Modeling of SolidLiquid Equilibria for Polyethylene

and Polypropylene Solutions with Equations of State

G. M. N. Costa,1 S. Kislansky,1 L. C. Oliveira,1 F. L. P. Pessoa,2 S. A. B. Vieira de Melo,1
M. Embirucu1
1

Programa de Engenharia Industrial, Escola Politecnica, Universidade Federal da Bahia,

Rua Professor Aristides Novis, 2, Federacao, CEP 40210-630, Salvador-Bahia, Brazil
2
Departamento de Engenharia Qumica, Escola de Qumica, Universidade Federal do Rio de Janeiro,
Centro de Tecnologia, bloco E, S. 209, Ilha do Fundao, 21949-900, Rio de Janeiro RJ, Brazil
Received 9 May 2010; accepted 4 August 2010
DOI 10.1002/app.33128
Published online 10 March 2011 in Wiley Online Library (wileyonlinelibrary.com).
ABSTRACT: We modeled solidliquid equilibria (SLEs)
in polyethylene and polypropylene solutions with a
SoaveRedlichKwong (SRK) cubic equation of state
(EOS) and a perturbed-chain statistical associating fluid
theory (PC-SAFT) EOS. Two types of mixing rules were
used with SRK EOS: The WongSandler mixing rule and
the linear combination of the Vidal and Michelsen mixing
rules (LCVM), both of which incorporated the Bogdanic
and Vidal activity coefficient model. The performance of
these models was evaluated with atmospheric-pressure
and high-pressure experimental SLE data obtained from
literature. The basic SLE equation was solved for the
equilibrium melting temperature instead of for the

composition. The binary interaction parameters of SRK

and PC-SAFT EOS were estimated to best describe the
experimental equilibrium behavior of 20 different polymersolvent systems at atmospheric pressure and 31
other polymersolvent systems at high pressure. A
comparison with experimental data showed that SRK
LCVM agreed very well with the atmospheric SLE
data and that PC-SAFT EOS was more efficient in highC 2011 Wiley Periodicals, Inc. J Appl
pressure conditions. V

INTRODUCTION

and, thus, stop them from occurring, a task that cannot be done with empirical models,5 especially when
one is dealing with complex systems, such as solution and slurry polymerization reactions.6,7
Two basic approaches can be used to model the
liquid-phase nonideality in SLE: excess Gibbs freeenergy (GE) models and equations of state (EOSs).
Some researchers have attempted to describe the
liquid-phase nonidealities using the universal quasichemical (UNIQUAC) functional-group activity coefficient and entropic free volume (entr-FV) models4
or by applying perturbed-chain statistical associating
fluid theory (PC-SAFT) EOS to the modeling of SLE
of binary and ternary systems with a solid-complex
phase formation.8 Some work has also been done on
the development of algorithms for the real-time prediction of SLE in the solution polymerization of PE
based on PC-SAFT EOS and to study the effects of
monomer and polymer polydispersity in solution
polymerization processes.9 Furthermore, copolymer
PC-SAFT EOS have been used to model SLE in systems containing PE, m-xylene, and amyl acetate.10
Polymer solutions are not adequately described by
cubic EOSs and their van der Waals type one-fluid
mixing rules. However, over the past 2 decades, a
large number of mixing rules have been proposed to
improve the performance of these kinds of equations

Polyethylene (PE) and polypropylene (PP) are global

commodity polymers. During the industrial processing of PE and PP, deposition of the polymer on the
reactor surface, heat exchangers, and flash drums
frequently occurs, and in pipelines, this can cause
clogging. Therefore, the solubility behavior of the
polymer is key information for solvent selection and
polymer processing in industry. For instance, in the
solution polymerization of ethylene, the PE produced in the reactors should remain in solution
because its precipitation from the solution is undesirable.13 Solvent selection is also crucial and is
dependent on the polymer solubility because polymer recycling requires detailed screening and identification of possible solvents for the polymers in the
waste material.4 The modeling of solidliquid equilibrium (SLE) is a useful method for gaining a better
understanding of these industrial polymer problems
Correspondence to: M. Embirucu ([email protected]).
Contract grant sponsors: Fundacao de Amparo a`
Pesquisa do Estado da Bahia, Conselho Nacional de
Desenvolvimento Cientfico e Tecnologico.
Journal of Applied Polymer Science, Vol. 121, 18321849 (2010)

C 2011 Wiley Periodicals, Inc.

V

Polym Sci 121: 18321849, 2011

Key words: phase behavior; phase diagrams; phase

separation; poly(propylene); (PP); polyethylene (PE)

MODELING OF SOLIDLIQUID EQUILIBRIA

1833

in representing the phase behavior for highly nonideal

mixtures, such as PE and PP solutions.6,11 Mixing rules
for cubic EOSs derived from GE models, called EOS/
GE models, such as the linear combination of the Vidal
and Michelsen mixing rules (LCVM)12 and the Wong
Sandler (WS)13 mixing rule, have been proposed extensively for the prediction of vaporliquid equilibrium,
but they are rarely used to describe SLE behavior.
In this study, only the EOS approach was used to
investigate the liquid-phase nonideality in SLE
for polymer systems. PC-SAFT and SoaveRedlich
Kwong (SRK)14 EOS were chosen, with LCVM and WS
mixing rules. On the basis of a previous article, in
which the evaluation of several activity coefficient
models, both at infinite dilution and finite concentration, were reported, the Bogdanic and Vidal (BV) activity coefficient model was selected to be incorporated
into LCVM and WS mixing rules of the SRK EOS.15
These three EOS-based models (PC-SAFT, SRKWS,
and SRKLCVM) were used to solve the equilibrium
equation for temperature. Experimental atmospheric
(for both nonassociated and associated systems) and
high-pressure SLE data available in the literature
were used to compare their performances. The NewtonRaphson (NR) and Regulafalsi (Rf) numerical
methods1619 were used and compared to solve this
equilibrium equation. For both atmospheric and highpressure experimental data, the performances of the
models with the estimated interaction parameters
averaged over all temperatures were investigated,
and the feasibility of obtaining generalized correlations among the binary interaction parameters and
the independent variables, such as polymer molecular
weight (MW) and polymer concentration, was evaluated. Finally, for high-pressure conditions, the influence of the solvent, pressure range, and polymer MW
on the performance of the models and the parameter
correlation interpolating and extrapolating capabilities
were examined through the charting of the experimental and model results.

stands for the polymer in solution, the superscript 0

refers to the pure liquid polymer, x is the molar fraction (solubility) in solution, DHu is the enthalpy of
melting per moles of crystal units, R is the gas constant, Tm is the polymer melting temperature, T is
the temperature, Dv is the polymer volume change,
P is the pressure, c is the crystallinity fraction, and u
is the number of monomer units in the polymer
backbone. The / values are calculated by an EOS.
PC-SAFT EOS
The PC-SAFT EOS is based on a perturbation theory,
where the hard-chain reference term represents the
repulsive forces and the perturbation terms reflect
the various attractive interactions, as described in
detail elsewhere.20,21 To associate fluids, this EOS is
conveniently written in terms of the reduced residual Helmholtz free energy as follows:
Ares
Ahc
Adisp Aassoc

NkT NkT NkT

NkT

where A is the Helmholtz free energy per number of

molecules, where the superscripts res, hc, disp, and
assoc identify the residual, hard-chain, dispersive
attraction, and associative attraction contributions,
respectively; N is the total number of molecules; and
k is the Boltzmann constant. The three terms on the
right-hand side of eq. (2) correspond to the hardchain, dispersion, and association contributions,
respectively. All of the required expressions for
these terms in eq. (2) and the pure-component
parameters for many substances can be found in the
literature.20,21 When PC-SAFT is used for mixtures,
the conventional BerthelotLorenz combining rules
used to calculate the mixture properties in the hard
chain and the dispersion term are applied, with only
one adjustable binary interaction parameter of the
PC-SAFT EOS (Kij) introduced to account for the
dispersive interactions, as follows:

THERMODYNAMIC MODELS

1
rij ri rj
2
p
eij ei ej  1
Kij

Solubility basic equation

Following the hypothesis of Harismiadis and
Tassios4 about the fugacity ratio and its relationship
with crystallinity, Pan and Radoz10 assumed the
solid phase to be pure solute and ignored the two
heat-capacity terms, proposing a SLE model on the
basis of an EOS, which is given as follows:
ln

uLp xp
u0p



DHu
Tm
DvP



1
 cu

RTm
T
RT

(2)

(3)
(4)

where r is the segment diameter, e is the dispersion

energy parameter, and the subscripts i and j refer to
specific components or groups.
SRK EOS

(1)

where / is the fugacity coefficient, the subscript p

stands for the polymer (solute), the superscript L

The SRK EOS14 is expressed as follows:

P

RT
aT


v
b vv b

(5)

Journal of Applied Polymer Science DOI 10.1002/app

1834

COSTA ET AL.

where a and b are parameter of the SRK EOS. To

extend the SRK EOS to polymersolvent systems,
the following mixing rules were considered: WS
and LCVM.
WS mixing rule
The WS mixing rule13 is a representative rule based
on the condition of infinite pressure, given as
follows:
X xi ai

ab
P P
b

1

GE
d

j xi xj b
a=RTij

bi

xi ai =bi RT
GE =dRT

(6)
(7)

with


b



a  1 
a  
a 
1
WSij
b

b

RT ij 2
RT i
RT j

12

hX
i
a

xi ai
bRT

proposed the LCVM mixing



 
k
1
k GE

Av
AM RT



1
kX
b
9
xi  ln

AM
bi

where Av
ln 2 and AM
0.53 for the SRK EOS

and k 0.36, as originally proposed by Boukouvalas
et al12. The conventional linear mixing rule is used
for parameter b.
GE model

ln centr
FV
i

FV 
b
bFV
1
i
ln i
xi
xi

(11)

where bFV
and xi are the free-volume fraction and
i
the molar fraction of the segment, respectively. FV is
defined as
vFV;i vi
vW;i

(12)

where vi and vW,i are the liquid molar volume and

the van der Waals volume, respectively.
For the calculation of the residual term (ln cres
i ), the
mixture is considered a solution of segments and the
molar fraction of each segment (X) is calculated by:
i

xi d
i
xk Pncomp
Pnseg k j
xj dm
m
j

(13)

where x is the component (i or j) molar fraction, the

subscripts k and m refer to segments and d refers to
the number of segments in a component (i or j). The
summations are extended to the total number of
components (ncomp) and to the total number of segments (nseg). The segment activity coefficients are
calculated with the UNIQUAC model, which contains an interaction parameter denoted as UQij.
EXPERIMENTAL DATA AND GENERAL
APPROACH FOR CALCULATING SLE
Experimental data

Both the LCVM and WS mixing rules require a G

model, and as mentioned previously, the BV22
model is the best choice.15 BV is a segment-based
model containing combinatorial, free volume (FV),
and energy contributions for GE. It is derived from
the entr-FV model and follows the idea of relating
the nonideality of polymersolvent mixtures with
the polymer segmentsolvent interaction parameters.
The energetics contribution is based on interactions
among individual segments (repeating units) of the
polymer (or copolymer) and solvent molecules. The
segment activity coefficients are calculated through
the UNIQUAC model.
Journal of Applied Polymer Science DOI 10.1002/app

(10)

where the first term on the right-hand side of eq.

(10) represents the entr-FV contribution and the
second one is the residual contribution.
As in the original model, the combinatorial and
FV contributions are combined in a single term, the
so-called entr-FV part:

Pncomp

LCVM mixing rule

ln cres
ln ci ln centr
FV
i
i

(8)

where d is a numerical constant equal to (
ln 2) for

the SRK EOS and WSij is the binary interaction
parameter.

Boukouvalas et al.
rule as follows:

The molar activity coefficient of component i (ci)

in the mixture is given by

Although the crystallization of polymer solutions

has received much attention recently, most studies
have focused on the kinetics of the process and on
the morphology of the crystals. Equilibrium data of
solid polymer solutions are often scarce.
Experimental SLE data at atmospheric pressure
available in the literature are presented in this article as temperature versus polymer solubility, and
only those systems that presented a continuous SLE
curve were selected. Tables I and II show the database for the 20 systems investigated with the temperature interval covering the whole range of
weight fractions.

MODELING OF SOLIDLIQUID EQUILIBRIA

1835

TABLE I
Data Points Contained in the Database Used
for the SolidLiquid Calculations at Atmospheric
Pressure: Nonassociated Systems

TABLE II
Data Points Contained in the Database Used
for the SolidLiquid Calculations at Atmospheric
Pressure: Associated Systems

Number
of data Temperature
points
range (K) Reference

System
1. PE (17,000)/xylene
2. PE (13,600)/heptane
3. PE (20,000)/dyphenil
4. PP (243,000)/tetradecane
5. PP (243,000)/eicosane
6. PP (28,000)/n-octane
7. PP (28,000)/cis-decalin
8. PE (13,600)/cetene
9. PE (31,000)/xylene
10. PE (32,600)/xylene
11. PE (13,600)/paraffin wax

8
5
6
3
3
3
3
7
6
6
7

345376
341351
377383
435452
445455
331357
321348
356362
370380
376384
357370

System

Number
of data
points

Temperature
range (K)

Reference

3
4

446456
371384

25
26

360382

26

369382

26

357368

26

4
4
3

357368
358366
443453

26
26
25

444454

25

12. PP (243,000)/C32H66
13. PP (28,000)/n-amyl
alcohol
14. PP (28,000)/n-hexyl
alcohol
15. PP (28,000)/n-octyl
alcohol
16. PP (28,000)/
isoamylacetate
17. PP (28,000)/phenetol
18. PP (28,000)/anisole
19. PP (243,000)/
n-eicosanoic
20. PP (243,000)/
n-pentadecanoic

23
23
24
25
25
26
26
23
27
27
23

The high-pressure SLE experimental data are

shown in Tables III and IV with the corresponding
polymer concentrations, pressures, and temperature
ranges. The selected solvents at these pressures and
temperatures were in a supercritical state.

Dv was determined from the densities of an amorphous polymer (qa) and a crystalline one (qc): Dv
1/qa
1/qc. For PE, qa 0.853 g/cm3 and qc
1.004 g/cm3.10 u was calculated with the polymer
and monomer MWs. Although c of the polymer
could have been used as an adjusted parameter, we

Polymer properties for thermodynamic models

The polymer properties used in eq. (1), such as DHu
and polymer Tm, were obtained from van Krevelen.31

TABLE III
Data Points Contained in the Database Used for the SolidLiquid Calculations
at High Pressure

System
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.

PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE
PE

(7000)/ethane
(7000)/ethane
(23,625)/ethane
(23,625)/ethane
(52,000)/ethane
(7000)/propane
(7000)/propane
(7000)/propane
(7000)/propane
(13,600)/propane
(13,600)/propane
(13,600)/propane
(13,600)/propane
(23,625)/propane
(23,625)/propane
(23,625)/propane
(42,900)/propane
(42,900)/propane
(52,000)/propane
(52,000)/propane
(52,000)/propane
(59,300)/propane
(119,600)/propane
(121,000)/pentane
(7000)/ethylene
(23,625)/ethylene

Polymer
concentration

Number
of data
points

Pressure
range (atm)

Temperature
range (K)

Reference

2.00
20.00
2.00
20.00
20.00
2.00
2.70
7.40
17.50
0.25
2.00
5.00
10.00
2.00
6.90
17.00
0.06
3.00
2.00
7.30
17.00
2.00
7.50
5.00
2.00
2.00

6
5
4
4
4
6
6
6
5
3
3
4
4
5
6
11
3
4
6
5
5
11
3
17
4
4

8081501
9021503
12051509
12061511
12081513
6051509
4011502
5971502
5031502
448587
550688
520725
475656
3971500
8041500
7041510
616686
628702
8041513
6671508
7051513
8041500
667735
81537
12001501
17021969

372381
375383
372377
372378
399403
364371
364372
365375
366378
383384
387
387388
387388
365372
366374
366376
382383
387
394404
393405
393405
389395
391392
371374
378383
380382

28
28
28
28
28
28
28
28
28
29
29
29
29
28
28
28
29
29
28
28
28
28
29
30
28
28

Journal of Applied Polymer Science DOI 10.1002/app

1836

COSTA ET AL.

