polymers
Article
Thermodynamic Modeling and Validation of the Temperature
Influence in Ternary Phase Polymer Systems
Marta Romay, Nazely Diban *






Citation: Romay, M.; Diban, N.;
Urtiaga, A. Thermodynamic
Modeling and Validation of the
Temperature Influence in Ternary
Phase Polymer Systems. Polymers
2021, 13, 678. https://doi.org/
10.3390/polym13050678
Academic Editors: David Mecerreyes,
Rafael Antonio Balart Gimeno,
Ana María Díez-Pascual, Vicente
Compañ Moreno, Angels Serra and
Rebeca Hernandez Velasco
Received: 4 February 2021
Accepted: 20 February 2021
Published: 24 February 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Polymers 2021, 13, 678. https://doi.org/10.3390/polym13050678
and Ane Urtiaga
Department of Chemical and Biomolecular Engineering, ETSIIyT, University of Cantabria, Avda. Los Castros s/n,
39005 Santander, Spain; romaym@unican.es (M.R.); urtiaga@unican.es (A.U.)
* Correspondence: dibann@unican.es; Tel.: +34-942-206-778
Abstract: The effect of the temperature, as a process variable in the fabrication of polymeric mem-
branes by the non-solvent induced phase separation (NIPS) technique, has been scarcely studied.
In the present work, we studied the influence of temperature, working at 293, 313 and 333 K, on the
experimental binodal curves of four ternary systems composed of PVDF and PES as the polymers,
DMAc and NMP as the solvents and water as the non-solvent. The increase of the temperature
caused an increase on the solubility gap of the ternary system, as expected. The shift of the binodal
curve with the temperature was more evident in PVDF systems than in PES systems indicating the
influence of the rubbery or glassy state of the polymer on the thermodynamics of phase separation.
As a novelty, the present work has introduced the temperature influence on the Flory–Huggins model
to fit the experimental cloud points. Binary interaction parameters were calculated as a function of
the temperature: (i) non-solvent/solvent (g12 ) expressions with UNIFAC-Dortmund methodology
and (ii) non-solvent/polymer (χ13 ) and solvent/polymer (χ23 ) using Hansen solubility parameters.
Additionally, the effect of the ternary interaction term was not negligible in the model. Estimated
ternary interaction parameters (χ123 ) presented a linear relation with temperature and negative
values, indicating that the solubility of the polymers in mixtures of solvent/non-solvent was higher
than expected for single binary interaction. Finally, PES ternary systems exhibited higher influence
of the ternary interaction parameter than PVDF systems.
Keywords: Flory–Huggins theory; binodal curve; temperature influence; ternary interaction; Hansen
solubility parameter; polyvinylidene fluoride (PVDF); polyethersulfone (PES)
1. Introduction
Non-solvent induced phase inversion (NIPS) is a synthesis technique commonly used
to manufacture asymmetric polymeric membranes for separation processes [1], both in
laboratory and industrial scale [2,3]. The mechanism of membrane formation using NIPS
consists in the precipitation of a polymer solution when it is introduced in a non-solvent
bath by the exchange between solvent and non-solvent or coagulant [4]. The liquid–
liquid demixing that occurs in the NIPS process, and ultimately the membrane structure,
are strongly affected by the thermodynamics and kinetics of the system.
Ternary phase diagrams, which are temperature dependent, are employed to depict
the thermodynamics and kinetic characteristics of a certain NIPS process. Figure 1 shows a
ternary diagram in which the continuous and dashed blue curves represent the thermody-
namic binodal and spinodal curves, respectively. The area between both curves is called
the metastable region. The point where these curves meet is the critical point. The binodal
curve defines the boundary between a homogeneous phase and a heterogeneous phase.
When the precipitation pathway (yellow line) crosses the binodal curve (point B) the poly-
mer nucleation takes place and the polymer solution separates into two equilibrium phases,
polymer-rich (B0 ) and polymer-lean (B”) phases. These equilibrium compositions are lo-
cated on the binodal curve and depict a tie-line in the ternary phase diagram. The pathway
https://www.mdpi.com/journal/polymers
Polymers 2021, 13, x FOR PEER REVIEW 2 of 20

Polymers 2021, 13, 678 2 of 19

rium phases, polymer-rich (B′) and polymer-lean (B″) phases. These equilibrium compo-
sitions are located on the binodal curve and depict a tie-line in the ternary phase diagram.
The pathway
between between
points points
B-C defines theB-C defines theofprecipitation
precipitation of the entire
the entire membrane, and membrane,
finally, fromand
C to
finally, from C to D points the rest of the solvent is exchanged
D points the rest of the solvent is exchanged with the non-solvent.with the non-solvent.

Figure 1. (a) Theoretical ternary phase diagram representing the polymer (P), the solvent (S) and the non-solvent (NS)
Figure 1. (a) Theoretical ternary phase diagram representing the polymer (P), the solvent (S) and the non-solvent (NS)
compositions and (b) structures obtained depending on the precipitation pathway. Adapted from [5,6].
compositions and (b) structures obtained depending on the precipitation pathway. Adapted from [5,6].
The final membrane morphology is described by the distribution and shape of the
The final membrane morphology is described by the distribution and shape of the
pores created in the NIPS process. The position of the binodal curve and the rate of
pores created in the NIPS process. The position of the binodal curve and the rate of sol-
solvent/non-solvent exchange modify the pore shape. When the binodal curve is located
vent/non-solvent exchange modify the pore shape. When the binodal curve is located near
near the polymer–solvent axis, the precipitation pathway reaches more rapidly the binodal
the polymer–solvent
curve so big finger-like axis,
pores theareprecipitation pathway when
expected. Otherwise, reaches themore rapidly
binodal curvethe binodal
is displaced
curve
to theso bigside
right finger-like pores aresmaller
of the diagram, expected. Otherwise,
sponge-like pores when the binodal
are obtained. Thecurve
entryispoint
dis-
placed to the right side of the diagram, smaller sponge-like pores
(B) of the precipitation pathway into the heterogeneous phase affects the pore distribution.are obtained. The entry
point (B) of the
Depending precipitation
on the pathway followedpathwayduring
into the heterogeneous phase
solvent/non-solvent affects different
exchange the pore solid
dis-
tribution.
phase morphologies can occur (Figure 1b), i.e., dense membranes (pathway 1) weredif-
Depending on the pathway followed during solvent/non-solvent exchange ob-
ferent
tainedsolid
when phase morphologies
the pathway crossescantheoccur (Figure 1b), i.e.,region,
gel homogeneous dense at membranes
sufficiently (pathway
high initial1)
were obtained when the pathway crosses the gel homogeneous
polymer concentration causing slow exchange due to high viscosity; binodal decompositionregion, at sufficiently high
initial
occurspolymer concentration
if the precipitation causing
pathway slow
enters theexchange
heterogeneousdue tophase high through
viscosity;thebinodal de-
metastable
composition occurs if the precipitation pathway enters the
region, to form (pathway 2) open-cellular or closed-cellular membrane morphologies,heterogeneous phase through
the metastable
or (pathway 4) region,
nodulartopore form (pathway 2)
distributions; open-cellular
otherwise, spinodal or closed-cellular
decompositionmembrane
takes place
morphologies, or (pathway
if the precipitation pathway 4)enters
nodular thepore distributions;
heterogeneous phaseotherwise,
throughspinodal decompo-
the heterogeneous
sition
phase, takes place if(pathway
obtaining the precipitation pathway (lacy)
3) bicontinuous entersdistribution
the heterogeneous of poresphase through
[6]. It is deemedthe
heterogeneous
that, due to thephase, obtaining
significant (pathway
dependence 3) bicontinuous
of membrane (lacy) distribution
morphology of pores [6].
on thermodynamics and
Itkinetics
is deemed that, due to the significant dependence of membrane
of the ternary phase systems, the development of mathematical models able to morphology on ther-
modynamics and kinetics of the
predict the thermodynamic andternary
kinetic phase systems, of
characteristics thedifferent
development of mathematical
polymer/solvent/non-
models
solventable to predict
systems wouldthe thermodynamic
facilitate the decisionand on kinetic characteristics
the selection of different
of membrane poly-
synthesis
mer/solvent/non-solvent systems concentration,
variables, such as initial polymer would facilitate the decision
process temperatureon theorselection of mem-
even solvent/non-
brane
solventsynthesis variables, such as initial polymer concentration, process temperature or
combination.
even solvent/non-solvent
Different mathematical combination.
models have been developed to describe the thermodynamics
of ternary
Differentsystems, such as Flory–Huggins
mathematical models have been (FH)developed
and compressible
to describeregular solution theory
the thermodynam-
(CRS) [7,8]. FH theory, characterized by its relative mathematical
ics of ternary systems, such as Flory–Huggins (FH) and compressible regular solution simplicity and good
the-
prediction
ory of phase
(CRS) [7,8]. behavior
FH theory, [9], is based
characterized byonits the Gibbs
relative free energy of
mathematical a mixture
simplicity and and
goodthe
use of binary
prediction interaction
of phase behaviorparameters.
[9], is basedOneonofthetheGibbs
limitations of the of
free energy FHa model
mixtureisand thatthethe
binary
use interaction
of binary parameters
interaction are traditionally
parameters. One of the obtained
limitations in experimental
of the FH model assays.is Many
that the ef-
forts have
binary been made
interaction to obtain
parameters are these parameters
traditionally obtainedfromintabulated
experimentalsolubility
assays.parameters.
Many ef-
Hansen’s
forts have solubility
been madeparameters
to obtain theseare commonly
parameters usedfrombecause of thesolubility
tabulated huge amount of data
parameters.
available for polymers and solvents. However, solubility parameters for new materials
(i.e., green solvents or copolymers) are seldom accessible and need to be calculated with
Polymers 2021, 13, 678 3 of 19

group contribution methods such as Hoftyzer and Van Krevelen [10], Hoy [10], Just [11] or
Stefanis [12]. Nevertheless, solubility parameters calculated with this methodology do not
give accurate predictions yet.
Polyethersulfone (PES) and polyvinylidene fluoride (PVDF) are polymers classically
used to manufacture membranes for different industrial separation applications. Partic-
ularly PVDF is a polymer with interest to be used for the development of photocatalytic
membranes [4]. Temperature plays an important role in the manufacture of polymeric
membranes. From the thermodynamic viewpoint, increasing the temperature of the ternary
system will increase the demixing gap [2]. Meanwhile, temperature varies the diffusion
parameters of the solvent and non-solvent and therefore, the kinetics of exchange between
them. Thus changes in the processing temperature affect the final membrane morphol-
ogy [1]. Despite the fact that during the industrial synthesis of polymeric membranes the
temperature of the polymer solution or the coagulation bath is often set different than
room temperature, few works have addressed the experimental analysis of the effect of
temperature on the thermodynamics of ternary systems [6,13]. Furthermore, there is a
lack of validations of predictive models considering temperature for PVDF and few works
address the temperature effect for the synthesis of PES membranes. Kahrs et al., evalu-
ated the influence of temperature on the thermodynamics of PES ternary systems [6,13].
When they compared PES membranes fabricated with N-methyl pyrrolidone (NMP) and
2-pyrrolidone (2Pyr) as the solvents, the SEM images showed an increase in the number of
pores when the temperature was increased from 20 to 40 ◦ C, because the reduction of the
viscosity favored the nuclei formation in the lean-polymer phase [6].
The aim of this work is on the one hand, to develop an enhanced thermodynamic
model based on Flory–Huggins theory that incorporates the effect of temperature to predict
the binodal curves of different ternary phase diagrams. On the other hand, the model
was validated experimentally, considering four ternary systems, using PVDF and PES
as the polymers; N,N-dimethylacetamide (DMAc) and N-methyl pyrrolidone (NMP) as
the solvents; and water as the non-solvent/coagulant, at three temperatures in the range
293–313 K. The importance of the binary and ternary interaction parameters on the model
predictions was explored and discussed. Additionally, a potential influence of the solid
polymer state (rubbery or glassy) on behavior of the thermodynamics of the ternary systems
at different temperatures was observed.

