Fluid Phase Equilibria 231 (2005) 77–83

Extension of the NRTL and NRF models to multicomponent polymer

solutions: Applications to polymer–polymer aqueous two-phase systems
Rahmat Sadeghi a,∗ , Mohammed Taghi Zafarani-Moattar b
a Department of Chemistry, Kurdistan University, Sanandaj 66135, Iran
b Physical Chemistry Department, University of Tabriz, Tabriz, Iran

Received 23 June 2004; received in revised form 8 November 2004; accepted 11 January 2005

Abstract

The modified NRTL (nonrandom two-liquid) model for multisolvent, single polymer solutions proposed by Chen [C.C. Chen, Fluid
Phase Equilib. 83 (1993) 301–312] and the modified NRF (nonrandom factor) model for single solvent, single polymer solutions, proposed
by Zafarani-Moattar and Sadeghi [M.T. Zafarani-Moattar, R. Sadeghi, Fluid Phase Equilib. 202 (2002) 413–422] have been used for the
representation of the excess Gibbs energy of multicomponent polymer solutions. The models represent a synergistic combination of the
Flory–Huggins description for the configurational entropy of mixing molecules of different sizes and the NRTL or NRF theory for the local
composition contribution from mixing solvents and segments of polymers. These models have been applied for the correlation of the phase
behavior of some polymer–polymer aqueous two-phase systems. The results show that the models can accurately correlate a series of the
liquid–liquid equilibrium phase diagrams of aqueous two-phase systems at different polymer molecular weights.
© 2005 Elsevier B.V. All rights reserved.

Keywords: Excess Gibbs energy; NRTL; NRF; Aqueous two-phase systems; Liquid–liquid equilibrium; Polymer solutions

1. Introduction These models have been extended and applied to aqueous

Aqueous solutions of two incompatible polymers such
as polyethylene glycol (PEG) and dextran (DEX) form
liquid–liquid two-phase systems, in which each aqueous
phase is rich with respect to one of the phase-forming
polymers. These polymer–polymer aqueous two-phase sys-
tems provide a powerful method for separating mixtures of
biomolecules by extraction [1–4].
The thermodynamics of polymeric solutions has been
studied extensively. Principal models for phase equilibrium
calculations are those of Flory [5] and Huggins [6], who de-
veloped an expression based on lattice theory to describe
the nonidealities of polymer solutions and of Edmond and
Ogston [7], who modeled nonidealities with a truncated os-
motic virial expansion based on McMillan–Mayer theory [8].
∗ Corresponding author. Tel.: +98 6660075; fax: +98 6660075.
E-mail address: rahsadeghi@yahoo.com (R. Sadeghi).
two-phase systems by others [2,7,9–15]. Local composition
models UNIQUAC, UNIFAC, NRTL and NRF have also been
used to describe the thermodynamics of polymer solutions.
Among these local composition models, only the UNIQUAC
model [16] can be used for the modeling of the phase equi-
librium of polymer solutions without further modification.
Kang and Sandler [17,18] and Hartounian et al. [19] used
the UNIQUAC solution model to deal with the phase be-
havior of polymer–polymer aqueous two-phase systems. The
UNIFAC model [20] was extended to polymer solutions by
Oishi and Prausnitz [21]. Tan and Shen [22] used the poly-
mer UNIFAC model to represent the LLE phase diagrams of
polymer–polymer aqueous two-phase systems. Two different
versions of the NRTL model [23] have been used for polymer
solutions. Chen [24] developed a segment-based local com-
position model for multisolvent, single polymer solutions,
which uses a combination of the Flory–Huggins expression
for the entropy of mixing molecules of different sizes and the

0378-3812/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.fluid.2005.01.003
78 R. Sadeghi, M.T. Zafarani-Moattar / Fluid Phase Equilibria 231 (2005) 77–83

