Thermophysical and Electrochemical Properties of Ethereal Functionalised Cyclic Alkylammonium‐based Ionic Liquids as Potential Electrolytes for Electrochemical Applications

Abstract A series of hydrophobic room temperature ionic liquids (ILs) based on ethereal functionalised pyrrolidinium, piperidinium and azepanium cations bearing the bis[(trifluoromethyl)sulfonyl]imide, [TFSI]−, anion were synthesized and characterized. Their physicochemical properties such as density, viscosity and electrolytic conductivity, and thermal properties including phase transition behaviour and decomposition temperature have been measured. All of the ILs showed low melting point, low viscosity and good conductivity and the latter properties have been discussed in terms of the IL fragility, an important electrolyte feature of the transport properties of glass‐forming ILs. Furthermore, the studied [TFSI]−‐based ILs generally exhibit good electrochemical stabilities and, by coupling electrochemical experiments and DFT calculations, the effect of ether functionalisation at the IL cation on the electrochemical stability of the IL is discussed. Preliminary investigations into the Li‐redox chemistry at a Cu working electrode are also reported as a function of ether‐functionality within the pyrrolidinium‐based IL family. Overall, the results show that these ionic liquids are suitable for electrochemical devices such as battery systems, fuel cells or supercapacitors.


Introduction
Ionic liquids have gained considerable interest over the past few decades due to their numerousa ttractive properties such as extremely low vapour pressure, low flammability,h igh thermal stabilitya nd large liquid range. This hasl ed to an umber of groups researching ionic liquid applications particularly in the fields of catalysis [1][2][3] ,s eparation [4,5] and nanotechnology. [6] In comparison with molecular solvents, ionic liquidsa lso possess good ionic conductivity coupled with good electrochemical stability and, therefore, have been proposed as new electrolytesf or energy devices [7][8][9][10][11][12][13][14][15][16][17][18] and as solvents for electrodeposition of metals. [19,20] Within the fieldo fe nergy storaged evices,anumber of ILs have been developed with specific physiochemicala pplications. To date, however,m ost of these are based on imidazolium, [21] quaternarya mmonium, [22,23] pyridinium [24] and quaternary phosphoniumc ations. [25,26] Therein, the phosphonium,a liphatica nd cyclic ammonium cations show ah igher resistance to electrochemical reduction than corresponding imidazolium and pyridinium analogues and are, therefore, considered to be more promising as electrolytes. Within the cyclic aliphatic ammonium cations,t ypically pyrrolidinium, [27][28][29] piperidinium [28,30] and more recentlya zepanium [31,32] cationic based structures have been chosen for al arge number of investigationsf or potential electrochemical devices. However,t he downside to this improved stability is an increase in viscosity which results in significant drops in performance at mid-to-high current rates. [12,13] Attempts to reduce the viscosity of ILs based on cyclic alkylammonium- [TFSI] based ILs have been achieved by blending with appropriate organic solvents, [33][34][35] utilising alternative anion structures, [14,36,37] and functionalisation of the alkyl chains of the cyclic aliphatic ammoniumc ations with different group functionalities. In this regard, some studies have highlighted the positivei nfluence of ethereal functionalisation on the transportp roperties of acyclica nd cyclic ammonium-based ionic liquids. [37][38][39][40][41] In contrast, ionic liquidsc ontaining two or more ether groups appended to the cation have been somewhat understudied. Functionalised quaternary ammonium ILs based on di-, tri-and tetraethereal-based cations have recently been reported as promising candidates for energy applications. [42][43][44][45] In addition, as eries of novel pyrrolidiniuma nd pi-As eries of hydrophobic room temperature ionic liquids (ILs) based on ethereal functionalised pyrrolidinium, piperidinium and azepanium cations bearingt he bis[(trifluoromethyl)sulfonyl]imide, [TFSI] À ,a nion were synthesized and characterized. Their physicochemical properties such as density,v iscosity and electrolytic conductivity,a nd thermalp roperties including phase transition behaviour and decomposition temperature have been measured. All of the ILs showed low meltingp oint, low viscosity and good conductivity and the latter properties have been discussed in terms of the IL fragility,a ni mportant electrolyte feature of the transport properties of glass-forming ILs. Furthermore, the studied [TFSI] À -based ILs generally exhibit good electrochemical stabilities and, by coupling electrochemical experiments and DFT calculations, the effect of ether functionalisationa tt he IL cation on the electrochemical stability of the IL is discussed. Preliminary investigations into the Li-redox chemistry at aC uw orking electrode are also reported as af unctiono fe ther-functionality within the pyrrolidiniumbased IL family.O verall, the results showt hat these ionic liquids are suitable fore lectrochemical devices such as battery systems,fuel cells or supercapacitors.
peridiniumI Ls based on cations with two identicalm onoether groups have been developedw hich showed promising applications forenergy devices. [28] In order to develop ILs with good physiochemical properties for possible energy applications,w eh ave investigated the effect of cation structural modificationsf or af amily of ILs based on pyrrolidinium, piperidinium and azepanium cations functionalised with two or four ether groups paired with the benchmark [TFSI] À anion. The [TFSI] À anion was utilised, herein, owing to its reasonable cost and commerciala vailability,a nd also due to the very good promotion of hydrophobicity and thermal, chemical and electrochemical stability of the resulting ILs. The physiochemical and electrochemical properties of these new ILs have been compared against the corresponding non-functionalised ionic liquids.

