Experimental VLE data of methyl acetate

or ethyl acetate + 1-butanol at 0.6 MPa.

Predictions with Peng-Robinson EOS and
group contribution models

P. Susial*, D. García-Vera, R. Susial and Y.C. Clavijo

Escuela de Ingenierias Industriales y Civiles. Universidad de Las Palmas de Gran Canaria, 35017
Las Palmas de Gran Canaria, Canary Islands, Spain.

Datos experimentales del ELV para acetato de metilo o acetato de etilo + 1-butanol a 0.6 MPa.
Predicciones utilizando Peng-Robinson EoS y los modelos de contribución por grupos.

Dades experimentals de l’ELV per l’acetat de metil o acetat d’etil + 1-butanol a 0,6 MPa.
Prediccions utilitzant Peng-Robinson EoS i els models d’aportació per grups.

RECEIVED: 5 SEPTEMBER 2017; REVISED: 14 DESEMBER 2017; ACCEPTED: 19 JANUARY 2018

SUMMARY por grupos ASOG y tres versiones de UNIFAC fue-

Vapor-liquid equilibrium data were obtained with
a stainless steel ebulliometer at 0.6 MPa for methyl
acetate + 1-butanol and ethyl acetate + 1-butanol. The
experimental data for the binary systems were test-
ed and verified thermodynamically, showed positive
consistency when the point-to-point test of Van Ness
was applied. The group contribution models ASOG
and three versions of the UNIFAC were applied to
calculate the vapor-liquid equilibrium data and after,
these values were compared to the experimental data.
The approach f-f was applied by using the Peng-Rob-
inson equation of state, the classical attractive term
was employed. The quadratic and Wong-Sandler mix-
ing rules were verified and the adjustable parameter
of Stryjek-Vera was also applied.
Keywords VLE isobaric data, Methyl Acetate, Ethyl
Acetate, 1-Butanol
RESUMEN
Los datos del equilibrio líquido-vapor para el acetato
de metilo + 1-butanol y el acetato de etilo + 1-butanol
fueron obtenidos a 0.6 MPa utilizando un ebullóme-
tro de acero inoxidable. Los datos experimentales de
los sistemas binarios fueron comprobados y verifica-
dos termodinámicamente, observándose que presen-
tan consistencia positiva al ser aplicado el test punto
a punto de Van Ness. Los modelos de contribución
ron empleados para calcular los datos del equilibrio
líquido-vapor, posteriormente los valores calculados
fueron comparados con los datos experimentales. La
aproximación f-f fue utilizada aplicando la ecuación
de estado de Peng-Robinson, utilizando el término
atractivo clásico. Las reglas de mezclado cuadráticas
y las de Wong-Sandler fueron verificadas y se empleó
el parámetro ajustable de Stryjek-Vera.
Palabras clave: Datos isobáricos del ELV; acetato de
metilo; acetato de etilo; 1-butanol.
RESUM
Les dades de l’equilibri líquid-vapor per a l’acetat de
metil + 1-butanol i l’acetat d’etil + 1-butanol es van
obtenir a 0,6 MPa utilitzant un ebulloscopi d’acer in-
oxidable. Les dades experimentals dels sistemes bina-
ris es van comprovar i verificar termodinàmicament,
i es va observar que presenten consistència positiva a
l’aplicació del test de Van Ness punt a punt. Els mo-
dels d’aportació per grups ASOG i les tres versions
de UNIFAC van ser empleats per calcular les dades
de l’equilibri líquid-vapor, i posteriorment els valors
calculats van ser comparats amb els dades experi-
mentals. L’aproximació f-f es va utilitzar aplicant
* Corresponding author:psusial@dip.ulpgc.es

