Available online at www.sciencedirect.

com

Fluid Phase Equilibria 261 (2007) 248257

Application of PC-SAFT to glycol containing systems PC-SAFT

towards a predictive approach
Andreas Grenner, Georgios M. Kontogeorgis ,
Nicolas von Solms, Michael L. Michelsen
Centre for Phase Equilibria and Separation Processes (IVC-SEP), Department of Chemical Engineering,
Technical University of Denmark, DK-2800 Lyngby, Denmark
Received 9 February 2007; received in revised form 17 April 2007; accepted 19 April 2007
Available online 27 April 2007

Abstract
For equations of state of the SAFT type a major limitation is the procedure of obtaining pure compound parameters using saturated vapor pressure
and liquid density data. However, for complex compounds such data are often not available. One solution is to develop a group contribution scheme
for estimating pure compound parameters from low molecular weight data and extrapolate to complex compounds. For associating compounds this
is not trivial since the two parameters for association (association energy and association volume) need to be fixed for a group. In this work, which
focuses on glycols, new general pure compound parameters were obtained for PC-SAFT which are able to perform well for both vaporliquid
equilibria (VLE) and liquidliquid equilibria (LLE). Linear trends of non-association parameters were obtained with respect to the molar mass.
However, identical values for the association parameters were used for all glycol oligomers. This makes it possible to predict the pure compound
parameters of other oligomers. With the new estimated parameters the simplified PC-SAFT equation of state has been applied for correlation
and prediction of VLE and LLE in mixtures containing glycol oligomers + water, hydrocarbons, aromatic hydrocarbon, methane, N2 and CO2 . To
improve correlations for mixtures containing glycol and aromatic hydrocarbons solvation was explicitly accounted for.
2007 Elsevier B.V. All rights reserved.
Keywords: Modeling; Glycol; PC-SAFT; Liquidliquid equilibria; Vaporliquid equilibria

1. Introduction
The final aim of this project is to obtain a thermodynamic
model which can be used in development of complex products
such as pharmaceuticals, polymers, detergents or food ingredi-

Abbreviations:
AAD, absolute average deviation AAD(%) =

exptl
100/np ((Zicalcd Zi )/Zicalcd ) where np is the number of data
points and Z represents P or x; CPA, cubic plus association equation of
state; DEG, diethylene glycol; EoS, equation of state; VLE, vaporliquid
equilibria; LLE, liquidliquid equilibria; MEG, monoethylene glycol;
PC-SAFT, perturbed chain-SAFT
equation of state; RMS, root mean

P(%) = 100

square,

1/(np 1)



1/(np 1)

2
(ycalcd yexptl )

(Picalcd P exptl )/P exptl

2

,y=

where np is the number of data points;

SAFT, statistical associating fluid theory; TEG, triethylene glycol; TeEG,

tetraethylene glycol
Corresponding author. Tel.: +45 45 25 28 59; fax: +45 45 88 22 58.
E-mail address: gk@kt.dtu.dk (G.M. Kontogeorgis).
0378-3812/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.fluid.2007.04.025

ents. This type of modeling is challenging due to the complexity

of the molecules. The perturbed chain-statistical associating
fluid theory (PC-SAFT) of Gross and Sadowski [1] in the simplified form of von Solms et al. [2] is employed to model
complex phase equilibria for self-associating and solvating systems of glycol oligomers. New estimated parameters for glycol
oligomers are presented which use the same values for the two
pure compound association parameters. The purpose was to test
these parameters for their capabilities to correlate and predict
mixture properties of glycol-containing systems. This is part of
a bigger project, the final aim is to develop a group contribution scheme for PC-SAFT. The justification for this approach is
that pure compound parameters for complex compounds cannot
always be obtained by fitting vapor pressure and liquid density data as such data may not be available. If satisfying results
are obtained, the next step is to test real group contribution
results.
The use of glycols in the petroleum as well as chemical
industry has increased in the last years due to environmental

A. Grenner et al. / Fluid Phase Equilibria 261 (2007) 248257

pressure. Glycols are increasingly replacing methanol as

hydrate inhibitors. Monoethylene glycol (MEG) is mainly
used as antifreeze and gas hydrate inhibitor of natural
gas. It is of economic importance to know the amount of
inhibitor needed based on accurate thermodynamic modeling
of oilgaswaterinhibitor mixtures. Diethylene glycole (DEG)
and triethylene glycol (TEG) are applied as drying agents for
natural gas. Tetraethylene glycol (TeEG) is employed in solvent
extraction processes to obtain high purity aromatic hydrocarbons and 1,2-propylene glycol (PG) is used in the production of
resins as well as in the food industry. For all these processes it is
important to have a thermodynamic model which can describe
the phase equilibria of these glycol-containing systems with high
accuracy.
Relatively few studies exist for this type of compound with
the different SAFT variants. Li and Englezos [3] present VLE
of some glycol systems with the SAFT version of Huang and
Radosz [4]. The combining rules for systems with solvation were
the arithmetic mean for the association volume and the geometric mean for the association energy. Furthermore, an adjustable
parameter in the association energy term of the solvating mixtures is introduced.
Ai Bi + Aj Bj
,
2