Synthesis of Ether-functionalised Cyclic Alkylammonium ILs
Severals ynthetic strategies were undertaken to develop these ILs with the maximum yield and purity.I nitiala ttempts to synthesise these materials using conventional alkylation procedures employedf or imidazolium-basedI Ls proved problematic due to the low alkylation yield observed betweent he alkylamine and the ethereal precursors. Therein, low to moderate yields of deeply coloured ILs were achieved.T herefore, the proposed strategy was to firstly synthesise the corresponding aminoethereal based species and then perform the alkylation as the final step. These aminoethereal compounds were easily synthesised using the conventionalW illiamsonp rocedure under aqueous conditions (Scheme 1).
Initially,t he secondary cyclic amine (pyrrolidine, piperidine or azepane) was reacted with one molar equivalent of the desired alkylating agent (RBr,w here R = butyl, 2-methoxyethyl or 2-(methoxyethoxy)ethyl). The reactions were carriedo ut in water which upon completion resulted in the formationo f am onophasic system. The corresponding hydrobromide salt was then treated with potassium hydroxide and after stirring for ca. 24 ha tr oom temperature ab iphasic system formed. The resultingc rude amine was then fractionally distilled in vacuo resulting in good yields (61-71 %). Mono-functionalised ILs were obtained by reacting these tertiary amine compounds with dimethylsulfate generating the corresponding methylsulfate ILs in almostq uantitative yields. Di-functionalised ILs were obtained by reactingt he tertiary amine compounds with either 2-methoxyethyl or 2-(methoxyethoxy)ethyl bromide.
These reactions proceeded with lower reactivityd ue to the reduced nucleophilicty of the N-centre of the glyme-functionalised tertiarya mines. Metathesis of the bromide or methylsulfate anion-containing ILs with lithium bis[(trifluoromethyl)sulfonyl]imide (Li[TFSI]) in biphasic water/dichloromethane systems yielded the corresponding [TFSI] À salts in good yields (92-97 %). The prepared [TFSI] À salts were dried in vacuo fora t least 48 ha nd stored in aglove box.

Characterisation of Ether-functionalised Cyclic Alkylammonium ILs
This family of ILs were functionalised with different combinations of alkyl and ethereal chains in order to introduce ethereal-like properties to the liquid.T he abbreviations [Pyrr] + ,[ Pip] + and [Aze] + are used to express the cations with 5, 6a nd 7membered alkylammonium rings,r espectively.T he subscript text is used to express two additional groups bondedt ot he nitrogen of the ring;1 = methyl-, 4 = butyl-, (2o1) = methoxyethyl-,( 2o2o1) = methoxyethoxyethyl-. The chemical structure of all synthesised ether-functionalisedc ations,a nd their respectivea bbreviations are shown in Figure 1. Table 1s ummarises the thermal behaviour and physiochemical properties at 298.15Kof all the ILs.

Thermal Properties
The decompositiont emperatures (T d )o ft he studied ILs was determined by dynamic thermogravimetric analysis( TGA) wherein the temperature of the sample is increased at ar ate of 10 Kmin À1 .T he T d values for range of studied ILs, determined as the temperature at which a5%m ass loss is recorded, were in the range of 560-679 K. The TGA mass loss traces for each IL are shown in Figure S1 in the Supporting Information. For the appending alkyl group an increase in the number of ether linkages generally resulted in ad ecrease in thermal stability. Asimilar study involvingimidazolium and morpholinium cations also observed that the replacement of the alkyl group with an ether functionalised group also decreased the thermal stabilityo ft he corresponding ionic liquid. [46] The thermalb ehaviour of the synthesised ILs was also investigated by differential scanning calorimetry (DSC) within the temperature range of 183.13 Kt o3 23.15 K. Only the [Pyrr 14 ] [TFSI] IL showed observable transitions associated with crystallisation (T f )a nd melting( T m )( Figure 2). The exotherm (T f )o bserved on the heatingstep is associated with the crystallization of the supercooled liquid at 219 K. Further warming of the sample results in the observation of as mall endothermic process (T cc )a t2 43.7 Kp rior to the main endotherm associated with melting( T m )o ft he IL at 253 K. The endothermic feature T cc is associated with ac rystal-crystal phase transition between two metastable phases. The existence of these polymorphs in [Pyrr 14 ][TFSI] has been previously reported using DSC [47] and, more recently,byt emperature-dependent Ramanspectroscopy investigations of the phase changes occurring in this IL. [48] The results of the latter study suggested that the solid-solid phase transition occurs with as uppressiono ft he presence of the cisconformer of [ .T he T g value was estimated from the halfmaximum on the rise of the broad endothermic peak. The DSC heatingt races of the remaining ILs are presented in Figure S2. Functionalisation of ionic liquid alkyl chains with ether groups has been previously reportedt or esult in ad ecrease in melting point. [49] Therein, the reason for the observed T m and T g depression was thought to be due to the increasedr otational freedom and subsequent reduction in lattice energy as has been described elsewhere. [38] No such clear trend is observed   Figure 2). Unfortunately,t he positiono ft his feature is too close to the minimum temperature of the DSC measurement, ca. 183 K, to be accurately distinguished.

Viscosity
The viscosity of an electrolyte is an importantp arameterw hich contributest ot he transport capability of the active substances during electrochemical processes. To study the effect of the cation structure on this property,t he viscosity of the ILs was measured betweent he temperature range of ca. 293-363 K. The temperature dependences of the viscosity are presented in the form of Arrhenius-type plots in Figure 5. The numerical raw viscosity data is presented in Ta ble S3. These plots show that the majority of the ILs do not follow Arrhenius-type behaviour.A si sc ommonf or ILs, the temperature dependence of viscosity, h,c an be well described using the Vogel-Tammann-Fulcher(VTF) Equation (3): where h o , B h and T o h are the fitting parameters and T is the temperature. The best fit parameters for each IL are listed in Ta ble 3, where B h is relatedt ot he pseudo activation energy barrierl inked to the energy barriern eeded to be overcome for ions to move past each other (i.e. B h = E a /R, where R is the gas constant). [54] T o is relatedt ot he glass transition temperature of the ionic liquid and, by its position in the VTFe quation, the temperature at which the viscosity would approachi nfinity. Good correlation of the temperature dependence of viscosity of the ILs was achieved with the VTFequationw ithin the measuredr ange.
Furtherd iscussion of the coefficients of the VTF correlation is detailed in af ollowing section. The results demonstrate that the viscosity of the IL tends to increasea st he size of the alkyl ring of the cation is increased and in general the viscosityo ft he IL follows the trend [ (where Xr epresents ap articular set of functional groups). Comparing ILs of the same cation ring size, the data also shows that substituting the butyl group with am ethoxyethyl group reduces the measured viscosity of the IL. The ether group has been previously reported to reduce the viscosity of ILs relative  to similarly sized alkyl-functionalised equivalents. [38,39,50] This effect has been associated with the flexibility of ether moiety due to the free rotationo ft he CÀOÀCb ond. [39] Alternatively, the results of am olecular dynamics study of similarly sized butyl-and 2-methoxyethyl-functionalised acyclica mmonium ILs suggested that the reduction in IL viscosity is relatedt o changes in cation-cation interactions;t hat is, short range van der Waals interaction between alkyl chains are more efficient than for the alkoxy-group equivalent. [55]  [TFSI] is higher.T hese observed variations between the ILs with different ring sizes could possibly arise again from ac ombinationo fopposing factorsw hichp ositively or negatively impact the viscosity.I ncreasing ring size and ether chain length may increase the number of available low energy conformers, thus increasing the potentialv oid space promoting mass transport of the ions, that is, decreasing the viscosity. However,i ncreasing the molecular weight of the cation increases the van der Waals interactions, inhibiting the mass transportl eadingt oani ncreased viscosity.

