Tuning the glass transition of siloxane-based poly(ionic liquid)s towards high ion conductivity

Herein, we report a simple and versatile synthetic approach towards siloxane-based poly(ionic liquid)s (PILs) with unusually low glass transition tempera-tures ( T g ) down to (cid:1) 73 (cid:3) C, and thus “ liquid-like ” behavior at room temperature. We designed


| INTRODUCTION
Poly(ionic liquid)s (PILs) are permanently charged polymers, which carry ionic liquid (IL)-like species in their repeating units. Their excellent ion conduction properties and safe handling in particular make them promising candidates for quasi-solid state electrolytes in energy devices such as capacitors and transistors. 1 PILs are commonly considered as single-ion conductors, as either the anion or the cation species is covalently restricted to the polymer backbone, while the other one, the counter ion, is responsible for charge transport. 1 Hence, conductivity is mainly determined by the mobile counter ions. [2][3][4][5] This is in contrast to their ionic liquid (IL) monomeric analogues, where both ions are able to move. Consequently, conductivity of PILs is generally at least two orders of magnitude lower than those of ILs, which currently limits their application as polymer electrolytes. 2,6 However, the enormous variety of cationanion pairs, and the vast possibilities of macromolecular design give rise to PILs with tunable properties in an unusually wide window. 1,[7][8][9][10] In fact, judicious choice of the ions and polymer backbone allows for designing "liquid-like" PILs, as recently demonstrated by us and other groups, [11][12][13][14][15] which are of significant potential in polymeric ion conductors. For instant, Jourdain and coworkers elegantly combined a siloxane-based backbone with IL-like species as side-groups and the prepared ionic polymers with a T g of À62 C and strongly enhanced conductivities up to $6.7 Â 10 À5 S/cm at 30 C. 11 Such unique properties are due to the relatively small energy barrier for rotational motion around silicon-oxygen bonds, which promotes segmental movement of the siloxane units at low temperatures. 16 It is therefore possible to create at room temperature "liquid-like" solvent-free ionic polymer materials by combination of pliable polysiloxanes with IL-like species. The scope of this methodology is also demonstrated by Liang and coworkers, who reported a siloxane-based PIL including phosphoniumtype species that exhibited a T g even below À70 C. 17 To date, several groups have addressed this promising class of siloxane-based PILs and prepared polycations with ammonium, 18 imidazolium, 19 phosphonium 17 and triazolium. 11 Some works have also focused on polysiloxane copolymers with both ionic and non-ionic segments (e.g., dimethylsiloxane units). 17,[19][20][21][22] A low concentration of ionic repeating units reduces the charge density but is generally favorable to achieve a low T g , as polydimethylsiloxanes (PDMSs) exhibit the lowest T g s among hitherto known polymer classes. 19 It is worth noting that for such copolymers also repulsive interactions between the ionic and non-ionic segments could dramatically influence the thermal behavior and conduction properties.
In this article we present a fast and efficient route towards siloxane-based PILs with "liquid-like "behavior at room temperature due to their flexible backbone and a balanced amount of ionized repeating units. We further explore the impact of the type of counter anion and sidechain length on the thermal properties and succeed in tuning its T g over a rather wide range far beyond conventional polyelectrolytes.

| Synthetic route
The most convenient strategy towards covalently linking imidazolium-based IL-like moieties to a polysiloxane backbone is via post-polymerization modification of commercially available polysiloxane derivatives. We chose the supposedly random copolymer poly[dimethylsiloxane-co-(chloropropylmethyl)siloxane] P(DMS-co-CMS) as precursor, which contains 14.8 mol% of the CMS unit, as determined by proton nuclear magnetic resonance spectroscopy ( 1 H-NMR). The chloropropyl groups of P(DMS-co-CMS) were modified by reaction with nalkylimidazoles, as schematically shown in Figure 1. Specifically, 1-methylimidazole and 1-butylimidazole were used as the reagents. The former was utilized for an initial study to underline the feasibility of this approach, and its quaternization product is hereafter termed as C 1 -PIM-Cl.
