Exploring the effect of fluorinated anions on the CO2/N2 separation of supported ionic liquid membranes

The CO2 and N2 permeation properties of ionic liquids (ILs) based on the 1-ethyl-3-methylimidazolium cation ([C2mim] ) and different fluorinated anions, namely 2,2,2-trifluoromethylsulfonyl-N-cyanoamide ([TFSAM] ), bis(fluorosulfonyl) imide ([FSI] ), nonafluorobutanesulfonate ([C4F9SO3] ), tris(pentafluoroethyl)trifluorophosphate ([FAP] ), and bis(pentafluoroethylsulfonyl)imide ([BETI] ) anions, were measured using supported ionic liquid membranes (SILMs). The results show that pure ILs containing [TFSAM] and [FSI] anions present the highest CO2 permeabilities, 753 and 843 Barrer, as well as the greatest CO2/N2 permselectivities of 43.9 and 46.1, respectively, with CO2/N2 separation performances on top of or above the Robeson 2008 upper bound. The re-design of the [TFSAM] anion by structural unfolding was investigated through the use of IL mixtures. The gas transport and CO2/N2 separation properties through a pure [C2mim][TFSAM] SILM are compared to those of two different binary IL mixtures containing fluorinated and cyano-functionalized groups in the anions. Although the use of IL mixtures is a promising strategy to tailor gas permeation through SILMs, the pure [C2mim][TFSAM] SILM displays higher CO2 permeability, diffusivity and solubility than the selected IL mixtures. Nevertheless, both the prepared mixtures present CO2 separation performances that are on top of or above the Robeson plot.


Introduction
The development of supported ionic liquid membranes (SILMs) for CO 2 separation has been widely investigated in recent years mainly due to their easy preparation and versatility. [1][2][3] In contrast to traditional liquid membranes, which are produced by impregnating a porous membrane support with common organic solvents, 4 SILMs use ionic liquids (ILs) and thus benefit from negligible displacement of the liquid phase from the membrane pores through evaporation, 5,6 due to the low volatility of ILs. 7 It should also be emphasized that within the CO 2 separation context, the most important features of ILs are their high CO 2 affinity over light gases [8][9][10] and their inherent designer nature that enables the tailoring of IL properties by proper selection of cations and/or anions or via the addition of specific functional groups.
Numerous works have investigated the effect of IL chemical structure on the gas permeation properties of SILMs. A broad diversity of cations, such as imidazolium, 11 triazolium, 12 thiazolium, 13 pyridinium, 14 cholinium, 15 ammonium, 16 and phosphonium, 17 combined with halogens, sulfonates, carboxylates, fluorinated or cyano-functionalized anions, have been studied. Other works, mostly focusing on imidazolium-based ILs, have also explored the effect of alkyl, 18 fluoroalkyl, 19 etoxyalkyl, 20 and aminoalkyl 21 -functionalized cations. Since IL anions have a stronger influence on the CO 2 separation performance of SILMs than IL cations, 1 they deserved from the start a closer look. The first studies on SILMs made use of fluorinated anions such as bis(trifluoromethylsulfonyl)imide [NTf 2 ] À , tetrafluoroborate [BF 4 ] À , and hexafluorophosphate [PF 6 ] À and enabled drawing conclusions about the CO 2 -phylic behaviour and high CO 2 permeabilities of these anions. 22 More recently, low viscous ILs with cyano-functionalized anions, such as tricyanomethanide [C(CN) 3 ] À and tetracyanoborate [B(CN) 4 ] À , [23][24][25] have been recognized as better candidates for the development of improved SILMs, because of their superior CO 2 permeabilities and permselectivities when compared to the most used [NTf 2 ] À anion. Task-specific ILs bearing amine groups, such as those containing amino acid anions, [26][27][28] have also been proposed to prepare SILMs, since amine groups can chemically bond CO 2 and act as carriers for CO 2 facilitated transport through SILMs at low pressures. However, the high viscosity of these task-specific ILs is undoubtedly a key limitation, as CO 2 diffusion is strongly compromised.
In an effort to improve the CO 2 permeability and permselectivity properties of SILMs, our recent studies explored the use of IL mixtures by fixing the [C 2 mim] + cation and researching on different anion chemical structures. Initially, SILMs based on IL mixtures combining anions with different CO 2 solubility behaviours were investigated: thiocyanate ([SCN] À ), dicyanamide ([N(CN) 2 ] À ) and bis(trifluoromethylsulfonyl)imide ([NTf 2 ] À ) that present physical solubility; acetate ([Ac] À ) and lactate ([Lac] À ), which additionally have chemical solubility. 29 3 ] À and different amino acid anions, so that one IL component maintains the low viscosity, while the other provides the desired chemical characteristics for the active transport of CO 2 . 31 The overall results of these studies showed that mixing anions with specific chemical features allows variations in IL viscosity and molar volume that significantly impact the gas transport through SILMs, and thus tailored permeabilities and permselectivities can be achieved. [29][30][31] In the present work, the gas permeation properties and CO 2  Additionally, this work investigates the impact on gas transport through SILMs of using a pure IL versus a structurally similar IL mixture as the liquid phase. Inspired by the fact that the [TFSAM] À anion has an unusual asymmetric chemical structure, which combines both fluorinated and cyano functionalities, the re-design of the chemical structure of a pure [C 2 mim][TFSAM] IL through the use of IL mixtures is explored. For that purpose, different pairs of ILs, based on the [C 2 mim] + cation and anions containing fluorinated or cyano functionalities, were selected and their gas permeation properties were compared to those of the pure [C 2 mim][TFSAM] SILM. One of the IL mixtures contains [NTf 2 ] À and [N(CN) 2 ] À anions, whose gas permeation properties were previously determined, 29 whereas the other IL mixture is based on [OTf] À and [SCN] À anions and its gas transport properties are reported here for the first time.