= (Ai Bi Aj Bj )(1 kijAB )

Ai Bj =
Ai Bj

(1)

Pedrosa et al. [5] employ the soft-SAFT of Blas and Vega

[6] also for VLE of glycol oligomers but no LLE calculations
are shown. Soft-SAFT uses LennardJones as reference fluid
while PC-SAFT uses a hard-chain reference fluid. In addition,
quadrupole interactions are accounted for directly by a separate
term in soft-SAFT. Similar to other SAFT versions the generalized LorentzBerthelot combining rules are applied, with an
adjustable binary parameter in each rule; ij in the combining
rule for the segment diameter and ij in the combining rule for
the dispersion energy:
ij = ij

ii + jj
,
2

ij = ij ii jj

(2)

Pedrosa et al. [5] use the same values of association parameters

for all ethylene oligomers. Extensive studies have been done
with the cubic-plus-association (CPA) EoS by Derawi et al. [7,8]
and by Folas et al. [9,10]. These authors also presents new LLE
data for glycol systems [11,12]. The CPA EoS [13] combines the
SoaveRedlichKwong EoS with an association term essentially
equivalent to SAFT. Comparisons to these previous studies are
presented in this work.
The presence of hydrogen bonds in the systems studied,
means that it is essential to take into account the association
contribution. This is done using a contribution obtained from
Wertheims theory [1417] and is common to all SAFT-type
equations. All glycol oligomers and water were considered to
have four associating sites identical to the 4C scheme, representing two hydrogen donor sites and the two hydrogen acceptor
sites, following the terminology of Huang and Radosz [4].

249

2. Theory
The simplified PC-SAFT [2] was developed with the goal
to reduce the computational and programming effort without
essentially disturbing the performance of the model. Thus, the
simplified PC-SAFT [2] is identical to the original PC-SAFT
[1] in case of single non-associating compounds, but not for
multicomponent mixtures in general and not for pure associating
compounds.
The simplified PC-SAFT [2] uses simpler mixing rules taking
into account that the segment diameters are usually similar for
segments belonging to different molecules.
The Helmholtz energy for a mixture of associating molecules
is:
a

A
= a id + a hc + a disp + a assoc
NkT

(3)

where a id is the ideal gas contribution, a hc is the contribution of

the hard-sphere chain reference system, a disp and a assoc are the
contributions of dispersion forces and association. The expression for the contributions from the ideal gas and dispersion were
used as in the original PC-SAFT version. The two modifications
of von Solms et al. [2] affect only the hard-sphere chain and the
association term. First, it is assumed that all segments in the
mixture have the same diameter. This diameter is calculated by
setting the volume fraction 3 where
=

d 3 
xi mi
6

(4)


d=

xi mi di3
i

1/3

i xi m i

Using Eq. (5) we recalculate n with Eq. (6)


n = xi mi din
6

(5)

(6)

This approach yields a simpler expression for ghs (radial distribution function) and the expression for a hs (hard-sphere term of
the Helmholtz energy) reduces to the CarnahanStarling equation [18] as shown in Table 1. Simplified PC-SAFT uses these
hs , however we calculate as usual
terms of ghs () and a CS
=


xi mi di3
6

(7)

Furthermore, we employ a slightly different expression for the

association strength
Ai Bj than in the original PC-SAFT [1], as
also shown in Table 1.
More detailed, we replace, consistent with the earlier simplifications, the radial distribution function with the simpler
expression. Finally, we replace the cubed diameter by the sphere
volume. The replacement of gijhs by ghs () makes the composition dependence of Ai Bj (required for evaluation of the fugacity
coefficient) much simpler, since the compositions dependence
is contained only in .

250

A. Grenner et al. / Fluid Phase Equilibria 261 (2007) 248257

The LorentzBerthelot combining rules are employed. The

binary interaction parameter k12 is introduced to correct the
dispersion potentials of unlike molecules.

ij = (1 kij ) ii jj ,

ij =

ii + jj
2

(8)

The combining rule used in this work for solvating systems

needed for the association strength

 Ai Bj 

Ai Bj = Nav ij3 ghs ()Ai Bj exp

1
(9)
6
kT
is opposite to the rule used by Li and Englezos [3]. We employ
the arithmetic mean for the association energy and the geometric
mean for the association volume:
Ai Bj =

Ai Bi + Aj Bj
,
2

Ai Bj =

AB A B
i i j j

(10)