In a typical experiment, both precursors P(DMS-co-CMS) and 1-methylimidazole (20 mol eq. per Cl) were dissolved in n-butanol and the reaction mixture was heated up stepwise to its boiling point and kept under these conditions overnight (see Exp. section for details). To isolate the ionic product the reaction mixture was precipitated into diethyl ether, followed by at least two dissolution-reprecipitation steps in dichloromethane and diethyl ether, respectively. Finally, the oily product was dried at 80 C under high vacuum (10 À3 mbar) to yield an almost colorless, viscous copolymer. The viscous nature of C 1 -PIM-Cl is demonstrated in Figure S1, which shows photographs of the ionic copolymer in an upsidedown glass container.
The chemical structure of C 1 -PIM-Cl was determined by 1 H-and 13 C-NMR spectroscopy. Figure 2(A) depicts the 1 H-NMR spectra of the ionic product (top) and its two precursors, the 1-methylimidazole and the neutral P(DMS-co-CMS) copolymer. The imidazolium proton signals (Im-2 and Im-4 + 5) are found only in C 1 -PIM-Cl (the top curve in Figure 2(A)) at 10.67 and 7.72 ppm, respectively. Proton signals associated with the Cl-(CH 2 ) 3 -group are shifted from 0.62, 1.78, and 3.49 ppm in the precursor P(DMS-co-CMS) (the mid curve in Figure 2 (A); signals a, b, and c) to 0.51, 1.91, and 4.32 ppm in C 1 -PIM-Cl (the top curve in Figure 2(A); signals A, B, and C). The same occurs to the CH 3 -group on the imidazolium ring when comparing 1-methylimidazole (the bottom curve in Figure 2(A); signal d) to C 1 -PIM-Cl (the top curve in Figure 2(A); signal D). Clearly, signals in the product are at different chemical shifts than in the starting materials, which underline a quantitative conversion of all chloropropyl units into dialkylimidazolium cations. The signal around 0.06 ppm is assigned to the -Si-CH 3 protons. The 13 C-NMR spectra of C 1 -PIM-Cl and the precursor P(DMS-co-CMS) are shown in Figure 2(B).
The product spectrum (the top curve in Figure 2(B)) reveals signals associated with the imidazolium ring (Im-2 and Im4 + 5) at 138.6 and 124.1 ppm, respectively, thereby confirming its presence in the ionic copolymer. These signals are fully absent in the spectrum of the precursor P(DMS-co-CMS) (the bottom curve in Figure 2 (B)). In addition, the small shoulder peak at 0 ppm is assigned to the -Si-CH 3 species at the chain end. The chemical structures of both the precursors P(DMS-co-CMS) and 1-methylimidazole, and the product C 1 -PIM-Cl are shown in Figure 2(C).
F I G U R E 1 General synthetic approach towards a polysiloxane-based copolymer of PDMS and an imidazolium-type PIL. R is methyl or butyl substituent F I G U R E 2 1 H-(A) and 13 C-NMR (B) spectra of C 1 -PIM-cl and its precursors recorded at 400 and 101 MHz, respectively, in CD 2 Cl 2 . (C) Chemical structures with signals being labeled The formation of the imidazolium cation was also underlined by attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy, as shown in Figure 3. The combinational vibration of N-CH/N-CH 2 / N-CH 3 stretching, typical for aromatic imidazolium rings, is observed in the spectrum of the product C 1 -PIM-Cl around 1170 cm À1 (the top curve in Figure 3). [23][24][25] This signal is absent in the spectrum of the precursor P(DMSco-CMS) (the bottom curve in Figure 3). All characteristic siloxane-type bands found in the starting material are also present in the product, for example, ν˜(Si-CH 3 ) ≈ 1256, 856, and 792 cm À1 , as well as ν˜(Si-O-Si) ≈ 1057 and 1008 cm À1 . 26 It is worth noting that the spectrum is dominated by Si-(CH 3 ) x bands, and therefore modes corresponding to C-Cl groups are not detectable. Overall, these data confirmed the formation of imidazolium rings and the high purity of the product.