TABLE IV
Data Points Contained in the Database Used to Test the Correlations
for the SolidLiquid Calculations at High Pressure

System
47.
48.
49.
50.
51.

PE
PE
PE
PE
PE

Polymer
concentration

Number
of data
points

Pressure
range (atm)

Temperature
range (K)

Reference

1.90
1.00
1.90
0.25
3.50

4
2
5
2
2

9111504
688721
7051503
609613
652687

369380
385
393404
388389
389390

28
29
28
29
29

(23,625)/propane
(42,900)/propane
(52,000)/propane
(119,600)/propane
(119,600)/propane

used the values reported with the experimental data

or, when these were missing, the values assumed by
Harismiadis and Tassios.4
As mentioned previously, three EOS were used to
calculate /Lp and /0p : PC-SAFT, SRKLCVM, and
SRKWS. Calculations with the SRK EOS require
pure component parameters, which use critical data
as input for their computation; however, for polymers, these critical data are not available. Therefore,
in this study, the pure polymer parameters were
determined with Kontogeorgis et al.s method.32
According to this method, the polymer parameters a
and b are fitted to two experimental volumetric data
at essentially zero pressure. However, in this study,
these parameters were obtained with the group
contribution method group contribution volume
(GCVOL)33 because no experimental volumetric data
were available. van der Waals volume data were
taken from Bondi.34
GCVOL is a group contribution method for the
prediction of the liquid densities for solvents,
oligomers, and polymers as a function of temperature. Elbro et al.33 proposed the following model for
predicting the molar volume (v) of a liquid:
v

ni Dvi

(14)

where ni is the number of group i in the GCVOL

method and the temperature dependence of the
molar group Dvi is given by the following simple
polynomial function (the group volume temperature
constants, Ai, Bi, and Ci, were available for 36 different groups):33
Dvi Ai Bi T Ci T2

(15)

As the SRK EOS polymer parameters a and b are

functions of MW and temperature, different values
of a and b for various values of these two properties
could be calculated with the PE monomer constants:
Ai 12.520 cm3/mol, Bi 0.01294 cm3 mol
1 K
1,
and Ci 0.0 cm3 mol
1 K
2.
Numerical solution procedure
Solving eq. (1) for the solubility of the polymer is
often a very hard task because of the instability of
the numerical procedure.4 To overcome this problem
and to avoid multiple solutions, eq. (1) should be
solved for temperature and not for composition. The
Rf and NR methods1619 were used to find the root
of this equation.
When the Rf and NR methods were applied for
the first 11 nonassociated systems at atmospheric

TABLE V
Average and Maximum Percentile Temperature Errors for the Nonassociated Systems in the SLE
at Atmospheric Pressure
PC-SAFT
System
1
2
3
4
5
6
7
8
9
10
11

AAD (%)
3.90
3.25
4.15
7.39
7.48
5.97
3.19
4.18
1.15
1.47
6.80













4

10
10
4
10
4
10
2
10
1
10
4
10
2
10
4
10
3
10
3
10
4

SRKLCVM

AMD (%)

1

4.66  10
3.92  10
1
6.14  10
1
2.21  10
1
2.24
8.49  10
1
9.38  10
2
5.61  10
1
2.12  10
3
4.11  10
3
1.16

Method
NR
NR
NR
NR
NR
NR
NR
NR
NR
NR
NR

AAD (%)
3.69
4.03
3.97
3.42
2.42
4.30
4.62
3.74
3.79
3.76
3.89













4

10
10
4
10
4
10
4
10
4
10
4
10
4
10
4
10
4
10
4
10
4

AMD (%)
5.05
5.79
4.94
3.79
2.70
5.13
4.98
4.84
4.66
4.17
4.89













1

10
10
1
10
1
10
1
10
1
10
1
10
1
10
1
10
1
10
1
10
1

SRKWS
Method
NR
NR
NR
NR
NR
NR
NR
NR
NR
NR
NR

AAD (%)
2.86
4.45
3.42
5.56
3.32
3.41
3.98
4.09
5.76
3.89
4.81













2

10
10
4
10
4
10
4
10
4
10
4
10
4
10
4
10
2
10
4
10
4

AMD (%)

Method

1.01
5.72  10
1
3.91  10
1
1.11
4.58  10
1
4.27  10
1
5.55  10
1
5.85  10
1
3.42  10
1
4.96  10
1
6.24  10
1

Rf
NR
NR
Rf
NR
NR
NR
NR
Rf
Rf
NR

AMD max[(|Tical
Tiexp|/Texp)  100], where i 1, N; AAD (1/N)  R(|Tcal
Texp|/Texp)  100.
Journal of Applied Polymer Science DOI 10.1002/app

MODELING OF SOLIDLIQUID EQUILIBRIA

1837

TABLE VI
Average and Maximum Percentile Temperature Errors for the Associated
Systems in SLE at Atmospheric Pressure
SRKLCVM
System
12
13
14
15
16
17
18
19
20

AAD (%)
2.55
2.35
3.30
3.69
3.34
3.20
3.32
3.07
2.61











4

10
10
4
10
4
10
4
10
4
10
4
10
4
10
4
10
4

AMD (%)
2.94
3.55
4.05
5.04
3.90
3.67
4.69
3.60
3.24











SRKWS
Method

1

10
10
1
10
1
10
1
10
1
10
1
10
1
10
1
10
1

NR
NR
NR
NR
NR
NR
NR
NR
NR

pressure (Table I), the results demonstrate that the

equilibrium Tm may have been strongly dependent
on the solving method for the SLE equation,
although this was not the case for SRKLCVM
because we observed that calculations with this
model were not sensitive to the solving method.
We also observed that Rf always converged, regardless of the thermodynamic model or the system,
whereas the NR method failed to converge for
some systems when the SRKWS EOS was used.
On the other hand, we observed that when it converged, NR always presented similar or better
results than Rf. For associated systems at atmospheric pressure (Table II), we observed that the
results were independent of the method used to
solve eq. (1), whereas for high-pressure SLE calculations (Table III), NR was the best method. Therefore, on the basis of these results, which cover the
whole set of systems, models, and conditions, our
recommendation is to use the NR method, and if it
fails to converge, the Rf must be used. This recommendation was applied to the results shown in the
latter sections of this article.

AAD (%)
3.38
3.46
3.64
3.19
4.24
3.27
3.52
2.98
3.31











4

10
10
4
10
4
10
4
10
4
10
4
10
4
10
4
10
2

AMD (%)
3.66
3.87
3.98
3.73
5.70
4.18
4.11
3.33
9.87











Method

1

10
10
1
10
1
10
1
10
1
10
1
10
1
10
1
10
1

NR
NR
NR
NR
NR
NR
NR
NR
Rf

For all systems (except for systems 4751 in

Table IV, which were used to test the correlations), the binary interaction parameters were considered tuning parameters for the selected models
through the minimization of the following objective function (OF) with the NelderMead simplex
method:35
OF

N
X
exp
Tj
Tjcal 2

(16)

j1

where superscripts exp and cal identify the experimental and calculated equilibrium temperatures,
respectively. The summation was extended to the
number of experimental data (N).

