Aqueous Biphasic Systems Using Chiral Ionic Liquids for the Enantioseparation of Mandelic Acid Enantiomers

ABSTRACT This work aims at extending the applicability of chiral aqueous biphasic systems (ABS) to enantioseparations by using chiral ionic liquids (CILs) simultaneously as phase forming agents and chiral selectors. After determining the ternary phase diagrams of ABS composed of CILs and salts, these were used to ascertain the CIL structure on the ABS aptitude to separate mandelic acid enantiomers. Representative CIL-based ABS were further employed in optimization studies, where the mandelic acid content, temperature, tie-line length (TLL), salt and phases weight ratio were studied. The influence of these parameters is shown to be highly dependent on the CIL-based ABS, however the results here reported suggest that the key driving the enantioseparation in these ABS is a combination of the enantiorecognition ability of a given CIL with the solubility of mandelic acid in the corresponding CIL-rich phase.


Introduction
In spite of having very similar physical and chemical properties (except for their optical rotation), the optical isomers of a molecule can be discriminated by the human body. This is crucial in drug development, since the pharmacological activity usually results from only one of the enantiomers (the eutomer), while the other (the distomer) may be inert, less potent or even toxic. [1] During the past years, the pharmaceutical industry has faced pressures from the regulatory bodies regarding the commercialization of chiral drugs, so that a shift from racemic to enantiopure drugs is demanded. [2] The production of enantiopure drugs remains a major challenge for the pharma industry due to the limited approaches available and their high cost. [3] The preferable and most powerful pathway is the asymmetric synthesis, where the direct synthesis of the pure isomer is carried. [4] Asymmetric synthesis usually entails laborious and prolonged development processes, with uncertain outcome. Moreover, it often requires the use of expensive enantiopure raw materials or specific enantioselective catalysts. [3,5] The synthesis of racemates followed by their chiral resolution is a simpler, more flexible and cheaper alternative to asymmetric synthesis. [6,7] Crystallization and chromatography are the most used techniques for enantioseparation, due to their simplicity of operation. [6] Still, the former is limited by the number of racemic mixtures able to form conglomerates (estimated values from 5 to 10%), the excessive solids handling and the need for additional steps of enantiomeric of us showcase the interest of their implementation in chiral ABS. [46] Their enantioseparation aptitude was further evaluated and optimized using mandelic acid, a key precursor in chiral pharmaceuticals manufacturing, as model chiral compound.
For the HPLC-DAD analysis of the mandelic acid enantiomers, copper (II) sulphate pentahydrate, CuSO 4 · 5H 2 O (purity > 98%), L-phenylalanine, L-Phe (purity > 98%), purchased from AnalaR and Alfa Aesar, respectively, and methanol (HPLC grade), acquired from Fisher Chemical, were used for the mobile phase. Ammonia solution at 25% was obtained from Chem-Lab. Ultra-pure water (double distilled and then treated with a Milli-Q plus 185 water purification apparatus) was used for the HPLC analysis. Syringe filters (0.45 μm) and regenerated cellulose membrane filters (0.45 μm), acquired at Specanalitica and Sartorius, respectively, were used during filtration steps.
