Simulation and measurement of water-induced liquid-liquid phase separation of imidazolium ionic liquid mixtures

Ionic liquids (ILs) have been intensely investigated in recent years due to their many favorable properties.¹ One of the most attractive features of ILs is that their physical properties can be adjusted by varying the identity and structure of the cation and anion. Recent work has shown that properties of ILs can be further fine-tuned by mixing ILs.^2–6 While in many cases the ILs are miscible, mixtures containing hydrophilic and hydrophobic ions have been shown to give rise to two distinct immiscible phases,⁷ which has implications for separation processes.^8,9 Miscibility of these ILs has been found to be temperature dependent such that a biphasic system is present at low temperature, but at high temperature the ILs are completely miscible.⁷ ILs have also been shown to form two aqueous-rich liquid phases when mixed with water and a “salting out” or kosmotropic inorganic salt such as K₃PO₄.¹⁰ These aqueous biphasic systems (ABSs) have been intensely studied for over a decade,¹¹ and have been found to be useful for liquid-liquid extraction of proteins, drug molecules, and dyes.^12,13 ABSs composed of only two ILs and water were first reported by the Ohno group,¹⁴ but less is known about these systems.

Herein, we demonstrate that miscible mixtures of two ionic liquids having a common cation (1-ethyl-3-methylimidaziolium or [C₂C₁im]) and different anions of varying hydrophilicity (chloride [Cl], acetate [OAc], and bis(trifluoromethylsulfonyl)amide [TFSI]) can be induced to separate into two liquid phases by the simple addition of water. One phase is rich in water and contains the cation and a predominance of the more hydrophilic anion, while the other phase is water-lean and contains the cation and a predominance of the more hydrophobic anion. We find that 1:1 molar mixtures of [Cl:TFSI] and [OAc:TFSI] exhibit this phase separation, while 1:1 molar mixtures containing [Cl:OAc] remain one phase at all water compositions.

This finding was first predicted by molecular dynamics simulations. Equimolar ionic liquid mixtures were modeled by placing 800 [C₂C₁im] cations in a simulation box along with 400 of one type of anion and 400 of another type of anion. Anion mixtures of [Cl:OAc], [Cl:TFSI], and [OAc:TFSI] were modeled at 293 K. Simulation details are provided in the supplementary material. When no water was present, each IL mixture was miscible, showing no signs of phase separation [Figs. 1(a) and 1(c)]. Upon the addition of 1200 water molecules (x = 0.60), the [OAc:TFSI] system began to show signs of phase separation, while the [OAc:Cl] system remained one phase. When more water was added (4000 molecules, x = 0.83), clear liquid-liquid phase separation was observed in the [OAc:TFSI] system [Fig. 1(b)], while the [OAc:Cl] system remained miscible [Fig. 1(d)]. The simulations also predicted that the [Cl:TFSI] system phase separates into two liquid phases upon the addition of water (see the supplementary material), with [Cl] going to the water-rich phase and [TFSI] going to a separate phase with less water.

Figure 1 clearly shows that the simulations predict that the cations and anions are uniformly dispersed throughout the liquid in the dry systems. When water is added to the [OAc:TFSI] system, the [OAc] anions segregate to the water-rich phase, while the more hydrophobic [TFSI] anions partition to the water-lean phase. To maintain charge neutrality, the [C₂C₁im] cations remain homogenously distributed between the two phases. For the mixture containing [OAc:Cl], both anions are hydrophilic and there is no preference for either of them to partition with water; water equally solvates both anions, and a single liquid phase is maintained at all water concentrations.

To test the simulation predictions experimentally, equimolar mixtures of these three ionic liquid systems were placed in borosilicate glass tubes and small amounts of water were slowly added at room temperature until a meniscus appeared. Each liquid phase was analyzed to determine the concentration of individual components.

At initial and low water concentrations, the 1:1 mole ratio [C₂C₁im][TFSI]:[C₂C₁im][OAc] mixture remained homogeneous as shown in Figs. 2(a) and 2(b). As water was initially added to the system, cloudiness was observed at points of contact between the ILs and water, and a noticeable amount of heat was released, resulting in a temperature increase which we estimate to be less than 10 K. This heat can be attributed to the exothermic process of hydrogen bonding between water and the acetate ions.^15,16 Gradual additions of water to the equimolar IL solution resulted in a liquid phase split, observed at 54 ± 2 mol. % H₂O, as shown in Fig. 2(c).

