Journal of Molecular Liquids 316 (2020) 113877

Contents lists available at ScienceDirect

Journal of Molecular Liquids

journal homepage: www.elsevier.com/locate/molliq

Extraction of ethanol from mixtures with n-hexane by deep eutectic

solvents of choline chloride + levulinic acid, + ethylene glycol, or +
malonic acid
Eun Jin Sa a, Bong-Seop Lee b,⁎, Byung Heung Park a,⁎
a
Department of Chemical and Biological Engineering, Korea National University of Transportation, Chungju, Chungbuk 27469, Republic of Korea
b
Department of Chemical Engineering, Kangwon National University, Chuncheon, Kangwon 24341, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: The separation of azeotropic mixtures is a great challenge for concentrating a desired chemical compound be-
Received 12 March 2020 cause it cannot be attained by a simple distillation process. Liquid-liquid extraction can be an alternative process
Received in revised form 16 July 2020 to resolve the separation problem if an appropriate solvent is selected to effectively partition the mixture. In this
Accepted 18 July 2020
work, the feasibilities of three different deep eutectic solvents (DESs) for breaking azeotropes of n-hexane and
Available online 21 July 2020
ethanol mixture were examined by means of liquid-liquid extraction. Choline chloride (ChCl) was chosen as a
Keywords:
base compound and three hydrogen bond donors (HBDs) of levulinic acid, ethylene glycol, and malonic acid
Deep eutectic solvent were added at their eutectic ratio to form DESs. Liquid-liquid equilibrium experiments for an azeotropic n-
Liquid-liquid equilibria hexane + ethanol system were carried out to determine binodal compositions and, then, evaluate the composi-
N-hexane + ethanol tions at both liquid phases at 298.2 K and atmospheric pressure by employing the prepared DESs. Refractive
Extraction index, density, and viscosity at the binodal compositions were measured to provide properties required for ex-
COSMO-SAC traction process establishment. The measured tie-lines were compared with the predicted results calculated by
COSMO-SAC model. Moreover, the ability for extracting ethanol was evaluated with distribution coefficient
and selectivity. It was found that the distribution coefficient and selectivity values are high enough to support
a scaling-up process by a liquid–liquid extraction.
© 2020 Elsevier B.V. All rights reserved.

1. Introduction industrial processes including petrochemical ones. In many cases, it is

necessary to obtain concentrated or pure ethanol from the mixtures.
In the last few decades, the demand for alternative fuels and renew- However, the separation of such azeotropic mixtures is a great challenge
able energy has significantly increased under the awareness of environ- and nearly impossible to achieve by a simple distillation process. To re-
mental concerns caused by the consumption of fossil fuels. Ethanol is solve the problem, advanced distillation techniques such as extractive
considered as a promising alternative fuel to the conventional fossil or azeotropic distillation [6,7] have been employed for recovering one
fuels and an important component to meet an increasing energy de- of the compounds beyond azeotropic composition. Most of these pro-
mand since biological processes are dominant in the production of eth- cesses require high pressure and/or temperature. In consequence of
anol from natural resources. Moreover, ethanol produced by biological the variation on the separation conditions, the advanced processes de-
means is commercially viable and profitable due to the advantages of mand a significant amount of energy and investment on equipment to
adopting various raw materials available for a feedstock. Ethanol can obtain a single fluid phase system. Alternatively, a liquid-liquid extrac-
be used as fuel in the internal combustion engine blended with gasoline tion process associated with a new solvent (entrainer) can be applied
[1,2] or as oxygenated additives for diesel to improve ignition and com- to separate the mixtures with greater energy efficiency. In the process,
bustion efficiency [3]. Furthermore, many researches have been carried the separation is usually carried out under ambient conditions after
out to directly use ethanol in fuel cells producing electric energy in an forming a liquid-liquid equilibrium (LLE) by the addition of a compo-
innovative energy sector [4,5]. nent interacting with the constituents of the azeotropic mixture [8].
The addition of ethanol to gasoline raises the vapor pressure because Ionic liquids (ILs) have emerged and drawn attention as an appeal-
ethanol produces azeotropes with many constituents of gasoline. Differ- ing replacement for conventional organic solvents because common in-
ent azeotropic mixtures of ethanol and alkanes are also found in various dustrial organic solvents such as tetraethylene glycol, propylene
carbonate, and sulfolane are usually toxic, volatile and flammable. ILs
⁎ Corresponding authors. are a kind of organic molten salts with properties of low melting
E-mail addresses: [email protected] (B.-S. Lee), [email protected] (B.H. Park). point, non-volatility, and inflammability [9]. The use of ILs has been

https://doi.org/10.1016/j.molliq.2020.113877
0167-7322/© 2020 Elsevier B.V. All rights reserved.
2 E.J. Sa et al. / Journal of Molecular Liquids 316 (2020) 113877

