Separation of Benzene ‐ Cyclohexane Azeotropes via Extractive Distillation Using Deep Eutectic Solvents as Entrainers

: The separation of benzene and cyclohexane azeotrope is one of the most challenging pro ‐ cesses in the petrochemical industry. In this paper, deep eutectic solvents (DES) were used as sol ‐ vents for the separation of benzene and cyclohexane. DES1 (1:2 mix of tetrabutylammonium bro ‐ mide (TBAB) and levulinic acid (LA)), DES2 (1:2 mix of TBAB and ethylene glycol (EG)) and DES3 (1:2 mix of ChCl (choline chloride) and LA) were used as entrainers, and vapor ‐ liquid equilibrium (VLE) measurements at atmospheric pressure revealed that a DES comprised of a 2:1 ratio of LA and TBAB could break this azeotrope with relative volatility ( α ij) up to 4.763. Correlation index suggested that the NRTL modelling approach fitted the experimental data very well. Mechanism of extractive distillation gained from FT ‐ IR revealed that with hydrogen bonding and π – π bond inter ‐ actions between levulinic acid and benzene could be responsible for the ability of this entrainer to break the azeotrope.


Introduction
Distillation is mostly used in industry to separate liquid mixtures. However, azeotropic mixtures (or approximate boiling point systems) require special distillation processes for effective separation processes. The separation of benzene and cyclohexane azeotrope is one of the most challenging processes in the petrochemical industry. The most attractive separation technique for benzene and cyclohexane is liquid−liquid extraction [1,2]. Traditional organic solvents are usually selected as extractants, such as furfuryl alcohol, ethylene glycol, N-methyl pyrrolidone (NMP). However, these solvents are typically toxic, flammable, volatile, and difficult to recover.
In this respect, ionic liquids (ILs) have been employed as effective entrainers for separations in the fine chemical industry. However, the toxicities, low purities, and relatively high costs of these entrainers have restricted their use in the food or pharmaceutical industries. Recently, deep eutectic solvents (DESs) that contain hydrogen bond acceptor (HBA) and hydrogen bond donor (HBD) have been proposed as an alternative class of entrainer [3]. Most DESs are comprised of an organic salt as an HBA (e.g., quaternary ammonium salt, quaternary phosphine salt, etc.) and a carboxylic acid, amide, or alcohol as an HBD. However, DESs cannot contain all kinds of these solvents, and many DESs have glass transition temperatures rather than eutectic melting points. Thus, DESs are also called low transition temperature mixtures (LTTMs) [4,5]. DESs also have been studied in the field of gas-gas/liquid−liquid extraction [6][7][8][9][10][11][12] and extractive distillations [13][14][15][16][17][18][19][20]. Deep eutectic solvents (DESs) as novel entrainers can be used for the separation of azeo-  [21][22][23]. Shang et al. suggested that deep eutectic solvent ChCl/Urea (1:2) could be used for extractive distillation for the separation of ethanol-water [24]. The study of Samarov [25] tested Choline chloride-based deep eutectic solvent (DES) for the separation of azeotropic mixtures of ethanol-ethyl acetate, n-propanol-n-propyl acetate and n-butanol-n-butyl acetate via liquidliquid extraction. Some researchers believed that azeotropic phenomenon was broken because hydrogen bond sites which exist in DES have strong solvation with polar substances, thus changing the relative volatility [26][27][28][29][30]. However, many of these extractive processes have not been fully optimized.
In this paper, a cheap and easily prepared DES as an entrainer in benzene-cyclohexane azeotropes extractive distillation processes was reported. Vapor-liquid equilibrium (VLE) data of all DES systems were determined to evaluate the extractive distillation effect. The separation process was simulated in Aspen Plus V8.4 to select the most suitable simulation method. Furthermore, the mechanism of extractive distillation was explored by FT-IR.

Chemicals
All the chemicals (>98% purity) used in this study were purchased from chemical suppliers and used without further chemical purification, with choline chloride (ChCl) stored in a vacuum desiccator before use.