Each liquid phase was weighed, sampled, and analyzed with a quantitative NMR approach. The applied NMR technique determines species concentrations by calculating relative peak integral ratios and has been previously used in the literature¹⁷ to determine unknown concentrations of bi-phasic solutions. In our study, this method was successfully applied using ¹H NMR to estimate the phase concentrations of the [C₂C₁im][TFSI], [C₂C₁im][OAc], and H₂O mixture, as [C₂C₁im], [OAc], and H₂O have clear proton peaks and integral representations. The concentration of [TFSI] was determined by difference as explained in the supplementary material. The analysis revealed that the top liquid phase was water-rich and the bottom liquid phase was rich in [C₂C₁im][TFSI], as shown in Fig. 2(d) and Table I. The study also determined that the top phase had a higher concentration of [C₂C₁im][OAc] as reported in Fig. 2(d). These results are reasonable as the [OAc] anion has a higher affinity for water than the [TFSI] anion, and the density of [C₂C₁im][TFSI] is greater than that of water and [C₂C₁im][OAc]. Therefore, the majority of [C₂C₁im][TFSI] was present in the bottom liquid phase. It was assumed that the [C₂C₁im] cations were distributed among both phases, according to the concentration of each anion, to maintain neutrality in each phase. Water content in each phase was also analyzed by Karl Fischer (KF) volumetric titration, and the results were within 1% of the water mass concentrations obtained using ¹H NMR, as shown in Fig. 2(d). Furthermore, the NMR results were used to estimate the overall mass phase partition, which resulted in the top and bottom phases containing 25.24 ± 2.47 wt. % and 74.76 ± 2.47 wt. % of the total mass, respectively. These results were within 1% of the actual measured phase partitions (26.16 ± 0.15 wt. % in the top phase and 73.84 ± 0.15 wt. % in the bottom phase) reported in Fig. 2(d).

In addition, the liquid-liquid equilibrium temperature dependence was observed. The two-phase mixture was heated in intervals of 5 K beginning at 298.15 ± 0.30 K and sequentially agitated and allowed to settle to identify the condition at which the meniscus disappeared (as described in the supplementary material). At 303.11 ± 0.30 K, the mixture returned to a single-phase system.

The second system in this study was the equimolar [C₂C₁im][TFSI] and [C₂C₁im][Cl] mixture. The [C₂C₁im][Cl] is present in the form of green-colored crystals at 295 K. Color in ionic liquids can originate from the presence of trace metal impurities. Measurements with inductively coupled plasma mass spectrometry (ICP-MS) determined the [C₂C₁im][Cl] contained 90-130 ppm Ni. The anhydrous salt nickel(II) chloride (NiCl₂) is known to be yellow and the more familiar hydrate NiCl·6H₂O is green. However, the trace levels of metal hydrate will have no effect on the liquid-liquid equilibria (LLE) measurements. When [C₂C₁im][TFSI] and [C₂C₁im][Cl] were initially combined, they remained in a heterogeneous solid-liquid solution as shown in Fig. 3(a), until the concentration of water was increased to about 20 mol. %. At this stage, all the [C₂C₁im][Cl] had dissolved and the mixture was a homogeneous green color. This color persisted as the system partitioned into a two-phase mixture at 39 ± 3 mol. % H₂O as shown in Fig. 3(c), indicating that [C₂C₁im][Cl] is present in both phases, though not necessarily at the same concentration.

The determination of the species concentration in each liquid phase required multiple analysis techniques. First, the mass of each layer was measured, and the volume was calculated as explained in the supplementary material. Quantitative ¹⁹F NMR analysis was applied to estimate the concentration of [C₂C₁im][TFSI] in each liquid layer. The NMR method was performed using an external fluorine standard in a coaxial tube to prevent the ionic liquid mixture and the standard solution from interacting. The volume effects in the NMR analysis were corrected as described by Henderson¹⁸ and are detailed in the supplementary material. Quantitative analysis by ¹H NMR was more difficult due to concentration effects and poor peak deconvolution. For this reason, the only method used to measure water concentration was Karl Fischer. To verify the mass balance, [C₂C₁im][Cl] content was measured using ion chromatography (IC) (see the supplementary material for more details).

The ¹⁹F NMR, IC, and KF results were used to estimate the mass percent of each phase to be 16.69 ± 1.47 wt. % in the top phase and 83.31 ± 1.47 wt. % in the bottom phase. The estimations were only about two percent different from the actual measured masses for each phase (14.93 ± 0.06 wt. % in the top phase and 85.07 ± 0.06 wt. % in the bottom phase); see Fig. 3(d). In addition, the results in Fig. 3(d) and Table II show that the top liquid phase was water rich and the bottom phase was rich in [C₂C₁im][TFSI]. Similar to the [C₂C₁im][TFSI]:[C₂C₁im][OAc]:H₂O system, it was expected that the bottom phase would be rich in [C₂C₁im][TFSI], and the top phase would be rich in water with a higher concentration of the hydrophilic ionic liquid. Temperature effects were also tested in the [Cl]:[TFSI] mixture, which was heated in 5 K intervals beginning at 298.15 ± 0.30 K and agitated until the two-phase meniscus disappeared at 323.11 ± 0.30 K.

