Diagrama de Flujo

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Journal of Thermal Analysis and Calorimetry

https://doi.org/10.1007/s10973-020-09562-6

Phase equilibria of binary and ternary polymer solutions using


modified UNIQUAC‑based local composition model
Amirhossein Amirsoleymani1 · Hamid Bakhshi1   · Seyed Reza Shabanian1 · Kamyar Movagharnejad1

Received: 26 August 2019 / Accepted: 13 March 2020


© Akadémiai Kiadó, Budapest, Hungary 2020

Abstract
In the present research, a modification on UNIQUAC activity coefficient model was done based on the local composition
concept. The model was applied for binary and ternary systems containing different polymers and protic solvents. The
newly modified model was compared with several local composition models such as Wilson, NRTL, UNIQUAC, polymer-
NRTL, UNIQUAC-LBY and UNIQUAC-DMD in vapor–liquid. Also, comparison of the model with NRTL, UNIQUAC and
UNIQUAC-LBY was done for liquid–liquid equilibrium data. In the presented model, a Freed-FV term was added to activity
coefficient expression to consider the free volume effects in the mixture. Also, new surface and volumetric structural param-
eters were applied for studied alcoholic and aqueous solutions. AARD % of the new model (known as M-UNIQUAC-LBY
model) was 1.98% for 28 vapor–liquid equilibrium data set (314 data points) of binary polymer–solvent systems. Moreover,
RMSD % of 14 studied liquid–liquid systems in equilibria (155 tie-lines) was equal to 10.6431. The results of presented
model were superior to previously presented ones in both VLE and LLE. Therefore, it can be successfully utilized to predict
phase equilibria in mixtures, which contain pertinent components for industrial applications.

Keywords  Local composition · Activity coefficient · Free volume · VLE · LLE · Polymer

List of symbols h Hard core


cal Calculated
Subscript
exp Experimental
i Component
j Component, phase Parameters
k Component, tie-line y Vapor phase mole fraction of component i
I Component x Liquid phase mole fraction of component i
J Component w Mass fraction of component i
f Fugacity of component i
Superscript
X The segment-based mole fraction of specie I for
sat Saturate
the poly-NRTL model
I Phase one
𝜇i Chemical potential of component i
II Phase two
mI The total number of segments in component I
Comb Combinatorial
𝜏ij Binary interaction parameters of model
Res Residual
aij Binary model parameter
FV Free volume
Δgij Interaction energy between molecules i and j
𝜙I Segment mole fraction polymer-NRTL model
𝜙i Different types of volume fraction in component

Electronic supplementary material  The online version of this


article (https​://doi.org/10.1007/s1097​3-020-09562​-6) contains
i for the UNIQUAC-DMD model
supplementary material, which is available to authorized users. 𝜙fv
i
 Free volume fraction of component i
𝜙hi Hard-core volume fraction of component i
* Hamid Bakhshi 𝜉i Volume fraction of UNIQUAC-LBY model
h.bakhshi@nit.ac.ir
𝜃i Surface area fraction
1
Faculty of Chemical Engineering, Babol Noshirvani R Molecular volume parameter of subgroups
University of Technology, Babol, Iran Q Molecular surface area parameter of subgroups

13
Vol.:(0123456789)
A. Amirsoleymani et al.

rj,J The number of j segment in component J in



Chen [8] developed the local composition concept based
polymer-NRTL model on the segment structure for polymer–solvent solutions.
ri Molecular volume parameter Furthermore, in other local composition models, group
qi Molecular surface parameter contribution procedure was applied for various solutions
vi Molar volume of component i and has become progressively valuable. The most common
vhi Hard-core volume of component i group contribution models for predicting phase equilibria
vfv Free volume of component i are ASOG [9], UNIFAC [10], modified UNIFAC (Lyn-
i
vvdw Van der Waals volume of component i gby), [11] and modified UNIFAC (Dortmund) [12]. The
i
mentioned group contribution methods are considered as
Abbreviations predictive models. However, calculated values of equilib-
VLE Vapor–liquid equilibrium rium compositions using such models deviate significantly
LLE Liquid–liquid equilibrium from experimental data. As a result of the deficiencies of
LBY Lyngby these models especially in predicting the phase behavior of
DMD Dortmund polymer solutions, several modifications on activity coef-
AARD Absolute average relative deviation ficient term have been performed. Mentioned models can
RMSD Root mean square deviation be considered as UNIQUAC-based local composition mod-
OF Objective function els which contain two terms (a combinatorial entropic and
VdW Van der Waals a residual term) for activity coefficient. Sayegh and Vera
PEG Polyethylene glycol [13] recommended the Guggenheim–Staverman expres-
DEX Dextran sion for considering of combinatorial entropic term. How-
ever, they suggest that in cases which the unrealistic results
would be obtained the Guggenheim–Staverman expression
Introduction could be replaced by Flory–Huggins model. Based on this
suggestion, Larsen et al. applied Flory–Huggins equation
An understanding of the phase equilibria of polymer solu- with applying a modified volume fraction term, instead of
tions is important in industrial applications such as removal Staverman–Guggenheim combinatorial term. They called
of residual solvents from monomers, oligomers and pro- the model as modified UNIQUAC model (UNIQUAC-LBY);
duced polymers [1]. In order to study phase behavior of they used this model successfully for calculation of binary
polymer solutions, enough experimental data should be and ternary VLE systems containing hexane, cyclohexane
generated. Although, the experimental processes for obtain- and methanol [11]. They have studied the UNIQUAC-LBY
ing such data are time-consuming and expensive. Therefore, model to show improvement of the results compared with
suitable thermodynamic models, such as activity coefficient original UNIQUAC model. Also, Larsen et al. improved the
expressions, could be applied in order to reduce available predictive results of original UNIFAC model and named the
data to proper correlations. In thermodynamic modeling modified model as UNIFAC-LBY. Several studies utilized
of polymer systems, activity of the associated solvent with UNIFAC-LBY as a predictive model for phase equilibria
polymer is a key parameter in vapor–liquid equilibria, which calculations. However, the studied systems did not contain
can be used as a kind of modified composition in the system polymers [14, 15]. Some researchers have tried to modify
[2]. Thermodynamic models for predicting phase behavior the UNIFAC-LBY model. Ye et al. modified UNIFAC-LBY
in polymer solutions have been obtained first from Van der by the use of Peng–Robinson EoS and Wong–Sandler mix-
Waals (VdW) and lattice theories [3, 4]. Later, various local ing rule. Their new model was called PRWS-UNIFAC-LBY
composition models were introduced. The most common and applied for phase equilibria of systems containing water,
local composition models for correlating phase equilibria are methanol, dimethyl ether and carbon dioxide [16].
Wilson [5], NRTL [6] and UNIQUAC [7]. A major short- One effective modification involved considering free
coming of the Wilson model is its limitation in predicting volume in the lattice of solutions. According to Patterson
the phase equilibria of partially miscible or non-miscible expression [17], phase behavior of polymer–solvent solu-
systems. Renon and Prausnitz [6] solved the weakness of tions is different from ordinary organic solvent mixtures,
Wilson’s model by considering local composition concept because the sizes of both components in polymer mixture
based on enthalpy of the mixture. In 1975, Abrams and are very different. Polymer has lower free volume space in
Prausnitz [7] presented UNIQUAC model, which contains the cell of postulated lattice than the small molecule of the
two terms known as combinatorial and residual contribu- solvent. Since this effect arises from the difference in struc-
tions. UNIQUAC model was satisfactory in a wide range ture and size of pertinent polymer and solvent, it is associ-
of multi-component systems. Many researches have been ated with a combinatorial part of the activity coefficient.
conducted based on the mentioned models. For instance, Several studies have utilized this concept to develop activity

