A redox-active ionic liquid manifesting charge-transfer interaction between a viologen and carbazole and its effect on the viscosity, ionic conductivity, and redox process of the viologen†

Redox-active ionic liquids (RAILs) are gaining attention as a material that can create a wide range of functions. We herein propose a charge-transfer (CT) RAIL by mixing two RAILs, specifically a carbazole-based ionic liquid ([CzC4ImC1][TFSI]) as a donor and a viologen-based ionic liquid ([C4VC7][TFSI]2) as an acceptor. We investigated the effect of CT interaction on the physicochemical properties of the CT ionic liquid (CT-IL) using the results of temperature-dependent measurements of UV-vis absorption, viscosity, and ionic conductivity as well as cyclic voltammograms. We employed the Walden analysis and the Grunberg–Nissan model to elucidate the effect of the CT interaction on the viscosity and ionic conductivity. The CT interaction reduces the viscosity by reducing the electrostatic attraction between the dicationic viologen and TFSI anion. It also reduces the ionic conductivity by the CT association of the dicationic viologen and carbazole. The electrochemically reversible responses of the viologens in [C4VC7][TFSI]2 and CT-IL are consistent with the Nernstian and the interacting two-redox site models. Notably, the transport and electrochemical properties are modulated by CT interaction, leading to unique features that are not present in individual component ILs. The inclusion of CT interaction in RAILs thus provides a powerful means to expand the scope of functionalized ionic liquids.


Introduction
Charge transfer (CT) interaction between a donor and acceptor has been of great interest because it can determine the material structures, electric conduction, and photoconduction in solid state. 1 In addition, it can give rise to unique dynamical physicochemical properties of Stoddart's supramolecules, 2,3 intermediate complex of chemical reaction, 4 and photochemically active species [5][6][7][8] in solution phase. CT interaction never emerges in its individual component species but appears as a synergetic effect. The CT interaction depends on the combination of a donor and acceptor, and one can exploit such combinations in ionic liquids to produce new photoactivity and to modulate their physicochemical properties.
A myriad of ionic liquids (ILs) with various physicochemical properties have been synthesized. Because ILs are designer solvents 9 that can be readily created by combining organic cations and anions, one can straightforwardly tune their viscosity, ionic conductivity, and melting point, as well as the solubility of substrate molecules. This leads to some notable advantages over typical organic solvents. For example, ILs are oen preferred due to the higher solubility of solutes such as cellulose 10 and CO 2 . 11,12 Therefore, they are frequently applied to extraction and separation chemistry. 13 ILs also provide interesting functionalities rarely found in organic molecular solvents, 14 leading to unique IL-based liquid materials. In addition, organic cations can be modied to further transform ILs into functionalized ionic liquids (FILs) or task-specic ionic liquids. 15 Extensive developments of FILs have led to functional liquid materials that are not merely a solvent or medium. The high concentrations in FILs (around 1-2 M) lead to strong interactions between functional groups and ions; these interactions can play dominant roles in new IL-based functional materials. For example, magnetic ILs consisting of metal complexes have been reported. [16][17][18] Hisamitsu et al. reported photo-functionalized ionic liquids with photon-up-conversion characteristics, 19,20 with concentrated chromophores being a critical design in their FILs. Murray et al. reported conductive redox-active ILs (RAILs) that comprise metal complexes and viologens. [21][22][23] In addition, Saielli et al. reported the thermal properties of various viologen-based ILs and ionic liquid crystals, which are correlated with their structure. [24][25][26] These ndings demonstrate the potential of viologens to be a potent basic component for developing and creating FILs.
