Ind. Eng. Chem. Res.

2001, 40, 957-964 957

From UNIFAC to Modified UNIFAC (Dortmund)†

1 rgen Lohmann,‡ Ralph Joh,§ and Ju
Ju 1 rgen Gmehling*,‡
Lehrstuhl für Technische Chemie (FB9), Carl von Ossietzky Universität Oldenburg, Postfach 2503,
D-26111 Oldenburg, Germany, and Axiva GmbH, Industriepark Höchst, D-65926 Frankfurt/Main, Germany

The group contribution method UNIFAC has become very popular because of its availability
via commercial process simulators, its reliably predicted vapor-liquid equilibrium (VLE) results,
its simple usage, and, above all, its large range of applicability. Nevertheless, the original
UNIFAC shows a few weaknesses. For example, there is no chance to describe simultaneously
VLE data and excess enthalpies with the required accuracy. This means that the temperature
dependence of the activity coefficient following the Gibbs-Helmholtz equation cannot be described
correctly. Furthermore, predictions of the real phase behavior in the dilute region and for
asymmetric systems can lead to poor results. The aim of this paper is to show how the mentioned
weaknesses were overcome by using modified UNIFAC (Dortmund) instead of the original
UNIFAC method, in which the required temperature-dependent parameters are fitted simul-
taneously to a large database covering various types of reliable phase equilibrium and excess
property data. Furthermore, 10 new modified UNIFAC (Dortmund) parameter pairs for the new
main group “epoxides“ and the interactions between aromatics and perfluorinated compounds
are given.

Introduction (usually measured for liquid mole fractions between 0.02

Group contribution methods are of major interest for
the synthesis, design, and optimization of thermal
separation processes and a large number of other
applications of industrial interest. In particular, they
can be successfully applied to predict phase equilibria
and excess properties which are necessary requirements
for a reliable synthesis and design of chemical processes,
in particular separation processes. Especially during the
development of chemical processes, group contribution
methods are most useful.1 Different group contribution
methods have been suggested, in which UNIFAC2,3 has
become one of the most popular models.
The main objective during the development of the
original UNIFAC was the prediction of vapor-liquid
equilibrium (VLE) data for a limited temperature range
from 290 to 400 K. With the help of UNIFAC, the VLE
behavior of a large number of mixtures (binary and
higher) can be predicted using van der Waals volumes
(Rk) and surface areas (Qk) as well as the fitted binary
group interaction parameters anm and amn. Extensive
tables for the mentioned van der Waals properties and
group interaction parameters have been published.4
Nevertheless, weaknesses of the original UNIFAC
method are also well-known. Predictions of the real
phase behavior for mixtures at high (>400 K) and low
(<290 K) temperatures can lead to poor results. Prob-
lems also can occur for other types of thermodynamic
data [e.g., enthalpies of mixing (hE) or solid-liquid
equilibria (SLE)]. Furthermore, the real phase behavior
in the diluted region of the compounds to be separated
is not always in agreement with the experimental data.
These limitations are not surprising, because VLE data
†
Dedicated to Professor Dr. Peter Köll (University of Old-
enburg) on the occasion of his 60th birthday.
* To whom correspondence should be addressed. E-mail:
gmehling@tech.chem.uni-oldenburg.de. www: http://www.
uni-oldenburg.de/tchemie.
‡ Carl von Ossietzky Universität Oldenburg.
§ Axiva GmbH.
10.1021/ie0005710 CCC: $20.00 © 2001 American Chemical Society
Published on Web 01/05/2001
and 0.98) do not provide information about the dilute
region and the temperature dependence of the activity
coefficients. This information can be obtained from
activity coefficients at infinite dilution (γ∞ data) and
enthalpies of mixing (hE), respectively. Additionally,
complete VLE data for strong asymmetric mixtures
(containing molecules of very different shapes and sizes)
are rare and give only reliable experimental γi informa-
tion for the low-boiling component.
However, several attempts have been carried out in
order to overcome these weaknesses. A modified version
of UNIFAC has been introduced: modified UNIFAC
(Dortmund).5,6 Later on a similar approach [modified
UNIFAC (Lyngby)] was published by Larsen.7 In both
models, the combinatorial part was modified in order
to improve the results for asymmetric systems. Then,
temperature-dependent parameters were introduced to
describe the temperature dependence of the activity
coefficient correctly. The group interaction parameters
of modified UNIFAC (Dortmund) are fitted simulta-
neously to all reliable thermodynamic mixture data.
