Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

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

Solvent selection for extractive distillation

processes to separate close-boiling polar systems

Lisette M.J. Sprakel, Peter Kamphuis, Anna L. Nikolova, Dylan J. Keijsper,

Boelo Schuur ∗
University of Twente, Faculty of Science and Technology, Sustainable Process Technology Group, The Netherlands

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

Article history: Solvent selection is key in extractive distillation process development and solvent effects are
Received 2 November 2018 often predicted based on the activity coefficients at infinite dilution. For close-boiling polar
Received in revised form 23 January systems with strong or specific interacting species, standard simulation tools, e.g. using
2019 UNIFAC or COSMO-RS, often predict poor as the activity coefficients at infinite dilution not
Accepted 27 January 2019 always reflect the selectivity in the process. For these systems, a heuristic solvent selection
Available online 5 February 2019 method in which molecular properties such as acidity, hydrogen bonding and polarity are
applied is desired as first estimate in the solvent selection. To explore the key parameters
Keywords: for such a first selection, solvent effects on the relative volatility (˛) were measured for three
Extractive distillation different industrially relevant polar mixtures, valeric acid – 2-methylbutyric acid, diethyl-
Solvent selection methylamine – diisopropylether, and 2-butanol – 2-butanone. For each of the cases the effect
Solvent effect of potential solvents on ˛ was measured in an ebulliometer. For the acids, the difference
Solvent in pKa of 0.1 was too small to separate based on acidity with a moderately basic solvent.
Stronger basic solvents resulted in thermal and chemical instability. Although the solvent
methyl-2-methyl butyrate is not suitable as a solvent because of reactivity, this structurally
similar solvent showed selectivity, indicating also in extractive distillation the like dissolves
like phenomenon can be applied to induce selectivity. A larger difference in basicity of the
mixture components (amine–ether mixture) and a difference in hydrogen bonding affinity
between the mixture components (ketone–alcohol mixture) allowed for increasing ˛ based
on differences in acidity and hydrogen bonding, respectively.
© 2019 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1. Introduction separation of azeotropic mixtures (Doherty and Knapp, 2004; Graczova

and Vavrusova, 2018; Kockmann, 2014). In extractive distillation a sol-
The climate target of the European Union for the year 2030 involves vent is applied that interacts with a preferred affinity for one of the
40% cuts in greenhouse gas emissions and a 27% improvement in mixture components, by which the relative volatility of the mixture is
energy efficiency (European-Council, 2014). In 2016 roughly 25% of the changed or an azeotrope is ‘broken’ (Doherty and Knapp, 2004). For a
total energy within the EU was used for industrial purposes (Eurostat, wide range of mixtures from very polar to apolar extractive distillation
2018). Hence, to reach the goals for 2030, reducing the industrial energy has been used, e.g. ethanol–water mixtures using either ethylene glycol
demand has an enormous potential. Because distillation is the most (EG) (Kiss and Ignat, 2012) or ionic liquids as entrainers (Kulajanpeng
widely applied separation process for homogeneous fluid separations et al., 2014; Quijada-Maldonado et al., 2013), ethyl acetate–ethanol
(Smith and Jobson, 2000), including many mixtures that are diffi- mixtures (Berg and Ratanapupech, 1986; Li et al., 2009), mixtures of
cult to separate by distillation due to pinch-points or close boiling monochloroacetic acid and dichloroacetic acid (Jongmans et al., 2012b),
points, energy-saving opportunities include improving these difficult mixtures of ethylbenzene and styrene (Jongmans et al., 2012a) and
distillation processes. Extractive distillation is a technology that can mixtures of aromatics/aliphatics (Canales and Brennecke, 2016).
potentially improve the separation of close-boiling mixtures and enable The choice of solvent is key for the viability of an extractive
distillation process, and solvent selection can be an expensive and

∗
Corresponding author at: Drienerlolaan 5, Meander 221, 7522 NB Enschede, The Netherlands.
E-mail address: [email protected] (B. Schuur).
https://doi.org/10.1016/j.cherd.2019.01.024
0263-8762/© 2019 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
124 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134