NRTL equation for weak local physical interactions between the sum of two contributions:
solvent and segments of polymer chains. More recently, Wu Config.
et al. [25] developed the modified NRTL model for the repre- ln γI = ln γI + ln γILC (2)
sentation of the Helmholtz energy of polymer solutions. They In this work, the Flory–Huggins expression was used for the
used a slightly different version of the NRTL equation; and entropic contribution and the modified NRTL and modified
for the entropic contribution, the truncated Freed [26] cor- NRF equations were used for the local composition contri-
rection to the Flory–Huggins expression as first correction bution to the excess Gibbs energy.
was used. Later, Wu et al. [27] further extended their poly-
mer NRTL model to describe the LLE of PEG–DEX aqueous 2.1. Entropic contributions to the excess Gibbs energy
two-phase systems.
Recently, Haghtalab and Vera [28] developed an NRF The Flory–Huggins equation for the configurational en-
model for electrolyte solutions. It was shown [28–30] that tropy of mixing of multicomponent polymer solutions can be
the model consistently produces better results than the NRTL written as:
model and reproduces the experimental value from the dilute
GEX,Config.
φI
region up to saturation. In our previous paper [31], the NRF = nI ln (3)
model was extended to single solvent, single polymer solu- RT xI
I
tions and applied to correlate the solvent activity of some
polymer solutions. It was shown [31] that the model con- therefore
 
sistently produces better results than the polymer NRTL, Config. φI φI
ln γI = ln +1− (4)
Flory–Huggins and UNIQUAC models and reproduces the xI xI
experimental value from the dilute region up to saturation.
The polymer NRF model [31] consists of two contributions where
r I nI
due to the configurational entropy of mixing, represented by φI =  (5)
the Flory–Huggins relation, and to the enthalpic contribu- J rJ n J
tion, represented by local compositions through nonrandom In these relations, nI , xI and rI are the number of moles, the
factors. mole fraction and the number of segments of the component
In this work, the modified NRTL model for multisolvent, I. Following Chen [24] we assume that the molar volume of
single polymer solutions [24] and the modified NRF model all segments and solvents are constant, thus for the solvents
for single solvent, single polymer solutions [31] have been rI = 1.
further developed for the representation of the excess Gibbs
energy of multicomponent polymer solutions. To obtain the 2.2. Local composition contributions to the excess Gibbs
necessary expression for excess Gibbs energy for polymer so- energy
lutions we considered two contributions. The modified NRTL
and NRF equations were used for weak local physical inter- Following Chen [24], for the modification of the NRF
actions between solvents and segments of polymer chains and NRTL models to polymer solutions the local composi-
(the local composition contribution). The Flory–Huggins ex- tion concept is applied to the individual segments and solvent
pression for the configurational entropy of mixing was used molecules, not the polymer chain. This approach reflects the
as an entropic contribution to the excess Gibbs energy. The vision that each segment should exert unique local physical
applicability of the developed models has been tested us- interaction characteristic with its immediate neighboring sol-
ing experimental liquid–liquid equilibrium phase diagrams vent species or segments. This unique physical interaction of
of some polymer–polymer aqueous two-phase systems. the segments determines the most favorable local environ-
ment around a segment. Following the NRTL derivation, the
effective local mole fractions Xji and Xii of species j and
2. Outline of the thermodynamic model i, respectively, in the immediate neighborhood of a central
species i are related by
As in our previous work [31], we assumed that the ex-  
Xji Xj
cess Gibbs energy of a polymer solution can be expressed = Gji (6)
as the sum of the configurational entropy of mixing (en- Xii Xi
tropic contribution), GEX,Config. , and the local composition where
contribution,GEX,LC : xI ri,I σI
Xi =   (7)
J j xJ rj.J
GEX = GEX,Config. + GEX,LC (1)
Gji = exp(−αji τji ) (8)
Therefore, the activity coefficient of component I (polymers gji − gii
τji = (9)
or solvents) in a polymer solution can also be considered as RT
 R. Sadeghi, M.T. Zafarani-Moattar / Fluid Phase Equilibria 231 (2005) 77–83 79