extended to various technologies including electrochemistry, catalysis, by COSMO-SAC thermodynamic method to elucidate the applicability of
synthesis, biotechnology, and separation process [10,11] supported by the model to the DES extraction process.
measured and calculated physico-chemical property data of mixtures
containing ILs [12,13]. The thermophysical properties of ILs can also 2. Experimental
make important contributions to azeotropic breaking by extractive dis-
tillation, liquid-liquid extraction, and supported liquid membranes [14]. 2.1. Materials
Azeotrope systems studied using ILs as separating agents cover various
mixtures [15–17]. As for ethanol + alkanes, ILs of ethyl sulfate [EtSO4] Choline chloride (ChCl) was dried in a vacuum chamber at 363 K for
and other anions combined with imidazolium-based cations have more than 6 h before mixing. The moisture content of the dried ChCl
been adopted for ethanol + n-hexane [18–21] and ethanol + n- was measured less than 0.5 wt% by a coulometric Karl Fischer titration
heptane [18,22,23] to present high separation efficiency. However, de- method using Metrohm 870 KF Titrino plus. The other chemicals were
spite their advantages over conventional organic solvents, the synthesis purchased from chemical suppliers and used without any further puri-
and preparation of ILs cost very high compared to molecular solvents, fication. Three different DESs were prepared by measuring ChCl and re-
which is regarded as one of the drawbacks associated with the commer- spective HBDs of levulinic acid, ethylene glycol, or malonic acid at their
cialization of chemical processes using ILs. In addition, the environmen- known eutectic mole ratios and mixing them under stirring in a gentle
tal friendliness of ILs is a disputable issue since their impact on the heating condition [31]. Recently, the definition of eutectic composition
environment depends on ionic components consisting of ILs [24,25]. for DES was closely examined and it was shown that the eutectic com-
Recently, low transition temperature mixtures (LTTMs) or deep eu- position is merely thermodynamic [32]. However, to date, the eutectic
tectic solvents (DESs), discovered in the early 2000s [26,27], have compositions were frequently given by molar ratios. The prepared
emerged as versatile alternatives to ILs. The physico-chemical proper- DESs were referred to as DES1 (ChCl:levulinic acid = 1:2), DES2
ties of DESs are compatible with those of ILs with regard to negligible (ChCl:ethylene glycol = 1:2), and DES3 (ChCl:malonic acid = 1:1), re-
volatility, high conductivity, inflammability, and superior chemical sta- spectively. A homogeneous and colorless liquid phase for each DES was
bility. DESs are generally composed of a quaternary ammonium salt and found after being melted. Details of the chemicals are summarized in
a hydrogen bond donor (HBD). In practice, various salts such as ammo- Table 1 with suppliers. The chemical compounds were weighed using
nium, phosphonium, and sulfonium salt can form different types of an analytical balance (Hansung, HS224S) having a precision of 10−3 g.
DESs depending on the nature of the complexing agent [28] but the ma-
jority of studies have focused on the quaternary ammonium halide salts. 2.2. LLE and physico-chemical property measurement
The mixed solvents were entitled as DESs because the melting point at
eutectic composition is 200 to 300 K lower than those of the individual The ternary LLE experiments were carried out for the determination
components. Hydrogen bonding between anion of salts and the of the binodal curves. Turbidimetric titration method was applied to ob-
hydrogen-donor moiety gives rise to the charge delocalization inducing tain the binodal data of all the n-hexane + ethanol + DES systems. The
the decrease of the melting point. In some researches, the solvents of experiments were performed in a thermostatic water bath at 298.2 K,
off-eutectic compositions were used but still referred to as DESs. DESs which was controlled within ±0.1 K. The temperature in an equilibrium
have some merits over ILs. High purity DESs can be easily prepared at cell was measured using a digital thermometer SDT25 from SUMMIT
relatively low cost and the preparation procedure is so simple that the (Korea) connected to a K-type thermocouple. The thermometer was
mixing of two precursors at the predetermined molar ratio under gentle calibrated against a standard mercury thermometer (0.1 K). In a glass
heating is enough to produce DESs. cell containing a magnetic stirrer, an immiscible binary mixture of n-
DESs have been successfully applied to azeotropic mixtures for ex- hexane + DES was loaded and the immiscible two liquids were stirred
traction of alcohol from alcohol + ester [29] and alcohol + alkane to make one to be dispersed as droplets in the second phase. Ethanol
[30] mixtures because they are able to dissolve hydrogen bond donor was added to the solutions until the ternary mixture became a homoge-
molecules. Therefore, when DES is applied to mixtures containing alco- neous phase. The miscibility change during the incremental addition of
hol, alcohol is readily dissolved into DES and other non-hydrogen-bond ethanol was monitored by observing the laser scattered image on a
donor molecules remain as a separated second phase in contact with the screen behind the cell. A binodal point was determined by calculating
DES-rich phase. the masses of components in the homogeneous solution. The uncer-
In this work, we carried out the liquid-liquid extraction experiments tainty of the binodal composition was estimated to 0.0003 for the
for recovering ethanol from mixtures with n-hexane by DESs consisting mass fraction. The refractive index, density, and viscosity of those ter-
of choline chloride and three different hydrogen bond donors (levulinic nary mixtures in a single liquid phase were determined in triplicate.
acid, ethylene glycol, and malonic acid). The physico-chemical proper- The tie-lines of LLE were also experimentally determined for the ter-
ties of solutions such as density and viscosity were measured to provide nary systems. Mixtures of known initial composition were prepared in
data for process development. The experimental results were calculated glass vials which were vigorously stirred in a thermostatic bath at

Table 1
Description of chemical compounds.

