the inlet.
A constant phase element (CPE) was proposed for the impedance of the electrochemical double layer at the PEDOT/electrolyte interface, where ω represents the frequency, and Q and β represent the parameters of the CPE.The constraint of 0 ≤ β ≤ 1 was applied for the fitting of experimental data.A short-circuit Warburg model was applied for the diffusion impedance parallel to the CPE, where ω is the frequency, R W the diffusion resistance and T W the diffusion time constant, which are defined as with the absolute temperature T, the gas constant R, cell spacing δ, number of charges z, Faraday constant F, area A and concentration c, as well as diffusion coefficient D for the transport-limiting species (here the Cu II ) [44].Resistances were inserted at the beginning of the circuit as well as before the diffusion element.
Fitted parameters are summarized in Table 4. Electrochemical impedance spectra of symmetrical PEDOT-PEDOT cells generally exhibit two semicircles, originating from the charge transfer resistance at the electrolyte/PEDOT interface and from the diffusion impedance.The charge transfer resistance dominates the impedance response for the Cu(oxabpy) system while this contribution becomes less significant for the Cu(tmby) 2 electrolyte, where a more distinct diffusion impedance arises.The high charge transfer resistance in the case of the Cu(oxabpy) redox system corresponds to a slower charge transfer rate at the electrolyte/PEDOT interface.This can be directly related to the lower fill factor in the solar cell devices.
Diffusion coefficients in the redox electrolytes were deduced from the fitted diffusion resistances R W according to Equation (5).The estimated diffusion coefficient for the diffusion-limiting Cu II (tmby) 2 of 22.4 cm 2 /s is found in excellent agreement with previously reported values and supports the proposed equivalent circuit model [20,41].A similar circuit model was applied to fit the impedance spectrum of Cu(oxabpy).The agreement between experimental data and the equivalent circuit fit indicates that diffusive mass transport is the determining charge transport mechanism.The obtained diffusion coefficient for Cu II (oxabpy) of 63.7 cm 2 /s exceeds the diffusion coefficient of Cu II (tmby) 2 by a factor of 2.8.With respect to the higher viscosity of the Cu(oxabpy) electrolyte, this result supports the above-proposed superposition of diffusive mass transport with a hole-hopping mechanism.
Both cyclic voltammetry and electrochemical impedance spectra indicate a significantly enhanced charge transport in the Cu(oxabpy) electrolyte with respect to Cu(tmby) 2 .Nonetheless, higher photocurrents were, at full illumination, measured with the Cu(tmby) 2 redox electrolyte.This indicates that the mesoporous TiO 2 hinders charge transport in the Cu(oxabpy) system.

UV/VIS Spectroscopy
UV/VIS spectra were measured using a HR-2000 spectrophotometer (Ocean Optics, Largo, FL, USA) with baseline correction from acetonitrile solutions of the complexes.

Viscosity Measurements
Viscosities were studied with a capillary and ViscoClock (Schott, Mainz, Germany).The test solutions for both Cu I/II (tmby) 2 as well as Cu I/II (oxabpy) were prepared with 10 mg/mL in chloroform.The complexes were mixed in a 5:1 ratio such as in the electrolyte (Cu(tmby) 2 : 7.85 mg Cu I , 2.14 mg Cu II ; Cu(oxabpy): 7.82 mg Cu I , 2.17 mg Cu II ).LiTFSI as well as tBP were excluded from the measurements.

Raman Spectroscopy
Raman spectra were collected using an InVia spectrometer (Renishaw, New Mills, UK) in confocal mode with a 50× objective, a frequency doubled Nd:YAG laser operating at 532 nm, and a Rayleigh line filter cutting 80 cm −1 into the Stokes part of the spectra.A 2400 lines/mm grating was used and the 520.5 cm −1 line from Si was used as a calibration giving a resolution of 1 cm −1 .The Raman spectra were recorded at different spots to confirm the homogeneity of the material.Similar results were obtained upon repetition of the measurements with varying laser intensities.
Dye solution was prepared with 0.1 mM Y123 (Dyenamo, Stockholm, Sweden) and 0.2 mM chenodeoxycholic acid in an acetonitrile 1:1 tert-butanol mixture.TiO 2 substrates were immersed in dye solution for 16 h after the aforementioned heating step.
Cells were assembled with a 25 µm Surlyn frame.Electrolyte solution was vacuum-injected through a pre-drilled hole in the counter electrode.After injection, the cells were sealed with a thermoplastic Surlyn cover and a glass coverslip.

Solar Cell Characterization
Current-Voltage measurements were carried out under AM 1.5G illumination using a 91160 solar simulator (Newport, Irvine, CA, USA).A 2400 voltage source (Keithley, Cleveland, OH, USA) was used to assess solar cell performance.After calibration with a certified silicon reference cell (Frauenhofer ISE, München, Germany), a mask was employed to confine active solar cell area to 0.16 cm 2 .Neutral density filters were placed between the light source and each cell to perform low light intensity measurements.