The choice of the combining rule used in this work was made
based on the results of Derawi et al. [8]. In that study six different kinds of combining rules were applied and tested for their
performance in glycol + water systems. The combining rule of
Eq. (10) performed best overall, with k12 -values close to zero
and small deviations.
It is known that solvation occurs between the lone electron pair of self-associating compounds and the -electrons
in aromatic hydrocarbons [19]. This was considered using the
approach of Folas et al. [9] where solvation is taken into account
by extending the combining rule (10). By assuming that the aromatic hydrocarbon has no association energy, the associating
energy of the mixture is obtained by dividing the association
energy parameter of the self-associating compound by 2. The
association volume is obtained by fitting to experimental data.
Ai Bj =

selfassociating compound
,
2

Ai Bj
Ai Bj = fitted

(11)

New pure compound parameters for the glycol oligomers were

estimated. Usually pure compound parameters were fitted to
vapor pressure and liquid densities, either using the actual experimental points or data correlation such as the DIPPR database
[20] in a temperature range of about 0.5 Tr 0.9. However,
the temperature range of the actual experimental data points
used in DIPPR is limited for the glycol oligomers. For that reason the pure compound parameters have been fitted only to a
smaller temperature range which is covered by the real experimental data. The glycol parameters obtained are summarized in
Table 2. The trend of the parameter with respect to the molar
mass is clearly linear for m 3 and m/k. There is one outlier for

Fig. 1. Trend of glycol pure compound parameters with respect to the molar
mass. Association parameters are considered to be constant for all glycol
oligomers.

the m parameter (m of TEG), as shown in Fig. 1. The equations

representing the lines in Fig. 1 are:
m = 0.0192M + 0.7924

(12)

m 3 = 1.3121M + 8.5441
m
= 6.1866M + 224.49
k

(13)
(14)

where M is the molar mass in g/mol, m denotes the number of

and /k is
segments, represents the segment diameter in A
the dispersion energy in K. The PC-SAFT parameters of other
compounds used are taken from Gross and Sadowski [1] and
Grenner et al. [21] and are also listed in Table 1.
3. Vaporliquid equilibria in glycol containing
self-associating systems
At first the simplified PC-SAFT EoS was employed to correlate VLE in systems of glycol + aromatic hydrocarbons, hexane,
CO2 , N2 and methane. The deviations between experimental
and calculated data and the k12 -values are given in Table 3.
TEG/TeEG + aromatic hydrocarbon were considered without

Table 1
Modifications of the simplified PC-SAFT EoS [2] compared to the PC-SAFT EoS [1]
PC-SAFT [1]
gijhs =

1
13

a hs =

1
0

Ai Bj =

di dj
di +dj

32
(13 )2

di dj
di +dj

Simplified PC-SAFT [2]

222
(13 )3

23
3
+ 22 0 ln(1 3 )
3 (13 )2
3
Nav ij3 ghs Ai Bj [exp(Ai Bj /kT ) 1]
31 2
13

ghs () =
hs
a CS
=

1/2
(1)3

432
(1)2

Ai Bj = Nav 6 ij3 ghs ()Ai Bj [exp(Ai Bj /kT ) 1]

A. Grenner et al. / Fluid Phase Equilibria 261 (2007) 248257

251

Table 2
PC-SAFT pure compound parameters. All glycols and water are modeled with the 4C association scheme
Compound

Ref.

(A)

/k (K)

AB (K)

AB

PS AAD (%)

AAD (%)

Tr range

MEG
DEG
TEG
TeEG
PG
Methane
Hexane
Heptane
Benzene
Toluene
CO2
N2
Water

This work
This work
This work
This work
This work
[1]
[1]
[1]
[1]
[1]
[1]
[1]
[21]

3.5914
3.6143
4.0186
3.8138
3.6351
3.7039
3.7983
3.8049
3.6478
3.7169
2.0729
3.3130
2.6273

325.23
310.29
333.17
309.22
284.62
150.03
236.77
238.4
287.35
285.69
169.21
90.96
180.3

1.90878
3.05823
3.18092
4.7509
2.33917
1.000
3.0576
3.4831
2.4653
2.8149
2.7852
1.2053
1.50

2080.03
2080.03
2080.03
2080.03
2080.03

1804.22

0.0235
0.0235
0.0235
0.0235
0.0235

0.0942

0.56
0.67
1.48
1.73
2.10
0.36
0.31
0.34
0.64
2.41
2.78
0.34
2.62

1.63
0.19
0.40
0.81
1.36
0.67
0.76
2.10
1.42
1.35
2.73
1.50
0.93

0.500.89
0.490.63
0.480.70
0.500.77
0.450.80
0.50overcrit
0.351.00
0.33overcrit
0.501.00
0.301.00
0.710.99
0.501.00
0.500.90

explicit accounting for solvation as the results are very satisfactory using a single interaction parameter (k12 ). Similar results
were obtained by Folas et al. [9] with the CPA EoS. Even with
k12 = 0 (prediction) the percentage deviation in temperature for
these isobaric VLE data are in all considered cases lower than
2.5%. The results for the VLE in TeEG + aromatic hydrocarTable 3
Glycol VLE in self-associating systems. Temperature-independent k12 values
and deviations using simplified PC-SAFT