Compounds Supplier Remark

Choline chloride Sigma-Aldrich Purity ≥98%, dried before use

Levulinic acid Sigma-Aldrich Purity ≥99%

Ethylene glycol Sigma-Aldrich Purity ≥99.8%

Malonic acid Sigma-Aldrich Purity ≥99%

Ethanol Sigma-Aldrich Purity ≥99.5%

n-Hexane Sigma-Aldrich Purity ≥99%
 E.J. Sa et al. / Journal of Molecular Liquids 316 (2020) 113877 3

298.2 K. The mixtures were severely stirred for at least 1 h and settled The Staverman-Guggenheim term for the combinatorial contribu-
down for at least 12 h in the thermostat bath to confirm the two distinct tion is considered by Eq. (6) with ϕI = xiqi / (∑i xiqi), θi = xiri / (∑i
liquid phases. The samples from the two separate phases were taken by xiri), li = z / 2(ri − qi) − (ri − 1) and z = 10. The ri and qi are the normal-
a syringe and physico-chemical properties including refractive index ized volume (ri = Vi / r) and surface area parameter (qi = Ai / q), respec-
were measured for both phases. tively, derived from the cavity volume Vi and the cavity surface area Ai.
The measurements of refractive indexes were carried out at 298.2 K All the required parameters are obtained from the COSMO calculation
and atmospheric pressure, using an automated Refractometer Abbemat [34,35].
200 from Anton Paar with resolution ±0.0001 and accuracy ±0.0001 of
ϕi z θ ϕ X
the refractive indexes. The densities were measured by using an oscillat- ln γi comb ¼ ln þ q ln i þ li − i xl ð6Þ
ing U-tube density meter (Rudolph Research Analytical, DDM2911) of xi 2 i ϕi xi j j j
which accuracy and temperature uncertainty are 5 × 10−5 g/cm3 and
0.03 K, respectively. Ultrapure water and dry air at atmospheric pres-
sure were applied to calibrate the instrument. The controllable built-in 4. Results and discussion
thermostat was available to maintain the temperature at 298.2 K. The
viscosities of the solutions were measured by Sine-Wave Vibro Viscom- The experimental results of binodal compositions were presented
eter SV-10 of A&D. The instrument was calibrated at two points of ter- from Tables 2 to 4 with measured refractive index, density, and viscosity
tiary distilled deionized water and viscosity standard solution (500 cP, of the solutions. The density of the DES-rich phase was greater than that
Brookfield) before experiments. The measurements are based on the of the hexane-rich phase and, thus, the extract formed a lower (bottom)
tuning-fork vibration method and, thus, care was taken to remove the phase at LLE experiments. DESs are usually denser and more viscous
external vibrations. The accuracy associated with the measured viscos- than conventional solvents. In general, the density and viscosity of mix-
ity data was within ±3%. The uncertainties of refractive index, density, tures which are composed of DES and a molecular solvent decreased
and viscosity were determined to ±0.0002, ±0.0002 g/cm3, and ± with the increase of the molecular solvent composition [36]. As shown
0.04 mPa·s/mPa·s, respectively. in Tables 2 to 4, the density and viscosity decrease from DES-rich to
hexane-rich phase. However, their relationship with composition in
3. COSMO-SAC model the ternary systems has not been clarified and needs to be further
investigated.
The equilibrium concentrations of components at a given tempera- The refractive indexes along the binodal compositions were fitted by
ture and pressure can be determined by a thermodynamic model. In the following equations to be applied for the determination of the tie-
the present work, COSMO-SAC [33] was used to calculate the composi- lines.
tions in two liquid phases at equilibrium. The segment activity coeffi-
cient from the conductor-like screening model is defined by the sum nD ¼ Aw1 þ Bw21 þ Cw31 þ Dw41 þ Ew2 þ Fw22 þ Gw32 þ Hw42 þ Iw3 ð7Þ
of residual (res) and combinatorial (comb) contributions. þJw23 þ Kw33 þ Lw43