Here ri,I is the number of the segment i in the component I; 2.2.2. The NRF theory
αji the nonrandomness factor, which was fixed at 0.25; gji and As mentioned above the reference states for the modified
gii the energies of interaction between j–i and i–i species, re- NRF equation are the random case for all the species. In the
spectively. Both gji and αji are inherently symmetric (gji = gij random case [28]
and αji = αij ). The species I and J can be solvent molecules

or polymers and species i and j can be solvent molecules or giref,NRF = Xj gji (16)
segments. For a multicomponent polymer solution, the local j
composition contributions to the excess Gibbs energy can be
Considering the above relations, we obtained the NRF equa-
expressed as:
tion for multicomponent polymer solutions as
GEX,LC

 
= nI ri,I giEX (10) GEX,NRF

X j G ji τ ji

RT = nI ri,I  
j
− Xj τji 
I i
RT k Xk Gki
I i j
where giEX is the molar contribution of species i to the excess (17)
Gibbs energy of the whole solution and can be defined with
respect to a reference state as: therefore

giEX = gi − giref (11)

Xj Gij
j Xj Gji τji
ln γINRF = ri,I   + 
where i k Xk Gki j k Xk Gkj

 
gi = Xji gji (12) 
k Xk Gkj τkj
j × τij − 
k Xk Gkj
The major difference between the NRTL model and the NRF   
model is in their reference states. The reference states for



+ ri,I  Xj Xk τkj − τij − τji 
the modified NRTL model are pure liquids for solvents and
i j k
a hypothetical segment aggregate state for segments [24].
(18)
But the reference states for the modified NRF model are the
random case for all the species [31]. The models require two binary interaction parameters, τ ij
and τ ji , for each of the solvent–solvent interaction pairs, the
2.2.1. The NRTL theory solvent–segment interaction pairs, and the segment–segment
As mentioned above, the reference states for the modified interactions pairs. These binary interaction parameters be-
NRTL model are pure liquids for solvents and a hypotheti- come the correlation variables in treating the thermody-
cal segment aggregate state for segments [24]. In this hypo- namic properties of polymer solutions. The expressions for
thetical aggregate state, all segments are surrounded by seg- the excess Gibbs energy and activity coefficient of species
ments of the same type. The reference states for solvents are for a ternary polymer–polymer–water system are given in
characterized by the solvent–solvent physical interactions. Appendix A.
The reference states for segments are characterized by the
segment–segment physical interactions. Thus
3. Results and discussion
giref,NRTL = gii (13)
The applicability of the proposed models has been
Considering the above relations, we obtained the NRTL equa-
tested using experimental liquid–liquid equilibrium data
tion for multicomponent polymer solutions as
for a variety of polymer–polymer aqueous two-phase

GEX,NRTL

j Xj Gji τji
systems. The aqueous two-phase systems described
= nI ri,I  (14) in this work are PEG–DEX–water, Ficoll–DEX–water,
RT k Xk Gki
I i polyvinyl alcohol (PVA)–DEX–water, polyvinyl pyrroli-
therefore we have done (PVP)–DEX–water and PEG–polyvinyl methyl ether
 (PVME)–water for different molecular weights of polymers.

j Xj Gji τji

Xj Gij The values for rI were calculated using rI = VI /V3 , where
ln γINRTL = ri,I   +  VI is the molar volume of component I and V3 the molar
i k Xk Gki j k Xk Gkj
volume of water. The molar volumes of polymers have been
   calculated from the specific volumes of polymers. The spe-
k Xk Gkj τkj
× τij −  (15) cific volume of PEG, DEX, Ficoll, PVA, PVME and PVP are
k Xk Gkj 0.838 [32], 0.602 [27], 0.8 [33], 0.766 [33], 0.898[33] and
0.799 [34], respectively.
80 R. Sadeghi, M.T. Zafarani-Moattar / Fluid Phase Equilibria 231 (2005) 77–83