Several applications of RAILs in electrochemical reactions and devices have also been reported; they are used as components in supercapacitors, 27,28 redox ow batteries, 29 and electrochromic devices, [30][31][32] and as electron mediators for photochemical catalysts. 33 Additional examples have been given in recent reviews. 34 between anions of ILs and electron accepting additives in a photo-induced electron transfer reaction between the species. 38 They concluded that the ILs play an active role in such photochemical systems rather than being an inert medium. With the incorporation of CT interactions into ILs, we may create a new functionality that is not possessed by individual donors and acceptors, and thereby modulate the physicochemical properties of the materials. In this study, we aim to exploit the functionality of viologens to design FILs based on CT interaction. In this regard, CT interactions between donors and viologens have been applied for assembling molecules, 2,39 photo-induced electron transfer reactions, [5][6][7][8]40,41 alkali metal sensing, 42 and thermochromism. 43 These studies demonstrated the potential of incorporating CT interactions into viologenbased ILs to pave a new path in IL research. To the best of our knowledge, research into RAILs with externally introduced CT interaction has not been attempted.
Intermolecular and interionic interactions in ILs have been investigated to understand the relationship between the structures and dynamic transport properties such as viscosity and ionic conductivity. In these species, electrostatic, van der Waals, hydrogen bonding, dipole-dipole, and p-p interactions all play important roles in the formation of microstructures, leading to unanticipated transport features and functionalities. 44,45 Understanding these strong and weak interactions will provide new insights into the transport properties and functionalities in ILs. In this regard, it is noteworthy that the effect of CT interaction in ILs on the physicochemical and electrochemical properties is presently unclear.
In this study, we present a "RAIL with CT interaction" (CT-IL) originating from two redox centers, a carbazole and viologen; it is based on an equimolar mixture of a viologen-based IL and a carbazole-based IL. As we shall see, in the CT-IL, the CT interaction between the carbazole and viologen modulates the physicochemical properties such as viscosity and ionic conductivity, as well as redox properties including redox reversibility of viologens. Our approach provides a convenient means for creating and tuning of CT-IL, with properties that are distinct from those of the components.

Experimental
Chemical structures of the carbazole-based RAIL ([CzC 4 ImC 1 ] [TFSI]) and the viologen-based RAIL ([C 4 VC 7 ][TFSI] 2 ) in this study are shown in Scheme 1, with a photo of the combined CT-IL straightforwardly prepared from an equimolar mixture between them. Synthetic procedures of the RAILs, water content in the RAILs, and experimental setups are described in the ESI. † All measurements in this study were conducted in neat RAILs, namely without any solvent, unless otherwise mentioned.
[C 4 VC 7 ][TFSI] 2 remains in a supercooled liquid state for a few hours at room temperature aer it was heated above the melting point (52 C). 46 Although the melting points of [CzC 4 -ImC 1 ][TFSI] and CT-IL could not be determined by differential scanning calorimetry (DSC) because of the absence of any peak assignable to the melting point, their solidication was not observed at room temperature (see the DSC thermograms in   2 , and the CT-IL in the neat system ( Fig. 1(a)). The neat [CzC 4 ImC 1 ][TFSI] and [C 4 VC 7 ][TFSI] 2 have no absorption band over the visible region as well as their diluted acetonitrile solutions (Fig. S2 †). For the neat CT-IL, a new broad absorption band at 427 nm was observed. The absorption band is attributable to the absorption by the CT complex between the carbazole and viologen. 7 Specically, our absorption spectrum is in good agreement with the CT absorption spectrum reported by Yonemura et al. 5 for a diluted carbazole-tethered viologen, in terms of not only peak wavelength but also line width. Thus, the formation of additional species such as higher aggregates of the CT complex in CT-IL is unlikely. The CT absorption band of CT-IL at 427 nm monotonically decreased with increasing the temperature, holding the peak wavelength invariant. This further supports a simple equilibrium of 1 : 1 complex of carbazole and viologen, the same as in the diluted system. Taking these considerations into account, the temperature dependence can be explained by the association/dissociation equilibrium of the carbazole and viologen (Cz + V ++ # CzV ++ ).
As a rst approximation, we do not consider carbazolecarbazole interaction and viologen-viologen interaction. The total concentration (c 0 ) of the carbazoles and viologens of CT-IL here was 0.92 M in the 1 : . The concentration of the CT complex (c CT ) can be described by the thermodynamic equilibrium constant.