The molar content of the quaternized repeat unit in C 1 -PIM-Cl was calculated to be 24 mol% from the 1 H-NMR spectrum (cf. ESI, Figure S2). To our surprise, this value is higher than the 14.8 mol% of the (chloropropyl) methylsiloxane repeat unit in the neutral P(DMS-co-CMS), determined by 1 H-NMR spectroscopy (cf. ESI, Figure S2). Such difference is presumably caused by the purification procedure, during which the polar reaction mixture was precipitated into diethyl ether. Only the oily precipitate (the imidazolium-rich polymer fraction) was collected, while the soluble or dispersible imidazoliumpoor polymer fraction was removed. Note that the values for the molar content of quaternized repeat units given above are the average values obtained from 1 H-NMR spectroscopy, that is, the samples of product and precursor may comprise a mixture of polymer chains with different imidazolium contents.
The polymeric nature of C 1 -PIM-Cl was confirmed by size-exclusion chromatography (SEC) in aqueous medium. The apparent average molecular weight is M n = 25.6 kDa (M w = 78.5 kDa) with a dispersity of 3.07 ( Figure S3). The reference SEC measurement of the precursor P(DMS-co-CMS) performed in THF as eluent showed a M n = 11.7 kDa (M w = 22.8 kDa). These values are not to be compared directly due to different calibration standards in different SEC systems, but prove the polymeric nature of our samples and the polydispersity index of the molecular weight of these polymers.
Differential scanning calorimetry (DSC) measurements were performed in order to characterize the thermal properties of C 1 -PIM-Cl, including its T g . All measurements were conducted in a temperature range of À160 to 100 C under nitrogen flow. Generally speaking, the shape of DSC curves and consequently the thermal characteristics are affected by the heating/cooling rates. 27,28 We used a heating rate of 20 C/min for all DSC measurements as it provided the most distinctive T g ( Figure S4). Figure 4 shows the DSC traces of C 1 -PIM-Cl measured under these conditions (orange). For reasons of comparison, the starting material P(DMS-co-CMS) is depicted as well (gray). In addition, Figure 4 shows the cooling traces prior to heating. C 1 -PIM-Cl exhibits a T g of À49 C, which we calculated as local maximum of the first order derivative of heat flow. This value already represents an unusually low T g for a Cl À -containing PIL copolymer, as chloride ions are known to associate strongly with imidazolium cations due to their point charge character. Typically, this leads to high T g s, as for instance in imidazolium-type polystyrene 29 (T g = 180 C).
F I G U R E 3 ATR-FTIR spectra of C 1 -PIM-cl and P(DMS-co-CMS). Transmittance was recalculated on basis of reflection. Modes characteristic for imidazolium and polysiloxane are labeled F I G U R E 4 DSC heating (20 C/min) and cooling traces (10 C/ min) for C 1 -PIM-cl and the precursor P(DMS-co-CMS). T g was calculated as local maximum of first order derivative of heat flow The fact that our product C 1 -PIM-Cl shows a significantly lower T g is rooted in the flexible nature of the polysiloxane backbone. 11,17 Surprisingly, the DSC trace of C 1 -PIM-Cl reveals a second, less pronounced phase transition at À110 C ( Figure S5), which is characteristic for polysiloxanes, 16 as observed in the precursor P(DMS-co-CMS). We assume that this weak T g is caused by a PDMS-rich microphase in the product C 1 -PIM-Cl, especially since dimethylsiloxane repeat units account for 76 mol% within the copolymer. Microphase separation is a phenomenon generally found in block copolymers carrying incompatible blocks. 6,30 Depending on the composition, such materials might exhibit individual glass transitions for each block. 31,32 Though not common, random copolymers are also capable of microphase separation, as theoretically predicted 33,34 and experimentally underlined. [35][36][37][38][39] It is important to note that, although we refer to our siloxane-based precursor as a supposedly randomly distributed copolymer, we cannot rule out compositional gradients within this commercial system, which might have a dramatic effect on its phase properties.