APRIL- JUNE 2018 | 97

l’equació d’estat de Peng-Robinson, utilitzant el terme
atractiu clàssic. Les regles de barrejat quadràtiques i
les de Wong-Sandler van ser verificades i es va utilit-
zar el paràmetre ajustable de Stryjek-Vera.
Paraules clau: Dades isobàriques de l’ELV; acetat de
metil; acetat de etil; 1-butanol.
IINTRODUCTION
Esters and alcohols are frequently used in different
industrial processes. Methyl acetate is used in organic
synthesis and is also an excellent solvent for resin and
paints, while ethyl acetate is used in the food, photo-
graphic, textile and paper industries. 1-butanol is em-
ployed in chlorination processes, as a dehydrating
agent, in paints, lacquers and varnishes; in the phar-
maceutical industry, for tanning of hides, in the pho-
tographic industry and in perfumes. 1-butanol has also
been studied as biodiesel due to the energy demand.
Consequently, the study of the behavior of these sub-
stances, mixtures and the determination of the va-
por-liquid equilibrium (VLE) has a scientific interest
and is necessary in many industrial processes. That is
why as in previous works1,2 we determined VLE data
for ester/alcohol binary mixtures at moderate pressure.
Data were determined at 0.6 MPa for (1) methyl acetate
+ (2) 1-butanol (MA1B) and (1) ethyl acetate + (2) 1-bu-
tanol (EA1B). These systems were previously studied by
different authors3-6 under several operating conditions.
VLE of MA1B has been studied at 74.66 and 127.99
kPa by Susial and Ortega3, at 101.3 kPa by Belousov
et al.3 , Esteller et al.3 , Ortega and Susial3, and Pat-
lasov et al.4; and at 0.3 MPa by Susial et al.5. VLE of
EA1B has been studied isothermally by Alsmeyer
and Marquardt3 and isobarically at 70.5 and 94 kPa
by Darwish and Al-Khateib3, at 97.3 kPa by Mainkar
and Mene6 at 101.3 kPa by Belousov et al.3 , Ortega et
al.4 and Shono et al.4; and at 0.3 MPa by Susial et al.5.
Azeotropes have not been reported in these systems.
In this study experimental data were verified by ap-
plying the test of Van Ness7 using the FORTRAN pro-
gram8 of Fredenslund et al. The g-f approach enables to
analyze the efficiency of the different group contribu-
tion models9-12. In addition, by using the f-f approach,
experimental data were correlated with the Peng-Rob-
inson13 (PR) equation of state (EOS) using quadratic
mixing rules or Wong-Sandler14 (WS) mixing rules.
The classical attractive term or the adjustable parame-
ter of Stryjek-Vera15 (SV) were also used in both cases.
EXPERIMENTAL
Chemicals
Methyl acetate, ethyl acetate and 1-butanol (99%,
99.9% and 99.9% mass purity, respectively) from Pan-
reac Química S.A. were used. The physical properties,
normal boiling point (Tbp), density (ρii) and refrac-
tive index (nD) at 298.15 K have been previously pub-
lished1,2,5. These products were used as received.
98 | AFINIDAD LXXV, 582
A Kyoto Electronics DA-300 vibrating tube density me-
ter with an uncertainty of ±0.1 kg·m-3 was employed for
density determinations of both pure components and
VLE data. In addition, a Zusi 315RS Abbe refractometer
with an uncertainty of ±0.0002 units was used for the
refractive index determinations of pure components.
Equipment and procedures
An ebulliometer made of stainless steel (2 mm thick-
ness) was employed (Fig. 1) to obtain experimental VLE
data. The apparatus operates with a 400 cm3 capacity.
It has been built to work at moderate or high pressures.
Flooding does not occur when working with about 800
cm3 of mixing liquid. The liquid mixture is heated in
a double-walled inverted vessel (E). Liquid and vapor
circulate due to the Cotrell pump effect that takes place
when the liquid is heated inside the boiling flask (E).
The liquid phase is circulated across the funnel (B) to
be collected in the C valve. The condensed vapor in the
cooler (F) is circulated to be collected in the D valve.
The general description of the equilibrium ebulliom-
eter and the disposal of the different elements in the
installation can be consulted in previous papers1,2,5.
Fig. 1. Schematic diagram
Fig. 1. Schematic diagram of theof the equilibrium
equilibrium recirculation
recirculation still used for VLE
measurements.
still used for VLE measurements.
The experiments began with the cleaning of the equip-
ment. For this, about 500 cm3 of ethanol were introduced
in the ebulliometer and the electric resistance located at 16
E (see Fig. 1) was switched on; ethanol was boiled and
kept under recirculation at atmospheric pressure for
45 min. After this, ethanol was removed and, with the
equipment still hot, the system was left under vacuum
at 10 kPa (absolute pressure) for 45 min. Next, the ebul-
liometer was loaded with about 500 cm3 of acetone and
the electric resistance located at E (see Fig. 1) was turned
on; boiling of acetone was maintained under recircu-
lation for 45 min at atmospheric pressure. Thereafter,
acetone was removed and before the equipment cooled,
vacuum was applied at 10 kPa (absolute pressure) for
45 min. Finally, the ebulliometer was closed at negative
pressure (taking manometric pressure as reference), and
dry nitrogen was introduced using a separate line1 until
a pressure of about 150 kPa was reached. Before loading
the equipment with the substances to be studied, pres-
sure was reduced to 101 kPa; this enables to introduce
the substances without contamination. Next, the equip-
 ment was closed and charged with dry nitrogen, ready to from VLE data and using Eq. 1 as was previously indi-
operate at moderate pressure1,2,5,16. cated and by using the properties of Table 2. Literature
The equilibrium still typology is of those in which data1,20,21 were employed to obtain the Antoine con-
both phases are recirculated. The ebulliometer oper- stants (see Table 2). A moderate positive deviation
ates dynamically by using the vapor lift pump effect. from Raoult’s Law can be observed, probably due to a
The equipment is made to work in co-currents flow, molecular association via hydrogen bonds.
and thus, the equilibrium condition depends on the
contact time between the non-miscible phases: the Table 1 Experimental VLE data for binary systems at 0.6
mass transfer process is a function of residence time. In MPa. Calculated values of second virial coefficients and
other words, the equilibrium condition does not only activity coefficients of the liquid phasea
depend on constant temperature and pressure. Con- T
x1 y1
B11 B22 B12
γ1 γ2
sequently, the statistical value of these two properties K L/mol L/mol L/mol
methyl acetate (1) + 1-butanol (2)
cannot be taken as the equilibrium criterion; the con- 452.24 0.000 0.000 -0.4855 -0.5675 1.00
stant composition of both phases must be also verified. 448.61 0.034 0.108 -0.4958 -0.5855 -0.5426 1.05 1.00
For this ebulliometer, the input/output flow of each 448.17 0.036 0.116 -0.4970 -0.5877 -0.5442 1.07 1.00
447.31 0.044 0.140 -0.4995 -0.5921 -0.5474 1.08 1.00
phase was evaluated under different operating condi- 446.09 0.054 0.175 -0.5031 -0.5985 -0.5520 1.12 1.00
tions and with different binary mixtures16. With a flow 444.77 0.068 0.213 -0.5070 -0.6056 -0.5570 1.11 1.00
around 25 cm3/min it was possible to obtain a com- 444.40 0.072 0.224 -0.5081 -0.6076 -0.5584 1.11 1.00
443.38 0.084 0.251 -0.5111 -0.6132 -0.5624 1.08 1.01
position in each phase that remained practically con-
441.51 0.100 0.290 -0.5168 -0.6236 -0.5697 1.08 1.02
stant when the renovation time was greater than 75 439.49 0.120 0.343 -0.5230 -0.6353 -0.5778 1.11 1.02
min. For this reason, the mixtures studied in this work 436.74 0.151 0.416 -0.5317 -0.6517 -0.5891 1.12 1.00
were kept at boiling conditions for 90 min to ensure the 435.16 0.166 0.437 -0.5368 -0.6615 -0.5957 1.10 1.03
433.67 0.186 0.476 -0.5416 -0.6710 -0.6021 1.10 1.02
stationary state. Once the steady state was reached
For this ebulliometer, the input/output flow of each phase was evaluated underthe different
433.02 0.192 0.486 -0.5438 -0.6752 -0.6049 1.11 1.03
vapor and liquid phase were both sampled. 16 Next, the 3 427.78 0.259 0.599 -0.5614 -0.7108 -0.6284 1.11 1.01
operating conditions and with different binary mixtures . With a flow around 25 cm /min 425.72it was
equilibrium was disturbed by adding one of the sub- 0.288 0.635 -0.5686 -0.7257 -0.6380 1.11 1.01
possible tostances
obtain atocomposition in each phase that remained practically constant 423.03
when 0.323
the 0.678 -0.5782 -0.7459 -0.6509 1.11 1.01
the mixture in the ebulliometer. 417.51 0.418 0.761 -0.5987 -0.7907 -0.6786 1.07 1.03
A digital
renovation time recorder
was greater than 75Dostmann
min. For thisElectronic
reason, the GmbH
mixtures p655
studied in this 417.11
work were0.425 0.762 -0.6002 -0.7941 -0.6807 1.07 1.05
and conditions
kept at boiling two Pt100 forprobes
90 min with ±0.03
to ensure the K uncertainty
stationary werethe steady415.14
state. Once 0.456 0.784
state was -0.6078 -0.8114 -0.6911 1.06 1.07
employed. The calibration of the system was done by 413.56 0.492 0.802 -0.6140 -0.8257 -0.6996 1.04 1.10
reached the vapor and liquid phase were both sampled. Next, the equilibrium was disturbed by 0.801
413.19 0.501 -0.6155 -0.8291 -0.7017 1.03 1.14
Dostmann Electronic GmbH. Proper operation of the 410.07 0.578 0.837 -0.6281 -0.8588 -0.7191 1.00 1.22
adding oneprobes
of the substances
installedtointhethe
mixture in the ebulliometer.
equipment was verified by mea- 406.15 0.659 0.880 -0.6444 -0.8988 -0.7420 1.00 1.26
suring
A digital the boiling
recorder Dostmannpoint of distilled
Electronic GmbH p655 water. Pressure
and two Pt100 probes with 404.97
±0.030.690
K 0.893 -0.6495 -0.9115 -0.7491 0.99 1.29
402.47 0.753 0.914 -0.6604 -0.9394 -0.7645 0.98 1.41
was controlled with a pressure regulating valve (Binks
uncertainty were employed. The calibration of the system was done by Dostmann 401.16 Electronic
0.790 0.926 -0.6662 -0.9546 -0.7728 0.98 1.49
MFG Co.) included in the nitrogen supply line. Pres- 400.18 0.812 0.937 -0.6707 -0.9662 -0.7791 0.98 1.47
GmbH. Proper operation of the probes installed in the equipment was verified by measuring the
sure was measured with a digital transducer 8311 from 397.34 0.895 0.963 -0.6837 -1.0013 -0.7978 0.98 1.70
boiling point of distilled
Burket Fluidwater.
ControlPressure was controlled
Systems, with anwith a pressure
operating regulating valve
range (Binks
396.45 0.920 0.971 -0.6879 -1.0128 -0.8038 0.98 1.81
396.02 0.931 0.974 -0.6899 -1.0184 -0.8067 0.98 1.91
MFG Co.) from
included0.0 to 4.0 MPa (uncertainty ±0.004 MPa).
in the nitrogen supply line. Pressure was measured with a digital transducer
395.56 0.943 0.977 -0.6921 -1.0244 -0.8099 0.98 2.08
8311 from A calibration
Burket curveSystems,
Fluid Control of composition vs. density
with an operating rangehad from 0.0 to395.01
4.0 MPa0.956 0.984 -0.6947 -1.0318 -0.8137 0.99 1.91
been previously obtained at 298.15 K for the systems 394.91 0.961 0.986 -0.6952 -1.0331 -0.8144 0.98 1.89
(uncertainty ±0.004 MPa).5 394.33 0.974 0.990 -0.6979 -1.0409 -0.8184 0.99 2.07
of this work . Mole fraction (x i) vs. density (ρij) data 394.10 0.979 0.991 -0.6991 -1.0441 -0.8200 0.99 2.32
A calibration
were verifiedcurve ofbycomposition
the adequatevs. density had been previously
correlation of the ex-obtained at393.01
298.15 K
1.000 1.000 -0.7043 -1.0592 1.00
5
for the systems of this workThe
cess volumes. . Mole fraction (xi)was
uncertainty vs. density (ρij) data
estimated to be were verified
a by the uncertainties U(0.95 level of confidence) are:
Expanded
less thanof0.003
adequate correlation in mole
the excess fraction
volumes. of vaporwas
The uncertainty phase.
estimated to be less thanU(T)=0.03
0.003 K, U(p)= 0.004 MPa, U(x1)=U(y1)=0.003
in mole fraction of vapor phase.
VLE data were tested for thermodynamic consisten-
RESULTS AND DISCUSSION cy by using the point-to-point test of Van Ness et al.7
Results and discussion VLEResults
data were indicate
testedthatfor the experimentalconsistency
thermodynamic data for MA1B by using the point-
Treatment Treatment
and Prediction of and VLEPrediction
data of VLE data and7 EA1B binary systems at 0.6 MPa satisfy the Fre-
The VLE
The VLE data
data T-x T-x1-y1 forandMA1B
1-y1 for MA1B EA1B at and0.6EA1B
MPa are atVan 0.6Ness
shown
al. Results
etdenslund
in Table
indicate that the experimental data for MA1B and EA1B
1. The et al. criterion. Results of the Redlich-Kister
8