Conductivity
The electrolytic conductivity, s,o ft he ether-functionalised ILs was measured acrosst he temperature range of 293-363K.T he experimental results, organised by the size of the alkyl ring of the cations, are presented as Arrhenius-type plots in Figure 6. The numerical conductivity data are presentedi nT able S4. The solid lines represent the non-linear fitting of the data with respect to the VTF Equation [Eq. (4)]:  where T is the absolute temperature and s o ,B s and T o s represent the fitting parameters. Like viscosity,t he VTF fitting equation is commonly used to accurately fit the temperature dependency of the conductivity of ionic liquidsw hich do not follow the Arrhenius Law. [56] Analogous to the parameters of the VTF equation used to describe IL viscosity in the previouss ection, B s is related to ap seudo-activatione nergy for conduction (B s = E a /R, where R is the gas constant)a nd T o is relatedt ot he glass-transition temperature of the IL, theoretically representing the temperature at which the conductivity of the system would disappear. Good correlation of the temperature dependenceo ft he conductivity was achieved by correlation with Equation (4). The VTF correlation parameters for the conductivity data of the studied ILs are presented in Table 4.
The reported data shows that increasing the alkyl ring size of the cation has an egative effect on the resulting conductivity.T his observation corresponds with the increasei nv iscosity (Table 3, Figure 5) and the fact that increasing the overall cation size is typically linked to ad rop in conductivity.S ince conductivity is dependent on ionicm obilities (as well as the number of charge carriers), the inverse relationship between the two quantities (conductivity and viscosity) is expected. The effect of substituting the butyl group for the methoxyethyl group results in an increase in conductivity regardless of the alkylammonium ring size. These observations correlate well with the observed reduction in IL viscosity and is attributed to the increased ionic mobility.
Increasing the size of the ether chain, from methoxyethylt o methoxyethoxyethyl, has an egative effect on the conductivity across the measured temperature range. Furthermore, increasing the overall cation size by substituting the methylg roup for as econd ether functional group yields as mall drop in the observed conductivity for the piperidinium-based ILs but asignificant dropi nt he measured conductance of the pyrrolidinium and azepanium-based ILs. This observation is likely to be the result of at rade-offb etween increasing the degrees of rotational freedom of the flexible ether chainsa nd increasingt he size of the cation which reduces the mobility of the cation and hinders the conductivity.A sd iscussed in terms of the measured IL viscosities, there will exist ab alance between factors either positively or negatively impacting the conductivity of the IL. In the case of the majority of the cations functionalised with two ether chains, the increased cation size appears to dominatet hese factors leading to lower conductance than the equivalent IL functionalised with only one ether chain. This is also consistentw ith previously reported observations which demonstrate the positive effect of the small flexible alkoxy chains. [39,50]

Ionicity
At ypical observation made when characterising ILs is the aforementioned inverse relationship between the viscosity and conductivity of the liquid.T he cooperative dependency of these behaviours is unsurprising given that viscosity describes al iquids resistance to flow whilet he conductance of an ionic liquid relies on the mobility of the ions carrying the charge. However,c onsidering the data discussed in the previous sections, the lower viscosities of certain ILs did not directly translate to higher conductivities.   Figure 8(a-c) are the ideal KCl lines representing the ideal Walden behaviour taken from extrapolation of the behaviour of a0 .01 mol dm À3 KCl aqueous solution,astronge lectrolyte where the ions are knownt ob ef ully dissociated and considered equallym obile. [57] Deviations below this calibration line, that is, where the conductivity of the IL is lower than may be expected at ag iven viscosity according to the Walden rule, imply the existence of significant ionic associationb etween the cation and anion with the tendency to behave as ion pairs. The magnitude of this deviation, DW (where DW represents the vertical displacement from the ideal KCl line)h as been used to classify the ionicity of the liquid. [57] For example, ILs for which DW is greater than one order of magnitude below ideal Walden behaviour (DW > 1, where DW = 1i sr epresented by the diagonal dashed lines in Figure 8) lie towards the bottom-right of the Walden plot and   can be classifieda sp oori onic liquidss ince the molar conductivity is much lower than possibly expected from ag iven low viscosity (high h À1 ). [58] Conversely,i ft he Walden behaviour of an IL lies closer to the ideal KCl line, where DW is small, it can be understood that these liquids exhibit more dissociated ionic character in which the constituent ions are more independently mobile. The data in Figure 8s hows that all the ILs in this study lie below the ideal calibration line, highlighting ac ertain degree of ionic association or aggregation between the constituent ions of the IL. However,t he deviation is not so strong as to classify any of the ILs as poor and DW [TFSI] andp ossibly account for the fact that this IL exhibits al ower conductivity even though it is less viscous than the analogous piperidinium IL functionalised with the shorter chain.
Furthermore,t he temperature dependant Walden behaviour of the ILs (Figure 8) shows that the apparent degree of dissociation is not independento ft he temperature according to the Walden rule (i.e. the slope of the data points is not parallel to the ideal KCl line). To account for these variations as af unction of the temperature an additional exponent, a W ,c an be introduced to give the fractionalW alden rule [Eqs. (5) and (6)]: log 10 L ðÞ ¼ log 10 c 0 ðÞ þ a W log 10 h À1 ðÞ ð6Þ where c' is analogoust ot he aforementioned Walden product. The linear fitting coefficients of Equation (6) and the calculated %ionicity as calculated from the fractional Walden rule for each IL are presented in Ta ble 5.
As ar esult of extrapolationo ft he seemingly ideal behaviour of aqueous KCl solutions, a W is considered to be unity for the ideal KCl line. Conversely,t he ionicity of majority of the ILs reported in this work appearst emperature dependanta nd a W values between 0.9-0.96 are exhibited, showingt hat, at increasingt emperatures, DW increases and the IL viscosity decreasesa tg reater rate than the increase in conductivity that may be expected from so-called ideal Walden behaviour.H owever,c omparison between the a W values for ILs and dilute electrolytic salt-in-solvent solutions has been previously described as largely meaningless;p articularly since real a w values relatingthe viscosity and conductivity of aqueous KCl solutions are found to roughlye qual 0.87. [59] Nevertheless, relative comparisono ft he temperature-dependant Walden behaviour of ILs shows that the relationship between IL conductivity and fluidity,w hereby a w values roughlye quatet o0 .9-0.95, is in reasonable agreement with previouslyr eported ILs;f or example, [EMIm][TFSI], a w = 0.906; [59] [EMIm][dicyanamide], a w = 0.95; [59] [BMIm][TFSI], a w = 0.998; [51] and[ Pyrr 14 ]][TFSI], a w = 0.971. [51] The effects contributing to ion-pair formation or aggregation will relate partly to delocalisation of the ionic charges urrounding the cation/anion centre and structuralc ontributions of bulky side chains contributing to steric hindrance of coulombic interaction. Within this familyo fI Ls, the negative chargeo ft he bulky [TFSI] À anion is highly delocalised via the electron withdrawing (trifluoromethane)sulfonylg roups, reducing affinity for ion-pair formation. Within the cyclic alkylammonium cations, generally the apparent %ionicity is greatestf or the alkyl-functionalised cations. Where ether functionality is introduceda nd the cation remains of similar size (i.e. butyl groups substituted by the 2-methoxyethyl group), electron donation from the lone pairs of the ethereal oxygen atoms would be expectedt o contributes omewhat to the localisation of the positive charge and, in turn, increaset he propensityf or ionic interaction with the counter ion. We reported as imilarr eduction in apparent ionic dissociation for ether vs. alkyl functionalised cyclic sulfonium [TFSI] À -based ILs based on the tetrahydrothiophenium cation. [60] For larger degrees of ether functionalisation, the tendency for ion-ion pair formation would be furtherd ependant on additional factors including steric hindrance, organisation/ freedoma nd interaction of flexible ether chains,v arying ring conformers and anion solvation/interaction in competition with ethereal groups and cation centres. As such, ac omplex variationint he %ionicity is not unexpected.

