Design Optimization of Deep Eutectic Solvent Composition and Separation Performance of Cyclohexane and Benzene Mixtures with Extractive Distillation

: Deep eutectic solvents (DESs) have properties that make them suitable candidates to be used as entrainers for extractive distillation. In the previous work, it was proven that DES(1:2) (tet-rabutylammonium bromide: levulinic acid, 1:2, molar ratio) can break the cyclohexane-benzene azeotrope. In the present work, the HBA and HBD ratio and molar concentration of DES were optimized to obtain a better constitute and condition of DES to be utilized in cyclohexane and benzene extractive distillation. The physical properties and structure of the prepared DESs were character-ized. Vapor–liquid equilibrium data of the ternary system (benzene + cyclohexane + DESs) were also measured at atmospheric pressure. All experimental equilibrium data were correlated with Wilson, nonrandom two-liquid (NRTL), and universal quasichemical (UNIQUAC) activity coefficient models, from which the coefficient of determination (R2) of the three pseudo-ternary systems fitting was calculated. From the obtained results, the best HBA and HBD ratio in the DESs is eluci-dated as 1:2, the best molar concentration of DES is 0.1, and the NRTL model predicts the experimental data more accurately than the Wilson and UNIQUAC models. From the derived mechanism, the formation of stronger hydrogen bond and π–π bond interactions between DES and benzene is obtained when HBA and HBD ratio in DES is 1:2. In other conditions, the azeotrope cannot be broken, or the efficiency is low. The present work provides an environmentally friendly method to separate aromatic/aliphatic mixtures and act as a guide for further study of DESs in extractive distillation.

In the previous work, it was proven that the DES(1:2) (tetrabutylammonium bromide: levulinic acid, 1:2, molar ratio) can break the cyclohexane-benzene azeotrope. In the present paper, the influence of HBA and HBD ratio and molar concentration of DES on the extractive effectiveness is studied to obtain the optimum input parameters for cyclohexane and benzene extractive distillation using DES.
In this paper, the HBA and HBD ratio (tetrabutylammonium bromide: levulinic acid) is fixed as 1:1, 1:2, 1:3, and 1:4, and the molar concentration of DES is selected as 0.05, 0.1 0.15, and 0.2. The DESs required for the tests are prepared, and their properties and structures are elucidated. VLE data of the ternary (benzene-cyclohexane-DESs) mixtures have been measured to investigate if the azeotrope can be broken to assess the DESs extraction ability. FT-IR and NMR are used in the study for determining the extractive mechanism.

Materials
The purity and source of all the chemicals used in the present work are listed in Table 1.

Preparation of DESs
DESs are prepared by following the procedure reported in the literature [3]. Mixtures are weighed using a Mettler AX205 balance having a precision of ±0.02 mg. Both HBD and HBA are added to a flask and are heated under constant stirring. The uncertainty in the molar mixing ratio is ±2%, evaluated based on the uncertainty in the balance reading and the purity of the chemicals used. The synthesized DESs are dried in a vacuum oven at 100 °C for 24 h. According to the above synthesis method, four kinds of deep eutectic solvents were synthesized using TBAB as a hydrogen bond acceptor and LA as a hydrogen bond donor, as shown in Table 2.

Characterization of DES
The melting point and glass transition temperature of the prepared DESs are investigated using differential scanning calorimetry (DSC). DSC analysis is carried out for all the prepared samples using a METTLER TOLEDO Differential Scanning Calorimeter type DSC1. The DSC measurements are performed under a nitrogen (N2) atmosphere to prevent oxidation of the samples. The N2 flow rate is fixed as 45 mL/min, where the continuous heating is carried out at a rate of 10 °C·min −1 until the specimens are melted completely. Further, the density and viscosity of the synthesized DESs are measured at different temperatures using an Anton Paar, instrument type DMA 5000-AMVn. The structure of chemical constituents and mutual interactions between different constituents are investigated by Fourier-transformed infrared (FT-IR) type T27-Hyperion-Vector22 and nuclear magnetic resonance (NMR) type AVANCE III.