Synthesis of the chiral ionic liquids based on quinine, L-proline, and L-valine The six CILs here used were synthesized in our laboratory according to well-established protocols. [46] Briefly, an alkylation reaction between dimethyl sulfate and quinine yielding [C 1 Qui][C 1 SO 4 ] was performed, L-valine-based CILs were obtained in a three step synthesis entailing reduction of L-valine, Eschweiler-Clark reaction and N-alkylation and L-proline-based CILs were synthesized by alkylation reactions between L-proline and iodomethane, bromoethane, or dimethyl sulfate affording [C 1 C 1 C 1 Pro]I, [C 2 C 2 C 2 Pro]Br or [C 1 C 1 C 1 Pro][C 1 SO 4 ], respectively. Relevant features of CILs, namely melting temperature (T m ,°C), decomposition temperature (T d ,°C) and specific rotations ( α ½ 25 D ) can be consulted in Supplemental Material, Table S1. [46] Determination of the phase diagrams and tie-lines The ternary phase diagrams of the ABS composed of CILs and K 3 PO 4 , K 2 CO 3 , or K 2 HPO 4 were determined through the cloud point titration method at 25 (±1)°C. [41] To aqueous solutions containing ca. 6-70 wt% of CILs, the alternate drop-wise addition of an aqueous solution of salt at ca. 40-50 wt% and of pure water was performed under constant stirring. The repetition of this procedure allows, by turns, entering the biphasic region (turbid solution) and reaching the monophasic region (clear solution), respectively. By weight quantification (±10 −4 g) after the addition of each solution, the ternary systems compositions of the phase diagram were determined. The experimental binodal curves were fitted by Equation 1 [47] CIL where [CIL] and [salt] are the CIL and salt weight fraction percentages, respectively, while A, B, and C correspond to the fitting parameters. The tie-lines were gravimetrically determined, as originally proposed by Merchuk et al. [47] A ternary mixture composition formed by CIL + salt + water located at the biphasic region was prepared within ±10 −4 g, vigorously stirred and left to equilibrate at 25 (±1)°C for at least 12 h. Both phases were then separated and weighed. The lever-arm rule by the relationship between the top CIL-rich phase and the overall system weights allowed calculating each tie-line. Detailed guidelines on the tie-lines determination can be found elsewhere. [41] Separation of mandelic acid enantiomers using ABS Mixture points localized in the biphasic region of the phase diagrams were selected to conduct studies on racemic mandelic acid enantioseparation. The systems were gravimetrically prepared (within ±10 −4 g) by adding the correct amounts of CIL, salt and water along with equal amounts of two aqueous solutions of R-mandelic acid and S-mandelic acid both prepared at the same concentrations (viz. 10, 50, or 100 mg.mL −1 ) to yield the desired final content in the ABS. Throughout this work, the evaluation of distinct conditions was carried out: CIL's structure, enantiomers content, temperature, TLLs, salt, and mixture points along the same TL. The overall mixture compositions and conditions are detailed in Supplemental Material, Table S2. The CILs were placed in contact with the mandelic acid enantiomers in aqueous solution for at least 12 h under constant stirring (300 rpm) at the desired temperature, to promote specific interactions between the CIL and the target enantiomers, as recommended elsewhere. [44] The salt was added after such a period to induce liquid-liquid demixing. To this a period of equilibration of at least 12 h under the desired temperature followed, to guarantee complete separation of the phases and partition of the enantiomers among phases. The phases, the top being CIL-rich and the bottom being salt-rich, were then separated and weighed (within ±10 −4 g). CIL-rich phases were submitted to HPLC-DAD analysis for mandelic acid enantiomers quantification. In order to estimate the average extraction/enantioseparation parameters and the corresponding standard deviations, triplicates were performed. The percentage extraction efficiencies of R and S-mandelic acid (EE R-MA and EE S-MA , %) were separately determined according to the Equation 2: where m CIL R=SÀMA is the mass of R or S-mandelic acid present in the CIL-rich phase and m 0 R=SÀMA is the mass of R or S-mandelic acid originally added to the ABS.
The enantiomeric excess (e.e., %) present in the CIL-rich phase was calculated in accordance to Equation 3: in which m CIL SÀMA and m CIL RÀMA are the mass of S and R-mandelic acid present in the CIL-rich phase, respectively.

