The ABS studied in this work were prepared using aqueous solutions of K₃PO₄ (≥98 wt% pure, Sigma) and aqueous solutions of individual ILs. The ILs studied were phosphonium-based ILs, namely triisobutyl(methyl)phosphonium tosylate ([P_i(444)1][Tos]) > 95 wt % pure, tributyl(methyl)-phosphonium methylsulphate ([P₄₄₄₁][MeSO₄]) > 99 wt % pure and tetrabutylphosphonium bromide ([P₄₄₄₄]Br) > 96 wt % pure. The studied phosphonium-based ILs were kindly provided by Cytec Industries, Inc. The selected ILs cover different cations and anions and merely represent a proof of principle of the versatility of ILs that can be used for forming ABS. All the ILs’ molecular structures and respective designations are shown in Figure 1.

For the validation of the experimental procedure adopted 1-butyl-3-methylimidazolium chloride ([C₄mim]Cl) (>99 wt % pure, Iolitec) was used. The partitioning solutes l-tryptophan, >99.0 wt % pure, and β-carotene, ≥97.0 wt % pure, were obtained from Fluka. Rhodamine 6G, >95.0 wt % pure, was acquired from Merck and caffeine, ≥99.5 wt % pure, was obtained from José M. Vaz Pereira, SA. The water used was ultra-pure water, double distilled, passed by a reverse osmosis system and further treated with a Milli-Q plus 185 water purification apparatus.

The phase diagrams were determined at 298 K (±1 K) by the cloud point titration method [25,26]. Aqueous solutions of K₃PO₄ at 40 wt % and aqueous solutions of the different hydrophilic ILs at 80 wt % were prepared and used for the phase diagrams determination. Repetitive drop-wise addition of the aqueous inorganic salt solution to each IL aqueous solution was carried out until the detection of a cloudy and further biphasic solution, followed by the drop-wise addition of ultra-pure water until the detection of a monophasic region (clear and limpid solution). Drop-wise additions were carried out under constant stirring. The ternary systems compositions were determined by the weight quantification of all components added within an uncertainty of ± 10⁻⁵ g. Further details on the experimental procedure can be found elsewhere [25,26]. The experimental procedure was validated by comparison of the phase diagram obtained for the [C₄mim]Cl and K₃PO₄ aqueous system against literature data [25,26], which showed them to be in close agreement.

Tie-lines (TLs) were determined by a gravimetric method originally proposed by Merchuck et al. [43]. For the determination of TLs, a mixture from the biphasic region was selected, vigorously agitated and allowed to reach the equilibrium, by the separation of both phases, for 12 h and at 298 K. After a careful separation step, both top and bottom phases were weighed. Finally, each individual TL was determined by application of the lever rule to the relationship between the top mass phase composition and the overall system composition [43]. The experimental binodal curves were fitted using Equation 1 [43]: (1) Y = A   exp [ ( B X 0.5 ) − ( C X 3 ) ]where Y and X, are respectively, the IL and salt weight percentages, and A, B and C are constants obtained by the regression.

For the TL determination the following system of four equations (Equations 2 to 5) and four unknown values (Y_T, Y_B, X_T and X_B) was solved [43]: (2) Y T = A   exp [ ( B X T 0.5 ) − ( C X T 3 ) ] (3) Y B = A   exp [ ( B X B 0.5 ) − ( C X B 3 ) ] (4) Y T = ( Y M / α ) − ( ( 1 − α ) / α ) Y B (5) X T = ( X M / α ) − ( ( 1 − α ) / α ) X Bwhere subscripts M, T, and B denote respectively the mixture, the top phase and the bottom phase, X is the weight fraction of inorganic salt, Y the weight fraction of IL and α is the ratio between the mass of the top phase and the total mass of the mixture. The system solution results in the concentration of the IL and inorganic salt in the top and bottom phases, and thus, TLs can be simply represented.

For the calculation of the tie-lines length (TLL) Equation 6 was used as follows: (6) TLL = ( X T − X B ) 2 − ( Y T − Y B ) 2where subscripts T and B denote respectively the top and bottom phases, and X and Y are the weight fraction of inorganic salt and IL, respectively. It should be pointed out that the top phase is the IL-rich phase, while the bottom phase is the K₃PO₄-rich phase.