This requires a large and reliable database which covers
a wide temperature range.
Because of the extensive research work in recent
years, the range of applicability was strongly increased
and the reliability of modified UNIFAC (Dortmund) was
steadily improved.8-10
This paper provides examples which show the supe-
riority of modified UNIFAC (Dortmund) compared to the
original UNIFAC and other group contribution methods
such as ASOG11,12 and modified UNIFAC (Lyngby).7
Fitting Procedure, Mixture Data, and Examples
The main intention of developing UNIFAC was the
reliable prediction of VLE. For this reason, the database
used for fitting the required group interaction param-
eters was mainly built up by VLE data. The required
van der Waals properties Rk and Qk were obtained
directly from the values given by Bondi.13,14 Because the
958 Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001
Figure 1. Experimental (b, ∆) and predicted (s, UNIFAC) VLE
data of systems containing alkanes and ketones: (A) n-hexane
(1) + 3-pentanone (2); (B) n-pentane (1) + acetone (2); (C)
cyclohexane (1) + cyclohexanone (2); (D) acetone (1) + n-heptane
(2); (E) acetone (1) + n-decane (2); (F) 2-butanone (1) + cyclohexane
(2).
temperature range covered by the VLE data sets
measured is at or close to atmospheric pressure, tem-
perature-independent parameters are used in the origi-
nal UNIFAC method:
Ψnm ) exp[-anm/T]
In most cases, this approach leads to satisfying predic-
tions for VLE data. Figure 1 shows typical results of
the original UNIFAC for six alkane/ketone systems. The
experimental data were taken from the the Dortmund
Data Bank (DDB).15
Using only VLE data and temperature-independent
parameters anm and amn, a satisfying simultaneous
description of the different phase equilibria [VLE,
liquid-liquid equilibria (LLE), and solid-liquid equi-
libria (SLE) of eutectic systems and azeotropic data] and
excess properties (hE and cPE) cannot be achieved.
Therefore, in modified UNIFAC (Dortmund) the follow-
ing expression was chosen:
[ ]
anm + bnmT + cnmT2
Ψnm ) exp -
T procedure. In particular, for the newly introduced main
The experimental and predicted excess enthalpies in
Figure 2 show the improvement when going from
UNIFAC to modified UNIFAC (Dortmund). The original
UNIFAC is not able to describe the hE behavior (in
particular, the temperature dependence) correctly; the
Figure 2. Experimental (b, ∆) and predicted (s, - - -) hE data for
the system 1,2-epoxybutane (1) + methanol (2): (A) UNIFAC; (B)
modified UNIFAC (Dortmund) (data taken from ref 41).
Table 1. Thermodynamic Mixture Data Stored in the
DDB (April 2000) Which Are Used for Fitting the
Required Group Interaction Parameters of Modified
UNIFAC (Dortmund)
type of thermodynamic no. of data sets/data points
mixture data stored in the DDB
VLE 20 000 data sets
activity coefficients at 36 000 data points
infinite dilution (γ∞)
excess enthalpies (hE) 14 500 data sets
excess heat capacities (cPE) 1 200 data sets
LLE 11 000 data sets
SLE 8 200 data sets
azeotropic data (AZD) 45 000 data points
usage of modified UNIFAC (Dortmund) leads to much
better results. This means that the introduction of
temperature-dependent parameters allows a more reli-
able temperature extrapolation and the extension of the
range of applicability: Up to seven different types of
thermodynamic mixture data are used to fit simulta-
neously the required group interaction parameters (see
Table 1). As mentioned above, the required experimen-
tal information is mainly taken from the DDB. A large
part of these data is also available in printed form in
the DECHEMA Chemistry Data Series (VLE, ref 16; hE,
ref 17; γ∞, ref 18; LLE, ref 19) or were published
elsewhere (azeotropic data in ref 20). At the same time,
a large number of systematic measurements were
carried out to enlarge the necessary database.
(1) To find the parameters which allow the best simul-
taneous description of all phase equilibria and excess
properties, thermodynamically inconsistent VLE data
have to be excluded from the fitting procedure. Fur-
thermore, weighting factors are used to set the emphasis
to the desired direction. The application of the various
thermodynamic mixture data types not only allows one
to cover a large temperature range but also takes into
account different important information about the real
mixture behavior. Figure 3 shows the typical temper-
ature ranges covered by thermodynamic mixture data
used for fitting the parameters. Each type of different
mixture data fulfills a special task during the fitting
procedure.