2. Theoretical Section
2.1. Thermodynamic Model
In this study, Flory–Huggins theory was selected to describe the thermodynamic
model of a ternary system. FH theory is based on the Gibbs free energy of a mixture (∆GM ),
Equation (1). The last term, that represents the ternary interactions of the components, is
usually neglected to simplify the calculations [14]. In this work, the effect of neglecting this
term will also be addressed.
∆G M
= n1 lnφ1 +n2 lnφ2 +n3 lnφ3 + g12 n1 φ2 + χ13 n1 φ3 +χ23 n2 φ3 +χ123 n1 φ2 φ3 (1)
RT
In Equation (1), the subscripts 1, 2 and 3 refer to the non-solvent, solvent and polymer,
respectively. Equation (2) defines the molar volume fraction (φi ) as function of the number
of moles (ni ) and the molar volume (Vi ) of the component i.

ni Vi
φi = (2)
n1 V1 +n2 V2 +n3 V3

The binary interaction parameters (gij or χij ) represent the interaction between each
pair of components. Solvent–polymer (χ23 ) and non-solvent–polymer (χ13 ) are interaction
parameters independent of the concentration. The non-solvent–solvent interaction parame-
ter (g12 ) is a solvent concentration-dependent parameter, function of u2 , which is defined
ers 2021, 13, x FOR PEER REVIEW 4 of 20

Polymers 2021, 13, 678 4 of 19

parameters independent of the concentration. The non-solvent–solvent interaction param-
eter (g12) is a solvent concentration-dependent parameter, function of u2, which is defined
in Equation (7). The ternary interaction parameter (χ123) was estimated through fitting
modeled curves to the experimental
in Equation cloud interaction
(7). The ternary points [15]. parameter (χ123 ) was estimated through fitting
The equilibrium
modeledphases,
curves B′to and B″, connectedcloud
the experimental by the tie-line
points have the same chemical
[15].
potential for each The equilibrium
component (μi), phases,
EquationB0 (3),
andbut B”,differ
connected by the tie-line
in composition. Thehave the same chemical
chemical
potential
potential of each for each
component, component
Equations (µi ), is
(4)–(6), Equation
derived (3),
from but differ in(1).
Equation composition. The chemical
potential of each component, Equations (4)–(6), is derived from Equation (1).
∆μ'i = ∆μ″i (3)
∆µi0 = ∆µi
00
(3)
∆μ1 dg12
= lnϕ1 +1 − ϕ1 − sϕ2 − rϕ3 +(g12 ϕ2 +χ13 ϕ3 )(ϕ2 +ϕ3 ) − χ23 sϕ2 ϕ3 − u1 u2 ϕ2 ( ) +χ123 ϕ2 ϕ3 (1 − 2ϕ1 )
RT du2 (4)
∆µ1
 
dg12
= lnφ1 +1 − φ1 − sφ2 − rφ3 + ( g12 φ2 +χ13 φ3 )(φ2 + φ3 ) − χ23 sφ2 φ3 − u1 u2 φ2 +χ123 φ2 φ3 (1 − 2φ1 ) (4)
∆μ2 RT dg12 du2
s = slnϕ2 +s − sϕ2 − ϕ1 − rϕ3 +(g12 ϕ1 +χ23 sϕ3 )(ϕ1 +ϕ3 ) − χ13 ϕ1 ϕ3 +u1 u2 ϕ1 ( ) +χ123 ϕ1 ϕ3 (1 − 2ϕ
 )
RT ∆µ2 du2 dg12 2 (5)
s = slnφ2 +s − sφ2 − φ1 − rφ3 + ( g12 φ1 +χ23 sφ3 )(φ1 + φ3 ) − χ13 φ1 φ3 +u1 u2 φ1 +χ123 φ1 φ3 (1 − 2φ2 ) (5)
RT du2
∆μ3
r = rlnϕ3∆µ
+r 3− rϕ3 − ϕ1 − sϕ2 +(χ13 ϕ1 +sχ23 ϕ2 )(ϕ1 +ϕ2 ) − g12 ϕ1 ϕ2 + χ123 ϕ1 ϕ2 (1 − 2ϕ3 ) (6)
RT r = rlnφ3 +r − rφ3 − φ1 − sφ2 + (χ13 φ1 +sχ23 φ2 )(φ1 + φ2 ) − g12 φ1 φ2 + χ123 φ1 φ2 (1 − 2φ3 ) (6)
RT
ϕi
ui = φi (7)
ϕ1 +ϕ2 ui = (7)
φ1 + φ2
In the above In equations,
the above s and r are the
equations, molar
s and volume
r are relations
the molar volumeV1/Vrelations
2 and V1/V V31,/V
re-2 and V 1 /V 3 ,
spectively. It respectively.
has been reported
It has that
beenatreported
sufficiently
that high molecular
at sufficiently weight
high of the weight
molecular polymer of the polymer
the phase
the phase diagram diagram
is fairly is fairly insensitive
insensitive to the choice to the choice
of V of V 3 (polymer
3 (polymer molar volume)
molar volume) pro- provided
vided that V1/V 3 isVsufficiently
that 1 /V 3 is sufficiently
small small
[16]. The [16].
last The
term last
of term of
Equations Equations
(4)–(6) (4)–(6)
corresponds corresponds to
the partial derivative of the term for ternary interaction. Most
to the partial derivative of the term for ternary interaction. Most of the literature neglects of the literature neglects
ternary
ternary interaction interaction
among among the components.
the components. In this work, theIn this work,interaction
ternary the ternary interaction
term will term will
be evaluated be evaluated
under under two(1)
two hypotheses: hypotheses:
omitting it(1)oromitting it or (2)it.considering it.
(2) considering

2.2.
2.2. Influence of theInfluence of theinTemperature
Temperature the Model in the Model
Temperature Temperature
is an important is variable
an important
of thevariable of the phase
phase inversion processinversion process
that affects the that affects
chemical potential, the solubility parameters (δ), the binary interaction parameters, and parameters,
the chemical potential, the solubility parameters (δ), the binary interaction
the components anddensity
the components
(ρ) and thus density (ρ)molar
their and thus their molar
volume. volume.
The next The (Figure
diagram next diagram
2) (Figure 2)
schematizes theschematizes the effect of over
effect of temperature temperature over thefunctions
the mentioned mentioned andfunctions and parameters.
parameters.

Figure
Figure 2. Influence
2. Influence of temperature
of the the temperature
on on different
different parameters
parameters that
that areare partofofthe
part thethermodynamic
thermodynamicmodel.
model.
The temperature is directly introduced in the expression of chemical potentials.
The temperature
The molar is directly
volume introduced in the
(Vi ) changes withexpression of chemical
the temperature potentials.
because of the The
temperature de-
molar volumependence
(Vi) changes with the temperature because of the temperature dependence
of the density (ρ), Equation (8). Density values of the solvents and non-solvent at
of the densitydifferent
(ρ), Equation (8). Density
temperatures values
were of the from
retrieved solvents
the and non-solvent
database at different
of the software Aspen Plus V 9 .
temperaturesTable
were 1retrieved
compilesfrom the database
the relevant of the
data and softwareofAspen
properties Plus V9.
the solvents andTable 1
non-solvent for the
compiles the model
relevantin data and properties of the solvents and non-solvent for the model
this study.
in this study. ρ( T )
Vi ( T ) = (8)
Mw
 Polymers 2021, 13, x FOR PEER REVIEW ρ(T) 5 of 20
Polymers 2021, 13, x FOR PEER REVIEW Vi (T) = 5 of 20 (8)
Mw
Polymers 2021, 13, 678 5 of 19
ρ(T)
Table 1. Chemical structure and relevant properties of solventsVand =non-solvents. Data at 298 K are reference(8)
i (T) ρ(T) (ref) values.
Vi (T) = Mw (8)
Mw
TableSolvent Formulaand relevant
Structure Mw and
(g/mol) T (K)at 298 K areρreference
(g/cm ) (ref) values.
Vi (cm3/mol) 3
1. Chemical structure properties of solvents non-solvents. Data
Table 1. Chemical structure and relevant properties of solvents and non-solvents. Data at 298 K are reference (ref) values.
Table 1. Chemical structure and relevant properties of solvents and non-solvents. Data 298
at 298 K are reference0.9365
(ref) values. 93.031
Solvent Formula Structure M (g/mol) T (K) (g/cm 3) V (cm3 /mol)
Solvent
Solvent Formula
Formula Structure
Structure M (g/mol)
w w
Mw (g/mol) T
T (K)(K) ρ (g/cm
293ρ (g/cm3) ρ 3) V
0.9408 i (cm
Vi (cm3/mol)
3/mol)
i 92.599
DMAc C4H9NO 87.12 298298298 0.9365
313 0.9365 0.9233
0.9365 93.031 93.03194.355
93.031
293293293 0.9408
0.9408 92.599 92.599
0.9408 92.599
DMAc DMAcC4 H9CNO
DMAc C4H9NO
4H9NO 87.12
87.12
87.12 313313313 333 0.9233
0.9233 0.9054 94.355 94.35596.224
94.355
0.9233
333333333 298 0.9054 0.9054
0.9054 96.224 96.22496.600
96.224
1.0262
298298298 293 1.0262 1.0302
1.0262
1.0262 96.600 96.60096.231
96.600
NMP C5H9NO 99.13 293293293
313 1.0302 1.0302
1.0302 96.231
1.014196.231 96.23197.754
NMP NMP
NMP C5 H9CNO
C5H
5H9NO
9NO 99.13
99.13
99.13 313313313 1.0141 1.0141 97.754
1.0141 97.754 97.754
333333333 333 0.9977 0.9977 0.9977
99.357 99.35799.357
0.9977 99.357
298298298 298 0.9945 0.9945
0.9945
0.9945 18.114
18.114 18.11418.114
293 293 293 0.9965 0.9965 18.079
0.9965 18.07918.079
Water H O H2O 18.01 293 0.9965 18.079
WaterWaterWater H2 2OH2O 18.01
18.01
18.01 313 313 0.9888 0.9888 18.219
313 313 0.9888 0.9888 18.219 18.21918.219
333 0.9797
333333 333 0.9797 0.9797 18.388 18.388
0.979718.388 18.388
Over the years, the authors have deepened in the concentration dependence of the
binaryOver the years,
interaction the authors
parameters, which have deepened
initially in the concentration
were considered constant. Later, dependence
Altena andof the
Over the
binaryOver years,
the the
interaction authors
years, the
parameters, have
whichdeepened
authors have
initially wereinconsidered
deepened the concentration
in the concentration
constant. dependence
Later, Altena of the of the
dependence
Smolders [17] determined the concentration dependency of g12, and Yilmaz et al. [18] an- and
binary interaction
binary
Smolders
alyzed the [17] parameters,
interaction parameters,
determined
influence which initially
which
the concentration
of concentration were
initially considered
dependency
on χ23, although were
both studies constant.
considered
of g12, agreed
and Yilmaz Later,
constant.
et al. χ
to consider [18]
13
Altena
Later, and and
an- Altena
SmoldersSmolders [17] determined the concentration dependency12of g12, and Yilmaz et [18]
as [17]
alyzed determined
the
constant influence
parameter.of the concentration
concentration
Currently, it is on χ
widely 23 , dependency
although
agreed that both
only of
studies
g 12 isga , and
agreed Yilmaz
to consider
concentration-de- et
χ al.
13 al. [18] an-
analyzed as constant
pendent
the parameter.
parameter
influence Currently, it ison
[15,19–22].
of concentration widely agreed that only g12 is a concentration-de-
alyzed the influence of concentration χon 23 , although
χ23, although bothboth studies agreed
studies agreedto consider
to consider χ13
pendent parameter
Interaction [15,19–22].
parameters χ13 and χ23 can be obtained experimentally; light scattering,
χ13 as constant
asvapor
constant parameter.
parameter. Currently,
Currently, it isitwidely
isbewidely agreed agreed thatthat onlyonly g12 isg12a is
concentration-
a concentration-de-
pressure
Interaction depression
parametersand χ13 membrane
and χ23 canosmometry obtained are techniques
experimentally; commonly pro-
light scattering,
dependent
pendent
posedparameter
to parameter[15,19–22].
determine χ [15,19–22].
, and equilibrium swelling for
vapor pressure depression and membrane osmometry are techniques commonly pro-
23 χ 13 . However, reliable estimations
Interaction
can be
posed parameters
obtained
Interaction using
to determine χEquation
23, χ
parameters 13 and
and (9),χdefined canby
χ1323and
equilibrium χbe
swelling
23 obtained
Hansen
can [23]
be
for and. experimentally;
χobtained
13 widelyexperimentally;
However, used light
litera- scattering,
in theestimations
reliable light scattering,
turebe
vapor pressure
can [15,24–29].
obtained This
depression expression
using and
Equation uses
membrane
(9), Hansen
defined solubility
osmometry
by Hansen parameters
are
[23] that
techniques
and are used
widely composed
commonly
in theoflitera-
proposed
vapor pressure depression and membrane osmometry are techniques commonly pro-
three[15,24–29].
ture contributions: dispersive (δd),uses
polarHansen
(δp) andsolubility
hydrogen parameters
bonding (δh).that Equation (9) also of
to determine
posed , andThis
χ23determine
to expression
equilibrium
χ23, and swelling for χ13 . However, for χ13.reliable are composed
estimations can
includes the correction factor (αij). equilibrium swelling However, reliable estimations
three using
be obtained contributions:
Equation dispersive
(9), (δd), polarby
defined (δp) Hansen
and hydrogen [23]bonding
and widely (δh). Equation
used (9) also
in used
the litera-
can be obtained
includes using
the correction Equation
Vi factor (αij). 2 (9), defined2by Hansen 2[23] and widely in the litera-
ture [15,24–29]. This χij expression
ture [15,24–29]. This = αij
RT
[(δi,d −uses
expressionδj,d ) Hansen
+0.25(δi,p −
uses solubility
Hansenδj,p ) +0.25(δparameters
solubility δ ) ] that are composed
(9)
i,h j,h parameters that are composed of of
three contributions: dispersiveV i (δ ), polar
2 (δ ) and hydrogen
2 bonding
2 (δ ). Equation (9)
three Table
contributions:
χ =
2 compilesα ij the [(δ
dispersive − d δ )
(δd), polar
+0.25(δ p − )
i,p (δp)j,pand
δ forhydrogen
+0.25(δ δ ) ]
j,h bonding h (9)
(δh). Equation (9) also
also includes thethe
ij
correction RT Hansen’s
i,d j,dsolubility parameters the i,h
polymers and solvents
of interest
includes in the present factor
correction study,
factorat(αthe ).reference
ij(α ij). temperature of 298 K [23].
Table 2 compiles the Hansen’s solubility parameters for the polymers and solvents
of interest
Table in theV
2. Hansen’s present Vstudy,
 parameters
solubility i
at the
at reference
2298 K [23].
2 temperature

