Experimental Measurements of Octanol-Water Partition Coefficients of Ionic Liquids

Ionic liquids (ILs) are low melting point organic salts, most of which are liquids at room temperature. In the past decade they have generated a huge interest in research and industrial fields for their capacity to serve as chemical and biochemical reaction media. In addition, ILs are of interest because they constitute a new group of polar and nonaqueous solvents, whose most important advantage is their negligible vapor pressure [1]. It is mainly for this reason that they are considered “green solvents” compared to conventional volatile organic compounds (VOCs). Due to their high chemical and thermal stability, ILs can resist high temperatures. However, for their potential applications, ILs are mainly valued because of the possibility of modulating their physical and chemical properties, such as melting point, viscosity, density, hydrophobicity and polarity by selecting the anion, the cation or substituent present in their specific structure. The number of different combinations of anions and cations that can be used to form potential ILs is enormous.


Introduction
Ionic liquids (ILs) are low melting point organic salts, most of which are liquids at room temperature. In the past decade they have generated a huge interest in research and industrial fields for their capacity to serve as chemical and biochemical reaction media. In addition, ILs are of interest because they constitute a new group of polar and nonaqueous solvents, whose most important advantage is their negligible vapor pressure [1]. It is mainly for this reason that they are considered "green solvents" compared to conventional volatile organic compounds (VOCs). Due to their high chemical and thermal stability, ILs can resist high temperatures. However, for their potential applications, ILs are mainly valued because of the possibility of modulating their physical and chemical properties, such as melting point, viscosity, density, hydrophobicity and polarity by selecting the anion, the cation or substituent present in their specific structure. The number of different combinations of anions and cations that can be used to form potential ILs is enormous.
Due to the non-volatile character of ILs, they do not contribute to atmospheric pollution [2]. However, despite their significant degree of solubility in water, their effect in this respect has not been studied in depth [3][4][5]. Moreover, because of their high stability, ILs could represent serious pollutants in aqueous waste streams or accidental spills. For this reason, it is very important to quantify this effect by means of toxicological parameters such as the octanol-water partition coefficient (K ow ) that we study in this work. The K ow is known to be one of the quantitative physical properties that best correlate with biological activity because the water-saturated octanol system is considered a reasonable model of the physico-chemical environment in living organisms [6,7]. K ow classifies ILs according to their hydrophobicity or hydrophilicity, the latter characteristic being closely linked to the lipophilicity of a chemical compound and this parameter constitutes an essential physicochemical property in medicinal chemistry. For instance, it plays a crucial role in the absorption, distribution, metabolism and excretion (ADME) characteristics of drugs [8]. In fact, K ow is used to model blood/lipid partition in toxicology to understand Abstract Interest in ionic liquids (ILs) has increased due to their promising use as "green solvents" because of their negligible vapor pressure. However, their solubility in water could lead to their dispersion into the environment through liquid effluents, generating an important toxicological effect in soils and seawater. One of the most relevant parameters related to the assessment of environmental risk is the octanol-water partition coefficient (K ow ). With this parameter is possible to estimate some ecosystem risk factors such as bioaccumulation, sorption to soils and sediments and toxicity in fish by the usage of experimental correlations. Shake-flask and slow-stirring methods are the most currently used methods for determining the K ow of a chemical compound. The former has the disadvantage that equilibrium might not be reached quickly, while the slow-stirring method is not always suitable for ILs, since some of them may decompose after continuous contact with water. We have developed a combined version of both methods. Here, we present measurements of the K ow of twenty-four ILs at 30°C, using the three experimental methods. The types of anion and alkyl chain length of the cation are among the parameters studied. The K ow of ILs used in this study range between 0.0017 and 3.6567 at 30°C. The K ow of ILs studied is lower than in commonly used industrial solvents. *Corresponding author: Villora G, Faculty of Chemistry, Regional Campus of International Excellence "Campus Mare Nostrum", University of Murcia, P.O. Box 4021, Campus of Espinardo, E-30071, Murcia, Spain, Tel: +34-868-887-363; E-mail: gvillora@um.es the tendency of a compound to cross biological membranes [9]. Therefore, K ow is the key parameter for use in experimental correlations to estimate some parameters related to bioconcentration [10][11][12][13] and toxicity in fish [14,15], as well as sorption to soils [16][17][18].