VTF Parameters and Fragility
Ta ble 6s hows ac omparison of the T g derived from DSC measurements and the VTF derived parameter, T o h/s ,f rom temperature dependant viscosity and conductivity measurements. The VTF parameters should agree if both charge transport properties, viscosity andc onductivity,a re closely coupled. However, the derived VTF fitting parameters are very sensitivet os mall deviations within the measurements and, as such, direct comparisonc an become problematic. Nevertheless, in many cases, both T o values showareasonable matchf or the studied ILs and the most significant discrepancies appear with the diether-functionalised ILs. Furthermore, due to the intricater elationship between IL viscosity and conductivity (or more specifically molar conductivity), differences in the activation energy barriers for the two processess hould also be relatedt oc hanges in the apparent ionicity of the ILs with temperature (i.e. a W ). [61] Since typical Arrhenius plots [ln(h)o rln(L)v s. 1/T]o ft he viscosity and molar conductivity of ILs yield non-linear behaviour, it is inferred that the apparent activation energy derived from Arrhenius equations [Eqs. (7) and (8)] is temperature dependant. /E a L represent the pre-exponential factor andt he activation energy barrier,r espectively for the two quantities, viscosity and molar conductivity.Assuch, following aprocedure describedp reviously, [61] the Arrhenius plotsf or each respective quantity were firstly fitted by at hird-order polynomial within the range of 293-343 K. By subsequent differentiation of the derived polynomial equation, the gradient (E a /R)was calculated at as eries of tangentst ot he curve. The activation energy (i.e. E a = R·gradient) was then taken as an average of the calculated values within the stated temperature range. The coefficients of the polynomial expansions and the derived average E a values are shown in Ta bles S6 and S7. Interestingly, the ratio between the two activation energies, E a L /E a h (detailed in Table 6) closely represent the derived values for the exponent of the fractional Walden rule, a W .I nf act, the majority of the values for the studied ILs differ by less than 2% (bracketed values in Ta ble 6). This is in agreement with the observations, and the originalh ypothesis of this relation,d etailed by Schreiner et al. [61] As mentioned previously,t he T o values derived from VTF correlationo fv iscosity and conductivity data are related to the IL'sglass transition temperature, T g (as measuredbyDSC). Typically for many ILs, T o usually falls approximately between 10 and 50 Kb elow the T g of the IL. [61,62] However,t he larger differences between T g and the theoretical T o are considered to depend on the liquid fragilityo rs trength. Furthermore, while lower values of T g would be expected to yield lower viscosity, or higherc onductivity,i th as been shown for ILs under ambient conditions that these properties are dependent on both fragilitya nd T g . [58,63] The concept of fragility, introduced by Angell, [64,65] describes the rate at which the transport properties and relaxation dynamics of ag lass-forming materialc hanges as the temperature approaches T g .M aterials described as highly fragile, or low strength,e xhibit highers usceptibility to structural change and, in turn, large changes in transport properties over small temperature variations close to the T g .
Conversely,s trong materials are consideredm ore impervious to structuralc hanges upon heating above their T g . [61] Ionic liquids, materials knownf or undergoing glass transitions and commonly inadequately described by normalA rrhenius-type temperature dependence (in terms of transport properties), have been frequently described as exhibiting intermediate to high fragility, such that their transport and relaxation dynamics change severalo rders of magnitude over relativelys mall temperaturedeviations close to their T g . [58,61,63,66,67] To assess the apparent fragility of the studied ionic liquids, quantified by the fragilityi ndex m,t he viscosity and conductivity data was further correlated as af unction of temperature by am odified VTF equation [Eqs. (9) and (10)]: where D h/s represents the strengthp arameter, inverselyp roportionalt ot he fragility index, m,a nd the remaining parameters are equivalent to the previously described VTF equation [Eqs.
(3) and( 4);a dditional notation of as uperscript f is included here to differentiate between parameters used in these and the previous equations].F or each dataset, an additional point of h(T g ) = 10 15 mPa so rs(T g ) = 10 À12 mS cm À1 was included in the correlation leading to slightly differentc orrelation parameters. These values represent an approximation of the sample viscosity andc onductivity,r espectively,a tt he T g of the material. [61,63] The correlation parameters of these fits, detailed in Ta ble S8,a re then used to extrapolate the temperature dependant behaviour of viscosity/conductivity towards the T g . This is presented in the form of an Angell-type plot (Figure 9), where the temperature along the x-axis is scaled by the material's T g .T he fragilityi ndex, m,c alculated using Equation (11) (and by differentiation of the fitted slope as T!T g ), is representative of the gradient of approacho fe ach respective property towardst he T g (higher m corresponds to as teeper approach, greaterc hanges in viscosity/conductivity as T approaches T g and ah igherf ragility). Full derivation of Equation (11) is explained in Ref. [61].
Within the Angell plots of Figure 9, an approximation of ideal Arrhenius-type behaviour of the transportp roperties of as trong (or non-fragile) materiali sr epresented by the diagonal dash and dash-dot-dot straight lines. For viscosity,t his represents am inimum fragility of 16 determined by the logarithmic ratio of the viscosity at T g and an approximation of the limiting viscosity (m min = Log 10 . This trend is supported by the behaviour of strong glass formers, for example, SiO 2 and GeO 2 . [58] For the conductivity plot,s ince the limiting conductivity is generally considerably more varied, the y-axis intercepto ft he hypothetical Arrhenius line on the Angellp lots was approximatedf rom the average s o f correlation parameter for the studied ILs [i.e. s o f = 2.86 mS cm À1 ,E q. (10)].T his yieldeda na pproximate value of 15 for the fragility index of the ideal Arrhenius line for conductivity.
The data presented in Figure 9s hows the studied ILs all exhibit very similar, high fragility behaviour.T his similarity is expectedc onsidering the shared anion and the closely related cationic structures. The values for m h/s derived from the viscosity and conductivity data are shown in Table 6a nd, in ranging from 115-170,a re of as imilar order to values previously reported forf ragile ILs. [61,63,68] Furthermore, the values for the fragility indices derived from the two different quantities, viscosity (m h )a nd conductivity (m s ), show reasonable agreement between their magnitudesa nd respective trends. Within the different ring-size families of ILs studied, no clear trend is observed. However,f or all cation ring sizes, the bis(2methoxyethyl) functionalised ILs (e.g. [Pip (2o1)2 ][TFSI]) appear to exhibit the highest fragilities;e xplaining why the ambient temperaturef luidity/conductivity of theseI Ls is not too drastically low,d espite possessing some of the highest T g values (particularly the piperidiniuma nd azepaniumI Ls).
Conversely,t he fragilityi ndiceso ft he butyl/methyl functionalised ILs (e.g. [Pyrr 14 ][TFSI]) tend to lie at the lower end of the measured values. Though the deviation of derived values is relatively small, and all studied ILs would be classed as highly fragile,t he lower apparent fragilityo ft he dialkyl-functionalised cations could be attributed to more efficient van der Waals interactions between alkyl substituents which are effectively dampened by substitution with ether groups.
Following the analysiso ft he data by Equations (9) and (10), values of the derived coefficients T o hf and T o sf for viscosity and conductivity,r espectively,c onsistently lie within ar ange of 17.6-27.4 Kbelow the measured T g .Furthermore, the ratios between these coefficients and the measured T g of the studied ILs (T g /T o hf and T g /T o sf )a re shown in Ta ble 6. An average value for T g /T o hf of 1.13 AE 0.004 is obtained from both viscosity and conductivity derived coefficients. All of these ratios satisfy the originalo bservation of Angell and described by Equation (12): where m min referst ot he minimum fragility index for each respectivep roperty (m min = 16 and1 5f or the viscosity and conductivity data, respectively). [65]