MPa are shown in Table 1. The activity coefficients of MPa satisfy

at 0.6 and Herington test were
the Fredenslund al.8 criterion.D=75.97
et respectively: Results of > 10%
the Redlich-Kister
activity coefficients of the liquid phase (γi) for each system were determined by using the
the liquid phase (gi) for each system were determined and ABS(D-J)=53.36 > 10 for MA1B; D=59.50 > 10% and
following equation:
by using the following equation: test were respectively:
ABS(D-J)=46.83 D=75.97 > 10>for 10%
EA1B.and Therefore
ABS(D-J)=53.36
both system > 10 for MA1B; D
  p o − p  v Land
 fails with area
ABS(D-J)=46.83 > 10test.forInEA1B.
addition, bibliographic
Therefore data were
both system
3-5
fails with area te
yp  p   po B i
 i  (1)
γ i = i o exp  2∑ y j Bij − ∑∑ yi y j Bij  − i ii +   (1) employed
3-5 for data verification of this work, and for this
xi pi

 RT  j i j


RT RT bibliographic
 purposedata all
weredataemployed for datatoverification
were correlated a polynomial of equa-
this work, and for t
 
tion presented
data were correlated in a previous
to a polynomial equation 16
as follows,in a previous paper16 as f
paper presented
The virial state equation truncated at the second term was employed and the second virial
The virial state equation truncated at the second term
x1
x1 = x1 (1 − x1 )∑ Ak ZT
17 k
Bij) were obtained
(Bii,employed
coefficientswas and the bysecond
means ofvirial
the Hayden and O’Connell
coefficients (Bii, Bij) method (see y1 − Table with ZT =
were obtained byofmeans of the Hayden and O’Connell 17 x1 + R T (1 - x1 )
1). The liquid molar volumes pure compounds were estimated from the equation of Yen and k =0

method (see Table 1). The liquid molar volumes of pure

Woods18. Table 1 includes the γi values calculated from VLE data being Ak and
Eq. 1Ras parameters. The fitting curves and(2)
and using T adjustable
was experimental dat
compounds were estimated from the equation of Yen1,20,21 being Ak and RT adjustable parameters. The fitting
previouslyand
indicated and by using the properties of Table
Woods18. Table 1 includes the gi values calculated 2. Literature data were
Figs. 2 and 3. employed
A significant
curves compressive
and experimental dataeffect can beinobserved
are shown Figs. 2 and as a consequence
to obtain the Antoine constants (see Table 2). A moderate positive deviation from Raoult's Law
pressure. Data from this study agrees with that in literature3-5.
can be observed, probably due to a molecular association via hydrogen bonds.
Next, considering the excess Gibbs function,
APRIL- JUNE 2018 | 99
E
5 G
3. A significant compressive effect can be observed as 0 .5

a consequence of the applied pressure. Data from this

study agrees with that in literature3-5.
VLE data were tested for thermodynamic
0 .4
consistency by using the point-to-point test
Table 1 Continueda Van Ness et al.VLE
7 data indicate
Results were tested for experimental
that the thermodynamicdataconsistency by using
for MA1B and EA1B the point-to-p
binary system
T B11 B22 B 7
Van
γ1 Ness etthe
al. Results indicate 8that the experimental data for MA1B and EA1B bin
0 .3
x1 y1 at120.6 MPa satisfy
γ2 VLE Fredenslund
data were et al.
tested for criterion. Results
thermodynamic of the Redlich-Kister
consistency and Heringto
by using the point-to-point test