Ta (K);

Pb (%)

yb

System

Ref. of
data

Condition

k12

TEG + benzene
TEG + toluene
TeEG + benzene
TeEG + toluene
TeEG + o-xylene
TEG + hexane

[22]

101.32 kPa
101.32 kPa
101.32 kPa
101.32 kPa
101.32 kPa
473 K

0.0114
0.0039
0.0055
0.0156
0.0153
0.0064

3.78
1.19
2.28
2.24
3.65
9.07

[25]

323.15 K
373.15 K
398.15 Kc

0.120

10.21
14.60
31.88

MEG + CO2 d

[26]

323.15 K
373.15 K
398.15 K

0.0099

43.63
14.97
19.51

MEG + CH4

[27]

323.15 K
373.15 K
398.15 K
283.2 K
293.2 K
303.2 K

0.0728

6.21
3.30
5.14
32.64
13.19
4.73

298.15 K
323.15 K
348.15 K
373.15 K
398.15 K

0.0116

16.07
3.71
4.97
3.63
3.31

MEG + N2

[23]

[24]

[26]

DEG + CO2

[27]

0.0006
0.0009
0.0076
0.0063
0.0023
0.0225

T = 1/np
|Ti calcd Ti exptl | where np is the number of data points.
b Deviations as AAD.
c MEG + N , 398.15 K: with a k = 0.086 the deviation can be reduced to
2
12
about 14%.
d MEG + CO : with temperature-dependent interaction parameters (323.15 K,
2
k12 = 0.0020; 373.15 K, k12 = 0.0146; 398.15 K, k12 = 0.0170) the deviation
can be reduced to about 13% for all three temperatures.
a

bons are shown in Fig. 2. Furthermore, the VLLE of the system

TEG + hexane was considered; Li and Englezos [3] report deviation for the pressure which is about 5% (RMS) lower than that in
this work. The binary systems MEG + CO2 /N2 /methane at three
temperatures were studied since they are interesting from the
practical point of view as dissolved gases reduce the capability
of glycols as gas hydrate inhibitors [25]. As Figs. 3 and 4 show,
the overall performance of simplified PC-SAFT is good. For
the systems with MEG + CO2 /methane the model can describe
the systems also with a rather small value of a temperatureindependent k12 -value without much increase of the deviations,
for the case of MEG + N2 larger k12 -values are needed and for
398.15 K a large deviation is obtained. For MEG + methane, the
interaction parameters follow a linear trend with respect to temperature for the data of Wang et al. [26]. The k12 -values for the
same system optimized from the data of Zheng et al. [25] are in
the same order of magnitude but do not follow the linear trend.

Fig. 2. VLE for TeEG + aromatic hydrocarbon. Lines are correlations simplified PC-SAFT; TeEG + benzene, k12 = 0.0055; TeEG + toluene, k12 = 0.0156;
TeEG + o-xylene, k12 = 0.0153. Experimental data are taken from Yu et al. [23].

252

A. Grenner et al. / Fluid Phase Equilibria 261 (2007) 248257

much higher at even lower pressure compared to MEG which

makes it easier to obtain reliable experimental data for the DEG
system. The k12 -values of DEG + CO2 have a linear trend with
the temperatures. We can reduce the number of parameters for
the system DEG + CO2 using the following equation:
k12 = 0.0000456 T/K 0.0268, 298 T, K 373

(15)

4. Vaporliquid equilibria in glycol solvating systems

For the MEG + water system extensive data are available from
343 to 395 K [29,30]. Small deviations with small k12 -values
are obtained. Fig. 5 shows results for correlations and predictions (k12 = 0). The deviations are summarized in Table 4. The
k12 -values decrease almost linearly with temperature. We can
generalize the interaction parameter for MEG + water by using
the following equation
k12 = 0.000553 T/K + 0.162357
Fig. 3. Methane solubility in MEG. Lines are correlations with the simplified PC-SAFT using a temperature-independent interaction parameter of
k12 = 0.0728. Experimental data from Zheng et al. [25].

The k12 -values of Pedrosa et al. [5] are for the data of Zheng et
al. about 0.25 higher than those used here. Also Li and Englezos
[28] apply a large k12 -value for MEG + methane (k12 = 0.5156)
and obtain still rather high deviations for the pressure (15.3%
RMS). The k12 -values of MEG + CO2 and DEG + CO2 systems
of this work are in the same order of magnitude but the deviations for DEG + CO2 are much smaller, as shown in Table 3.
A possible explanation is that the solubility of CO2 in DEG is

Fig. 4. CO2 solubility in DEG. Lines are correlations with the simplified PCSAFT using a temperature-independent interaction parameter k12 = 0.0116.
Experimental data are taken from Jou et al. [27].