13
Phase equilibria of binary and ternary polymer solutions using modified UNIQUAC‑based local…

coefficient models. Elbro et al. [18] proposed an entropic vapor–liquid data points. Also, liquid–liquid equilibria of
model in which the volume fraction for combinatorial term 14 different PEG–DEX–water systems have been calculated
of Flory–Huggins was replaced by free volume fraction. using the presented model. Necessary binary interaction
Later, Kannan et al. [19] obtained an expression for free parameters of investigated models were obtained through
volume activity coefficient of components by subtracting correlating available data in vapor–liquid equilibria. When-
the combinatorial term of Flory–Huggins from Elbro model. ever such data were unavailable in the literature, binary
They applied their free volume model on the 43 non-aqueous parameters were calculated directly from LLE data points.
and 11 aqueous polymer solvent mixtures to show the per-
formance of the model in the wide range of data points.
Radfarnia et al. [20, 21] proposed free volume combinatorial
term based on modification of Freed-Flory–Huggins model Thermodynamic modeling
to calculate vapor–liquid equilibrium of solvent–polymer
mixtures (Freed-FV model). Another local composition Vapor–liquid equilibria
model utilized UNIQUAC-NRF equation by considering
segments of molecules to calculate VLE of polymer solu- As a result of considering different experimental data, there
tions [22]. This model suggested a new segment-based ther- are different approaches for calculating VLE. The ( 𝛾 − 𝜑 )
modynamic model including the combinatorial and ener- approach was chosen in this study for VLE calculations.
getic contributions of excess Gibbs energy. In 2005, Sadeghi Due to the equality of fugacity of components into vapor
developed Wilson’s equation for non-electrolyte solutions to and liquid phases, it follows that:
calculate excess Gibbs energy of multi-component polymer yi 𝜑i p = xi 𝛾 i psat (1)
solutions [23]. In this model, the weak local physical inter-
i

actions between solvent and segment of polymer chain have where psati
, 𝛾i , xi , yi , 𝜑i stand for saturated vapor pressure,
been considered. Other modifications of Wilson and NRTL activity coefficient, liquid phase mole fraction, vapor phase
models have been performed by Sadeghi for calculation mole fraction and fugacity coefficient of component i,
of solvent activity in polymer solutions [24]. It was found respectively. Since no polymer enters vapor phase, the mole
that both models are appropriate in representing the excess fraction of solvent is set to unity in this phase. Thus, the
enthalpy for polymer solution. Madeira et al. presented a pressure or activity of solvent could be considered as a target
modified Wilson’s model for multi-component aqueous two- for VLE calculation. In addition, at low pressure, the vapor
phase solutions containing polymer and salt and successfully phase has been assumed to be ideal; hence, the fugacity coef-
correlated the phase behavior of the mentioned systems [25]. ficient is approximately one. Therefore, the total pressure
Also, a local composition model was presented by Pazuki and activity of solvent can be obtained directly as follows:
et al. [26] to correlate vapor–liquid and liquid–liquid equi-
libria in polymer–polymer aqueous two-phase systems. p = xi 𝛾 i psat
i (2)
Moreover, Sadeghi et al. [27] proposed another combina-
torial model to predict vapor–liquid equilibrium of some ai = xi 𝛾i . (3)
semi-dilute binary polymer solutions at different tempera-
tures. Thermodynamic modeling of binary phase equilib- Since the activity coefficient models contain binary inter-
ria led to fitting of proper binary interaction parameters. action parameters, the absolute average deviation percent
Polymers are inherent components in aqueous two-phase (AARD %) defined as follows is used as an objective func-
system, as an important tool for extraction of biomolecules tion to minimize the deviation between experimental and
which could be considered a liquid–liquid equilibria sub- calculated activities of solvent i:
ject [28, 29]. Therefore, it is essential to obtain best set of
100 ∑ || ai − ai ||
cal exp
such parameters to extend their application in more complex AARD% = | |. (4)
n | aexp |
solutions. Attempts have been done in this case with some | i |
new extension of different models, which are applicable for
The saturated vapor pressure of solvent i can be calcu-
both polymer–solvent binaries and polymer–salt–solvent or
lated using the following equation:
polymer–polymer–solvent ternaries [30–35].
In present study, a new thermodynamic activity coeffi- ( )
C2
(5)
sat C5
cient model has been proposed for polymer solutions con- pi (Pa) = exp C1 + + C3 ln T + C4 T .
T
taining protic solvents. Optimized adjustable parameters
of the model have been reported. The new model has been C1 to C5 are extended Antoine coefficient constants,
used for correlation of 28 different binary systems contain- which are shown in Table 1 for each component in the cur-
ing different polymers and protic solvents which involve 314 rent study.

13
A. Amirsoleymani et al.

Table 1  Constants of extended Solvent C1 C2 C3 C4 C5 Tmin/K Tmax/K


Antoine equation for studied
solvents [36] Methyl alcohol 39.205 −1324.4 −3.4366 3.10E − 05 2 90.69 190.56
Ethyl alcohol 73.304 −7122.3 −7.1424 2.89E − 06 2 159.05 514
Propyl alcohol 84.6642 −8307.2 −8.5767 7.51E − 18 6 146.95 536.8
Isopropyl alcohol 96.094 −8575.4 −10.292 1.67E − 17 6 185.26 508.3
Water 73.649 −7258.2 −7.3037 4.1653E − 06 2 273.16 647.096

Liquid–liquid equilibria Δ𝜇 Ii = Δ𝜇 IIi . (8)

In this study, we applied the suggested methods by Kang and The presented flowchart algorithm by Haghtalab and
Sandler to calculate the liquid phase compositions in equilib- Asadollahi [38] for calculating the liquid–liquid equilib-
rium [37]. The following criteria were considered for each rium for polymer solution was used in the current study, as
component: shown in Fig. 1. The algorithm utilizes two objective func-
tions in order to limit the number of trials for optimiza-
fiI = fiII (6) tion of adjustable parameters. The first objective function
­(OF1), which narrows down the initial guesses of needed
aIi = aIIi (7) parameters, is written as follows:

Fig. 1  Flowchart of LLE calcu-


lation [38]

13
Phase equilibria of binary and ternary polymer solutions using modified UNIQUAC‑based local…

∑ ∑(
m 3
)2 Moreover, the polymer-NRTL model as a modification of
OF1 = Δ𝜇 Iik − Δ𝜇 IIik . (9) original NRTL has been applied for previously mentioned sys-
k i tems in order to consider the combinatorial effect of big poly-
Afterward, minimization of second objective function mers in the solution. The equations of polymer-NRTL model
­(OF2) leads to the best set of calculated compositions using [8] are presented in “Appendix.”
the following equation: Another investigated local composition model in the cur-
rent study is the UNIQUAC model, which could be written as:
2 ( )2

m

3
∑ ( ) ( ( ))
OF2 = Min wExp − w Cal
. (10) 𝜙 𝜙i z 𝜙 𝜙i
ijk ijk ln 𝛾icomb = 1 − i + ln − 1 − i + ln
k i j xi xi 2 𝜃i 𝜃i
(14)
Since the molecular mass of polymers is usually high,
it is better to use mass fractions instead of mole fractions � �m � �m 𝜃j 𝜏 ij
ln 𝛾ires = − qi ln 𝜃 𝜏ji + qi ∑ (15)
in calculating the compositions, as suggested by Kang and j=1 j m
𝜏
j=1 k=1 𝜃k kj
Sandler [37]. To assess the results, RMSD % defined as fol-
lows is used to compare different models as used by previous
ri xi
researchers [33, 37, 38]: 𝜙i =

m
(16)
� rj xj

� N 2 3 ⎛ Cal ⎞
2 j=1
�� � � wijk − wExp
100 � ⎜ ijk
⎟ (11)
RMSD% = �
N ⎜ w Cal
k=1 j=1 i−=1 ⎝ ijk + w Exp ⎟
qi x i
ijk ⎠
𝜃i =
∑m
(17)
qj xj
where N is the number of tie-lines and the superscripts of j=1
i, j and k are defined as number of components, phases and
tie-lines, respectively. ∑
ri = v(k)
i Rk (18)
Activity coefficient models k


The activity coefficient of component i could be divided into qi = v(k)
i
Qk (19)
combinatorial and residual terms as follows: k

( )
ln 𝛾i = ln 𝛾icomb + ln 𝛾ires . (12) Δuij aij
ln 𝜏ij = − = − . (20)
RT T
NRTL model does not include a combinatorial part which
can be assumed as limitation of it in comparison with other xi is mole fraction of component i, while θi and 𝜙i are the
LC models [2], because the difference in the sizes of com- surface and volume fractions of compound i, respectively. qi
ponents should be considered which could be defined by and ri as structural parameters of pure component i are cal-
combinatorial part of LC models. Additionally, in Wilson culated from physical functional group parameters Q and R,
equation, the combinatorial term of activity coefficient is respectively. v(k) is the number of k functional group in mol-
not determined clearly, but this term is specified obviously
i
ecule i. R and Q are obtained from the literature for constituent
in UNIQUAC and poly-NRTL equations [2]. In this paper, functional groups of each component [10]. Obtained values
different binary polymer systems of PPO, PEG, PEO, PVAC of structural volumes (r) and surface parameters (q) used for
and PVAL mixed with various solvents have been investi- different solvents in the present study were calculated from
gated to evaluate the local composition activity coefficient Eqs. (18) and (19), respectively, and are shown in Table 2.
of components. Wilson and NRTL models were applied for Also, aij is considered an adjustable parameter for UNIQUAC
modeling of mentioned systems. The equations of these model. z parameter is the coordination number and is set to 10.
models as well-known local composition models are not pre- The combinatorial term of UNIQUAC-LBY as the next
sented here. It is important to note that the binary interaction studied model can be written as [11]:
parameter of NRTL model ( 𝜏 ij) is considered as follows: ( )
( ) 𝜉i 𝜉
Δgij 𝛾
ln i comb
= 1 − + ln i (21)
𝜏ij = − = aij . (13) xi xi
RT

13
A. Amirsoleymani et al.

Table 2  The structural Component UNIQUAC [10] UNIQUAC-LBY UNIQUAC-DMD M-UNIQUAC-


parameters of solvents used in [43] [12] LBY
different models (this work)
r q r (z/2)q r q r (z/2)q