RESULTS AND DISCUSSION FOR SLE AT

ATMOSPHERIC PRESSURE
As highlighted by Pan and Radoz,10 the second
term on the right-hand side of eq. (1) vanishes for
low-pressure systems. To illustrate this, we evaluated the two terms of the right hand of eq. (1),

TABLE VII
Parameter Correlation with the Temperature for the Nonassociated Systems
in SLE at Atmospheric Pressure
System

D1

SRKLCVM (k)
1

4.0000  10
4
8
1.2300  10
2
11

1.8700  10
2
SRKWS (WS12)
1

8

1.2700  10
2
11

8.5000  10
3
PC-SAFT (K12)
1
0
8

1.0000  10
4
11
0

E1

F1

CC

2.7940  10
1

8.8830
1.3529  104

4.8196  101
1.6080  103

2.4405  103

0.8927
0.9909
0.9096

1.6786  103

1.2585  103

0.9874
0.9968

4.0891  10
1

1.8486  101

2.2300  10
1

0.9000
0.9200
0.9300

9.2445
6.5342
1.1600  10
3
1.0300  10
1
5.3000  10
4

Journal of Applied Polymer Science DOI 10.1002/app

1838

COSTA ET AL.

TABLE VIII
Average and Maximum Percentile Temperature Errors and Average Parameters
for SLE at High Pressure
PC-SAFT
Kij

System
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46

3.25
3.34
2.36
2.13
6.14
1.10
1.27
1.31
1.50
3.96
4.11
4.29
4.55
1.21
9.32
8.92
3.45
3.94
4.31
4.58
4.87
3.57
4.50
1.66
3.75
2.67




























SRKLCVM

AAD (%)

AMD (%)

UQij

AAD (%)

AMD (%)

0.103
0.200
0.047
0.075
0.160
0.700
0.761
0.598
0.527
0.118
0.205
0.274
0.240
0.745
0.472
0.554
0.077
0.114
0.453
0.525
0.445
0.536
0.065
0.476
0.065
0.248

0.264
0.445
0.096
0.151
0.321
1.277
1.367
1.198
1.125
0.175
0.304
0.460
0.369
0.998
0.914
0.932
0.116
0.200
0.714
0.833
0.920
1.473
0.097
1.167
0.096
0.456

0.169
0.177
0.346
0.344
0.052
0.355
0.226
0.255
0.272

0.122

0.059

0.080

0.134
0.352
0.345
0.403
0.025

0.006
0.106
0.063
0.048
0.171

0.031

0.149
0.425
CNA

0.999
0.809
0.319
0.563
0.929
1.030
0.658
0.906
0.655
0.785
1.163
0.950
1.137
1.037
0.667
0.357
0.414
1.539
0.832
0.831
0.997
0.844
1.143
1.358
2.043
CNA

1.492
1.213
0.591
0.443
0.700
2.476
3.704
2.494
2.148
0.880
0.864
1.080
1.040
3.643
2.083
1.876
0.235
0.366
1.572
2.148
1.833
1.599
0.268
1.554
0.067
CNA

3.100
2.350
0.850
0.687
1.110
4.630
7.210
5.980
4.310
1.280
1.270
1.680
1.600
5.680
4.870
3.330
0.356
0.626
3.080
3.600
3.290
3.870
0.400
3.670
0.239
CNA

2

10
10
2
10
2
10
2
10
2
10
2
10
2
10
2
10
2
10
2
10
2
10
2
10
2
10
2
10
3
10
3
10
2
10
2
10
2
10
2
10
2
10
2
10
2
10
2
10
2
10
2

CNA convergence not achieved.

calculation and the prediction of the temperature.

Table V shows the percentage absolute average deviation (AAD) and the percentage absolute maximum
deviation (AMD) calculated for the first 11 nonassociated systems at atmospheric pressure, as described
in Table I. The table also shows the method used to
solve eq. (1) according to the recommendation
described earlier. For these nonassociated systems
evaluated at atmospheric pressure, the results indicate
SRKLCVM as the most accurate model. It may be
rather surprising that SRKLCVM showed better
behavior than PC-SAFT, as the latter has been widely
and successfully applied to polymer systems. It seems
that at low pressure, the complexity of PC-SAFT, which
is more adequate for describing polymer systems, was
less important than the number of adjusted parameters.
Therefore, the models with more adjusted parameters
(two in the SRK models) presented smaller errors than

using the PEheptane system at 348 K and atmospheric pressure:


DHu
Tm

1 474:46

RTm
T
DvP
1:7288  10
4
RT
Therefore, we carried out atmospheric pressure
calculations, neglecting the second term on the righthand side of eq. (1).
Parameter estimation for nonassociated systems
For each system, the parameters were fitted for each
experimental datum, and an average parameter value
for each system was obtained by the division of its
sum by the number of experimental data. These average parameter values were applied once more to the

TABLE IX
PC-SAFT Parameter Correlation at High Pressure
System
PEpropane
PEethane

G1

H1

1

5.0070  10

5.9346  101

Journal of Applied Polymer Science DOI 10.1002/app

I1

8.9686  10

3.3825  102
1

J1

7

1.1680  10
3.2729  10
8

3.5030  10
3

1.6157  10
2

MODELING OF SOLIDLIQUID EQUILIBRIA

1839

TABLE X
SRKLCVM Parameter Correlation at High Pressure
System
Parameter
A2
B2
C2
D2
E2
F2
G2
H2

PEpropane
1.0574

2.7166
1.7504

2.2670

3.6472

2.2398
1.8620
1.6290










PEethane

10
102
101
10
6
10
1
10
1
10
7
10
1

1.9132

4.1530  10
5
1.0276  10
1

8.8248  103
1.8130  10
5

2.6840  10
2
1.8930  1015
0

PC-SAFT, where only one parameter was fitted. Moreover, the sophisticated mixing rules of SRK are based
on models that are very suitable for describing the lowpressure equilibrium of complex systems.
An interesting result shown in Table V is system
11 because, in this case, the solvent (paraffin wax)

was a mixture of unknown composition. Only its

chemical structure and MW were available, and SLE
calculations were carried out with only this information; notwithstanding, good results were obtained,
regardless of the model.
Parameter estimation for the associated systems
Table VI shows the average and maximum percentage temperature errors obtained for the associated
systems at atmospheric pressure, with the same
averaging procedure adopted over the estimated
parameters described earlier. All of the calculations
were carried out with only the SRK equation
because of the best performance achieved with the
nonassociated systems at atmospheric pressure. The
results show that the calculated data were in excellent agreement with the experimental data for all
selected systems, mainly when the complex chemical
nature of the solvents was highlighted.

TABLE XI
Average Percentile Temperature Error Obtained with the Correlated Parameters for
SLE at High Pressure
PC-SAFT

System
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

SRKLCVM

(Kij)ad

AAD
(%)

AMD
(%)

kad

 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
3
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
 10
2
a
a
a
1.41  10
2
4.14  10
2
4.03  10
2
4.13  10
2
4.13  10
2

0.137
0.258
0.199
0.154
0.159
3.299
2.840
2.445
1.509
2.944
3.280
3.602
3.989
0.752
0.607
0.577
1.493
0.449
0.628
1.019
1.315
0.980
0.716
a
a
a
1.789
0.389
0.809
0.177
0.375

0.401
0.412
0.294
0.245
0.322
4.632
4.260
3.682
2.653
3.099
3.559
3.937
4.287
1.452
1.524
0.948
1.606
0.637
1.037
1.593
1.770
2.469
0.770
a
a
a
2.573
0.458
1.202
0.200
0.574

1.700 
1.758 
3.438 
3.467 
5.200 
2.340 
2.795 
2.203 
3.948 
1.125 
2.191 
2.974 
6.038 
1.963 
2.082 
2.662 

8.683 
2.119 
1.320 
1.236 
2.019 
1.155 

4.027 
a
a
a
1.88 
5.17 
1.24 

2.29 
5.41 

3.32
3.26
2.27
2.21
6.14
2.48
2.47
2.42
2.30
2.59
2.57
2.52
2.45
1.40
1.20
7.86
4.14
4.16
4.03
4.06
4.10
4.01
4.13

10
1
10
1
10
1
10
1
10
2
10
1
10
1
10
1
10
1
10
2
10
1
10
1
10
2
10
1
10
1
10
1
10
2
10
1
10
1
10
1
10
1
10
1
10
2

10
1
10
2
10
1
10
1
10
2

(UQij)ad

AAD
(%)