Electrochemical Stability
The electrochemical stability of the alkyl and ether functionalised cyclic alkylammonium [TFSI] À ILs wasd etermined by cyclic voltammetry (CV) at a3mm diameter glassy carbon (GC) macrodisk electrode. Under dry,a nd inert conditions of an Ar-filled glovebox, the potential of the working electrode cycled at 2mVs À1 until ac urrent density boundaryo fAE 0.5mAcm À2 was recorded, at which point the potential sweep was reversed.
The temperature of all CV measurements was ca. 303 AE 1K as dictated by the internal atmosphere of the glovebox. The electrochemical windowso ft he studied ILs are presented in full in Figure S3 in wherein the reductive currentl imits are sequentially increased. The CV traces show that the magnitude of the impurity peak at ca. 0.7 Vv s. Ag[NO 3 ]/Ag increases as more charge is passed on reduction. In fact, there appears to be ap roportionality between the two processes (as shown in Figure S4a, inset graph) and the oxidative integral charge of the impurity peak corresponds to ca. 22-25 %oft he total charge passed on the reductive sweep. The electrochemical stability window,d efined as the potential range between the reductive and oxidative decomposition potentials of the electrolyte, is indicative of the stable operating voltage of the materiali ne lectrochemical systems. The wide potential windowe xhibited by many different IL structures is one of the mosta ttractive features of applying ILs as electrolytes in electrochemical energy storage devices like Libatteries and high-voltage electrochemical doublelayer capacitors (EDLCs). [16,[69][70][71][72] For EDLC devices in particular,e xtension of the maximum operative voltage( U max )r ange, typically limited by the electrolyte formulation, can result in direct enhancement in device specific energy (E Sp )a nd specific power( P)c apabilities due to the proportionalities; E Sp = 0.5·C·U max 2 and P = U max 2 /(4·ESR)w here C and ESR represent the capacitance and equivalent series resistance, respectively. [70] The onsetp otentials of reductive and oxidative decomposition (E a and E c ,r espectively) of the IL electrolytes studied herein are approximated using ac ut-off current density of AE 0.1 mA cm À2 .A ss uch, the electrochemical stability window (DE)i sd efined by the difference between these upper andl ower potential limits (i.e. DE = (E a ÀE c )V ). The numerical potential limits determined by cyclic voltammetry are presented in Ta ble 7a nd represented graphically in Figure 10. With the exceptiono ft he bis(methox-  [TFSI]), the studied ILs show electrochemical windows greater than 5V , wherein reductived ecomposition occurs close to the Li + /Li reductionp otential and oxidative decompositiono ccurs at ca. 4.9-5.9 Vv s. Li + /Li. As described in the experimental section, an internal ferrocener edox couple was used to normalise the reference potentialv s. the Li + /Li reduction potential following determination of electrochemical windows (based on the approximation E Fc + /Fc % 3.2 Vv s. E Li + /Li ). An exemplary CV of the ferrocener edoxc ouple versus the Ag[NO 3 ]/Ag reference and versus aL i-metalr eference electrode is shown in Figure S5a in the ESI to justify this approximation. The data presented in Ta ble 7a nd Figure 10 straightforwardly shows some apparent trends in the electrochemical stabilities of the studied ILs. Firstly,a sm ay  , exhibit the largestr eductive and oxidative stabilities and, in turn, the largest electrochemical windows( ca. 5.9 V). Beyond this, increasing the degree of ethereal group functionalisation on the three alkylammonium rings appears to yield subsequent reductionsi nt he magnitude of the electrochemical windows, DE. Specifically,s ubstitutiono fb utyl groups for the 2methoxyethyl group reduces E c slightly but affects al arger reductioni nt he oxidative stability, E a .L engthening of this ethereal group (e.g. from [Pyrr 1(2o1) ] + to [Pyrr 1(2o2o1) ] + )f or all three ring sizes promotes af urther significant reduction in the oxidative stabilities of the studied ILs. This, and the large degree of variation in oxidative stability of the remaining ILs, is interesting since the more traditional consensus aligns that the oxidative decomposition is generallya ttributed to oxidation of the anion (and viceversa for the cathodic reductivel imits). [73,74] However,m olecular dynamics (MD)s imulation and density functional theory (DFT) calculations have suggested this may not alwaysb es os traightforward. [75][76][77] To comparet he trends in oxidative and reductive stabilities of the studied ILs as af unctiono ft he variation in cation structure, the HOMO/ LUMO energy levelsofmany geometrically optimized conformers of cations, anions and ion pairs were determined from DFT calculations. An extensive summary of the calculated HOMO/ LUMO energy levels and surface diagrams of single ions and ion pairs is provided in Ta ble S8 in the ESI.
Within the cation ring size families, increasing ether functionality has almost identical effects on the computed HOMO and LUMO values based on ion calculations;t hat is, HOMO: However,t he functional group modifications described affect am uch larger change in HOMO values (e.g. % 3eVrange between lowest cation, [Pyrr 14 ] + = À12.39 eV,a nd highest cation, [Pyrr (2o2o1)2 ] + = À9.46 eV) than the computed single ion LUMO values (where the range is % 0.7 eV). For the analogous functional group pairing, the effect of the cation rings ize on the computed values is generally small but the LUMO follows the trend of [Aze X ] + 0 [Pip X ] + 0 [Pyrr X ] + ,i na pproximate agreement with trends in observed E c values. Furthermore, the agreementsi nt he computed LUMO values and E c limits can be further extended to include our previously reported experi-mental and computational observations for as eries of [TFSI] À ILs based on ether-functionalised acyclics ulfonium cations. [77] These sulfonium based ILs undergo reductive decomposition at ca. 1Vhigher under identicalc onditions and the computed LUMO values for the sulfonium cations are ca. 1eVm ore negative than calculated for the cyclic alkylammonium cations reportedh ere.
Additionally,t he trend in cation HOMO values is the same approximate trend observedf or E a .Ap lot of the computed cation HOMO values vs. the experimentally observed oxidative limits (see Figure S6) for each IL displays good linear correlation;w herein only [Pip (2o1)2 ][TFSI] appears an outlier. However, given that the electrochemical window of an IL is limited by the least stable component in both oxidative or reductive potential regions, the variation in cation HOMO values does not alone justify the observed E a trends since the single ion HOMO values computed for the [TFSI] À anion component are much higher (i.e.l ess stable towards oxidation; À4.19 eV, À3.97 eV and À3.89 eV for the three [TFSI] À conformers). However,w e recently reported similare xperimentala nd computational observations for the series of ether-functionalised acyclics ulfonium [TFSI] À ILs. [77] There are several possible explanations for the prominente ffect of the cation structureo nt he oxidative stabilityo ft he IL. Firstly,t he cation may undergo direct oxidation at the electrode interface. Given the well-structured natureo fI Le lectrical double layers,t he presence of cationic species directly at highly positive electrode surfacea ppears counter-intuitive. However,t he long and flexible 2-(methoxyethoxy)ethyl group will somewhat change the organisationo f any IL double layer with this ether group. Further, the observed E a of these ILs is similar to the oxidative stabilityofanalogousg lyme type solvents (e.g. tri/tetra ethylene glycol dimethyl ether;4 .6-4.9 Vv s. Li + /Li) [80,81] indicatingt he possibility of the terminal etherealo xygen (where the cation HOMO resides, see surfaced iagrams in Ta ble S8) behaving more as af ree ether group.
As econd rationale relates to many possible changes in cation-anion interactions as af unction of the increasing ether functionality.I naprevious section, we demonstrated using the Walden behaviour that, with af ew exceptions, the ionicity (or degree of dissociation)o ft he ILs decreased with the introduction of more ethereal groups. Electron donation from the ethereal oxygen atoms into the nitrogen centre of the cation decreases the delocalisation of the positive charge and, in turn, the increased tendency for the formation of ion pairs and the associated restricted freedom of the ions may increase the possibility of direct participation of the cation during oxidative decomposition. Additionally,i ncreased localisation of the cation charge contributes to reductions in the partial positive charge of the alpha methyl/methylenep rotons and reductions in the computed cation LUMO values. However,i nt erms of oxidative stabilities, the increased Coulombic sharing of available orbitals between cation and anion trough ion-pair formation would be expected to increase HOMO energy levels relative to single ionc omputed levels, that is, reducing the anodic potential barrier of HOMO oxidation and reducing oxidative stability of the IL. Additionally,i no ur observations reportedf or the ether-functionalised sulfonium [TFSI] À ILs, [77] the DFT calculations of ion pairs indicatedt hat when increasing the level of ether-functionality on the sulfonium cation, the cis-conformer of the [TFSI] À anion becomes more favoured over the transconformer.B ased on the single ion DFT calculations, cis-[TFSI] À should exhibit al ower electrochemicalo xidative stability (by 0.22 eV) than the trans alternative. However,w hile the comparative ion-pair simulations of the alkyl ande ther functionalised ammonium-based ILs did not reveal ac lear-cut change in the global minimum energy conformer of the [TFSI] À anion,t he relative energy difference between trans-and cis-c onformers was lower for [Pyrr 1(2o2o1) ][TFSI] than for [Pyrr 14 ][TFSI] (see Ta ble S9). We relate this observation, in conjunction with the reduced apparent ionicity throughi ncreased cation-anion interaction, to the restriction of the anion positioning around the cation centre as ar esult of the ether chain presence;t hat is, interaction between anion and etherealo xygen atoms is unfavourable and, thus, the freedomo fi on coordination around the cation centre is reduced. Consequently,t he probability of the less electrochemically stable cis-[TFSI] À conformers existing in the ether functionalised IL is highert han mayb ee xpected for the alkyl functionalised analogue. ) can possibly account for the reduced oxidative stabilities observed. At hird consideration relates to the possibility of the ether-functionalised cations readily reacting with any radical products of [TFSI] À oxidation. Such ap rocess would be autocatalytic and somewhat enable ar eductioni nt he oxidative potentialb arrier required to proceed the reactions.U ltimately,t he overall processesm ay be combi-nationso ft he discussed ideas and we are currently devising experimental procedures to understand the reactions taking place during decomposition to makes ense of the prominent effectsofc ation structure.