time-consuming procedure. Next to manipulating the relative volatil- Three specific industrially relevant binary mixtures were chosen
ity in the extractive distillation process, it is essential that the solvent that cover a significant part of the field of polar compounds in terms
is regenerated easily to recycle the solvent back to the extractive of functional groups and each show different behavior regarding their
distillation column. In apolar systems the solvent typically induces relative volatility, i.e. the mixture of A) diethylmethylamine (DEMA)
positive deviations from Raoult’s law, whereas in polar systems, such as and diisopropylether (DIPE), B) valeric acid (HVal) and its isomer 2-
monochloroacetic acid and dichloroacetic acid (Jongmans et al., 2012b), methylbutyric acid (2MBA) and C) 2-butanol (2-BuOH) and 2-butanone
the solvent induces negative deviations. In this work the focus is on (2-BuO). Next to differences in functional groups, for each case there
polar close-boiling mixtures, and solvents that have an attractive effect are other differences regarding boiling point, and acidity (or basicity)
on the high-boiling component, thereby inducing negative deviations of the components, see Table 1.
from Raoult’s law. The combination of being effective in the extractive The case with HVal and 2MBA was chosen to study on the lim-
distillation and being regenerable requires a certain affinity window iting pKa for separation based on extractive distillation, in which
that is unfortunately not straight forward to define. Ideally, extensive results were compared to the separation of monochloroacetic and
experimental screening of solvents is avoided, which might be realized dichloroacetic acid reported by Jongmans et al. (2012b) (Tboil , of 5 ◦ C,
by making use of known solvent scales to characterize intermolecu- and pKa of 1.6). HVal is an important industrial building block (DOW,
lar affinity between solvent and mixture constituents. An example is 2014) that is produced from 1-butene, where impurities lead to produc-
the proton affinity scale, which is a scale that describes the basicity tion of 2MBA (Bahrmann et al., 1997; DOW, 2014; Kubitschke et al., 2000).
of solvents, where the proton affinity is equal to the enthalpy that The components in the case with DEMA–DIPE are widely applied in
is measured during the deprotonation reaction of the base in the gas industry as building blocks (Roose et al. (2015)) and as solvents (Sakuth
phase (Laurence and Gal, 2009). Stronger bases therefore have a larger et al., 2010). DEMA and DIPE are an ether and an amine and have there-
proton affinity. For this scale, as well as for other scales applied for fore a different acidity. This case was chosen because of the very low
solvent characterization, the databases that are created list properties relative volatility of the mixture which makes distillation infeasible.
of the solvents based on their interactions with a (scale-)specific com- The case with 2-BuOH–2-BuO (relatively large Tb ) involves two com-
ponent. Other scales often applied for solvent characterization are the ponents with different functional groups (hydrogen bond donating and
BF3 -affinity-scale (Laurence and Gal, 2009; Laurence et al., 2011) and the accepting for 2-BuOH, and only hydrogen bond accepting for 2-BuO).
hydrogen bond basicity scale pKHBX (Laurence and Gal, 2009; Laurence For each of these cases solvents were selected from different classes
et al., 2011). For most of these scales the properties or interactions and the effect on the relative volatility was measured and related to
of solvents with the components need to be determined experimen- the molecular structures, interactions and activity coefficients of the
tally, but also parametrizations based on the linear solvation energy components in the mixtures. Aiming at further increasing the under-
relationship (LSER) (Kamlet et al., 1983; Taft et al., 1985) are applied, standing of affinities in the systems, isothermal titration calorimetry
using solvatochromic and/or Kamlet–Taft parameters (Bízek et al., 1993; (ITC) was applied to measure the heat of interaction and excess
Meyer and Maurer, 1995). enthalpies of mixing.
The direct prediction of the solvent effect on the relative volatil-
ity of a mixture that is to be separated can also be based on infinite
dilution activity coefficients (Bastos et al., 1985; Lei et al., 2003; Seader
2. Materials and methods
et al., 2011). These predictions are mainly applied on apolar systems of
alkanes/alkenes or alkane/aromatics, but also acetone/methanol mix- 2.1. Chemicals
tures (Kossack et al., 2008; Schult et al., 2001) and for separation of
hydrocarbons with ionic liquids (Zhu et al., 2007). Software for solvent Chemicals were obtained from Sigma–Aldrich, unless
selection and prediction of VLE behavior in extractive distillation is also mentioned otherwise, i.e. acetic acid (≥ 99.5%), Acetone
®
widely applied in which different types of chemical calculations can Lichrosolv (≥ 99.8, Merck KGaA), beta-cyclodextrin (≥ 98%,
be combined. Examples are COSMO-RS (Eckert and Klamt, 2002), UNI- Acros Organics), 1,8-bis(dimethylamino)naphthalene (99%,
FAC (Fredenslund et al., 1975) and Computer Aided Molecular Design
Proton Sponge® ), butanoic acid (99%, HBu), 2-butanol (>99.5%,
(CAMD) on the basis of the selectivity at infinite dilution with soft-
® ® 2-BuOH), 2-butanone (>99.5%, 2-BuO), m-cresol (≥ 98%),
ware packages ProCAMD , a tool that is a part of the ICAS framework
18-crown-6 (TCI-GR), cyclohexanol (≥ 99%), cyrene (99.0%),
(Harper and Gani, 2000). These predictive methods based on activity
coefficients are successfully applied for solvent selection for separation
dibutylamine (≥ 99.5%, DBA), dibutylether (≥ 99.3%, DBE),
of various mixtures, including C8 – aromatics mixtures, C4 mixtures, diethylene glycol (≤ 100%, Merck KGaA, DEG), diethylene gly-
aromatics and non-aromatics, ethanol–water, ILs, acids, oxychemicals, col dibutylether (≥ 99%, DEGDBE), diethylene glycol dimethyl
and systems of ketone, esters and alcohols (Brignole et al., 1986; Lei ether (≥ 99.5%, diglyme), N,N-diethylmethylamine (≥ 98%,
et al., 2003; Lek-utaiwan et al., 2011; Pretel et al., 1994). Acros Organics and TCI-GR, DEMA), diisopropylether (≥ 99%,
The consideration of thermodynamics, such as the selectivity at DIPE), DMSO (99.9%), dodecane (≥ 99%), ethylbenzene (Fluka,
infinite dilution, does however not always lead to the selection of the 98%), ethylene glycol (≥ 99.8%, EG), glycerol (99.0%), n-heptane
right solvent as the selectivity at infinite dilution is not in all cases anhydrous (99 +%, Alfa Aesar), isovaleric acid (99%, isoHVal),
related to the selectivity at finite dilution and a comparison is prob-
itaconic acid (≥ 99%), 2MBA (≥ 98%), 5-methyl-2-hexanone
lematic in case the distillate product changes (Kossack et al., 2008).
(5-m-2-one, 98.0%), methyl isobutyl ketone (≥ 99.7%, MIBK),
The selectivity only focuses on the extractive distillation column, but
methyl-2-methylbutyrate (≥ 98%, MeMeBu), methyl levulinate
for a viable process the regeneration column, e.g. the enthalpy of evap-
oration of the solvent, should also be included (Kossack et al., 2008). (98.0%, MeLev), octanoic acid (≥ 98%, HOct), 1-octanol (≥ 99%),
Predictions based on group contribution methods are also limited by oleic acid (≥ 99.5%, Fischer Scientific), phenol (≥ 99.5%,
the database they were fitted on and therefore the design of new PhOH), 1,2-propane diol (≥ 99%), propylene glycol (99.0%, PG),
solvents is not straightforward. There are also other cases in which tribenzylamine (≥ 99%, TBzA), tributylphosphate (99 +%, TBP),
application of methods based on infinite dilution activity coefficients 2,2,4-trimethylpentane (≥ 99.5%, iso-octane), trioctylamine
or group contribution methods is not straightforward, especially in the (98%, TOA), HVal (≥ 99%) and p-xylene (Honeywell, 99%).
case of very strong interactions and in case of specific interactions or
formation of azeotropes, which are all common in close-boiling sys-
tems of polar compounds. For these polar systems the prediction of
2.2. Experimental methods
VLE behavior is challenging, although a first indication for selecting
the right type of solvent for a specific separation is desired. Therefore, 2.2.1. Ebulliometer
in this work an experimental study is described that focuses on solvent Fischer Labodest VLE602 ebulliometers were used for the VLE
effects in extractive distillation of close-boiling polar compounds. experiments. The equilibrium temperature of the mixtures
 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134 125

Table 1 – Physical properties (boiling point Tb , melting point Tm , acidity pKa and octanol–water partitioning coefficients
(logP) of components of the three binary mixtures applied in this study.
Substance Abbreviation Tb (◦ C) Tm (◦ C) Structure Density (g/mL) pKa logP

Valeric acid HVal 187 −34 0.938 4.71 1.39

2-Methylbutyric acid 2MBA 177 −90 0.936 4.80 1.05

Diethylmethylamine DEMA 63-65 −196 0.72 3.65 (pKb ) 0.958

Diisopropylether DIPE 68.5 −60 0.725 ≈ 17 (pKb ) 1.52

2-Butanol 2-BuOH 99.1 −115 0.806 ≈ 16 0.683

2-Butanone 2-BuO 79.5 −86.7 0.805 ≈ 20 0.29

was measured with a Pt-100 thermocouple after setting the perature of injection and FID were 280 ◦ C, initial column
pressure. For each experiment the measurement cell was temperature 50 ◦ C with a linear increase after 1 min to 130 ◦ C
filled with 80 mL of sample, equilibration required between at 14 min, and after 1 min again a linear increase to 180 ◦ C
30–90 min, depending on the mixture and the temperature of in 1.3 min after which it was increased linearly to 190 ◦ C in
the previous measurement. After equilibration, 1 mL aliquots 3.3 min and then kept constant for 2 min. For the samples that
were taken from the liquid phase and the condensed vapor also contain octanoic acid (HOct) the procedure was extended
phase to determine the compositions. Solvent regeneration with a linear temperature increase from 190 ◦ C to 235 ◦ C in
was investigated by measuring binary VLE data of the high- 1.2 min followed by a last linear increase to 245 ◦ C in 3.3 min
boiling component of the mixture and the solvent. For each where the temperature was kept constant for 2 min. For the
case, the measurement and analysis errors were determined 2-BuOH–2-BuO mixture samples were analyzed with the pre-
as they also depend on the temperature and pressure at which viously mentioned method with the Varian GC, using heptane
the measurement is carried out. as an internal standard in this case. The injection and FID tem-
perature were at 280 ◦ C, initial column temperature of 45 ◦ C
with a linear increase after 4 min to 150 ◦ C at 15.5 min, after
2.3. Analytical methods
which there was a second linear increase to 250 in 4.5 min.