Table 1
The parameters of the modified NRTL model and modified NRF model for some polymer–polymer aqueous two-phase systems
Number System τ 12 τ 21 τ 13 τ 31 τ 23 τ 32
NRTL
1 PEG (1)–DEX (2)–water (3) −1.8270 4.5550 −1.2420 1.7740 0.3180 0.2737
2 Ficoll (1)–DEX (2)–water (3) −1.6641 6.2408 −0.8553 1.3778 0.3180 0.2737
3 PVA (1)–DEX (2)–water (3) −1.5981 7.6454 −0.8735 1.3536 0.3180 0.2737
4 PVP (1)–DEX (2)–water (3) −0.5939 0.7070 −0.3061 0.5693 0.0743 0.1210
5 PEG (1)–PVME (2)–water (3) −1.4023 2.6417 −1.2420 1.7740 1.3295 −0.1949
NRF
1 PEG (1)–DEX (2)–water (3) −3.2350 −1.3791 7.1017 −8.4868 0.0108 −1.5199
2 Ficoll (1)–DEX (2)–water (3) 1.5062 −3.3634 −1.5104 −3.5125 0.0108 −1.5199
3 PVA (1)–DEX (2)–water (3) −0.0471 −6.786 × 10−4 2.766 × 10−5 −1.5972 0.0108 −1.5199
4 PVP (1)–DEX (2)–water (3) −0.0549 0.0343 −0.8420 1.0029 −1.1348 −1.3815
5 PEG (1)–PVME (2)–water (3) −3.9980 −2.2730 7.1017 −8.4868 −0.0440 −1.6640

The model parameters of the modified NRTL and NRF where wp,l,I is the weight percent of the component I in
models were estimated by minimizing the following objective the phase p for lth tie-line. The liquid–liquid equilibrium
function and have been listed in Table 1: data were correlated using the obtained parameters given in
Table 1 and the equilibrium condition:



exp
OF = p,l,I − wp,l,I )
(wcal (19)
p l I (xI γI )top = (xI γI )bot (20)