y1 - x1
K L/mol L/mol L/mol
al.8experimental
7
ethyl acetate (1) + 1-butanol (2) test at 0.6
were MPa satisfy
Van Ness
respectively: al.the Fredenslund
Results
etD=75.97 > indicate
10% and etABS(D-J)=53.36
that the criterion. Results
data of MA1B;
for MA1B
> 10 for theand
Redlich-Kister
EA1B an
binary >syste
D=59.50 10
0 .2
at 0.6 MPa satisfy the Fredenslund et al.8 criterion. Results of the Redlich-Kister and Hering
452.24 0.000 0.000 -0.6539 -0.5675
test were
1.00
and ABS(D-J)=46.83respectively:
> 10 forD=75.97 > 10% and
EA1B. Therefore ABS(D-J)=53.36
both > 10
system fails with fortest.
area MA1B; D=5
In additio
451.47 0.008 0.018 -0.6566 -0.5712 -0.6220 1.13 1.00
test were respectively: D=75.97 > 10% and ABS(D-J)=53.36 > 10 for MA1B; D=59.50 > 10
3-5
451.31 0.013 0.025 -0.6571 -0.5720 -0.6227
and
bibliographic ABS(D-J)=46.83
0.97 1.00
dataVLE
were > 10
dataemployed
were 0 for
. 1 for
tested EA1B.
for data Therefore
verification
thermodynamic both
of
consistency thissystem
work,
by using fails
and with
thefor thisarea test.
purpose
point-to-point testa
450.82 0.018 0.044 -0.6588 -0.5744 -0.6247 1.24 1.00and ABS(D-J)=46.83 > 10 for EA1B. Therefore both system fails with area test. In additi
3-5 7 16
data were correlated
Van Nesstoetadata
bibliographic polynomial
were employedequation presented in a data previous paper and as follows,
employed for dataverification
verification for of this
work,work, and forandthisfor this
450.66 0.023 0.049 -0.6594 -0.5752 -0.6253 1.08 1.00 al. Results indicate that the experimental MA1B EA1B binary system
450.11 0.034 0.073 -0.6613 -0.5779 -0.6276 1.10 1.00bibliographic3-5 data were for data of this purpose
8
at 0.6 MPa satisfy the Fredenslund et al. criterion. Results of the Redlich-Kister and Heringt
16
449.13 0.050 0.108 -0.6648 -0.5828 -0.6317
data were
1.13 1.00
ydata correlated
were correlated to aAtopolynomial
( )∑
0
ak polynomial 0 . 2 equation
equation 0 presented
0 . 4Z = presented
x01. 8 in ain a previous
1previous paper16 paper
as as follo
(2)follows,
1 − x1 = x1 1 − x1 k ZT with
0 .6
447.81 0.069 0.147 -0.6695 -0.5896 -0.6373 1.14 1.01
446.97 0.089 0.180 -0.6726 -0.5939 -0.6409 1.10 1.01
test were respectively:
k =0
D=75.97 > 10% and
T
xABS(D-J)=53.36
1 x 1 + R T 1 - x1 ( > )
10 for MA1B; D=59.50 > 10
x1 x
445.99 0.105 0.214 -0.6761 -0.5991 -0.6452 1.13 1.01 y − x = x (1 −
y1 − x1 = x1 1 − >x1 10 for x( ) A)∑∑kk
Z
Ak ZEA1B. with
with both Z =
Z x=system 1 (2) (2)
+(1R- Tx1 )1 (- x1) at data are shown
and ABS(D-J)=46.83
1 1 1 1 k T
Therefore T
fails with area test. In additio
445.12 0.123 0.245 -0.6793 -0.6037 being A1.12
-0.6490 k and1.01RT adjustable
Fig. 3 parameters.
Experimental k =0
k = 0 The
T
points fitting −x1) vs.Txand
of (y1curves + R TEA1B
1 x1experimental
1 for ( )
3-5
444.25 0.139 0.268 -0.6825 -0.6084 -0.6528 1.11 1.02bibliographic0.6 R
Ak and data
MPa. were employed
Literature datafor data fitting
at The
70.5 verification
( ) and ofand
94.0 (this)experimental
work,
kPa byand fordata thisare
purpose
443.27 0.158 0.298 -0.6861 -0.6138 Figs. 2being
-0.6572 1.10 3.Abeing
and 1.03 A
k
significant
and R Darwish
T
adjustable
T compressive
adjustable and
parameters.
parameters.
Al-Khateib effect3, can The
101.3 be(fitting
observed
) kPa
curves
curves
by as a and
Ortega consequence
et experimental
al. 4 16
of the shown
applie
data a
441.18 0.202 0.365 -0.6940 -0.6255 -0.6666 1.10 1.04
data were correlated to a polynomial
Figs. 2 and 3. A significant compressive effect can equation presented be in a previous paper as
observed as a consequence of the appl follows,
and 0.3 ( ) MPa by Susial et al. 3-5
with fitting curves.
437.30 0.284 0.478 -0.7089 -0.6483 pressure. Data
-0.6848 Figs.
1.11 1.05 from
2pressure. this
and 3. A study
significant agrees with
compressive that in literature
effect
5
can .
be observed
x1 as a consequence of
∑
3-5
x1 (1 −study
x1 ) A
k
436.66 0.301 0.506 -0.7114 -0.6522 -0.6878 1.12 1.03 Data from
y1 − x1 = this agrees
k ZT withwith that inZliterature
T = . (2)
Next, considering the excess Gibbs function, x 1 + R T (1 - x
3-5
1 )
434.05 0.363 0.570 -0.7218 -0.6685 -0.7005 1.10pressure. Data
1.06 Next,
from
Next, thisconsidering
study
considering theGibbs
theagrees
excess excess
with Gibbs
that
k =0
function, function, .
in literature
431.25 0.433 0.642 -0.7331 -0.6869 -0.7146 1.10 1.07
being Ak and RT adjustable E parameters. The fitting curves and experimental data are shown
1.08 1.10Next, considering the G excess
= x1 lnGγ1Gibbs function,(3)
429.17 0.496 0.691 -0.7417 -0.7010 -0.7253 E

427.48 0.565 0.738 -0.7488 -0.7129 -0.7341 1.05 1.13Figs. 2 and 3. A significant RT compressive
+=xx2 lnln γγ12+ xcan
1 effect 2 ln γbe

2 observed
(3)
(3) as a consequence of the appli
RT
426.96 0.584 0.749 -0.7527 -0.7195 -0.7390 1.05 1.16 E 3-5
pressure. Data from this were
study G with that in literature .
agrees
425.24 0.649 0.790 -0.7584 -0.7292 the activity
-0.7462 coefficients
1.02 1.20the activity were
the correlated
coefficients
activity with
correlated
coefficients the 1 following
with
= xwereln γ1the 2 ln thermodynamic
+ xfollowing
correlated γ 2 withthermodynamic
the
(3) fol- models: thetheWilson
models: Wilso
424.6 0.674 0.808 -0.7612 -0.7340 -0.7497 1.02 1.20
Next, considering
lowing the
23thermodynamic excess RT Gibbs function,
models: 24 the Wilson 22 model,
423.98 0.701 0.829 -0.7639 -0.7387 model, 1.02
-0.7531 the NRTL model, model
1.18
23 the NRTL and model and the UNIQUAC
the23 UNIQUAC 24
model. model. 24
423.83 0.709 0.827 -0.7645 -0.7398 -0.7539 the
1.01 activity
1.24 the NRTL
coefficients were model 22 and
correlated E the UNIQUAC
with the following model. thermodynamic models: t
In the Wilson 22 model the
G 22 excess Gibbs function(3) is represented by the follow
423.49 0.727 0.835 -0.7660 -0.7424 -0.7558 In the
1.00 1.27 Wilson modelIn23
the Wilson the excess model RT
Gibbs
= x1the +function
ln γ1excessx2 ln γ 2 Gibbs
24 is represented
function is by the followin
423.07 0.750 0.854 -0.7679 -0.7456 -0.7581 model,
1.00 1.24 the NRTL
equation, model and
represented by the the following
UNIQUAC equation, model.
422.52 0.777 0.863 -0.7703 -0.7499 equation,
-0.7612 0.99 1.33the activity coefficients were22correlated with the following thermodynamic models: the Wilson
0.99 1.31In the Wilson model the excess 24Gibbs function is ∆λrepresented by the
E
422.05 0.798 0.879 -0.7724 -0.7536 -0.7639 G Vj ij 
421.89 0.806 0.887 -0.7731 -0.7548 -0.7648 E model, the NRTL
0.99 G1.28 = − x231 ln (x1 + Λand
model 12 x 2 )the ln(x2 + Λ 21 x1 ) model.
− x2UNIQUAC with Λ ij =
Vj exp −
 i ∆λij RT 
 (4)
RT V
421.30 0.830 0.899 -0.7757 -0.7595 -0.7681 equation, (
In the Wilson model the
) (
0.99 1.33= − x1 ln x1 + Λ12 x 2 − x 2 ln x 2 22 ) + Λ 21 x1 with Λ =
excess Gibbsij function
exp −
is RT