ln γi COSMO−SAC ¼ ln γi res þ ln γi comb ð1Þ M þ Nw1 þ Ow21 þ Pw31 þ Qw41

w3 ¼ ð8Þ
R þ S expðTw1 Þ þ U expðVw2 Þ
The non-ideality of fluid due to differences in the electrostatic inter-
actions between molecules is described in the residual term and molec- w3 ¼ 1−ðw1 þ w2 Þ ð9Þ
ular interactions are described by the σ-profile, p(σ). The activity
coefficient from the residual contribution is defined by Coefficients A to V in Eqs. (7) and (8) are adjustable parameters to fit
nhb;OH;OT
X      t  the equations with the experimental data and w1, w2, and w3 corre-
ln γi res
¼ ni ∑t pti σ tm ln ΓtS σ tm − ln Γti σm ð2Þ spond to the mass fraction of n-hexane, ethanol, and DES, respectively.
σm
The equations were used to estimate compositions of the respective liq-
where Γts and Γti are activity coefficient segment in the solution s and in uid phase from experiments for tie-line measurement. The refractive
pure liquid i, respectively. The superscript t indicates a segment type
which is designated as the non‑hydrogen bond (nhb), the hydroxyl Table 2
(OH) group, and all other types of hydrogen bond donating and Experimental data for binodal composition, refractive index, density, and viscosity of n-
hexane(1) + ethanol(2) + DES1(3) ternary systems at 298.2 Ka.
accepting atoms (OT). The segment activity coefficient (SAC) of seg-
ment m is determined from Composition (wt. fraction) Measured properties
( ) w1 w2 nD ρ (g/cm3) η (mPa·s)
  X
nhb;OH;OT X        
lnΓtS σ tm ¼ − ln psS σ sn ΓsS σ sn exp −ΔW σ tm ; σ sn =RT ð3Þ 0.0092 0.0964 1.4569 1.0995 31.49
s σn
0.0149 0.1763 1.4475 1.0100 26.03
0.0172 0.2504 1.4369 0.9914 37.87
The term ΔW means the interaction energy between two surface 0.0155 0.2958 1.4312 1.0303 36.44
segments, and is expressed as the following equation: 0.0250 0.3497 1.4236 0.9878 21.89
0.0594 0.4199 1.4128 0.9375 8.373
   2   2 0.1025 0.4861 1.4014 0.8924 4.667
ΔW σ tm ; σ sn ¼ cES σ tm þ σ sn −chb σ tm ; σ sn σ tm −σ sn ð4Þ 0.1477 0.5122 1.3934 0.8597 0.901
0.1946 0.5137 1.3884 0.8351 0.974
where the electrostatic interaction parameter, cES, is a temperature- 0.2430 0.5144 1.3837 0.8120 1.455
dependent parameter (i.e., cES = AES + BES/T2) and the hydrogen bond 0.2990 0.4994 1.3817 0.7919 1.225
interaction parameter (chb) is given by 0.3687 0.4704 1.3774 0.7710 1.191
0.4537 0.4265 1.3748 0.7475 1.026
0.5775 0.3534 1.3743 0.7180 1.096
8 0.7286 0.2494 1.3698 0.6878 0.6765
>
> cOH−OH if s ¼ t ¼ OH and σ tm  σ sn b0 0.8086 0.1830 1.3698 0.6759 0.6362
  <
cOT−OT if s ¼ t ¼ OT and σ tm  σ sn b0 0.8725 0.1255 1.3707 0.6669 –
chb σ tm ; σ n s
¼ ð5Þ
>
> c if s ¼ OH; t ¼ OT; and σ tm  σ sn b0
: OH−OT a
Standard uncertainties u are: u(T) = 0.2 K; u(w) = 0.0003; u(nD) = 0.0002; u
0 otherwise (ρ) = 0.0002 g/cm3; ur(η) = 0.04 mPa·s/mPa·s.
4 E.J. Sa et al. / Journal of Molecular Liquids 316 (2020) 113877