The molar absorption coefficient (3 CT ) of the CT complex between carbazoles and viologens was reported to be around 400 AE 70 M À1 cm À1 at 420 nm. 7 However, the absorption coef-cient is invalid in our CT-IL because if we use it as an approximated value, the concentration of the CT complex is estimated to be over 1.0 M at 298 K with the Lambert-Beer relation (Abs CT (1) and (2). The absorbances were proportional to the concentration of the diluted species. In these cases, i.e. with a large excess of C 4 VC 7 or CzC 4 ImC 1 , the absorbance of the CT complex can be described as follows.  ) in CT-IL is 0.48 M at 298 K with an assumption of 3 CT to be 880 M À1 cm À1 in the CT-IL, which is taken as an average of the absorption coefficients in Fig. 1(c) and (d). Around 50 mol% carbazole and viologen in the CT-IL achieve the CT complexation. Its thermodynamic parameters are DG ¼ À2.2 kJ mol À1 , DH ¼ À13.9 kJ mol À1 , and DS ¼ +39.4 J mol À1 K À1 (see Fig. S3 and Table S1 †). The theoretical evaluation using density functional theory (DFT) showed DG ¼ À3.6 kJ mol À1 , DH ¼ À49.8 kJ mol À1 and DS ¼ +155 J mol À1 K À1 at 298 K with a model compound of CzV ++ in an IL medium (see the ESI †). The theoretical evaluation provided reasonable DG for CT-IL, although the assumption of the isolated CT complex surrounded by the continuous IL dielectric medium was far different from the real system. [TFSI] is not clear but it is beyond our focus in this study. On the other hand, we observed no emission in CT-IL system. This can be attributed primarily to the fast uorescence quenching by photo-induced electron transfer from the excited carbazole to the viologen. 5,47 As can be seen above, neat CT-IL shows rather different optical properties from individual RAILs and the diluted CT-IL in acetonitrile.  Fig. 2(a) and (b) 2 from the viscosity. The viscosity of CT-IL is the lowest among the three RAILs, while the conductivity of CT-IL is not the highest among them. To rationalize the trends, the activation energy was evaluated using the Arrhenius equations (eqn (4)), and the approximated interpolation equations (eqn (5)) were expressed by the Vogel-Fulcher-Tammann (VFT) formalism:

Viscosity and ionic
The best t parameter-sets are listed in Table 1. The Arrhenius plot of the viscosities and ionic conductivities (Fig. S6 †) shows a little deviation from straight lines. The coefficient of determination (R 2 ) for the viscosity of CT-IL is 0.9914, and that for the ionic conductivity of [ energy (E ha ) of the viscosity of CT-IL is reduced by mixing the two RAILs. To examine the characteristics of the ionic transportation, we use the Walden analysis and the Grunberg-Nissan 48-51 analysis below.

Walden analysis.
For the investigation of transportation parameters of ILs, the Walden relation (hs ¼ constant) has oen been used, even though it originally holds in a diluted ionic solution. The Walden product, hs or hs/c, remains constant, independent of the temperature in the absence of inter-ionic interactions. Fig. 2(c) shows the temperature dependence of the Walden product of the RAILs obtained by using the formal concentration (c F ), while the full Walden plot is shown in Fig. S7. † The density of ILs and thus c F of the ions slightly decreased with increasing temperature because of volumetric expansion. However, the changes are negligibly small. 5,31 Therefore, we assumed the formal concentrations are temperature-independent constants for the Walden analysis.
In comparison, the Walden product of CT-IL increased with increasing temperature, while that of [  almost constant. 52 In CT-IL, the ionic association and the CT complex formation can be deeply involved in the transport properties. Therefore, the temperature dependent Walden product of CT-IL is attributable to both equilibria: association and dissociation of (1) ionic components (between the viologen, imidazolium, and TFSI), and (2) CT complex (between the carbazole and viologen).

Grunberg-Nissan analysis.