Vapor-Liquid Equilibrium Device
The Vapor-liquid balance device is shown in Figure 1, and the experimental process is as follows. A 30 mL aliquot of the liquid mixture is loaded in the equilibrium kettle. Then, the cooling water is turned on, and gradually, the mixture is heated at the beginning stage. The heating power is adjusted in such a way that the return rate of condensed water of 2-3 drops per second is achieved. To stabilize the reflux action, the device is refluxed for 0.5 h, and then, the temperature is noted every 5 min. After 0.5 h of the stabilization period, the temperature becomes stable, and the liquid and vapor phases are believed to have achieved equilibrium. Then, 0.6 μL of the sample from the liquid phase and vapor phase is collected and tested using chromatography. Each test measurement is repeated three times to decline the test error. As the outlet of the condenser shown in the sketch map is connected with the outside ambient atmosphere, the pressure in the VLE experiment is kept at atmospheric pressure. The samples are analyzed using chromatography with an FID detector employing the programmed temperature technique. The initial column temperature is kept at 50 °C for 1 min. Then, it is heated to 220 °C at a heating rate of 10 °C·min −1 and held for 10 min. The gasification chamber temperature is maintained at 220 °C, where the temperature of the detector is 250 °C, to which high purity N2 as a carrier gas is introduced with a flow rate of 30 mL·min −1 . Further, hydrogen and air are introduced with flow rates of 30 mL·min −1 and 300 mL·min −1 , respectively. Gas chromatographic parameters are shown in Table 3.

DESs Characterization
The melting point and glass transition temperature values evaluated for the four DESs are summarized in Table 4. The density and the viscosity of the four DESs are measured at atmospheric pressure in the temperature range from T/K = 293.15 to 353.15, with a step increase of 10 K. The experimental values for the density and the viscosity at different temperatures are tabulated in Table 5. To study the interactions that exist in HBA and HBD, the FT-IR analysis is carried out for DESs, HBD, and HBA. All the prepared DESs display a glass transition temperature, i.e., when the temperature reaches lower than their freezing point, they become supercooled liquids instead of undergoing crystallization process, and finally, they transform into a glassy state. The melting point of the pure HBD component, which forms DES, has a lower melting point, and therefore, all the glass transition temperatures of DES are compared with the melting point of HBD. The glass transition temperatures of all DES are lower than the melting point of HBD, substantiating that the hydrogen bonding is formed in the prepared DESs, which is very essential. The glass transition temperature of DES(1:4) is the least of all the DESs, indicating that the hydrogen bonds that exist in DES(1:4) are the strongest.  (Figure 2). From the figure, it is inferred that within the temperature range of 293.15 to 353.15 K, the densities of the samples are found to cover the range between 1.04 and 1.12 g/cm 3 , which is close to the density of water, and also, the density tends to decrease with an increase in temperature. The obtained results follow the general liquid density characteristic. Further, with the increase in HBD content, the density increases.  Figure 3, it is inferred that between 293.15 and 353.15 K, the viscosity values of DES are found in the range from 1000 to 10 mPa·s, which follows the same trend as observed in the ionic liquid. Till the temperature is 333.15 K, the viscosity decreases enormously with each step increase in temperature rise. After reaching 333.15 K, the change in viscosity is slow, showing a gradual decrease. Furthermore, it is to be noted that with the increase in HBD content, the density decreases.  In Figure 4, the FT-IR spectra of LA, TBAB, and DESs display the hydroxy peak v(OH) at 3200 cm −1 . Compared to LA, the peak intensity becomes stronger and wider in DES(1:1-1:4) and is shifted to 3000 cm −1 . These interpretations show that the hydrogen bonds are formed between LA and TBAB since there are active hydrogens in LA that can act as HBD and high electron density in the Br atom of TBAB that can act as HBA. At different HBD and HBA ratios, the hydroxy peak intensity strength and width are different, which substantiates that the strength of the hydrogen bond is different for DES(1:1-1:4) and is developed by different ratios.

VLE Experiment
A series of DESs developed using LA and TBAB at different mole ratios, including 1:1, 1:2, 1:3, and 1:4, are used to separate the azeotrope (cyclohexane + benzene) at 101.325 kPa where the mole concentration of DESs is fixed as 0.1.
For a better understanding of the pseudo-ternary VLE behavior, both relative volatility (aij) and selectivity (S) are assessed using the following expressions: where y is the molar fraction of the vapor phase, x the molar fraction of the liquid phase, and the subscripts i and j refer to the more volatile component (cyclohexane) and the less volatile component (benzene), respectively. The estimated relative volatility and selectivity values are provided in Table 6. Table 6 summarizes the Vapor-liquid equilibrium data of the pseudo-ternary systems cyclohexane (1)+benzene (2)+DES (3) at a pressure of 101.325 kPa, where X3 is the DES molar fraction in the liquid phase, X1 is the cyclohexane molar fraction in the liquid phase, X'1 is the DES-free cyclohexane molar fraction in the liquid phase, Y1 is the cyclohexane molar fraction in the vapor phase, T is the temperature, αij is the relative volatility, and S is the selectivity. Table 6. Vapor-liquid equilibrium data of the pseudo-ternary systems (cyclohexane (1) + benzene (2) + DES (3)). At the azeotropic point, the value of the relative volatility is found to be one, which means that both the components in the mixture have the same vapor pressure. As the boiling points of the two components are close to each other, the relative volatility of the two components can be increased by adding a low eutectic solvent. Table 7 and     The local composition activity coefficient models UNIQUAC, WILSON, and NRTL are used for regression analysis of the VLE data of three ternary systems. For each system, the coefficient of determination is estimated to perceive the goodness of fit of the models. The obtained results are provided in Table 7. In the systems containing DES(TBAB:LA) (1:2) and DES(ChCl:LA) (1:2) [31], the prediction is satisfactory as the coefficient of determination is >0.99 assessed from all the three models. In the system containing DES(TBAB:LA) (1:2), the prediction is not satisfactory since the coefficient of determination is <0.99 using the three models.