(16)

The data at 343.15 K represent an exception, where smaller deviations could be obtained if a larger negative k12 -value, outside
the linear trend, is used. Moreover, a temperature independent interaction parameter k12 = 0.046 can be used, resulting
in deviations lower than 4.3% AAD for the bubble pressure,
except for 395 K where the deviation for the bubble pressure
is 6.20% AAD. Those deviations are, however, 12% higher
than those results obtained with k12 -values from Eq. (16). A
comparison with the work of Derawi et al. [8] shows that
the results of this study are, in general, slightly better for the
system MEG + water. But rather large k12 -values are required
with CPA to minimize the deviations in the DEG/TEG + water
systems.

Fig. 5. VLE for MEG + water. Lines are correlations with simplified PCSAFT; dashed line, prediction (k12 = 0); solid line, fit to experimental data
(k12 = 0.046). Experimental data are taken from Chiavone-Filho et al. [29].

A. Grenner et al. / Fluid Phase Equilibria 261 (2007) 248257

253

Table 4
Predictions and correlations of VLE in solvating systems with the simplified PC-SAFT. Temperature-independent k12 values and deviations are presented. Deviations
are expressed as AAD
System

Ref. of data

T (K)

k12

MEGwater

[29]

0.000

[30]

343.15
363.15
371.15
383.15
395.15

DEGwater

[31]

TEGwater

P (%)

k12

P (%)

17.11
16.75
15.51
24.19
28.57

0.0021
0.003
0.0165
0.0267
0.0262

0.046

2.39
3.67
2.59
3.32
6.20

0.0015
0.0054
0.0050
0.0091
0.0148

393.15

77.42

0.0255

0.127

12.00

0.0086

[32]

297.60
332.60

116.62
64.34

0.147

8.35
8.85

PGwater

[30]

371.15
383.15
395.15

28.51
27.30
33.12

0.0631
0.0543
0.0390

0.069

2.41
2.07
1.71

0.0105
0.0042
0.0059

MEG + PG

[30]

371.15
383.15
393.15

3.97
2.15
3.50

0.0103
0.0124
0.0100

0.0039

4.46
1.79
2.48

0.0089
0.0097
0.0102

Gross and Sadowski [1]. For MEG/DEG/PG + heptane, deviations lower than 20 AAD% in both phases were obtained. For
TEG and TeEG larger deviations occur, up to 60 AAD%. Some
results are presented in Figs. 7 and 8, while deviations for all
systems studied are summarized in Table 5. Derawi et al. [7]
obtained excellent results for these glycol + heptane systems
with the CPA EoS using the k12 and the pure compound parameters optimized to the LLE data. Derawi et al. [7] showed that
the glycol parameter sets and k12 -values which are based on this
LLE data in a temperature range between approximately 305
and 360 K can correlate the LLE data [33,34] over an extended
temperature range. Data for DEG, TEG and TeEG with heptane
are in literature available. The data of Rawat and Prasad [34]
for TEG + heptane, are at low temperatures, not as reliable as
those of Derawi et al., as illustrated for TEG in Figs. 9 and 10.

Fig. 6. VLE for PG + water. Lines are fit to experimental data with simplified
PC-SAFT using a temperature-independent interaction parameter k12 = 0.069.
Experimental data are taken from Lancia et al. [30].

The system PG + water plotted in Fig. 6 can be modeled with

a temperature independent k12 -value of k12 = 0.0685 whereas
the deviation in pressure are, for all three temperatures, lower
than 3.7%. Furthermore, the system MEG + PG was studied. The
k12 -values are small and therefore no large difference between
prediction (k12 = 0) and the k12 fit is observable. Li and Englezos
[28] obtain for MEG + water at 371.15 K with two fitted parameters (Eq. (1)) higher deviations (about 2% higher as RMS for
the pressure) compared with this work.
5. Glycol + heptane liquidliquid equilibria
The glycol + heptane LLE data of Derawi et al. [11] were
modeled using simplified PC-SAFT and the new glycol parameters of this work. The heptane parameters were taken from

Fig. 7. LLE for DEG + heptane. Lines are correlation with simplified PC-SAFT
using k12 = 0.032. Experimental data are taken from Derawi et al. [11].

254

A. Grenner et al. / Fluid Phase Equilibria 261 (2007) 248257

Fig. 8. LLE for TEG + heptane. Lines are correlation with simplified PC-SAFT
using k12 = 0.045. Solid line, correlations with simplified PC-SAFT. Experimental data are taken from Derawi et al. [11].
Table 5
Deviations and k12 -values for glycol (1) + heptane (2) LLE with the simplified
PC-SAFT. Deviations are AAD/%; I, glycol rich phase; II, heptane rich phase
System

Ref.