Mandelic acid enantiomers quantification
Mandelic acid enantiomers were quantified by HPLC-DAD using an analytical method adapted from Yue et al. with modifications. [48] The liquid chromatograph HPLC Elite LaChrom (VWR Hitachi) used for this purpose was equipped with a diode array detector (DAD) l-2455, column oven l-2300, auto-sampler l-2200 and pump l-2130. A C 18 reversed-phase analytical column (LiChrospher 100 RP-18, 5 µm, 250 mm × 4 mm i.d.) linked to a guard column (5 µm, 4 mm × 4 mm) with the same stationary phase was used. The column oven and autosampler were operated at controlled temperature of 22°C and 25°C, respectively. The mobile phase was made up of water:methanol [85:15 (v/v)], 2 mM L-phenylalanine and 1 mM CuSO 4 at pH = 4.00 (±0.02), adjusted by adding an ammonia aqueous solution at 5 wt%. The separation was carried out using isocratic elution at a flow rate of 0.8 mL.min −1 and the injection volume was 20 µL. DAD was set to measure the spectrum from 200 to 600 nm, with a specific wavelength of 270 nm being used for R-mandelic acid and S-mandelic acid quantification. Calibration curves were previously determined using stock solutions prepared in water:methanol [85:15 (v/v)] at concentrations of 10-500 µg.mL −1 of each enantiomer. The R enantiomer elutes first, at a retention time of around 11.7 min, followed by S eluting at approximately 13.2 min. The LOD and LOQ were, respectively, 5 µg.mL −1 and 10 µg.mL −1 for both enantiomers. Intra-and inter-day precisions were 0.27-3.29 % and 1.

Ternary phase diagrams and tie-lines
The knowledge of the CIL-based ABS phase diagrams is essential for the development of enantioseparations. To accomplish this, the ternary phase diagrams composed of five CILs, [C 1 C 1 C 1 Val]I, [C 1 C 1 C 1 Val][C 1 SO 4 ], [C 2 C 2 C 2 Pro]Br, [C 1 C 1 C 1 Pro]I, and [C 1 Qui][C 1 SO 4 ], and K 3 PO 4 , a strong salting-out agent, [41] were measured at 25 (±1)°C. [C 1 C 1 C 1 Pro][C 1 SO 4 ] was not able to form ABS with K 3 PO 4 . Two additional salts, K 2 CO 3 and K 2 HPO 4 , were paired with [C 2 C 2 C 2 Pro]Br to evaluate the role of salt type upon ABS formation.
The ternary phase diagrams are shown in Figures 1 and 2 in weight fraction. All detailed experimental data related (ternary phase diagrams weight fraction compositions, Equation 1 regression parameters and TL information) are provided as Supplemental Material (Tables S3-S11). The ternary phase diagrams determined in this study provide information on the CILs and salt role upon ABS formation (Figures 1 and 2). The biphasic zone is placed above the binodal curve meaning that the broader this is the more prone is the CIL to form ABS.
As observed in Figure 1, the CILs' ability to form ABS with K 3 PO 4 can be ranked as follows (at fixed CIL weight fraction composition of 10 wt%): SO 4 ]) and anion role ([C 1 C 1 C 1 Val]I vs. [C 1 C 1 C 1 Val][C 1 SO 4 ]) on the ABS formation. The cation effect is driven by the hydrophobicityhydrophilicity of the cation, where the order [C 1 Qui] + > [C 1 C 1 C 1 Val] + > [C 1 C 1 C 1 Pro] + directly correlates with the octanol-water partition coefficients of their precursors (log K o/w of 3.44, −0.08 and −0.10 for quinine, valinol and proline methyl ester, respectively [49] ). In close agreement with previous studies, [41] the more hydrophobic the CIL the higher is its aptitude to form ABS. It should be highlighted that, although valinol and proline methyl ester possess similar log K o/w , [C 1 C 1 C 1 Pro][C 1 SO 4 ] failed to induce phase separation in presence of K 3 PO 4 . This can be attributed to the higher hydrophobicity of the cation when compared to the parent compound, due to alkylation: while the addition of three methyl groups is done to valinol, only two methyl groups are added to proline methyl ester (cf. Figure S1 in Supplemental Material).  