As an alternative method of analysis, we evaluated the viscosity of binary mixtures using the Grunberg-Nissan model. [48][49][50][51] where h 1 , h 2 , and h mix are the viscosities of compounds 1, 2, and their mixture, respectively; x 1 and x 2 are the molar fractions, and W 12 is the interaction parameter between compounds 1 and 2. If the interaction parameter W 12 is negligibly small, the viscosity of the mixture is a fraction-weighted logarithmic average, that is, it signies a non-associated liquid. When W 12 is non-zero positive/negative, the viscosity of the mixture should be greater/smaller than those of the individual compounds. The experimentally obtainable interaction parameters for viscosity, and those for ionic conductivity, in the analogy to the viscosity, can be evaluated using the following expression. Fig. 2(d) shows the temperature dependence of Dln h and Dln s À1 for CT-IL, obtained using the continuous data interpolated with the VTF formula of the viscosity and conductivity as well as the tted parameters in Table 1. The interaction parameters were negative, and the magnitudes of both viscosity and conductivity decreased monotonically with increasing the temperature. The curvatures of Dln h and Dln s À1 are different, where Dln s À1 shows more convex behavior.
In the CT-IL system, two kinds of equilibria should be considered: (1) electrostatic cation-anion pair association and dissociation and (2) formation of the CT complex. In ionic liquids, electrostatic attraction between anions and cations dominates the attractive interaction that leads to the association of cations and anions ( Fig. 3(a) and (b)). The associations can be examined in terms of ionicity by the impedance technique and nuclear magnetic resonance. 53

The decreased viscosity in CT-IL upon mixing [CzC 4 ImC 1 ][TFSI] and [C 4 VC 7 ]
[TFSI] 2 can be attributed to decreasing electrostatic attraction because of the formation of the CT complex between the carbazole and the viologen (Fig. 3(c)). Our DFT calculation supports the decrease of the interionic attraction. We employed a fragment molecular orbital (FMO) calculation to evaluate the CT interaction which contributes to the charge donation from the carbazole to the viologen. We found that the carbazole donates 0.163 electrons to the viologen resulting in decreasing its cationic charge, i.e. the formal charge can be depicted as Cz +0.163 V +1.837 . Besides, the attraction enthalpy difference   between the viologen and TFSI is 63.4 kJ mol À1 with carbazole and 65.9 kJ mol À1 without carbazole (geometries of the contact ionic pairs are shown in Fig. S17 †). Thus, the CT attraction can weaken the electrostatic attraction between the viologen and TFSI. Therefore, we may expect that decreased viscosity would lead to increasing ionic conductivity. However, the ionic conductivity of CT-IL was in fact smaller than that of [C 4 VC 7 ] [TFSI] 2 .
To explain the smaller ionic conductivity of CT-IL than [C 4 VC 7 ][TFSI] 2 , we note that the actual concentration of free ionic species is important for ionic conductivity. Formation of the CT complex would lead to decreasing concentration of the free dicationic viologens and suppressing the movement and migration of the viologens and imidazoliums linked with carbazole. Because increasing temperature leads to dissociation of both the ions and the CT complex, this results in a steeper temperature dependence on Dln s À1 than on Dln h.
The viscosity and ionic conductivity of those RAILs can be affected by the temperature-dependent microstructural and dynamic heterogeneities of the ILs. 44 Fig. 4. A reversible redox wave of the viologen was observed around À0.6 V (region A). Anodic currents originating from the oxidation of carbazoles were found around +0.7 to +1.5 V (region B). Although no anodic current was found around +0.5 V during the rst positive scan, a redox pair in that region appeared in the second and subsequent scans resulting from the redox reaction of electro-polymerized carbazole produced by the oxidation of monomeric carbazole in potential region B. 54,55 The similar responses of carbazole and poly-carbazole were likewise seen in neat [CzC 4 ImC 1 ][TFSI] (Fig. S8 †). However, the redox responses of the carbazole and the electropolymerized carbazole were complicated and the analysis was difficult. Therefore, we focused our attention on the redox response of the viologen.