(a) VLE and Azeotropic Data. VLE data and
azeotropic points provide the information about the
composition dependence of the activity coefficients.
Therefore, these data are the basic data for the fitting
(2)
groups (see Figure 8), a large number of additional
measurements were necessary. In a few cases also
confidential VLE data provided from members of the
UNIFAC consortium were used.
(b) Activity Coefficients at Infinite Dilution.
Activity coefficients at infinite dilution (γ∞) can be

Figure 3. Typical temperature ranges of thermodynamic mixture
data.
Figure 4. Experimental (b) and predicted (s) activity coefficients
at infinite dilution of n-heptane in several n-alkanes at 25 °C: (A)
UNIFAC; (B) modified UNIFAC (Dortmund).
Figure 5. Experimental (b) and predicted (s) activity coefficients
at infinite dilution of the system n-hexane + ethanol: (A) UNIFAC;
(B) modified UNIFAC (Dortmund).
measured with the help of gas-liquid chromatography
(GLC),21 headspace gas chromatography,22,23 ebullio-
metry or dilutor technique.24 These data (determined
at various temperatures) not only give information
about the temperature dependence of the activity coef-
ficients but also provide the most important information
about the real phase behavior in the dilute region. Only
with the help of reliable γ∞ data, correct separation
factors at infinite dilution (R12∞) can be obtained. Figure
4 shows the improved predictions of the activity coef-
ficients at infinite dilution for n-heptane in several
n-alkanes as a function of the chain length, where it
has to be mentioned that for systems containing only
one main group (e.g., a mixture consisting only of
alkanes) the deviation from Raoult’s law is predicted
by the combinatorial part of the model only.
Figure 5 gives an impression of the progress in
describing the temperature dependence of the activity
Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001 959
Figure 6. Number of hE data sets stored in the DDB as a function
of temperature.
coefficients at infinite dilution for the system n-hexane
+ ethanol. Using original UNIFAC, in particular the
temperature dependence of the γ∞ values of ethanol in
n-hexane is not in agreement with the experimental
values (2).
(c) Excess Enthalpies. The temperature dependence
of the activity coefficient is described quantitatively by
the Gibbs-Helmholtz equation:
( )
∂ ln γi
∂(1/T) P,x
)
h iE
h
R
(5)
Following this equation, a direct relationship between
the temperature dependence of the activity coefficient
(γi) and the partial molar excess enthalpy (hh iE) is given.
For this reason, excess enthalpies (measured at different
temperatures) are most valuable for the revision and
extension of group contribution methods. Furthermore,
hE data sets at higher temperatures (up to 450 K) are
required as supporting data at high temperatures for
fitting temperature-dependent group interaction pa-
rameters.
Unfortunately, the major part of all published hE data
was determined around ambient temperature (Figure
6). To overcome the lack of data at higher temperatures,
systematic excess enthalpy measurements are carried
out in our laboratory using isothermal flow calorime-
try.25,26
(d) Excess Heat Capacities. Excess heat capacities
(cPE) provide quantitative information about the tem-
perature dependence of the excess enthalpies:
(∂hE/∂Τ)P,x ) cPE (6)
(e) SLE. While hE data sets are used as supporting
data at high temperatures, SLE of eutectic systems
fulfill this task at low temperatures (lower than 290 K).
A simple equation27 provides the relationship among the
solubility (xi), system temperature (T), and activity
coefficient (γi), if the melting temperature (Tm,i), en-
thalpy of fusion (∆hm,i), enthalpy of transition (∆htrs,i),
and transition temperature (Ttrs,i) of component i are
known:
( ) ( )
∆hm,i T ∆htrs,i T
ln xiγi ) - 1- - 1- (7)
RT Tm,i RT Ttrs,i
A large number of systematic SLE measurements28,29
have been carried out to extend the required database.
At the same time, missing pure-component properties
(enthalpies of fusion and transition) have been deter-
960 Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001
Table 2. Modified UNIFAC (Dortmund) Rk and Qk Parameters and Group Assignment for the New Main Group 53
“Epoxides”
main group subgroup no.
53 epoxides H2COCH 107
HCOCH 109
Figure 7. Experimental (b) and predicted LLE of systems
containing various compounds (1) and cyclohexane (2) (- - -,
UNIFAC; s, modified UNIFAC (Dortmund); ‚‚‚, LLE UNIFAC).