AMD
(%)

9.150  10
1
9.101  10
1
4.338  10
1
4.290  10
1
9.376  10
1
9.108  10
1
9.283  10
1
8.814  10
1
5.681  10
1
6.597  10
1
9.194  10
1
9.332  10
1
8.203  10
1
9.326  10
1
9.112  10
1
5.981  10
1
4.773  10
1
9.785  10
1
9.713  10
1
9.349  10
1
6.200  10
1
9.816  10
1
1.014
a
a
a
9.29  10
1
8.92  10
1
9.68  10
1
8.20  10
1
1.09

1.486
1.155
0.518
0.530
0.699
3.628
3.256
2.875
2.801
2.713
5.631
7.307
3.756
4.818
5.074
2.955
2.502
4.590
1.741
2.329
3.061
1.838
0.332
a
a
a
7.267
0.476
2.170
6.596
8.981

3.070
2.390
0.927
0.759
1.110
6.940
5.640
6.330
5.360
4.070
7.210
9.420
5.290
8.250
7.550
5.530
2.810
5.380
2.970
4.530
5.740
5.030
0.487
a
a
a
9.730
0.719
3.640
8.120
9.500

a
The experimental polymer concentration was insufficient to obtain a parameter
correlation.

Journal of Applied Polymer Science DOI 10.1002/app

1840

COSTA ET AL.

Figure 1 Solidfluid phase transitions of the PEpentane

system: MW 121,000 and WT 5 (system 44).

Predictive correlations
SLE calculations can be very useful to predict equilibrium Tm values for compositions where experimental data are not available. Therefore, for systems
1, 8, and 11, which presented the largest numbers
of experimental data, parameter correlations as a
function of temperature were obtained with standard least-squares procedures applied to the optimal
interaction parameter found earlier. The parameter
functional forms are presented in eqs. (17)(19), and
their coefficients and correlation coefficients (CCs)
for each model are shown in Table VII (the UNI-

Figure 3 Solidfluid phase transitions of the PEethane

system: (a) MW 7000 and WT 2 and 20 (systems 21
and 22, respectively) and (b) MW 23,625 and WT 2
and 20 (systems 23 and 24, respectively). [Color figure
can be viewed in the online issue, which is available at
wileyonlinelibrary.com.]

QUAC UQ12 parameters, in general, did not correlate well with the temperature):
k D1 T2 E1 T F1
WS12 D1 T 2 E1 T F1

(17)
(18)

K12 D1 T 2 E1 T F1

(19)

where D1, E1, and F1 are constants. For the SRKLCVM

model, parameter correlations (ks) were quadratically
dependent on the temperature and gave good results.
The same observations were made for the SRKWS
model (WS12 parameter) with systems 8 and 11, whereas
for system 1, it was not possible to obtain a good parameter correlation because of the poor results obtained, as
shown in Table V. The results for the PC-SAFT model
show good simple correlations for the adjusted K12 parameter. For nonassociated systems, insufficient experimental data were available to develop good correlations.
RESULTS AND DISCUSSION FOR SLE
AT HIGH PRESSURE
Parameter estimation
Figure 2 Solidfluid phase transitions of the PEethylene
system: (a) MW 7000 and WT 2 (system 45) and (b)
MW 23,625 and WT 2 (system 46).
Journal of Applied Polymer Science DOI 10.1002/app

The three models, PC-SAFT, SRKWS, and SRK

LCVM, were applied to the systems and are

MODELING OF SOLIDLIQUID EQUILIBRIA

1841

Figure 4 Solidfluid phase transitions of the PEethane

system: MW 52,000 and WT 20 (system 25).

presented in Table III for high-pressure SLE calculation. The same parameter averaging procedure
described previously was used, but for SRKWS, the
parameter range for each evaluated system was too
large and, therefore, did not allow the use of average
parameter values. This was unexpected because the
WS mixing rule was originally developed to extend
the cubic EOS for high-pressure applications. Few
differences were observed among the parameter
Figure 6 Solidfluid phase transitions of the PEpropane
system: (a) MW 13,600 and WT 0.25 and 2 (systems
30 and 31, respectively) and (b) MW 13,600 and WT 5
and 10 (systems 32 and 33, respectively). [Color figure
can be viewed in the online issue, which is available at
wileyonlinelibrary.com.]

values at moderate pressures, but at the high-pressure level, it was impossible to achieve good results,
regardless of the estimated initial parameters; therefore, we did not take the SRKWS results further
(see the Appendix for these results).
Table VIII summarizes the analyses of the results
from the PC-SAFT and SRKLCVM EOSs. Although
average parameters were used, the corresponding
AAD was extremely low. The calculated results presented in Table VIII show that LCVM gave good
results when the GE mixing rule model and the
equation solving method were efficient, although
this mixing rule was not originally developed for
high-pressure SLE. Another aspect worth highlighting is that only one parameter was sufficient to give
good results with the PC-SAFT EOS.
Predictive correlations
Figure 5 Solidfluid phase transitions of the PEpropane
system: (a) MW 7000 and WT 2 and 2.7 (systems 26
and 27, respectively) and (b) MW 7000 and WT 7.4
and 17.5 (systems 28 and 29, respectively). [Color figure
can be viewed in the online issue, which is available at
wileyonlinelibrary.com.]

From a practical viewpoint, it is important for an

EOS to be robust with regard to extrapolation, as is
shown in the section on Interpolating and Extrapolating Parameter Correlations for the Systems Not
Used in the Parameter Estimation. Nevertheless, even
if one does not intend to extrapolate the application of
Journal of Applied Polymer Science DOI 10.1002/app

1842

COSTA ET AL.

were MW and WT. When data from other alkane

solvents were available, we could even try a correlation with the number of carbons or a group contribution approach.
The same experimental data were used to obtain
correlations for the LCVM parameters k and UQij.
The resulting expressions are shown in eqs. (21) and
(22), and their corresponding coefficients are shown
in Table X:


kMW; WT A2








WT 3
WT
WT
C2
B2
100:0
100:0
100:0

D2 MW 21



E2 100
WT
H2  ln
UQij MW; WT
(22)
100
F2 G2 MW
WT

where A2, B2, C2, D2, E2, F2, G2, and H2 are constants.
Table XI presents the adjusted model parameters
and average and maximum percentage temperature
errors for SLE. The developed correlations obtained
from a wide data set were applied to each system,
and the average percentage temperature error was
calculated. With this calculation, we could analyze
Figure 7 Solidfluid phase transitions of the PEpropane
system: (a) v 23,625 and WT 2 and 6.9 (systems 34
and 35, respectively) and (b) MW 23,625 and WT 17
(system 36). [Color figure can be viewed in the online
issue, which is available at wileyonlinelibrary.com.]

the model or doubt its predictive capabilities, parameter correlations are a very suitable method for modeling interpolations. Therefore, we looked for predictive
parameter correlations for the PC-SAFT and SRK
LCVM models.
The fitted parameters were put into groups according to the selected polymersolvent systems (PE
ethane and PEpropane), and all of the respective
experimental data presented in Table III were used.
Then, a correlation was generated with MW and percentage weight polymer concentration (WT) as independent variables. Equation (20) shows the PCSAFT parameter correlation for the PEpropane
(96 experimental data points) and PEethane (23
experimental data points) systems. The numerical
values of the constants are shown in Table IX:
Kij MW; WT

WT
G1 100:0

H1 I1  MW2

J1 ln MW

(20)

where G1, H1, I1, and J1 are constants. It is important

to note the large MW range considered in this adjustment. As a result of these parameter correlations, the
calculation became a completely predictive task
because the only properties that had to be known
Journal of Applied Polymer Science DOI 10.1002/app

Figure 8 Solidfluid phase transitions of the PEpropane

system: (a) MW 42,900 and WT 0.06 and 3 (systems
37 and 38, respectively) and (b) MW 52,000 and WT
2, 7.3, and 17 (systems 39, 40, and 41, respectively). [Color
figure can be viewed in the online issue, which is available
at wileyonlinelibrary.com.]