Measurements of viscosity and density were performed in the temperature range from 298.15 K to 318.15 K, and at atmospheric pressure, using an automated SVM 3000 Anton Paar rotational Stabinger viscometer-densimeter. Further details on the equipment can be found elsewhere [41]. The dynamic viscosity has a relative uncertainty within 0.35%, while the absolute uncertainty of the density is ±5 × 10⁻⁴ g·cm⁻³. Density and viscosity measurements were carried out at selected biphasic regions. Individual mixtures at the biphasic region were prepared by weight, vigorously agitated and allowed to reach the equilibrium by the separation of both phases for at least 12 h and at 298 K. The proper phase separation was noticeably indicated by the absence of turbidity in each phase. After the separation step, viscosity and density measurements were performed for both aqueous rich phases.

Partition coefficients (K_i) between the IL-rich and inorganic salt-rich phase were determined at 298 K for the following biomolecules: l-tryptophan (K_Trp), β-carotene (K_βcarot), rhodamine 6G (K_Rhod) and caffeine (K_Caf). The partition coefficients are defined as the ratio of the concentration of each biomolecule in the IL- and in the K₃PO₄-rich phase, and as described by Equation 7 (taken as example the biomolecule l-tryptophan): (7) K Trp = [ Trp ] IL [ Trp ] K3PO4where [Trp]_IL and [Trp]_K3PO4 are, respectively, the concentrations of l-tryptophan in the IL and in the K₃PO₄ aqueous-rich phases.

A mixture in the biphasic region was selected, prepared by weight, and used to evaluate the biomolecules partitioning. For this purpose, aqueous solutions with a concentration of approximately 0.78 g·dm⁻³ (3.8 × 10⁻³ mol·dm⁻³) for l-tryptophan, 0.15 g·dm⁻³ (2.8 × 10⁻⁴ mol·dm⁻³) for β-carotene, 0.015 g·dm⁻³ (3.1 × 10⁻⁵ mol·dm⁻³) for rhodamine 6G and 5.0 g·dm⁻³ (2.5 × 10⁻² mol·dm⁻³) for caffeine, were used. These aqueous solutions were used at a mass fraction composition that corresponds to the water mass fraction composition at the biphasic mixture composition selected. The solute’s initial concentration was carefully choosen after carrying out some preliminary tests aimed at achieving absorbance values in an adequate range. All biomolecules aqueous solutions can be considered at infinite dilution and completely solvated in aqueous media, thus avoiding specific interactions between themselves. The biphasic solution was left to equilibrate at (298 ± 1) K for 12 h (a time period established in previous optimization experiments) to achieve equilibrium conditions for the biomolecules partitioning between the two phases. Due care was taken with β-carotene, which undergoes isomerisation on exposure to light, maintaining the samples covered by aluminium foil during the time necessary for equilibration. The experimental procedure for each biomolecule extraction, as well as qualitative evidence for the coloured compounds partitioning to the IL-rich phase, is depicted in Figure 2.

The solute quantification, in both phases, was carried out by UV-Vis spectroscopy using a Shimadzu UV-1700, Pharma-Spec Spectrometer, at wavelengths of 279 nm, 512 nm, 527 nm and 274 nm for l-tryptophan, β-carotene, rhodamine 6G and caffeine, respectively. Calibration curves were previously established for each individual compound. All the wavelengths used for the biomolecules quantification correspond to the maximum absorption peaks of each solute. Interferences of both the inorganic salt and the IL with the analytical method were taken into account and found to not be significant at the magnitude of the dilutions performed. Three samples of each aqueous phase were precisely quantified and the respective standard deviations determined. Moreover, both phases were additionally weighted and the corresponding TLs were obtained as previously described.

The experimental phase diagrams for each IL + K₃PO₄ + H₂O systems at 298 K and atmospheric pressure are presented in Figure 3. All data are presented in molality units for a more detailed understanding of the ILs aptitude for ABS formation (the experimental weight fraction data is provided in the Supplementary). Results obtained for the [C₄mim]Cl IL are also presented in Figure 3 for comparison purposes. In addition, data taken from literature [26] for the ILs [C₄mim]Br and [C₂mim][MeSO₄] are also included for a deeper inspection of the cation influence in promoting ABS.

Although there are several reports in the literature describing ABS comprising imidazolium-based ILs [12–29], this work is the first evidence that phosphonium-based ILs also undergo phase separation in inorganic salt aqueous systems. In addition, the studied phosphonium-based ILs display a greater ability for ABS formation than imidazolium-based ILs when evaluating both classes with similar anions. The higher the affinity for water and/or hydrophilic nature of the IL, the less effective is the IL in promoting phase separation. Therefore, the studied phosphonium-based ILs present lower affinity for water, and are thus more hydrophobic than imidazolium-based ILs. The quaternary phosphonium cations present four alkyl chains and no aromatic character, which is in turn responsible for their lower affinity for aqueous phases. Their water miscibility essentially derives from the polar water hydrogen bonding to the anions.