Methyl alcohol 1.4311 1.432 1 1 0.8585 0.9938 1 1


Ethyl alcohol 2.575 1.972 2.575 1.972 2.4952 2.6616 2.105 1.356
Propyl alcohol 3.249 3.128 3.249 3.128 3.1277 3.7224 2.779 2.512
Isopropyl alcohol 3.022 3.124 3.022 3.124 2.9605 3.3433 2.77 2.508
Water 0.92 1.4 0.92 1.4 1.7334 2.4561 0.8154 0.904

xi r2∕3 Eq. (15). According to Eq. (25), new correction to Staver-


𝜉i = ∑m i 2∕3 (22) man–Guggenheim model was used for the combinatorial
x r
j=1 j j part. The structural volume and surface parameters for this
model have been regressed using functional group param-
(z∕2)qi xi eters (R and Q) defined and tabulated in UNIFAC-Dortmund
𝜃i = ∑ m (23) model [12]. The calculated values of mentioned parameters
x
j=1 (z∕2)qj j for studied solvents in the current study are presented in
Table 2. Volume fraction (𝜙i ) and surface fraction (𝜃i ) of
∑ ( )
qi = v(k)
z each component are defined in UNIQUAC equation.
i 2
Qk . (24)
k
Current work
In UNIQUAC-LBY model, different definitions for vol-
ume fraction ( 𝜉i ) have been applied as suggested by Kikic We have presented a new version of modified UNIQUAC
et al. [39], as shown in Eq. (22). In this model, the combina- equation for aqueous and alcoholic polymer mixtures. Three
torial term of Staverman–Guggenheim in UNIQUAC model improvements have been applied with respect to previous
has been replaced by a term from Flory–Huggins model. The models. First of all, the exponent of volumetric structural
residual term of UNIQUAC-LBY is the same as residual parameters in the definition of volume fraction has been
contribution of UNIQUAC model. The structural param- optimized by fitting available experimental data. Secondly,
eters of each solvent (r and q) have been calculated using the Freed-FV term has been added to activity coefficient
Eqs. (18) and (24) and are presented in Table 2. Functional expression to consider free volume effect. Finally, new sur-
group structural parameters R and (z/2)Q were taken from face and volumetric structural parameters have been applied
the original presentation of UNIQUAC-LBY model [11]. for alcoholic and aqueous solutions (protic solvents) of poly-
Moreover, another modification of the UNIQUAC model mers. Also, another activity coefficient model has been pre-
was suggested here. In presented activity coefficient model, sented by modifications of combinatorial and residual terms
the combinatorial part of UNIFAC-Dortmund model [12] of modified UNIFAC model.
has been added to residual term of original UNIQUAC According to Larsen et al. [11], since the Flory–Huggins
model [7]. We named this model UNIQUAC-DMD. The model did not consider free volume effect of postulated lat-
combinatorial contribution of presented UNIQUAC-DMD tice, a Freed-FV free volume expression has been added to
is as follows: UNIQUAC-LBY in order to cover this deficiency. Therefore,
� ( �) ( ( )) the presented new model is:
𝜙i 𝜙i z 𝜙 𝜙i
comb
ln 𝛾i =1− + ln − 1 − i + ln
xi xi 2 𝜃i 𝜃i ln 𝛾i = ln 𝛾icomb + ln 𝛾ires + ln 𝛾ifv . (27)
(25)
The combinatorial part of this model is similar to UNI-
� xi ri3∕4 QUAC-LBY as shown in Eq.  (21). The residual part of
𝜙i = ∑m . (26) this model is the same as that of UNIQUAC. The modified
x r3∕4
j=1 j j Freed-FV equation proposed by Radfarnia et al. [21] can be
In UNIFAC-Dortmund model, an empirical definition written as:
of volume fraction ( 𝜙i ) has been added to combinato-

rial term of activity coefficient as shown in Eq. (26). The


residual term of UNIQUAC-DMD model was given by

13
Phase equilibria of binary and ternary polymer solutions using modified UNIQUAC‑based local…

( ) ( ) ( )
𝜙fv
i
𝜙hi − 𝜙fv
i
vvdw
i
= 15.17 ri (34)
ln 𝛾ifv = ln +
𝜙hi xi
where ri and vi are the volumetric structural parameter and
( )2 ( )2
𝜙fv 𝜙hi molar volume of molecule i, respectively. The volumetric
i
+ 0.2 1 − −0.2 1 − and surface structural parameters of each component have
xi xi
(28) been calculated by Eqs. (18) and (34), respectively, simi-
lar to the UNIQUAC-LBY model. However, to improve the
xi vfv
i
results, new physical values of functional group structural
𝜙fv
i = ∑m (29) parameters (R and Q) for water and OH groups, suggested by
x vfv
j=1 j j Radfarnia et al. [21], have been applied. Table 2 shows the
calculated values of mentioned parameters for solvents by
xi vh using the basic structural parameters presented in Table 3.
𝜙hi = ∑m i h . (30) In addition, another improvement has been considered for
x v
j=1 j j
the model. In fact, the exponent of volumetric parameter of
The free volume vfv is the space that is not occupied by each component in the definition of volume fraction ( 𝜉i ) has
been set to an adjustable parameter “n”:
i
the molecules and is responsible for the mobility of long
chain of polymer in a solution [40]. The vfv
i
can be written as: xi rn
𝜉i = ∑m i n . (35)
vfv
i
= vi − vhi (31) x r
j=1 j j

where vhi is the hard-core volume of component i. Some Thus, the new model known as M-UNIQUAC-LBY was
researchers equate it to the Van der Waals volume. How- formed by the summation of Eqs. (15), (21) and (28) as
ever, in the present study, according to comments by Kon- residual, combinatorial and free volume terms of activity
togeorgis et al. [41] and Van Krevelen and Te Nijenhuis [42] coefficient, respectively. For achieving a general formula, it
for liquid and rubbery polymer, the hard-core volume was is necessary to calculate a unique quantity for the adjustable
considered larger than Van der Waals volume; therefore, vhi parameter n. The value of n, in presented model, has been
was determined using the following equation: regressed by minimizing the AARD % of solvent activities
( ) presented in Eq. 4. The optimized value of n was obtained
vhi = 1.3 vvdw
i . (32) as 0.7.
For LLE calculation, NRTL, UNIQUAC, UNIQUAC-
In the present paper, the molar volume of polymer has
LBY and M-UNIQUAC-LBY models have been applied on
been regressed by suggested temperature dependent equation
the basis of mass fraction instead of mole fraction, which
by Van Krevelen [42]:
was perfectly explained by Kang and Sandler [37].
( )
vi = vvdw
i
1.3 + 10−3 T . (33)

T is temperature in kelvin. Van der Waals volume of spe-


cies i, vvdw
i
 , has been determined as:

Table 3  Basic structural Group name UNIQUAC [10] UNIQUAC-LBY [43] M-UNIQUAC-LBY


parameter values in each model (this work)
and the monomer structure of
used polymers R Q R (z/2)Q R (z/2)Q

CH 0.4469 0.4469 0.4469 0.228 0.4469 0.228


CH2 0.6744 0.6744 0.6744 0.540 0.6744 0.540
CH3 0.9011 0.9011 0.9011 0.848 0.9011 0.848
ACH 0.5313 0.5313 0.5313 0.400 0.5313 0.400
CHO(ether) 0.6908 0.6908 0.6908 0.650 0.6908 0.650
CH2 O(ether) 0.9183 0.9183 0.9183 0.780 0.9183 0.780
CH3 COO(ester) 1.9031 1.9031 1.9031 1.728 1.9031 1.728
OH 1.0000 1.2000 1.0000 1.2000 0.52999 0.5840

13
A. Amirsoleymani et al.

Table 4  Binary polymer– No. System Mn.10−3 Temp range/K Np Reference


solvent systems studied in the
present work 1 PEG/methyl alcohol 0.6 303.15 6 [44]
2 PEG/ethyl alcohol 0.6 303.15 6 [44]
3 PPO/ethyl alcohol 2 303.15 7 [44]
4 PPO/methyl alcohol 2 303.15 5 [44]
5 PPO/propyl alcohol 2 303.15 6 [44]
6 PVAL/methyl alcohol 116 363–383 6 [45]
7 PVAc/methyl alcohol 113 338–373 10 [45]
8 PEG/isopropyl alcohol 1.96 298.15 18 [46]
9 PEG/isopropyl alcohol 2.96 298.15 20 [46]
10 PEG/isopropyl alcohol 9.95 298.15 13 [46]
11 PEG/water 0.2 298 11 [47]
12 PEG/water 0.3 303.15–338.15 24 [48]
13 PEG/water 0.4 298 11 [47]
14 PEG/water 0.6 298 11 [47]
15 PEG/water 1 298 7 [47]
16 PEG/water 1.45 298 9 [47]
17 PEG/water 3 338.15 9 [48]
18 PEG/water 3.35 298 7 [47]
19 PEG/water 4 308.15–338.15 16 [49]
20 PEG/water 5 333.15–338.15 16 [48]
21 PEG/water 6 298 7 [47]
22 PEG/water 8 298 7 [47]
23 PEG/water 1 298 7 [47]
24 PEG/water 20 298 7 [47]
25 PVAc/water 113 313.15 7 [45]
26 PVAL/water 116 338.15–373.15 16 [45]
27 PEO/water 6 298.15–308.15 39 [50]
28 PPO/water 2 303.15 6 [44]
Overall 314