MODELING OF SOLIDLIQUID EQUILIBRIA

1843

this model was not originally developed for SLE calculation, these errors could be considered small. For
the PEethane systems (2125), the PC-SAFT EOS
once again showed better results than the SRK
LCVM EOS; the highest AAD percentage was 0.258
for system 22, which also showed an AMD percentage equal to 0.412, although the performance of the
SRKLCVM EOS was quite good, with a maximum
AAD percentage equal to 1.486 and a AMD percentage equal to 3.070 (both for system 21).

Graphic analysis of the solvent, pressure range,

polymer MW, and composition effects on the
model performance
Calculations with the average estimated parameters

Figure 9 Solidfluid phase transitions of the PEpropane

system: (a) MW 59,300 and WT 2 (system 42) and (b)
MW 119,600 and WT 7.5 (system 43).

the behavior in each of the systems investigated and,

to some extent, evaluate the fitting quality. The
results demonstrate good performance of the PCSAFT equation, and it was worthwhile to highlight
the performance in describing systems 4751, which
were not used in the fitting procedure. With regard
to systems 4446 (pentane and ethylene systems),
there were no results because various data of polymer concentration and/or MW were not at our
disposal, so a correlation could not be developed.
We performed a detailed assay by analyzing the
AAD percentage values for each system. For systems
2643 (PEpropane systems used on correlation
development), with the PC-SAFT equation, the maximum AAD percentage with fitted parameters was
3.989 for system 33, and the AMD percentage was
4.632 for system 26. Even systems with a high polymer concentration and large pressure and temperature ranges, such as systems 41 and 42, exhibited
ADD percentages equal to 1.315 and 0.980 and AMD
percentages equal to 1.770 and 2.469, respectively.
Comparing the SRKLCVM and PC-SAFT models,
we observed that SRKLCVM EOS gave a larger
AAD percentage, and its maximum value was 7.307;
this was achieved with system 32, which also
showed an AMD percentage equal to 9.420. Because

To further evaluate the average parameters estimated in the Parameter Estimation section, some of
these parameters were applied to the corresponding
system to observe the influence of the solvent, pressure range, and polymer MW on the performance
of the models. The PEpentane and PEethylene
systems were selected for these calculations as only
one MW and one polymer concentration data point
were available for the first system and only one
polymer concentration data point was available for

Figure 10 Solidfluid phase transitions of the PEpropane

system: (a) magnification of Figure 6(a) and (b) magnification of Figure 6(b). [Color figure can be viewed in the
online issue, which is available at wileyonlinelibrary.com.]
Journal of Applied Polymer Science DOI 10.1002/app

1844

Figure 11 Solidfluid phase transitions of the PEpropane

system: (a) magnification of Figure 8(a) and (b) magnification of Figure 9(b). [Color figure can be viewed in the
online issue, which is available at wileyonlinelibrary.com.]

COSTA ET AL.

Figure 3(a) shows the behavior of both models

with a low polymer MW. The simulation results
indicate excellent performance with the PC-SAFT
model with parameters from the developed correlation. With LCVMSRK, although we observed a
reasonable performance at the lower bound of the
considered pressure range, the overall behavior was
poor because this model failed in the qualitative
description of the phenomenon. The same general
observations applied to Figures 3(b) and 4, and
therefore, we observed that for this PEethane system, the model performances were insensitive with
regard to the polymer MW.
Figure 5(a) shows the excellent behavior of the
PC-SAFT equation in describing this PEpropane
system. Despite the small difference between the
weight fractions of the two data sets, the model
captured this difference for the whole high-pressure
range. The performance of the LCVM was not good,
with a systematic slope mismatch that was also
observed throughout virtually all of the other simulations, shown in Figures 59. Figure 5(b), with the
same polymer MW and higher concentrations shown
in Figure 5(a), also shows good simulation results
with the PC-SAFT equation for both polymer concentrations. With SRKLCVM, the error between the

the second system. It was, therefore, not possible to

develop a parameter correlation, as stated earlier
and as shown in Table III.
Figure 1 shows the solidfluid phase transition
calculations for the PEpentane system. The results
indicate that there was no difference between the
experimental data and those of the PC-SAFT equation
and the LCVMSRK model. The ethylene system
results are shown in Figure 2. Figure 2(a) shows similar performances for the two models. Figure 2(b)
shows the excellent performance of the PC-SAFT
model and the good performance of the cubic LCVM
SRK EOS.
Calculations with the predictive correlations
To further evaluate the fitting procedures developed
in the Predictive Correlations section, the developed
parameter correlations were applied to the ethane
and propane systems. To show the influence of the
selected model, the polymer MW, and the polymer
weight fraction, Figures 3 and 4 (ethane systems)
and Figures 59 (propane systems) show the solid
fluid phase transitions of the analyzed systems.
Figures 10 and 11 show magnifications of some of
the graphs in Figures 59 to emphasize their details.
Journal of Applied Polymer Science DOI 10.1002/app

Figure 12 Solidfluid phase transitions of the PEpropane systems not used in the parameter estimation: (a)
MW 23,625 and WT 1.9 (system 47) and (b) MW
42,900 and WT 1 (system 48).

MODELING OF SOLIDLIQUID EQUILIBRIA

1845

Figure 13 Solidfluid phase transitions of the PEpropane systems not used in the parameter estimation: (a)
MW 52,000 and WT 1.9 (system 49) and (b) MW
119,600 and WT 0.25 and 3.5 (systems 50 and 51, respectively). [Color figure can be viewed in the online issue,
which is available at wileyonlinelibrary.com.]

experimental data and the simulation results at the

highest concentration increased as the pressure
decreased and the opposite behavior was observed
at the lowest polymer concentration. As shown
Figures 6(a) and 10(a) (the magnification), with a

higher polymer MW, the PC-SAFT equation was not

able to precisely distinguish the difference between
the polymer weight fractions. This models low sensitivity with regard to weight fraction is also shown
in Figure 6(b) [with the same polymer MW and
higher concentrations than those shown in Figs. 6(a)
and 10(b) (the magnification) but, in this case, in accordance with experimental data]. As the polymer
MW increased, as shown in Figure 7(a,b), we
observed excellent results with the PC-SAFT simulations. With the LCVMSRK model, there was an
increase in the simulation error as the pressure
increased. This same trend was observed with other
polymer MWs and other polymer concentrations.
Whereas Figures 57 show low- and medium-MW
polymer data, Figures 8 and 9 show the results for
higher MW systems. Figures 8(a) and 11(a) (the
magnification) show excellent results for the PCSAFT model, especially for the 3% polymer weight
fraction. There were different trends with regard to
the LCVMSRK model: an overestimation with the
smallest polymer weight fraction and an underestimation with the highest one. Figure 8(b) shows
the excellent performance of the PC-SAFT model for
all three polymer concentrations. As shown in Figure 9(a), there was an increasing difference between
the experimental and simulated temperatures with
increasing pressure for both correlation models.
A remarkable model mismatch was observed with
the LCVMSRK equation, shown in Figure 9(a),
although, as shown in Figures 9(b) and 11(b) (the
magnification), with the highest investigated MW,
we observed excellent behavior with this model.
In summary, Figures 511 show that PC-SAFT had
a very good overall performance, clearly superior to
that of LCVMSRK. Moreover, the PC-SAFT model
performed better at higher polymer concentrations,

TABLE XII
PEPropane Temperature Calculation with the Correlated Parameters
System

MW

Polymer
concentration

Texp
(K)

Tcalc
(PC-SAFT)

Tcalc
(SRKLCVM)

Pressure
(atm)

47

23,625

1.9

48

42,900

49

52,000

1.9

50

119,600

0.25

51

119,600

3.5

365.307
367.535
369.763
371.377
385.954
385.961
404.734
402.375
399.024
394.308
393.067
389.020
389.230
389.858
390.858

369.428
373.081
376.970
380.931
387.723
387.199
404.988
399.495
394.226
390.482
388.754
389.799
388.625
389.166
388.614