represented
(4)
by the followi
RT Vi  
420.02 0.880 0.925 -0.7814 -0.7698 -0.7755 0.99 1.45

where Ethe Wilson functions Λij depends on the adjustable interaction energy (4) parameters ∆λij a
418.70 0.938 0.957 -0.7874 -0.7807 -0.7831 0.99 1.68equation,
G Vj  ∆λ ij 
418.00 0.964 0.976 -0.7906 -0.7866 where
-0.7873 the Wilson
0.99 1.64 functions
the molar=volumes Λ
− x1 ln x1ijof+Λ( )
depends
pure ( on
12 x 2components, the
− x2 ln x2 +VΛi 21adjustable
andx1 Vj, which )
interaction
with have energy
Λ ij =been exp  −parameters
calculated

 the∆λ
with
 ij an
Yen (a
417.82 0.976 0.983 -0.7915 -0.7881 -0.7883 0.99 1.76 RT 18 Gwhere the Wilson functions Lij depends V
E
onj theV λ ij  RT 
i ∆ad-
the molar Woods
1.00 volumes equation
of pure (
(see
components, Table )
2).
12 2 Vi2 and ( 2 Vj,21which ) haveijbeen  
DlViji calculated  with the Yen an
= − x ln x + Λ x − x ln x + Λ x with Λ = exp −
 the (4)
417.58 0.982 0.988 -0.7926 -0.7901 -0.7898 1.67 justable
RT
1
interaction
1
energy parameters
1
and RT 
417.33 0.990 0.994 -0.7937 -0.7923 -0.7912 where
1.00 the Wilson
1.51 The NRTL functionsmodel23Λappliesij depends on the adjustable
the following equation for interaction
the excess Gibbs energy paramet
free energy,
Woods1.00equation molar volumes 18 of pure components, V and Vj, which
417.02 1.000 1.000 where(see
-0.7951 Table functions
the Wilson
-0.7923 2). Λij depends on the adjustablei interaction energy parameters ∆λij a
the molar volumes have
G 23of pure
E been calculated
 G21τcomponents,
with the Yen and Woods18 equa- calculated
G12 τ12 Vi and Vj, which have been ∆ g ij with
a Theare:
Expanded uncertainties U(0.95 level of confidence) NRTL model
the molar volumes
RT
tion applies
= x1(see
x
purethe
x 2 of Table
+ G
21
x
following
components,
+
2). x + G x
equation
 i and
V =for
expthe
withVGj,ij which (−have τ ij )been
α ijexcess Gibbs
τ ij = free
andcalculated
RT
energy,
with the
(5) Yen a
18  1 1
Woods
U(T)=0.03 K, U(p)= 0.004 MPa, U(x1)=U(y1)=0.003 equation
18 The(seeNRTL Table 21 2
2).2). 2 23 12applies
model the following equation
G Woods equation (see Table
E
 G  ∆g ij
where
= x x  α ij =α
for
21 τ
ji is
the
21 a +
G
23
12 τ
non-random
excess Gibbs
12  parameter
free
with of
energy,
G = the(− mixture
exp α ij τ ij ) and
whichτ ij is= associated with
(5) molecu
0 .7
RT The NRTL
1 2
The NRTL model model
 x1 + G21 x 2 x 2 + G12 x1  applies
23 
applies the
the following
ij
following equation
equation for for
the the
excess excess
RT Gibbs
Gibbs free free e
energy,
organization.
E
The adjustable parameters, τij, depend on a temperature function with interact
E ∆g
x1 x2G 21G τ12 
= (exp (− α) τand
where αij=αjienergies G
isGa non-random 21 τ 21
τ parameter GG1212τof
12 the
∆g ij(5)
 mixture which
− α ij τ ij is associated
) τijand= τ with molecul
0 .6 ij
 with
x ijx=between
=∆g   an RT1 2  + + of
21 i-j pair
+xG
1 + G21 x 2
withGijG= exp
molecules. x 2 + G12 x1  =ij ij ij RT ij
RT  x1 parameters,
21 2 x 2 + G12 x1  x RT
0 .5 organization. The adjustable
In the UNIQUAC model24 τij,the
depend
excess on a temperature
Gibbs functionofwith
function is composed interactio
the combinator
where αij=αji is a non-random parameter of the mixture which is associated with molecu
where α
and =α is
residual
a non-random
parts as follow, parameter of the mixture which(5)is associated with
energies ∆gij between
ij ji an i-j pair of molecules.
0 .4 organization. The adjustable
where aij=aji is aparameters,
non-random τij, depend on a temperature
parameter function with interacti
of the mix-
y1 - x1

E E E
organization.
In theenergies
UNIQUAC The
∆gture adjustable
whichan
model
ij between theparameters,
24is associated
i-j
G Gibbs
excess
pair τij,molecular
= with
GComb
of molecules. depend
G Re sid
+function on a(6)temperature
isorganization.
composed of thefunction with
combinatori
0 .3
The adjustable parameters, RT RTt , depend
RT on a tempera-
energies
and residual ∆gas
parts between
ij Infollow,
the
ture an i-j pair
UNIQUAC
function of24interaction
model
with molecules.
the excessij Gibbs
energiesfunction
Dgijisbetween
composed of the combinator
0 .2
and
In residual
the UNIQUAC i-j pair
anparts ofmodel
as follow,Emolecules.
24
G the G excess Gibbs function is composed of the co
E E
G
0 .1 In the UNIQUAC = Comb model
E +
Re
24 sid
E the excess
E Gibbs function
(6)
and residual parts RT RTG GRT
as follow, of the combinatorialComb G Re sid
is composed RT
=
RT
+
RT
and residual
(6) parts
0 as follow, E E E
0 0 .2 0 .4 0 .6 0 .8 1
G G G (6)
x1 = Comb + Re sid (6)
RT RT RT
Fig. 2 Plot of experimental (y1−x1) vs. x1 data for MA1B ( E
GComb Φ  Φ   Z  θ  θ 
) at 0.6 MPa. Literature data at 74.66 ( ) and 127.99 ( = x1 ln 1  + x 2 ln 2  +   q1 x1 ln 1  + q 2 x 2 ln 2  (7)
RT  x1   x2   2    Φ1   Φ2 
) kPa by Susial and Ortega3 , 101.3 ( ) kPa by Ortega
and Susial3 , and 0.3 ( ) MPa by Susial et al. 5 with E
G Re sid
fitting curves. = −q1 x1 ln(θ1 + θ 2 τ 21 ) − q 2 x 2 ln( θ 2 + θ1τ12 ) (7)(8)

