Aqueous Biphasic Systems Composed of Random Ethylene/Propylene Oxide Copolymers, Choline Acetate, and Water for Triazine-Based Herbicide Partitioning Study

ABSTRACT Aqueous biphasic systems comprising ethylene/propylene oxide copolymer (EOPO-2500) and choline acetate ([Cho][OAc]) were used for extraction of three herbicides (simazine, cyanazine, and atrazine) from aqueous samples. The influence of phase composition and temperature on phase separation as measured by tie-line lengths (TLLs) and thus, on the partition coefficient (K), of the herbicides was investigated. The herbicides preferred the EOPO-rich phase. The partition coefficients decreased with higher concentrations of EOPO-2500 and [Cho][OAc] and with increasing temperatures (i.e., K decreases with higher TLLs). The proposed system was more environment-friendly and fast compared to traditional liquid–liquid extraction, considering that no organic solvent was consumed.


Introduction
Triazine-based herbicides are widely applied in agricultural production due to their low cost and high selectivity on many important weeds. [1,2] However, their prolonged use causes a risk to nontarget environments and organisms. [1,3] For example, atrazine is reported to be an endocrine disruptor for animals and human beings at low concentrations (~1.0-2.5 mg/kg day). [3][4][5] Their effect at such low concentrations emphasizes the need for sensitive methodologies for their determination. However, these analyses in real samples present challenges due to their complex matrices, and thus, the sample preparation step becomes indispensable to isolate the analytes.
The traditional method used for the extraction of triazine-based herbicides is liquid-liquid extraction (LLE), [6] which presents several drawbacks, including that it is tedious, time consuming, and hazardous to the environment, since large amounts of organic, hydrophobic solvents are consumed. As an alternative, solid-phase extraction (SPE) [7][8][9][10] has become the preferred technique for sample preparation, because it is faster and uses lower amounts of volatile solvents than that of LLE. Still, SPE requires costly cartridges and organic solvents for the extraction and purification of the analytes. As alternatives, several microextraction techniques were proposed for the extraction of the herbicides, for example, solid-phase microextraction, [11] single-drop microextraction, [12] stir-bar sorptive extraction, [13] hollow-fiber-based liquid-phase microextraction, [14] and dispersive liquid-liquid microextraction. [15][16][17][18][19][20][21][22][23][24][25][26] However, poor reproducibility and carryover problems are still drawbacks that most of these techniques need to overcome.
Aqueous biphasic systems (ABS) have been proposed as greener alternatives to conventional LLEs, considering the low toxicity of the ABS components and the shorter times required. [27,28] Based on their composition, ABSs can be classified as polymer-polymer, [29,30] ionic liquid (IL)-salt, [31][32][33] polymer-salt, [34,35] and polymer-IL ABS. [36] The selection of one of these will depend on the nature of the analytes and the matrix of the sample. In addition, an assessment of the environmental and human safety of the ABS components should also be included. [37,38] Due to their low toxicity and easy biodegradability, choline-based ILs have been explored as components of ABS. [36,[39][40][41][42][43][44][45][46][47][48][49][50][51][52][53][54][55] When combined with polymers, the formation of the ABS depends on the hydrophobicity of the polymers. [41] More hydrophobic polymers have lower affinity with water and thus are more easily salted-out by choline-based ILs. In particular, ethylene oxide/propylene oxide random copolymers (EOPO), with both highly hydrophobic (PO) and highly hydrophilic (EO) components, have been used for ABS together with inorganic salts (e.g., K 3 PO 4 , Na 2 SO 4 , (NH 4 ) 2 SO 4 , etc.) or renewable hydrocarbon polymers (e.g., dextran and starch derivatives). [56][57][58] The use of the EOPO polymers is particularly useful in that their balance of hydrophilicity and hydrophobicity is tunable by number and location of the EO an PO units. [59] Although these ABS showed high extraction efficiency for organic compounds and macromolecules, such as enzymes or antibiotics, [60][61][62][63] the combination of EOPO-2500 and ILs in ABS for pesticides extraction has been scarcely investigated.
Among the different options, the combination of a polymer and a hydrophilic IL [39][40][41]55,64,65] can satisfy the requirements for the analysis of herbicides, taking into account the extraction efficiency of the weak or nonpolar pesticides by different ABS. In the present study, the IL choline acetate ([Cho][OAc]) was selected as the salting-out component of an ABS due to its low cost and availability. The relatively high polarity difference between EOPO-2500 and [Cho][OAc] favors ABS formation and the extraction of the selected triazine-based herbicides, with medium polarity into the tunable polymer-rich phase. Here, we use high-performance liquid chromatography (HPLC) to investigate the partitioning behavior of the herbicides under different conditions.