Li-redox in Pyrrolidinium ILs
One of the most desirable applications for electrochemical stable ILs, in addition to EDLC electrolyte components, is as electrolytes for alkali metal batteries, notably Li and Na. To utilise Li/Na-metal anodes in place of traditional graphitic anodes, the electrolyte must facilitate deposition and stripping of the alkali metal ions with ah igh cycling stability. In the context of electrodeposition and stripping of Li Li[TFSI] in these ILs, [82] and the [FSI] À also shows evidence of good SEI (solid electrolyte interphase) formation during Li anode plating/stripping processes [12,36] and also during intercalation into Li-iong raphitic anodes. [30,82] Figure 11  sition at potentials more negative than À3.2 Vv s. Fc + /Fc which is succeeded by an anodic stripping current peak on the reverses can. As for the measurement of the electrochemical windowsa tt he GC electrode, the reference potentials were normalised initially vs. an internal Fc + /Fc couple in each sample and further normalised using the approximation E Fc + /Fc % 3.2 Vv s. E Li + /Li .L ikewise, exemplary CVs of the Fc + /Fc couple vs. the Ag[OTf]/Ag and Li-metal reference electrodes at the Cu workinge lectrode are presented in Figure S5b in the ESI to support this approximation.
During the first cycles,i ndependently of the Li-salt component, the onset of Li electrodeposition occurs at ca. - This behaviour prompts the consideration that the secondary ethereal oxygen in the cation functional groups (i.e. the oxygen furthest from the positive nitrogen centre) is involved significantly in the Li + -ion solvation complex in these ILs. Raman/NMR investigations have shown previously that substitution of the butyl group for the short methoxyethyl group enables direct IL-cation-Li + interactions (via the weak Lewis basic ethereal oxygen atom), reducingt he coordination number of [TFSI] À anions which form the Li + -solvation sphere. [83] Furthermore, the presence of the longere ther groups in a N-ethyl-N,N,N-tri[2-(2-methoxyethoxy)ethyl]ammonium-[TFSI] IL has been recently reported to promote the solubility of Na[TFSI] and display correlated relaxation dynamics between IL and Na + ions, indicative of significant interactions, and potential cation-cation cross-linking, between alkali metal cation andI L cation. [43] As such based on these interesting observations, and the apparent increased cycling stabilityo ft he di[2-(2-methoxyethoxy)ethyl]-functionalised IL (relative to [Pyrr 1(2o2o1) ][TFSI]), furtheri nvestigations in the IL/Li-salt interactions andt he effects of IL-salt concentrations in these ILs are ongoing.
Beyond the first cycle in all studied ILs, the onset of Li electrodeposition shiftedt ol ess cathodic potentials slightly but the potential to drive the electrodeposition process at higher currents shiftedt om ore cathodic potentials, associated with evolution of higher overpotentials required to drive the process. Similarly,w ith progressive cycling, anodic shifts of the strippingp eak can be observed for most of the IL formulations. However,w ith both Li Figure 10, Ta ble 7) at the GC electrode. Evidence for the apparent stabilisation of some [TFSI] À -based ILs towards reductived ecomposition in the presence of dissolved Li-salts has, however,b een reported numerous times previously. [28,41,79] The best cycling performance, taking into consideration stability on repeat cycling and the maximum utilised current densities, appearst ob et he combination of [Pyrr 1(2o1) ][TFSI] with Li[FSI] salt. This may be expected given this IL exhibits the best transport properties from the all ILs described in this work, the positive Li + conducting SEI film-forming properties of the [FSI] À anion, [12,36] and the good stripping/plating performance reported previously for this IL. [79] However,t he applicability of these ILs in for Li-metal systems (and possibly Na systems also) could be further optimised by investigation of alternative salts, salt concentrationr anges and SEI filmf orming additive compounds.