RT

xi ri xi q i  ∆u ij 
Φi = ; θi = and τ ij = exp − 
 (9)
xi ri + x j r j xi q i + x j q j  RT 
100 | AFINIDAD LXXV, 582
being Z the coordination number, Φ the molecular fraction of segments and θ the molec
 E
GComb Φ   Φ   Z  θ  θ 
= x1 ln 1  + x 2 ln 2  +   q1 x1 ln 1  + q 2 x 2 ln 2  (7)
RT  x1   x 2   2    Φ1   Φ2 
Table 2 Properties of literature19 and fromE this work.
Table 2 Properties of literature19 and from this work. G Re sid
= −q1 x1 ln(θ1 + θ 2 τ 21 ) − q 2 x 2 ln(θ 2 + θ1τ12 ) (8)
Tc19 Pc19 RD19 µ19 Zc19 A B RTC Vi r q
KTc MPaPc ÅRD Zc A B C Vi r q
19 19 19 19 19
Dm xi ri L/mol xi q i  ∆u ij 
K MPa Å D Φi = ; θ iL/mol
= and τ ij = exp −  (9)
methyl acetate
methyl acetate
xi ri + x j r j xi q i + x j q j  RT 


506.80
506.80 4.694.69 2.996
2.996 1.679
1.679 0.254
0.254 6.7347
6.7347 1529.38
1529.38 6.59
6.59 2.8042 2.576
2.8042 2.576 0.0798
0.0798
being Z the coordination number, Φ the molecular fraction of segments and θ the molecular
ethyl acetate
ethyl acetate
523.25
523.25 3.833.83 3.468
3.468 1.781
1.781 0.252
0.252 7.0337
7.0337 surfaces. τij -22.19
fraction of1869.43
1869.43 are the adjustable
0.0985parameters
0.0985 and 3.116
3.4786
3.4786 ij represents the average interaction
∆u3.116
1-butanol
1-butanol
energy of molecules. The volume and area of groups of van der Waals are used to calculate r
562.93 4.4127
562.93 4.4127 3.251
3.251 1.66
1.66 0.259
0.259 6.4296
6.4296 1261.325
1261.325 106.43 0.0936 3.4543 3.052
and q, volume and area106.43
parameters, 0.0936 3.4543
of the UNIQUAC model3.052
(see Table 2).
o B
log10 Pi /kPa = AThe
− adjustable parameters in each of these models (Table 3) were obtained using the
T/K − C 25
Nelder and Mead method . Deviation in the sum of the squares of activity coefficient was
Table 3 Correlation parameters of GE/RT vs. x1, mean absolute deviations and standard deviations
minimized for both substances during optimization of the parameters. For MA1B the NRTL
Table 3 Correlation parameters of GE/RT vs. equation x1, mean23absolute
yieldeddeviations
the lowestand
meanstandard deviations
absolute deviations (MAD) as well as standard deviations
model
model parameters
parameters MAD(y
MAD(y1)1) MAD(T)/K
MAD(T)/K SD(γ SD(γ) SD(γ
1 1)
) SD(G
SD(γ
2 2)
E
SD(G/RT)
E
/RT)
methyl acetate (1) +(SD) between
1-butanol (2) experimental
at 0.6 MPa and calculated values for temperature and vapor compositions.
Wilson22 methyl
Δλ12 = -2824.4 J·mol -1 acetate (1) + 1-butanol
Δλ21 = 6962.7 J·mol-1the (2) at 0.6 MPa 2.04
0.019 0.11
However, Wilson model22 yields the best 0.14 0.015
correlation for EA1B, with the lowest MAD in
Wilson22 Δλ12 = -2824.4 J·mol-1 -1 Δλ21 = 6962.7 J·mol-1-1 0.019 2.04 0.11
NRTL (α = 0.47) 23
Dg = 5996.6 J·mol Dg = -2183.6 J·mol 0.018 2.04 0.10 0.120.14 0.018 0.015
NRTL (α = 0.47) 23 12 -1
∆g12 = 5996.6 J·mol -1 21
both temperature
∆g21 = -2183.6 -1
J·mol -1 and vapor phase
0.018 2.04mole fraction.
0.10 0.12 0.018
UNIQUAC (Z =2410) Δu12 = 3935.6 J·mol
24
-1 Δu21 = -2104.9 J·mol-1 0.020 2.07 0.11 0.14 0.016
UNIQUAC (Z = 10) Δu12 = 3935.6 J·mol Δu21 = -2104.9 J·mol 0.020 2.07 0.11 0.14 0.016
ethyl acetate (1) + 1-butanolTemperature, pressure, vapor phase composition and the calculated activity coefficients
(2) at 0.6 MPa
ethyl-1acetate (1) + 1-butanol-1(2) at 0.6 MPa
Wilson 22
Δλ12 = -1693.4 J·mol Δλ21 = 4206.9were J·mol
compared 0.013
with the 0.53
theoretical 0.07 0.09
predictions of VLE 0.010
obtained 9
Wilson 22 -1
Δλ12 = -1693.4 J·mol -1 -1
0.010 with the ASOG model , the
NRTL (α = 0.47)23 Dg = 4396.8 J·mol Dg21Δλ=21-1896.1
= 4206.9 J·mol -1
J·mol-1 0.013
0.013 0.53
0.61 0.07
0.07 0.080.09 0.012
NRTL (α = 0.47)23 ∆g12
12 = 4396.8 J·mol
-1
mod.
∆g21 = -1896.1 UNIFAC-Lyngby
J·mol 0.013 model10 proposed
0.61 by
0.07 Larsen
0.08 et al., 0.012
the original UNIFAC model 8
with
UNIQUAC (Z =2410) Δu12 = 3185.2 J·mol
24 -1
Δu21 = -1922.2 J·mol-1 -1
0.014 0.57 0.07 0.09 0.011
UNIQUAC (Z = 10) Δu12 = 3185.2 J·mol-1 Δu21 = -1922.2 J·mol 0.014 0.57 0.07 0.09 0.011
Hansen et al. parameters11 and the mod. UNIFAC-Dortmund model12 proposed by Gmehling et
E
  Z   MAD( θ 1
n
 (F − F )
al.∑1 exp cal ;
n 2
GComb Φ  Φ Fexp − Fθ
 +   q1 x1 ln F )1 =n+− 2q∑
E
cal2; SD( F ) = F ≡ y1 , T , p, γ1 , γ 2 , G /RT
= x1 ln 1  + x 2 ln 2 21 x 2 ln  n(7)
−2
RT G E  x1   Φ   x 2  2  Φ
 Φ 2   Z    1 θ1  Φ
  2θ2  In the group contribution models the activity coefficient of the liquid phase are
Comb
= x1 ln 1  + x 2 ln x1 ln 4 Results
 +   q1Table ln  using
 + q 2ofx 2predictions group(7)
contribution models and PR-EOS
RT E  x1   x 2 mod. UNIFAC-
2   Φ1 
UNIFAC-1991 8,11  Φ 2 
mod.
calculated
UNIFAC-
with the following equation:
ASOG-19799
G Re sid Lyngby 10
Dortmund 12
PR 13
PRSV 13,15
+ ln γ i PRWS
Comb
Re sid 13,14
PRSVWS13-15
= E− q1 x1 ln(θ1 + θ 2 τ 21OH/COOC
) − q 2 x2 ln(θ 2 +OH/CCOO
θ1τ12 ) (8) Differences in lnthe models arise from the interpretation
OH/COO γ i = ln γi (10)
(8)
OH/CCOO
RT G Re sid methyl acetate (1) + 1-butanol (2) atin
0.6ineach
MPa
= −q x ln(θ + θ τ ) − q x ln(θ + θ τ )
MAD(y11)1 1 2 0.011
21 2 2 (8)
20.0281 12
given
Differences
0.023
one about
the models
0.046
the
arise from thecombinatorial
0.007
interpretation givenand
0.014
residual
in each one about the
0.012 0.013
RT 9 9 model the combinatorial part is obtained
xi ri MAD(T)/K xi q i 0.93 3.32  ∆u ij  2.76 5.93 contributions. In the ASOG model the combinatorial
combinatorial and residual contributions. In the ASOG
0.82 0.47 0.37 0.45
Φ i E= ; θi = and τ ij = exp − ∆u (9) Flory-Huggins equation
GComb xi rΦ + x r  xi1ri MAD(p)/MPa
Φ  Φ x 2 q  +xZ xqi q0.011
i  θ1  0.044   θRT 
   −2 ij 12.78