Table 2
The predicted results and its comparison with the experimental ones for some polymer–polymer aqueous two-phase systems
Number System Mn (1) Mn (2) T (K) Dev(NRTL)a Dev(NRF) Ref.
1 PEG4000 (1)–DEX 17 (2) 3810 13239 298.15 1.16 0.48 35
PEG4000 (1)–DEX 19 (2) 3810 27705 298.15 0.43 0.69 35
PEG4000 (1)–DEX 37 (2) 3810 52066 298.15 0.68 0.52 35
PEG4000 (1)–DEX 48 (2) 3810 88252 298.15 0.76 0.35 35
PEG6000 (1)–DEX 17 (2) 5318 13239 298.15 0.44 0.68 35
PEG6000 (1)–DEX 19 (2) 5318 27705 298.15 0.50 0.62 35
PEG6000 (1)–DEX 37 (2) 5318 52066 298.15 1.28 0.65 35
PEG6000 (1)–DEX 48 (2) 5318 88252 298.15 1.33 0.77 35
PEG10000 (1)–DEX 17 (2) 10216 13239 298.15 1.06 0.78 35
PEG10000 (1)–DEX 19 (2) 10216 27705 298.15 1.30 0.62 35
PEG10000 (1)–DEX 37 (2) 10216 52066 298.15 1.15 0.41 35
PEG10000 (1)–DEX 48 (2) 10216 88252 298.15 1.37 0.84 35
PEG20000 (1)–DEX 17 (2) 19081 13239 298.15 1.04 1.24 35
PEG20000 (1)–DEX 19 (2) 19081 27705 298.15 0.86 0.64 35
PEG20000 (1)–DEX 37 (2) 19081 52066 298.15 0.68 0.57 35
PEG20000 (1)–DEX 48 (2) 19081 88252 298.15 1.29 0.40 35
PEG3400 (1)–DEX 40 (2) 3400 24400 295.15 1.32 1.02 9
PEG3400 (1)–DEX 70 (2) 3400 38400 295.15 1.75 1.64 9
PEG3400 (1)–DEX 500 (2) 3400 234200 295.15 1.74 1.55 9
PEG8000 (1)–DEX 40 (2) 8000 24400 295.15 1.23 0.38 36
PEG8000 (1)–DEX 70 (2) 8000 38400 295.15 1.14 0.55 36
1 PEG8000 (1)–DEX 500 (2) 8000 234200 295.15 1.32 0.77 36
PEG20000 (1)–DEX 40 (2) 20000 24200 295.15 1.75 1.75 36
PEG20000 (1)–DEX 70 (2) 20000 38400 295.15 1.78 2.38 36
PEG20000 (1) –DEX500 (2) 20000 234200 295.15 1.20 1.12 36
PEG3350 (1)–DEX 70 (2) 3690 37000 298.15 1.27 1.02 13
PEG8000 (1)–DEX 500 (2) 8920 167000 298.15 1.15 1.32 13
PEG6000 (1)–DEX 70 (2) 6000 28700 296.15 1.04 0.63 3
2 Ficoll 400 (1)–DEX 70 (2) 400000 28700 296.15 0.095 0.064 3
3 PVA (1)–DEX 70 (2) 55000 28700 295.15 0.17 0.21 3
4 PVP (1)–DEX 70 (2) 12700 28700 281.15 0.38 0.30 3
5 PEG4000 (1)–PVME100 (2) 3520 51000 293.15 0.90 0.63 37
PEG7500 (1)–PVME100 (2) 10100 51000 293.15 0.66 0.65 37
PEG35000 (1)–PVME100 (2) 29400 51000 293.15 1.22 1.38 37
a Dev = OF/6N where N is the number of tie-lines.
 R. Sadeghi, M.T. Zafarani-Moattar / Fluid Phase Equilibria 231 (2005) 77–83 81

Table 2 summarizes the absolute deviations of all aque- x mole fraction

ous two-phase systems described in this paper, along with X effective local mole fraction
the sources of the experimental phase diagrams. The ordi-
nal numbers on the first column of Table 2 correspond to Greek letters
those in Table 1. That is to say, the LLE of all PEG–DEX α nonrandomness factor
aqueous two-phase systems (27 systems) numbered with 1 γ activity coefficient
in Table 2 are simultaneously predicted with the parame- τ NRTL or NRF parameters
ters numbered with 1 in Table 1, that of the 1 system num- φ segment fraction
bered with 2 in Table 2 is predicted with the parameters
numbered with 2 in Table 1, and so on. Also, we assumed Subscripts
that at the small temperature range 293.15–298.15 the in- i, j, k any species, solvents or segments
teraction parameters are constant. As mentioned above, the I,J any species, solvents or polymers
models used in this paper are segment based. The advan- l tie-lines
tage of the segment-based models over conventional mod- p phase (top and bottom)
els for the correlation of polymer solution experimental 1,2 polymers
data is that, unlike the classical models, they can cover a 3 solvent
wide range of polymer molar masses with a series of in-
teraction parameters, which provides a predictive capability. Superscripts
The results show that, for both models the calculated pa- bot bottom phase
rameters are independent of the polymer molecular weights cal calculated value
and can also give a good representation of the LLE phase Config. configurational
diagram of the polymer–polymer aqueous two-phase sys- exp experimental value
tems at different polymer molecular weights. On the basis EX excess
of the obtained deviations, given in Table 2, we conclude LC local composition
that the two models have similar behavior in the LLE rep- NRF local composition contribution, represented by the
resentation for aqueous two-phase systems and the calcu- nonrandom factor equation
lation accuracy of LLE data being around 1 wt.% for both NRTL local composition contribution, represented by the
models. nonrandom two liquid equation
ref reference state
top top phase
4. Conclusion