2.3.1. GC and GC–MS

For the DEMA–DIPE mixture samples were analyzed with a 2.3.2. Isothermal titration calorimetry (ITC)
Varian CP-3800 GC with FID (accuracy > 98%) and Agilent DB- ITC was performed using a TA Instruments TAM III
WAX column (60 m × 0.25 mm × 0.25 ␮m), using 100 L sample Microcalorimeter with 1 or 4 mL stirred sample vials at 20 ◦ C.
and 100 L internal standard (dodecane), diluted in acetone The automatic syringe was filled with 300 ␮L of titrant and
to a total of 1.7 mL. The temperature profile is as follows: connected with the sample cell through a cannula. The titrant
injection and FID at 280 ◦ C, initial measurement tempera- was injected periodically. For complex formation between
ture of 40 ◦ C with a linear ramp of 5 ◦ C/min after 1 min to HVal and TOA, the sequential reaction model described by
50 ◦ C, and after 1 min at 50 ◦ C, again a ramp in 3 min to Sprakel and Schuur (2018) was applied for fitting of the exper-
250 ◦ C at which the temperature was kept constant for 3 min. imental data from which the stoichiometry, enthalpy and
Analysis was also performed with 7890A Ms 5975C Agilent equilibrium constant of the reaction could be determined. The
GC–MS with FID and helium as carrier gas (accuracy > 98%) experimental data were corrected with the heat of injection of
and Agilent HP-5Ms, HP19191S-433 column. The sample con- the last injection of a measurement series.
centration was approximately 5% in methylisobutylketone
(MIBK). The injection and FID were kept at 280 ◦ C, the ini- 2.3.3. Error analysis
tial column temperature was 45 ◦ C with a linear increase For the study on HVal, the error in the equilibrium temperature
after 4 min to 160 ◦ C at a measurement time of 32 min, and measured was 0.4 ◦ C. Based on results of experiments in triplo,
after 2 min again a linear increase to 250 ◦ C at 47 min, after standard deviations were calculated for the composition of the
which it was kept constant for 2 min. For the HVal – 2MBA vapor and liquid phase. At 300 mbar the standard deviation
mixture samples of binary VLE experiments were analyzed in the liquid composition was 1.5 × 10−3 and 1.8 × 10−3 in
with a HPLC 1200 series with RID (accuracy > 99%), with either the vapour composition. At 900 mbar this was 2.1 × 10−3 and
an Agilent Hi-plex H+ Column (300 × 7.7 mm) or GROM Resin 3.1 × 10−3 , respectively. For the solvent screening VLE mea-
H+IEX (250 × 8 mm) and a 0.6 mL/min 5 mM H2 SO4 mobile surements the standard deviation of the GC analysis that was
phase. Samples of pseudo-binary VLE experiments were ana- performed in triplo was 3.2% for. For the case with DEMA-DIPE,
lyzed with a Varian CP-3800 GC with FID (accuracy > 98%) the standard deviation in the temperature equaled 0.4 ◦ C at
and VF1701ms (60 m × 0.25 mm × 0.25 ␮m) column. The tem- 1000 mbar and 0.7 ◦ C at 300 mbar. For ˛ the standard devia-
126 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134

ponents with and without the presence of the solvent (Doherty

and Knapp, 2004). In this equation Psat
i
is the saturated vapor
pressure of a component at a given temperature, yi is the vapor
mole fraction and xi is the liquid mole fraction.
 
i psat  
  i i
˛ij j psat j
j
Sij = s
= s ∼
= s
(1)
i psat i
˛ij i j
j psat
j

For DEMA–DIPE, the separation is complicated due to ˛

being close to unity. In those cases, generally there are posi-
tive deviations from Raoult’s law (i.e. both activity coefficients
greater than or equal to 1) and the ratio of activity coefficients
is larger than the ratio of saturated vapor pressures. Thus, see
Eq. (1), a solvent is desired that interacts ideally with the high-
Fig. 1 – Pure component boiling temperatures for DIPE ()
boiling component and repels the low-boiling component (or
and DEMA (), lines are calculated with the Antoine
interact ideally with the low-boiling component and attracts
equation and parameters for the pure components
the high-boiling component) (Doherty and Knapp, 2004). Next
(Gmehling et al., 1991–2014).
to improving ˛ of the binary mixture, the solvent should be
chemically and thermally stable, non-toxic and non-corrosive,
tion equaled 0.024 (GC) and 0.052 (GC–MS) at 1000 mbar and inexpensive, readily available and not reacting with one of the
0.013 (GC) and 0.060 (GC–MS) at 300 mbar. The resulting stan- mixture components (Kockmann, 2014; Doherty and Knapp,
dard deviation based on GC–MS experiments was larger since 2004). Furthermore, it should be possible to regenerate the
in GC a triple injection was performed on each sample. For the solvent so it can be recycled to the process (Jongmans et al.,
2-BuOH–2-BuO mixture, the absolute error in the equilibrium 2012b). For the regenerability of the solvent moderate inter-
distillate temperature was 0.4 ◦ C and 0.2 kPa in the pressure of actions are favored over very strong interactions (Jongmans
the system. The absolute error in ˛ was calculated to be 0.2, et al., 2012b). Also, the solvent should not form an azeotrope
which is next to errors in the sample preparation (0.001 g error with one of the components (Doherty and Knapp, 2004), which
in mass) and in the GC analysis (0.5% in wt. fraction), mainly can in general be achieved by keeping a boiling point differ-
a result of a significant error in the solvent-to-feed ratio of ence of at least 30 ◦ C and for the economic viability a relatively
±10% that was calculated based on feed samples that were low heat of evaporation is favored. For binary mixtures with
also analyzed with GC. a low ˛ it is in general easier to respect the natural difference
in vapor pressure of the components and not select a solvent
3. Results and discussion that will invert the relative volatility of low-boiling and high-
boiling component (Doherty and Knapp, 2004; Seader et al.,
3.1. The system N,N-diethylmethylamine (DEMA) – 2011).
diisoproylether (DIPE) Different types of solvents were applied, several sol-
vents were selected based on similarity and homologues
3.1.1. Verification of ebulliometry measurements (dibutylether (DBE), diglyme, dibutylamine (DBA), dipropy-
The equilibration and temperature measurement of the ebul- lamine (DPA)), others on acidity (acetic acid (HAc), HVal, oleic
liometer was verified by measuring the pure component acid), polarity and the potential of hydrogen bond formation.
boiling temperatures of DEMA and DIPE at various pressures, Hydrogen bonding is expected to play an important role as
see Fig. 1. From the figure it can be concluded that the both DEMA and DIPE contain significantly different hydro-
experimental data correspond well with the theoretical val- gen bond accepting groups which most likely show different
ues calculated with the Antoine equation (Gmehling et al., affinity with a solvent. This potential of differences in hydro-
1991–2014). gen bond formation was also studied based on qualitative
predictions of deviations from Raoult’s law for interaction of
3.1.2. Binary mixture VLE measurements different compounds classes, see Table 2 (Robbins, 1980). Ide-
Binary mixture VLE data measured for DEMA and DIPE at 300 ally, the solvent shows a different effect on the deviations of
and 1000 mbar are displayed in Fig. 2, in which it is clear that Raoult’s law to induce and increase in ˛, as is the case for acids,
˛ of this mixture is very low. At 1000 mbar the average ˛ of alcohols, secondary amines, and paraffins. Solvents in each of
DEMA over DIPE equals 1.05 ± 0.024. A decrease in pressure these categories were selected. Especially for the alcohols, pos-
increased the average ˛ to 1.09 ± 0.013 at 300 mbar. An ideal itive deviations from Raoult’s law are expected for interaction
equilibrium curve with ˛ = 1.09 is shown in Fig. 2a. Further- with the ether and ideal behavior with the amine making this
more, although no clear azeotrope can be observed at this low an interesting group of solvents, a result of the difference in
relative volatility, there is a possibility of a minimum boiling electronegativity between the nitrogen and oxygen atom. Also
azeotrope based on the Txy-diagrams (Fig. 2b and d). In this for glycols interesting solvent properties in terms of hydrogen
separation case Tb is very small, but the difference in acidity bonding possibilities are expected, and in addition they are
(basicity) and molecular structure between the components is cheap and commercially available on large scales.
large. Pseudo-binary VLE data were measured for all selected sol-
For this case solvents are selected with the aim to achieve vents at a S/F ratio of 1, the results for the effect on the relative
a high selectivity, see Eq. (1). The solvent selectivity can be volatility in the system are displayed in Fig. 3, based on only
written as the ratios of the activity coefficients of both com- the fractions of the mixture components and not the solvent.
 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134 127

Fig. 2 – Binary VLE data for DEMA and DIPE at 300 mbar (a) xy-diagram from Sprakel et al. (2018) and dotted line represents
ideal curve with ˛ = 1.09, (b) Txy-diagram from Sprakel et al. (2018), and at 1000 mbar (c) xy-diagram and (d) Txy-diagram.
In (b) and (d):  equilibrium composition and temperature of the vapor phase and  of the liquid phase. Error in xDEMA ,
yDEMA equals 0.0027 at 300 mbar and 0.0056 at 1000 mbar, see also Section 2.3.3.