0.036by using the 0.078 part is obtained by using the Flory-Huggins equation
q1 x1 ln and + q17.81
ln 1) θi i i=+  j  j 4.12
=i i x=1 lnj j  + x;2MPD(γ exp
τ2ij x=2 ln (7)
(9) 32.96
RT xiri x+1 xj r j MPD(γ x j q j  Φ1  4.13  Φ 2 RT
 2x) 2  xi q2i +12.75



3.70 4.31

 c

 c
eing Z the coordination number, Φ the molecular fraction of segments and θ the molecular
ethyl acetate (1) + 1-butanol (2) at 0.6 MPa iln γ
Comb
= 1 + ln
 ϑi


c 
−
ϑi
(11) (11)
being Z the coordination number,
MAD(y ) Φ the molecular fraction of segments and θ the molecular ∑ ∑
c
E  x ϑ  x jϑ j
G Re sid 1 0.012 0.020
(9) 0.012 0.023 0.008 0.008 j j
 0.012 0.006
raction of surfaces. τij are the= −adjustable 1 x1 ln(θ1 + θ 2 τ 21 ) − q 2 x 2 ln ( ∆u represents ) 0.38 
parameters and the average interaction
j j
qMAD(T)/K 1.17 θ 2 + θ12.16
τ12
ij (8)0.70 2.39 0.74 0.55 0.26
beingRTZτijthe
fraction of surfaces. coordination
are MAD(p)/MPa
the adjustable number,
parameters
0.011 andthe
F ∆u molecular
represents the average interaction
0.028 ij c
0.009being ϑj the0.031
number of atoms (non hydrogen atoms) in the molecule j. The residual part is

calculate rr
nergy of molecules.
energy
fraction
The volume
of molecules.
of segments and area
MPD(γ ) and ofq10.74
1 the
groups
5.33 molecular
of van fraction
der Waals
Waalsare
17.45 of areused used to
6.16
to calculate
being J the number of atoms (non hydrogen atoms) in
21.18 c
xi ri The volume MPD(γ ) and xi qarea
i
of groups of van der 4.13  ∆u ij  3.51 determined as4.2 follow,
j
Φ i = surfaces. ; tij areθ i the adjustable parameters and Du τ ij = exp −  the molecule j. The residual part is determined as follow,
2
= and (9)
 ij 100 n Fexp − Fcal
nd q, volume and area
and q, volume ri +parameters,
xiand xarea
j r j parameters, of the qi UNIQUAC
xi of +thex j qUNIQUAC model model (see(seeTable 2).2). ∑ RT
represents the average j
interaction energy of Table
mole-   F) =
MPD(
n−2 1 Fexp
= ∑ ϑ ki  ln Γk − ln Γk 
Re sid i
being The adjustable cules.parameters
The volume inΦeach
and of these
area of groups models of van (Tableder 3)3)were
Waals andobtained using
using lnthe theγi (12) (12) 7
Z the The adjustable
coordination parameters
number, in each
the molecular these models
fraction of(Table
segments were obtained
θ the molecular  
are used 25 to calculate
25 r and q, volume and area param-
Nelder and
fraction of Mead
Nelder surfaces.method
and Mead method
τij are . Deviation
the . Deviation
adjustable inparameters
thein the sumsum of the
of∆u
andTable thesquares
squaresofofactivity
ij represents
activity coefficient
coefficient was(non
was 26
eters, of the UNIQUAC model (see 2). being ϑthe average
ki the number
being interaction
ofJatoms theRenumber
ki ln γ i ln γ = ∑
sid hydrogen
=ϑ
of
ϑ
 ln
∑ki   k i atoms
Re sid
lnΓ Γ
atoms)
−
− ln
ln Γ Γ (non
 