In general, the ability of an IL anion to create ABS is related with the decrease in their hydrogenbond accepting ability (β). [50,51] The aforementioned rank places I − as a better two-phase formation inducer than [C 1 SO 4 ] − , in good agreement with their relative position in the scale of hydrogen bond basicity of ILs proposed by Cláudio et al. [52] Although [C 2 C 2 C 2 Pro]Br vs. [C 1 C 1 C 1 Pro]I do not allow direct comparisons, it should be noted that Br − is a stronger hydrogen-bond acceptor than I −, [52] being thus expected to yield smaller biphasic regions. Since the opposite is observed, the effect of longer alkyl chains in [C 2 C 2 C 2 Pro]Br may overwhelm that of the anion nature (cf. Figure S1 in Supplemental Material). Figure 2 shows the ability of three salts to promote phase separation, which can be rated as follows (at fixed CIL weight fraction composition of 10 wt%): K 3 PO 4 ≈ K 2 HPO 4 > K 2 CO 3 . This ranking follows the Hofmeister series as previously established in the literature for ABS composed of ILs and salts. [53,54] CILs-based ABS: evaluating the impact of CILs structures in enantioseparation An initial screening comprising the five CIL-based ABS developed (Supplemental Material, Table S2) was done in order to understand the role of the cation/anion structures on the enantioseparation of R-and S-mandelic acid structures. The extraction efficiencies (EE R-MA and EE S-MA ) as well as enantiomeric excesses (e.e.) obtained are depicted in Figure 3 and detailed in Supplemental Material (Table S12). The EE R/S-MA values reveal a similar partition of mandelic acid between the two phases, or a preferential partition of mandelic acid for the salt-rich phase. Under the conditions adopted (initial biphasic mixture compositions, temperature and mandelic acid content - Table S2 in Supplemental Material), all CILs exhibited preferable chiral recognition for the S-mandelic acid over the R enantiomer, with modest e.e. (1.61 ± 0.92 % to 17.37 ± 1.92 %). Moreover, valine and prolinebased CILs seem to be more promising than the quinine-based CIL.
In general, electrostatic interactions between mandelic acid and the CIL cations play an important role in the "three-point model"-based enantiorecognition process, since mandelic acid is deprotonated (pKa 1 = 3.75 and pKa 2 = 13.57, Figure S2 in Supplemental Material) [49] under the alkaline pH induced by K 3 PO 4 . Given the chemical structures of the CILs and mandelic acid (cf. Figure S1 in Supplemental Material) and the results found, it seems that other interactions can act in the  acid enantiomeric discrimination by valine and proline-based CILs. Moreover, and contrarily to what is reported in literature when aromatic chiral recognition agents and solutes are present, [55][56][57] in this specific case, π-π stacking does not seem to contribute to the enantioseparation of racemic mandelic acid, since [C 1 Qui][C 1 SO 4 ] (the only CIL bearing aromatic rings) yielded the lowest e.e.
It has been previously shown that factors other than the CIL structure may affect the enantioseparation ability and that such impact is dependent on the ABS phase formers. [44,45] Bearing this in mind, optimization studies will be carried for two CILs, the least and the most performant ones, aiming to gain further insight on the phenomena governing enantioseparations in these CIL-based ABS.
[C 1 Qui][C 1 SO 4 ]-based ABS: evaluating the impact of mandelic acid content, temperature and TLL in enantioseparation SO 4 ], here identified as the weakest enantiorecognition agent, was used for further optimization to understand whether its enantioseparation ability could be improved by modifying the operational conditions. Mandelic acid content, TLL (varied by changing the mixture point) and temperature were evaluated, as presented in Supplemental Material (Table S2). The results obtained are depicted in Figure 4 and detailed in Supplemental Material (Table S12) and suggest that, although having distinct effects on the extraction and separation of mandelic acid enantiomers, the parameters evaluated lead to better enantioseparations [from nearly 0 to a maximum e.e. of 7.88 ± 0.70% obtained with [C 1 Qui][C 1 SO 4 ]-based ABS at 15 (±1)°C].