1: (A) acetonitrile; (B) N,N-dimethylformamide; (C) nitromethane;
(D) methanol; (E) aniline; (F) methyl formate.
mined with the help of a Setaram calorimeter. It has to
be mentioned that eq 7 is a simplified form because the
terms which consider the differences in the molar heat
capacities are neglected. However, a comprehensive
investigation30 showed that nearly the same SLE results
were achieved by considering or neglecting the differ-
ence in heat capacities. This may be caused by the
relatively small temperature differences (Tm,i - T).
(f) LLE. Experimental LLE data can be used to
enhance the description of strong real systems. How-
ever, the description of the miscibility gap (distribution
coefficients and binodal curve) is much more difficult
compared to the prediction of VLE and SLE, because
not only the composition but also the temperature
dependence of the real behavior has to be described by
the activity coefficients. In the case of VLE and SLE
data, pure-component properties such as vapor pres-
sures, melting temperatures, transition temperatures,
heats of fusion, and heats of transition are mainly
responsible for the temperature dependence. Neverthe-
less, the introduction of temperature-dependent param-
eters allows a better description of the miscibility gap’s
course.
Figure 7 shows the improvement for LLE data when
going from UNIFAC to modified UNIFAC (Dortmund)
Rk Qk sample group assignment
1.3601 1.8031 propylene oxide 1 H2COCH, 1 CH3
0.9104 0.6538 2,3-epoxybutane 1 HCOCH, 2 CH3
Table 3. New or Revised Modified UNIFAC (Dortmund)
Interaction Parametersa
n m anm/K bnm cnm/K-1 amn/K bmn cmn/K-1
1 53 269.67 0.0000 0.000 00 -41.00 0.0000 0.000 00
2 53 9.39 0.0000 0.000 00 57.86 0.0000 0.000 00
3 40 123.26 -0.6133 0.000 00 419.04 -0.1703 0.000 00
3 53 392.56 0.0000 0.000 00 -260.22 0.0000 0.000 00
4 40 29.57 0.0000 0.000 00 153.32 0.0000 0.000 00
4 53 164.05 0.0000 0.000 00 102.04 0.0000 0.000 00
5 53 495.28 -2.6496 0.001 59 700.05 0.2237 -0.002 44
6 53 137.95 -1.4112 0.001 72 524.52 0.4314 -0.001 66
7 53 499.44 -6.6553 0.016 56 952.57 -0.8437 -0.004 80
42 53 286.54 0.1879 0.000 00 58.43 -0.5292 0.000 00
a The main groups 1-46 have been defined previously.8,9
for six systems containing cyclohexane. In all cases, the
predictions with modified UNIFAC (Dortmund) (solid
lines, using the previously published parameters8) are
more accurate compared to the results of original
UNIFAC (dashed lines). For the systems shown in
Figure 7E,F (aniline + cyclohexane and methyl formate
+ cyclohexane), original UNIFAC does not even predict
the miscibility gap which was observed experimentally.
Additionally, the superiority of modified UNIFAC (Dort-
mund) becomes obvious, if the results are compared with
the results of LLE UNIFAC (dotted lines),31 a group
contribution method based on the original UNIFAC
which was especially developed for the prediction of LLE
data at ambient temperature.
With the help of the various thermodynamic phase
equilibria and excess properties stored in the DDB, a
lot of gaps in the parameter matrix of modified UNIFAC
(Dortmund) were filled (see Figure 8).
Furthermore, many new main groups with compo-
nents of immense industrial interest (various mono- and
dialkylated amides, carbonates, anhydrides, sulfones,
epoxides, refrigerants, etc.) have been introduced into
the parameter matrix of modified UNIFAC (Dort-
mund).10,29,32-34 To keep this figure legible, the 10 new
additional main groups for refrigerants with about 100
new group interaction parameters33 are not shown.
In this paper, the parameters for the newly introduced
main group 53 “epoxides” and two revised parameter
sets for aromatics with perfluorinated compounds
(3/40 and 4/40) are given. The parameters and van der
Waals values (see Tables 2 and 3) allow accurate
predictions of systems containing epoxides such as
propylene oxide, 2,3-epoxybutane, or 1,2-epoxybutane.
In all cases a reliable reproduction of the experimental
information was achieved. Unfortunately, predictions of
systems containing ethylene oxide, the most important
representative of this substance class, are predicted only
qualitatively. This is caused by the special behavior of
ethylene oxide, which is the first member of the homo-
logue epoxides. Therefore, the introduction of a new
main group for the ethylene oxide would be desirable,
but up to now the corresponding database is very
limited.