Fig. 3 – Pseudo-binary VLE data for DEMA–DIPE, relative volatility ˛ of DIPE over DEMA for 300 mbar (black, data was
previously reported in Sprakel et al. (2018) and 1000 mbar (grey), with (a) S/F of 1 and (b) S/F of 3.

For the carboxylic acids the interactions with the mixture of affinity for the two components is too small to have a signifi-
DEMA and DIPE were too strong and no equilibrium could be cant effect on ˛. Secondary amines are able to form hydrogen
obtained in the experiments. When m-cresol was applied as bonds, nevertheless it can be concluded that the hydrogen
solvent, a miscibility gap was observed, so for this solvent no bond basicity of the secondary amines is too weak to affect
˛ could be measured at S/F = 1. For all other solvents except ˛. The alcohols cyclohexanol, PhOH and propylene glycol (PG)
when diglyme was applied, the measured volatility of DIPE each have a positive effect on ˛. Interaction with each of these
was larger than that of DEMA. solvents is based on hydrogen bonding and the increased ˛
The secondary amines, alkanes and ethers had no to little confirms the positive deviations from Raoult’s law that were
effect on ˛, indicating that for these solvents the difference in expected for interaction with the ether and ideal behavior with
128 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134

Table 2 – Predicted qualitative deviations (positive +,

negative − or none (0)) from Raoult’s law based on
functional groups and hydrogen bonding interactions for
the selected case of tertiary amine and ether, data are
adapted from (Robbins, 1980).
Solvent Solute

Hydrogen bonding Tertiary amine Ether

Phenol – –
Donors Acid – 0
Alcohol 0 +
Ketones + +
Tertiary amines 0 0
Secondary amines + 0
Acceptors Primary amines + +
Ether, oxides 0 0
Esters + +
Aromatics and olefins 0 0 Fig. 4 – Integrated ITC data for interaction of HVal with
None Paraffins 0 + DEMA (0.13 M) at 20 ◦ C in diluent  DIPE and 䊉 dodecane,
with the isotherms fitted by the sequential reaction model
(dotted lines, ····) (Sprakel and Schuur, 2018). Fitted
parameters for the sequential reaction model are: in
dodecane K1,1 = 21, H1,1 = −19.5 kJ/mol, Kn + 1,1 = 112,
H1,1 = −12.5 kJ/mol and n = 1.58; and in DIPE K1,1 = 9.5,
the amine (see Table 2). These differences are a result of the H1,1 = −21.1 kJ/mol, Kn + 1,1 = 11, H1,1 = −16.4 kJ/mol
difference in electronegativity between the nitrogen and oxy- and n = 1.59.
gen atom. The effect of PhOH is significantly larger than the
other two solvents, which is a result of the more acidic proton 3.1.3. Interaction between DEMA and DIPE
of PhOH. With isothermal titration calorimetry (ITC) two different
The increased relative volatility of DIPE over DEMA indi- experiments were performed with the aim to obtain more
cates that there is a larger hydrogen bonding basicity of insight in the interactions between DEMA and DIPE, as well
the tertiary amine as compared to the ether, which can be as with the solvent HVal, because no equilibrium could be
expected based on the nitrogen atom in DEMA. Increasing the reached in experiments with HVal. In the first ITC experiment
solvent to feed ratio (Fig. 3b) also increases the relative volatil- HVal was titrated to a mixture of DEMA in the apolar and inac-
ity, although based on that also for PhOH at 1000 mbar a higher tive diluent dodecane, whereas in the second experiment HVal
˛ would be expected. The ˛ at increased S/F-ratio for PhOH was titrated to a mixture of DEMA in DIPE as a diluent. The
may be influenced by more pronounced self-aggregation of the integrated ITC results are shown in Fig. 4.
PhOH at higher S/F-ratios instead of hydrogen bonding with Based on the integrated data and the sequential reaction
the amine. Moreover, at 1000 mbar the equilibrium tempera- model described by Sprakel and Schuur (2018) the param-
ture is increased, resulting in less strong hydrogen bonding eters that describe the interactions and complex formation
in general. Because of the possibility of a minimum boiling between the mixture components were fitted. For interac-
azeotrope in the binary VLE data (Fig. 2), for the three solvents tion of HVal with DEMA diluted in dodecane the calculated
that successfully increased the relative volatility (cyclohex- fit parameters are K1,1 = 21, H1,1 = −19.5 kJ/mol, Kn + 1,1 =
anol, PhOH and PG), VLE data were also measured at xDEMA of 112, H1,1 = −12.5 kJ/mol and n = 1.58; and in DIPE K1,1 =
0.05 and 0.95 and all showed a relative volatility above 1 (see 9.5, H1,1 = −21.1 kJ/mol, Kn + 1,1 = 11, H1,1 = −16.4 kJ/mol
Table 3, data was previously reported (Sprakel et al., 2018)). and n = 1.59. These parameters indicate that the stoichiom-
Based on this the presence of an azeotrope in the pseudo- etry of the complexes is not affected by the choice of diluent.
binary system could be excluded. Also in these results, the Furthermore, the moderate reaction enthalpies (< 25 kJ/mol)
solvent effect of PhOH is more pronounced. At higher S/F-ratio indicate hydrogen bonding, there is no indication of the
miscibility gaps appeared, because of which no equilibrium stronger proton-transfer in the carboxylic acid-amine com-
data could be measured. The most straightforward method plex (Tamada and King, 1990a, b). The enthalpy of reaction is
to test the regeneration is by measuring the binary VLE data higher in case the complexes are diluted in DIPE, indicating
of the high-boiling component (DEMA) and the solvent. For a more favorable environment for complexation in the polar
all three solvents a very high or infinite relative volatility was and hydrogen bond accepting DIPE and therefore suggesting
obtained as either no or a very small signal was obtained for interactions between DEMA and DIPE. The experimental sta-
DEMA in the GC chromatogram. Based on this the recovery of bility problems and not obtaining a vapor-liquid equilibrium
the solvents is not expected to be problematic. can therefore not be explained by too strong interactions and
From the results on DIPE–DEMA, it can be concluded that may be a result of the formation of a second liquid phase.
even for mixtures of components with (almost) equal volatil-
ity, a large difference in hydrogen bonding affinity offers 3.2. The system valeric acid (HVal) and
opportunities for hydrogen bond donating solvents to increase 2-methylbutyric acid (2MBA)
˛. An advantage of a mixture of components with (almost)
equal volatility is that the natural difference in volatility is Fig. 5 shows the binary VLE data for the system of HVal and
not important and solvents may be selected to target either 2MBA at both 300 and 900 mbar. The average ˛ is low and
of the mixture components, based on the most appropriate equals 1.35 ± 0.02 at 300 mbar and 1.30 ± 0.03 at 900 mbar. At
attractive or repulsive interactions. all pressures the xy-diagrams (Fig. 5a and c) show the typi-
 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134 129

Table 3 – VLE data (relative volatility ˛) for the three most promising solvents for separation of DEMA and DIPE. (n.e. = no
equilibrium achievable). Data previously reported in Sprakel et al. (2018).
˛

Pressure (mbar) 1000 300

S/F 1 3 1 3

Solvent xDEMA 0.05 0.95 0.05 0.95 0.05 0.95 0.05 0.95
– 1.3 1.0 1.1 1.1
Propylene glycol 3.7 n.e. n.e. n.e. 4.0 n.e. n.e. n.e.
Cyclohexanol 1.5 1.6 n.e. 1.1 1.8 1.9 n.e. 1.8
Phenol 7.6 3.6 8.0 n.e. 15 5.0 58 n.e.