While a decline of about 15 % on EE R/S-MA is observed, an increase of ca. 3.8 times in e.e. occurs by raising mandelic acid content in the system (Figure 4a). So, the enantioseparation seems to be ruled by a compromise between the solubility of mandelic acid in the CIL-rich phase and the more favorable interactions between the CIL and S-mandelic acid. The TLL effect is marginal under the conditions studied in this work (Figure 4b). Temperature, in turn, has a significant impact in the EE R/S-MA , likely as a result of the increasing solubility of mandelic acid in the CIL-rich phase at 45°C (Figure 4c). Furthermore, Figure 4c provides support of a trade-off between EE R/S-MA and e.e. parameters. e.e. is higher for lower temperatures (e.g. 15°C) where the molecular motions are slower, thus favoring the "S-mandelic acid-[C 1 Qui][C 1 SO 4 ]" interactions. The same behavior was previously observed in a work on the chiral separation of phenylalanine enantiomers with CIL-based ABS. [44] [C 2 C 2 C 2 Pro]Br-based ABS: evaluating the impact of mandelic acid content, temperature, TLL, salt and phases' weight ratio in enantioseparation Since the best enantioseparations were achieved with [C 2 C 2 C 2 Pro]Br and [C 1 C 1 C 1 Val][C 1 SO 4 ], the role of the operational conditions on the performance of these ABS was further studied. The [C 2 C 2 C 2 Pro]Br was used as a model chiral selector to evaluate the influence of mandelic acid content, temperature, TLL, salt and the phases' weight ratio, as specified in Table S2 in Supplemental Material. Figure 5 overviews the results obtained with ABS composed of [C 2 C 2 C 2 Pro]Br and K 3 PO 4 and reveals a complex scenario regarding the impact of distinct operational conditions upon mandelic acid enantioseparation (detailed data provided as Supplemental Material, Table S12). When mandelic acid content is increased from 0.17 wt% to 1.7 wt% in the ABS, a 37% drop in EE R/S-MA is observed. The maximum enantioseparation is achieved at intermediate mandelic acid concentration (e.e. = 17.37 ± 1.92 %) - Figure 5a -this seeming to be the optimal mandelic acid/CIL compositions to favor "[C 2 C 2 C 2 Pro]Br-S-mandelic acid" interactions. A shift towards R-enantiomer higher partitions to the CIL-rich phase seems to occur (negative e.e. values are merely indicative of a R-mandelic acid enriched mixture) at cost of e.e. at the lowest (e.e. of 4.91 ± 1.34 %) and highest (e. e. of 6.40 ± 2.92 %) concentrations ascertained. Although this behavior may seem odd, similar results are found in literature, showing that low to intermediate concentrations are enantioseparation boosters. [20,24,37,39,58] It should however be emphasized that the type of system, chiral selector, operational conditions and racemic compound may lead to different dependencies. [26,45,48,59] The lowest concentration of mandelic acid investigated seems to restrict the occurrence of "[C 2 C 2 C 2 Pro] Br-S-mandelic acid" interactions, thus limiting enantioseparation. [45] The TLL influences both the extraction and enantioseparation performance of [C 2 C 2 C 2 Pro] Br-based ABS, as shown in Figure 5b. Mixture points yielding longer TLLs, i.e. higher concentrations of both CIL and K 3 PO 4 in both top and bottom phases, respectively, promote the extraction of mandelic acid toward the CIL-rich phase. This may be explained in the light of hydrophobic interactions occurring between the mandelic acid and the [C 2 C 2 C 2 Pro]Br in the top phase. However, enantioseparations are less efficient under such conditions, indicating that the relative amounts of CIL/salt in the top phase and mandelic acid in the system are crucial to design efficient CIL-based ABS. Contrarily to what was observed for [C 1 Qui][C 1 SO 4 ], the temperature does not significantly affect EE R/S-MA or e.e. (Figure 5c). Therefore, both the solubility of mandelic acid in the [C 2 C 2 C 2 Pro]Br-rich phase and the specific interactions taking place between the CIL and the S-enantiomer seem to be unaffected on the temperature range studied.