2 of 2298 K [23]. 2 2
i (9)
χij = αijχij = αijδi,d 1/2[(δ δ j,d− δ+j,d0.25
− i,d ) +0.25(δ
δi,p −i,p δ− +)0.25
j,p δj,p +0.25(δ δj,h ) ]
δi,h i,hδ j,h (9)
Table 2. Hansen’sRT
RTparameters
δd (MPa)
solubility δat
p (MPa)1/2
298 K [23]. δh (MPa)1/2 δT (MPa)1/2
PVDF 17.2 12.5 9.2
Table 2 compiles the Hansen’s solubility parameters for23.167
the polymers and solvents
Table 2 compiles
PES theδdHansen’s
(MPa)1/2 solubility
19.6 δ10.8
p (MPa) 1/2
parameters δ9.2 (MPa)
h for 1/2
the polymers δT (MPa)
24.196 and1/2solvents of
of interest
PVDF
DMAc
in the present
17.2
16.8
study, at the reference
12.5temperature
11.5
temperature of
9.2298 K [23].
10.2of
298
22.771
K
23.167[23].
interest in the present study, at the reference
PES
NMP 19.6
18 10.8
12.3 7.2 9.2 22.959 24.196
TableDMAc
2. Hansen’s solubility
Water 16.8 parameters
15.5 at 298 K [23].42.310.2
1611.5 47.807 22.771
Table 2. Hansen’s solubility parameters at 298 K [23].
NMP 18 12.3 7.2 22.959
δ15.5
The dependency of
Water
(MPa)1/2
dsolubility parameters δp (MPa)
16with1/2
1/2
the temperature δh (MPa)
42.3 can be1/2
1/2
calculated with
47.807
δT (MPa)1/2
(MPa) 1/2 δTK)(MPa)1/2
δd [23].
Equations (10)–(12) p (MPa)
Molar volumesδ(V δh (MPa)
and Vref, at the reference temperature of 298
PVDF 17.2 12.5 9.2 23.167
for solvents and non-solvent
PVDF 17.2 can be found in12.5
Table 1. For polymers,9.2the molar volumes in23.167
The dependency of solubility
PES (10)–(12) refer to19.6 parameters with the temperature can be calculated with
Equations the molar volume of 10.8the monomer (V* and V* 9.2
ref) calculated
24.196
PES 19.6 Molar
Equations (10)–(12) [23]. volumes (V 10.8 9.2 temperature
and Vref, at the reference 24.196
of 298 K)
DMAc
using 16.8
Equations (13) and (14) [10]. Equation (13) is 11.5
employed for the rubbery10.2
polymer and 22.771
DMAc 16.8
for solvents and non-solvent can be found11.5 10.2 the molar volumes
in Table 1. For polymers, 22.771
in
NMP NMP 18
18 to the molar volume
Equations (10)–(12) refer 12.3
12.3 of the monomer7.2 7.2 22.959 22.959
(V* and V*ref) calculated
Water Water
using and 15.5
Equations (13)15.5 (14) [10]. Equation 16employed for42.3
16(13) is 42.3 polymer
the rubbery and 47.807
47.807

The dependency
The dependency of solubility
of solubility parameters
parameters with thewith the temperature
temperature can be calculated
can be calculated with with
Equations (10)–(12) [23]. Molar volumes (V and V ref, at the reference temperature of 298 K)
Equations (10)–(12) [23]. Molar volumes (V and Vref , at the reference temperature of 298 K)
for solvents
for solvents and non-solvent
and non-solvent can becan be found
found in Tablein Table
1. For1.polymers,
For polymers, the molar
the molar volumesvolumes in
Equations (10)–(12) refer to the molar volume of the monomer (V*
in Equations (10)–(12) refer to the molar volume of the monomer (V* and V*ref ) calculated and V* ref) calculated

using Equations
using Equations (13)(14)
(13) and and[10].
(14) Equation
[10]. Equation
(13) is(13) is employed
employed for the
for the rubbery
rubbery polymer and
polymer
and Equation (14) for glassy polymers. These equations are function of the van der Waals
volume (Vw ), the temperature and, in the case of glassy polymers, the glass transition
temperature (Tg ).
Vre f 1.25
 
δd = δd,re f (10)
V
 ume (Vw), the temperature and, in the case of glassy polymers, the glass transition temper-
ature (Tg).
Polymers 2021, 13, x FOR PEER REVIEW 1.25 6 of 20
Vref
δd = δd,ref ( ) (10)
V
Polymers 2021, 13, 678 6 of 19
Equation (14) for glassy polymers. These equations are function of the van der Waals vol-
Vref 0.5 the glass transition temper-
ume (Vw), the temperature and, in the case of glassy polymers,
ature (Tg). δp = δp,ref ( ) (11)
1.25
V
Vref 0.5
Vre f

δd = δd,refδ ( = )δ (10)
p V p,re f 0.5 (11)
-3
V V ref
δh = δh,ref ⁄exp " [-1.22·10 (Tref -T)- ln (( ) )]  !# (12)
Vref 0.5   V V 0.5
δp = δp,ref ( ) −3 re f (11)
δh = δh,re f / exp −1.22V·10 Tre f − T − ln (12)
∗ -3
V
Rubbery → Vi (T) = Vw [1.30+10 T] (13)
-3
Vref h0.5 i
δh = δh,ref ⁄Rubbery
exp [-1.22·10 (T ∗-T)- ln (( ) )] − 3 (12)
→ V ref ( T ) = V w 1.30 + 10
i V T (13)
Glassy → V∗i (T) = Vw [1.30+0.55 h 10 -3
T g + 0.45 10 -3
T] i (14)
→ V i∗→(VT∗i )(T)
Glassy Rubbery == V 1.30+
Vww[1.30+10 -3
0.55 10−3 Tg + 0.45 10−(13)
T] 3
T (14)
Table 3 compiles the chemical structure and parameters relevant for the mathemati-
cal model
Table 3ofcompiles
the monomer
Glassy the unit
(T) = Vwof
→ V∗ichemical the polymers.
structure
[1.30+0.55 10-3and
Tg + 0.45 10-3 T]
parameters relevant for (14)the mathematical
modelTable
of the monomer unit of the polymers.
3 compiles the chemical structure and parameters relevant for the mathemati-
Table 3.cal
Chemical structure and properties of monomer units of the polymers.
model of the monomer unit of the polymers.
Table 3. Chemical structure and properties of monomer units of the polymers.
Polymer Formula Structure 3 ρ (g/cm ) Mw* (g/mol)
3
Table 3. Chemical structure and properties of monomer units of the polymers.
Vw (cm /mol) T (K) Vi* (cm3/mol)
3 Mw * (g/mol) 3 298 Vi *35.896
(cm3 /mol)
Polymer Formula
Polymer Formula Structure
Structure ρ 3(g/cm
ρ (g/cm ) Mw)* (g/mol) Vw (cm3/mol) Vw (cm Vi* (cmT3/mol)
T (K)/mol) (K)
298 293
35.896
298 35.839
35.896
PVDF C2H2F2 1.76 64.03 25.03
293 313
35.839
293 36.064
35.839
PVDF PVDF C2 H2C
F22H2F2 1.76 1.76 64.03 64.03 25.03 25.03
313 36.064
313 36.064
333
333
36.290
36.290
333 36.290
298 298
298
190.997 190.997
190.997
293 293
293
190.745 190.745
190.745
PES
PES C12
PES C HC88SO
12H SO
12H833SO3 1.38
1.38 1.38 232.25 232.25 112.15
232.25 112.15
112.15
313 313
191.754
313 191.754
191.754
333 333
192.763 192.763
333 192.763
* Molecular weight and molar volume of the monomer unit.
* Molecular weight and molar volume of the monomer unit.
* Molecular weight and molar volume of the monomer unit.
Equation (18) presents the effect of concentration on the binary interaction parameter
Equation
between solvent(18)
Equation and presents
(18) presentsthe
non-solvent. It effect
the iseffectof
ofconcentration
obtained on
onthe
from the relationship,
concentration binary
Equation
the interaction
binary(15), be-
interactionparameter
parameter
tween thesolvent
excess of and
Gibbsnon-solvent.
free energy (GE), It
between
between solvent non-solvent. ItEquation (16), andfrom
isisobtained
obtained the Gibbs
from thethefree energy for a Equation (15),
relationship,
relationship, Equation (15), be-
binary mixture (ΔGM), Equation (17) [30]. E
between theexcess
tween the excessofofGibbs
Gibbsfree
freeenergy
energy(G (GE),),Equation
Equation(16),
(16),and
andthe
theGibbs
Gibbsfree
freeenergy
energyfor
fora
N
abinary
binarymixture
mixture(ΔG(∆GMM ),
),∆G Equation
Equation (17)
(17)
M = ∆GM, ideal
E [30].
+G [30].
= RT ∑ xi ln xi +G E
(15)
i=1

E E NN
RT∑ (15)
E
∆G = ∆GM, ideal +GE = = RT
∆G M =MG∆G =M,
x1 ln
RT
γ +
ideal +xG
2 ln γ
1
∑i=1xxi ln
2
i ln
xix+
i +G
GE (16) (15)
i =1
∆GM E
lnEϕG
= x1G +x2 ln ϕ2 +g12 x1 ϕ2 (17) (16)
= =x1xln1 ln
γ1γ+ +x ln γ
RT 1
RT 1 x 2 2ln γ2 2 (16)
RT
GE
g12 = ( ∆G
+x lnM
x1 +x ln x -x ln ϕ -x ln ϕ )⁄x ϕ2 (18)
RT 1 ∆G M2 x 2ln 1φ +1x 2ln φ 2 + g 1 x
= 2 +g12 x11φ (17)
(17)
RT RT = 1x1 ln 1ϕ +x
1
2 2 ln ϕ
2 12
ϕ2
2
GE is calculated from the
 activity coefficient (γ) of vapor–liquid equilibrium
 data, ob-
E
G E
tained with the modified UNIFAC-Dortmund methodology based on groups contribution
g12 = G+ x1 ln x1 + x2 ln x2 − x1 ln φ1 − x2 ln φ2 /x1 φ2 (18)
g12 =RT
( +x1 ln x1 +x2 ln x2 -x1 ln ϕ1 -x2 ln ϕ2 )⁄x1 ϕ2 (18)
RT
GE is calculated from the activity coefficient (γ) of vapor–liquid equilibrium data,
GE iswith
obtained calculated from theUNIFAC-Dortmund
the modified activity coefficient (γ) of vapor–liquid
methodology equilibrium
based on groupsdata, ob-
contri-
tained of
bution with
thethe modified
activity UNIFAC-Dortmund
coefficients methodologyisbased
[31]. This methodology on groups
described contribution
in Appendix A.
A polynomial expression suggested by Tompa [32] was used to fit the g12 as a function of
the solvent concentration, Equation (19).

g12 (φ2 ) = ao + a1 φ2 + a2 φ22 + a3 φ23 . . . (19)

2.3. Calculation Procedure

In this model there are six variables corresponding to the volume fraction of the three
components in the two phases. The set of equations is formed by the three equilibrium
Polymers 2021, 13, 678 7 of 19

relationships for each component (Equation (3)) and the two mass balance equations in
each phase, Equation (20).
3 3
∑ φi0 = ∑ φi
00
= 1 (20)
i =1 i =1

The system of equations is solved using KNITRO solver in GAMS Development

Corporation 27.3.0, USA by minimizing the objective function F, Equation (21), as the equi-
librium is found when the chemical potential of each component is the same in both phases.
0
The polymer lean-phase molar volume fraction (φ3 ) is selected as an independent variable.