Fig. 5 – Binary VLE data of mixture of HVal and 2MBA at 300 mbar (standard deviation in x2MBA = 0.0015 and
y2MBA = 0.0018) (a) xy diagram from Sprakel et al. (2018) and (b) Txy diagram, and 900 mbar (standard deviation in x2MBA =
0.0021 and y2MBA = 0.0031, see Section 2.3.3. for more information) (c) xy diagram and (d) Txy diagram. In (b) and (d): 
equilibrium composition and temperature of the vapor phase and  of the liquid phase.

cal shape of ideal binary data. The Txy-diagrams (Fig. 5b and tive deviations from Raoult’s law by double hydrogen bonding
d) show a narrow plateau in the region around an equimolar or acid-base interactions, respectively. On the other hand,
composition, this effect is stronger at the lowest pressure of for the apolar solvents and the solvents based on similar-
300 mbar. ity that may induce steric hindrance, the aim is to induce
The narrow plateau in (Fig. 5b) may be caused by the dimer- positive deviations from Raoult’s law. Because of the struc-
ization of acids in the vapor phase (Miyamoto et al., 1999). In tural difference between HVal and 2MBA also other solvents
literature, similar xy and Txy-diagrams are reported for binary were selected based on steric hindrance, e.g. for tribenzy-
mixtures of HBu and isobutyric acid, and as well for HBu and lamine (TBzA) that contains three aromatic rings a difference
isoHVal (Sewnarain et al., 2002). Also in those cases, the rela- in affinity is expected with the linear HVal and the branched
tive volatility is low and the slopes in the temperature profile structure of 2MBA. A similar effect may be expected for
®
are flattening close to equimolar compositions. the Proton Sponge where the cavity for interaction with
Also for this case, potential solvents for improving ˛ were the solvent is also relatively narrow based on the molecular
selected based on different aspects, i.e. a group of basic sol- structure. Pseudo-binary VLE experiments to determine the
vents, acidic solvents, solvents based on similarity with one solvent effect on the relative volatility of the system were per-
of the components of the mixture and apolar solvents. With formed with all selected solvents at 300 and 900 mbar, see
both the acidic and basic solvents the aim is to induce nega- Fig. 6.
130 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134

as the equilibrium temperature is approximately 30 ◦ C higher

at 900 mbar.
Applying the solvent MeMeBu increased ˛ to 1.84 ± 0.18 at
300 mbar. Especially in the range of low relative volatility this is
an increase that can strongly reduce the heat duty in extractive
distillation processes (Blahušiak et al., 2018). The solvent is an
ester with a lower boiling point than the mixture components
and will therefore be leaving an extractive distillation process
with the distillate stream. Because 2-methylbutyrate is also
a product of the hydrolysis of MeMeBu additional validation
was performed to exclude that the increased ˛ measured is
not a result of that. The maximum water content in the ini-
tial sample was 0.5 wt% and therefore the increased relative
volatility cannot be a result of the hydrolysis, since even if
based on this all the formed 2MBA would hypothetically end
up in the vapor phase without being part of the equilibrium,
the relative volatility would only increase to 1.6. However, at
Fig. 6 – Pseudo-binary VLE data for HVal and 2MBA at 900 mbar at which the temperature is higher, which results in
300 mbar (black, data from Sprakel et al. (2018)) and more ideal behavior based on which no significant decrease in
900 mbar (grey), S/F = 1/1 for all cases except 18-crown-6 ˛ is expected, the measured value for ˛ is lower for MeMeBu.
®
ether and Proton Sponge (S/F = 0.2) and tribenzylamine The decreased ˛ is most likely a result of transesterification
(S/F = 0.1). of the solvent with HVal, which means that MeMeBu is not
a good solvent because of reactivity. Nevertheless, there is
still an effect of MeMeBu on ˛, as the additional 2MBA that
is formed is simply part of the equilibrium and only slightly
With both the acidic and basic solvents the aim is to induce decreases the S/F-ratio.
negative deviations from Raoult’s law by double hydrogen From the observations with the various solvents inves-
bonding or acid-base interactions, respectively. On the other tigated for HVal and 2MBA, it can be concluded that the
hand, for the apolar solvents and the solvents based on sim- difference of 0.1 in pKa is too small to induce a significant
ilarity that may induce steric hindrance, the aim is to induce increase in the relative volatility through attractive interac-
positive deviations from Raoult’s law. However, most solvents tions by acid-base pairing or hydrogen bonding. Although
(Fig. 6) have no or only limited effect on ˛. For the apolar sol- applying an even stronger basic solvent may possibly increase
®
vents, apparently the difference affinity with HVal that has a ˛, as observed with the Proton Sponge , the interactions with
longer carbon chain is too small to increase ˛. Moreover, at the mixtures components are so strong that in those cases
900 mbar the temperature is increased which results in more thermal and chemical stability is compromised. Jongmans
ideal behavior, as the activity coefficients with dodecane are et al. (2012b) also report that solvents with the strongest
expected to be >1, a decreased ˛ is expected at 900 mbar. The interactions with the mixture components could not be regen-
expected difference in hydrogen bonding ability between the erated, but in their case the difference in pKa (1.6) was large
mixture components is also not strong enough for increased ˛ enough to apply moderately strong solvents to increase the
with the ether-based solvents. For tribenzylamine (TBzA) that relative volatility by inducing negative deviation from Raoult’s
contains three aromatic rings a difference in affinity could be law. Further study of mixtures with differences in acidity is
expected based on differences in steric hindrance with the lin- required to determine the minimum difference in pKa that
ear HVal and the branched structure of 2MBA, however ˛ is not allows for applying moderately strong solvents that can be
increased. For tributylphosphate (TBP) no color change was regenerated.
visible, however additional peaks appeared in the GC chro-
matogram, indicating decomposition or instability. 3.2.1. Interaction between valeric acid and TOA
®
The two solvents increasing ˛ are Proton Sponge and The heat of interaction in the system of HVal, 2MBA and TOA
MeMeBu. MeMeBu was selected based on steric hindrance was measured with ITC by performing titrations of the acids
because of the structural difference between HVal and 2MBA on samples with compositions that resemble the sample com-
and also because of structural resemblance with 2MBA, which position in the extractive distillation process, see Fig. 7. Based
results in either increased interaction between the solvent and on these results the interactions between TOA and the acids
2MBA or positive deviations from Raoult’s law for HVal. An can be explored. Fig. 7a shows the interaction of HVal with a
effect based on steric hindrance may also be expected for the mixture of 2MBA and TOA in a polar environment (1-octanol),
®
Proton Sponge , a solvent that is also strongly basic, where the where the molar ratio is the ratio of total acid concentration
cavity for interaction with the solvent is also relatively nar- over the amine concentration (and since there was already
row based on the molecular structure. For the experiments acid in the sample, the curves do not start at zero). The
®
with Proton Sponge (as well as with trioctylamine (TOA)) energy released upon interaction is between 10 and 20 kJ/mol,
the color of the mixture significantly changed and after the which is a moderate interaction energy in the range of val-
experiments the sample was completely dark, indicating a low ues reported for hydrogen bonding interactions (Sprakel and
thermal stability which is most likely a result of too strong Schuur, 2018; Tamada and King, 1990a). In a second experi-
interactions with the acids, as was previously also reported ment, see Fig. 7b where the molar ratio is the ratio of total
for monochloroacetic acid and dichloroacetic acid (Jongmans acid concentration over amine concentration, the stronger
et al., 2012b). Thermal instability may also be the reason that acid lactic acid (50 vol% in 1-octanol) was titrated to a sample
®
no increased ˛ was measured at 900 mbar for Proton Sponge , consisting of HVal, TOA and 1-octanol. This type of experiment
 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134 131