T (K)

k12

x2 in (I)

x1 in (II)

MEG + heptane
DEG + heptane

[11]
[11]
[11,33]
[11]
[11,34]
[11]
[11,34]
[11]

316352
313353
313448
309351
309410
306356
306403
308352

0.0400
0.0320
0.0320
0.0450
0.0450
0.0327
0.0327
0.0100

16.9
5.6
5.2
29.1
29.1
63.3
63.3
11.2

20.3
8.3
16.6
50.0
39.2
60.2
51.5
10.3

TEG + heptane
TeEG + heptane
1,2 PG + heptane

Fig. 10. LLE for TEG + heptane. Lines are correlation with simplified PC-SAFT
using k12 = 0.045. Experimental data are taken from Derawi et al. [11] and Rawat
and Prasad [34].

Thus, they were omitted in the calculations of the deviations.

The deviations which include the higher temperatures are also
presented in Table 5.
6. Glycol + aromatic hydrocarbon liquidliquid
equilibria
Binary mixtures of one of the glycols MEG, DEG and
TEG with either benzene or toluene were considered. The data
were taken from Folas et al. [12], Srensen and Arlt [35] and
Mandik and Lesek [36]. Because of solvation, the mutual solubility in glycol + aromatic hydrocarbons systems is higher than
in the corresponding system with an aliphatic hydrocarbon.
Thus, a mutual solubility which is too low is obtained with
the model if solvation is not accounted for. Figs. 11 and 12
show typical results for MEG/DEG + benzene. Two correlations
are presented, with and without accounting for the solvation
by the modified combining rule (Eq. (11)). The results are reasonably accurate when solvation is accounted for. Only for
DEG + benzene and TEG + toluene larger deviations occur. The
Ai Bj ) and the deviations for all
interaction parameters (k12 and fitted
glycol + aromatic hydrocarbons LLE are presented in Table 6.
Folas et al. [9,12] studied extensively the LLE in these systems with CPA EoS and obtained results which have, in some
systems, about 15% AAD lower deviation compared to this
study.
7. Prediction for ternary systems

Fig. 9. LLE for DEG + heptane. Lines are correlation with simplified PC-SAFT
using k12 = 0.032. Experimental data are taken from Derawi et al. [11] and
Johnson and Francis [33].

Finally, we present prediction of VLE and LLE in two

ternary systems. The ternary VLE data [29] in the system
water + PG + MEG at 371.15, 383.15 and 395.15 K were predicted with temperature independent kij -values fitted to binary

A. Grenner et al. / Fluid Phase Equilibria 261 (2007) 248257

255

Table 6
Ai Bj values (Eq. (11)) for glycol (1) + aromatic hydrocarbon (2) LLE with the simplified PC-SAFT. Deviations are AAD/%; I, glycol rich phase;
Deviations, k12 and fitted
II, aromatic rich phase
System

Ref.

T range (K)

k12

Ai Bj
fitted

x2 in (I)

x1 in (II)

MEG + benzene
MEG + toluene
DEG + benzene
DEG + toluene
TEG + benzene
TEG + toluene

[12]
[12]
[35]
[36]
[12]
[12]

279342
279361
293353
307386
278288
279345

0.0199
0.0242
0.0113
0.0199
0.0490
0.0181

0.043
0.049
0.070
0.080
0.498
0.470

8.7
6.7
22.7
8.9
30.5
15.3

9.8
6.6
33.5
7.9
8.0
55.1

Fig. 11. LLE for MEG + benzene. Lines are correlations with simplified PCSAFT. Dotted line, without accounting for solvation (k12 = 0.02); solid line,
Ai Bj = 0.043). Experimental data are
accounting for solvation (k12 = 0.0199 fitted
taken from Folas et al. [12].

systems, as shown in the previous sections. The results were

compared to the predictions of Li and Englezos [3] but only
for vapor phase composition since they presented no bubble
pressure deviations. The results for predicting the vapor phase
compositions are very satisfactory. The maximum deviations
are 0.0257 RMS for all cases considered. Li and Englezos [3]
obtain deviations which are about 0.01 RMS higher than those
in this work. Larger deviations occur for the bubble pressure
especially at 383.15 K (around 15% RMS). For the other temperatures better results were obtained. The deviations are shown
in Table 7.
For the system water + DEG + benzene the LLE [33] at
298.15 K and atmospheric pressure was predicted. Binary LLE
data of DEG + benzene were taken from Srensen and Arlt
[35]. The kij -values of the two binary systems, DEG + water and
DEG + benzene were previously presented (Tables 4 and 6). For
the system benzene + water, the LLE data of Tsonpoulos [37]
were modeled to obtain the interaction parameter. The modified combining rule (Eq. (11)) was used to account for solvation
(k12 = 0.061 and ijfitted = 0.16). The results of the LLE in the
ternary system depend strongly on the modeling quality of the
binary system DEG + benzene. The prediction is presented in
Fig. 13. The two results shown are both with the same k12 values for the binary systems benzene + water and DEG + water
but different k12 -values for the system DEG + benzene corresponding to those results shown in Fig. 12. If the glycol rich
phase is not well correlated (solid line) the ternary LLE will not
be satisfactory predicted. Better results will be obtained if the
glycol rich phase is correlated with smaller deviations (dotted
line).
Table 7
Predicted VLEa in the system water (1) + PG (2) + MEG (3) [30] with the simplified PC-SAFT (this work) and SAFT (Li and Englezos) [3]
T (K)