Besides filling parameter gaps using new available
experimental data, also a comprehensive revision of the
previously published parameters has been carried out.
The revisions were necessary because in some cases poor

Figure 8. Current parameter matrix of modified UNIFAC (Dortmund) (April 2000).
Figure 9. Experimental (b, ∆, () and with modified UNIFAC
(Dortmund) predicted hE data of the system benzene (1) +
perfluoro-2-methylpentane (2): (A) previously published param-
eters;35 (B) revised parameters (this paper; data taken from ref
41).
results were obtained when predictions were carried out
outside the temperature range, which was covered by
the database during the fitting procedure. Figure 9
shows two typical examples of revised parameters. It
is obvious that the revised parameters (3/40) lead to
much more reliable results than the parameters pub-
lished previously.35
Besides the work on the revision and extension of
modified UNIFAC (Dortmund), an extensive comparison
of the prediction capability concerning various types of
thermodynamic mixture data (VLE, SLE, hE, γ∞, LLE,
Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001 961
and azeotropic data) was carried out.34 To ensure an
objective comparison and a reproduction of the results
by the reader, only published parameters of the group
contribution methods and a large database have been
used.
In the case of VLE data, the corresponding database
was formed by 2400 thermodynamically consistent VLE
data sets. These are the complete data sets (x, y, T, and
P are given) which can be described with all selected
group contribution methods and which have passed two
consistency tests3,36,37 at the same time. Because of the
small parameter matrices of modified UNIFAC (Lyngby)
and ASOG, the database is mainly limited to relatively
simple compounds (alcohols, ethers, esters, aldehydes,
etc.). UNIFAC and modified UNIFAC (Dortmund) would
allow predictions of approximately 3200 consistent data
sets.34 Figure 10 shows the calculated mean deviations
in mole fraction (∆y), temperature (∆T), and pressure
(∆P) for the reduced common database.
Besides the predicted results using the four group
contribution methods original UNIFAC, modified UNI-
FAC (Dortmund), modified UNIFAC (Lyngby), and
ASOG, UNIQUAC parameters14 were fitted to each data
set. The resulting deviations can be assumed as the
minimum deviations, which can be obtained using a gE
model or group contribution method. It is obvious that
the step from UNIFAC to modified UNIFAC (Dort-
962 Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001
Figure 10. Mean deviations between experimental and predicted
VLE data for 2400 consistent VLE data sets.
Figure 11. Mean deviations between experimental and predicted
activity coefficients at infinite dilution for 13 500 data points.
mund) leads to a considerable improvement of the
predicted results. For example, the mean deviation for
the mole fraction (∆yabs) using modified UNIFAC (Dort-
mund) is just 0.38 mol % higher than the correlated
results using UNIQUAC, while for the original UNI-
FAC, the deviation is 0.88 mol % higher. Similar
conclusions for systems containing ethers (methyl tert-
butyl ether and methyl tert-amyl ether) were found by
Coto.38 Applying the group contribution methods UNI-
FAC, modified UNIFAC (Lyngby), and ASOG for the
same VLE database leads to results which are obviously
worse.
Comparable superior results for modified UNIFAC
(Dortmund) were also found for activity coefficients at
infinite dilution (Figure 11). Again a large database
(13 500 data sets) was used for the comparison. From
all binary data sets stored in the DDB (see Table 1),
the experimental data sets with γ∞ values larger than
100 have been excluded from the comparison because
these data measured by different authors often differ
very strongly (differences by a factor of 10 are not
unusual). Furthermore, it is known that the very large
activity coefficients of hydrocarbons in water and the
corresponding water solubilities predicted with UNIFAC
and its two modified versions are wrong by a factor of
10 (predicted xhydrocarbon in water ≈ 10-3 instead of experi-
mental values of ≈10-4).39 This was accepted to ensure
a satisfying description of the different thermodynamic
properties (e.g., VLE) for water/alcohol systems.
Other limiting factors are again the small parameter
matrices of modified UNIFAC (Lyngby) and ASOG.
These models allow calculations only for approximately
60% of those data which can be predicted by UNIFAC
or modified UNIFAC (Dortmund).
Also in case of the activity coefficients at infinite
dilution, modified UNIFAC (Dortmund) leads to more
reliable predictions. The mean deviations are 40-50%
smaller than the results of the other methods.