Based on the results presented so far, we have shown that the combination of a highly flexible PDMS backbone with pendant imidazolium chloride ion pairs presents a simple strategy towards low-T g ionic copolymers. It is worth noting that, among many other factors, a low T g can lead to very high ionic conductivities in PILs. 11,40 In the following section we describe the optimization of the chemical structure of our PIL-based copolymers towards further lowering down its T g and ultimately promoting the ionic conduction behavior.

| Optimization of the chemical structure
In general, there are three strategies to alter the chemical structure of our copolymer C 1 -PIM-Cl, which can be exploited towards achieving a lower T g : (a) Exchanging the Cl À with a bulkier anion; (b) Introducing long spacers between the backbone and the cation to increase the mobility of the ionic species; (c) Attaching alkyl chains of optimized length to imidazolium and thereby plasticizing the polymer matrix. All these approaches are well known to dramatically affect the properties of PILs, in particular the glass transition. [40][41][42][43] We carried out a new set of experiments with the same precursor, and followed the above-described approaches (a) and (c) in order to find the lowest possible T g .
First, we exchanged chloride ions in C 1 -PIM-Cl with TFSI À , PF 6 À , and BF 4 À , respectively, through anion metathesis. Thereby, we were able to determine the anion with the strongest impact on T g . Second, we replaced the methyl group attached to the imidazolium ring with a longer alkyl chain, specifically with n-butyl, and then exchanged Cl À by the anion-of-choice (as determined through our first set of experiments). This allowed us to approach the lowest possible T g .

| Anion metathesis reactions
Anion metathesis reactions were performed with AgX salts (with X denoting TFSI À , PF 6 À , or BF 4 À ) to assist the anion exchange. 41 In a typical experiment, 1 eq. of C 1 -PIM-Cl was dissolved in ethanol and, while stirring, a slight excess of dissolved silver salt was added (see Exp. section for details). The anion-exchanged products C 1 -PIM-X (with X denoting TFSI À , PF 6 À , or BF 4 À ) remain soluble, despite being more hydrophobic than the chloride containing C 1 -PIM-Cl. After removal of silver chloride, the samples were isolated and purified. The absence of chloride in the anion-exchanged products was confirmed by titrating C 1 -PIM-X against silver nitrate. The fact that no silver chloride precipitate was found allowed us to conclude that the previous anion metatheses had been completed. All anion-exchanged products visibly showed a lower viscosity than their precursor C 1 -PIM-Cl at room temperature. Compared to C 1 -PIM-Cl, which is soluble in water (see Table S1), the anionexchanged products are not, given their increasingly more hydrophobic anions. Independent of the types of anions used, both 1 H-and 13 C-NMR spectra ( Figure S6) remained similar to that of C 1 -PIM-Cl. No additional signals compared to C 1 -PIM-Cl were observed in the 1 H-NMR spectra of C 1 -PIM-TFSI, C 1 -PIM-PF 6 , and C 1 -PIM-BF 4 , respectively, which was expected since TFSI À , PF 6 À , and BF 4 À are free of hydrogen atoms. Yet, new peaks associated with TFSI À (CF 3 -group) appeared in the 13 C-NMR spectrum of C 1 -PIM-TFSI. ATR-FTIR spectroscopy confirmed the successful anion exchange by the presence of stretching modes characteristic for TFSI À , for example, ν˜a s (SO 2 ) ≈ 1349 cm À1 , ν˜a s (CF 3 ) ≈ 1186 cm À1 , and ν˜s (SO 2 ) ≈ 1135 cm À1 , 44,45 as well as for PF 6 À , for example, ν˜s(P-F) ≈ 820 cm À1 , 46 see Figure S6.