The action of two additional salts (K 2 HPO 4 and K 2 CO 3 ) on the enantioseparation ability of [C 2 C 2 C 2 Pro]Br-based ABS was evaluated. It has been widely shown that the salt used has an important influence in the extraction and separation of a solute using ABS, [41] in particular if enantiomeric separations are targeted. [45] Figure 6 shows that also here the enantioseparation is dependent on the salt used. Under the conditions assessed (see Table S2 in Supplemental Material), K 3 PO 4 ranks first (e.e. = 17.37 ± 1.92 %), followed by K 2 CO 3 (e.e. = 5.79 ± 0.12 %), while K 2 HPO 4 completely failed to separate mandelic acid enantiomers (e.e. = 0.82 ± 0.18 %) (detailed data provided as Supplemental Material, Table S12). These salts create an alkaline pH in ABS (pH CIL K 2 HPO 4 = 8.32 ± 0.02, pH CIL K 2 CO 3 = 11.68 ± 0.02 and pH CIL K 3 PO 4 = 12.99 ± 0.02), so that mandelic acid is deprotonated (pKa 1 Figure 5. Impact of mandelic acid content (a), TLL (b), and temperature (c) on the extraction efficiencies (EE R-MA , yellow bars and EE S-MA , green bars) and enantiomeric excesses (e.e., diamonds) obtained with ABS composed of [C 2 C 2 C 2 Pro]Br and K 3 PO 4. = 3.75 and pKa 2 = 13.57). [49] In presence of K 3 PO 4 mandelic acid deprotonates displaying a distribution of approximately 79% and 21% for mono-and divalent ions, respectively. [49] The amount of divalent ions further decreases with decreasing pH, down to around 1.34 % (K 2 CO 3 ) and completely vanishes at pH 9.2 ( Figure S2 in Supplemental Material). [49] Better recognition ability seems to be accomplished when divalent mandelic acid ions are present in solution, in agreement with previous insights gathered in the chiral separation of mandelic acid in micellar systems containing copper-β-cyclodextrin-complexes as chiral selector. [21] Alongside, mandelic acid partitions majorly to the CIL-rich phase (more hydrophobic) when K 2 HPO 4 is used (EE R−MA = 81.90 ± 5.59 % and EE S−MA = 81.63 ± 5.08 %), while an almost equivalent distribution of mandelic acid between the two phases is observed for K 3 PO 4 (EE R−MA = 44.62 ± 2.50 % and EE S−MA = 61.58 ± 3.96 %) and K 2 CO 3 (EE R−MA = 55.13 ± 0.03 % and EE S−MA = 62.61 ± 2.31 %). Divalent mandelic acid is more polar than its monovalent congener, what may explain this extraction profile. It should be highlighted that the enantioseparation in these ABS may be additionally influenced by specific interactions promoted by the salt ions or distinct solubility profiles exhibited by mandelic acid in the corresponding ABS phases. The body of results hitherto reported provides some evidence that the performance of CIL-based ABS in the enantioseparation of mandelic acid may be improved by manipulating the solubility of mandelic acid in the CIL-rich phase. At first glance, the preferential partition of the S-enantiomer to the [C 2 C 2 C 2 Pro]Br-rich phase is enhanced by constraining the mandelic acid solubility. To confirm this hypothesis, partition studies were carried out along the same TL for two distinct ABS (K 3 PO 4and K 2 HPO 4 -based), meaning that different weight ratios were used while the phases compositions were kept constant (Figure 7 and Table S12 from Supplemental Material). As shown in Figure 7a for the K 3 PO 4 -based ABS, almost complete partition of mandelic acid towards the [C 2 C 2 C 2 Pro]Br-rich phase occurs for systems possessing larger CIL-rich phases (EE R-MA = 87.17 ± 7.70% -94.51 ± 5.94% and EE S-MA = 86.63 ± 8.15% -92.81 ± 5.79%). When the phases weight ratio is decreased, the CILrich phase becomes saturated, as revealed by the decreasing of mandelic acid partition (EE R−MA = 37.32 ± 2.74% and EE S−MA = 48.10 ± 2.62). In previous studies, distinct solubility profiles of the phenylalanine enantiomers in CILs phases were also observed. [44,60] A completely distinct pattern was observed by replacing K 3 PO 4 by K 2 HPO 4 , where neither EE R/S-MA nor e.e. significantly vary along the tie line (Figure 7b), what must be related with the effect of the pH upon the charge of the mandelic acid as discussed above. Critical assessment of chiral ABS application in the enantioseparation of mandelic acid As aforementioned, the satisfactory compromise between cost-effectiveness, broad applicability, easy operation and scale-up [18,19] has placed ABS in the spotlight. Four works using chiral ABS for chiral separation of mandelic acid enantiomers were previously reported. While most are based on the introduction of an extra chiral