00
2
2 
2
F = ∆µ10 −∆µ1 + s ∆µ20 −∆µ200 + r ∆µ30 −∆µ300

(21)

As a summary, Table 4 collects the physical parameters of the mathematical model

that have been calculated or estimated by fitting to the experimental cloud points.

Table 4. Calculated physical parameters and fitting parameters use in the calculation procedure.

Calculated Parameters Fitted Parameters

χ13 Equation (9) α13 Equation (9)
χ23 Equation (9) α23 Equation (9)
g12 Equation (19) χ123 Equations (4)–(6)

3. Materials and Methods

3.1. Materials
The polymers used were polyvinylidene fluoride (PVDF) Kynar® 761A (Arkema
Inc., Colombes, France) with melt viscosity of 32 Kps@100s−1 and polyethersulfone (PES)
Sumikaexcel® 5200P (Sumitomo Chemical Europe Inc., Machelen, Belgium) with a reduced
viscosity of 0.52 (1 (w/v)% PES dissolved in DMF). The viscosity data of the polymers are
supplied by the companies. N,N-Dimethylacetamide (DMAc) and N-methyl-2-pyrrolidone
(NMP) (Acros Organics) were used as solvents and ultrapure water was used as the
non-solvent.

3.2. Cloud Point

The binodal curve is experimentally obtained by cloud point titration. Cloud point
experiments were performed for both solvents (DMAc and NMP) using water as the
non-solvent and PVDF and PES as the polymers. First, polymers were dried in an oven
overnight at 333 K. Then, polymer–solvent binary mixtures were prepared at several
polymer concentrations. Polymeric solutions, of an initial volume of 60 mL, were kept
under constant magnetic stirring. Drops of water were added using a digital burette
(Tittrete® , Brand GMBH + CO KG, Wertheim, Germany) to a polymer solution until
permanent turbidity was obtained for 1 h. Experiments were performed in duplicate at
different polymer concentrations in the range between 2 and 15 wt% using DMAc and
NMP as solvents. Higher polymer concentrations were not tested as the high viscosity of
the solution hindered its adequate stirring. The influence of the temperature was studied
at 293, 313 and 333 K. Student’s t-test was used to statistically compare the significance of
the difference between the experimental cloud points of each ternary system at the three
temperatures.

4. Results and Discussion

4.1. Experimental Cloud Point
Cloud point results are represented in Figure 3 and Table A3 of Appendix A compiles
the values. Although the experimental points of the four systems studied appeared very
proximate at 293 K the position of the cloud points were significantly different between
solvents (i.e., PVDF/DMAc/water vs. PVDF/NMP/water and PES/DMAc/water vs.
PES/NMP/water) and between polymers (i.e., PVDF/DMAc/water vs. PES/DMAc/water
 4. Results and Discussion
4.1. Experimental Cloud Point
Cloud point results are represented in Figure 3 and Table A3 of Appendix B compiles
the
Polymers 2021, 13, 678 values. Although the experimental points of the four systems studied appeared very 8 of 19
proximate at 293 K the position of the cloud points were significantly different between
solvents (i.e., PVDF/DMAc/water vs. PVDF/NMP/water and PES/DMAc/water vs.
PES/NMP/water) andand between polymers
PVDF/NMP/water vs.(i.e., PVDF/DMAc/water
PES/NMP/water). Cloud vs.point
PES/DMAc/water
experimental data and obtained
PVDF/NMP/water vs. PES/NMP/water). Cloud point experimental data obtained
between 293 and 295 K (room temperature) are widely reported in the literature for the between
293 and 295 Kfour
(room temperature)
systems studied are widely reported
[7,14,28,33–46]. in the literature
Experimental resultsfor
ofthe
thisfour
studysystems
fall within the
studied [7,14,28,33–46]. Experimental
range of results results
previously of thisatstudy
reported roomfall within the range
temperature. However,of results pre-
it is remarkable the
viously reported
hugeat room temperature.
dispersion However,
of data reported foritPVDF
is remarkable
systems,the huge dispersion
especially of data
in the PVDF/NMP/water
reported for PVDF
systemsystems,
[41,44].especially in thebetween
A relationship PVDF/NMP/water
the differentsystem
polymer[41,44]. A relationship
molecular weight characteris-
tics of the
between the different polymers
polymer used inweight
molecular these studies with theofvariation
characteristics of theused
the polymers cloudinpoints
these reported
was
studies with the not found.
variation of theOn the points
cloud other hand,
reportedthiswas
could
notbe moreOn
found. rationally
the otherattributed
hand, thisto the dif-
could be moreficulty observed
rationally to determine
attributed the experimental
to the difficulty observed tochange in turbidity
determine for PVDF systems
the experimental
change in turbidity for PVDF systems compared to the clear change in PES systems, for points
compared to the clear change in PES systems, for which literature cloud which are more
consistent [35,36].
literature cloud points are more consistent [35,36].

Figure
Figure 3. Cloud 3. Cloud
points at 293,points at 333
313 and 293,K313 and
of (a) 333 K of (a) PVDF–DMAc–water,
PVDF–DMAc–water, (b) PVDF–NMP–water,
(b) PVDF–NMP–water, (c) PES–DMAc–water(c) and
PES–DMAc–water and (d) PES–NMP–water systems. Duplicated experimental results are repre-
(d) PES–NMP–water systems. Duplicated experimental results are represented by  and *. Significant statistical difference
sented by □ and *. Significant statistical difference with Student’s t-test (p < 0.05) was found in all
with Student’s t-test (p < 0.05) was found in all systems comparing points between 293 and 313 K and between 313 and
333 K. Significant difference was also found at 293 K between systems comparing the use of DMAc and NMP with the same
polymer and between PVDF and PES with the same solvent.

For PVDF and PES systems, a significant influence of the temperature that displaces
the cloud points to the right is observed. However, for PES systems, the displacement of
the cloud point curves with the temperature was smaller than for PVDF. PES is classified
as a glassy polymer with glass transition temperature (Tg = 489–505 K) above the typical
temperature used in phase inversion processes (298–333 K), while PVDF is in rubbery
 the cloud points to the right is observed. However, for PES systems, the displacement of
the cloud point curves with the temperature was smaller than for PVDF. PES is classified
as a glassy polymer with glass transition temperature (Tg = 489–505 K) above the typical
temperature used in phase inversion processes (298–333 K), while PVDF is in rubbery
Polymers 2021, 13, 678 9 of 19
solid state (Tg = 206–278 K) at this range of processing temperatures. As the cloud points
represent a pseudo stable liquid phase point in the proximities to solid precipitation, this
difference can be attributed to the different state of the polymers in the solid phase, being
solid state (Tgand
PVDF a rubbery-polymer = 206–278
PES aK) at this range of processing
glassy-polymer. temperatures.
The specific volume ofAsthe theglassy
cloud points
represent a pseudo stable liquid phase point in the proximities to solid precipitation,
polymers suffers a thermal expansion when temperature rises but there is a characteristic
this difference can be attributed to the different state of the polymers in the solid phase,
rigidity of polymer chains. In the rubbery state, there is a higher impact of the temperature
being PVDF a rubbery-polymer and PES a glassy-polymer. The specific volume of the glassy
on the increasepolymers
of the specific
suffers volume,
a thermalas chain mobility
expansion is more important
when temperature rises but than
there for
is a glassy
characteristic
polymers [2]. Hence, the penetration of the solvents and therefore the solubility might be
rigidity of polymer chains. In the rubbery state, there is a higher impact of the temperature
more relevant on forthe
polymers in the
increase of thespecific
rubbery state (PVDF)
volume, than forisglassy
as chain mobility polymersthan
more important (PES).
for glassy
polymers
These experimental [2]. Hence,
results indicatethethat,
penetration
from theof the solvents and therefore
thermodynamic point ofthe solubility
view, tuningmight be
the NIPS process temperature could be considered an interesting approach to tailor the (PES).
more relevant for polymers in the rubbery state (PVDF) than for glassy polymers
These experimental
membrane morphology of rubbery results indicate
polymers that,as
(such from the thermodynamic
PVDF). point not
However, it might of view,
be atuning
the NIPS process temperature could be considered an interesting approach to tailor the
reasonable choice for membranes prepared of glassy polymers such as PES.
membrane morphology of rubbery polymers (such as PVDF). However, it might not be a
reasonable choice for membranes prepared of glassy polymers such as PES.
4.2. Mathematical Modeling
4.2. Mathematical
The obtained Modeling at different temperatures are compiled in Table
solubility parameters
A4, Appendix B. ItThe canobtained
be seensolubility parameters
for all the componentsat different
that thetemperatures
solubilityare compiled with
decreased in Table A4,
the temperature. This reduction of the solubility parameters will affect the values of χwith
Appendix A. It can be seen for all the components that the solubility decreased ij the
temperature. This reduction of the solubility parameters will affect
used in the simulations, so it is confirmed the necessity of calculating them considering the values of χ ij used in
the simulations, so it is confirmed the necessity of calculating them considering the effect
the effect of temperature.
of temperature.
Figure 4 presents the
Figure g12 curves
4 presents theasg12a curves
functionas aoffunction
solventofconcentration (ϕ2) for(φthe
solvent concentration 2 ) for the
binary systemsbinary
(a) DMAc–water and (b) NMP–water
systems (a) DMAc–water and (b) NMP–water calculated using
calculated UNIFAC-Dort-
using UNIFAC-Dortmund
mund methodology. Table Table
methodology. 5 presents the the
5 presents parameters
parametersthatthatresults fromthe
results from thefitting
fitting of the
of the calculated
calculated g12 values to the
g12 values to thepolynomial
polynomialEquation
Equation(19).(19). ItIt can seenthat
can be seen thatthetheinteraction
interaction between
DMAc and water decreased with the solvent concentration,
between DMAc and water decreased with the solvent concentration, while the NMP–wa- while the NMP–water behaved
as the opposite. Additionally, an increase
ter behaved as the opposite. Additionally, an increase of the of the g values with the temperature
12 g12 values with the tempera- is observed
in both cases.
ture is observed in both cases.