Further comparisons for other types of thermody-
namic mixture data can be found elsewere.34 A paper
which provides some of these results (excess enthalpies
and azeotropic data) as well as various applications of
industrial interest using modified UNIFAC (Dortmund)
will be published soon.40
The examples show that modified UNIFAC (Dort-
mund) is superior if compared with other group contri-
bution methods. Therefore, it can be applied successfully
for a variety of additional applications of industrial
interest and for prediction of various types of thermo-
dynamic mixture data. Therefore, it is used by a large
number of chemical engineers during their daily work
on the design and optimization of chemical processes
using the different commercial process simulators.
Outlook
The research work on the revision and extension of
modified UNIFAC (Dortmund) will be continued with
the support of the members of the UNIFAC consortium.
By introduction of new main groups for compounds of
technical interest (isocyanates, lactones, lactames, etc.)
and extension of the parameter matrix (which is nearly
the only limiting factor for the application of group
contribution methods), the range of applicability of
modified UNIFAC (Dortmund) will be further extended.
Therefore, a large number of additional measurements
(VLE, hE, SLE, γ∞, LLE, azeotropic data, etc.) will be
performed in our laboratory. At the same time, the
continuously growing DDB and the confidential data of
the consortium members will be used for the further
development.
During the planned research work on modified UNI-
FAC (Dortmund), also efforts will be undertaken to
minimize the well-known weaknesses of group contribu-
tion methods such as the description of isomeric com-
pounds (e.g., xylenes or butanols) or proximity effects
(effects of nearby polar groups containing strong elec-
tronegative atoms such as oxygen or fluorine).
Furthermore, the introduction of a hydrophobic CH2
group is planned to improve predictions of the water
solubilities for the mentioned water/hydrocarbon (al-
kanes, alkenes, and naphthenes) systems.
Actual information about the current status of the
research work is available via the Internet (http://
www.uni-oldenburg.de/tchemie/consortium/).
Acknowledgment
The authors thank the members of the UNIFAC
consortium for financial support of the continuing
research work and for the experimental data, which
were placed kindly at our disposal. Furthermore, we
thank the DDBST GmbH (Oldenburg, Germany) for
providing the latest version of the DDB.
Notation/Abbreviations
anm ) UNIFAC group interaction parameters between
main groups n and m
anm, bnm, cnm ) modified UNIFAC (Dortmund) group
interaction parameters between main groups n and m
abs ) absolute deviation
AZD ) azeotropic data
cPE ) molar excess heat capacities [J mol-1 K-1]
cP,i ) molar heat capacities of component i [J mol-1 K-1]
F ) objective function
hE ) molar excess enthalpy [J mol-1]
h iE ) partial molar excess enthalpy [J mol-1]
h
∆hm,i ) molar enthalpy of fusion of component i [J mol-1]
∆htrs,i ) molar enthalpy of transition of component i [J
mol-1]

i ) component
k ) subgroup
LLE ) liquid-liquid equilibrium
n, m ) main groups
P ) total pressure [kPa]
PiS ) saturation vapor pressure of component i [kPa]
Qk ) relative van der Waals surface of subgroup k
Rk ) relative van der Waals volume of subgroup k
SLE ) solid-liquid equilibrium
T ) absolute temperature [K]
Tm ) melting temperature [K]
Ttrs ) transition temperature [K]
VLE ) vapor-liquid equilibrium
xi ) mole fraction of component i in the liquid phase
yi ) mole fraction of component i in the vapor phase
γ∞ ) activity coefficient at infinite dilution
γi ) activity coefficient of component i
φi ) fugacity coefficient of component i
Ψnm ) UNIFAC/modified UNIFAC (Dortmund) tempera-
ture term (eqs 1 and 2)
Literature Cited
(1) Gmehling, J. Group contribution methodssideal tools for
the synthesis and design of separation processes. Pure Appl. Chem.
1999, 71, 939-949.
(2) Fredenslund, Aa.; Jones, R. L.; Prausnitz, J. M. Group
contribution estimation of activity coefficients in nonideal liquid
mixtures. AIChE J. 1975, 21, 1086-1099.
(3) Fredenslund, Aa.; Gmehling, J.; Rasmussen, P. Vapor-
liquid equilibria using UNIFACsA group contribution method;
Elsevier: Amsterdam, The Netherlands, 1977.
(4) Hansen, H. K.; Rasmussen, P.; Fredenslund, Aa.; Schiller,
M.; Gmehling, J. Vapor-liquid equilibria by UNIFAC group
contribution. 5. Revision and extension. Ind. Eng. Chem. Res. 1991,
30, 2352-2355.