Materials
The three triazine-based herbicides (purity ≥99.0%, Figure 1), choline hydroxide solution (46% in water), glacial acetic acid, EOPO-2500 copolymer (MW 2500), acetonitrile (HPLC grade), and methanol (HPLC grade) were obtained from Sigma-Aldrich (Oakville, ON, Canada). All chemicals and reagents were of analytical grade or higher and were used without further purification. The water used in the experiments was obtained from a Millipore purified water system (resistivity 18.2 MΩ cm, 25°C, Milli-Q Academic, Millipore, Canton, MA, USA).
[Cho][OAc] was synthesized via neutralization of the base with acetic acid, following a reported procedure. [66][67][68] [41][42][43][44]55] A known mass of water was subsequently added to make the mixture clear again. This procedure was repeated to obtain sufficient data for the development of the binodal curve of the ABS. The concentrations of the phase components were determined by weight quantification of all the components added within an uncertainty of 0.001 g.
The water content of [Cho][OAc] (5,000 ± 231.0 ppm) and EOPO-2500 copolymer (1,500 ± 134.7 ppm) was measured using a Karl Fisher Titrator (C20, Mettler Toledo, Switzerland), and the values were taken into account for the calculation of the compositions of the ABS mixtures. The experimental binodal curves were correlated using empirical Eq. (1) [69] to obtain a best fit result: where X is the mass fraction of the [Cho][OAc], Y is the mass fraction of EOPO, and a, b, c, d are the fitted parameters. The equation's parameters were determined by the experimental binodal data at different conditions. Low standard deviations (SDs) were obtained when the experimental values and the calculated data were compared (information in ESI). The tie-lines (TLs) were calculated for the different mass fractions of EOPO-2500 and [Cho][OAc] selected in the biphasic region. For the TL determination, a mixture at the biphasic region was prepared in a 10 mL glass vial, vigorously stirred, and allowed to achieve equilibrium by separation of both phases for 12 h at 25°C. After phase separation, both top and bottom phases were carefully collected. The bottom phase was the EOPO-rich solution, and the top phase was the [Cho][OAc]-rich solution. The composition of the top and bottom phases was determined using the gravimetric method described by Merchuk et al. [70] and confirmed using 1 H NMR. 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.
Eqs. (2) and (3) were used to calculate the tie-line length (TLL) and the slope of the TLs (STL), respectively: where X and Y are the weight fraction of IL and polymer, respectively, and the subscript T and B indicate the top and bottom phases, respectively. TLLs are calculated in mass fractions and are reported as 100TLL (100 times the TLL).
In order to confirm the TL's measured compositions, the Othmer and Tobias equation (Eq. (4)) [71] was used:

is mass fraction of [Cho][OAc] in the [Cho]
[OAc] rich-phase, F and G are adjustable parameters, and w B is mass fraction of EOPO-2500 in the EOPO-2500 rich-phase.
Partition of the target herbicides using the ABS Stock herbicide standard solutions (1.0 mg/mL) were prepared in methanol. Intermediate mixture solutions (100 μg/mL in methanol) were prepared from the stock solutions and stored at 4°C for up to 1 month. Standard solutions with lower concentrations were prepared by serial dilutions with methanol.
For the partitioning study of the herbicides using the ABS, in 10.0 mL centrifugal tubes, known masses of 80 wt% EOPO, 80 wt% [Cho][OAc] aqueous solutions, water, and the target herbicide solutions were sequentially added to form the ABS with different mass fractions of EOPO-2500 and [Cho] [OAc]. The mixtures were then vigorously stirred for 2 min by a vortex at room temperature and centrifuged at 3,500 rpm for 10 min (Sorvall ST8, Thermo Scientific, Canada). After centrifugation, the centrifuge tube was left for 2 h at room temperature to form the ABS. The top phase was primarily composed of [Cho][OAc] and water, while the bottom phase was mainly composed of EOPO-2500, water, and the target herbicides. A blank sample was also prepared and treated according to the procedure mentioned above without adding the target herbicide solutions. The top and bottom phases were carefully separated using plastic syringes, and their respective volumes and masses were recorded. (Since the polymer phase is very viscous, the volume and mass of the EOPO-rich phase were calculated by subtracting the volume and mass of the top phase to the total volume and mass of the mixture, i.e., by difference.) The top phases were filtered with 0.45 µm filter membranes (VWR International, QC, Canada) and injected into the HPLC, while the bottom phases were diluted by methanol, filtered, and then injected into the HPLC.
Partition coefficients (K) and enrichment factor (EF) of the target compounds were calculated using the following equations (Eqs. (5) and (6)): where subscripts IL, EOPO, and 0 represent the IL phase, EOPO-2500 phase, and the spiked sample solution, respectively, and C represents the concentration of the target compounds in the different solutions.

Determination of the solubility of the herbicides in different aqueous solutions
The solubilities of the three herbicides, simazine, cyanazine, and atrazine in different aqueous solutions were determined. The selected herbicide standard solutions (500 µL of 100 µg/mL each herbicide, dissolved in methanol) were dried under N 2 at room temperature, and the residues were redissolved in 1 mL of 40

Determination of the selected herbicides in the top and bottom phases
An Agilent 1260 series HPLC system with quaternary pump, autosampler, and multi-wavelength UV-vis detector (Agilent Technologies, Santa Clara, CA, USA) was used for the quantification of the studied herbicides. The chromatographic separations were performed using a XDB C 18 column (Zorbax Eclipse 250 mm × 4.6 mm, 5 µm, Agilent Technologies, USA) at 30°C. The flow rate was 1.0 mL/min, and the injection volume was 20 μL. Gradient elution was performed with water and acetonitrile (CH 3 CN) as mobile phase for the selected herbicide separations. The CH 3 CN concentration was increased linearly (decreasing water) from 30% and reached 50% after 20 min. The column was then washed by increasing the proportion of CH 3 CN from 50% to 85% in 5 min and then to 95% in 1 min. This composition was held for a further 4 min before being returned to 30% CH 3 CN, followed by a re-equilibration time of 5 min. All solutions to be injected into the HPLC system were filtered using a 0.45 μm membrane filter. The detection wavelength was 220 nm.

Results and discussion
The herbicides simazine, cyanazine, and atrazine ( Figure 1) have pKa values of 1.62, 0.63, and 1.70, [72,73] respectively, and they are stable in solutions between pH 5.0 and 9.0. Since hydrolysis is observed in strong acids and alkalis, [73] the systems used for their extraction, such as the ABS components, must be chosen carefully. EOPO-2500 is completely miscible with water, moderately hydrophobic and can afford neutral pH conditions, [56][57][58] while [Cho][OAc] could provide a suitable pH range and is also environmentfriendly. [40,43,45] The choice of this combination is further based on the high affinity of [Cho][OAc] with water and its high ability to trigger phase splitting. [50] Up to now, this combination has not been considered for ABS formation.