Figure 4. Influence of solvent concentration in g12 binary interaction parameter at 293, 313 and 333 K for (a) DMAc–water
Figure 4. Influence of solvent concentration in g12 binary interaction parameter at 293, 313 and 333
and (b) NMP–water binary systems.
K for (a) DMAc–water and (b) NMP–water binary systems

Table 5. Coefficient of solvent/non-solvent interaction parameters for the equation g12 (φ2 ) = ao + a1 φ2 + a2 φ22
+ a3 φ2 + a4 φ2 + a5 φ25 + a6 φ26 .
3 4

System T (K) a0 a1 a2 a3 a4 a5 a6
293 0.2265 −0.0758 0.2781 −1.1099 1.4731 −0.815
DMAc/Water 313 0.3396 −0.0269 0.3184 −1.1455 1.5387 −0.8392
333 0.4223 0.0945 −0.358 1.5048 −3.2626 3.316 −1.3648
293 0.3442 0.0751 −0.0107 0.0549 −0.0411
NMP/Water 313 0.4617 0.1069 −0.0234 0.0871 −0.0693
333 0.5438 0.1373 −0.0365 0.1273 −0.1003
Polymers 2021, 13, x FOR PEER REVIEW 10 of 2

Table 5. Coefficient of solvent/non-solvent interaction parameters for the equation g12 (ϕ2 ) =
Polymers 2 3 4 5 6
ϕ +a13,
ao +a12021,
2
678
2 ϕ +a3 ϕ +a4 ϕ +a5 ϕ +a6 ϕ .
2 2 2 2 2
10 of 19

System T (K) a0 a1 a2 a3 a4 a5 a6
293 0.2265 −0.0758
When the value 0.2781 the interaction
of g12 increases, −1.1099between 1.4731 −0.815
the solvent and non-solvent
DMAc/Water 313 is0.3396 −0.0269
lower, favoring 0.3184
the displacement −1.1455
of the binodal 1.5387
curve to the −0.8392 axis,
polymer–non-solvent
333 and therefore extending
0.4223 0.0945 the region of homogeneous
−0.358 1.5048liquid phase [47]. In Figure
−3.2626 4, it can be
3.316 −1.3648
293 seen that,
0.3442 overall, NMP-water
0.0751 g presents
−0.0107
12 higher
0.0549values than for
−0.0411 DMAc–water systems
(worse interaction). Moreover, for the system DMAc–water, the interaction was favoured at
NMP/Water 313 0.4617 0.1069 −0.0234 0.0871 −0.0693
increasing concentrations of DMAc and at lower temperatures, while for the NMP–water
333 0.5438 the high0.1373
system, −0.0365 and high
NMP concentrations 0.1273 −0.1003
temperatures decreased the interaction
between the components. Accordingly, Figure 3 shows the expected displacement to the
When
right of the value
the cloud pointsof g12 the
with increases, the interaction
temperature, between
and the larger the solvent and
liquid homogeneous non-solven
region
for the NMP
is lower, systems.the displacement of the binodal curve to the polymer–non-solvent axis
favoring
and therefore extending the region of homogeneous liquid phase [47]. In Figure 4, it ca
4.2.1. Modeling Binodal Curves Not Considering the Contribution of the Ternary
be seen that,
Interaction overall, NMP-water g12 presents higher values than for DMAc–water system
Term
(worse interaction). Moreover, for the system DMAc–water, the interaction was favoure
In this section it is considered a negligible effect of the ternary interaction parameter
at increasing concentrations
on the phase diagram of modeling
(χ123 = 0). This DMAc and at lower
approach temperatures,
has been while
the most widely for the NMP
reported
water
in system, [17,18,27,48].
the literature the high NMP concentrations
Firstly, and high temperatures
χ13 and χ23 parameters were calculateddecreased
accordingthe inter
action between the components. Accordingly, Figure 3 shows the expected
to Equation (11) considering α13 = α23 = 1 (Table 6). It can be seen that all χ13 values displacemen
decreased
to the rightwith
oftemperature. On thewith
the cloud points otherthe
hand, χ23 values increased
temperature, and the with temperature
larger for
liquid homogeneou
all binary systems except
region for the NMP systems. for PVDF/NMP. Figure 5 shows the solubility parameters in the
Hansen space and the radius of interaction between each pair of compounds. Interestingly,
PVDF
4.2.1. was closer to
Modeling the solvents
Binodal than
Curves notPES, especially PVDF
Considering was more soluble
the Contribution of theinTernary
DMAc Interac
(Ra = 1.62). Besides, PES was closer to NMP than to DMAc (Ra = 4.06). The solvation could
tion Term
be easier for PVDF since its monomer size was smaller compared with PES, and on the
In this
other hand thesection
phenylitgroups
is considered a negligible
in PES polymer effect
contribute toof
thethe ternary
steric interaction paramete
hindrance.
on the phase diagram (χ123 = 0). This modeling approach has been the most widely re
Table 6. Binary interaction parameters
ported in (χ13
theandliterature
χ23 ) at different temperatures
[17,18,27,48]. from solubility
Firstly, χ13 andparameters with α13 =were
χ23 parameters α13 = 1calculated ac
and χ123 = 0. cording to Equation (11) considering α13 = α23 = 1 (Table 6). It can be seen that all χ13 value
Temperature (K) decreased 293with temperature. On the 313 other hand, χ23 values increased
333 with temperatur
Interaction for all binary systems except for PVDF/NMP. Figure 5 shows the solubility parameters i
χ13Hansen space
the χ23and the radius
χ13 of interaction
χ23between each
χ13 pair of compounds.
χ23 Interest
Parameter
PVDF/DMAc/Water ingly,
2.106 PVDF was closer to the
0.023 solvents than
1.879 PES, especially
0.031 1.684 PVDF was more soluble i
0.043
PVDF/NMP/Water DMAc
2.106 (Ra = 1.62). Besides, PES
0.069 was closer to0.053
1.879 NMP than to 1.684
DMAc (Ra = 4.06).
0.042 The solvatio
PES/DMAc/Water 2.237 be easier 0.299
could for PVDF since 2.003
its monomer0.337 1.802compared0.380
size was smaller with PES, and o
PES/NMP/Water 2.237 0.157 2.003 0.174 1.802 0.195
the other hand the phenyl groups in PES polymer contribute to the steric hindrance.

2
2
Figure5.5. Hansen
Figure Hansen space and radius radius of
of interaction
interaction (Ra) according
according to Ra))22 =
to ((Ra = 4(δ d,2-−
4 δd,2 δd,1) +(δ
δd,1 +p,2 -
2δ

2 2 −δ

2
δ − +
δ ) +(δ -δ ) [23].
p,2 p,1 δh,2 h,1 [23].
p,1 h,2 h,1
 Table 6. Binary interaction parameters (χ13 and χ23) at different temperatures from solubility pa-
rameters with α13 = α13 = 1 and χ123 = 0.
Polymers 2021, 13, 678 11 of 19
Temperature (K) 293 313 333
Interaction
χ13 χ23 χ13 χ23 χ13 χ23
Parameter
Figure A1 of Appendix A depicts the modeled binodal curves of ternary systems
PVDF/DMAc/Water
obtained 2.106
under the above assumptions 0.023
(i.e., neglecting1.879
the ternary0.031 1.684
interaction parameter). 0.043
PVDF/NMP/Water
Overall, the model correctly 2.106 0.069
predicts the displacement 1.879 0.053
of the binodal 1.684
curve to the right at 0.042
increasing temperature,
PES/DMAc/Water as it happens
2.237 with the
0.299 experimental
2.003 data. However,
0.337 the simulated
1.802 0.380
curves did not fit adequately2.237
PES/NMP/Water of experimental
0.157points. It2.003
is observed0.174
that all binodal curves
1.802 0.195
were shifted to the left (solvent–polymer axis) in contrast to the experimental cloud points.
Different aspects, such as the polymer having a broad molecular weight distribution,
Figure
or swelling A1 of Appendix
or plasticizing B depictsbetween
effects occurring the modeled binodal
the polymer andcurves of ternary
the solvent systems o
for binary
tained under the above assumptions (i.e., neglecting the ternary interaction
mixtures have been reported to alter significantly the values of the Flory–Huggins binary parameter
Overall, the
interaction model correctly
parameters predicts
[48–50]. Previous the displacement
works of the
have reported the use binodal curvefactor
of a correction to the right
αijincreasing temperature,
in the calculation of binaryasinteraction
it happensparameters
with the experimental data. However,
χ13 and χ23 , Equation the simulate
(11), to amend
the deviation
curves observed
did not between the
fit adequately modeled and experimental
of experimental cloud points.
points. It is observed thatTherefore,
all binodal curve
a were
fittingshifted
procedure to left
to the estimate the correction factors
(solvent–polymer axis) inwas used obtaining
contrast the χ13 and cloud
to the experimental χ23 point
values Different
presented aspects,
in Table 7. Simulated and experimental curves using this approach
such as the polymer having a broad molecular weight distributio are
depicted in Figure 6. Due to the fitting approach, the simulated curves are now fairly
or swelling or plasticizing effects occurring between the polymer and the solvent for b
predicting the experimental points, as expected.
nary mixtures have been reported to alter significantly the values of the Flory–Huggin
Table 7. Corrected binary interaction
binary interaction parameters
parameters χij with[48–50]. Previouscorrection
the corresponding works have reported
constants αij . the use of a corre
tion factor αij in the calculation of binary interaction parameters χ13 and χ23, Equation (11
293 K
to amend the deviation observed 313 K
between 333 K
the modeled and experimental cloud point
α13 Therefore,
χ13 α23a fitting
χ23procedure
α13 to estimate the correction factors was used obtaining
χ 13 α 23 χ 23 α 13 χ 13 α 23 χ 23 the χ
PVDF/DMAc/Water 0.7 and 1.474 1
χ23 values 0.0234
presented 0.75in Table
1.409 7. Simulated
1 0.0307 and0.83 1.397
experimental 1
curves 0.0429
using this ap
PVDF/NMP/Water 0.8 1.685
proach are1 depicted
0.069 0.87
in Figure 1.635
6. Due to the1 0.053
fitting 0.93
approach, 1.566
the simulated 1 0.042
curves are no
PES/DMAc/Water 0.64 1.432 0.1 0.030 0.75 1.502 0.1 0.034 0.85 1.531 0.2 0.076
PES/NMP/Water 0.64 fairly
1.432 predicting
0.5 the
0.078 experimental
0.75 1.502 points,
0.5 as expected.
0.087 0.85 1.531 0.5 0.098

Figure 6. Cloud points and corrected theoretical binodal curves (BC) at 293, 313 and 333 K of (a)
Figure 6. Cloud points and corrected theoretical binodal curves (BC) at 293, 313 and 333 K of (a)
PVDF–DMAc–water, (b) PVDF–NMP–water, (c) PES–DMAc–water and (d) PES–NMP–water systems.
PVDF–DMAc–water, (b) PVDF–NMP–water, (c) PES–DMAc–water and (d) PES–NMP–water sys
Duplicated experimental results are represented by  and *.
tems. Duplicated experimental results are represented by □ and*.
Polymers 2021, 13, 678 12 of 19

Table 7 shows that for PVDF systems the α13 correction factors were near to 1, and χ23
calculation did not require of a correction factor (α23 = 1). For χ13 a linear tendency with
the temperature is observed with a slope of 0.0033 for both PVDF/DMAc and PVDF/NMP
systems. Instead, PES systems needed significant correction in both χ13 and χ23 binary
interaction parameters. The value of α13 is the same for both PES systems and it increased
with temperature. However, the solvent–polymer interaction, represented by χ23 , needs a
higher correction with α23 values between 0.1 and 0.5.
Other authors propose different αij values. Lindving et al. [51] consider a general value
of 0.6, independent of the system. Wei et al. [27] estimated fitting values of α13 = 0.83 and
α23 = 0.08 for both PES/NMP/water and PES/DMAc/water systems at 298 K; these values
differ from the results obtained in this study because of the revised expression of g12 and
the solubility parameters for PES employed in the present work. Although the model
adequately predicts the experimental points, the use of correction factors entails important
limitations and concerns: (1) low polymer concentration points are not well adjusted to
the binodal curve; (2) two fitting parameters are needed (α13 and α23 ) and (3) no reason-
able relation is found in the adjustment of the alpha correction factors, thus limiting its
extrapolation to other systems or finding a tendency. In this study, the necessity of a α23
correction factor for PES systems may be attributed to a higher cosolvency effect between
the three components, this phenomenon will be studied in detail in the following section.
According to Barth et al. [52], who studied PES/DMF/water systems, the use of α13 in the
calculation of χ13 points to the occurrence of ternary interactions between polymer, solvent
and non-solvent. This approach, that is described below, has not been explored so far for
the present ternary systems.