Both room temperature systems [C₂C₁im][TFSI]: [C₂C₁im][OAc]:H₂O and [C₂C₁im][TFSI]:[C₂C₁im][Cl]: H₂O exhibited a liquid-liquid phase separation with the addition of water. The mole ratio of H₂O:[OAc] in the top phase was 2.87:1 and 2.29:1 in the bottom phase, while the mole ratio of H₂O:[Cl] was 1.58:1 in the top phase and 1.18:1 in the bottom phase. This behavior is consistent with the water-[OAc] and water-[Cl] interactions in the top phase and with the repulsive hydrophobic interactions of the [TFSI] anion with water in the bottom phase.

The third system investigated was the 1:1 mole ratio mixture of [C₂C₁im][Cl] and [C₂C₁im][OAc]. The solution was initially a heterogeneous solid-liquid mixture with no water added [Fig. 4(a)], similar to the [TFSI]:[Cl] system, as [C₂C₁im][Cl] is a solid at room temperature. When the system reached 43 ± 1 mol. % H₂O, the [C₂C₁im][Cl] dissolved and the solution became light green in color [Fig. 4(b)]. More water was added to the system, until the water concentration was 83.6 ± 0.1 mol. %. Heat was released as the solution was mixed, and the color changed to an amber-like tint, as shown in Fig. 4(c). No liquid-liquid separation was observed at room temperature. This phase behavior is consistent with the molecular dynamics simulations and is expected, as both anions are hydrophilic and have similar dimensions. Considering [OAc] and [Cl] have comparable van der Waals radii of 1.6 Ź⁹ and 1.75 Å,²⁰ respectively, it can be concluded that the anions share similar charge densities and are equally miscible with water.

Previous literature has shown that a color change in imidazolium based ionic liquids with [Cl] anions is due to the thermal decomposition of the ionic liquid.^21,22 However, the present study indicates that the color change occurs at room temperature when the H₂O to [C₂C₁im][Cl] ratio is increased from 1.5:1 to 10:1. To understand this color change, the liquid structure in these mixtures was studied by creating spatial distribution functions (SDFs) from the molecular dynamics simulations. The results for [OAc] and [Cl] around a central [C₂C₁im] are shown in Fig. 5. At 40 mol. % and 60 mol. % H₂O, [OAc] prefers to stay close to the C2 position of [C₂C₁im] and locates on both sides of the imidazolium ring. On the other hand, [Cl] locates close to the C4 and C5 positions, with less intensity on the C5 side due to repulsion from the cation ethyl group. At 83 mol. % H₂O, the solvation structure dramatically changes: [Cl] is now most likely to be found on top of the C2 carbon and has a slightly lower probability around C4. On the other hand, [OAc] is still close to C2 but is clearly pushed further to the sides of the imidazolium ring. It is very likely that the color change observed in experiments is related to these solvation structure changes when the H₂O mol. % was increased from 60% to 83%. Further studies are needed to verify this behavior.

In summary, evidence of a water-induced liquid-liquid phase separation of ionic liquid mixtures having a common [C₂C₁im] cation but pairs of anions with different degrees of hydrophilicity ([Cl]-[TFSI] and [OAc]-[TFSI]) was observed in molecular dynamics simulations. The simulations also predicted that an ionic liquid mixture containing the [C₂C₁im] cation with equimolar amounts of hydrophilic anions ([Cl] and [OAc]) would remain one phase at all water concentrations. All three of these predictions were subsequently confirmed experimentally. Quantitative concentration measurements were made for each phase using a combination of NMR, IC, and KF. In addition, a color change from pale green to amber was observed upon the addition of water to the [C₂C₁im][OAc]:[C₂C₁im][Cl] mixture. At such water concentrations, the simulations depict an abrupt change in the distribution of [Cl] ions about the [C₂C₁im] cation, and therefore suggest a possible reason for the color variation.

See supplementary material for the following: simulation details including computational approach and force fields; experimental details including materials, procedures, and error calculation.

We acknowledge Dr. David Minnick for providing directions on the mixture content analysis, Dr. Justin T. Douglas and Sarah Neuenswander for their assistance in running NMR samples, and Rasha Faraj for performing the ion chromatography analysis. Y.Z. and E.J.M. acknowledge the support by the U.S. Department of Energy, Basic Energy Science, Joint Center for Energy Storage Research under Contract No. DE-AC0206CH11357 and the Air Force Office of Scientific Research under Contract No. AFOSR FA9550-18-1-0321. Computational resources were provided by the Center for Research Computing (CRC) at the University of Notre Dame.