Preparation of DES
DES were prepared following the procedure reported in Ref [30]. Specifically, mixtures of the HBD and HBA were stirred and heated until they melted into a homogeneous liquid. Then the resultant DES was dried in a thermostatic drier box at 100 °C for 24 h. The DES were classified as DES1 (1:2 mix of tetrabutylammonium bromide (TBAB) and levulinic acid (LA)), DES2 (1:2mix of TBAB and ethylene glycol (EG)) and DES3 (1:2mix of ChCl and LA). Figure 1 showed the representative images of the DES that were prepared.

Characterization of DES
The melting point and glass transition temperature of the selected DES were determined using differential scanning calorimetry (DSC) using a METTLER TOLEDO Differential Scanning Calorimeter type DSC1. All DSC experiments were carried out under a N2 atmosphere to avoid oxidation of the samples, using a N2 flow rate of 45 mL/min and a heating rate of 10 °C•min −1 until the sample had melted completely. The densities and viscosities of the selected DES were measured at a series of temperatures using an Anton Paar viscosity and density instrument (DMA 5000-AMVn). IR analysis was carried out using a Fourier-transformed infrared (FT-IR) spectrometer (T27-Hyperion-Vector22).

VLE Measurements
VLE measurements were conducted using an all-glass dynamic circulation vaporliquid equilibrium still (a modified Othmer still). There was 30 mL of a liquid sample injected into an equilibrium kettle fitted with a condenser and the rate of heating adjusted until a condensation return of 2-3 drops per seconds was achieved. After refluxing for 0.5 h, the temperature was then measured every 5 min until the temperature of the distillate was constant (around 0.5 h) to ensure that the liquid and vapor phases had reached equilibrium. Next, 0.4 uL samples of the liquid and vapor phases were sampled and subjected to chromatographic analysis, with each experiment repeated three times to ensure consistency. The outlet of the condenser was maintained at ambient atmospheric pressure throughout the course of the VLE experiment. Samples were analyzed using gas chromatogram fitted with a KB-FFAP column (30 m × 0.32 mm × 0.25 um) and an FID detector using a two-stage programmed temperature method, with a slit ratio of 200. The initial column temperature was maintained at 70 °C for 1 min; then heated to 150 °C (at a rate of 15 °C•min −1 ); maintained at 150 °C for 2 min; heated to 220 °C (at a rate of 20 °C•min −1 ); and then maintained at 220 °C for 5 min. The oven temperature was maintained at 230 °C, the temperature of the detector was 250 °C, with high purity N2 used as a carrier gas at a flow rate of 30 mL•min −1 H2, and air flow rates of 30 mL•min −1 and 300 mL•min −1 , separately.

Characterization of DES
The results obtained for the melting points and glass transition temperatures of the three DES are shown in Table 1. From Table 1, it can be seen that only DES1 and DES2 were found to exhibit glass transition temperatures, which occurred between 102.2-107.2 °C and 39.1-41.1 °C, which were lower than the melting points of their pure components. Furthermore, the glass transition temperature of DES1 decreased more than that of DES2, which means that the hydrogen bonding interactions in DES1 could be stronger.
The densities and viscosities measured at atmospheric pressure in duplicate from 293.15-353.15 K at temperature intervals steps of 10 K were shown in Figures 2 and 3. The relative standard deviations of the densities and the viscosities of the DES were found to be u(ρ) = 0.002 and u(u) = 0.003, respectively. The densities of DES1 and DES2 were measured at different temperatures (293.15-353.15 K) at atmospheric pressure, affording density values between 1.0~1.1 g/cm 3 that are similar to the density of water. The viscosities of DES1 and DES2 measured over the same temperature range (293.15 K~353.15 K) at atmospheric pressure decreased as the temperature increased, with their 100~1000 mPa•S viscosity values similar to those exhibited by ionic liquids. DES3 was a solid at room temperature and therefore a glass transition temperature could not be measured.