4.2.2. Modeling Binodal Curves Considering the Contribution of the Ternary

Interaction Term
To investigate the influence of the ternary interaction term of the systems, the χ123
ternary interaction parameter was included in the thermodynamic model. Given the lack
of any methodology reported in the literature to calculate the ternary interaction parameter
(χ123 ), we decided to estimate it by adjusting the experimental cloud point data to the
model, Equations (4)–(6). Figure 7 presents the values of χ123 and the resulting simulated
binodal curves.
It can be seen that the fitted χ123 values were negative indicating a strong interaction
between the components. This ternary interaction was not considered in previous literature
approaches to the systems studied, where only binary interactions were deemed. Few au-
thors have taken into account this parameter for polycaprolactone/dimethylformamide/
water system [15] and poly(ethylene-co-vinyl alcohol) (EVAL)/2-propanol (2P)/water
system [53]. The negative sign of χ123 implies a cosolvency between the solvent and
water. Cosolvency is a phenomenon that has been broadly reported in ternary systems
where polymers were soluble on binary mixtures of two non-solvents EVAL/2P/water
and polystyrene/acetone/diethylether [53,54]. In the present study, it is hypothesized
that at certain range of solvent/non-solvent compositions, the solubility of the polymer
is higher than expected simply considering binary interaction between the components,
which can be attributed to a cosolvency effect (solvent/non-solvent mixtures acting as a
solvent). Higher absolute values of χ123 means higher cosolvency influence. PES presents
higher degree of cosolvency (higher χ123 estimated values) than PVDF systems, probably
due to its worse solubility on the solvents (Figure 5).
rs 2021, 13,Polymers 13, 678
2021, REVIEW
x FOR PEER 13 of 20 13 of 19

Figurecurves
Figure 7. Binodal 7. Binodal curves
using usingχ123
adjusted adjusted χ123 interaction
interaction parameterparameter (a) PVDF–DMAc–water,
(a) PVDF–DMAc–water, (b)
(b) PVDF–NMP–water,
PVDF–NMP–water, (c) PES–DMAc–water and (d) PES–NMP–water systems. Duplicated experi-
(c) PES–DMAc–water and (d) PES–NMP–water systems. Duplicated experimental results are represented by  and *.
mental results are represented by □ and *.
It is observed a linear trend of χ123 with temperature, in the range 293–333 K, as shown
It can beinseen
the that theequations
fitting fitted χ123inserted
values were negative
in Figure indicating
6. The absolutea strong interaction
χ123 value follows the order
between the components.
PVDF/NMP/water This ternary interaction was not
> PVDF/DMAc/water > considered
PES/NMP/waterin previous litera-
> PES/DMAc/water.
Theto
ture approaches increase of χ123
the systems with temperature
studied, indicates
where only binary that cosolvency
interactions remains,Few
were deemed. although this
authors haveeffect
takenis diminished
into account with temperature
this parameterasforindicated by the reduction in the absolute values
polycaprolactone/dimethylforma-
of χ123 .[15]
mide/water system Now thepoly(ethylene-co-vinyl
and curves adjust better to the experimental
alcohol) cloud points even
(EVAL)/2-propanol at low polymer
(2P)/wa-
concentrations.
ter system [53]. The negative sign of χ123 implies a cosolvency between the solvent and
water. CosolvencyThe is aresults obtainedthat
phenomenon in the
has present study highlight
been broadly reported inthat, despite
ternary being the most
systems
where polymerswidely
werespread
solublemethodology, neglecting
on binary mixtures thenon-solvents
of two ternary interaction among the
EVAL/2P/water andcomponents
polystyrene/acetone/diethylether [53,54]. In the present study, it is hypothesized that atpresenting
of the phase inversion process might be erroneous, particularly for polymers
certain rangelow solubility in traditional solvents. It is also envisaged that the ternary interaction
of solvent/non-solvent compositions, the solubility of the polymer is higher
term would have an important contribution in describing the equilibrium of systems that
than expected simply considering binary interaction between the components, which can
incorporate solvents with low solubility properties, i.e., typical green solvents such as
be attributed to a cosolvency effect (solvent/non-solvent mixtures acting as a solvent).
dimethylsulfoxide (DMSO) [55], PolarClean® [56] or 2Pyr [6].
Higher absolute values of χ123 means higher cosolvency influence. PES presents higher
degree of cosolvency (higher χ123 estimated values) than PVDF systems, probably due to
its worse solubility on the solvents (Figure 5).
Polymers 2021, 13, 678 14 of 19

5. Conclusions
The influence of temperature in the processing of polymeric membranes by NIPS has
not been sufficiently studied. In this work, four systems composed by PVDF and PES as
the polymers, DMAc and NMP as the solvents and water as the non-solvent were studied
at three temperatures (293, 313 and 333 K). The experimental cloud points of the systems
were obtained and used to validate a mathematical model based on the Flory–Huggins
theory that included the effect of the temperature.
The experimental cloud points showed that as the temperature increased the solubility
region of the ternary systems was enlarged. The displacement of the binodal curve toward
the polymer/non-solvent axis was more evident for the PVDF (rubbery polymer) than in
the PES (glassy polymer) systems.
For the first time, the thermodynamic model developed in the present work suc-
cessfully incorporated the effect of temperature to predict the binodal curves of four
polymer/solvent/water ternary systems. We found that the model considering only binary
interaction parameters critically deviated from experimental data. Therefore, we evaluated
the incorporation to the model of a ternary interaction term that includes a ternary interac-
tion parameter χ123 , which was estimated for each system by fitting the experimental data
to the proposed model. Negative values of χ123 indicated the strong interaction between
the three components and the presence of a cosolvency phenomenon that increased the
homogeneous region where the polymer is still soluble. In all systems, the magnitude of
the ternary interaction parameter decreased linearly at increasing temperatures.
The main disadvantage of the implementation of the ternary interaction term in the
model is that χ123 parameter must be estimated by fitting to experimental data. To expand
the universal application of this model, future works should aim at broadening the experi-
mental validation with other polymer systems and the search of methodologies to estimate
reliable χ123 values for different ternary systems.

Author Contributions: Conceptualization, N.D. and A.U.; methodology, M.R., N.D. and A.U.;
investigation, M.R.; writing—original draft preparation, M.R.; writing—review and editing, N.D. and
A.U.; supervision, N.D. and A.U.; funding acquisition, A.U. and N.D. All authors have read and
agreed to the published version of the manuscript.
Funding: This research was funded by the Spanish Ministry of Economy, Industry and Competi-
tiveness (MINECO-FEDER) through project CTM2016-75509-R and by the Spanish Research Agency
through the projects X-MEM (PCI2018-092929, project call APCIN 2018) and PID2019-105827RB-
I00/AEI/10.13039/501100011033.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Acknowledgments: Marta Romay is grateful to an FPI contract grant (BES-2017-081112). Arkema
Inc. and Sumitomo Chemical Europe Inc. for the supply of polymers.
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A
The UNIFAC modified (Dortmund) methodology is a way to obtain activity coefficient
of binary mixtures from group contribution data.
Firstly, contribution groups are assigned to the compounds involved in the mixture.
For a multitude of compounds this step can be easily done with the Dortmund data bank
tool of assignation http://www.ddbst.com/unifacga.html (accessed on 7 September 2020).
Secondly, volume and surface area data of each group and the binary interaction be-
tween all the groups involved in a mixture are compiled from http://www.ddbst.com/
PublishedParametersUNIFACDO.html (accessed on 7 September 2020).
Polymers 2021, 13, 678 15 of 19

Then the modified Dortmund UNIFAC methodology is applied. The activity coeffi-
cient γ of each compound is calculated using Equation (A1), this equation is composed by
the combinatorial part Equation (A2) and the residual part Equation (A3).

ln γi = ln γic + ln γiR   (A1)

Vi
Combinatorial ln γic = 1 − Vi0 + ln Vi0 − 5 qi 1 − Fi + ln VFii (A2)
h i
(i ) (i )
Residual ln γiR = ∑ vk ln ΓK − ln ΓK (A3)
(i )
Γk : residual activity coefficient of group k in a reference solution k
containing molecules of type i    
Θm Ψkm
Γk : group residual activity coefficient ln Γk = Qk 1 − ln ∑(Θm Ψmk ) − ∑ ∑n Θn Ψnm
(A4)
m m
q
Surface area/mole fraction ratio Fi = ∑ qij x j (A5)
j
Volume/mole fraction ratio Vi = ∑ rrij x j (A6)
j
ri3/4
Vi0 = 3/4 (A7)
∑j rj xj
(i )
Volume of the pure component (i) ri = ∑ v k R k (A8)
k
(i )
Surface area of the pure component (i) qi = ∑ v k Q k (A9)
k
Area fraction of group (m) Θm = ∑QmQXn X m
n
(A10)
n
(i )
∑j m xj
v
Mole fraction of group m in the mixture Xm = (i ) (A11)
∑ j ∑n vn x j 
2
Group interaction parameter Ψmn = exp − amn +bmnTT +cmn T (A12)

Polymers 2021, 13, x FOR PEER REVIEW Table A1 compiles the information of the group assignment and volume
16 of (R
20k ) and area
numbers (Qk ) for the compounds used in this study. Table A2 presents the interaction
parameters between the groups listed in Table A1 [31,57].
TablePolymers
A1. Group assignment,
2021, 13, volume and area numbers of different compounds based on UNIFAC-Dortmund
x FOR PEER REVIEW methodology.
16 of 20
Table A1. Group assignment, volume and area numbers of different compounds based on UNIFAC-Dortmund methodology.
GROUP Assignment
Compound Formula
Table A1. Group Structure
assignment, volume and area numbers of different compoundsGROUP Group-Sub
based on UNIFAC-Dortmund
Assignment methodology. Qk Rk
Group ×Vik
Compound Formula Structure GROUP Assignment Group No. Qk Rk
Group-Sub
Compound Formula Structure Group × VGroup-Sub 1-1Qk No. Rk1.0608
Group CH×Vik3 × 1 ik
Group 0.6325
Group No.
DMAc C4H9NO CH3 × 1 1-1 1.0608 0.6325
CON(CH
CH33)2×× 11 48-1011-1 1.9643 1.06082.4748 0.6325
DMAc DMAcC4 H9 NOC4H9NO
CON(CH 3)2 × 1 48-101 1.9643
48-101 2.4748
CON(CH 3 )2 × 1 1.9643 2.4748
cy-CH2 × 3 42-78 0.8635 0.7136
cy-CH2 × 3 42-78 0.8635 0.7136
NMP C5H9NO
NMPC5 H9 NOC5H9NO
cy-CH2 × 3 42-78 0.8635 0.7136
NMP cy-CON-CH
cy-CON-CH3 ×1
3 × 1 46-86 46-86
3.2 3.9813.2 3.981
cy-CON-CH3 × 1 46-86 3.2 3.981
Water H2O H2O × 1 7
Water
Water H2O
H2 O H2H O2×O1× 1 72.45617 1.7334
2.4561 2.45611.7334 1.7334
Table A2. Interaction parameters between UNIFAC groups.
Table A2. Interaction
n,m
parameters
anm,1
between
anm,2
UNIFAC
anm,3
groups.
amn,1 amn,2 amn,3
n,m anm,1 anm,2 anm,3 amn,1
Table A2. Interaction 1,7 parameters between UNIFAC groups.
n,m 1391.3anm,1−3.6156 anm,2 0.001144 −17.253
anm,3 0.8389 0.000902
amn,1 amn,2 amn,3
Water–DMAc 1,48 1529.52 −6.2025 0.00975 82.6 −0.615 −0.00062
1,7 n,m
7,48 1391.3
64.43999 anm,1
−0.0094−3.6156 - anm,20.001144
−439.58 −17.253
anm,3
0.3142 - 0.8389
amn,1 0.000902
amn,2 amn,3
Water–DMAc 7,42 1,48 274.37 1529.52 −0.5861−6.2025
−0.0003 0.00975
1632.9 82.6 0.003455
−2.8719 −0.615 −0.00062
Water–NMP 7,46 1,7 1391.3 −
- 3.6156 0.001144 −17.253 0.8389 0.000902
7,48 659.22 64.43999 −1.8841−0.0094 −588.21
- 0.9707
−439.58 - 0.3142 -
Water–DMAc42,46 1,48 298.46 1529.52
−0.6823 −
- 6.2025 499.59 0.00975
−0.8158 - 82.6 −0.615 −0.00062
7,42 7,48 274.37 64.43999 −0.5861 −0.0094 −0.0003 -1632.9 −439.58 −2.8719 0.003455
0.3142 -
Water–NMP
Table A3. Average 7,46
values of cloud point data at the 659.22 −1.8841
three temperatures - systems.
for the four studied −588.21 0.9707 -
7,42 274.37 −0.5861 −0.0003 1632.9 −2.8719 0.003455
ΦPVDF 42,46 298.46
ΦDMAc −0.6823 - 499.59
ΦWater −0.8158 -
Water–NMP 7,46 659.22 − 1.8841 - − 588.21 0.9707 -
0.0183 ± 2.00 × 105 0.8837 ± 5.90 × 104 0.0988 ± 1.95 × 104
42,46 298.46 4 −0.6823 - 499.59 −0.8158 -
Table A3.0.0286
Average ±values1.00
of cloud
× 10 point0.8769
5 data at the
± three temperatures
2.15 × 10 for the four
0.0946 ± studied
2.25 × systems.
10 4