Table 3 curves and the tie-lines were indicated in triangular diagrams as

Experimental data for binodal composition, refractive index, density, and viscosity of n-
Figs. 1 to 3 for the respective ternary system. The broad two-phase re-
hexane(1) + ethanol(2) + DES2(3) ternary systems at 298.2 Ka.
gion in the diagrams presented the great immiscibility between n-
Composition (wt. fraction) Measured properties hexane and DESs and applicability of the selected DESs to break the
w1 w2 nD ρ (g/cm3) η (mPa·s) azeotrope of n-hexane + ethanol system. The negative slope of tie-
0.0013 0.0464 1.4653 1.0957 35.31 lines implies that the extraction of ethanol into the DES-rich phase is fa-
0.0031 0.0932 1.4594 1.0801 28.34 vored. This behavior is explained by the strong interaction between eth-
0.0064 0.1357 1.4544 1.0638 21.68 anol and components of DES.
0.0105 0.2318 1.4422 1.0245 12.23 In Table 5 and Figs. 1 to 3, the experimental LLE data were compared
0.0156 0.3006 1.4325 0.9953 9.606
0.0216 0.3322 1.4291 0.9831 8.371
with calculation results obtained from COSMO-SAC [38,39]. Standard
0.0374 0.4036 1.4188 0.9501 6.120 deviation (σ) between experimental and calculated properties was de-
0.0608 0.4625 1.4088 0.9150 4.394 fined by the following equation [40] and used to estimate the accuracy
0.0841 0.4954 1.4030 0.8931 3.563 of the model.
0.1016 0.5123 1.3992 0.8790 3.207
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
m 2  exp 
0.1420 0.5334 1.3943 0.8589 2.756
n cal 2
0.1817 0.5457 1.3879 0.8297 2.065 ∑k¼1 ∑i wexp ik
−wcal ik Phase I þ wik −wik Phase II
0.2269 0.5460 1.3844 0.8125 1.600 σ ð%Þ ¼ 100 ð10Þ
2mn
0.2747 0.5318 1.3806 0.7934 1.430
0.3244 0.5081 1.3788 0.7791 1.320
0.3614 0.4908 1.3770 0.7674 1.160 where n is the number of tie-lines and m is the number of components.
0.4015 0.4700 1.3757 0.7591 1.067 The subscript i denotes component and k indicates tie-line.
0.4324 0.4524 1.3750 0.7838 0.8166 The standard deviations of COSMO-SAC modeling were calcu-
0.4762 0.4269 1.3737 0.7412 0.9524 lated to 6.66, 5.78, and 2.07% for DES1, DES2, and DES3 systems,
0.5283 0.3928 1.3724 0.7208 0.7214
0.6037 0.3420 1.3710 0.7143 0.5880
respectively. The model is a quantum chemistry-based thermody-
0.6608 0.3010 1.3704 0.7005 0.4568 namic model presenting a remarkable advance in the prediction
0.7383 0.2438 1.3696 0.6866 – and calculation of fluid phase equilibria. It contains only a few uni-
0.7836 0.2081 1.3692 0.6793 – versal parameters independent of species, which means the model
0.8469 0.1482 1.3696 0.6725 –
is predictive and no experimental data are required for the predic-
0.8946 0.1020 1.3704 0.6660 –
0.9422 0.0565 1.3713 0.6606 – tion of thermodynamic properties and phase equilibrium. Therefore,
a
the calculation results shown in Figs. 1 to 3 deviated from the exper-
Standard uncertainties u are: u(T) = 0.2 K; u(w) = 0.0002; u(nD) = 0.0002; u
(ρ) = 0.0002 g/cm3; ur(η) = 0.04 mPa·s/mPa·s.
imental data at the region of high ethanol composition (w2 N 0.5) in
the lower phase. At the low ethanol concentration region, COSMO-
SAC model satisfactorily regenerated the LLE behaviors for both
indexes for determining the composition of a liquid mixture had already
phases when compared to the experimental data. At high ethanol
been validated in ternary systems and the uncertainty in the composi-
concentration region (w2 N 0.5), the hydrogen-bonding phenomena
tion is estimated to be up to ±0.006 [30,37] and the similar uncertainty
in choline chloride, hydrogen bonding donor, and ethanol molecules
of ±0.006 was also estimated for the present work.
may exist more strongly than at the low ethanol concentration re-
The compositions of a raffinate (upper) and an extract (lower) phase
gion. Due to this reason, the predicted results with COSMO-SAC
using DESs to recover ethanol from the mixtures with n-hexane were
model show some deviation from experimental data. For the more
presented in Table 5. The LLE experiment results including the binodal
sophisticated description for hydrogen-bonding phenomena
COSMO-SAC model has been developing for a long time [33,41–44].
Table 4 The extraction efficiency is measured by the percentage of solute
Experimental data for binodal composition, refractive index, density, and viscosity of n- moving from one phase to the other. In practice, two parameters used
hexane(1) + ethanol(2) + DES3(3) ternary systems at 298.2 Ka. in the assessment of the extraction efficiency are the distribution coeffi-
Composition (wt. fraction) Measured properties cient and the selectivity. In this study, the distribution coefficient of eth-
w1 w2 nD ρ (g/cm3) η (mPa·s)
anol (β2) and the selectivity (S) defined by the weight fraction were
used as follows:
0.0001 0.1825 1.4584 1.1253 137.3
0.0050 0.2551 1.4489 1.0943 66.05
wII2
0.0129 0.3168 1.4391 1.0556 35.31 β2 ¼ ð11Þ
0.0207 0.3790 1.4301 1.0241 22.50 wI2
0.0245 0.4091 1.4250 1.0076 16.59
0.0327 0.4423 1.4195 0.9845 11.42 ! !
0.0496 0.4763 1.4144 0.9582 8.516 wI1 wII2
S¼ ð12Þ
0.0686 0.4975 1.4117 0.9338 6.889 wII1 wI2
0.0997 0.5303 1.4031 0.9009 4.737
0.1331 0.5535 1.3955 0.8723 3.788
0.1744 0.5538 1.3897 0.8515 2.726 The liquid phases were designated as I and II for the hexane-rich and
0.2249 0.5507 1.3855 0.8226 2.055 the DES-rich phase, respectively. The distribution coefficient gives a
0.2679 0.5459 1.3818 0.8027 1.631 measure of the affinity of the solute for the two liquid phases. The selec-
0.3016 0.5325 1.3795 0.7900 1.473
tivity which is defined as the ratio of the distribution coefficient of one
0.3441 0.5145 1.3772 0.7768 1.262
0.3812 0.5007 1.3751 0.7623 1.090 solute to the distribution coefficient of the other solute means the
0.4337 0.4750 1.3734 0.7470 0.8709 ease with which a solvent extracts the desired solute from a feed
0.4961 0.4395 1.3712 0.7303 0.7208 stream. High values of the distribution coefficient imply a high affinity
0.5218 0.4260 1.3700 0.7237 0.6601
of the DES for ethanol permitting lower solvent/feed ratios. High values
0.5885 0.3812 1.3688 0.7091 0.5309
0.6584 0.3217 1.3687 0.6969 0.4936 of selectivity indicate a reduced number of equilibrium stages in an ex-
0.7379 0.2584 1.3684 0.6838 – traction process.
0.8465 0.1510 1.3692 0.6704 – The distribution coefficients and the selectivity were plotted with
a
Standard uncertainties u are: u(T) = 0.2 K; u(w) = 0.0003; u(nD) = 0.0002; u the composition of ethanol in Figs. 4 and 5, respectively. As shown in
(ρ) = 0.0002 g/cm3; ur(η) = 0.04 mPa·s/mPa·s. Figs. 4 and 5, the distribution coefficient and the selectivity decrease
 E.J. Sa et al. / Journal of Molecular Liquids 316 (2020) 113877 5

Table 5
Compositions for LLE tie-line of ternary systems at 298.2 K.

Experimenta COSMO-SACb

Upper phase Lower phase β2 S Upper phase Lower phase

w1 w2 w1 w2 w1 w2 w1 w2

n-hexane(1) + ethanol(2) + DES1(3)

0.524 0.387 0.125 0.496 1.28 5.39 0.786 0.210 0.123 0.543
0.674 0.290 0.087 0.466 1.61 12.45 0.833 0.165 0.099 0.511
0.775 0.207 0.059 0.431 2.09 27.32 0.877 0.121 0.077 0.467
0.853 0.143 0.041 0.396 2.77 58.29 0.910 0.089 0.062 0.421
0.891 0.106 0.027 0.345 3.27 109.15 0.939 0.061 0.048 0.367
0.909 0.088 0.019 0.284 3.21 154.60 0.960 0.040 0.038 0.311
0.935 0.064 0.014 0.218 3.40 224.53 0.978 0.022 0.030 0.243
0.979 0.017 0.010 0.120 7.02 684.60 0.995 0.005 0.022 0.127
a b
Standard uncertainties u are: u(T) = 0.2 K; u(w) = 0.006. Standard deviation, σ = 6.66%.