agent to the ABS, [20,21,36] one proposes the use of chiral phase formers. [37] The phase formers are focused on polymer-salt, [20] alcohol-salt, [36] and micellar systems [21] incorporating β-cyclodextrin derivatives as well as on polymer-polymer introducing a chiral compound acting as both phase former and chiral agent. [37] The enantioseparation ability is commonly evaluated by determining e.e. (Eq. 3) and/or enantioselectivity (α, the ratio between the partition coefficients of mandelic acid enantiomers). Overall, the enantioseparation abilities hitherto reported are highly dependent on the type of ABS and conditions adopted: ethanol-(NH 4 ) 2 SO 4 Figure 7. Extraction efficiencies (EE R-MA , yellow bars and EE S-MA , green bars) and enantiomeric excesses (e.e., diamonds) obtained with ABS composed of [C 2 C 2 C 2 Pro]Br + K 3 PO 4 (a) and [C 2 C 2 C 2 Pro]Br + K 2 HPO 4 (b) at 25 (±1)°C and at distinct initial compositions along the same TL: binodal curve (dashed line), TL (solid line), and initial mixture composition (triangles). + sulfonated-β-cyclodextrin with α = 1.69 and e.e. = 16 % [36] ; poly(ethylene) glycol-(NH 4 ) 2 SO 4 + βcyclodextrin with α = 2.46 and e.e. = 42 % [20] ; and Triton X-114 + copper-β-cyclodextrin complex with e.e. = 68 %. [21] Yet, the implementation of chiral phase formers yields less efficient enantioseparations (α = 1.27). [37] With this strategy, also employed in this work, the technological simplicity, target enantiomer polishing and ABS constituents recycling and reuse are enhanced. [37] Moreover, with the ability to overcome technological limitations of polymeric ABS [37] (e.g. viscosity of the phases, limited hydrophobicity-hydrophilicity range and difficulty to find pairs of polymers able to form ABS), the CIL-based ABS here developed are somehow more efficient (maximum e.e. of 17.37 ± 1.92 %). In addition, the possibility of using chiral cationic and/or anionic groups [43] affords the opportunity of designing specific CILs structures able to interact more specifically with some particular enantiomeric structures, avoiding the use of complex extraction systems, since these can act simultaneously as chiral selectors and solvents. Most purification systems reported in literature lack specificity towards the enantiomers structures, which leads to low specificity on their separations. Moreover, ILs can be combined with a large range of phase formers to generate ABS, [41] providing an extra degree of tailoring the enantioseparations. Such features offer unique opportunities to plan ABS to match specific enantioseparations. [41] Finally, some CILs show great promise in the field as they can be synthesized by simple, practical and benign routes, [43] using natural precursors [43] and available in large scale. [5] However, given the limited understanding on their enantiorecognition mechanisms, the design of efficient chiral ABS platforms still relies on case-bycase studies and the broad applicability of ABS for these separations remains challenging.

Conclusions
On the search for alternative enantioseparation techniques, this work proposes the implementation of CILs as chiral phase formers in ABS to resolve racemic mandelic acid. The ternary phase diagrams of ABS constituted by five CILs bearing chirality in the cation and salts were ascertained under ambient conditions, with the hydrophobicity of the CIL cation and the salting out aptitude of the salt dictating the two-phase separation aptitude. After an initial screening where all five CILs were ranked according to their relative ability to separate mandelic acid enantiomers, a maximum e.e. of 17.37 ± 1.92 % was achieved with [C 2 C 2 C 2 Pro]Br. The most and least promising CILs were object of detailed optimization, comprising the parameters mandelic acid content, temperature, TLL, salt and phases' weight ratio. With the CIL structure playing a central role, all remaining conditions were shown to influence the enantioseparation. Such impacts are highly dependent on the ABS nature: while temperature was the main factor improving the enantioseparation ability of [C 1 Qui][C 1 SO 4 ]based ABS, [C 2 C 2 C 2 Pro]Br-based ABS was mainly influenced by the salt used and the phases weight ratio. Based on the optimization results it seems that the saturation of the CIL-rich phase rules the enantioseparation: S-mandelic acid (the enantiomer with higher affinity for this set of CILs) remains in the CIL-rich phase, while R-mandelic acid partitions to the K 3 PO 4 -rich phase.