Mn is number-average relative molar mass of the polymer

Results and discussion AARD % between experimental and calculated activi-


ties of solvents for 28 studied binaries which include 314
VLE systems equilibrium data points has been reported for seven thermo-
dynamic models and is presented in Table 5. In Table 5, the
To assess the seven studied models, experimental data of deviation of models for each system and average deviation
polymer solution have been gathered from the literature of each model for all 28 studied binaries were also observed.
[44–50]. Twenty-eight investigated vapor–liquid equilibrium The order of arrangement for systems presented in this table
data of polymer–solvent binary mixtures investigated in pre- is the same as Table 4.
sent study are listed in Table 4 along with their references It can be seen that the two sets of solvents (alcohols and
and number of data points (Np). Vapor–liquid equilibrium water) have been studied with various polymers. In general
data sets were applied to obtain the calculated activity of point of view, it is clear that considering the free volume
solvent for comparison with reported available experimental term in M-UNIQUAC-LBY as our new model decreased the
data. The same set of data for each system were utilized to average AARD %. Among the studied models, Wilson and
fit necessary binary interaction parameters of investigated NRTL had worst results. It is obvious that these two models
models. contain no term for combinatorial effect of activity coeffi-
cient, which seem to be important in modeling big polymer

13
Phase equilibria of binary and ternary polymer solutions using modified UNIQUAC‑based local…

Table 5  Absolute average deviation percent of seven investigated models for binary polymer–solvent systems
No. System AARD % Ref.
Wilson NRTL Polymer -NRTL UNIQUAC​ UNIQUAC-LBY UNI- M-UNI-
QUAC- QUAC-LBY
DMD (this work)

1 PEG/methyl alcohol 4.45 17.92 4.06 2.19 5.32 4.06 3.32 [44]
2 PEG/ethyl alcohol 3.75 4.64 3.65 3.35 3.26 3.3 3.24 [44]
3 PPO/ethyl alcohol 13.22 23.19 11.36 7.33 5.55 13.03 4.28 [44]
4 PPO/methyl alcohol 6.06 13.07 7.72 9.06 9.29 9.68 8.41 [44]
5 PPO/propyl alcohol 4.53 7.03 4.73 4.56 4.64 4.49 4.53 [44]
6 PVAL/methyl alcohol 16.88 8.56 7.67 7.78 9.78 7.68 8.66 [45]
7 PVAc/methyl alcohol 19.67 34.93 6.07 6.86 9.87 6.32 7.04 [45]
8 PEG/isopropyl alcohol 0.08 0.15 0.08 0.08 0.08 0.08 0.09 [46]
9 PEG/isopropyl alcohol 0.1 0.1 0.1 0.1 0.1 0.1 0.1 [46]
10 PEG/isopropyl alcohol 0.06 0.12 0.07 1 0.06 0.06 0.08 [46]
11 PEG/water 0.95 3.05 0.61 2.23 0.19 0.96 0.58 [47]
12 PEG/water 10.49 27.2 19.54 3.43 2.75 13.42 1.8 [48]
13 PEG/water 1.81 4.2 1.03 3.14 0.84 1.79 0.27 [47]
14 PEG/water 2.13 3.79 2.09 3.23 1.42 2.14 0.54 [47]
15 PEG/water 0.18 0.11 0.12 0.08 0.13 0.11 0.24 [47]
16 PEG/water 0.23 0.62 0.23 0.22 0.36 0.53 0.22 [47]
17 PEG/water 6.6 16.39 4.35 3.45 1.44 2.36 1 [48]
18 PEG/water 0.13 0.05 0.06 0.06 0.06 0.05 0.19 [47]
19 PEG/water 0.05 0.19 0.19 0.19 0.19 0.19 0.1 [49]
20 PEG/water 5.41 10.67 1.43 2.05 0.73 1.3 0.19 [26]
21 PEG/water 0.2 0.08 0.1 0.02 0.11 0.08 0.09 [47]
22 PEG/water 0.18 0.06 0.09 0.05 0.09 0.07 0.24 [47]
23 PEG/water 0.19 0.11 0.12 0.11 0.12 0.11 0.25 [47]
24 PEG/water 0.21 0.08 0.11 0.05 0.12 0.08 0.28 [47]
25 PVAc/water 4.28 12.84 3.43 7.75 3.54 3.7 3.92 [45]
26 PVAL/water 8.16 27.5 2.37 2.7 29.39 7.32 3.37 [45]
27 PEO/water 0.13 0.13 0.17 0.11 0.11 0.11 0.1 [50]
28 PPO/water 2.91 5.95 2.49 2.32 2.39 6.91 2.35 [44]
Average of AARD % 4.04 7.95 3 2.63 3.28 3.21 1.98

molecules. Moreover, it can be seen that UNIQUAC model (last term of Eq. 14), while this term has been omitted from
presented better results compared to UNIQUAC-LBY and the UNIQUAC-LBY model. Surface area plays an impor-
UNIQUAC-DMD. It seems that without considering the free tant role in contact of polymer segments and solvents.
volume effect in activity coefficient calculations, they do When molecules are large, their contact area effect is more
not make significant improvement to the UNIQUAC model. dominant. This could be the reason why the UNIQUAC-
The original UNIQUAC model contains a surface area DMD shows better presentation of data in comparison
effect term in the combinatorial part of activity coefficient with UNIQUAC-LBY. Additionally, as earlier discussed,

Table 6  Comparison of System Wilson NRTL Poly-NRTL UNIQUAC​ UNI- UNI- M-UNI-


absolute average relative QUAC- QUAC- QUAC-
deviation percent (AARD %) LBY DMD LBY
of models for alcoholic and (this
aqueous solvents work)

Alcohol–polymer 6.88 10.97 4.55 4.23 4.79 4.88 3.98


Water–polymer 2.46 6.28 2.14 1.73 2.44 2.29 0.87

13
A. Amirsoleymani et al.

the UNIQUAC-DMD is a modification of UNIFAC-DMD, of different activity coefficient models more precisely. It
which is based on functional groups. However, availability could be observed that an improvement in the results by
of adjustable parameters in this model leads to better results M-UNIQUAC-LBY model in aqueous polymer solution was
as can be seen in Table 5. Moreover, since the polymer- significant by an AARD % equal to 0.87. According to the
NRTL model does not involve the surface area effect, its investigated temperatures of solutions, the assumption of
results were not better than UNIQUAC and its modifications. an ideal behavior of gas phase for water–polymer mixture is
Comparison of deviations between various models could be more realistic and gives better understanding of the differ-
done by dividing solvents into alcoholic and aqueous groups ences between various models.
as tabulated in Table 6. As seen in Table 6, all studied mod- The calculated activities of some solvents by studied
els represented lower AARD % for polymer–water binaries models in the present work along with experimental values
in comparison with alcoholic mixtures. It is known that alco- are depicted in Fig. 2. Comparison of results shows that
hols are more volatile solvents compared to water. Therefore, considering free volume and new structural parameters leads
non-ideality of vapor phase rises when an alcohol is present to superiority of presented model in this study.
in the solution. Since we assumed that the fugacity coef- Figure  3 shows the experimental and calculated
ficient of the vapor phase is equal to one, thermodynamic vapor–liquid equilibrium data of PEG (300)—water binary
modeling of water–polymer systems reflects the capabilities system at three temperatures. M-UNIQUAC-LBY and

(a) 1.0 (b) 1.0 (c)


1.0
0.9
0.8 0.9
0.8
Activity

0.6 0.8
Activity

NRTL NRTL
0.7 NRTL

Activity
Wilson Wilson
Wilson
0.4 UNIQUAC
0.6
UNIQUAC 0.7 UNIQUAC
Poly-NRTL Poly-NRTL Poly-NRTL
UNIQUAC-LBY UNIQUAC-LBY
0.5 0.6 UNIQUAC-LBY
0.2 UNIQUAC-DMD UNIQUAC-DMD UNIQUAC-DMD
This work This work This work
Experimental 0.4 Experimental 0.5 Experimental
0.0
0.0 0.2 0.4 0.6 0.8 0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3
Wt fraction of solvent Wt fraction of solvent Wt fraction of solvent

(d) (e) (f)