Octanol and lipids have similar molecular structures containing both polar and hydrophilic oxygen at the end of a long hydrophobic alkyl chain and also similar physical properties. For this reason, correlations between natural ecosystems and K ow are highly useful [19]. A high number of empirical correlations have been developed to estimate bioconcentration (BCF) and bioaccumulation (BAF) factors, soil sorption coefficients (K OC ), and toxicity. Bioconcentration is the process that causes an increased chemical concentration in an aquatic organism compared to that observed in water, due to the absorption of chemicals by different metabolic routes. The bioconcentration factor, BCF, is the equilibrium ratio of the chemical's concentration in the organism, in μg/kg lipid, to the concentration in the water, in μg/L, when exposure is only to the chemical in the water [20]. Pollutants also enter the biota through the food chain, which is referred to as biomagnification. The sum of bioconcentration and biomagnification is referred to as bioaccumulation [21]. The soil sorption coefficient, K OC , describes soil-water partitioning. The K OC is the ratio of the mass of a chemical adsorbed per unit weight of organic carbon in a soil to the concentration of the chemical in a liquid phase. Toxicity is typically reported in terms of mortality to various species quantified by the LC 50 in μmol/L, which is the concentration lethal to half of the species population [22]. Examples of equations based on experimental data relating K ow with BCF, K OC and LC 50 are shown in Table 1 [23,24].
The activity of a compound, a i , in the water-rich phase and the octanol rich-phase will be, by definition, the same ( ) , where γ i is the activity coefficient and x i is the mole fraction in the water rich ( w ) and octanol rich ( o ) phases, respectively, the following expression can be deduced: If the test compound solution is very dilute and pressure and temperature are constant, concentrations and mole fractions will be proportional. In addition, if the compound is extremely dilute in both phases, so-called "infinite dilution", activity coefficients can be considered not to change with small variations in the concentrations. As a result, the Nernst distribution law is followed: where K ow is the octanol-water partition coefficient and C i are sufficiently dilute concentrations in both phases [3,6].
Ideally, concentrations of the same solute species are measured in the two liquid phases when determining K ow values. This may be difficult for species such as acids or salts, in which the solute tends to dissociate more in the aqueous phase than in the octanol-rich phase. As shown in Figure 1, ILs [M v+ X v-] tend to have a greater tendency to dissociate in the water-rich phase [3].
When solute dissociation is expected, the K ow is calculated as the ratio between the concentration of the undissociated and dissociated species of the salt in the octanol-rich phase and in the water-rich phase. For this work, concentrations were measured in each phase using UV-Vis spectroscopy, which detects the imidazolium, pyridinium and ammonium group on the cation (whether dissociated or undissociated). Therefore, the K ow values reported are given by Eq.3: It is important to recognize that octanol and water are not completely immiscible. At 25°C, the solubility of water in octanol is approximately 0.275 mole fractions, but the solubility of octanol in water is 7.45 × 10 -5 mole fraction [25]. Since the solubility of water in octanol is high, the K ow cannot be represented simply as the ratio of saturation concentrations of a solute dissolved in separate solutions of octanol and water. The mutually saturated octanol and water phases affect the partitioning of the solute between the two solvents.