293 K 0.0482 ± 5.00 × 106 0.8673 ± 6.00 × 105 0.0845 ± 5.50 × 105
0.0677ΦPVDF
± 2.00 × 105 0.8516 ± ΦDMAc
8.50 × 105 0.0808 ± 1.05 × 10Φ4 Water

0.0183 0.1103 ± ± 2.00 9.00××10

1055 0.8837 ± ± 8.50 × 105.90 × 10
0.8084 5 4
0.0813 ±0.09881.75 × 10±4 1.95 × 104
0.0286 0.0184 ± ± 1.00 1.05××10
1055
0.8769 ± ± 1.25 × 102.15 × 10
0.8626 4 0.1190
4 ±0.09462.30 × 10±4
2.25 × 104
0.0282 ± 4.50 × 1065 0.8645 ± 4.30 × 104 0.1073 ± 4.80 × 10 4
293 K 0.0482 ± 5.00 × 10 5 0.8673 ± 6.00 × 105 0.0845 ± 5.50 × 105
313 K 0.0473 ± 6.50× 10 0.8512 ± 1.17 × 103 0.1016 ± 1.24 × 103
0.0677 0.0656 ± ± 2.00 3.45××10 0.8516 ± ± 4.55 × 108.50 × 10 ±0.08084.89 × 10± 1.05 × 104
5 5
104 0.8247 3 0.1098 3

0.1103 0.0926 ± ± 9.00 2.60××10

1054 0.8084 ± ± 6.80 × 108.50
0.7904 4
× 10
0.1169
5 ±0.08134.20 × 10±4
1.75 × 104
0.0180 ± 4.00 × 1055 0.8297 ± 1.93× 103 0.1523 ± 1.97 × 103
0.0184 ± 1.05 × 10 0.8626 ± 1.25 × 10 4 0.1190 ± 2.30 × 104
0.0449 ± 0.00 × 100 0.8172 ± 0.00 × 100 0.1379 ± 0.00 × 100
333 K
0.0282 ± 4.50 × 1050 0.8645 ± 4.30
0 × 104 0.1073 ±0 4.80 × 104
Polymers 2021, 13, 678 16 of 19

Table A3. Average values of cloud point data at the three temperatures for the four studied systems.

ΦPVDF ΦDMAc ΦWater

0.0183 ± 2.00 × 105 0.8837 ± 5.90 × 104 0.0988 ± 1.95 × 104
0.0286 ± 1.00 × 105 0.8769 ± 2.15 × 104 0.0946 ± 2.25 × 104
293 K 0.0482 ± 5.00 × 106 0.8673 ± 6.00 × 105 0.0845 ± 5.50 × 105
0.0677 ± 2.00 × 105 0.8516 ± 8.50 × 105 0.0808 ± 1.05 × 104
0.1103 ± 9.00 × 105 0.8084 ± 8.50 × 105 0.0813 ± 1.75 × 104
0.0184 ± 1.05 × 105 0.8626 ± 1.25 × 104 0.1190 ± 2.30 × 104
0.0282 ± 4.50 × 105 0.8645 ± 4.30 × 104 0.1073 ± 4.80 × 104
313 K 0.0473 ± 6.50× 105 0.8512 ± 1.17 × 103 0.1016 ± 1.24 × 103
0.0656 ± 3.45 × 104 0.8247 ± 4.55 × 103 0.1098 ± 4.89 × 103
0.0926 ± 2.60 × 104 0.7904 ± 6.80 × 104 0.1169 ± 4.20 × 104
0.0180 ± 4.00 × 105 0.8297 ± 1.93× 103 0.1523 ± 1.97 × 103
0.0449 ± 0.00 × 100 0.8172 ± 0.00 × 100 0.1379 ± 0.00 × 100
333 K
0.0630 ± 0.00 × 100 0.8017 ± 0.00 × 100 0.1354 ± 0.00 × 100
0.0915 ± 0.00 × 100 0.7756 ± 0.00 × 100 0.1329 ± 0.00 × 100
ΦPVDF ΦNMP ΦWater
0.0192 ± 1.00 × 105 0.8917 ± 2.45 × 104 0.0892 ± 2.55 × 104
0.0491 ± 5.00 × 106 0.8771 ± 1.25 × 104 0.0738 ± 1.35 × 104
293 K
0.0680 ± 3.00 × 105 0.8565 ± 1.80 × 104 0.0755 ± 2.05 × 104
0.0929 ± 2.05 × 104 0.8392 ± 2.90 × 104 0.0679 ± 4.95 × 104
0.0189 ± 3.00 × 105 0.8767 ± 1.73 × 103 0.1044 ± 1.77 × 103
0.0480 ± 4.00 × 105 0.8576 ± 7.30 × 104 0.0944 ± 7.75 × 104
313 K
0.0655 ± 7.00 × 105 0.8245 ± 7.30 × 104 0.1101 ± 8.05 × 104
0.0890 ± 2.10 × 104 0.8039 ± 4.15 × 104 0.1071 ± 6.20 × 104
0.0180 ± 2.00 × 105 0.8378 ± 1.07 × 103 0.1442 ± 1.09 × 103
0.0457 ± 1.40 × 104 0.8161 ± 2.44 × 103 0.1382 ± 2.58 × 103
333 K
0.0633 ± 8.00 × 105 0.7974 ± 8.55 × 104 0.1393 ± 9.30 × 104
0.0865 ± 1.65 × 104 0.7816 ± 4.50 × 105 0.1318 ± 2.10 × 104
ΦPES ΦDMAc ΦWater
0.0179 ± 2.00 × 105 0.8798 ± 6.50 × 105 0.1023 ± 9.00 × 105
0.0450 ± 7.00 × 105 0.8566 ± 4.90 × 104 0.0984 ± 5.60 × 104
293 K
0.0912 ± 3.65 × 104 0.8188 ± 7.50 × 104 0.0900 ± 1.11 × 103
0.1375 ± 0.00 × 100 0.7787 ± 0.00 × 100 0.0838 ± 0.00 × 100
0.0178 ± 3.50 × 105 0.8759 ± 6.45 × 104 0.1063 ± 6.80 × 104
0.0448 ± 8.00 × 105 0.8533 ± 7.85 × 104 0.1018 ± 8.70 × 104
313 K
0.0908 ± 3.80 × 104 0.8156 ± 9.20 × 104 0.0936 ± 1.30 × 103
0.0684 ± 0.00 × 100 0.7751 ± 0.00 × 100 0.0881 ± 0.00 × 100
0.0177 ± 2.50 × 105 0.8703 ± 3.05 × 104 0.1120 ± 3.30 × 104
333 K 0.0446 ± 1.05 × 104 0.8483 ± 1.14 × 103 0.1071 ± 1.24 × 103
0.0901 ± 3.55 × 104 0.8086 ± 7.50 × 10 0.1013 ± 1.11 × 103
ΦPES ΦNMP ΦWater
0.0174 ± 4.00 × 105 0.8511 ± 9.20 × 104 0.1315 ± 9.60 × 104
293 K 0.0429 ± 7.55 × 104 0.8341 ± 3.00 × 104 0.1230 ± 1.06 × 103
0.0890 ± 5.25 × 104 0.8025 ± 3.68 × 103 0.1135 ± 8.25 × 104
0.0173 ± 4.00 × 105 0.8468 ± 8.75 × 104 0.1359 ± 9.10 × 104
313 K 0.0426 ± 7.75 × 104 0.8286 ± 6.90 × 104 0.1287 ± 1.46 × 103
0.0881 ± 0.00 × 100 0.7926 ± 0.00 × 100 0.1194 ± 0.00 × 100
0.0171 ± 5.00 × 105 0.8382 ± 1.46 × 103 0.1447 ± 1.50 × 103
333 K 0.0419 ± 6.70 × 104 0.8138 ± 1.05 × 103 0.1444 ± 3.80 × 104
0.0863 ± 0.00 × 100 0.7792 ± 0.00 × 100 0.1345 ± 0.00 × 100
Polymers 2021, 13, 678 17 of 19

Table A4. Solubility parameters calculated at different temperatures from Hansen solubility parame-
ters at 298 K.

T (K) δd (MPa1/2 ) δp (MPa1/2 ) δh (MPa1/2 ) δT (MPa1/2 )

Polymer–PVDF
298 17.2 12.5 9.2 23.167
293 17.234 12.510 9.264 23.223
313 17.099 12.471 9.012 23.003
333 16.967 12.432 8.767 22.788
Polymer–PES
298 19.6 10.8 9.2 24.196
293 19.632 10.807 9.262 24.249
313 19.503 10.779 9.015 24.038
333 19.376 10.750 8.775 23.833
Solvent–DMAc
298 16.8 11.5 10.2 22.771
Polymers 2021, 13, x FOR PEER REVIEW 293 16.898 11.527 10.286 18 of 20 22.896
313 16.506 11.419 9.945 22.399
333 16.106 11.308 9.610 21.900
Solvent–NMP
Solvent–NMP
298
298 18 18 12.3 12.3 7.2 7.222.959 22.959
293
293 18.086
18.086 12.324 12.324 7.258 7.258
23.058 23.058
313
313 17.735
17.735 12.227 12.227 7.028 7.028
22.659 22.659
333
333 17.378
17.378 12.128 12.128 6.803 6.803
22.257 22.257
Non-solvent–Water
Non-solvent–Water
298
298 15.5 15.5 16 16 42.3 47.807
42.3 47.807
293
293 15.538
15.538 16.015 16.015 42.600 42.600
48.090 48.090
313
313 15.388
15.388 15.954 15.954 41.413 41.413
46.972 46.972
333
333 15.211
15.211 15.880 15.880 40.228 40.228
45.846 45.846

Figure Cloud
A1.Cloud
Figure A1. point
point and theoretical
and theoretical curves
curves at 293, 313atand
293,
333313
K ofand 333 K of (a) PVDF–DMAc–water,
(a) PVDF–DMAc–water, (b) (b) PVDF–
PVDF–NMP–water, (c) PES–DMAc–water and (d) PES–NMP–water systems (α13 = α13 = 1 and χ123 = 0).
NMP–water, (c) PES–DMAc–water and (d) PES–NMP–water systems (α13 = α13 = 1 and χ123 = 0).
References
1. Wang, D.M.; Lai, J.Y. Recent advances in preparation and morphology control of polymeric membranes formed by nonsolvent
induced phase separation. Curr. Opin. Chem. Eng. 2013, 2, 229–237.
2. Mulder, M. Basic Principles of Membrane Technology; Kluwer Academic Publishers: Enschede, The Netherlands, 1991; ISBN 0-
7923-0978-2.
Polymers 2021, 13, 678 18 of 19