Cyclohexane (1) + Benzene (2) + DES(1:1)(3), at 101.325 kPa
For a better understanding of the fitness of the three thermodynamic models, the coefficient of determination (correlation index, R 2 ) is calculated using the following expressions: where yi and ui are the experimental data and the estimated value of the vapor phase molar fraction, respectively, and y-is the average value of the experimental data of the vapor phase molar fraction. According to Table 8, the NRTL model displays the best fitness. For each system, the correlation index R 2 value follows the relation NRTL > UNIQUAC > Wilson, validating that the NRTL model is more suitable for predicting mixtures containing low eutectic solvents. Theoretical analysis demonstrates that the deep eutectic solvent is immiscible with benzene and cyclohexane, and thereby, a miscibility gap is created between the mixed solution resulting in the inefficient correlation by the Wilson model. At the same time, the correlation made by each model for the system containing extractant DES(1:2) is low, but the extraction effect of DES(1:2) is the best. This fact is consistent with the conclusion that the better the extraction effect is, the lower the applicability of the prediction model.

Selection of DES Concentration
In the process of the selection of HBD and HBA ratios, the results validate that DES(1:2) is the most effective choice from the prepared DESs. Therefore, the concentration of DES(1:2) should be selected next. In the present work, the molar concentrations of DES(1:2) is chosen as 0.05, 0.10, 0.15, and 0.20. The results of vapor-liquid equilibrium of the pseudo-ternary systems cyclohexanebenzene-DES(1:2) are summarized in Table 9, where X3 is the DES molar fraction in the liquid phase, X1 is the cyclohexane molar fraction in the liquid phase, X'1 is the DES(1:2)free cyclohexane molar fraction in the liquid phase, Y1 is the cyclohexane molar fraction in the vapor phase, T is the temperature, aij is the relative volatility, and S is the selectivity. Table 9. Vapor-liquid equilibrium data of the pseudo-ternary systems cyclohexane (1) + benzene (2) + DES(1:2) (3) at a pressure of 101.335 kPa. At the azeotropic point, the value of the relative volatility obtained will be one, which represents that both components in the mixture have the same vapor pressure. If the azeotrope is not broken, the relative volatility will be lower than one for all concentrations. However, if the azeotrope is broken, the relative volatility becomes higher than one at all concentrations. Table 9 and Figure 9 display that when the DESs are added as entrainers in a molar fraction of 0.05 to 0.2, the relative volatility (α ij ) of the system with 0.1, 0.15, and 0.2 molar concentrations of DES(1:2) is higher than one at all concentrations, whereas the relative volatility of system with 0.05 molar concentration of DES is lower than one. Figure  10 also depicts the same result, where in the pseudo-ternary system with 0.1, 0.15, and 0.2 molar concentrations of DES(1:2), Y1 is greater than X1 in all concentrations and, therefore, validates that the azeotrope is broken. Figure 9. Experimental X'Y diagram for the pseudo-ternary systems (cyclohexane (1) + benzene (2) + DES(1:2)(3)) at atmospheric pressure and constant DES molar fraction (X3 = 0.1), and the experimental XY diagram for binary system (cyclohexane (1) + benzene (2)).

VLE DATA Correlation
In most cases of DES(1:2) molar concentration, the NRTL model shows the best fitness. The coefficient of determination value (R 2 ) follows the trend NRTL > UNIQUAC > Wilson in Table 10. The model regression results for different molar concentrations of DES(1:2) are shown in Figures 12-15. The results depicted from the Figure 11 are also in accordance with the data of the relative volatility (α ij ) in Table 11. In the system added with 0.1, 0.15, and 0.2 molar concentrations of DES(1:2), the azeotrope is broken. For each system, the two binary interaction parameters of cyclohexane/benzene through the NRTL model are obtained, which are shown in Table 11.