Conclusions
The synthesisa nd systematic thermophysical and electrochemical investigationsof1 5c yclic alkylammonium[ TFSI] À -based ILs with and without varying degrees of ether functionality is reported. All the studied ILs exhibit low meting temperatures and high thermal stabilities. Ether functionality on the cation slightly reduces upper limits of thermal stabilities but all ILs exist in the liquid phase under wide range of potential operating temperatures for electrochemical devices (e.g. À20 to > 100 8C). The transport properties, relating to the IL viscosity and conductivity,a re strongly affected by cation structure. Initial substitution of alkyl groups for like-sized ether groups affords important reductions in IL viscosity ande nhances IL conductivities observed, in line with previous observations. However,i ncreasing the degree of ether functionality engages more complex trends in viscosity and conductivity depending on cation ring size. For example, the longer flexible ether chains may increase degrees of rotational freedoma nd reduce cation-cation interactions, increasing IL fluidity but, in most cases, reducing IL conductivity.T he bulky side chains obviously increasec ation size, likely reducing ionic mobilities (most prominentf or the smallp yrrolidinium cations), but also increase the propensity for ion-pair formation,a se videnced by the assessment of the Walden-type behaviour for the studied ILs. Herein, all but one of the studied ILs exhibited > 40 % ionicity and may be classed as good ionic liquids but, as stated, the introduction of ether functionality at the cation generally affected reductions in the observed ionicity.F urthermore, estimations of the IL fragilities were completed using extrapolation of both the viscosity andc onductivity measurements towards glass transition temperatures. These results furtherh ighlighted the coupled nature of both physicalp roperties and revealed the studied ILs to be highly fragile materials;adesirable quality for IL electrolytes.
Furthermore, all of the studied ILs exhibited wide electrochemicalw indows( > 4.5V), an important feature for potential IL electrolytes. Nevertheless, initial substitution of the alkyl groups for ether groups reduces the magnitude of the ob- served windows slightly and further lengthening of the ether group yields greater reductions. Most interestingly, despite the ILs sharing the equivalent[ TFSI] À anion, the modifications of the cation structure affect the greatest changes in the oxidative stabilityo ft he IL, contrary to the classical understanding. Calculations of the HOMO/LUMO energieso fs ingle ion conformers reveal the least oxidatively stable orbital of the cation lies on the terminal oxygen of the ether groups but the calculated magnitudes still indicate the lowest thermodynamic pathway towards IL oxidation should be removal of the electron from the HOMO of the [TFSI] À anion.H owever,i ncreasing ether functionalisation at the cation, firstly,i ncreasest he propensityf or ion-pair formation (Walden behaviour) and, secondly,i ncreases the geometric stability of the less electrochemically stable cis-[TFSI] À conformer (DFT calculations). These factors may contributet ot he observed reduction in the oxidative stability limits of the ILs but furthere xperimental work is required to fully understand the sequence of eventso ccurring at the electrode/electrolyte interfaced uringe lectrochemical decompositiono ft he IL. Preliminary investigations also revealed, or furthers upported, the applicability of these ILs for plating/ strippingp rocesses at Li-metal anodes. Some of these results were indicative of interesting IL-cation-Li-cation interaction effects observed with higher degree of ether-group functionalisation, prompting furtheri nvestigations into these IL/Li-salt compositions.