Many different experimental methods (direct and indirect) exist for determining K ow of a chemical compound, but two methods are the most commonly used: shake-flask and slow-stirring, both direct methods. In the traditional shake-flask method, octanol and water are mutually saturated for three days. A sufficiently dilute solution of the test compound and water-saturated octanol is brought into contact with the same quantity of saturated water and shaken for about five minutes to achieve equilibrium. Next, both phases are separated by centrifugation and the test chemical concentrations are measured in each phase. This method has speed as an advantage and the disadvantages of the possible formation of microdroplets after centrifugation and the fact that equilibrium is not guaranteed, then the evaluation of partition equilibrium can be done by monitoring the absorbance after the phase separation [3,[26][27][28]. Slow-stirring method is similar to the previous method. However, instead of rigorously shaking both phases, they are stirred slowly for an extended period of time (45 days, approximately). After that, each phase is analyzed in the same way as in the shake-flask method. The main advantage of this method is to avoid the emulsification due to the reduction of the stagnant diffusion layer between the two phases. Nevertheless, care must be taken with the length of time that the phases are in contact since some ILs may decompose after continuous contact with water [3,29].
Our research group is especially interested in working with ILs hence this paper is focused on the determination of the K ow for a group of ILs. Information about the toxicological risk of ILs is limited in the literature. Our research group has already measured several K ow values for some imidazolium-based ILs using our own method, which will be explained below [30]. Other authors, for example, Ropel et al. [3], Deng et al. [7,31] and Ventura et al. [32] used the slow-stirring method to measure K ow for imidazolium, pyridinium, ammonium and pyrrolidinium-based ILs. Others like Kaar et al. [33], Zhao et al. [34] and Lee and Lee [9] determined K ow values for imidazolium and pyrrolidinium-based ILs using the shake-flask method. The results reported by these publications for ILs have been inconsistent due to different experimental methods and range of IL concentrations established in the experiments. However, some tendencies can be extracted. For instance, it is clear that more hydrophobic anions such as [NTf 2 -] lead to higher K ow values and, K ow values increase with alkyl chain length in ILs with the same anion. Some studies have developed an indirect way to determine K ow from the ratio of IL solubility in water and in octanol but this value does not exactly correspond with the definition of K ow [35,36]. Other authors have preferred to use computational simulations [37,38] or DFT calculations (LFER parameters) [39] to determine K ow for ILs. Recently, a robust and automated method for measurement of K ow has been established. It consists of a syringe pump with a selection valve, a holding column, a silica capillary flow-cell and an in-line spectrophotometer. Distribution of the drug between the aqueous and octanol phases occurres by the oscillation movement of the syringe pump piston. The system has been applied to the determination of the K ow of some common drugs, achieving high precision with only one-phase measurement [40]. Other new experimental methodology to obtain K ow has been developed by the use of a bubble column set-up in combination with headspace concentration measurement [41]. Finally, a K ow prediction method based on the lipophilicity estimation of ILs by chromatographic methods was used by Stepnowski et al. [42], Ranke et al. [43] and Studzińska et al. [44]. Due to the quantity of ionic ILs synthesized and the lack of information data about their toxicological Correlation Reference Log (BCF)=0.79log (K ow )-0.40 [24] Log (BCF)=log (K ow ) -1.32 [10] For Log (K ow ) < 5; Log (BCF)=0.5 [13] Log (BCF)=0.791log (K ow )-0.729 [15] Log (K oc )=0.544log (K ow )+1.377 [16] Log (K oc )=0.989log (K ow )-0.364 [17] Log (K oc )=0.679log (K ow )+0.663 [18] Log (1/LC 50 )=0.854log (K ow )-1.74 a [15] Log (1/LC 50 )=0.629log (K ow )-0.489 a [15] Log (1/LC 50 )=0.89( ± 0.03)log (K ow )-1.75( ± 0.05) b [14]  a known concentration of IL (1 mM) was added to the centrifuge tube. The tube was shaken vigorously for about five minutes and then the octanol and water phases were separated by centrifugation. Samples were taken from the octanol-rich phase and the water-rich phase with a glass syringe and analyzed directly.
In all three methods, the IL concentration in each phase was analysed by UV-VIS spectrophotometry as described in the section Analytical Method. Determinations were made at least in triplicate to ensure repeatability of the tests and the mean values are reported. The samples were diluted until the absorbance was less than 1. The initial concentration of IL in the octanol phase was less than 1.2 × 10 -2 mol/L, but enough to ensure that the IL could be measured accurately. The final concentrations of IL in the octanol phase were between 1.77 × 10 -5 mol/L and 3.92 × 10 -3 mol/L and in the water phase between 2.75 × 10 -5 and 8.38 × 10 -3 mol/L. IL concentrations were kept at the dilution limit so that K ow values would be independent of concentration.
In the present paper, we have studied twenty-four ILs based on imidazolium, pyridinium and ammonium salts with different substituents:

Materials
The structures of the ILs used in this study are listed in Table 2 risk, more studies are required to ensure the safe and environmental friendly use of ILs by researchers and industries.
In the present paper, we measured the K ow of dilute samples of several imidazolium, pyridinium and ammonium-based ILs. The K ow values previously reported for some ILs have shown huge discrepancies, as can be seen below. For this reason, we have used three different experimental methods and compared the results obtained, which has not been done before now. Furthermore, the effect of the anion and cation alkyl chain length on the K ow values is discussed.

Experimental Experimental method
The K ow values of ILs were measured using three different methods, two of them explained in the previous section (the shake-flask and slowstirring methods) and the other one a combined version of the same which our research group developed in a previous paper [30]. Before to carry out each experiment, both water and octanol were mutually saturated by stirring for three days or more. The experimental setup for the slow-stirring method and the combined method is very similar, both consisting of a 22 mL glass vial with an open-top screw cap sealed with a silicone/Teflon septum. In the slow-stirring method and our combined method, approximately 10 mL of distilled, deionised water, presaturated with octanol was added to the glass vial. A 12-gauge Teflon tube was introduced into the vial, reaching just below the water surface. Approximately 10 ml of octanol-IL "stock" solution, consisting of octanol presaturated with water and containing a known concentration of IL (1 mM), was added to the vial. In the combined method, the vial was shaken vigorously for 5 minutes and then maintained at 30 ± 0.5°C in a thermostatic bath without stirring for 13 days. In the slow-stirring method, the vial was not shaken intensely, but was maintained in the same bath at 30 ± 0.5°C with a 1 cm Teflon coated magnetic stir bar which provided slow-stirring for 45 days. In both methods, samples were taken from the octanol-rich phase by penetrating the silicone septum with a stainless-steel needle fitted with a glass syringe. Samples were withdrawn from the water-rich phase by inserting the needle through the Teflon tubing directly into the aqueous phase to prevent octanol contamination. In both cases, samples of each phase were taken from the vials until three measurements coincided confirming that the concentrations in both phases had stabilized.
In the shake-flask method, the apparatus consists of a 12 mL centrifuge tube with cap. Approximately 5 mL of distilled, deionised water, presaturated with octanol and 5 mL of octanol-IL "stock" solution, consisting of octanol presaturated with water and containing

Analytical method
The concentration of the ILs in octanol and water was measured by UV-VIS spectrophotometry using a ThermoSpectronic UV-VIS recording spectrophotometer (Heλios α), which has a sensitivity of ± 0.001. A calibration curve was prepared for the different ILs at their maximum absorbance wavelength: 212 nm for imidazolium, 216 nm for pyridinium and 210 nm for ammonium-based ILs. The calibration curves were made in duplicate and mean values of the extinction coefficients are reported (Table 3). Samples were diluted if their concentration exceeded the calibration range.

Experimental technique evaluation
The combined experimental technique developed by our group was validated in a previous work [30] with benzaldehyde, which has a well-known log(K ow )=1.48. Benzaldehyde shows two characteristic peaks at 250 and 200 nm in water, and at 245 and 207 nm in octanol. The wavelengths at 250 nm and 245 were selected for absorbance measurements in water and octanol, respectively. The extinction coefficient in water and octanol is shown in Table 3. Log(K ow ) of benzaldehyde obtained was 1.43, which is very close to the value given in the literature (1.48), confirming the validity of the proposed method [30].