1.000 1.0

0.995 0.9 0.6


Activity

Activity
Activity

NRTL 0.8 NRTL NRTL


0.990 Wilson Wilson 0.4 Wilson
UNIQUAC 0.7 UNIQUAC UNIQUAC
0.985 Poly-NRTL Poly-NRTL Poly-NRTL
UNIQUAC-LBY
UNIQUAC-LBY
UNIQUAC-DMD
0.6 UNIQUAC-DMD
0.2 UNIQUAC-LBY
UNIQUAC-DMD
0.980 This work This work This work
Experimental 0.5 Experimental Experimental
0.0
0.6 0.7 0.8 0.9 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.00 0.02 0.04 0.06 0.08 0.10 0.12
Wt fraction of solvent Wt fraction of solvent Wt fraction of solvent

(g) (h)
0.8 0.8

0.6
NRTL 0.6 NRTL
Activity
Activity

Wilson Wilson
0.4 UNIQUAC UNIQUAC
Poly-NRTL Poly-NRTL
0.4 UNIQUAC-LBY
UNIQUAC-LBY
0.2 UNIQUAC-DMD UNIQUAC-DMD
This work This work
Experimental 0.2 Experimental
0.0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25
Wt fraction of solvent Wt fraction of solvent

Fig. 2  Comparison between experimental and calculated activities of [47], f PPO 2000  + methyl alcohol T = 303.15  K [44], g PPO
solvents. a PEG 300 + water T = 303.15 K [48], b PEG 3000 + water 2000 + ethyl alcohol T = 303.15  K [44] and h PPO 2000 + propyl
T = 328.15  K [48], c PEG 5000 + water T = 318.15  K [48], d PEG alcohol T = 303.15 K [44]
4000 + water T = 338.15  K [47], e PEG 600 + water T = 298.15  K

13
Phase equilibria of binary and ternary polymer solutions using modified UNIQUAC‑based local…

0.25
Experimental data at 303.15 K
M-UNIQUAC-LBY model correlates systems with less aver-
Experimental data at 325.15 K
Experimental data at 340.15 K
age molecular weight of polymer more precisely than others.
0.20
This work
UNIQUAC-LBY
As previously mentioned, all studied models involve
binary interaction parameters. These binary parameters
were fitted by minimizing AARD % of calculated solvent
0.15
activities and best set parameters of each model for 28
Activity

studied binaries, as presented in Table 7. These parameters


0.10 could be used for the mentioned models in multi-compo-
nent systems or other pertinent solutions which contain
0.05 polymers such as ATPS. Therefore, regressed parameters
from VLE data were applied in ternary systems of liq-
0.00 uid–liquid equilibria in the present work.
0.0 0.2 0.4 0.6
According to dispersion of all predicted activities from
Wt fraction of solvent
the diagonal line in Fig. 5, it can be seen that M-UNI-
QUAC-LBY model had the best results compared to other
Fig. 3  Correlating the PEG (300) + water systems at different tem- investigated models. Also, prediction of NRTL and Wilson
peratures using M-UNIQUAC-LBY (this work) and UNIQUAC-LBY
models. Experimental data were obtained from Malcolm and Rowlin- models deviated more from experimental values, due to
son [48] the fact that NRTL and Wilson models have been formed
based on considering simple small spherical molecules
in the lattice. While in the UNIQUAC and other studied
0.0320 segment-based models, differences in molecular sizes were
considered.
0.0315

0.0310 LLE systems


0.0305
Activity

For the calculation of liquid–liquid equilibrium, the pre-


Experimental PEG (1000)
0.0300 Experimental PEG (3350) cise binary interaction parameters of the molecules were
Experimental PEG (20000)
UNIQUAC-LBY (Dotted) required. The interaction parameters of PEG–water were
0.0295 This wtork (Dash)
obtained by fitting the VLE data, as presented in Table 7.
0.0290 It should be noted that in calculating the residual terms of
UNIQUAC model, the exact structural parameters of com-
0.0285
0.5 0.6 0.7 0.8 0.9 1.0 ponents play a crucial role to converge to a proper result in
Solvent wt fraction liquid–liquid equilibria assessment. Fourteen investigated
LLE systems assigned from A to N, as well as the tem-
Fig. 4  Correlating VLE of different molecular masses of PEG + water perature ranges of available experimental data and number
binaries at 298.15  K via M-UNIQUAC-LBY (this work) and UNI- of data points, are listed in Table 8. Also, The references
QUAC-LBY models. Experimental data were obtained from Ninni of experimental data points can be seen in Table 8. The
et al. [47]
RMSD % and interaction parameters of our presented model
(M-UNIQUAC-LBY) along with three other investigated
UNIQUAC-LBY equations were used for modeling. Figure 3 models in LLE (NRTL, UNIQUAC and UNIQUAC-LBY)
shows that our presented model correlates the results more are presented in Table 9. Moreover, average RMSD % values
precisely at lower temperatures. The same trend could be of four studied models for presenting all studied LLE data
seen for UNIQUAC-LBY model which is depicted for com- points are reported in Table 10.
parison. The effect of different molecular masses of PEG According to the average RMSD % values in Table 10,
can be seen in Fig. 4. Three different molecular masses of UNIQUAC and UNIQUAC-LBY models had relatively good
PEG (1000, 3350, 20,000) at 298.15 K were chosen. Fig- performance. In addition, their average RMSD % is nearly
ure 4 shows that the presented M-UNIQUAC-LBY model is the same (see systems L and E).
capable of correlating both low and high molecular masses In VLE calculations, binary interaction parameters were
of polymer better than UNIQUAC-LBY model, which is pre- regressed. Such parameters were used for calculating the
sented in Fig. 4 for comparison. High molecular masses of liquid–liquid equilibria for the studied systems at different
polymers involve more monomers. Therefore, more errors temperatures.
in the studied models are expected. Consequently, the

13

13
Table 7  Optimized interaction parameters of seven studied models for 28 polymer–solvent binary systems
System NRTL Poly-NRTL UNIQUAC​ UNIQUAC-LBY UNIQUAC-DMD M-UNIQUAC-LBY
number (this work)
a12 a21 a12 a21 a12 a21 a12 a21 a12 a21 a12 a21

1 3.504 − 4.743 0.718 − 1.143 − 271.138 460.641 3071.631 − 437.803 1313.679 − 364.071 − 294.036 43.38
2 1.192 − 1.412 − 1.55 2.02 − 80.828 228.025 − 201.818 330.946 − 153.398 246.391 − 262.312 393.17
3 2.317 − 3.139 − 1.751 1.998 − 191.114 474.923 − 273.97 457.183 47.346 − 45.974 − 291.114 512.132
4 2.304 − 3.202 − 3.772 2.396 29.248 209.241 1744.833 − 282.587 993.465 − 139.891 1004.259 − 367.38
5 2.095 − 2.692 − 1.322 2.096 − 105.366 211.03 − 211.001 325.52 − 154.382 235.091 − 87.715 142.814
6 15.87 − 17.989 − 1.662 1.854 2147.843 140.875 − 178.294 − 63.267 882.702 − 394.213 − 464.369 214.399
7 14.352 − 16.594 − 0.208 1.045 1377.483 192.634 734.224 − 491.411 1506.879 − 385.597 − 434.196 434.106
8 1.311 53.946 − 0.208 1.045 38.073 13.804 43.201 − 2.653 11.031 28.503 89.802 − 54.71
9 1.496 47.976 − 0.334 1.311 − 65.652 176.296 − 107.483 226.057 − 111.388 216.874 − 143.082 266.367
10 2.786 30.786 − 0.291 1.484 − 52.295 152.473 − 82.08 125.073 − 92.635 153.636 − 124.418 135.27
11 − 1.867 2.262 1.002 − 1.445 − 109.66 811.862 − 247.792 461.595 − 323.701 5192.162 111.001 70.005
12 − 120.209 − 7.404 − 69.607 − 3.323 − 287.059 627.098 − 246.123 539.114 606.973 − 607.003 − 372.21 783.001
13 − 2.326 2.782 1.188 − 1.823 − 96.779 695.435 − 307.929 5223.051 − 321.703 5495.202 − 283.541 832.911
14 2.84 − 3.37 − 0.447 0.476 − 317.738 5471.023 − 313.654 5475.107 − 322.598 5296.724 − 186.731 446.688
15 − 7.114 6.229 − 3.799 0.368 − 47.614 396.509 − 427.994 333.063 − 440.486 326.682 − 253.606 293.498
16 − 7.568 6.689 − 0.193 0.013 − 52.235 374.446 − 441.918 2272.589 − 432.347 4361.725 − 280.022 24.835
17 − 125.499 − 4.308 − 1.694 1.976 − 234.763 561.565 − 270.392 496.24 − 266.814 452.561 − 259.361 520.152
18 − 9.865 9.507 − 3.84 0.375 − 115.444 498.208 − 550.981 395.764 − 403.695 207.303 − 276.952 300.058
19 − 5.483 9.248 − 4.639 5.358 − 50.297 269.915 − 357.273 835.028 − 392.246 738.1 − 129.755 175.789
20 − 5.912 6.073 1.237 − 2.853 − 265.95 861.951 − 300.386 599.43 − 297.434 589.502 − 256.111 547.194
21 − 11.227 11.107 − 3.863 0.412 − 228.025 − 8.258 − 602.442 521.136 − 533.39 464.458 − 418.036 705.422
22 − 11.995 − 11.987 − 3.859 0.39 − 225.521 − 15.175 − 643.646 520.778 − 542.096 417.618 − 330.499 412.46
23 − 12.504 − 12.666 − 3.832 0.375 − 339.384 51.49 − 647.045 444.631 − 496.002 230.291 − 312.342 370.749
24 − 14.272 − 14.649 − 3.865 0.412 − 266.427 13.297 − 668.363 544.362 − 562.579 465.382 − 316.248 352.502
25 − 73.425 − 9.674 − 0.789 2.103 1268.002 31.067 1782.848 − 275.878 1781.654 − 303.546 1714.72 − 83.034
26 − 90.249 − 17.069 − 1.462 1.016 − 110.166 149.164 1047.88 − 472.926 − 424.505 770.807 − 444.184 306.945
27 − 78.134 11.082 − 54.37 87.071 156.886 − 206.022 − 104.502 − 27.37 − 20.602 − 79.934 − 184.823 184.882
28 − 1.501 2.019 − 1.136 2.219 − 121.615 525.131 − 182.378 488.995 130.38 − 26.684 − 50.953 668.213
A. Amirsoleymani et al.
Phase equilibria of binary and ternary polymer solutions using modified UNIQUAC‑based local…