382.576
391.447
399.756
407.497
385.054
383.185
415.196
404.676
393.359
383.912
378.755
420.614
408.971
356.873
353.717

911.628
1110.594
1309.561
1504.910
721.723
688.462
1503.423
1250.667
997.963
803.969
705.081
690.245
613.038
687.050
652.734

Journal of Applied Polymer Science DOI 10.1002/app

1846

COSTA ET AL.

regardless of the polymer chain length. A slightly

better performance of the PC-SAFT model in describing ethane systems (Figs. 3 and 4) than in describing
propane systems (Figs. 511) was also observed.
Interpolating and extrapolating parameter
correlations for the systems not used
in the parameter estimation
Figures 12 and 13 present results with systems not
used in the fitting procedure, and Table XII details
these results and shows the calculated and experimental values. With these results, it was, therefore,
possible to evaluate the fitting performance. The
results of systems 4748 and 5051 represent weight
polymer concentration extrapolations with regard to
the parameter correlations developed, whereas the
results of system 49 represent an interpolation calculation. The simulation results show the excellent
behavior of the PC-SAFT equation. The same evaluation could not be carried out with the SRKLCVM
model, except for the simulation of the 42,900 polymer MW system, where the results were considered
excellent. A number of factors must have influenced
these results. For example, with the lowest MW, a
higher pressure level was observed, and given this,
the mixing rule was unable to cope with the equilibrium description. Even with the PC-SAFT equation,
there was an increased simulation error with
increasing pressure for this system. When the results
with the 52,000 polymer MW system were analyzed,
no systematic behavior with regard to pressure
variations was observed. On the other hand, despite
the low pressure level with the 119,600 polymer MW
system, the SRKLCVM simulation errors were considered relevant.
CONCLUSIONS
SLEs for PE and PP solutions with different solvents
were modeled with the SRK cubic EOS (with LCVM
and WS mixing rules) and PC-SAFT EOS. The focus
of this study was to compare the performances of
these models in describing SLE phase behavior. This
new approach is different from those usually found
in the literature, in which the effects of crystallinity
and MW on SLE were investigated.
Regardless of the model used to simulate SLE, a
key variable in the analysis of the results is the pressure level. As pointed out by Pan and Radoz,10 the
second term on the right-hand side of eq. (1)
vanishes for low-pressure systems. So with these
simulations, the pressure effect is restricted to the /Lp
and /0p , which are related to the selected EOS. On
the other hand, high-pressure conditions are quite
difficult to model because the pressure influences
both sides of the SLE equation.
Journal of Applied Polymer Science DOI 10.1002/app

The PC-SAFT and SRK EOSs, with two different

mixing rules (LCVM and WS), were used and compared. An experimental database at atmospheric
pressure containing 20 polymersolvent systems was
considered. The influence of the method used for
solving the SLE equation to obtain Tm was also
investigated. The calculated results indicate SRK
LCVM as the best model at atmospheric pressure
conditions. However, it is important to highlight
that with the SRKLCVM model, two adjusted
parameters were used in the simulation: one related
to the mixing rule and another related to the GE
model selected. When the predictive ability of the
SRKLCVM model was evaluated, we observed that
it was not easy to generate a good correlation of the
GE parameter (WT12) with the temperature. This
difficulty was not observed with the PC-SAFT
EOS because the only fitted parameter correlated
very well with the temperature for all of the systems
analyzed.
Another database containing 31 polymersolvent
systems at high pressure was used, and under these
conditions, the sensitivity of the EOS with regard to
pressure was noticeable as the pressure influence on
the right-hand side of the SLE equation had to be
taken into consideration. Surprisingly, the SRKWS
model was not appropriate for high-pressure SLE
calculation in these systems. Interaction parameter
correlations as a function of MW and polymer concentration were developed for the PC-SAFT and
SRKLCVM EOSs. The PC-SAFT EOS with parameter correlation provided the best performance. The
results were excellent for the PC-SAFT equation
with various hydrocarbon solvents, especially for
higher polymer concentrations, regardless of the
polymer chain length or the pressure range. The predictive abilities of the PC-SAFT and SRKLCVM
EOSs were evaluated by the testing of five systems
not used in the fitting procedure. Once more, PCSAFT and the developed parameter correlation presented good results and showed suitable interpolating and extrapolating features.

NOMENCLATURE
a
A
A2, B2, C2,
and D2
Ai, Bi,
and Ci
AM and Av
AAD
AMD
b

parameter of the SRK EOS

Helmholtz free energy per number of
molecules
constants [eq. (21)]
group volume temperature constants in
the GCVOL method
parameters of the LCVM mixing rule
percentage absolute average deviation
percentage absolute maximum deviation
parameter of the SRK EOS

MODELING OF SOLIDLIQUID EQUILIBRIA

BV
c
CC
d
D1, E1,
and F1
E2, F2, G2,
and H2
entr-FV
EOS
FV
G1, H1, I1,
and J1
GCVOL
GE
DHu

1847

Bogdanic and Vidal

crystallinity fraction
correlation coefficient
numerical constant of the WS mixing rule
[eq. (6)]
constants [eqs. (17)(19)]

WT
x
X

constants [eq. (22)]

b
d
e
c
/
qa
qc
r

entropic free volume

equation of state
free volume
constants [eq. (20)]

group contribution volume

Gibbs free energy
enthalpy of melting per moles of crystal
units
k
Boltzmann constant
adjustable binary interaction parameter
Kij
of the PC-SAFT EOS [eq. (4)]
LCVM
linear combination of the Vidal and
Michelsen mixing rules
MW
molecular weight
N
total number of molecules or number of
experimental data
number of group i in the GCVOL method
ni
ncomp
number of components
NR
NewtonRaphson
nseg
number of segments
OF
objective function
P
pressure
PC-SAFT
perturbed-chain statistical associating fluid
theory
PE
polyethylene
PP
polypropylene
R
gas constant
Rf
Regulafalsi
SLE
solidliquid equilibrium
SRK
SoaveRedlichKwong
T
temperature
melting temperature
Tm
u
number of monomer units in the
polymer backbone
UNIQUAC Universal quasi-chemical
interaction parameter of the LCVM mixing
UQij
rule [eq. (22)]
v
molar volume
Dv
polymer volume change
van der Waals volume
VW,i
WS
WongSandler
binary interaction parameter of the WS
WSij
mixing rule [eq. (8)]

percentage weight polymer concentration

molar fraction (solubility)
molar fraction of the segment

Greek letters
a and k

parameters of the LCVM mixing rule

[eq. (9)]
FV fraction
number of segments in the component
dispersion energy parameter
molar activity coefficient
fugacity coefficient
density of an amorphous polymer
density of a crystalline polymer
segment diameter

Superscripts
0
ad
assoc
cal
disp
exp
hc
L
res

pure liquid polymer

adjusted
association contribution
calculated
dispersion contribution
experimental
hard chain contribution
polymer in solution
residual

Subscripts
i
component or group i
j
component or group j
k, m, and n segments
p
polymer
W
van der Waals

APPENDIX
SLE calculations at high pressure with
the SRKWS model
Table XIII shows the PEpropane temperature calculations with the SRKWS model with adjusted parameters
for each set of MW/polymer concentration data, that is,
without the correlation of the adjusted parameters with
MW and polymer concentration. It was impossible to
obtain a parameter correlation with this model because of
the large scattering of fitted parameters. For this reason,
the results are presented in a different fashion, only
through tabulated values instead of the graphic displays
presented for the previous results. The simulation results
showed a systematic performance: higher deviations from
the experimental data were observed with increased
polymer MW.

Journal of Applied Polymer Science DOI 10.1002/app

1848

COSTA ET AL.