Ionic Liquid Synthesis
Full description of the synthesis of the 15 studied ILs is provided in the Supporting Informatoin in conjunction with results of NMR, Microanalysis, and Li-content analysis.

Preparation of Materials for Physiochemical Measurements
Prior to all physical and electrochemical measurements, all ILs were dried under vacuum (ca. 10 À3 mbar) at 353-373 Kw ith stirring for at least 2days. This treatment was found to reduce the water content to below 100 ppm, as measured by Karl Fischer coulometric titration. The resolution of the water content measurements was 0.001 wt/wt %( or 10 ppm) and measurements were completed in duplicate. Once dried, the IL samples were stored in the Ar-filled glovebox with moisture levels less than 3ppm water.L ithium content of the ILs was analysed by inductively coupled plasma optical emission spectroscopy (ICP-OES) on an Agilent 5100 ICP-OES, and along with Microanalysis, was performed by Analytical Services at Queen's University,B elfast. 1 Ha nd 13 CNMR spectra were recorded at 293 Ko naBruker Avance DPX spectrometer at 300 MHz and 75 MHz, respectively. 1 Ha nd 13 CNMR spectra for all [TFSI] À ILs are shown in Figures S7 to S36 in the ESI.

Physical Measurements
Density measurements were performed using aD M40 oscillating tube density meter (Mettler To ledo, AE 1 10 À4 gcm À3 )i nt he range of 293.15-363.15 K( AE 0.1 K). The instrument was cleaned using acetone and dried using dehumidified air prior to any measurements. The viscosity of the ILs was measured using aB ohlin Gemini Rotonetic Drive 2c one and plate rheometer (AE 1%)f rom 293-363 K( AE 0.01 K) at atmospheric pressure. The rheometer was calibrated using ultra-pure water and an oil viscosity standard (ASTM Oil Standard S600, Cannon, 1053 mPa sa t2 98.15 K). Dynamic thermal decomposition profiles were collected by thermogravimetric analysis (TGA) using aT GA Q5000 (TAI nstruments). Heating profiles were performed at ar ate of 10 Kmin À1 under nitrogen flow from room temperature to 773 K( AE 1K). Thermal phase transitions were recorded using differential scanning calorimetry (DSC) analysis on aD SC Q2000 (TAI nstruments) with ca. 5mgI Ls amples in hermetically sealed Al pans. Ah eating gradient of 5Kmin À1 was used and 183 Ka nd 323 K( AE 0.1 K) were used as the lower and upper temperature limits, respectively.
Conductivity measurements were conducted using as ensION + EC71 benchtop meter with a3 -pole platinum sensION + 5070 conductivity probe (< 0.5 %o fr ange) with an in-built Pt1000 temperature probe (Hach Lange). Regular calibration of the probe was completed using aqueous KCl standard conductivity solutions (147 mScm À1 ,1 413 mScm À1 ,a nd 12.88 mS cm À1 at 298 K). The immersion and sealing of the conductivity probe in the liquid sample was carried out in an Ar-filled glovebox. The conductivity probe, when disconnected from the meter,w as immersed in the liquid sample (ca. 3cm 3 )i nside ag lass sample tube also containing as mall magnetic stirrer bar.T he sample was sealed with the probe in the glass tube using an O-ring seal and Parafilm. The conductivity of the sample was then recorded with stirring as af unction of temperature (using the temperature reading built into the conductivity probe). The temperature of the sample was varied from 293-363 K(AE 0.2 K) using asmall oil bath and ahot-plate with athermocouple control. The sample conductivity and temperature were recorded when the observed values were stable for ca. 1min.

Electrochemical Measurements
All electrochemical measurements of the electrochemical windows and the Li-redox chemistry were completed using cyclic voltammetry using either an Autolab PGSTAT 302 workstation (Metrohm) or aV MP3 multi-channel workstation (BioLogic). Ac ustom 3p ort glass cell was prepared in-house which allowed completion of all electrochemical measurements with ca. 1cm 3 of electrolyte. Cell preparation and electrochemical measurements were conducted inside the Ar-filled glovebox. The temperature of the electrochemical measurements was ca. 303 K, as determined by the internal atmosphere of the glovebox. For the determination of electrochemical windows, the working electrode was a3mm diameter glassy carbon macrodisk electrode (ALS Co.,L td). For Li-redox chemistry, the working electrode was a1 .6 mm diameter copper macrodisk electrode (ALS Co.,L td). Prior to all measurements, the working electrodes were cleaned with acetonitrile and ultra-pure water and then polished using alumina slurries of decreasing grain size