Formation of EOPO-2500/[Cho][OAc] ABS
Solutions with different mass fractions of the polymer (0.5-79.8%) were placed into a tube, and IL solutions (2.2-51.3%) and water were sequentially added to develop the phase diagram (using cloudpoint titration [41][42][43][44]55] ) at 25°C and atmospheric pressure. In the biphasic region, the upper phase is the IL-rich phase, while the lower phase is the polymer-rich phase. [Cho][OAc] is highly hydrophilic with strong ion hydration in solution and thus exhibits a high ability to form ABS in the presence of EOPO-2500.
The temperature effect on the phase diagram of the EOPO-2500/[Cho][OAc] ABS was then evaluated. In previous studies, temperature normally increased the biphasic region of an ABS due to enhancement of the salting-out effect. For example, in a study of an ABS composed of PPG-400 and imidazolium or choline-based ILs, [47,74] increasing the temperature drove water out of the PPG-rich phase to IL-rich phase. This had the effect of decreasing the mass fraction of the ILs required for ABS formation as temperature increased. However, the opposite trend was observed in certain PEG-IL [64] or IL-inorganic salt ABS. [75] In these latter systems, the immiscibility region decreased with higher temperatures. Indeed, the effect of temperature on the phase behavior of any ABS is complex and depends on the different mechanisms of ABS formation. Thus, it is still necessary to empirically determine the effects of temperature on a given ABS.
The resulting phase diagrams for the EOPO-2500/[Cho][OAc] ABS at 20, 25, 30, and 40°C are presented in Figure 2 (the binodal data are provided in Table S1 in the ESI). Temperature within the range studied had little effect on the phase diagrams in the developed ABS. The biphasic regions did expand, but only slightly, with increasing temperature, a phenomenon that was previously reported. [47,74] Although the data suggest that the salting-out ability of [Cho][OAc] can be enhanced by increasing the temperature, considering the negligible difference and to keep the procedure simple, all subsequent work was performed at ca. 25°C.
The TLs of the EOPO-2500/[Cho][OAc] ABS were measured by a gravimetric method, [70] and the respective length (TLL) and slope (STL) of these ABSs were also determined (ESI Tables S3 and S4). The TLL is often used to express the difference in compositions of the top and bottom phases. With increasing TLL, the mass fraction of IL increases in the top phase (from~40% to 51%) and decreases in the bottom phase (from~9% to 2%), while the concentration of polymer in each phase presents the opposite trend (from~5% to 0.6% and~57% to 80% in the top and bottom phases, respectively). TLL is therefore often used to evaluate the effect of phase compositions on the partitioning behavior.