We then performed DSC measurements to study the impact of anion exchange on T g . Figure 5(A) depicts the heating curves for all C 1 -PIM-X polymers, plus the curve of the precursor P(DMS-co-CMS). Clearly, the choice of anion strongly affects the glass transition. For example, C 1 -PIM-TFSI exhibits a T g of À67 C. Surprisingly, both C 1 -PIM-BF 4 and C 1 -PIM-PF 6 reveal significantly higher T g s of À47 and À32 C, respectively. These values are even higher than the one for C 1 -PIM-Cl. This is somewhat counter-intuitive, considering the fact that both PF 6 À -and BF 4 À -ions have lower coordination tendencies towards the imidazolium cation than chloride. 47 To our surprise, the DSC trace of C 1 -PIM-TFSI revealed only one phase transition. It seems that the presence of the more hydrophobic TFSI À anion improves the compatibility between the ion-rich and the ion-poor domains within the copolymer. This is in stark contrast to C 1 -PIM-Cl, which showed an additional T g at À110 C. Also, both C 1 -PIM-BF 4 and C 1 -PIM-PF 6 revealed a clear second phase transition in the same region. As discussed earlier, we attributed this phenomenon to the PDMS-rich microdomains, which we hypothesize to be phaseseparated from the ion cluster domains in the copolymer. Overall, the introduction of bulky, weakly coordinating anions (TFSI À , PF 6 À , and BF 4 À ) provides a simple means to tune the properties of PILs, in particular their thermal properties. We found the strongest effect for TFSI À . Next, we focused on replacing the methyl group attached to the imidazolium cation by a longer alkyl chain.

| Tuning the length
Tuning the length of the alkyl chain attached to the imidazolium cation, and thereby replacing the methyl group for a longer alkyl substituent, is a simple way to modify T g of our copolymer product. On the one hand, a longer alkyl chain will increase the size of the cation and eventually weaken its interaction with the anion; on the other hand, it can also amplify the hydrophobic interactions, and very long alkyl chains could even crystallize at room temperature. We followed a similar synthetic route as earlier, but utilized n-butylimidazole as a starting compound. Thereby, we prepared a new ionic copolymer with a longer alkyl chain and were able to study its impact on T g . As expected, the use of n-butylimidazole engendered an increase in reaction time to reach a full conversion and the resulting ionic copolymer became increasingly more hydrophobic. The formation of imidazolium rings was confirmed by 1 H-and 13 C-NMR spectroscopy ( Figures S8 and S9). The molar ratio of quaternized repeat unit was calculated as 21% and thus was similar to the one in C 1 -PIM-Cl (24%). The successful formation of imidazolium chloride ion pairs was underlined by ATR-FTIR-spectroscopy. The characteristic modes related to both imidazolium and PDMS, as similarly discussed earlier, are present in C 4 -PIM-Cl. The resulting ionic copolymer exhibited increased viscosity, which points out stronger hydrophobic interactions. The corresponding DSC measurement is depicted in Figure 5(B) (cf. Figure S10 for the cooling period) and clearly shows that T g shifts from À49 to À39 C upon replacing the short methyl with the longer n-butyl moiety. Still, the sample exhibits a relatively low T g . Also, a second T g at lower temperature was observed, similar to that in C 1 -PIM-Cl. Overall, it was indeed possible to tune T g of the ionic copolymer by replacing the methyl group attached to the imidazolium cation with a longer alkyl substituent. However, we did not observe the desired decrease in T g . At the same time, we demonstrated the versatility of our synthetic approach.