(5) Weidlich, U. Experimentelle und theoretische Untersuchun-
gen zur Erweiterung der Gruppenbeitragsmethode UNIFAC.
Ph.D. Thesis, University of Dortmund, Dortmund, Germany, 1985.
(6) Weidlich, U.; Gmehling, J. A modified UNIFAC model. 1.
Prediction of VLE, hE and γ∞. Ind. Eng. Chem. Res. 1987, 26,
1372-1381.
(7) Larsen, B. L.; Rasmussen, P.; Fredenslund, Aa. A modified
UNIFAC group-contribution model for the prediction of phase
equilibria and heats of mixing. Ind. Eng. Chem. Res. 1987, 26,
2274-2286.
(8) Gmehling, J.; Li. J.; Schiller, M. A modified UNIFAC model.
2. Present parameter matrix and results for different thermody-
namic properties. Ind. Eng. Chem. Res. 1993, 32, 178-193.
(9) Gmehling, J.; Lohmann, J.; Jakob, A.; Li, J.; Joh, R. A
modified UNIFAC (Dortmund) model. 3. Revision and extension.
Ind. Eng. Chem. Res. 1998, 37, 4876-4882.
(10) Lohmann, J.; Joh, R.; Nienhaus, B.; Gmehling, J. Revision
and extension of the group contribution method Modified UNIFAC
(Dortmund). Chem. Eng. Technol. 1998, 21, 245-248.
(11) Kojima, K.; Tochigi, K. Prediction of vapor-liquid equi-
libria by ASOG method; Kodansha-Elsevier: New York, 1979.
(12) Tochigi, K.; Tiegs, D.; Gmehling, J.; Kojima, K. Determi-
nation of new ASOG parameters. J. Chem. Eng. Jpn. 1990, 23,
453-463.
(13) Bondi, A. Physical properties of molecular crystals, liquids
and glasses; Wiley: New York, 1968.
(14) Abrams, D. S.: Prausnitz, J. M. Statistical thermodynam-
ics of liquid mixtures: A new expression for the excess Gibbs
energy of partly or completely miscible systems. AIChE J. 1975,
21, 116-128.
(15) Gmehling, J.; Menke, J.; Fischer, K.; Rarey, J. R. Ind. Eng.
Chem. Res. 2001, in preparation.
(16) Gmehling, J.; Onken, U.; Arlt, W.; Grenzheuser, P.; Kolbe,
B.; Weidlich, U.; Rarey, J. R. Vapor-Liquid Equilibrium Data
Collection I (21 parts); DECHEMA Chemistry Data Series; DECHE-
MA: Frankfurt, Germany, 1977-2000.
(17) Gmehling, J.; Christensen, C.; Rasmussen, P.; Weidlich,
U.; Holderbaum, T. Heats of Mixing Data Collection (4 parts);
Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001 963
DECHEMA Chemistry Data Series; DECHEMA: Frankfurt,
Germany, 1984-1991.
(18) Gmehling, J.; Tiegs, D.; Medina, A.; Soares, M.; Bastos,
J.; Alessi, P.; Kikic, I.; Menke, J.; Schiller, M. Activity Coefficients
at Infinite Dilution (4 parts); DECHEMA Chemistry Data Series;
DECHEMA: Frankfurt, Germany, 1986-1994.
(19) Soerensen, J. M.; Arlt, W.; Rasmussen, P.; Macedo, M. P.
A. Liquid-Liquid Equilibrium Data Collection (4 parts); DECHE-
MA Chemistry Data Series; DECHEMA: Frankfurt, Germany,
1979-1987.
(20) Gmehling, J.; Menke, J.; Krafczyk, J.; Fischer, K. Azeo-
tropic Data (2 parts); VCH-Verlag: Weinheim, Germany, 1994.
(21) Weidlich, U.; Gmehling, J. Measurement of γ∞ using GLC,
Part I. Results for the stationary phases 1-octacosane, 1-docosanol,
10-nonadecanone and 1-eicosene. J. Chem. Eng. Data 1987, 32,
138-142.
(22) Castells, C. B.; Eikens, D. I.; Carr, P. W. Headspace gas
chromatographic measurements of limiting activity coefficients of
eleven alkanes in organic solvents at 25 °C. 1. J. Chem. Eng. Data
2000, 45, 369-375.
(23) Castells, C. B.; Eikens, D. I.; Carr, P. W. Headspace gas
chromatographic measurements of limiting activity coefficients of
eleven alkanes in organic solvents at 25 °C. 2. Accuracy and
Precision. J. Chem. Eng. Data 2000, 45, 376-381.