Physical Interactions between the DES
Physical interactions between the components of DES1, DES2 and DES3 were analyzed using FT-IR spectra (see . As shown in Figure 4, in the infrared spectrum of LA, peaks near 3200 cm −1 were identified as the -OH stretching vibration peak. The peaks near 2870 cm −1 and near 2960 cm −1 were identified as the symmetric and antisymmetric stretching vibration of -CH3. The peak near 2850 cm −1 and 2930 cm −1 were considered as symmetrical and anti-symmetric stretching vibration peaks stretching vibration peaks of methylene (-CH2-). The peak near 1722 cm −1 was regarded as carbonyl group (-C=O) stretching vibration peak, and peak near 1500 cm −1 was identified as the antisymmetric stretching vibration peak of carboxyl group (-COO-). In the infrared spectrum of TBAB, the peaks near 2870 cm −1 and near 2960 cm −1 were identified as the symmetric and antisymmetric stretching vibration of -CH3. The peak near 2850 cm −1 and 2930 cm −1 were considered as symmetrical and anti-symmetric stretching vibration peaks stretching vibration peaks of methylene (-CH2-). The peak around 1200~900 cm −1 was regarded as stretching vibration peak of CN. As for DES1, the infrared spectrum was almost the superposition of LA and TBAB, except the hydroxyl group (-OH) near 3200 cm −1 that became broader and larger, red-shifted to 3000 cm −1 . It could be the evidence that hydrogen bonds could be formed by LA and TBAB. As shown in Figure 5a, in the infrared spectrum of EG, peaks near 3200 cm −1 were identified as the -OH stretching vibration peak. The peak near 2850 cm −1 and 2930 cm −1 were considered as symmetrical and anti-symmetric stretching vibration peaks stretching vibration peaks of methylene (-CH 2 -). The peak around 1050~1150 cm −1 was regarded as stretching vibration peak of -COH. In the infrared spectrum of DES2, the red-shifted hydroxyl group (-OH) peak near 3200 cm −1 (shown in Figure 5b) suggested that hydrogen bonds could be formed by TBAB and EG. As shown in Figure 6a, in the infrared spectrum of ChCl, peak near 3200 cm −1 were identified as the -OH stretching vibration peak. The peaks near 2870 cm −1 and near 2960 cm −1 were identified as the symmetric and antisymmetric stretching vibration of -CH3. The peak near 2850 cm −1 and 2930 cm −1 were considered as symmetrical and anti-symmetric stretching vibration peaks stretching vibration peaks of methylene (-CH2-). The peak around 1200~900 cm −1 was regarded as stretching vibration peak of CN. In the infrared spectrum of DES3, the red-shifted hydroxyl group (-OH) peak near 3200 cm −1 (shown in Figure 6b) suggested that hydrogen bonds could be formed by ChCl and LA. Therefore, hydrogen bond was proved to exist in the three DES because of active hydrogen atoms in hydrogen bond donor and high electron density of hydrogen bond acceptors.

Vapor-Liquid Equilibrium and Selection Profile of the DES
Vapor-liquid equilibrium experiment for benzene-cyclohexane-DES system was conducted experiment at atmospheric pressure. The molar concentration of extractant was 0.1. The results were listed in Table 2. Specifically, X3 is the molar fraction of low eutectic solvent in the liquid phase, X1 is the molar fraction of cyclohexane in the liquid phase at the vapor-liquid equilibrium, X'1 is the molar fraction of cyclohexane in the liquid phase at the vapor-liquid equilibrium (without DES content), Y1 is the molar fraction of cyclohexane in the gas phase at the vapor-liquid equilibrium, and T is the temperature at the vapor-liquid equilibrium. Furthermore, relative volatility (αij) and selectivity factor (S) were calculated according to the following Equations (1) and (2): where, Y is the molar fraction of the vapor phase, X is the molar fraction of the liquid phase, and the subscripts i and j referred to the most volatile component (cyclohexane) and the less volatile component (benzene), respectively. The Y1-X1 and α-X1 graphs derived from the experimental data were shown in Figures 7 and 8. As can be seen from Figure 7, when DES1 is added to the mixture system, the euboiling-point disappears. The DES2 or DES3 mixture has the same boiling point as the benzene-cyclohexane binary system. The experiment shows that DES2 and DES3 have a weak interaction with benzene and cyclohexane without changing the azeotropic point, DES1 has strong interaction with benzene or cyclohexane, which can break the azeotrope. Therefore, DES1 can be used for extractive distillation of benzene-cyclohexane. It is known that azeotropic point occurs when the value of the relative volatilities of the components of a distillate are equal to 1. If the azeotrope was broken, the relative volatility should be greater than 1. On the basis of Figure 8, the relative volatility was greater than 1 in the full concentration range only when extraction agent DES1 was added.