Fig. 7 – Heat of interaction at 20 ◦ C, of (a) titrating HVal to a sample containing 2MBA (1.4 mmol), TOA (4.3 mmol) and
1-octanol (6.2 mmol) and (b) titrating lactic acid in 1-octanol (50 vol%) to a sample containing HVal (0.49 mmol), TOA
(0.91 mmol) and 1-octanol (0.92 mmol).

is called a displacement experiment, because the acid-base the data of Tanaka et al. (1992). The differences with the work
interaction for the stronger acid is preferred over the inter- of Miller and Huang (1972) may be a result of the presence of
action of the base with the weaker acid, and complexes with impurities in their system that affect the relative volatility.
the weaker acid are displaced by complexes with the stronger For this case there is a significant difference in functional
acid. These experiments may be used to distinguish plateaus groups as one of the components is an alcohol and one a
in the isotherm, a method that can be applied in ITC of higher ketone. Although the average ˛ is already 1.9 for the case of
affinity interactions (Krainer and Keller, 2015). For this exper- 2-BuO–2-BuOH, partly induced by the relatively large Tb of
iment however there is only one step (two plateaus) in the approximately 20 ◦ C, solvents were selected with the aim to
measured energy instead of the three expected plateaus that further increase ˛, in which the effect of different types of sol-
are expected as a result of (1) additional loading of the com- vents could be compared. In the mixture of 2-BuO and 2-BuOH,
plexes with acid titrant, (2) the replacement of acid that was the activity coefficient of 2-BuOH is around 1 and that 2-BuO
already present with acid titrant, and (3) the solvation of acid is larger than one 1 (Miller and Huang, 1972). Thus, a good
titrant in the sample with fully loaded complexes. The sin- solvent should increase the ratio of the activity coefficients of
gle step indicates that there is similar interaction energy of 2-BuO and 2-BuOH to improve ˛. This is achieved by making
the stronger lactic acid and HVal upon interaction with TOA. the interaction with the OH group stronger or more proba-
Moreover Fig. 7b shows a steep decrease in the energy around ble and/or the interaction with the >C O group weaker or less
a stoichiometry of 2, indicating the formation of complexes probable.
with overloading of acid, which also confirms the hydrogen Because of the OH group in the higher-boiling 2-BuOH,
bonding that was concluded based on the results in Fig. 7a. the solvent should have attracting interaction with this OH
The spread in the first few data points is most likely a result group (hydrogen bond acceptor), therefore oxides (DMSO),
from errors in the injection volume as a consequence of the alcohols (EG, PG, 1-octanol, glycerol), esters (methyl levulinate
high viscosity of the lactic acid–1-octanol mixture. Thus, the (MeLev)), ethers (diglyme, DBE), acids (HBu) and ketones (5-
interactions in the mixtures are (mainly) based on hydrogen methyl-2-hexanone, cyrene) were selected. Especially for the
bonding and the formed complexes are overloaded with acid components with a larger functional group density a strong
compared to amine. effect is expected, as for example in the case of the di-alcohol
EG the probability of hydrogen bonding is increased com-
pared to 2-BuOH itself, whereas for 1-octanol with a lower
3.3. The system 2-butanol (2-BuOH) and 2-butanone
functional group density the probability is decreased. DMSO
(2-BuO)
and HBu are expected to have strong effects as they are able
to form stronger hydrogen bonds with their C O group in
Fig. 8 shows the binary VLE data for the system of 2-BuO and
comparison to the oxygen atoms in 2-BuO and 2-BuOH. The
2-BuOH at 1000 mbar, with an average ˛ of 1.9 at 1000 mbar.
-bond in aromatic rings creates a quadrupole that interacts
The xy-diagram as well as the Txy-diagram have the typi-
with the dipoles of the alcohol and ketone groups. Since the
cal shape of those of an ideal mixture and the average ˛ is
dipole moment of 2-BuO larger than that of 2-BuOH, a stronger
not very low in this case. Because of this relative volatility,
interaction between the aromatic ring and 2-BuO is expected.
distillation is an intrinsically feasible process for this sepa-
Therefore p-xylene and ethylbenzene are predicted to show
ration case, however the required energy input can still be
attractive interaction with 2-BuO and to be suitable as sol-
significantly reduced when ˛ could be increased to above 3
vents a very strong difference in affinity is required in order
(Blahušiak et al., 2016). Therefore the application of extractive
to invert the relative volatility of the binary mixture and have
distillation is also potentially interesting in this case. The xy-
2-BuOH as the top stream (considering the relative volatility
diagram of the VLE data (Fig. 8a) measured in this study was
of the binary mixture) (Kossack et al., 2008).
compared with the data of Miller and Huang (1972) and Tanaka
The effect on the relative volatility of the systems was
et al. (1992) that were not consistent with each other, although
tested for all solvents in pseudo-binary VLE experiments of
measured at almost equal pressure of 1 atm and 1000 mbar,
which the results are shown in Fig. 8c, except for the solvent
respectively. The data in this study is in good agreement with
132 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134

Fig. 8 – VLE data of mixture of 2-BuO and 2-BuOH (a) binary xy-diagram 䊉 this work, 1000 mbar  data from Miller and
Huang (1972), 1 atm and  data from Tanaka et al. (1992), 1000 mbar, (b) binary Txy-diagram, data from this work (boiling
points () and dew points (), (c) (pseudo-binary) relative volatility ˛ (±0.2, see Section 2.3.3 for more information) in the
presence of solvent (S/F = 1) at 1000 mbar, for x2−BuO of 0.4 (black bars) and 0.6 (grey bars), and (d) detailed VLE data for 
binary system and pseudo-binary with  ethylene glycol and  DMSO. Dashed curves are a guide for the eye.

glycerol that appeared to be not fully miscible with the binary selected with their boiling points between 130–200 ◦ C with the
mixture. As expected based on the interaction with the aro- aim to ensure recoverability by having a boiling point differ-
matic ring, p-xylene and ethylbenzene decrease ˛. However, ence of more than 30 ◦ C with the mixture components. The
the difference in solvent affinity with the mixture components recoverability of the solvent EG was investigated in the same
is not large enough to allow for a top stream of 2-BuOH, a way as for Case B and full recovery of EG is possible, as no sol-
result of the interaction between the aromatic ring of the sol- vent could be detected in the condensed vapor phase of the
vent and the alcohol group. A decreased ˛ is also observed for VLE measurements covering the range of (initial) EG volume
MeLev, diglyme, DBE and 5-methyl-2-hexanone, which means fractions of 0.19–0.93. By optimizing the solvent-to-feed ratio
their affinity towards 2-BuOH is not sufficiently larger than an even further improvement of ˛ can be expected compared
the affinity towards 2-BuO. For 1-octanol and HBu there is no to the ˛ of approximately 3 that was obtained with EG at a vol-
or only a very small effect, indicating similar affinity with the ume based ratio of 1, although there is a limitation as a two
mixture components. For 1-octanol this may be a result of the phase system occurred in experiments with a 5:1 solvent-to-
low functional group density. The interaction of HBu may be feed ratio. For the industrial process the stream of 2-BuO and
not specific enough because of the strongly interacting acid 2-BuOH that requires purification will most likely not have an
group. EG, PG and DMSO all increase ˛, indicating their larger equimolar composition. For a more realistic feed stream with
affinity for (hydrogen bond) interaction with 2-BuOH than with larger concentration of 2-BuO the effect of the extractive dis-
2-BuO. VLE data over the whole range of compositions are tillation process on the energy input of the process is expected
required for a more detailed process design of extractive dis- to be smaller.
tillation. To further explore the solvent effect on the relative The results for this system indicate that improving ˛ is pos-
volatility of the mixture, the pseudo-binary xy-diagram was sible based on difference in hydrogen bonding affinity of the
measured over the whole range of compositions for EG and solvent and mixture components. As the differences between
DMSO, see Fig. 8d. Both solvents appear to be successful and the mixture components are larger for this system and also the
EG is suggested as the best solvent because of price, availabil- binary VLE data show a moderate relative volatility, interac-
ity and the larger effect on ˛ compared to PG. Solvents were tions with the solvent are required to be less strong. However,
 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134 133