TABLE XIII
PEPropane Temperature Calculation with the SRKWS
Model Without Correlation of the Adjusted Parameters
with MW and Polymer Concentration

MW
7000

Polymer
concentration
(system)
2 (26)

2.7 (27)

7.4 (28)

17.5 (29)

13,600

0.25 (30)

2 (31)

5 (32)

10 (33)

23,625

2 (34)

6.9 (35)

17 (36)

Texp
(K)

Tcalc
(SRKWS)

Pressure
(atm)

371.453
369.471
367.709
365.066
364.185
364.185
364.691
364.403
365.267
367.283
369.300
372.757
365.843
366.419
367.572
369.300
371.604
375.061
366.995
367.283
369.876
372.181
378.806
384.074
383.926
383.960
387.000
387.015
387.060
387.857
388.062
388.091
387.946
388.062
387.917
387.948
388.140
365.066
364.845
367.488
369.471
372.995
366.285
368.000
369.142
369.428
372.000
374.285
366.857
369.714
371.428
373.714
376.285
365.116
366.511
369.302
371.162
373.488
376.279

371.452
367.485
360.629
361.385
364.184
364.184
364.689
364.401
365.264
359.038
368.199
368.199
365.840
366.417
367.566
364.260
364.260
361.038
366.993
367.279
369.874
368.383
375.979
383.483
383.537
382.629
383.172
382.135
385.663
379.446
381.779
385.752
384.226
385.054
384.408
384.200
384.348
365.062
364.844
360.195
361.755
361.755
366.280
367.995
367.995
367.435
367.435
363.385
366.854
369.154
369.154
367.957
367.957
365.113
366.509
367.169
370.889
367.362
367.362

1509.604
1310.734
1102.824
903.954
705.084
605.649
401.219
606.097
802.439
1015.853
1203.658
1502.439
597.560
700.000
802.439
1007.317
1203.658
1502.439
503.658
606.097
810.975
1007.317
1502.439
587.995
525.488
448.704
688.126
652.412
550.628
725.664
654.244
588.174
520.311
656.030
586.381
516.739
475.676
397.740
596.610
1003.389
1202.259
1500.564
804.790
897.005
997.604
1098.203
1299.401
1500.598
704.191
905.389
1106.586
1307.784
1500.598
591.751
713.452
912.655
1111.756
1310.908
1510.111

Journal of Applied Polymer Science DOI 10.1002/app

TABLE XIII. Continued

MW

Polymer
concentration
(system)

42,900

0.06 (37)

3 (38)

52,000

2 (39)

7.3 (40)

17 (41)

59,300

119,600

2 (42)

7.5 (43)

Texp
(K)

Tcalc
(SRKWS)

Pressure
(atm)

383.666
383.166
382.000
387.000
387.000
387.000
387.333
404.722
401.481
400.555
398.703
395.925
394.537
393.606
395.901
398.483
401.065
405.368
405.116
402.325
398.604
395.348
393.488
389.444
390.370
391.296
392.685
393.611
395.022
393.920
392.599
392.190
391.042
391.055

382.358
382.051
381.720
385.912
383.806
383.004
384.225
385.869
382.067
368.725
368.725
381.743
382.642
383.860
380.073
380.073
380.073
380.073
385.942
385.591
385.448
385.562
385.438
378.099
381.555
381.555
368.546
368.546
353.985
353.985
364.008
378.537
376.355
385.051

686.780
650.861
616.623
702.307
668.119
651.025
628.817
1513.198
1313.945
1114.947
1015.295
915.541
804.890
667.103
803.139
1007.088
1253.462
1508.529
1513.298
1258.846
1015.341
805.036
705.383
815.377
914.927
1014.477
1114.078
1313.075
1500.564
1310.734
1102.824
735.454
701.548
667.599

References
1. Pontes, K. V.; Maciel, R.; Embirucu, M. J Appl Polym Sci 2008,
109, 2176.
2. Pontes, K. V.; Maciel, R.; Embirucu, M.; Hartwich, A.;
Marquardt, W. AIChE J 2008, 54, 2346.
3. Embirucu, M.; Pontes, K.; Lima, E. L.; Pinto, J. C. Macromol
React Eng 2008, 2, 161.
4. Harismiadis, V. I.; Tassios, D. P. Ind Eng Chem Res 1996, 35,
4667.
5. Fontes, C. H. O.; Embirucu, M. Comput Chem Eng 2001, 25,
191.
6. Costa, G. M. N.; Guerrieri, Y.; Kislansky, S.; Pessoa, F. L. P.;
de Melo, S. A. B. V.; Embirucu, M. Ind Eng Chem Res 2009,
48, 8613.
7. Embirucu, M.; Fontes, C. Chem Eng Sci 2006, 61, 5754.
8. Tumakaka, F.; Prikhodko, I. V.; Sadowski, G. Fluid Phase
Equilib 2007, 260, 98.
9. Abbas, S.; Mukherjee, R.; De, S.; Ganguly, S. Chem Eng Process 2004, 43, 1449.
10. Pan, C.; Radoz, M. Fluid Phase Equilib 1999, 155, 57.
11. Poling, B.; Prausnitz, J.; OConnell, J. The Properties of Gases
and Liquids, 5th ed., McGraw-Hill: New York, 2001.
12. Boukouvalas, C.; Spiliots, N.; Coutsikos, P.; Tzouvaras, N.;
Tassios, D. P. Fluid Phase Equilib 1994, 92, 75.
13. Wong, D. S. H.; Sandler, S. I. AIChE J 1992, 38, 671.

MODELING OF SOLIDLIQUID EQUILIBRIA

14. Soave, G. Chem Eng Sci 1972, 27, 1197.

15. Costa, G. M. N.; Dias, T.; Cardoso, M.; Guerrieri, Y.; Pessoa,
F. L. P.; de Melo, S. A. B. V.; Embirucu, M. Fluid Phase Equilib
2008, 267, 140.
16. Henrici, P. Elements of Numerical Analysis; Wiley: New York,
1964.
17. Ralston, A.; Rabinowitz, P. A First Course in Numerical Analysis, 2nd ed.; McGraw-Hill: New York, 1978.
18. Atkinson, K. E. An Introduction to Numerical Analysis, 2nd
ed.; Wiley: New York, 1989.
19. Stoer, J.; Bulirsch, R. Introduction to Numerical Analysis, 3rd
ed.; Springer-Verlag: New York, 2002.
20. Gross, J.; Sadowski, G. Ind Eng Chem Res 2001, 40,
1244.
21. Gross, J.; Sadowski, G. Ind Eng Chem Res 2002, 41, 5510.
22. Bogdanic, G.; Vidal, J. Fluid Phase Equilib 2000, 173, 241.
23. Richards, R. B. Trans Faraday Soc 1946, 42, 10.
24. Nakajima, A.; Fujiwara, H.; Hamada, F. J Polym Sci Part A-1:
Polym Chem 1966, 4, 507.

1849

25. Kim, S. S.; Lloyd, D. R. Polymer 1992, 33, 1036.

26. Nakajima, A.; Fujiwara, H. J Polym Sci Part A-1: Polym Chem
1968, 6, 723.
27. Agarwal, R.; Prasad, D.; Maity, S.; Gayen, K.; Ganguly, S.
J Chem Eng Jpn 2004, 37, 1427.
28. Gregorowicz, J. J. Supercrit Fluids 2007, 43, 357.
29. Condo, P. D., Jr.; Colman, E. J.; Ehrlich, P. Macromolecules
1992, 25, 750.
30. Zhang, W.; Dindar, C.; Bayraktar, Z.; Kiran, E. J Appl Polym
Sci 2003, 89, 2201.
31. Van Krevelen, D. W. Properties of Polymers, 3rd ed.; Elsevier
Scientific: Amsterdam, 1990.
32. Kontogeorgis, G.; Harismiadis, V. I.; Fredeslund, A.; Tassios,
D. P. Fluid Phase Equilib 1994, 96, 65.
33. Elbro, H. S.; Fredesnlund, A.; Rasmussen, P. Ind Eng Chem
Res 1991, 30, 2576.
34. Bondi, A. Physical Properties of Molecular Crystals, Liquids
and Glasses; Wiley: New York, 1968.
35. Nelder, J. A.; Mead, R. Comput J. 1964, 7, 308.

Journal of Applied Polymer Science DOI 10.1002/app