1.2 UNIQUAC-LBY models is shown in Fig. 6. As observed,


the UNIQUAC-LBY model had better result compared to
1.0
NRTL model, which had the highest deviation between
models. UNIQUAC-LBY model considers the combina-
Predicted activity

0.8
torial effects in activity coefficient which is missing in
0.6 NRTL model. Since the UNIQUAC-LBY had better cor-
NRTL relation, this model was compared with UNIQUAC model,
0.4
Wilson
UNIQUAC
as shown in Fig. 6. This comparison showed better predic-
0.2 Polymer-NRTL tion of UNIQUAC toward UNIQUAC-LBY. Further com-
UNIQUAC-LBY
UNIQUAC-DMD parison between UNIQUAC and M-UNIQUAC-LBY (our
0.0 M-UNIQUAC-LBY
Diagonal line study) revealed that the proposed model (M-UNIQUAC-
LBY) had the best performance among all models. At the
0.0 0.2 0.4 0.6 0.8 1.0 1.2
center of Fig. 6, the comparison of four mentioned models
Experimental activity
was shown in one single plot. These results indicate that
the accuracy of calculated activity coefficient could be
Fig. 5  Comparison between experimental and calculated activities for improved by considering the free volume effect as well
various models
as decent structural parameters in order to predict compo-
nents mass fraction in each phase.
As observed in Table 9, studied systems showed better Moreover, the assessment of binodal curves is illus-
performance for M-UNIQUAC-LBY compared to other trated in Fig. 7 for different molecular weights of PEG
models in most cases. In some cases, there was little dif- at 293.15 K (systems C and D) via M-UNIQUAC-LBY
ference between deviations of models from experimental and UNIQUAC-LBY models. Finally, the compari-
data (system L). In contrast to other models, NRTL had son of binodal curves for ATPS of DEX and PEG with
less accuracy for RMSD %. The average RMSD % of all different molecular weights of DEX at 298.15 K using
studied data points for each model is shown in Table 10. M-UNIQUAC-LBY and UNIQUAC-LBY models is shown
These values indicate that M-UNIQUAC-LBY model pre- in Fig. 8 (systems I, J, K and L in Table 10). According
sents the data better than other investigated ones. to isothermal plots in Figs. 7 and 8, the deviation from
As an example, to illustrate the capability of the model experimental data points had a direct relationship with
in phase behavior of liquid–liquid systems, Fig. 6 com- the molecular weight of polymers, which means that in the
pares the calculated values of system A to previously specific system, when one polymer like PEG or DEX has
reported data. The comparison between NRTL and higher molecular weight, the deviation from experimental

Table 8  Properties of the System T (K) Number PEG DEX Mn× 10−3 Mn× 10−3 Reference
ternary systems applied in this of tie-line PEG DEX
work
A 298.15 8 PEG 8000 DEX T-500 8.92 167 [51]
B 298.15 6 PEG 3350 DEX T-70 3.35 37 [51]
C 277.15–313.15 19 PEG 6000 DEX-500 6.23 179.347 [52]
D 277.15–313.15 16 PEG 35000 DEX-500 39.005 179.347 [52]
E 277.15 4 PEG 4000 DEX T-40 3.4 24.4 [53]
F 313.15 4 PEG 4000 DEX T-40 3.4 24.2 [53]
G 277.15–295.15 8 PEG 4000 DEX T-70 3.4 38.4 [53]
H 277.15–295.15 8 PEG 4000 DEX T-500 3.4 234.2 [53]
I 277.15–313.15 12 PEG 4000 DEX-10 3.8 13.2 [54]
J 277.15–313.15 12 PEG 4000 DEX-40 3.8 27.7 [54]
K 277.15–313.15 12 PEG 4000 DEX-110 3.8 52.1 [54]
L 277.15–313.15 12 PEG 4000 DEX-500 3.8 88.2 [54]
M 298.15–313.15 12 PEG 6000 DEX-40 5.3 27.7 [54]
N 281.15–323.15 22 PEG 6000 DEX-70 6 28.7 [55]

13
A. Amirsoleymani et al.

Table 9  The RMSD % and System Model aPEG–DEX aDEX–PEG aDEX–water awater–DEX RMSD %
binary interaction parameters
for fourteen studied LLE A 1 5.890 − 1.628 8.939 615.426 30.26
systems
2 − 7066.682 460.764 − 1197.795 − 959.98 20.4
3 − 9241.307 429.814 − 16,781.17 − 873.22 24.78
4 − 871.91 451.666 − 1390.359 − 1017.121 15.16
B 1 − 4.272 3.514 4 30.570 16.53
2 − 9420.79 − 325.737 − 1516.671 654.194 8.24
3 − 9237.551 398.732 − 9457.408 − 756.538 12.82
4 − 52.997 377.626 − 8698.904 − 891.149 6.50
C 1 − 5.698 3.569 5.005 29.988 20.25
2 − 3183.742 352.531 − 9585.716 − 715.224 14.17
3 − 9818.936 449.098 − 9701.098 − 677.461 15.51
4 − 89.49 434.678 − 5238.028 − 955.606 9.38
D 1 4.88 − 2.498 25.813 6.825 16.13
2 − 778.48 369.324 − 8243.914 − 821.523 10.81
3 − 234.306 476.772 − 7559.354 − 1033.578 23.37
4 − 234.306 476.771 − 1555.227 − 1033.578 14.95
E 1 4.208 1.613 2.745 28.585 14.24
2 − 9821.347 − 1280.38 391.58 − 805.44 6.89
3 − 6621.977 − 60.66 − 9450.416 490.525 6.34
4 − 3061.955 − 877.039 497.271 − 5992.532 4.46
F 1 3.388 − 4.335 3.967 27.309 13.73
2 − 1200.777 − 8575.998 369.281 − 763.183 4.54
3 − 49.81 − 9726.979 465.299 − 9112.236 5.94
4 − 590.247 − 6640.161 468.778 − 1193.497 3.79
G 1 3.424 − 0.069 3.065 28.543 9.81
2 − 797.414 − 8749.171 377.486 − 770.15 19.65
3 − 9571.649 − 9433.985 369.538 − 628.692 3.33
4 − 8928.787 − 14,394.72 468.246 − 1047.769 5.77
H 1 10.272 − 4.607 3.277 28.593 31.01
2 − 2641.481 − 8828.168 365.524 − 749.938 28.34
3 − 7096.74 − 9771.713 378.633 − 678.543 37.00
4 − 8955.73 − 1545.669 435.901 − 879.923 24.27
I 1 4.15 0.214 179.21 24.677 6.46
2 − 1339.084 − 9820.143 339.975 − 649.587 22.71
3 − 8549.52 − 9234.384 556.878 − 347.186 17.36
4 − 1422.227 − 1028.969 440.784 − 1010.996 5.05
J 1 6.548 0.124 175.592 24.905 21.81
2 − 2884.935 − 1010.352 336.862 − 651.914 26.13
3 − 9827.142 − 9983.706 564.297 − 445.814 9.69
4 − 7458.069 − 1411.062 439.872 − 969.584 8.10
K 1 6.706 0.01 890.93 25.299 28.59
2 − 8057.456 − 1007.464 341.799 − 674.644 6.84
3 − 9420.155 − 1019.646 544.028 − 409.613 21.73
4 − 9845.034 − 8102.091 430.045 − 906.276 14.21
L 1 6.181 0.034 313.781 25.196 22.62
2 − 9660.21 − 8117.172 340.061 − 669.204 30.31
3 − 9043.4 − 7933.153 535.642 − 382.669 30.79
4 − 8004.987 − 7487.17 429.804 − 902.547 21.28