(24) Gruber, D.; Krummen, M.; Gmehling, J. The determination
of activity coefficients at infinite dilution with the help of the
dilutor technique (inert gas stripping). Chem. Eng. Technol. 1999,
22, 827-831.
(25) Gmehling, J. Excess enthalpies for 1,1,1-trichloroethane
with alkanes, ketones and esters. J. Chem. Eng. Data 1993, 38,
143-146.
(26) Gmehling, J.; Krentscher, B. The excess enthalpy of
tetrahydrofuran (oxolane) + water, a binary liquid mixture with
closed miscibility curve, below (343.15 K), between (383.15 K), and
above (416.29 K) the consolute temperature. Int. Electron. J. Phys.-
Chem. Data 1995, 1, 291-296.
(27) Prausnitz, J. M. Molecular thermodynamics of fluid phase
equilibria; Prentice-Hall: Englewood Cliffs, NJ, 1969.
(28) Jakob, A.; Joh, R.; Rose, C.; Gmehling, J. Solid-liquid
equilibria in binary mixtures of organic compounds. Fluid Phase
Equilib. 1995, 113, 117-126.
(29) Lohmann, J.; Röpke, T.; Gmehling, J. Solid-liquid equi-
libria of several binary systems with organic compounds. J. Chem.
Eng. Data 1998, 43, 856-860.
(30) Jakob, A. Thermodynamische Grundlagen der Kristalli-
sation und ihre Anwendung in der Modellentwicklung. Ph.D.
Thesis, University of Oldenburg, Oldenburg, Germany, 1995.
(31) Magnussen, T.; Rasmussen, P.; Fredenslund, Aa. UNIFAC
parameter table for prediction of liquid-liquid equilibria. Ind. Eng.
Chem. Process Des. Dev. 1981, 20, 331-339.
(32) Joh, R. Erweiterung der Anwendungsgebiete der Grup-
penbeitragsmethode Modified UNIFAC (Dortmund). Ph.D. Thesis,
University of Oldenburg, Oldenburg, Germany, 1998.
(33) Nienhaus, B. Erweiterung und Weiterentwicklung der
Gruppenbeitragsmethode Modified UNIFAC (Dortmund) zur Be-
schreibung und Vorhersage des Phasen gleichgewichtsverhaltens
aromatischer und kältemittelhaltiger Systeme. Ph.D. Thesis,
University of Oldenburg, Oldenburg, Germany, 2000.
(34) Lohmann, J. Revison, Erweiterung und Evaluation von
Gruppenbeitragsmethoden. Ph.D. Thesis, University of Oldenburg,
Oldenburg, Germany, 2000.
(35) Kreutz, J. Experimentelle und theoretsiche Untersuchun-
gen zur Kristallisation eutektischer Systeme. M.S. Thesis, Uni-
versity of Oldenburg, Oldenburg, Germany, 1996.
(36) Redlich, O.; Kister, A. Thermodynamics of nonelectrolyte
systems (x-y-T relations in a binary system). Ind. Eng. Chem.
1948, 40, 341-348.
(37) Van Ness, H. C.; Byer, S. M.; Gibbs, R. E. Vapor-Liquid
Equilibrium: Part I. An appraisal of data reduction methods.
AIChE J. 1973, 19, 238-244.
(38) Coto, B.; Pando, C.; Renuncio, J. A. R. Prediction of phase
equilibria for binary and ternary mixtures involving tert-butyl
methyl ether and tert-amyl methyl ether. Ind. Eng. Chem. Res.
2000, 39, 767-774.
(39) Wienke, G.; Gmehling, J. Prediction of octanol-water
partition coefficients, Henry coefficients and water solubilities.
Toxicol. Environ. Chem. 1998, 65, 57-86 (erratum: Toxicol.
Environ. Chem. 1998, 67, 275).
964 Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001

(40) Lohmann, J.; Gmehling, J. Modified UNIFAC (Dort- Received for review June 13, 2000
mund): A reliable model with a wide range of applications. J. Revised manuscript received September 26, 2000
Chem. Eng. Jpn. 2001, in press. Accepted October 23, 2000
(41) Lohmann, J.; Bölts, R.; Gmehling, J. Excess Enthalpy Data
for Seven Binary Systems at Temperatures from 50 to 140 °C. J.
Chem. Eng. Data 2000, submitted for publication. IE0005710