Extractive Distillation Mechanism
Extraction mechanisms are very important to understand the extractive distillation process and will provide significant insights for the development of molecule design and the smart design of DES entrainers. FT-IR and NMR are employed to elucidate the extraction mechanism. In the present research, benzene acts as the heavy component and is kept at the bottom along with DESs, where the interaction of benzene and DESs is studied.

Selection of HBD and HBD Ratio
FT-IR analysis is used to elucidate the mechanism for the process of selection of HBD and HBD ratios and to validate the results of the selection. The infrared spectrograms of the test samples are provided in Figures 16-18.  From Figure 17, the FT-IR spectra of DES(1:1-1:4), benzene and DES(1:1-1:4) + benzene shows the peak corresponding to the hydroxy v(OH) group at 3200 cm −1 and the peak ascribed to the carboxyl group v(C=O) at 1722 cm −1 . In the mixture containing DES(1:1-1:4) + benzene, the hydroxy and carboxy peaks become weaker. This fact is attributed to the formation of hydrogen bond and π-π bond interaction between DES(1:1-1:4) and benzene that have destroyed the hydrogen bond within the DES(1:1-1:4) system, and subsequently, the strength of each system is found to be different due to different molar ratios. The results validate that the C=O double bond in LA of DES interacts with benzene by forming a π-π bond, and the O-H bond in LA of DES interacts with benzene by forming the hydrogen bond. However, the change in peak intensity of DES(1:2) is the largest, substantiating that the hydrogen bond and π-π bond interaction between DES(1:2) and benzene is the strongest, leading to the breaking of the azeotrope.

Selection of DES Concentration
Since the change in concentration of the constituents cannot be exactly revealed by FT-IR analysis, NMR is employed in the selection of the DES concentration study. Using ChemDraw software, the magnetic resonance hydrogen spectrum of tetrabutylammonium bromide, levulinic acid, and benzene is simulated, and chemical shift values are obtained, as shown in Figures 19-21. The experimental test of 1 H NMR spectra and magnified view of a particular range are shown in Figures 22 and 23. From Figures 21 and 22, it is inferred that the NMR spectrum simulated from the ChemDraw software highly corroborates with the test measured spectra, confirming that the test samples are highly pure and they are target material. From the expanded spectra shown in Figures 22 and 23, it is inferred that compared to pure DES(1:2), different concentrations of DES(1:2) all have apparent spectral peak shifts and are shifted towards the low field. This is ascribed to the formation of hydrogen bonds between different concentrations of DES(1:2) and benzene. As a result, the electron cloud density around the hydrogen nucleus has declined due to the induced effect. As can be seen from Figure 23, the chemical shift of spectral peak offset value is different for different concentrations of DES(1:2); the offset distance is positively correlated with the DES(1:2) concentration. This substantiates that the hydrogen bonding strength formed between different concentrations of DES(1:2) and benzene is different, and it is positively correlated with the concentration. From these interpretations of the NMR results, the vapor-liquid equilibrium results are verified, i.e., low concentrations of DES(1:2) cannot break the benzene-cyclohexane azeotrope, but high concentrations of DES can break the azeotrope.

Conclusions
The eutectic solvent is prepared, and its physical properties are characterized. Through the vapor-liquid phase equilibrium experiments, the X-Y phase diagram and relative volatility diagram are compared, and the optimal HBA and HBD molar ratio and optimal molar concentration are estimated as 1:2 and 0.1, respectively. Other DES systems changed the azeotropic point but did not break the azeotrope. The results show that the extraction effect of DES is dependent on the system to be separated, the type and ratio of DES, and the DES concentration.
Here, three thermodynamic models are used to fit vapor-liquid phase equilibrium data. By comparing the model-related indices (R 2 ), the correlation degree of the NRTL model is found to be the highest, and the binary interaction parameters of each system are obtained. The results proved the NTRL model is more suitable for predicting the vaporliquid equilibrium behavior of a system containing DESs. In the best extraction system, cyclohexane-DES(1:2) (molar concentration 0.1) system, R 2 is 0.8771, NRTL model binary interaction parameters: aij, −17.895; aji, 26.168; bij, 10000; bji, −10000; cij, 0.1717 (i = cyclohexane, j = benzene).
The screening and optimized results of DES are verified by FT-IR and NMR analyses. It is proven that there is a formation of a strong hydrogen bond and π-π bond between benzene and DES with an extraction effect, and this effect can break the azeotrope. There are weak or no obvious hydrogen bonds and π-π bonds between benzene and DES without an extraction effect, and some can only change the azeotropic point but cannot break the azeotrope. These interpretations substantiate the results of the vapor-liquid phase equilibrium experiments. Combined with thermodynamics research, the present work provides a theoretical basis for the design and optimization of DES.

Conflicts of Interest:
The authors declare no conflict of interest.