13
Phase equilibria of binary and ternary polymer solutions using modified UNIQUAC‑based local…

Table 9  (continued) System Model aPEG–DEX aDEX–PEG aDEX–water awater–DEX RMSD %

M 1 6.053 − 0.365 917.174 24.737 24.37


2 − 8069.668 − 385.382 83.79 − 607.49 26.54
3 − 9432.855 9824.5 − 352.153 − 566.491 19.56
4 − 8871.054 − 8642.553 399.973 − 774.092 9.97
N 1 5.724 − 5.669 399.577 25.174 7.83
2 − 2100.867 − 9337.741 366.307 − 659.623 17.66
3 − 9608.612 − 9122.928 347.821 − 556.597 8.96
4 − 2066.037 − 1045.49 393.293 − 728.01 6.10

1, NRTL; 2, UNIQUAC; 3, UNIQUAC-LBY; 4, M-UNIQUAC-LBY

Table 10  The average RMSD  % in each model for fourteen studied points for this model will be higher than that with a lower
LLE systems molecular weight. It could be seen that the new model can
Model NRTL UNIQUAC​ UNIQUAC- M-UNIQUAC- represent the data in different ranges of molecular weights
LBY LBY of polymers in ATPS very well and the results are compa-
rable with UNIQUAC-LBY model.
Average of 18.83 17.37 16.94 10.64
RMSD %

DEX
0.0 1.0
Experimental Data
M-UNIQUAC-LBY
UNIQUAC

0.1 0.9

0.2 0.8

PEG 0.3 0.7 Water


0.0 0.1 0.2 0.3

0.00
1.00
Experimental
UNIQUAC
UNIQUAC-LBY
This Work
NRTL

DEX
DEX 0.0 1.0
0.0 1.0
Experimental Data
Experimental Data UNIQUAC-LBY
UNIQUAC NRTL
UNIQUAC-LBY
0.1 0.9
0.1 0.9
0.25
0.75
0.00 0.25
0.2 0.8
0.2 0.8

PEG 0.3 PEG 0.3 0.7 Water


0.7 Water 0.0 0.1 0.2 0.3
0.0 0.1 0.2 0.3

Fig. 6  Comparison of calculated mass fraction of M-UNIQUAC-LBY model with three other models (UNIQUAC, NRTL and UNIQUAC-
LBY) for system A (introduced in Table 8) at 298.15 K. Experimental data from King et al. [51]

13
A. Amirsoleymani et al.

0.18 PEG–DEX–water systems in liquid–liquid equilibria. After-


0.16 Exp PEG 6000 ward, the proposed model’s capability for correlating the phase
Exp PEG 35000
This work (Dash) behavior of liquid–liquid systems was determined by com-
0.14 UNIQUAC-LBY
paring it with three local composition models (NRTL, UNI-
0.12
QUAC and UNIQUAC-LBY). The total average RMSD based
Wt PEG

0.10 on LLE calculation for NRTL, UNIQUAC, UNIQUAC-LBY


0.08 and M-UNIQUAC-LBY (this work) was 18.83, 17.37, 16.94
0.06 and 10.64, respectively. Finally, results showed the superior-
0.04
ity of this new model compared to six other presented local
composition models. Hence, it can be successfully utilized for
0.02
correlating the systems that contain polymers and water or
0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
alcohols.
Wt DEX
Acknowledgements This work was financially supported by
Babol Noshirvani University of Technology through Grant No.
Fig. 7  Comparison of experimental and calculated mass fractions by BNUT/390058/1396. Author appreciate Dr. Jefferson Stephen for his
M-UNIQUAC-LBY model for ATPS of DEX-40–water–PEG (at dif- helps for editing the manuscript.
ferent molecular weights) at 293.15 K (systems C and D in Table 10)
Appendix

0.16 The polymer-NRTL equation can be written as follows [8]:


( ) ( )
0.14 DEX-10
𝜙I ∑ 𝜙J
DEX-40
DEX-110 ln 𝛾Icomb = ln + 1 − mI (36)
0.12 DEX-500 xI J
mJ
This work
UNIQUAC-LBY
0.10
Wt PEG

∑ ∑
0.08 ⎛ 𝜏ji Gji Xj ⎛ Xk 𝜏kj Gkj ⎞⎞
� �⎜ j � Xj Gji ⎜ k ⎟⎟
0.06
ln 𝛾Ires = ri,I ⎜ ∑ + ∑ ⎜𝜏ij − ∑ ⎟⎟
i ⎜ Gki Xk j Gkj Xk ⎜ Gkj Xk ⎟⎟
0.04
⎝ k k ⎝ k ⎠⎠
(37)
0.02 ( )
Gji = exp − 𝛼ji 𝜏ji (38)
0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Wt DEX x m
𝜙I = ∑ I I
xJ mJ (39)
Fig. 8  Comparison of experimental and calculated mass fractions by J
M-UNIQUAC-LBY model for ATPS of PEG 4000–water–DEX (at
different molecular weights) at 298.15°K (systems I, J, K and L in �
Table 10) xI ri,I
Xi = ∑ ∑ � (40)
xJ rj,J
J j
Conclusions
∑ �
mI = ri,I (41)
Based on obtained results of seven studied activity coefficient
i
models for vapor–liquid equilibria of 28 various binary poly-
mer solvent solutions, it could be concluded that considering where I and J represent any of the species including solvent
the free volume effect increases the capability of the model to or polymer. Species i and j could be solvent molecules or
predict data points. The newly modified M-UNIQUAC-LBY segments. The Xi parameter is the segment-based mole frac-
model was superior to six other investigated local composi- tion of specie i. The XJ parameter is the mole fraction of
tion models. Applying different modifications to this model component J, while 𝜙I is the segment mole fraction of com-
ponent I. Also, rj,J parameter shows the number of j segment

like considering different exponents of volume fraction along
with free volume term of activity coefficient reduced the aver- in component J, while mI is the total number of segments.
age AARD % of 28 binary systems to 1.9810. This model The energy and interaction parameters of this model are the
was also applied for calculating the phase behavior of various same as NRTL model.