Extinction coefficients
To determine the concentrations of IL in each phase and the resulting K ow values, the IL extinction coefficients (ε) in octanol and water were obtained at the wavelength of maximum absorption of the imidazolium (λ max = 212 nm), pyridinium (λ max = 216 nm) and ammonium (λ max = 210 nm) groups (  has been widely used in recent decades; however, several parameters need to be taken into account for the measurement to be valid. For example, pre-saturation of the two solvents with each other is essential to obtain accurate numbers for the volume of each phase. Shaking has to be carried out manually or mechanically for a sufficiently long time to homogenize the solution and, since octanol and water form an emulsion, the two phases have to be separated. This problem may be overcome by centrifugation, but, even so, micro droplets might still remain in each phase, which can introduce large errors in the final measurements, especially in the case of hydrophobic ionic compounds.
The K ow values obtained by the shake-flask method are generally the lowest, probable because the equilibrium had not been achieved.
The slow stirring method can avoid the measurement errors introduced by the emulsion between octanol and water, especially for hydrophobic compounds. This method is generally considered a better representation of ecological conditions, whereby oil and water are equilibrated and the IL is present in very dilute amounts. Nevertheless, it is not appropriate for all ILs, since some may decompose after continuous contact with water for a long period of time. The values obtained by the slow stirring method are generally much higher than those obtained by the other methods.
The combined version of both methods seeks to avoid the disadvantages of each of the conventional methods. As can be seen in Table 4, the values obtained with the combined method are generally intermediate with respect to the other two procedures. However, depending on the IL, different situations can be found. For some ILs the combined method provides results very similar to those obtained with the shake-flask method and, for others they are similar to the slowstirring results. This may be due to the different rates of degradation in water and/or the different tendencies to form microemulsions, and demonstrates the difficulty involving in using the technique.
However, while the three methods deliver different results, some trends can be identified. The K ow values increase with increasing alkyl chain length of the cation (Figure 2), which is consistent with the findings of other authors [3,5,7,30,37,39,52]. The ILs with a bis(trifluoromethylsulfonyl)imide anion have higher K ow values and the ILs with an ammonium cation have the lowest values. There are some differences between the K ow values calculated by the two molecular simulation methods, but the trends observed are similar and agree with the experimental results.

Conclusions
The K ow of ILs is an important parameter because describes the lipophilicity of ILs and can be used to determine properties such as are bigger than the others, perhaps as a result of different interactions with the solvents.

K ow Values
The values of K ow measured for each ionic liquid and obtained with the three methods indicated in Experimental Method Section are shown in Table 4. The K ow values found in literature for some of the ILs have been added to compare the results, distinguishing between the K ow values obtained by the shake-flask or slow-stirring method. The uncertainties reported are standard deviations of multiple tests, as detailed in the Experimental section.
The K ow of a given chemical structure can also be estimated. Most methods divide the molecule into fragments or groups of atoms, each with its corresponding empirical constant and structural factor. Using experimental data, a database of the different contributing fragments and structure factors possible can be computed. We used two of these approaches to estimate the K ow values of ILs, both using SMILES notation (Simplified Molecular Input Line Entry System): web page molinspiration.com [45] and Bio-Loom Software [46]. The estimated values are also shown in Table 4 [47][48][49].

Discussion
The ILs studied here based on imidazolium, pyridinium and ammonium, are fairly hydrophilic so their K ow were expected to be low. Indeed, the experimental values shown in Table 4 are lower than those in commonly used industrial solvents (i.e., K ow(ethanol) =0.479 [50]; K ow(nheptane) =31623 [51]).
On the other hand, the three methods used here to determine the K ow of ILs have their advantages and drawbacks.
The shake-flask method is a classical method to measure the K ow that    environmental effects, liquid/liquid solute partitioning and miscibility with other solvents. However, the values of the K ow of ILs differ very much according to the experimental method used for their determination. It is possible that the differences in K ow values obtained by different authors for the same ionic liquid might be also due to differences in ionic liquid concentrations.
The K ow of ILs used in this study range between 0.0017 (ETAN) The values of K ow are lowest for the most hydrophilic ILs and increase with the cation alkyl chain length. Since all of the K ow values are very small, we conclude that none of the ILs studied will accumulate or concentrate in the biota.