Experimenta COSMO-SACb

Upper phase Lower phase β2 S Upper phase Lower phase

w1 w2 w1 w2 w1 w2 w1 w2

n-hexane(1) + ethanol(2) + DES2(3)

0.630 0.325 0.158 0.540 1.66 6.64 0.854 0.143 0.157 0.585
0.753 0.234 0.101 0.513 2.19 16.38 0.892 0.106 0.123 0.556
0.813 0.183 0.074 0.484 2.64 29.16 0.916 0.082 0.101 0.524
0.883 0.112 0.056 0.456 4.09 64.12 0.941 0.058 0.080 0.477
0.927 0.072 0.039 0.411 5.73 137.59 0.959 0.040 0.064 0.423
0.949 0.049 0.029 0.374 7.70 250.36 0.970 0.030 0.055 0.379
0.963 0.035 0.020 0.323 9.21 436.04 0.980 0.020 0.047 0.326
0.975 0.023 0.013 0.252 11.11 864.53 0.989 0.011 0.039 0.255
0.981 0.014 0.008 0.187 13.49 1683.33 0.995 0.005 0.036 0.189
a
Standard uncertainties u are: u(T) = 0.2 K; u(w) = 0.006. bStandard deviation, σ = 5.78%.

Experimenta COSMO-SACb

Upper phase Lower phase β2 S Upper phase Lower phase

w1 w2 w1 w2 w1 w2 w1 w2

n-hexane(1) + ethanol(2) + DES3(3)

0.811 0.188 0.140 0.550 2.92 16.99 0.766 0.233 0.075 0.551
0.850 0.144 0.086 0.525 3.65 36.11 0.820 0.179 0.058 0.511
0.874 0.126 0.065 0.502 3.99 53.98 0.844 0.156 0.050 0.488
0.892 0.108 0.046 0.469 4.36 85.42 0.872 0.128 0.042 0.456
0.897 0.103 0.033 0.432 4.20 115.24 0.892 0.108 0.036 0.428
0.902 0.098 0.023 0.390 3.98 159.16 0.913 0.087 0.031 0.395
0.928 0.071 0.014 0.338 4.80 316.23 0.940 0.060 0.024 0.342
0.961 0.039 0.008 0.289 7.42 854.77 0.961 0.039 0.019 0.286
0.979 0.020 0.002 0.205 10.10 6410.64 0.983 0.017 0.014 0.205
a b
Standard uncertainties u are: u(T) = 0.2 K; u(w) = 0.006. Standard deviation, σ = 2.07%.

w2 (EtOH)

0.0
1.0

0.1
0.9

0.2
0.8

0.3
0.7

0.4
0.6

0.5
0.5

0.6
0.4

0.7
0.3

0.8
0.2

0.9
0.1

1.0
0.0
w3 (DES1) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 w1 (Hexane)

Fig. 1. Binodal curves for LLE of n-hexane + ethanol (EtOH) + DES1 system with tie-lines; experimental (■) and COSMO-SAC calculation (□) results.
6 E.J. Sa et al. / Journal of Molecular Liquids 316 (2020) 113877

w2 (EtOH)

0.0
1.0

0.1
0.9

0.2
0.8

0.3
0.7

0.4
0.6

0.5
0.5

0.6
0.4

0.7
0.3

0.8
0.2

0.9
0.1

1.0
0.0
w3 (DES2) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 w1 (Hexane)

Fig. 2. Binodal curves for LLE of n-hexane + ethanol (EtOH) + DES2 system with tie-lines; experimental (■) and COSMO-SAC calculation (□) results.

as the composition of ethanol in the aliphatic-rich phase increases, ethanol from mixtures with n-hexane would be preferred over the
which indicates that the extraction of ethanol from n-hexane to DES other DESs.
more preferable at low concentration of ethanol. As for the distribution The comparisons of S with literature data [21,37,40,45] for other DES
coefficient, DES2 exhibited the highest value compared to the other and IL solvents were given in Fig. 6. The values of S change with exper-
DESs. DES2 adopted an alcoholic product as HBD while the other DESs imental conditions and it is difficult to find data at the same tempera-
were made from acidic HBDs. Therefore, it was supposed that the ture and composition with the present work. The data shown in Fig. 6
ethylene glycol in DES2 attracts n-hexane as well as ethanol into the were collected at around 2 wt% of ethanol composition in the n-
DES-rich phase, which is reflected to high distribution coefficient and hexane rich phase and at temperatures close to 298.2 K. Among the sys-
relatively low selectivity. The results obtained using DES3 (ChCl:malonic tems compared, DES systems were found to be better than IL systems.
acid = 1:1) were better than DES1 (ChCl:levulinic acid = 1:2). In gen- From Fig. 6, it can be also noticed that the efficiency of the studied
eral, higher solute distribution coefficients are more profitable three DESs in the separation of n-hexane + ethanol is comparable to
since lower solvent-to-feed ration could be attained using extract other DESs. The literature data presented in Fig. 6 suggested that the se-
agents with high distribution coefficients, which implies lower en- lectivity would be changed by temperature. However, the effect of tem-
ergy requirements. Therefore, the usage of DES2 for the recovery of perature is different depending on the systems. As for the [ChCl][oxalic

w2 (EtOH)

0.0
1.0

0.1
0.9

0.2
0.8

0.3
0.7

0.4
0.6

0.5
0.5

0.6
0.4

0.7
0.3

0.8
0.2

0.9
0.1

1.0
0.0
w3 (DES3) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 w1 (Hexane)