F I G U R E 5 (A, B) DSC heating traces for C 1 -PIM-X and C y -PIM-cl (X = cl, TFSI, PF 6 , or BF 4 anion, y = 1, 4), respectively. Heating curves were obtained at a heating rate of 20 C/min. T g values are indicated next to the curves and were calculated as local maxima of first order derivative of heat flow

| Structure synergy and conductivity
As evident from the previous section, the use of both the TFSI À -anions and short alkyl substituents (in particular methyl) favors a low T g . Yet, C 4 -PIM-Cl showed relatively low T g of À39 C. Hence, we became interested in combining this copolymer with TFSI À -anions too, in order to assess whether we could achieve a comparably low T g as the combination of methyl and TFSI À in C 1 -PIM-TFSI. Consequently, we prepared C 4 -PIM-TFSI by anion metathesis of C 4 -PIM-Cl with AgTFSI. The 1 H-and 13 C-NMR, as well as the ATR-FTIR spectra (cf. Figure S11) confirmed its chemical structure. Most excitingly, the DSC trace ( Figure 6(A), cf. Figure S12) revealed a T g at À73 C for C 4 -PIM-TFSI, which is the lowest T g ever reported for an imidazolium-based PIL homo-or copolymer. 42,48 Previous records of low-T g polyimidazoliums include Choi's 42,48 radically polymerized imidazolium acrylates (À47 C), and Jourdain's 11 siloxane-based poly (1,2,3-triazolium) PIL (À62 C). It is worth noting, that C 4 -PIM-TFSI exhibits a slightly higher T g than Liang's 17 phosphonium-based polysiloxanes.
Clearly, C 4 -PIM-TFSI reveals a slightly lower T g (À73 C) than its shorter homologue C 1 -PIM-TFSI (À67 C). We previously mentioned that a longer alkyl chain promotes hydrophobic interactions. Here, the presence of the bulky and shape-anisotropic TFSI À -anion manifests in a plasticizing effect, which is more dominant than the hydrophobic interactions. This agrees with a recent report by Choi 48 and coworkers, in which they managed a lower T g of imidazolium-based polyacrylates by increasing the side-chain length to a certain extent. The authors used TFSI À and PF 6 À as anions and attributed the lower T g to an increasing size of charged repeat units. 42,43 In another article, they observed this trend independent of the types of anions used, 49 which is different from our findings here, as we noticed a strong dependence of T g on the choice of anions in the current polysiloxane imidazolium system. It is generally accepted that a low T g is crucial to achieve high ionic conductivities in PILs. 11 C 4 -PIM-TFSI showed the lowest T g of all imidazolium-based PILs and is thus promising as ion conducting material. In order to determine its anhydrous (water-free) temperaturedependent direct current (DC) conductivity (σ DC ), broadband dielectric spectroscopy (BDS) was employed. Since the anion exchange strongly affects T g , the ionic conductivity is expected to follow a similar trend. To underline this assumption, BDS measurements were also conducted with C 4 -PIM-Cl. Measurements were carried out in a temperature range from 160 to 480 K and in a frequency window between 10 À2 and 10 +7 Hz. The values of σ DC were derived from the distinct plateaus in the spectra of the real part of the complex conductivity (cf. Figures S13, S14, and S15). Figure 6(B) reveals the ionic conductivity of both samples against reciprocal temperature. The temperature dependence of σ DC follows a Vogel-Fulcher-Tammann (VFT) relation above T g . A dotted line in Figure 6(B) represents 300 K, where C 4 -PIM-Cl showed an ionic conductivity of 2.5 Â 10 À7 S/cm. This is relatively high for a halide-containing polyimidazolium. Usually, PILs carrying imidazolium chloride ion pairs show considerably lower values around 10 À10 S/cm. 2 Excitingly, C 4 -PIM-TFSI revealed a DC-conductivity of 4.8 Â 10 À5 S/cm. This is among the highest values ever reported for PILs. 11,50 Precisely F I G U R E 6 (A) DSC traces of C 4 -PIM-TFSI and C 4 -PIM-cl. (B) Temperature-dependent direct current (DC) conductivity (σ DC ) obtained from BDS measurements through fitting the Dyre formula 51,52 to the distinct plateau and the AC conductivity contribution to the real part of the complex conductivity spectra. Curves were fitted with a Vogel-Fulcher-Tammann (VFT) dependence speaking, to the best of our knowledge, this is the highest DC-conductivity of any imidazolium-based PIL so far. 2,19 As an example, our polymer exceeds the recent work by Karlsson, who pursued a similar strategy, that is, starting from the same precursor and using "click-chemistry" to obtain PILs ($2 Â 10 À5 S/cm). 19 Among other polycations, the conductivity of our polymer is comparable to Frenzel's 50 sulfonium-based polyacrylate containing bulky ammonium ions (3-6 Â 10 À5 S/cm depending on molecular weight). However, Jourdain and coworkers recently reported an even higher ion conductivity for their 1,2,3-triazolium-based polysiloxane ($6.7 Â 10 À5 S/cm). 11 In addition to the plateau of the DC conductivity contribution in the real conductivity spectra (σ'), the conductivity functions of both samples C 4 -PIM-Cl and C 4 -PIM-TFSI exhibit a shoulder at lower frequencies. The temperature-dependent frequency of this shoulder and the strength of the dielectric permittivity at this particular frequency suggest that it arises from polarization within the sample distinct from polarization at the sampleelectrode interface. 52 Thus, this observation strongly indicates that the samples might undergo microscopic phase separation as discussed above.