Results and discussion
Gas permeation through SILMs having fluorinated anions The structures of the pure ILs bearing fluorinated anions are depicted in Fig. 1.
The water content (wt%), molar mass (M), viscosity (Z), density (r) and molar volume (V m ) values of the pure ILs used as liquid phases in the studied SILMs are summarized in Table 29 It is important to mention that in order to attain stable SILMs, both hydrophilic and hydrophobic supports were used according to the hydrophobicity of ILs, and the results are compared in this section, irrespective of the support used. From Table 2, the same trend in gas permeability is valid for all the studied SILMs: PCO 2 c PN 2 , as expected. Regarding the influence of the fluorinated-based anions, SILMs having the [FSI] À , [TFSAM] À and [FAP] À anions present higher CO 2 permeabilities of 843, 753 and 624 Barrer, respectively, than the SILM containing the [NTf 2 ] À anion, which is well-known for its high CO 2 permeability (589 Barrer). 29 It should be noted that in spite of the similar structures of [NTf 2 ] À and [FSI] À anions, in which the difference consists in two extra À CF 3 groups in the [NTf 2 ] À anion structure (Fig. 1 . Notice that the [FAP] À anion has the most different chemical structure among all the IL anions studied in this work, consisting of a phosphorus atom surrounded by fluorine atoms, without sulfonyl functional groups (Fig. 1). Moreover, taking a closer look at the gas permeabilities obtained through SILMs immobilized with the remaining ILs, it can be seen that higher CO 2 permeabilities are achieved for ILs with anions bearing a smaller number of fluorine elements, such as [TFSAM] À and [FSI] À anions ( Table 2).
Gas diffusivity (D) is a mass transfer property that directly accounts for gas permeability (eqn (1)). Typically, the higher the gas diffusivity, the faster is the gas flux through the SILM. The experimental CO 2 and N 2 diffusivity values obtained through the prepared SILMs are presented in Table 3. The CO 2 diffusivities of the SILMs with fluorinated anions can be ordered as follows: which fully corresponds to the IL anion order observed for CO 2 permeabilities ( Table 2). As for N 2 diffusivities the subsequent order is attained: , which is nearly the same IL anion order observed for N 2 permeabilities, but different from that obtained for CO 2 diffusivities and permeabilities.
The relationship between IL viscosity and gas diffusion is in fact the basis of the dependence of permeability on viscosity that we have shown above. A number of works have reported the inversely proportional relationship between gas diffusivity and IL viscosity. 3,16,17,41 Along this line of thought, the relationship between experimental CO 2 diffusivities and IL viscosity for the studied SILMs is depicted in Fig. 2.
In agreement to what was previously observed in the literature for other SILMs, 4    Over the past few years, a number of correlations have been proposed with the intention of understanding the relationships between CO 2 solubility and the intrinsic properties of ILs. [42][43][44] The proposed models showed that CO 2 solubility increases with increasing IL molecular weight, molar volume and free volume. 10 Taking into consideration the gas solubility results obtained in this work (  46 it was recently reported after critical analysis that no special effect of the fluorination upon the CO 2 solubility has been observed for both perfluorocarbon and heavily fluorinated ILs. 50 In fact, the introduction of fluorination into the anions of the ILs studied in this work does not significantly affect the obtained gas solubility values (  (Fig. 3), which is studied here for the first time, and [C 2 mim][N(CN) 2 ][NTf 2 ], whose gas permeation and thermophysical properties were previously determined by us. 29 Both these mixtures have IL anions that show structural similarities to the [TFSAM] À anion (Fig. 3) Table 5, while their gas permeability, diffusivity and solubility values are depicted in Fig. 4(a)-(c), respectively.     (Fig. 5).