experiments at 900 mbar and thus higher temperature showed

reactivity of the solvent in the form of transesterification.
The difference of 0.1 in pKa in combination with the small
Tb is too small to allow for separation using a moderately
basic solvent, and applying a stronger basic solvent resulted
in chemical instability. For this specific case crystallization
would be a promising alternative process. For the case with
2-BuOH–2-BuO, Tb is relatively large and a difference in func-
tional group allowed for separation through hydrogen bonding
with the solvent. All successful solvents are hydrogen bond
acceptors that show a larger affinity for 2-BuOH than for 2-
BuO.
Generalizing, for mixtures of very similar components in
terms of acidity, boiling point and structure, stronger inter-
action with the solvent is required (e.g. complex formation,
proton transfer), which may be problematic in terms of chem-
Fig. 9 – Energy of mixing interaction upon mixing of ical stability. Therefore, before applying a solvent in the actual
2-BuOH and 2-BuO over the full range of concentrations at pseudo-binary VLE measurements, the thermal and chemical
20 ◦ C, data represent  titrating of 2-BuO to 2-BuOH and  stability with the mixture should be checked. For less simi-
titrating of 2-BuOH to 2-BuO. lar components, weaker interactions with the solvent such as
hydrogen bonding are sufficient to affect ˛. Solvents should
another consequence is that the natural difference in volatil- be selected for which specific attractive or repulsive interac-
ity, i.e. 2-BuO being the most volatile component, should be tions with the mixture components are expected, taking into
taken into account in the solvent selection. Therefore attrac- account the extent of the natural difference in volatility of
tive interactions with the least volatile component and/or the mixture components. Knowing this, isothermal titration
repulsive interactions with the most volatile component are calorimetry (ITC) may be applied for initial solvent effect pre-
required. diction based on measurement of excess enthalpy of mixing.

3.3.1. Interaction between 2-BuOH and 2-BuO Acknowledgement

As the interaction energy between the solvents that are
selected for this separation case and the mixture components This is an ISPT (Institute for Sustainable Process Technology)
is relatively low compared to the acid-base interactions of the project.
other cases, the mixing interaction of the two mixture compo-
nents was measured with ITC instead of their interaction with References
a solvent. The heat of mixing of 2-BuOH and 2-BuO (Fig. 9),
is relatively low and positive in sign, i.e. endothermic, which Bahrmann, H., Frohning, C.D., Heymanns, P., Kalbfell, H., Lappe,
means that the energy of intermolecular interactions is similar P., Peters, D., Wiebus, E., 1997. n-Valeric acid: expansion of the
for both compounds and the combination of compounds. This two phase hydroformylation to butenes. J. Mol. Catal. A:
also implies that the components do not (strongly) attract each Chem. 116, 35–37.
other and a successful solvent may be found based on moder- Bastos, J.C., Soares, M.E., Medina, A.G., 1985. Selection of solvents
for extractive distillation. A data bank for activity coefficients
ate or even weak repulsion of 2-BuO or attraction of 2-BuOH.
at infinite dilution. Ind. Eng. Chem. Process. Des. Dev. 24,
For very large mole fractions of either of the two components,
420–426.
the heat effects are more strongly endothermic. Berg, L., Ratanapupech, P., 1986. US Patent no. US4569726(A),
Process for the separation of ethyl acetate from ethanol and
4. Conclusion water by extractive distillation. Google Patents.
Bízek, V., Horáček, J., Koušová, M., 1993. Amine extraction of citric
acid: effect of diluent. Chem. Eng. Sci. 48, 1447–1457.
Solvent effects on the relative volatility were measured for
Blahušiak, M., Kiss, A.A., Kersten, S.R.A., Schuur, B., 2016. Quick
three different industrially relevant cases, that are separation assessment of binary distillation efficiency using a heat
of (a) diethylmethylamine and diisopropylether (Tboil = 3–5, engine perspective. Energy 116, 20–31.
˛ DEMA over DIPE of 1.05 ± 0.028 at 1000 mbar, ˛ = 1.20 ± 0.036 Blahušiak, M., Kiss, A.A., Babic, K., Kersten, S.R.A., Bargeman, G.,
at 300 mbar), (b) valeric acid and 2-methylbutyric acid (Tboil = Schuur, B., 2018. Insights into the selection and design of fluid
10, pKa = 0.1, ˛ = 1.35 ± 0.02 at 300 mbar and 1.30 ± 0.03 at separation processes. Sep. Purif. Technol. 194, 301–318.
900 mbar) and (c) 2-butanol and 2-butanone (Tboil = 20, Brignole, E.A., Bottini, S., Gani, R., 1986. A strategy for the design
and selection of solvents for separation processes. Fluid Phase
˛ = 1.9 at 1000 mbar). For each of the cases potential solvents
Equilib. 29, 125–132.
were selected based on different aspects, including acid- Canales, R.I., Brennecke, J.F., 2016. Comparison of ionic liquids to
ity/basicity, structural similarity, steric hindrance, polarity and conventional organic solvents for extraction of aromatics
predicted differences in hydrogen bonding affinity. from aliphatics. J. Chem. Eng. Data 61, 1685–1699.
For DEMA-DIPE, although Tb is very small, the difference Doherty, M.F., Knapp, J.P., 2004. Distillation, Azeotropic, and
in acidity and molecular structure of the components could Extractive. Kirk-Othmer Encyclopedia of Chemical Technology.
DOW, 2014. Product safety assessment: valeric acid. In: Company,
be used to find suitable solvents. PhOH (> 5 at S/F = 3 and
T.D.C.
300 mbar), cyclohexanol (˛ = 2) and PG (˛ = 3.1) each increase
Eckert, F., Klamt, A., 2002. Fast solvent screening via quantum
˛ of the binary mixture. For HVal–2MBA the solvent MeMeBu chemistry: COSMO-RS approach. AlChE J. 48, 369–385.
increased ˛ (˛ = 1.84 ± 0.18 at 300 mbar), a solvent based on European-Council, 2014. 2030 Climate and Energy Policy
structural similarity with the mixture components. However, Framework.
134 Chemical Engineering Research and Design 1 4 4 ( 2 0 1 9 ) 123–134