Fig. 3. Binodal curves for LLE of n-hexane + ethanol (EtOH) + DES3 system with tie-lines; experimental (■) and COSMO-SAC calculation (□) results.
 E.J. Sa et al. / Journal of Molecular Liquids 316 (2020) 113877 7

16
acide] system, selectivity was increased with temperature while [ChCl]
[malonic acid] presented decreased selectivity as temperature was in-
14
DES1 creased. Therefore, it is not easy to select a best solvent by simply com-
12 DES2 paring the selectivity and the distribution coefficient because the values
DES3 are depending on temperature and composition of feed for extraction.
10 In addition, it should be careful to quantitatively compare DES systems
since S is very sensitive to ethanol composition in low composition re-
8 gions as shown in Fig. 5.
β2

6 5. Conclusions

4 In the present work, three different DESs were examined as extracting

agents for breaking the azeotropic mixture of n-hexane and ethanol. The
2
DESs for this study were prepared based on choline chloride (ChCl) with
different HBDs containing hydroxyl or carboxylic group; DES1 (ChCl:
0
0.0 0.1 0.2 0.3 0.4
levulinic acid = 1:2), DES2 (ChCl:ethylene glycol = 1:2), and DES3
(ChCl:malonic acid = 1:1).
w2
Binodal compositions of ternary mixtures were determined for
the three systems of n-hexane + ethanol + DES at 298.2 K and atmo-
Fig. 4. Solute distribution coefficient measured from LLE experiments of n-hexane +
spheric pressure. The physico-chemical properties such as density,
ethanol + DES system with ethanol composition in n-hexane rich phase.
viscosity, and refractive index were measured for the homogeneous
phases. LLE data were obtained for the ternary systems of n-hexane
+ ethanol + DES based on the binodal data. The solute distribution
coefficient and selectivity have been calculated and compared to
other literature LLE data. As for the distribution coefficient, DES2
would be favored over other DESs for extracting ethanol. The high
10000
value of density and low value of viscosity of DES2 also suggest
that DES2 will be profitable when adopted to an extraction process.
DES1 The tie-line compositions of LLE, distribution coefficient, and selec-
DES2
1000 DES3 tivity data were predicted by COSMO-SAC thermodynamic model
which required no adjustable parameters. The calculation results
were satisfied with experimental data at region of low ethanol
S

concentration. However, it was found that the prediction of

100
COSMO-SAC overestimated n-hexane composition at high ethanol
concentration (w2 N 0.5) in DES-rich phase.
The evaluated distribution coefficient and selectivity values obtained
10 from ternary LLE experiments adopting the selected DESs are high
enough to support a scaling-up process by liquid-liquid extraction com-
pared to other DESs and ILs. The efficiency of extraction using the DESs
would be expected to be improved by changing the temperature condi-
0.0 0.1 0.2 0.3 0.4
tion. The experimental data obtained in this study also show that the
w2 DESs can be viewed as alternatives to ILs in the separation of the mix-
ture of n-hexane and ethanol considering the unique characteristics of
Fig. 5. Selectivity values measured from LLE experiments of n-hexane + ethanol + DES DESs such as low price, easy preparation, biodegradability, low toxicity,
system with ethanol composition in n-hexane rich phase. negligible vapor pressure, and recyclability.

CRediT authorship contribution statement

Eun Jin Sa: Data curation, Formal analysis, Investigation, Resources,

Validation. Bong-Seop Lee: Conceptualization, Methodology, Formal
DES1, 298.2 K analysis, Funding acquisition, Supervision, Validation, Writing - original
DES2, 298.2 K
draft. Byung Heung Park: Conceptualization, Methodology, Formal
DES3, 298.2 K
[ChCl][glycolic acid], 298.2 K [45]
analysis, Writing - original draft, Writing - review & editing.
[ChCl][lacic acid], 298.2 K [45]
[ChCl][oxalic acid], 293.2 K [40]
[ChCl][malonic acid], 293.2 K [40] Declaration of competing interest
[ChCl][oxalic acid], 313.2 K [40]
[ChCl][malonic acid], 313.2 K [40]
[MMim][MeSO4], 298.2 K [37]
The authors declare that they have no known competing financial
[BMim][NTf2], 298.2K [21] interests or personal relationships that could have appeared to influ-
[BMpy][NTf2], 298.2 K [21] ence the work reported in this paper.
[BMpy][TfO], 298.2 K [21]

0 1 2 3 4 5
Acknowledgment
log S
This research was supported by Basic Science Research Program
Fig. 6. Comparison of selectivity with different extracting solvent at around 2 wt% of through the National Research Foundation (NRF) of Korea funded by
ethanol composition in n-hexane rich phase. the Ministry of Education (Grant No. NRF-2020R1I1A2069224).
8 E.J. Sa et al. / Journal of Molecular Liquids 316 (2020) 113877