The modified NRTL model for multisolvent, single poly- Appendix A. Application of the modified NRTL and
mer solutions and the modified NRF model for single sol- NRF models for the polymer–polymer aqueous
vent, single polymer solutions have been further extended two-phase systems
for the representation of the excess Gibbs energy of mul-
ticomponent polymer solutions. These developed models • The Flory–Huggins equation for the configurational en-
have been applied to predict the LLE phase diagrams tropy of mixing
of some polymer–polymer aqueous two-phase systems. It For a polymer (1)–polymer (2)–water (3) system, Eq.
was found that, the proposed models represent the exper- (3) becomes:
imental LLE data of the polymer–polymer aqueous two-
GEX,Config. φ1 φ2 φ3
phase systems with good accuracy. Also, we found that the = n1 ln + n2 ln + n3 ln (A.1)
model parameters are independent of the polymer molecular RT x1 x2 x3
weight. • The NRTL theory
For a polymer (1)–polymer (2)–water (3) system, Eqs.
List of symbols (14) and (15) become:
g molar Gibbs energy, or energy parameter  
GEX,NRTL φ2 G21 τ21 + φ3 G31 τ31
G Gibbs energy = n 1 r1
n number of moles RT φ1 + φ2 G21 + φ3 G31
N number of tie-lines  
φ1 G12 τ12 + φ3 G32 τ32
OF objective function + n2 r2
r degree of polymerization φ1 G12 + φ2 + φ3 G32
R gas constant  
T temperature φ1 G13 τ13 + φ2 G23 τ23
+ n3 (A.2)
w weight percent φ1 G13 + φ2 G23 + φ3
82 R. Sadeghi, M.T. Zafarani-Moattar / Fluid Phase Equilibria 231 (2005) 77–83

 
1 φ2 G21 τ21 + φ3 G31 τ31 φ1 φ2 G12
ln γ1NRTL = 1− +
r1 φ1 + φ2 G21 + φ3 G31 φ1 + φ2 G21 + φ3 G31 φ1 G12 + φ2 + φ3 G32
   
φ1 G12 τ12 + φ3 G32 τ32 φ3 G13 φ1 G13 τ13 + φ2 G23 τ23
× τ12 − + τ13 − (A.3)
φ1 G12 + φ2 + φ3 G32 φ1 G13 + φ2 G23 + φ3 φ1 G13 + φ2 G23 + φ3

 
1 φ1 G21 φ2 G21 τ21 + φ3 G31 τ31 φ1 G12 τ12 + φ3 G32 τ32
ln γ2NRTL = τ21 − +
r2 φ1 + φ2 G21 + φ3 G31 φ1 + φ2 G21 + φ3 G31 φ1 G12 + φ2 + φ3 G32
   
φ2 φ3 G23 φ1 G13 τ13 + φ2 G23 τ23
× 1− + τ23 − (A.4)
φ1 G12 + φ2 + φ3 G32 φ1 G13 + φ2 G23 + φ3 φ1 G13 + φ2 G23 + φ3

 
φ1 G31 φ2 G21 τ21 + φ3 G31 τ31 φ2 G32
ln γ3NRTL = τ31 − +
φ1 + φ2 G21 + φ3 G31 φ1 + φ2 G21 + φ3 G31 φ1 G12 + φ2 + φ3 G32
   
φ1 G12 τ12 + φ3 G32 τ32 φ1 G13 τ13 + φ2 G23 τ23 φ3
× τ32 − + 1− (A.5)
φ1 G12 + φ2 + φ3 G32 φ1 G13 + φ2 G23 + φ3 φ1 G13 + φ2 G23 + φ3
• The NRF theory
For a polymer (1)–polymer (2)–water (3) system, Eqs. (17)
and (18) become:
   