Solubility studies
After selecting an appropriate ABS for this study, we turned our attention to determining the solubility of each herbicide in solutions near the final phase compositions observed in the studied TLLs, namely, aqueous solutions of 40% [Cho][OAc], and 60 and 80 wt% of EOPO-2500. In addition, the solubilities of the analytes were studied in 80% [Cho][OAc] solution, to evaluate the effect, if any, of the viscosity and concentration of the IL on their solubility. To determine the solubilities, the herbicide standard solutions were evaporated to dryness under N 2 , and the residues were redissolved in the aqueous solutions mentioned above to form saturated solutions of the selected herbicides. Then, after filtration, the concentrations of the herbicides dissolved in the solutions were determined by HPLC (Table 1).
From the data in Table 1 On the other hand, the highest solubility of these herbicides in any of the solutions was observed for 60 wt% aqueous EOPO-2500. All three herbicides exhibited a large decrease in solubility as the concentration of polymer increased from 60 to 80 wt%.   [76] 24.8 [77] 710.5 [78] 153.0 [77] The generally higher affinity for EOPO solutions is in agreement with the log octanol/water partition coefficients (log k o/w ) of the hydrophobic herbicides 2.18 (simazine), 2.22 (cyanazine), and 2.61 (atrazine). [79] However, the decrease in solubility of the herbicides in the 80 wt% EOPO-2500 solutions suggested a relatively lower solubility of the herbicides in EOPO-2500 itself, possibly related with an increase of the viscosity of the polymer solution, affecting the mass transfer of the organic compounds. [56,59] Partitioning studies The data in Table 1 suggest that the solubility of the herbicides in the EOPO-2500/[Cho][OAc] ABS is tunable and that the partition of the herbicides can be modified by the composition of the phases which can be followed by the TLLs. We therefore next studied the partition of these herbicides as a function of TLL at 25°C. For an accurate quantification of the herbicides and the volume ratio measured, the mass fraction of [Cho][OAc] was chosen to be in the range of 28-36 wt% with a constant EOPO-2500 concentration of ca. 25 wt%, or the mass fraction of EOPO-2500 was chosen to be in the range of 25-37 wt% with a constant mass fraction of [Cho][OAc] of ca. 28 wt%.
The components of each ABS were weighed according to the different mass fractions and then the herbicides (ca. 4.0 × 10 −6 g) were added to the ternary mixture. The mixtures were vigorously stirred for 2 min and left to achieve phase equilibrium for ca. 2 h. (Although centrifugation was used to speed up the separation of phases, it was necessary to keep the mixtures on the bench for 2 h to achieve equilibrium of the distribution of the analytes.) After phase separation, the concentrations of the herbicides in the two phases of the ABS were determined by HPLC. The partition coefficients were calculated as the ratio of the concentration of the analytes in the lower EOPO-rich phase to the concentration of the analytes in the upper IL-rich phase (see Eq. (5)).
The partition coefficients were first studied at different TLLs with a constant EOPO-2500 mass fraction of ca. 25 wt% (Figure 3a). It was found that ln K was higher than 2, indicating a higher affinity of the herbicides to the EOPO-2500 phase rather than to the [Cho][OAc] phase in all the studied TLs. However, the K values decreased as the TLL increased corresponding to an increase in partition to the IL-rich phase. An increase in TLL corresponds to an increase in the [Cho] [OAc] concentration in the upper phase and a decrease in its concentration in the lower phase. The opposite is true for the EOPO which would become more concentrated in the lower phase. As noted in Table 1, changing the EOPO concentration has a greater effect on herbicide solubility.
Here, it appears that the increase in EOPO concentration in the lower polymer-rich phase forced more of the herbicides into the upper IL-rich phase despite a similar but smaller trend in herbicide solubility. In other words, since the decrease of the solubility of the herbicides in the EOPO-2500rich phase was higher than that in the [Cho][OAc]-rich phase, the partition coefficients of the herbicides decreased with increasing TLL.
In order to confirm the effect of TLL on the partition coefficients, partition was further studied using five ABSs with the same mass fractions of [Cho][OAc] (ca. 28%) (Figure 4a). The composition of the phases and TLL and STL of these ABSs with different mass fractions of EOPO-2500 were determined (Table S4). As shown in Figure 4b, the K values decreased with increasing EOPO-2500 concentration and the corresponding increase of the TLLs. The trend was consistent with the results in Table 1, that is, the solubility of the selected herbicides decreased with increasing concentrations of EOPO-2500 solutions. This confirms that the partition coefficients of the herbicides always decrease with increasing TLL irrespective of the change in the total system concentration of EOPO-2500 or [Cho] [OAc].
The trend observed in Figure 4 was opposite to those in the previous reports, [80,81] in which the partition coefficients of the target compounds were enhanced with the increase of TLLs. In the reported systems, [80,81] the ABSs were formed by imidazolium ILs and salts, and the partition coefficients of the target compounds were the ratio of the concentration in IL-rich phase to the concentration in salt-rich phase. The target compounds were only soluble in the ILs solutions and almost immiscible in salt solutions. Therefore, with increase of the TLLs, the partition coefficients were enhanced due to their higher solubilities in the IL-rich phases.
The effect of different system compositions at the same TLLs on the partition coefficients was also investigated. The results ( Figure 5) indicated that the K values are nearly identical for a given TLL (100TLL = 60.5). This is consistent with the understanding that the EOPO-2500-rich and [Cho][OAc]-rich phases should have the same compositions at a given TLL regardless of the overall system composition.
Although it was observed that the influence of different temperatures on the binodal curves of the EOPO-2500/[Cho][OAc] ABS is negligible (Figure 2), temperature could still have an effect on the partition coefficients of the herbicides. Therefore, the partition coefficients of the herbicides were studied at temperatures from 20 to 40°C using the ABS with 25.0 wt% EOPO-2500 and 28.  (Higher temperatures were not evaluated, since cyanazine is not stable at higher temperatures under alkaline conditions. [82] ) As can be seen in Figure 6, the partition coefficients of simazine and cyanazine decreased significantly with increasing temperature, while the partition coefficients of atrazine decreased first and then remained almost constant with slight differences at higher temperature (40°C).