References [22] A.B. Pereiro, A. Rodríguez, Sep. Purif. Technol. 62 (2008) 733.
[23] A. Pereiro, E. Tojo, A. Rodriguez, J. Canosa, J. Tojo, Green Chem. 8 (2006) 307.
[1] A.E. Farrell, R.J. Plevin, B.T. Turner, A.D. Jones, M. O'hare, D.M. Kammen, Science 311 [24] M. Petkovic, K.R. Seddon, L.P.N. Rebelo, C.S. Pereira, Chem. Soc. Rev. 40 (2011) 1383.
(2006) 506. [25] S. Steudte, P. Stepnowski, C.-W. Cho, J. Thöming, S. Stolte, Chem. Commun. 48
[2] A.K. Agarwal, Prog. Energy Combust. Sci. 33 (2007) 233. (2012) 9382.
[3] N.M. Ribeiro, A.C. Pinto, C.M. Quintella, G.O. da Rocha, L.S. Teixeira, L.L. Guarieiro, M. [26] A.P. Abbott, G. Capper, D.L. Davies, R.K. Rasheed, V. Tambyrajah, Chem. Commun.
do Carmo Rangel, M.C. Veloso, M.J. Rezende, R. Serpa da Cruz, Energy Fuel 21 (2007) (2003) 70.
2433. [27] A.P. Abbott, D. Boothby, G. Capper, D.L. Davies, R.K. Rasheed, J. Am. Chem. Soc. 126
[4] C. Lamy, S. Rousseau, E. Belgsir, C. Coutanceau, J.-M. Léger, Electrochim. Acta 49 (2004) 9142.
(2004) 3901. [28] E.L. Smith, A.P. Abbott, K.S. Ryder, Chem. Rev. 114 (2014) 11060.
[5] S. Rousseau, C. Coutanceau, C. Lamy, J.-M. Léger, J. Power Sources 158 (2006) 18. [29] Z. Maugeri, W. Leitner, P.D. de María, Tetrahedron Lett. 53 (2012) 6968.
[6] R. Taylor, R. Krishna, Chem. Eng. Sci. 55 (2000) 5183. [30] F.S. Oliveira, A.B. Pereiro, L.P. Rebelo, I.M. Marrucho, Green Chem. 15 (2013) 1326.
[7] F.J. Novita, H.-Y. Lee, M. Lee, Korean J. Chem. Eng. 35 (2018) 926. [31] K.-S. Kim, B.H. Park, J. Mol. Liq. 254 (2018) 272.
[8] C. Hanson, Recent Advances In Liquid-liquid Extraction, Elsevier, 2013. [32] M.A. Martins, S.P. Pinho, J.A. Coutinho, J. Solut. Chem. 48 (2019) 962.
[9] R.D. Rogers, K.R. Seddon, Science 302 (2003) 792. [33] C.M. Hsieh, S.I. Sandler, S.T. Lin, Fluid Phase Equilib. 297 (2010) 90.
[10] R. Sheldon, Chem. Commun. (2001) 2399. [34] T. Banerjee, M.K. Singh, A. Khanna, Ind. Eng. Chem. Res. 45 (2006) 3207.
[11] N.V. Plechkova, K.R. Seddon, Chem. Soc. Rev. 37 (2008) 123. [35] L. Yang, S.I. Sandler, C.J. Peng, H.L. Liu, Y. Hu, Ind. Eng. Chem. Res. 49 (2010) 12596.
[12] B.H. Park, Korean J. Chem. Eng. 33 (2016) 2191.
[36] A.R. Harifi-Mood, R. Buchner, J. Mol. Liq. 225 (2017) 689.
[13] R. Soleimani, A.H.S. Dehaghani, N.A. Shoushtari, P. Yaghoubi, A. Bahadori, Korean J.
[37] A.B. Pereiro, A. Rodríguez, Chem. Eng. J. 153 (2009) 80.
Chem. Eng. 1 (2018).
[38] B.-S. Lee, S.-T. Lin, Chem. Eng. Sci. 121 (2015) 157.
[14] A. Pereiro, J. Araújo, J. Esperança, I. Marrucho, L. Rebelo, J. Chem. Thermodyn. 46
(2012) 2. [39] B.-S. Lee, S.-T. Lin, AICHE J. 63 (2017) 3096.
[15] N. Calvar, B. González, E. Gómez, Á. Domínguez, J. Chem. Eng. Data 51 (2006) 2178. [40] A.A. Samarov, M.A. Smirnov, M.P. Sokolova, E.N. Popova, A.M. Toikka, Fluid Phase
[16] C. Jork, C. Kristen, D. Pieraccini, A. Stark, C. Chiappe, Y. Beste, W. Arlt, J. Chem. Equilib. 448 (2017) 123.
Thermodyn. 37 (2005) 537. [41] S.-T. Lin, S.I. Sandler, Ind. Eng. Chem. Res. 41 (2002) 899.
[17] A. Pereiro, J. Canosa, A. Rodriguez, Fluid Phase Equilib. 254 (2007) 150. [42] S. Wang, S.I. Sandler, C.C. Chen, Ind. Eng. Chem. Res. 46 (2007) 7275.
[18] A. Pereiro, F. Deive, J. Esperança, A. Rodríguez, Fluid Phase Equilib. 291 (2010) 13. [43] W.-L. Chen, S.-T. Lin, Phys. Chem. Chem. Phys. 19 (2017) 20367.
[19] A.B. Pereiro, A. Rodriguez, Green Chem. 11 (2009) 346. [44] C.-K. Chang, W.-L. Chen, D.T. Wu, S.-T. Lin, Ind. Eng. Chem. Res. 57 (2018) 11229.
[20] A. Pereiro, A. Rodríguez, Fluid Phase Equilib. 270 (2008) 23. [45] N.R. Rodriguez, B.S. Molina, M.C. Kroon, Fluid Phase Equilib. 394 (2015) 71.
[21] S. Corderí, B. González, J. Chem. Thermodyn. 55 (2012) 138.