GEX,NRF φ2 G21 τ21 + φ3 G31 τ31 φ1 G12 τ12 + φ3 G32 τ32
= n 1 r1 − φ2 τ21 − φ3 τ31 + n2 r2 − φ1 τ12 − φ3 τ32
RT φ1 + φ2 G21 + φ3 G31 φ1 G12 + φ2 + φ3 G32
 
φ1 G13 τ13 + φ2 G23 τ23
+ n3 − φ1 τ13 − φ2 τ23 (A.6)
φ1 G13 + φ2 G23 + φ3

 
1 φ2 G21 τ21 + φ3 G31 τ31 φ1
ln γ1NRF = 1− + (φ1 − 1)(φ2 τ21 + φ3 τ31 )
r1 φ1 + φ2 G21 + φ3 G31 φ1 + φ2 G21 + φ3 G31
 
φ2 G12 φ1 G12 τ12 + φ3 G32 τ32
+ τ12 − + (φ1 − 1)φ2 τ12 + φ2 φ3 τ32
φ1 G12 + φ2 + φ3 G32 φ1 G12 + φ2 + φ3 G32
 
φ3 G13 φ1 G13 τ13 + φ2 G23 τ23
+ τ13 − + (φ1 − 1)φ3 τ13 + φ2 φ3 τ23 (A.7)
φ1 G13 + φ2 G23 + φ3 φ1 G13 + φ2 G23 + φ3

 
1 φ1 G21 φ2 G21 τ21 + φ3 G31 τ31
ln γ2NRF = τ21 − + (φ2 − 1)φ1 τ21 + φ1 φ3 τ31
r2 φ1 + φ2 G21 + φ3 G31 φ1 + φ2 G21 + φ3 G31
 
φ1 G12 τ12 + φ3 G32 τ32 φ2
+ 1− + (φ2 − 1)(φ1 τ12 + φ3 τ32 )
φ1 G12 + φ2 + φ3 G32 φ1 G12 + φ2 + φ3 G32
 
φ3 G23 φ1 G13 τ13 + φ2 G23 τ23
+ τ23 − + (φ2 − 1)φ3 τ23 + φ1 φ3 τ13 (A.8)
φ1 G13 + φ2 G23 + φ3 φ1 G13 + φ2 G23 + φ3

 
φ1 G31 φ2 G21 τ21 + φ3 G31 τ31
ln γ3NRF = τ31 − + (φ3 − 1)φ1 τ31 + φ1 φ2 τ21
φ1 + φ2 G21 + φ3 G31 φ1 + φ2 G21 + φ3 G31
 
φ2 G32 φ1 G12 τ12 + φ3 G32 τ32
+ τ32 − + (φ3 − 1)φ2 τ32 + φ1 φ2 τ12
φ1 G12 + φ2 + φ3 G32 φ1 G12 + φ2 + φ3 G32
 