In Figure 3 there are small, but still visible, quantitative differences in the binodal curves for the phosphonium-based ILs. Considering, for instance, the fixed distance between the binodal curves and the origin at 0.4 mol·kg⁻¹ of K₃PO₄, the ability of the ILs for phase separation can be described by the following order: [P₄₄₄₄]Br > [P_i(444)4][Tos] > [P₄₄₄₁][MeSO₄]. Thus, at such an IL concentration, [P₄₄₄₄]Br has the higher ability to induce phase separation in IL + K₃PO₄ + water ternary systems. Indeed this IL is the one with four straight butyl chains enhancing therefore the ABS formation ability. On the other hand, [P₄₄₄₁][MeSO₄] presents a shorter alkyl chain coupled to a salting-out inducing anion that consequently decreases the miscibility region of the ternary phase diagram. In addition, the aromatic ring in [P_i(444)4][Tos] is certainly the main responsible for its enhanced water affinity when compared to [P₄₄₄₄]Br. The results presented in Figure 3 show that, in general, [P₄₄₄₁][MeSO₄] is the strongest salting-out inducing phosphonium-based IL studied while [P₄₄₄₄]Br is the strongest salting-in inducing IL.

The experimental binodal data was fitted using the approach described by Merchuck et al. [43] (Equation 1). The coefficients of the polynomial equation A, B, and C were determined by regression analysis. These values along with the correlation coefficients (R²) are given in Table 1.

An example of the fitting of the experimental data using Equation 1 is provided in Figure 4. Results for the remaining ILs are provided in the Supplementary. From Figure 4 it is observed that the empirical equation correlates satisfactorily with the experimental data for the [P_i(444)4][Tos] + water + K₃PO₄ system. The same analogy occurs for the remaining ILs. Therefore, Equation 1 can be useful in determining the system composition at any particular point of interest.

Tie-lines (TLs) for each ternary system were determined by the gravimetric method described previously through application of Equations 2–5. The ternary mass fraction composition of the biphasic region used to determine the TLs, TLs adjusted parameters and respective tie-lines length (TLLs) are reported in Table 2.

As an example, the results obtained for the [P_i(444)4][Tos] + water + K₃PO₄ ternary system are represented in Figure 4. The TLs obtained for the ILs [P₄₄₄₄]Br and [P₄₄₄₁][MeSO₄] are graphically represented in the Supplementary. Since the TLL represents the difference between the IL and inorganic salt concentrations in the top and bottom phases, the higher the TLL, the higher is the IL concentration in the top phase and the salt concentration at the bottom phase.

The physical properties of phase forming systems at various concentrations and temperatures are indispensable requirements for the design and scale up of a wide range of processes. Therefore, density and viscosity measurements were performed on both the top and bottom phases for selected mass fraction compositions at the biphasic region. The mass fraction compositions for each ternary system are presented in Table 2. The experimental data is provided in the Supplementary. Figure 5 presents the density and viscosity data as a function of temperature.

Both density and viscosity for IL- and inorganic salt-rich phases are found to decrease as the temperature increased. Typical polymer-salt systems present lower viscosities compared to polymer-polymer systems enhancing a faster segregation of the two phases after extraction procedures, which could be highly advantageous in industrial processes [43]. From the results depicted in Figure 5 no significant differences in density values between IL-based ABS and typical polymer-inorganic salt ABS are observed [44]. Similarly, the bottom phase is the inorganic salt-rich phase while the top phase is the polymer- or IL-rich phase, with density values ranging from 1.0 to 1.2 g·cm⁻³ [44]. However, larger differences are observed in viscosity values between IL-based and polymer-based ABS. IL-rich phases are far less viscous (4–11 MPa·s) than typical PEG-rich phases (39 MPa·s) at mass fraction compositions close to those measured in this work [44]. Therefore, since IL-based ABS are less viscous it constitutes a major advantage for the mass transfer process between the phases and phase handling. The lower viscosity of IL-based ABS compared with polymer-based ABS is thus very important from an industrial point of view.