Materials
Lithium bis(pentafluoroethylsulfonyl)imide (Li(CF 3 CF 2 SO 2 ) 2 N, LiBETI, 98%, Chameleon Reagent) and lithium nonafluoro-1butanesulfonate (LiC 4 F 9 SO 3 , 4 95%, TCI Chemicals) were used without purification. Reagent-grade dichloromethane, acetonitrile, hexane and ethyl acetate were obtained from Aldrich or Merck and were dried by vacuum distillation over P 2 O 5 . N-Methylimidazole (98%, Aldrich) and bromoethane (98%, Acros) were distilled under an inert atmosphere over CaH 2 .    15,17,22,24,29,30,32,53  with dichloromethane (4 Â 40 mL). The combined CH 2 Cl 2 solution was washed with a small amount of water and dried over anhydrous MgSO 4 . The magnesium sulfate was filtered off and dichloromethane was stripped off under reduced pressure. The product was obtained as slightly yellow transparent fluid oil, which was finally dried at 323 K and 100 Pa for 12 h using a special flask filled with P 2 O 5 and introduced into the vacuum line. Yield: 5.61 g (68%); anal. calcd for C 10  , was prepared using an analytical high precision balance with an uncertainty of AE10 À5 g by syringing known masses of the IL components into a glass vial. Good mixing was ensured by magnetic stirring for 30 min at 298 K. Then, the IL mixture was dried at roughly 1 Pa and 318 K for at least 4 days immediately prior to use. The water contents of all IL samples were determined by Karl Fischer titration using a 831 KF Coulometer (Metrohm).
Density and viscosity determination. The density and viscosity measurements of the pure ILs and the [C 2 mim][SCN][OTf] IL mixture were performed at 293 K and atmospheric pressure using an SVM 3000 Anton Paar rotational Stabinger viscometerdensimeter, where the standard uncertainty for the temperature was 0.02 K. The repeatability of density and dynamic viscosity of this equipment was 0.0005 g cm À3 and 0.35%, respectively. Measurements of each sample were performed in triplicate to ensure accuracy and the reported results are average values. The highest relative standard uncertainty registered for the density and dynamic viscosity measurements was 1 Â 10 À4 and 0.03, respectively.
Gas permeation measurements. Porous hydrophobic poly-(vinylidene fluoride) (PVDF) membranes supplied by Millipore Corporation (USA), with a pore size of 0. 22 29 Ideal gas permeabilities and diffusivities through the prepared SILMs were measured using a time-lag apparatus. 36 First, each SILM was degassed under vacuum inside the permeation cell for 12 h. Then, CO 2 and N 2 permeation experiments were carried out at 293 K with a trans-membrane pressure differential of 100 kPa. All the permeation data were measured at least in triplicate on a single SILM sample. The highest relative standard uncertainty registered for gas permeability measurements was 0.03. The permeation cell and lines were evacuated until the pressure was below 0.1 kPa before each run. No residual IL was found inside the permeation cell at the end of the experiments. The thickness of the SILMs was assumed to be equivalent to the membrane filter thickness.
Gas transport through the prepared SILMs was assumed to follow a solution-diffusion mass transfer mechanism, 37 where the permeability (P) is related to diffusivity (D) and solubility (S) as follows: The permeate flux of each gas ( J i ) was determined experimentally using eqn (2), 38 where V p is the permeate volume, Dp d is the variation of downstream pressure, A is the effective membrane surface area, t is the experimental time, R is the gas constant and T is the temperature.
Ideal gas permeability (P i ) was then determined from the steady-state gas flux ( J i ), the membrane thickness (c) and the trans-membrane pressure difference (Dp i ), as shown in eqn (3). 38 Gas diffusivity (D i ) was determined according to eqn (4). The time-lag parameter (y) was calculated by extrapolating the slope of the linear portion of the p d vs. t curve back to the time axis, where the intercept was equal to y. 39 After P i and D i were known, the gas solubility (S i ) was calculated using the relationship shown in eqn (1). The ideal permeability selectivity (or permselectivity), a i/j , was obtained by dividing the permeability of the more permeable species i to the permeability of the less permeable species j. The permselectivity can also be expressed as the product of the diffusivity selectivity and the solubility selectivity:

Conclusions
In this work, ILs containing a common cation ( IL, not only in terms of thermophysical properties, but also regarding gas transport and CO 2 /N 2 separation performance.

Conflicts of interest
There are no conflicts to declare.