3.3.2 Redox reversibility of the viologen: the Nernstian and the interacting two-redox site models. CVs of the viologen in region A in Fig. 4 produced peaked diffusion-limited waveforms due to semi-innite planar diffusion to the electrode surface. For analytical evaluation of electrochemical reversibility of the viologen, we investigated CVs of the viologens at some concentrations with slow scan rates at which the CVs show a sigmoidal waveform. Reducing the scan rate brought the CVs to a steady state without current and shape changes. Thus, the CVs can be regarded as electrochemically reversible responses. Fig. 5  , and CT-IL. Although hysteresis was found due to a contribution of the planar diffusion (see the ESI †), all samples showed sigmoidal waveforms, whose forward and reverse scan curves were superimposable by shiing the potential axis less than 19 mV. That is, the systems again showed electrochemically reversible CVs whose line shapes did not vary at lower scan rates. However, the slope of the CVs at E ¼ E 1/2 was dependent on the system. In principle, when the electrochemical reaction is described by a simple electrochemical equilibrium (Ox + e À % Red) of a redox couple and the redox species are not interacting with each other, the Nernst equation is established and can provide the concentration ratio of the redox species on the electrode surface [Ox] surf /[Red] surf . In that situation, a steady-state voltammogram can be described by the ideal Nernst response eqn (9) under the assumption of equal diffusion coefficients D Ox ¼ D Red with n app ¼ 1.
where f ¼ F/RT, i ss is the steady-state limiting current whose sign is dened so that the cathodic current goes negative, E 1/2 is the half-wave potential and n app is an apparent number of electrons representing the redox equilibrium. Here, n app can take a noninteger value to appropriately describe the redox characteristics of CVs. In fact, the diluted system of 10 mM [C 4 VC 7 ][TFSI] 2 in [BMIM][TFSI] produced a Nernstian CV with n app ¼ 1.00. However, the other systems with higher concentrations of the viologens indeed produced Nernstian CVs with greater n app than unity, i.e. the slope of the CV at E ¼ E 1/2 is steeper than that with n app ¼ 1 (see Table 2). It implies that the reduced viologens are interacting attractively. The attractive interaction brings about a situation where the reduction potential of an Ox species close to a Red species shis positively when the Ox species accepts an electron from the electrode to be the Red species. The schematic representation of the attractive reaction can be proposed as follows.
V 2+ + e À (eld) % V + c (step 1) Step 3 is the dimerization reaction of the reduced viologen as discussed below. If DG in step 3 is negatively large and the reactions from step 1 to step 3 are regarded as a sequential redox process with two redox sites, we can approximately describe the reaction as a two-consecutive one-electron transfer process. In that case, the reduction potential difference DE ¼ E 1 À E 2 , the rst and second reduction potentials E 1 and E 2 , can negatively increase with increasing the attractive interaction which means the system prefers the dimerized viologens (V + c/ V + c). The two steps of the reduction reaction eventually cannot be distinguished from each other but seen to be coalescent apparently resulting in an indistinguishable two-electron reduction reaction. Under the assumption that diffusion coef-cients of all species involved in the reaction are the same, the steady-state voltammogram can be described as follows. 56,57 iðEÞ i ss ¼ 1 2 Â expð Àf ðE À E 1 ÞÞ þ 2 expð Àf ð2E À E 1 À E 2 ÞÞ 1 þ expð Àf ðE À E 1 ÞÞ þ expð Àf ð2E À E 1 À E 2 ÞÞ where E 1 and E 2 are the macroscopic redox potentials for the two redox steps. Here we dened an interaction parameter, DE ¼ E 1 À E 2 , which reects an inter-site interaction in the redox reaction as described above. If there is no inter-site interaction in the system, i.e. redox potentials of all steps are equal (3 1 ¼ 3 2 ¼ 3 3 ¼ 3 4 in ref. 56), it turns out that DE ¼ (2 ln 2)RT/F ¼ +39.8 mV at 333 K and n app becomes unity. If the inter-site interaction is repulsive/attractive, i.e. one reduced site discourages/encourages the reduction reaction on another site, then DE > +39.8 mV for repulsive and DE < +39.8 mV for attractive interaction. The results from analyzing the CVs in Fig. 5 are listed in Table 2. The diluted solution of 10 mM [C 4 VC 7 ][TFSI] 2 exhibited a typical Nernstian CV with n app ¼ 1 and DE ¼ +39.8 mV, which indicates the absence of interaction between viologens. Other  Table 2.