Fig. 12. LLE for DEG + benzene. Lines are correlation with the simplified PCSAFT; dotted line, without accounting for solvation (k12 = 0.0164); solid line,
Ai Bj = 0.070). Experimental data are
accounting for solvation (k12 = 0.0113, fitted
taken from Srensen and Arlt [35].

P (%)

y1

y2

y3

Combining rule Eq. (10)

Temperature independent kij (simplified PC-SAFT), this work
371
10.56
0.0138
0.0062
383
14.37
0.0235
0.0147
395
6.51
0.0233
0.0140

0.0094
0.0125
0.0140

Combining rule Eq. (1)

(SAFT) Li and Englezos [3]
371

383

395

0.012
0.0188
0.009

All deviations RMS.

0.0336
0.0379
0.0327

0.0226
0.0202
0.0263

256

A. Grenner et al. / Fluid Phase Equilibria 261 (2007) 248257

k
m
M
N
Ps
T
x, y

Fig. 13. LLE for DEG + water + benzene. Lines are correlations with simpliAi Bj = 0.16 and water + DEG
fied PC-SAFT; water + benzene k12 = 0.061, fitted
k12 = 0.127 for both calculations shown. For DEG + benzene different kij were
used according to the two different correlations. The solid line corresponds to
the solid lines in Fig. 12 and the dotted line corresponds to the dotted line,
respectively. Experimental data are taken from Johnson and Francis [33].

8. Conclusions
The simplified PC-SAFT equation of state was employed to
correlate and predict phase equilibria in ethylene/propylene glycol oligomers containing systems. Because of our future goal to
develop a group contribution scheme, new pure glycol compound parameters with identical association parameters based
on DIPPR correlations were estimated. Vaporliquid equilibria
in self-associating systems and solvating systems were modeled with satisfactory results. The performance of simplified
PC-SAFT is comparable or slightly better to other glycol studies with different SAFT types or the CPA EoS. For some systems
constant k12 -values can be used. Employing the same pure compound parameter as used for the VLE, LLE in glycol + heptane
and glycol + aromatic hydrocarbon systems were correlated. Satisfactory results are obtained for three glycols (MEG/DEG/PG)
with heptane. For the glycol aromatic systems, solvation was
explicitly accounted for by modified combining rule (Eq. (11))
with reasonable results. Finally, prediction for VLE and LLE in
two ternary systems was performed. The results for vapor phase
composition in the system water + PG + MEG could be predicted
successfully with very low deviations while larger deviations for
the bubble pressure occur. Also the binodal curve at 298.15 K in
the system DEG + water + benzene was predicted with satisfactory results. These results show that pure compound parameters
with identical association parameters can be used for VLE and
to some extend LLE.
List of symbols
A
Helmholtz energy (J)
a
reduced Helmholtz energy
d
temperature dependent segment diameter
g
radial distribution function

Boltzmanns constant, J/K/PC-SAFT binary interaction parameter

segment number
molar mass (g/mol)
number of molecules
saturated vapor pressure
temperature (K)
mole fraction (liquid, vapor)

Greek letters

AB
association strength
AB
association energy (K)

volume fraction/soft-SAFT binary interaction parameter for size

AB
association volume

density

segment diameter (A)