VLE Data Correlation
In this section, a continuous extraction process for separating benzene from cyclohexane was preliminarily simulated in Aspen Plus V8.4. UNIQUAC, Wilson and NRTL local composition activity coefficient models were used to fit the obtained VLE data. The correlation index R2 was calculated as the following Equation (3): where yi and ui are the experimental data and estimated values of the vapor phase molar fractions, respectively, and y is the average value of experimentally determined vapor phase molar fraction data. From Table 3, it can be seen that coefficient of determination values (R 2 ) of the three systems ranked as NRTL > UNIQUAC > Wilson. The R 2 of the same model for the three systems was found to be in order of DES3 > DES2 > DES1, indicating that the model application degree of the system was changed by extractant. The better the extraction effect, the lower the model application degree. Next, the vapor-liquid equilibrium data were simulated by NRTL model regression module, and the summary data were shown in Table 4. The phase diagram of Y1-X1 fitted by the cyclohexane(1)-benzene(2)-DES(3) system and the NRTL model is shown in Figures  9-11. It can be seen that the simulated data were very close to the experimental data.

Mechanism Analysis of Extractive Distillation
The mechanism of the extractive distillation process was investigated using FTIR. FT-IR spectra of DES and benzene were shown in Figures 12-14. The FT-IR spectra of DES1, DES2 and DES3 have been analyzed in  In the infrared spectrum of benzene, the peak at 3030 cm −1 was regarded as the C-H stretching vibration peak. Peaks round 1500~1800 cm −1 were assigned as the benzene ring skeleton vibration peak. The peak near 1000 cm −1 was attributed as in-plane C-H bending vibration. As can be seen in Figure 12, compared with DES1, in the DES1 + benzene mixture, the C=O peak band near 1722 cm −1 became narrow, which indicates π-π bond was formed between DES1 and benzene. At the same time, the hydroxyl stretching peak band near 3200 cm −1 became narrowed, which indicates that the hydrogen bond between DES1 and benzene was formed, while the inner hydron bond of DES1 was broken. These results are consistent with the previous study of Hou [14], who reported that similar DES entrainers acted through π-π interactions with the aromatic ring of toluene. Due to the hydrogen bond and π-π bond between DES1 and benzene, the separation of the mixture is achieved. Conversely, comparison of the FT-IR spectra of DES2 and DES3, benzene and DES2/DES3 + benzene (see Figures 13 and 14) reveal no change in the width and intensities of their OH and C=O absorptions, which demonstrated the vapor-liquid equilibrium data.

Conclusions
In this paper, DES1 (1:2 mix of tetrabutylammonium bromide (TBAB) and levulinic acid (LA)), DES2 (1:2 mix of TBAB and ethylene glycol (EG)) and DES3 (1:2 mix of ChCl and LA) were used as entrainers for the separation of benzene and cyclohexane. Based on the VLE data of benzene−cyclohexane−DESs, DES1 has been used as an entrainer for the successful extractive distillation of azeotropic mixtures of aromatic and non-aromatic hydrocarbons. In addition, the NRTL model was applied to correlate the experimental data, and the results show that it has good agreement with the experimental data. Therefore, a continuous extraction process for separating benzene from cyclohexane was preliminarily simulated in Aspen Plus V8.4. FT-IR spectra analysis revealed that the entrainment process used to break the azeotrope relies on selective hydrogen bonding and interactions between the LA component of the DES and benzene during the distillation process. This proves that DES1 is a very promising solvent for the separation of benzene−cyclohexane mixtures.