| CONCLUSIONS
With this contribution, we present a simple synthetic route towards siloxane-based PILs with pendant imidazolium-type IL-like species. We expanded the scope of our system towards an unusually low T g by introducing bulky anions and alkyl substituents of short and intermediate length. As a result, we obtained an imidazoliumcontaining copolymer with a record-low T g of À73 C. Such a low T g is a consequence of (a) the flexible polysiloxane backbone, 11 (b) the propyl spacer separating charges from the backbone, 2 (c) the low coordination tendency of TFSI À with the cation and its action as a plasticizer, 49 and (d) the increasing mobility of the repeat unit due to longer alkyl substituents, which reduces the cation-anion association. 42,43 At the same time, our product exhibits a strikingly high DC-conductivity of 4.8 Â 10 À5 S/cm at 300 K, elevating our novel PIL to one of the highest conducting ionic polymers so far. 11,19,50 4 | EXPERIMENTAL SECTION/ METHODS

| Substitution with 1-(n-alkyl) imidazole
In a typical experiment, the precursor P(DMS-co-CMS) was weighed into a three-necked round-bottom flask equipped with a reflux condenser. The apparatus was schlenked using nitrogen and the precursor dissolved in n-butanol (degassed prior to use). Subsequently, the corresponding 1-(n-alkyl)imidazole (20 mol eq. per Cl) was added via a syringe. The clear mixture was stirred at 50 C for 2 h, at 80 C for another 2 h, and finally refluxed (i.e., 117-118 C) overnight during which the solution turned yellowish. Reaction progress was determined by 1 H-NMR spectroscopy. The reaction mixtures were centrifuged and filtered to remove any insoluble substances, and finally concentrated at 80 C under reduced pressure. The yellowish phase was dissolved in dichloromethane and subsequently precipitated into diethyl ether. The supernatant was discarded and the oily phase was washed three times with diethyl ether. This process was repeated at least three times. Finally, all samples were dried at 80 C under high vacuum (10 À3 mbar) for 24 h, which gave almost colorless, clear, and viscous polymers.

| Anion exchange
1.0 eq. of C y -PIM-Cl (y = 1, 4) was dissolved in ethanol and, while stirring, a slight excess (1.03-1.31 eq.) of dissolved silver salt (AgX with X = TFSI À , PF 6 À , or BF 4 À ) was added at once under light protection. Immediate precipitation of silver chloride took place and stirring was continued at room temperature in the dark for 12 h. Subsequently, the mixture was centrifuged and the supernatant was collected. The precipitate was washed with ethanol and centrifuged again. The combined supernatants were filtered and concentrated at 80 C under reduced pressure. In case of C 1 -PIM-PF 6 and C 1 -PIM-BF 4 , the crude products were washed with dichloromethane three times. C 1 -PIM-TFSI was dialyzed in ethanol for 2 days (membrane with molecular weight cut-off of 3.5 kDa). Finally, all products were dried at 80 C under high-vacuum (10 À3 mbar) for 24 h to yield clear and viscous polymers. The absence of Cl À -ions in all samples was confirmed by silver nitrate titration.