Eurostat, 2018. Energy Statistics – An Overview. European Council. Miyamoto, S., Nakamura, S., Iwai, Y., Arai, Y., 1999. Measurement
Fredenslund, A., Jones, R.L., Prausnitz, J.M., 1975. of vapor-phase compressibility factors of monocarboxylic
Group-contribution estimation of activity coefficients in acids using a flow-type apparatus and their association
nonideal liquid mixtures. AlChE J. 21, 1086–1099. constants. J. Chem. Eng. Data 44, 48–51.
Gmehling, J., Onken, U., Arlt, W., Grenzheuser, P., Weidlich, U., Pretel, E.J., López, P.A., Bottini, S.B., Brignole, E.A., 1994.
Kolbe, B., Rarey, J., 1991–2014. Chemistry Data Series, Volume Computer-aided molecular design of solvents for separation
1: Vapor-Liquid Equilibrium Data Collection, in: Behrens, D. processes. AlChE J. 40, 1349–1360.
(Ed.), Vapor-Liquid Equilibrium Data Collection. Quijada-Maldonado, E., Aelmans, T.A.M., Meindersma, G.W., de
Graczova, E., Vavrusova, M., 2018. Extractive distillation of Haan, A.B., 2013. Pilot plant validation of a rate-based
acetone–methanol mixture using extractive distillation model for water–ethanol separation
1-ethyl-3-methylimidazolium trifluoromethanesulfonate. with the ionic liquid [emim][DCA] as solvent. Chem. Eng. J.
Chem. Eng. Trans. 70, 1189–1194. 223, 287–297.
Harper, P.M., Gani, R., 2000. A multi-step and multi-level Robbins, L.A., 1980. Liquid–liquid extraction: a pretreatment
approach for computer aided molecular design. Comput. process for wastewater. Chem. Eng. Prog. 76, 58–61.
Chem. Eng. 24, 677–683. Roose, Peter, Eller, Karsten, Henkes, Erhard, Rossbacher, Roland,
Jongmans, M.T.G., Hermens, E., Raijmakers, M., Maassen, J.I.W., Höke, H., 2015. Amines, Aliphatic, Ullmann’s Encyclopedia of
Schuur, B., de Haan, A.B., 2012a. Conceptual process design of Industrial Chemistry.
extractive distillation processes for ethylbenzene/styrene Schult, C.J., Neely, B.J., Robinson, R.L., Gasem, K.A.M., Todd, B.A.,
separation. Chem. Eng. Res. Des. 90, 2086–2100. 2001. Infinite-dilution activity coefficients for several solutes
Jongmans, M.T.G., Londoño, A., Mamilla, S.B., Pragt, H.J., in hexadecane and in n-methyl-2-pyrrolidone (NMP):
Aaldering, K.T.J., Bargeman, G., Nieuwhof, M.R., Kate, A.t., experimental measurements and UNIFAC predictions. Fluid
Verwer, P., Kiss, A.A., van Strien, C.J.G., Schuur, B., de Haan, Phase Equilib. 179, 117–129.
A.B., 2012b. Extractant screening for the separation of Seader, J.D., Henley, E.J., Roper, D.K., 2011. Separation Process
dichloroacetic acid from monochloroacetic acid by extractive Principles: Chemical and Biochemical Operations, 3rd ed.
distillation. Sep. Purif. Technol. 98, 206–215. Wiley, Hoboken, NJ.
Kamlet, M.J., Abboud, J.L.M., Abraham, M.H., Taft, R.W., 1983. Sewnarain, R., Ramjugernath, D., Raal, J.D., 2002. Isobaric
Linear solvation energy relationships. 23. A comprehensive vapor−liquid equilibria for the systems propionic
collection of the solvatochromic parameters,. pi.*,. alpha., acid + butyric acid, isobutyric acid + butyric acid, butyric
and. beta., and some methods for simplifying the generalized acid + isovaleric acid, and butyric acid + hexanoic acid at
solvatochromic equation. J. Org. Chem. 48, 2877–2887. 14 kPa. J. Chem. Eng. Data 47, 603–607.
Kiss, A.A., Ignat, R.M., 2012. Innovative single step bioethanol Smith, R., Jobson, M., 2000. Distillation. In: Wilson, I.D. (Ed.),
dehydration in an extractive dividing-wall column. Sep. Purif. Encyclopedia of Separation Science. Academic Press, Oxford,
Technol. 98, 290–297. pp. 84–103.
Kockmann, N., 2014. 200 years in innovation of continuous Sakuth, Michael, Mensing, Thomas, Schuler, Joachim, Heitmann,
distillation. Chembioeng Rev. 1, 40–49. Wilhelm, Strehlke, Günther, Mayer, D., 2010. Ethers, Aliphatic,
Kossack, S., Kraemer, K., Gani, R., Marquardt, W., 2008. A Ullmann’s Encyclopedia of Industrial Chemistry.
systematic synthesis framework for extractive distillation Sprakel, L.M.J., Schuur, B., 2018. Thermal activity in affinity
processes. Chem. Eng. Res. Des. 86, 781–792. separation techniques such as liquid–liquid extraction
Krainer, G., Keller, S., 2015. Single-experiment displacement analyzed by isothermal titration calorimetry and accuracy
assay for quantifying high-affinity binding by isothermal analysis of the technique in the molar concentration domain.
titration calorimetry. Methods 76, 116–123. Ind. Eng. Chem. Res., 12574–12582.
Kubitschke, J., Lange, H., Strutz, H., 2000. Carboxylic Acids, Sprakel, L.M.J., Kamphuis, P., Nikolova, A.L., Schuur, B., 2018.
Aliphatic, Ullmann’s Encyclopedia of Industrial Chemistry. Development of extractive distillation processes for
Wiley-VCH Verlag GmbH & Co. KGaA. close-boiling polar systems. Chem. Eng. Trans. 69, 529–534.
Kulajanpeng, K., Suriyapraphadilok, U., Gani, R., 2014. Taft, R.W., Abboud, J.-L.M., Kamlet, M.J., Abraham, M.H., 1985.
Ionic-Liquid Based Separation of Azeotropic Mixtures. Linear solvation energy relations. J. Solution Chem. 14,
Laurence, C., Gal, J.-F., 2009. Lewis Basicity and Affinity Scales. 153–186.
John Wiley & Sons, Ltd. Tamada, J., King, C., 1990a. Extraction of carboxylic acids with
Laurence, C., Graton, J., Gal, J.-F., 2011. An overview of Lewis amine extractants. III: effect of temperature, water
basicity and affinity scales. J. Chem. Educ. 88, 1651–1657. coextraction, and process considerations. Ind. Eng. Chem.
Lei, Z., Li, C., Chen, B., 2003. Extractive distillation: a review. Sep. Res. 29, 1333–1338.
Purif. Rev. 32, 121–213. Tamada, J.A., King, C.J., 1990b. Extraction of carboxylic acids with
Lek-utaiwan, P., Suphanit, B., Douglas, P.L., Mongkolsiri, N., 2011. amine extractants. 2. Chemical interactions and
Design of extractive distillation for the separation of interpretation of data. Ind. Eng. Chem. Res. 29, 1327–1333.
close-boiling mixtures: solvent selection and column Tanaka, H., Muramatsu, T., Kato, M., 1992. Isobaric vapor–liquid
optimization. Comput. Chem. Eng. 35, 1088–1100. equilibria for three binary systems of 2-butanone with
Li, Q., Zhang, J., Lei, Z., Zhu, J., Zhu, J., Huang, X., 2009. Selection 3-methyl-1-butanol, 1-butanol, or 2-butanol. J. Chem. Eng.
of ionic liquids as entrainers for the separation of ethyl Data 37, 164–166.
acetate and ethanol. Ind. Eng. Chem. Res. 48, 9006–9012. Zhu, J., Yu, Y., Chen, J., Fei, W., 2007. Measurement of activity
Meyer, P., Maurer, G., 1995. Correlation and prediction of partition coefficients at infinite dilution for hydrocarbons in
coefficients of organic solutes between water and an organic imidazolium-based ionic liquids and QSPR model. Front.
solvent with a generalized form of the linear solvation energy Chem. Eng. China 1, 190–194.
relationship. Ind. Eng. Chem. Res. 34, 373–381.
Miller, K.J., Huang, H.-S., 1972. Vapor–liquid equilibrium for
binary systems 2-butanone with 2-butanol, 1-pentanol, and
isoamyl alcohol. J. Chem. Eng. Data 17, 77–78.