References
1. Wang, D.M.; Lai, J.Y. Recent advances in preparation and morphology control of polymeric membranes formed by nonsolvent
induced phase separation. Curr. Opin. Chem. Eng. 2013, 2, 229–237. [CrossRef]
2. Mulder, M. Basic Principles of Membrane Technology; Kluwer Academic Publishers: Enschede, The Netherlands, 1991;
ISBN 0-7923-0978-2.
3. Prézélus, F.; Tiruta-Barna, L.; Guigui, C.; Remigy, J.C. A generic process modelling—LCA approach for UF membrane fabrication:
Application to cellulose acetate membranes. J. Memb. Sci. 2021, 618, 118594. [CrossRef]
4. Romay, M.; Diban, N.; Rivero, M.J.; Urtiaga, A.; Ortiz, I. Critical Issues and Guidelines to Improve the Performance of Photocat-
alytic Polymeric Membranes. Catalysts 2020, 10, 570. [CrossRef]
5. Baker, R.W. Membrane Technology and Applications, 3rd ed.; Wiley: West Sussex, UK, 2012; ISBN 9780470743720.
6. Kahrs, C.; Gühlstorf, T.; Schwellenbach, J. Influences of different preparation variables on polymeric membrane formation via
nonsolvent induced phase separation. J. Appl. Polym. Sci. 2020, 137, 48852. [CrossRef]
7. Sadeghi, A.; Nazem, H.; Rezakazemi, M.; Shirazian, S. Predictive construction of phase diagram of ternary solutions containing
polymer/solvent/nonsolvent using modified Flory-Huggins model. J. Mol. Liq. 2018, 263, 282–287. [CrossRef]
8. Keshavarz, L.; Khansary, M.A.; Shirazian, S. Phase diagram of ternary polymeric solutions containing nonsolvent/solvent/polymer:
Theoretical calculation and experimental validation. Polymer 2015, 73, 1–8. [CrossRef]
9. Ariono, D.; Aryanti, P.T.P.; Hakim, A.N.; Subagjo, S.; Wenten, I.G. Determination of thermodynamic properties of polysul-
fone/PEG membrane solutions based on Flory-Huggins model. AIP Conf. Proc. 2017, 1840, 090008.
10. van Krevelen, D.W.; te Nijenhuis, K. Properties of Polymers, 4th ed.; Elsevier: Amsterdam, The Netherlands, 2009;
ISBN 978-0-08-054819-7.
11. Just, S.; Sievert, F.; Thommes, M.; Breitkreutz, J. Improved group contribution parameter set for the application of solubility
parameters to melt extrusion. Eur. J. Pharm. Biopharm. 2013, 85, 1191–1199. [CrossRef] [PubMed]
12. Stefanis, E.; Panayiotou, C. Prediction of Hansen Solubility Parameters with a New Group-Contribution Method.
Int. J. Thermophys. 2008, 29, 568–585. [CrossRef]
13. Kahrs, C.; Metze, M.; Fricke, C.; Schwellenbach, J. Thermodynamic analysis of polymer solutions for the production of polymeric
membranes. J. Mol. Liq. 2019, 291. [CrossRef]
14. Xu, L.; Qiu, F. Simultaneous determination of three Flory–Huggins interaction parameters in polymer/solvent/nonsolvent
systems by viscosity and cloud point measurements. Polymer 2014, 55, 6795–6802. [CrossRef]
15. Nivedita, S.; Ahamed, D.; Joseph, S. Thermodynamic analysis of phase diagram of H2O-DMF-PCL system: Investigation on the
influence of inorganic additives TiO2/MMT. J. Mater. Sci. 2020, 55, 5431–5444. [CrossRef]
16. Cohen, C.; Tanny, G.B.; Prager, S. Diffusion-Controlled Formation of Porous Structures in Ternary Polymer Systems. J. Polym. Sci.
Polym. Phys. Ed. 1979, 17, 477–489. [CrossRef]
17. Altena, F.W.; Smolders, C.A. Calculation of Liquid-Liquid Phase Separation in a Ternary System of a Polymer in a Mixture of a
Solvent and a Nonsolvent. Org. Mol. Photophy. 1982, 15, 1558. [CrossRef]
18. Yilmaz, L.; McHugh, A.J. Analysis of nonsolvent–solvent–polymer phase diagrams and their relevance to membrane formation
modeling. J. Appl. Polym. Sci. 1986, 31, 997–1018. [CrossRef]
19. Tan, X.M.; Rodrigue, D. A review on porous polymeric membrane preparation. Part I: Production Techniques with Polysulfone
and Poly (Vinylidene Fluoride). Polymers 2019, 11, 1160. [CrossRef]
20. Wittig, R.; Lohmann, J.; Gmehling, J. Vapor-Liquid Equilibria by UNIFAC Group Contribution. 6. Revision and Extension.
Ind. Eng. Chem. Res. 2003, 42, 183–188. [CrossRef]
21. Kurada, K.V.; De, S. Polyaniline doped ultrafiltration membranes: Mechanism of membrane formation and pH response
characteristics. Polymer 2018, 153, 201–213. [CrossRef]
22. Kurada, K.V.; Mukherjee, M.; De, S. Permeability hysteresis of polypyrrole-polysulfone blend ultrafiltration membranes: Study of
phase separation thermodynamics and pH responsive membrane properties. Sep. Purif. Technol. 2019, 227, 115736. [CrossRef]
23. Hansen, C.M. Hansen solubility parameters. A User’s Handbook, 2nd ed.; CRC Press Taylor & Francis Group: Oxfordshire, UK, 2007;
Volume 118, ISBN 0-8493-7248-8.
24. Zhang, J.; Zhang, Y.; Zhao, J. Thermodynamic study of non-solvent/dimethyl sulfoxide/polyacrylonitrile ternary systems: Effects
of the non-solvent species. Polym. Bull. 2011, 67, 1073–1089. [CrossRef]
25. Tang, Y.H.; Ledieu, E.; Cervellere, M.R.; Millett, P.C.; Ford, D.M.; Qian, X. Formation of polyethersulfone membranes via
nonsolvent induced phase separation process from dissipative particle dynamics simulations. J. Memb. Sci. 2020, 599, 117826.
[CrossRef]
26. Kurada, K.V.; De, S. Modeling of solution thermodynamics: A method for tuning the properties of blend polymeric membranes.
J. Memb. Sci. 2017, 540, 485–495. [CrossRef]
27. Wei, Y.-M.; Xu, Z.-L.; Yang, X.-T.; Liu, H.-L. Mathematical calculation of binodal curves of a polymer/solvent/nonsolvent system
in the phase inversion process. Desalination 2006, 192, 91–104. [CrossRef]
28. Mohsenpour, S.; Safekordi, A.; Tavakolmoghadam, M.; Rekabdar, F.; Hemmati, M. Comparison of the membrane morphology
based on the phase diagram using PVP as an organic additive and TiO2 as an inorganic additive. Polymer 2016, 97, 559–568.
[CrossRef]
Polymers 2021, 13, 678 19 of 19

29. Hamta, A.; Ashtiani, F.Z.; Karimi, M.; Safikhani, A. Manipulating of polyacrylonitrile membrane porosity via SiO2 and TiO2
nanoparticles: Thermodynamic and experimental point of view. Polym. Adv. Technol. 2021, 32, 872–885. [CrossRef]
30. Sandler, S.I. Chemical and Engineering Thermodynamics, 3rd ed.; John Wiley and Sons Inc.: Hoboken, NJ, USA, 1940; ISBN
0-471-18210-9.
31. Published Parameters UNIFAC(Do)—DDBST GmbH. Available online: http://www.ddbst.com/PublishedParametersUNIFACDO.
html#ListOfMainGroups (accessed on 6 October 2020).
32. Tompa, H. Phase relationships in polymer solutions. Trans. Faraday Soc. 1949, 45, 1142–1152. [CrossRef]
33. Yeow, M.L.; Liu, Y.T.; Li, K. Isothermal Phase Diagrams and Phase-Inversion Behavior of Poly(vinylidene fluoride)/Solvents/
Additives/Water Systems. J. Appl. Polym. Sci. 2003, 90, 2150–2155. [CrossRef]
34. Fadaei, A.; Salimi, A.; Mirzataheri, M. Structural elucidation of morphology and performance of the PVDF/PEG membrane.
J. Polym. Res. 2014, 21, 1–8. [CrossRef]
35. Barzin, J.; Sadatnia, B. Theoretical phase diagram calculation and membrane morphology evaluation for water/solvent/
polyethersulfone systems. Polymer 2007, 48, 1620–1631. [CrossRef]
36. Lau, W.W.Y.; Guiver, M.D.; Matsuura, T. Phase separation in polysulfone/solvent/water and polyethersulfone/solvent/water
systems. J. Memb. Sci. 1991, 59, 219–227. [CrossRef]
37. Zeman, L.; Tkacik, G. Thermodynamic analysis of a membrane-forming system water/N-methyl-2-pyrrolidone/polyethersulfone.
J. Memb. Sci. 1988, 36, 119–140. [CrossRef]
38. Sadrzadeh, M.; Bhattacharjee, S. Rational design of phase inversion membranes by tailoring thermodynamics and kinetics of
casting solution using polymer additives. J. Memb. Sci. 2013, 441, 31–44. [CrossRef]
39. Wang, X.; Zhang, L.; Sun, D.; An, Q.; Chen, H. Formation mechanism and crystallization of poly(vinylidene fluoride) membrane
via immersion precipitation method. Desalination 2009, 236, 170–178. [CrossRef]
40. Sun, A.C.; Kosar, W.; Zhang, Y.; Feng, X. A study of thermodynamics and kinetics pertinent to formation of PVDF membranes by
phase inversion. Desalination 2013, 309, 156–164. [CrossRef]
41. Fashandi, H.; Yegane, A.; Abolhasani, M.M. Interplay of Liquid-liquid and Solid-liquid Phase Separation Mechanisms in Porosity
and Polymorphism Evolution Within Poly(vinylidene fluoride) Nanofibers. Fibers Polym. 2015, 16, 326–344. [CrossRef]
42. Sukitpaneenit, P.; Chung, T.S. Molecular elucidation of morphology and mechanical properties of PVDF hollow fiber membranes
from aspects of phase inversion, crystallization and rheology. J. Memb. Sci. 2009, 340, 192–205. [CrossRef]
43. Bottino, A.; Camera-Roda, G.; Capannelli, G.; Munari, S. The formation of microporous polyvinylidene difluoride membranes by
phase separation. J. Memb. Sci. 1991, 57, 1–20. [CrossRef]
44. Enayatzadeh, M.; Mohammadi, T. Morphology and performance of poly(vinylidene fluoride) flat sheet membranes: Thermody-
namic and kinetic aspects. J. Appl. Polym. Sci. 2018, 135, 46419. [CrossRef]
45. Zuo, D.; Li, H. Membrane formation mechanism for water-solvent-pvdf systems and membrane structure for ultrofiltration
application. Adv. Mater. Res. 2012, 578, 129–132. [CrossRef]
46. Mansourizadeh, A.; Mousavian, S. Structurally developed microporous polyvinylidene fluoride hollow-fiber membranes for
CO2 absorption with diethanolamine solution. J. Polym. Res. 2013, 20, 1–12. [CrossRef]
47. Mohsenpour, S.; Esmaeilzadeh, F.; Safekordi, A.; Tavakolmoghadam, M.; Rekabdar, F.; Hemmati, M. The role of thermodynamic
parameter on membrane morphology based on phase diagram. J. Mol. Liq. 2016, 224, 776–785. [CrossRef]
48. Chen, Q.P.; Xie, S.; Foudazi, R.; Lodge, T.P.; Siepmann, J.I. Understanding the Molecular Weight Dependence of χ and the Effect
of Dispersity on Polymer Blend Phase Diagrams. Macromolecules 2018, 51, 3774–3787. [CrossRef]
49. Lu, H.; Du, S. A phenomenological thermodynamic model for the chemo-responsive shape memory effect in polymers based on
Flory-Huggins solution theory. Polym. Chem. 2014, 5, 1155–1162. [CrossRef]
50. Lu, H.; Liu, Y.; Leng, J.; Du, S. Qualitative separation of the physical swelling effect on the recovery behavior of shape memory
polymer. Eur. Polym. J. 2010, 46, 1908–1914. [CrossRef]
51. Lindvig, T.; Michelsen, M.L.; Kontogeorgis, G.M. A Flory-Huggins model based on the Hansen solubility parameters.
Fluid Phase Equilib. 2002, 203, 247–260. [CrossRef]
52. Barth, C.; Wolf, B.A. Quick and reliable routes to phase diagrams for polyethersulfone and polysulfone membrane formation.
Macromol. Chem. Phys. 2000, 201, 365–374. [CrossRef]
53. Young, T.H.; Cheng, L.P.; Hsieh, C.C.; Chen, L.W. Phase behavior of EVAL polymers in water-2-propanol cosolvent. Macromolecules
1998, 31, 1229–1235. [CrossRef]
54. Wolf, B.A.; Molinari, R.J. True cosolvency. Acetone/diethylether/polystyrene. Die Makromol. Chemie 1973, 173, 241–245. [CrossRef]
55. Marino, T.; Galiano, F.; Simone, S.; Figoli, A. DMSO EVOLTM as novel non-toxic solvent for polyethersulfone membrane
preparation. Environ. Sci. Pollut. Res. 2019, 26, 14774–14785. [CrossRef]
56. Xie, W.; Tiraferri, A.; Liu, B.; Tang, P.; Wang, F.; Chen, S.; Figoli, A.; Chu, L.Y. First Exploration on a Poly(vinyl chloride)
Ultrafiltration Membrane Prepared by Using the Sustainable Green Solvent PolarClean. ACS Sustain. Chem. Eng. 2020, 8, 91–101.
[CrossRef]
57. DDBST—DDBST GmbH. Available online: http://www.ddbst.com/ (accessed on 6 October 2020).