The density of the inorganic salt-rich phase is higher than the corresponding equilibrated IL-rich phase. [P₄₄₄₁][MeSO₄] is the IL studied with smaller differences in density values among both phases. The density differences among phases generally increase with the increase of the TLLs since larger differences in the mixture composition are occurring. Indeed for [P₄₄₄₁][MeSO₄] the TLL, at the mixture composition selected, is the shorter one. On the other hand, viscosities of the IL-rich phase, and at the mass fraction compositions evaluated, decrease in the rank: [P₄₄₄₄]Br > [P_i(444)4][Tos] > [P₄₄₄₁][MeSO₄]. Nevertheless, the opposite order is verified for the inorganic-salt rich phases. Both differences in viscosities and densities at the K₃PO₄-rich phase indicate that there is some IL migration for this phase contributing for these deviations. However, differences among the inorganic salt-rich phases are significantly smaller than differences between distinct IL-rich phases. In fact, the trend on viscosities for the K₃PO₄-rich phases follows the trend on the IL affinity for water. From Figure 3, [P₄₄₄₁][MeSO₄] is the phosphonium-based IL with an higher affinity for water and as a consequence it will partition in a more pronounced way to the inorganic salt-rich phase, enhancing thus the viscosities of such phase.

Figure 2 provides qualitative evidence for the preferential migration of the coloured compounds to the IL-rich phase. The success of the extractive potential of ABS largely depends on the ability to manipulate phase properties aimed at obtaining appropriate partition coefficients and specific selectivities for the biomolecules of interest. There are several approaches to manipulate the partitioning of the solute in the studied systems, such as the application of different salting-out inducing salts and/or ILs, changing either the concentration of salt or IL, and the introduction of additional co-solvents, anti-solvents or amphiphilic structures. Therefore, the inorganic salt and IL compositions selected for the biomolecules partitioning, as well as the respective partition coefficients, TLs and TLLs are reported in Table 3.

From the results obtained it can be established that the addition of biomolecules into the water phase, at least at low enough concentrations, has no influence in the TLs and TLLs [26], and as graphically shown in Figure 4. In fact, these TLs and TLLs can be considered additional phase equilibrium data. These trends are in close agreement with previous results from our group [26].

During the partitioning of l-tryptophan, rhodamine 6G, β-carotene and caffeine there are several competing interactions between the IL, the inorganic salt, the biomolecules and water. Hydrogen-bonding, π··· π interactions, dispersive interactions, as well as electrostatic interactions between different compounds, are some examples of these interactions.

l-Tryptophan, rhodamine 6G and caffeine are considered quite hydrophilic biomolecules, while β-carotene is highly hydrophobic with a negligible solubility in water [45]. For rhodamine 6G the results indicate that the partitioning behaviour increased with the IL hydrophilic nature, being [P₄₄₄₁][MeSO₄] the most efficient in its extraction, followed by [P_i(444)4][Tos], and finally by [P₄₄₄₄]Br that presents indeed a negligible K_Rhod. Except for [P₄₄₄₄]Br, rhodamine 6G preferentially partitions to the IL-rich phase (K_Rhod > 1). The trend on the ILs’ extraction capability for fairly hydrophilic solutes follows their phase behaviour and affinity for water, and as depicted in Figure 3. In addition, for the IL [P₄₄₄₁][MeSO₄] the solutes affinity contribution in partitioning follows the order: l-tryptophan ≈ rhodamine 6G > caffeine. Nevertheless, comparing with previous results from our group using imidazolium-based ILs [25–27], the partition of l-tryptophan and caffeine are significantly lower when using the evaluated phosphonium-based ILs. For example, with the [C₄mim]Cl + K₃PO₄ + water system the K_Caf is ≈ 49 [27] while K_Tryp is ≈ 37 [25] (for a mass fraction composition of 25 wt % of IL and 15 wt % of K₃PO₄ at 298 K). This trend clearly reflects the strong hydrophobic nature of the studied phosphonium-based ILs and their lower ability for extracting hydrophilic solutes. Indeed, while imidazolium-based ILs present higher partition coefficients of caffeine compared to l-tryptophan, the opposite is verified using the phosphonium-based ILs. On the other hand, and since β-carotene is highly hydrophobic, its extraction is highly efficient with [P_i(444)4][Tos], achieving partition coefficients in the order of 61. In effect, the partition coefficient of β-carotene is highly significant compared to the remaining biomolecules.

In summary, it can be established that high K_i values can be obtained for hydrophobic solutes using phosphonium-based ABS, and that these systems may be a successful and a clean approach for biomolecule separation and purification in biotechnological processes. Furthermore, the large range obtained in the partition coefficients values by changing the IL indicates that the individual biomolecules extraction efficiency can be manipulated by the correct choice of the IL cation and/or anion.