Electrochemistry
Electrochemical characterization was carried out with an XR potentiostat (Ivium, Eindhoven, The Netherlands).Cyclic voltammetry was recorded using a three-electrode setup employing a 2 mm disk gold electrode and a porous graphite rod as working and counter electrode, respectively.Potentials were measured against a Ag/AgCl (1 M KCl, H 2 O) reference electrode.Voltammograms were recorded from a 5 mM acetonitrile solution of Cu I (oxabpy) with 0.1 M tetrabutylammonium hexafluorophosphate supporting electrolyte at a scan speed of 20 mV/s.All potentials were subsequently calibrated using the ferrocene/ferrocenium redox couple, measured in the same conditions.
For assessment of the charge transport in the electrolyte, symmetrical PEDOT-PEDOT cells were assembled.Therefore, the abovementioned electrochemical polymerization of 3,4-ethylene dioxythiophene was employed to deposit contacts on both anode and cathode.The cells were assembled, injected and sealed as described for the photovoltaic cells above.
Limiting currents in the electrolyte systems were determined through cyclic voltammetry of these symmetrical PEDOT-PEDOT cells.Voltammograms were recorded at a scan speed of 25 mV/s.

Figure 1 .
Figure 1.(a) Schematic diagram of charge transport processes in a Dye-sensitized Solar Cell (DSC).(b) Chemical structures of the Cu(oxabpy) complex (left) and the Y123 dye (right).One or two bis(trifluoromethane)sulfonimide (TFSI) ions (not shown) serve as counter-ions for the Cu I -and Cu IIbased complexes, respectively.

Figure 1 .
Figure 1.(a) Schematic diagram of charge transport processes in a Dye-sensitized Solar Cell (DSC).(b) Chemical structures of the Cu(oxabpy) complex (left) and the Y123 dye (right).One or two bis(trifluoromethane)sulfonimide (TFSI) ions (not shown) serve as counter-ions for the Cu I -and Cu II -based complexes, respectively.

Figure 3 .
Figure 3. Raman spectra of Cu(oxabpy)TFSI, Cu(oxabpy)TFSI2 and MgTFSI2.For clarity, the Cu I and Cu II spectra were offset by 8 × 10 3 and 11 × 10 3 counts, respectively.Black boxes highlight Raman modes of the TFSI counterion, red dashed boxes mark modes of the complex.

Figure 6 .
Figure 6.Electron lifetimes in Y123-sensitized TiO2 in Cu(tmby)2 (black) and Cu(oxabpy) electrolyte.The color code indicates the ratio of Cu II in the Cu(oxabpy) electrolyte (orange = standard electrolyte 0.2 M Cu I , 0.04 M Cu II ; blue = 1:1 mixture).To demonstrate the influence of the Cu II content in the electrolyte on recombination kinetics, measurements are presented for electrolyte compositions from the standard 5:1 electrolyte (orange) stepwise to a 1:1 mixture of Cu I/II (blue).Cu I concentration was kept constant in all compositions, the respective entry indicates the relative concentrations of Cu I and Cu II .The observation of a Fermi level shifts up to 70 mV for the Cu(oxabpy) standard electrolyte compared to the Cu(tmby)2 reference supports the hypothesis of slow recombination kinetics for the Cu(oxabpy) redox couple.The shift in the Fermi level follows a consistent trend towards lower potentials upon increase of Cu II concentration in the electrolyte, as a larger number of recombination sites becomes available for excited electrons.

Figure 7 .
Figure 7. (a) Photoinduced absorption and (b) transient absorption spectra of Y123-sensitized photoanodes in inert (black) as well as Cu(oxabpy) electrolyte.The color code indicates the ratio of Cu II in the Cu(oxabpy) electrolyte (orange = standard electrolyte 0.2 M Cu I , 0.04 M Cu II ; blue = 1:1 mixture).

Figure 7 .
Figure 7. (a) Photoinduced absorption and (b) transient absorption spectra of Y123-sensitized photoanodes in inert (black) as well as Cu(oxabpy) electrolyte.The color code indicates the ratio of Cu II in the Cu(oxabpy) electrolyte (orange = standard electrolyte 0.2 M Cu I , 0.04 M Cu II ; blue = 1:1 mixture).

Table 1 .
Optical and Electrochemical Characteristics of Cu I/II (oxabpy) Complexes.

Table 1 .
Optical and Electrochemical Characteristics of Cu I/II (oxabpy) Complexes.

Table 2 .
J-V characteristics for Y123-sensitized solar cells with Cu(tmby)2 and Cu(oxabpy) electrolyte under full AM 1.5G and under 10% light intensity.For detailed characterization, the reader is referred to Figures S6-S9 in the Supporting Information.Electrolyte Voc (mV) Jsc (mA•cm −2 ) Fill Factor PCE (%)

Table 2 .
J-V characteristics for Y123-sensitized solar cells with Cu(tmby) 2 and Cu(oxabpy) electrolyte under full AM 1.5G and under 10% light intensity.For detailed characterization, the reader is referred to Figures S6-S9 in the Supporting Information.

Table 3 .
Regeneration times and efficiencies for Y123-sensitized photoanodes in inert and Cu(oxabpy) electrolyte, by addition of Cu II into the electrolyte.

Table 3 .
Regeneration times and efficiencies for Y123-sensitized photoanodes in inert and Cu(oxabpy) electrolyte, by addition of Cu II into the electrolyte.

Table 4 .
Electrochemical characterization of symmetrical PEDOT-PEDOT cells.The first two rows list current density J L and diffusion coefficient D from the cyclovoltammograms.The equivalent circuit parameters for the electrochemical impedance spectra are: serial resistance R S , charge transfer resistance R CT , frequency-independent parameters Q and β for the constant phase element and resistance R W , as well as time constant T W , for the Warburg diffusion model.Diffusion coefficients from the electrochemical impedance spectra are shown in the last row.The diffusion coefficients were calculated from Equations (2) and (5), respectively.Previously reported values in parenthesis.