13
Phase equilibria of binary and ternary polymer solutions using modified UNIQUAC‑based local…

References 22. Radfarnia HR, Bogdanic G, Taghikhnai V, Ghotbi C. The


UNIQUAC-NRF segmental interaction model for vapor-liquid
equilibrium calculations for polymer solutions. POLIMERI-
1. Prausnitz JM, Lichtenthaler RN, de Azevedo EG. Molecular ther-
ZAGREB. 2005;26(3):115.
modynamics of fluid-phase equilibria. London: Pearson Educa-
23. Sadeghi R. Extension of the Wilson model to multicomponent
tion; 1998.
polymer solutions: applications to polymer–polymer aqueous
2. Kontogeorgis GM, Folas GK. Thermodynamic models for indus-
two-phase systems. J Chem Thermodyn. 2005;37(1):55–60.
trial applications: from classical and advanced mixing rules to
24. Sadeghi R. Extension of the segment-based Wilson and NRTL
association theories. Hoboken: Wiley; 2009.
models for correlation of excess molar enthalpies of polymer
3. Flory PJ. Thermodynamics of high polymer solutions. J Chem
solutions. J Chem Thermodyn. 2005;37(9):1013–8.
Phys. 1942;10(1):51–61.
25. Madeira PP, Xu X, Wu Y-T, Teixeira JA, Macedo EA. Liq-
4. Huggins ML. Solutions of long chain compounds. J Chem Phys.
uid–liquid equilibrium of aqueous polymer two-phase sys-
1941;9(5):440.
tems using the modified Wilson equation. Ind Eng Chem Res.
5. Wilson GM. Vapor-liquid equilibrium. XI. A new expression for
2005;44(7):2328–32.
the excess free energy of mixing. JACS. 1964;86(2):127–30.
26. Pazuki G, Azimaie R, Taghikhani V, Vossoughi M. Exten-
6. Renon H, Prausnitz JM. Local compositions in thermodynamic
sion of the Wilson-NRF Gibbs energy model in correlating
excess functions for liquid mixtures. AIChE J. 1968;14(1):135–44.
vapor–liquid and liquid–liquid phase behavior of polymer–
7. Abrams DS, Prausnitz JM. Statistical thermodynamics of liquid
polymer aqueous two-phase systems. J Dispersion Sci Technol.
mixtures: a new expression for the excess Gibbs energy of partly
2009;30(4):534–9.
or completely miscible systems. AIChE J. 1975;21(1):116–28.
27. Sadeghi R, Shahebrahimi Y, Zonnouri A. The chemical thermo-
8. Chen C-C. A segment-based local composition model for
dynamic models for calculating the solvent activity coefficient of
the Gibbs energy of polymer solutions. Fluid Phase Equilib.
semidiluted aqueous and nonaqueous polymer solutions in vapor–
1993;83:301–12.
liquid equilibrium. J Chem Eng Data. 2015;60(9):2701–8.
9. Derr E, Deal C, editors. Analytical solutions of groups: correla-
28. Hamta A, Mohammadi A, Dehghani, Feyzi F. Liquid–liquid equi-
tion of activity coefficients through structural group parameters.
librium and thermodynamic modeling of aqueous two-phase sys-
Institution of Chemical Engineers symposium series; 1969.
tem containing polypropylene glycol and ­NaClO4 at T = (288.15
10. Fredenslund A, Jones RL, Prausnitz JM. Group-contribution esti-
and 298.15) K. J Solution Chem. 2018;47(1):1–25.
mation of activity coefficients in nonideal liquid mixtures. AIChE
29. Guo B, Chen L, Liu Y, Yan Z. Modification of the extended UNI-
J. 1975;21(6):1086–99.
QUAC model to study the phase behavior of the [Bmim] Cl–
11. Larsen BL, Rasmussen P, Fredenslund A. A modified UNIFAC
K2HPO4 aqueous two phase system at 298.15 K. J Solution Chem.
group-contribution model for prediction of phase equilibria and
2017;46(5):1059–67.
heats of mixing. Ind Eng Chem Res. 1987;26(11):2274–86.
30. Jahani F, Shahebrahimi Y. Water activities of binary and ter-
12. Gmehling J, Li J, Schiller M. A modified UNIFAC model. 2. Pre-
nary aqueous aluminium salts and poly (ethylene glycol) solu-
sent parameter matrix and results for different thermodynamic
tions at T = 303.15 K: experimental and correlation. J Mol Liq.
properties. Ind Eng Chem Res. 1993;32(1):178–93.
2015;208:183–90.
13. Sayegh S, Vera J. Lattice-model expressions for the combina-
31. Mobalegholeslam P, Bakhshi H. A new model of excess
totial entropy of liquid mixtures: a critical discussion. Chem
gibbs energy for systems containing polymer–salt–water
Eng J. 1980;19(1):1–10.
applicable to aqueous two phase systems. J Solution Chem.
14. Maghami M, Seyf JY, Sadrameli S, Haghtalab A. Liquid–liquid
2016;45(12):1826–41.
phase equilibrium in ternary mixture of waste fish oil biodiesel–
32. Bakhshi H, Mobalegholeslam P. Phase equilibria calculations of
methanol–glycerol: experimental data and thermodynamic mod-
electrolyte solutions containing water–polymer–salt using a new
eling. Fluid Phase Equilib. 2016;409:124–30.
thermodynamic model, applicable in aqueous two phase systems.
15. Asoodeh A, Eslami F, Sadrameli SM. Liquid–liquid equilibria
Fluid Phase Equilib. 2017;434:222–32.
of systems containing linseed oil biodiesel + methanol + glyc-
33. Mobalegholeslam P, Moayyedi G, Majed AA. A new proposed
erol: experimental data and thermodynamic modeling. Fuel.
thermodynamic model for aqueous polymer solutions. J Mol Liq.
2019;253:460–73.
2018;253:53–60.
16. Ye K, Freund H, Sundmacher K. Modeling (vapour–liquid)
34. Sadeghi A, Nazem H, Rezakazemi M, Shirazian S. Predictive con-
and (vapour–liquid–liquid) equilibria of water (­ H2O) + metha-
struction of phase diagram of ternary solutions containing poly-
nol (MeOH) + dimethyl ether (DME) + carbon dioxide (­ CO2)
mer/solvent/nonsolvent using modified Flory-Huggins model. J
quaternary system using the Peng–Robinson EoS with Wong–
Mol Liq. 2018;263:282–7.
Sandler mixing rule. J Chem Thermodyn. 2011;43(12):2002–14.
35. Edrisi S, Bakhshi H, Rahimnejad M. Aqueous two-phase systems
17. Patterson D. Role of free volume in polymer solution thermo-
for cephalexin monohydrate partitioning using poly ethylene gly-
dynamics. Pure Appl Chem. 1972;31(1–2):133–50.
col and sodium tartrate dihydrate: experimental and thermody-
18. Elbro H, Fredenslund A, Rasmussen P. A new simple equa-
namic modeling. Korean J Chem Eng. 2019;36(5):780–8.
tion for the prediction of solvent activities in polymer solutions.
36. Perry RH, Green DW, Maloney JO. Perry’s chemical engineers’
Macromolecules. 1990;23(21):4707–14.
handbook. New York: McGraw-Hill; 1997.
19. Kannan D, Duda J, Danner R. A free-volume term based on the
37. Kang C, Sandler S. Phase behavior of aqueous two-polymer sys-
van der Waals partition function for the UNIFAC model. Fluid
tems. Fluid Phase Equilib. 1987;38(3):245–72.
Phase Equilib. 2005;228:321–8.
38. Haghtalab A, Asadollahi MA. An excess Gibbs energy model to
20. Radfarnia HR, Ghotbi C, Taghikhani V, Kontogeorgis GM. A
study the phase behavior of aqueous two-phase systems of polyeth-
modified free-volume-based model for predicting vapor–liquid
ylene glycol + dextran. Fluid Phase Equilib. 2000;171(1):77–90.
and solid–liquid equilibria for size asymmetric systems. Fluid
39. Kikic I, Alessi P, Rasmussen P, Fredenslund A. On the combinato-
Phase Equilib. 2005;234(1):94–100.
rial part of the UNIFAC and UNIQUAC models. Can J Chem Eng.
21. Radfarnia HR, Kontogeorgis GM, Ghotbi C, Taghikhani V. Clas-
1980;58(2):253–8.
sical and recent free-volume models for polymer solutions: a
comparative evaluation. Fluid Phase Equilib. 2007;257(1):63–9.

13
A. Amirsoleymani et al.

40. Choudalakis G, Gotsis A. Free volume and mass transport in 49. Eliassi A, Modarress H, Mansoori GA. Measurement of activity of
polymer nanocomposites. Curr Opin Colloid Interface Sci. water in aqueous poly (ethylene glycol) solutions (effect of excess
2012;17(3):132–40. volume on the Flory–Huggins χ-parameter). J Chem Eng Data.
41. Kontogeorgis GM, Kouskoumvekaki IA, Michelsen ML. A com- 1999;44(1):52–5.
ment on Guggenheim-like hard-core volume expressions. Ind Eng 50. Wohlfarth C. Vapour–liquid equilibrium data of binary polymer
Chem Res. 2002;41(18):4686–8. solutions(vapour pressures, henry-constants and segment-molar
42. Van Krevelen DW, Te Nijenhuis K. Properties of polymers: their excess gibbs free energies). PSD. 1994.
correlation with chemical structure; their numerical estimation 51. King RS, Blanch HW, Prausnitz JM. Molecular thermodynam-
and prediction from additive group contributions. Amsterdam: ics of aqueous two-phase systems for bioseparations. AIChE J.
Elsevier; 2009. 1988;34(10):1585–94. https​://doi.org/10.1002/aic.69034​1002.
43. Larsen BL, Rasmussen P, Fredenslund A. A modified UNIFAC 52. Grossmann C, Tintinger R, Zhu J, Maurer G. Partitioning of low
group-contribution model for prediction of phase equilibria and molecular combination peptides in aqueous two-phase systems
heats of mixing. Ind Eng Chem Res. 1987;26(3):2274–86. https​ of poly(ethylene glycol) and dextran in the presence of small
://doi.org/10.1021/ie000​71a01​8. amounts of K ­ 2HPO4/KH2PO4 buffer at 293  K: experimental
44. Kim J, Choi E-H, Yoo K-P, Lee CS. Measurement of activities results and predictions. Biotechnol Bioeng. 1998;60(6):699–711.
of solvents in binary polymer solutions. Fluid Phase Equilib. 53. Diamond AD, Hsu JT. Fundamental studies of biomolecule
1999;161(2):283–93. partitioning in aqueous two-phase systems. Biotechnol Bioeng.
45. Palamara JE, Zielinski JM, Hamedi M, Duda JL, Danner RP. 1989;34(7):1000–14.
Vapor–liquid equilibria of water, methanol, and methyl acetate in 54. Forciniti D, Hall C, Kula M-R. Influence of polymer molecular
poly (vinyl acetate) and partially and fully hydrolyzed poly (vinyl weight and temperature on phase composition in aqueous two-
alcohol). Macromolecules. 2004;37(16):6189–96. phase systems. Fluid Phase Equilib. 1991;61(3):243–62.
46. Zafarani-Moattar MT, Yeganeh N. Isopiestic determination of 55. Zaslavsky BY. Aqueous two-phase partitioning: physical chemis-
2-propanol activity in 2-propanol + poly (ethylene glycol) solu- try and bioanalytical applications. Boca Raton: CRC Press; 1994.
tions at 25°C. J Chem Eng Data. 2002;47(1):72–5.
47. Ninni L, Camargo M, Meirelles A. Water activity in poly Publisher’s Note Springer Nature remains neutral with regard to
(ethylene glycol) aqueous solutions. Thermochim Acta. jurisdictional claims in published maps and institutional affiliations.
1999;328(1):169–76.
48. Malcolm G, Rowlinson J. The thermodynamic properties of aque-
ous solutions of polyethylene glycol, polypropylene glycol and
dioxane. Trans Faraday Soc. 1957;53:921–31.

13

You might also like