Analytical figures of merit and comparison with other methods
In order to overcome the complicated operation and high pollution of the traditional methods in water sample analysis, the ABS based on [Cho][OAc] and EOPO-2500 was developed to extract the target herbicides. To test this approach, spiked water samples (spiked concentration 0.01 mg/kg) were prepared with tap water by adding a mixture of standard solutions of the herbicides. Each phase component was measured by weight fraction, and an appropriate amount of spiked water (which was used to substitute water in the ternary mixture) was added. The protocol described above was followed to form the ABS. The lower phase was collected, filtered, and injected into the HPLC for quantification. Blank samples were also prepared to evaluate possible interferences of the system. To study the use of the ABS method for determining herbicides in tap water samples, optimized conditions were chosen as 1.5% EOPO-2500 + 45.90% [Cho][OAc], providing a TLL of 80.5 and a volume ratio (ratio of the upper volume to the lower volume) of 65. A lower volume of the extracting polymer-rich phase was selected to enhance the EF (ratio of the concentrations of the analytes in the EOPO-2500 phase to their concentrations in the spiked samples), which was 50 at the selected conditions.
The developed method was evaluated by a validation procedure with spiked samples to obtain linear range, accuracy (expressed as recovery), precision (relative standard deviation [RSD]), limits of detection (LODs), and limits of quantification (LOQs). Under the optimized conditions, the calibration curves of the target herbicides were established using the measured peak areas of the herbicides against their concentrations. Good linearity was shown with correlation coefficients higher than 0.9910.
The limit of detection (LOD, S/N = 3) and the limit of quantification (LOQ, S/N = 10) were determined according to the ratio of signal to noise of the spiked samples. The LODs of the herbicides ranged between 0.5 × 10 −3 and 1.2 × 10 −3 mg/kg and LOQs in the range of 0.002-0.004 mg/kg. The recoveries of the targets were evaluated at three spiked concentrations (0.005, 0.01, 0.05 mg/kg). The extraction recoveries of simazine, cyanazine, and atrazine were 92(±6)%, 81(±6)%, and 83(±5)%, respectively, and the RSDs were all below 8.7%, indicating satisfactory accuracy and precision.
When compared with methodologies previously reported for the extraction and determination of triazine herbicides from real samples (Table 2), the proposed method presents advantages, such as simple operation without using special or expensive apparatus, and relatively low LODs in the same range of the current methods. However, the method required relatively long times for achieving phase equilibrium (ca. 2 h), and the extraction time was the highest among these methods. Due to the high tunability of ILs, a future evaluation of alternative ILs' components might allow minimization of the equilibration time.
The developed method showed relatively lower toxicity and is proposed to be more environmentally benign, since it uses an environment-friendly and biodegradable polymer [83,84] and IL rather than volatile organic solvents during the extraction. In addition, the phase components of the ABS could be recycled and reused. As a thermos-separating polymer, EOPO-2500 could be separated from aqueous solution and recycled at around 50°C. [85] Therefore, after phase separation and sample injection, the bottom phase was diluted with a certain amount of water and heated to 50°C for 12 h, resulting in EOPO separation from water. At the same time, the selected IL was recycled by evaporation of water.

Conclusions
In this study, an ABS composed of EOPO-2500 (MW2500) and [Cho][OAc] was studied and optimized for increasing partition coefficients of the herbicides: simazine, cyanazine, and atrazine. The partition coefficients of the herbicides decreased with increasing TLL and temperature, whereas the salting-out effect of [Cho][OAc] did not play a major role in the partition of the herbicides. The competition among the IL, polymer, and the herbicides for the water molecules seems to dominate the phase separation and the partition coefficients of the analytes. The developed ABS was optimized, and high extraction efficiencies of the target herbicides were achieved ranging between 81% and 92%. The results revealed that the herbicides had high affinities for the EOPO-2500 phase. The proposed method is simple, cheap, and an environment-friendly alternative to the conventional sample preparation methods for the determination of these herbicides in water samples.