soft-SAFT binary interaction parameter for energy

Superscripts
assoc association
calcd calculated
exptl
experimental
disp
dispersion
hc
hard chain
hs
hard sphere
id
ideal
Acknowledgement
The authors are grateful to the Danish Research Council for
Technology and Production Sciences for financial support of this
work as part of the project Advanced Thermodynamic Tools for
Computer-Aided Product Design.
References
[1] J. Gross, G. Sadowski, Ind. Eng. Chem. Res. 40 (2001) 12441260.
[2] N. von Solms, M.L. Michelsen, G.M. Kontogeorgis, Ind. Eng. Chem. Res.
(42) (2003) 10981105.
[3] X.-S. Li, P. Englezos, Ind. Eng. Chem. Res. 42 (2003) 49534961.
[4] S.H. Huang, M. Radosz, Ind. Eng. Chem. Res. 29 (1990) 22842294.
[5] N. Pedrosa, J.C. P`amies, J.A.P. Coutinho, I.M. Marrucho, L.F. Vega, Ind.
Eng. Chem. Res. 44 (2005) 70277037.
[6] F.J. Blas, L.F. Vega, Mol. Phys. 92 (1997) 135150.
[7] S.O. Derawi, M.L. Michelsen, G.M. Kontogeorgis, E.H. Stenby, Fluid
Phase Equilib. 209 (2003) 163184.
[8] S.O. Derawi, G.M. Kontogeorgis, M.L. Michelsen, E.H. Stenby, Ind. Eng.
Chem. Res. 42 (2003) 14701477.
[9] G.K. Folas, G.M. Kontogeorgis, M.L. Michelsen, E.H. Stenby, Ind. Eng.
Chem. Res 45 (2006) 15271538.
[10] G.K. Folas, S.O. Derawi, M.L. Michelsen, E.H. Stenby, G.M. Kontogeorgis, Fluid Phase Equilib. 228-229 (2005) 121126.
[11] S.O. Derawi, G.M. Kontogeorgis, E.H. Stenby, T. Haugum, A.O. Fredheim,
J. Chem. Eng. Data 47 (2002) 169173.
[12] G.K. Folas, G.M. Kontogeorgis, M.L. Michelsen, E.H. Stenby, E. Solbraa,
J. Chem. Eng. Data 51 (2006) 977983.
[13] G.M. Kontogeorgis, E.C. Voutsas, I.V. Yakoumis, D.P. Tassios, Ind. Eng.
Chem. Res. 35 (1996) 43104318.
[14] M.S. Wertheim, J. Stat. Phys. 35 (1984) 1934.
[15] M.S. Wertheim, J. Stat. Phys. 35 (1984) 3547.

A. Grenner et al. / Fluid Phase Equilibria 261 (2007) 248257

[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]

M.S. Wertheim, J. Stat. Phys. 42 (1986) 459476.

M.S. Wertheim, J. Stat. Phys. 42 (1986) 477492.
N.F. Carnahan, K.E. Starling, J. Chem. Phys. 51 (1969) 635.
R.L. Brinkley, R.B. Gupta, AIChE J. 47 (2001) 948953.
T.E. Daubert, R.P. Danner, Physical and Thermodynamic Properties of Pure
Chemicals: Data Compilation, Hemisphere, New York, 2003.
A. Grenner, J. Schmelzer, N. von Solms, G.M. Kontogeorgis, Ind. Eng.
Chem. Res. 45 (2006) 81708179.
S.K. Gupta, B.S. Rawat, A.N. Goswami, S.M. Nanoti, R. Krishna, Fluid
Phase Equilib. 46 (1989) 95102.
Y.-X. Yu, J.-G. Liu, G.-H. Gao, Fluid Phase Equilib. 157 (1999) 299307.
R.L. Rowley, G.L. Hoffman, AIChE Symp. Ser. 86 (1990) 619.
D.-Q. Zheng, W.-D. Ma, T.-M. Guo, Fluid Phase Equilib. 155 (1999)
277286.
L.-K. Wang, G.-J. Chen, G.-H. Han, X.-Q. Guo, T.-M. Guo, Fluid Phase
Equilib. 207 (2003) 143154.

257

[27] F.-Y. Jou, F.D. Otto, A.E. Mather, Fluid Phase Equilib. 175 (2000) 5361.
[28] X.-S. Li, P. Englezos, Fluid Phase Equilib. 224 (2004) 111118.
[29] O. Chiavone-Filho, P. Proust, P. Rasmussen, J. Chem. Eng. Data 38 (1993)
128131.
[30] A. Lancia, D. Musmarra, F. Pepe, J. Chem. Eng. Jpn. 29 (1996) 449
455.
[31] E.S. Klyucheva, N.L. Yarym-Agaev, Zh. Prikl. Khim. 53 (1980) 794796.
[32] M. Herskowitz, M.J. Gottlieb, Chem. Eng. Data 29 (1984) 173175.
[33] G.C. Johnson, A.W. Francis, Ind. Eng. Chem. 46 (1954) 16621668.
[34] B.S. Rawat, G.J. Prasad, J. Chem. Eng. Data 25 (1980) 227230.
[35] J.M. Srensen, W. Arlt, Liquidliquid equilibrium data collection, (Binary
Systems); DECHEMA Chemistry Data Series, vol.V, DECHEMA, Frankfurt, Germany, 1980, Part 1.
[36] L. Mandik, F. Lesek, Collect. Czech. Chem. Commun. 47 (1982)
16861694.
[37] C. Tsonopoulos, Fluid Phase Equilib. 186 (2001) 185206.