φ1 G13 τ13 + φ2 G23 τ23 φ3
+ 1− + (φ3 − 1)(φ1 τ13 + φ2 τ23 ) (A.9)
φ1 G13 + φ2 G23 + φ3 φ1 G13 + φ2 G23 + φ3
 R. Sadeghi, M.T. Zafarani-Moattar / Fluid Phase Equilibria 231 (2005) 77–83
References
[1] P.A. Albertsson, Partitioning of Cell Particles and Macromolecules,
3rd ed., Wiley-Interscience, New York, 1986.
[2] H. Walter, D.E. Brooks, D. Fisher, Partitioning in Aqueous Two-
phase Systems, Academic Press, New York, 1985.
[3] B.Y. Zaslavsky, Aqueous two-phase partitioning, in: Physical Chem-
istry and Bioanalytical Applications, Marcel Dekker, New York,
1995.
[4] M.-R. Kula, K.H. Kroner, H. Hustedt, Adv. Biochem. Eng. 24 (1982)
73–118.
[5] P.J. Flory, J. Chem. Phys. 9 (1941) 660–661.
[6] M.L. Huggins, J. Chem. Phys. 9 (1941) 440.
[7] E. Edmond, A.G. Ogston, Biochem. J. 109 (1968) 569–576.
[8] W.G. McMillan, J.E. Mayer, J. Chem. Phys. 13 (1945) 276–284.
[9] A.D. Diamond, J.T. Hsu, Biotechnol. Bioeng. 34 (1989) 1000–
1014.
[10] A.D. Diamond, J.T. Hsu, AIChE J. 36 (1990) 1017–1024.
[11] T. Hino, J.M. Prausnitz, J. Appl. Polym. Sci. 68 (1998) 2007–2017.
[12] M.T. Zafarani-Moattar, R. Sadeghi, Fluid Phase Equilib. 203 (2002)
177–191.
[13] R.S. King, H.W. Blanch, J.M. Prausnitz, AIChE J. 34 (1988)
1585–1594.
[14] H. Cabezas Jr., J.D. Evans, D.C. Szlag, Fluid Phase Equilib. 53
(1989) 453–462.
[15] C. Grossmann, J. Zhu, G. Maurer, Fluid Phase Equilib. 82 (1993)
182–275.
[16] D.C. Abrams, J.M. Prausnitz, AIChE J. 21 (1975) 116–128.
[17] C.H. Kang, S.I. Sandler, Fluid Phase Equilib. 38 (1987) 245–272.
[18] C.H. Kang, S.I. Sandler, Ind. Eng. Chem. Res. 28 (1989) 1537–
1542.
83
[19] H. Hartounian, E. Floeter, E.W. Kaler, S.I. Sandler, AIChE J. 39
(1993) 1976–1984.
[20] A. Fredenslund, R.L. Jones, J.M. Prausnitz, AIChE J. 21 (1977)
1086–1099.
[21] T. Oishi, J.M. Prausnitz, Ind. Eng. Chem. Proc. Des. Dev. 17 (1978)
333–339.
[22] T.-W. Tan, Z.-Y. Shen, Proceedings of the International Symposium
on Thermodynamics in Chemical Engineering and Industry, Chem-
ical Industrial and Engineering Society of China, National Natu-
ral Science Foundation of China, Beijing, May 24–27, 1994, pp.
171–175.
[23] H. Renon, J.M. Prausnitz, AIChE J. 14 (1968) 135–142.
[24] C.C. Chen, Fluid Phase Equilib. 83 (1993) 301–312.
[25] Y.-T. Wu, Z.-Q. Zhu, D.-Q. Lin, L.-H. Mei, Fluid Phase Equilib.
121 (1996) 125–139.
[26] J. Dudowicz, K.F. Freed, W.G. Madden, Macromolecules 23 (1990)
4803–4819.
[27] Y.-T. Wu, D.-Q. Lin, Z.-Q. Zhu, Fluid Phase Equilib. 147 (1998)
25–43.
[28] A. Haghtalab, J.H. Vera, AIChE J. 34 (1988) 803–813.
[29] M.T. Zafarani-Moattar, J. Jahanbin Sardoodi, Fluid Phase Equilib.
166 (1999) 207–223.
[30] M.T. Zafarani-Moattar, J. Jahanbin Sardoodi, Fluid Phase Equilib.
172 (2000) 221–235.
[31] M.T. Zafarani-Moattar, R. Sadeghi, Fluid Phase Equilib. 202 (2002)
413–422.
[32] L. Lepori, V. Mollica, J. Polym. Sci.: Polym. Phys. Ed. 16 (1978)
1123–1134.
[33] R. Zana, J. Polym. Sci.: Polym. Phys. Ed. 18 (1980) 121–126.
[34] R. Sadeghi, M.T. Zafarani-Moattar, J. Chem. Thermodyn. 36 (2004)
665–670.