i

i
group k in atoms)
in hydrogen
(12) (12)
molecule i, Γk the gro
minimized for both substances during optimization of the parameters. For MA1B the NRTL
i ki k
k  k
minimized for both substances
The adjustable during optimization
parameters
energy of molecules. The volume and area of groups of van der activity in of
each the
of parameters.
these models
Waals are For MA1B in the
group
used toofcalculate
coefficient NRTL k
Re sid
ln γ i k and
group r= ∑Γϑkki the
in molecule i 
 ln Γactivity i, G
k − ln Γkkcoefficient
 the group (12)of group k coef-
activity in a standard state (pu
being ϑki the number of atoms (non  hydrogen i atoms)  in group k in molecule i, Γk the group
quation equation
23 23 (Table 3) were obtained using the Nelder and Mead
yielded yielded the lowest mean absolute deviations (MAD) being
as ϑ
well the
as number
standardficient of of
atoms
deviations group (non k and
hydrogen G the
atoms) activityin group coefficient
k in molecule of i, Γk the gro
and q, volume andthearealowest
method parameters,
25
mean of
. Deviation
absolute
theinUNIQUAC deviations (MAD)
the sum ofmodel (see
the squares
as
beingTable
well
component2).
ϑkiofthe
kias standard
ac-number
i). Thecoefficient
activity ASOG
group
of atoms
deviationsmodel
k ofin(nongroup considers
k and Γki the atoms)
a standard
hydrogen
the group
activity
state
k
coefficient
(pure
in
activity
group of group
component k
coefficient
in kmolecule
in a standard
i).
is given
The i,state
by the Wils
Γk (pure
the group
i
(SD) between experimental and calculated values for temperature activity and vapor
coefficient compositions.
of group k and Γ the activity coefficient of group k in a standard state (pu
SD) between experimental
The adjustable tivityparameters and calculated
coefficient inwaseach ofvalues
minimized these models for
for temperature
both equation
substances
(Table and
wereasvapor
follows,
component compositions.
ASOG i). The ASOG
model model k considers the
considers the group activity coefficient group activity coefficient is given by the Wilson
22 activity3) coefficient obtained
of group usingk andthe Γki the activity coefficient of group k in a standard state (pure
However, theduring Wilsonoptimization
25model
22 yieldsofthethe bestparameters.
correlation for For EA1B,
component
MA1B with i).theThe
equationlowest
isASOG
asgiven MAD
follows, model
by the in considers
Wilson equation the group activity as follows, coefficient is given by the Wils
However,
Nelder and the Mead
Wilson model
method yields the
. Deviation best correlation
23in the sum of the squares
forcomponent
EA1B, of with
activitythe coefficient
lowest MAD wasin  the  X l alk
k = 1 − ln ∑ X l a lk  − ∑ X a
the NRTL equation yielded i). The ASOG model ln Γconsiders group  activity coefficient(13) is given by the Wilson
both temperature and vapor phase mole fraction.the lowest mean abso-
equation as follows,
l ∑
ln Γ = 1 − ln ∑ X a − ∑ X a
l lk
(13)
oth temperature
minimized and
for both vapor
substances phase mole
during fraction.
optimization of the parameters. For MA1B the NRTL   ∑
k l lk
l   X a m lm
lute deviations (MAD) as well as standard deviations
equation as follows,  l l m lm (13)
Temperature, pressure, vapor phase composition and the calculated activity coefficients  
m m
(SD) between experimental and calculated values for X l alk
∑ lk  ∑
23
Temperature,
equation yielded the pressure,
lowest vapor phase composition
mean absolute deviations (MAD) and the as calculated
well Xas
where activity
standard coefficients
deviations
where Xl represent
represent the ln
group Γ
the =
9 k group 1 − ln
 fraction
fraction
 X a
of ofgroup
l group
 − ll in in the liquid
alkthe
solution. (13)
In the above expression
were compared with the theoretical
temperature and vapor predictions
compositions. of VLE However,obtained with
thel the ASOG where model
Γk X ,− ln
the
= l19represent l the group X
∑
l l fraction X mliquid
alm of group solution. Inthe
l in the above expressi
(SD)compared
between with experimental and 22calculated (Eq. 13) thelnsummations

∑
extend

Xover
l a lk all ∑
− groups
l ∑
 and alk, alm are the (13) group interaction parameters.
were the theoretical
Wilson
mod. UNIFAC-Lyngby model predictions
model10yields proposed thevaluesbest of VLE
by Larsen
for temperature
correlation obtained
et al., the for with
(Eq. and
13)the
EA1B,
original
vapor
the ASOG
UNIFAC
compositions.
summations
liquidmodel
Themodel
8 , the
extend
solution.
UNIFAC8,10-12 with  over
l In allthe
models are generally
groups above and X mexpression
maalm lk, alm are the (Eq. group 13)interaction
the parameter
m based on the equations of the UNIQUAC model
However,
mod. with
the Wilson model
UNIFAC-Lyngby the lowest
2210
model11 yields MAD
proposed in
the best both temperature
correlation
byUNIFAC-Dortmund
Larsen et al.,forthe and
EA1B,wherevapor
original X
with represent
the summations
for lowest
lUNIFAC
The UNIFAC the
model MADgroup
8,10-12 8 extend
fraction
in being
with
models of over
group all l groups
in the and
liquid a , a
solution.
25 areIn the
the above expressi
et ofare forgenerally based on the equations lm of the UNIQUAC mod
12 8
Hansen et al.phase parameters and the mod. model proposed the combinatorial part, the clasical UNIFAC model, lk
mole fraction. where Xl represent thebygroupGmehling
group fraction
interaction groupparameters. l in the liquid solution. In the above expression
both temperature and vapor
11 phase mole fraction. (Eq. 13)combinatorial
for12the the summations part, extend
being over all groups  θ and alk Φ, alm8 model,are the group interaction parameter
ln γ et=for
 Φthe  clasical
Hansen et al. parameters and the mod. UNIFAC-Dortmund model proposed by Gmehling  + q   ln UNIFAC
Z 
 x ∑
Comb
al. Temperature, pressure, vapor phase (Eq. composition
13) the summationsThe extend UNIFAC x  models
8,10-12
ln  i
alkare
 + l −generally
, alm are
i
x l i
based (14) on the
8,10-12 over all groups  2and  the group interaction parameters.
i i i j j
 Φ
The UNIFAC models are generally based on the equations of the UNIQUAC mod
i i i j

l. Temperature,
In the and the calculated
pressure,
group vapor phase
contribution activity
models coefficients
composition
the activity and were compa-
the calculated
coefficient of the activity equations
liquid coefficients
phase of Φthe i 
UNIQUAC
 Z fraction  θ i  model  Φ i for  the combina-
= lnare 8Φ∑
8,10-12 Comb
The UNIFAC ln models are generally based on the equations 8j of the volume
model,UNIQUAC model
θ represents γthe molecular  + qiarea
 surface   ln  (see + liEq. − 9), x j lmolecular
is the (14) fraction
red with the theoretical predictions of VLE obtained for the the
combinatorial torial i
part, part,
9being being
 xfor
i  the for 2the
 clasical clasical Φ iUNIFAC UNIFAC
 isxi model,
 j
were In
compared
the group with the theoretical
contribution predictions
models of VLE obtained with ASOG model , the
calculated with the following model9,the
equation: theactivity coefficient of the liquid phaseforare
(see Eq. 9) and the pure-component lattice parameter, 8 l, function of van der Waals surface
with the ASOG mod. UNIFAC-Lyngby for the combinatorial part, being the clasical UNIFAC model,
10 area (q) and van der
8 Waals  Φ 
volume 
(r).Z 
For the θ 
mod. 
UNIFAC-LarsenΦ  10
model,
Φ   xi ∑ j j
mod. UNIFAC-Lyngby model proposed by Larsen etthe al.,original
the(10) θ represents
original UNIFAC the molecular
model Comb
surface
withln i  area + q  fraction  ln i (see + li −Eq.  i9),  Φx is l the molecular (14) volume fracti
with the model proposed ln γby Larsen +etln γal., UNI-
10 Comb Re sid ln γ i = (14)
alculated following equation: i = ln γ i  Φ   x Z  i 2θ i ψ Φ l i − 
= ln i  + qii  ln γ ln = ln ψ i ∑
i Comb
FAC model 8
with Hansen et al. parameters and
11 12
the ln γ i  + x l j (14)
Comb ii
11 (see proposed
Eq. 9) andbythe pure-component
i
l, is function of van der Waals surfa
Hansen et al.Differences
parameters in and thelnthemodelsmod.Comb UNIFAC-Dortmund model Gmehling  xi the et  2  lattice  parameter,
 Φ i  x   xxi  j

 i+
i 1 − j(15)
j

γ i = ln γ arise +from γ i the12interpretation given

Gme-in each qone about
Re sid i i
mod. UNIFAC-Dortmund lnmodel proposed (10) by represents the represents
molecular surface thearea molecularfraction (see surface Eq. area
9), Φ is fraction
the 10 (see
molecular
θarea model, volume fracti
i
al. combinatorialhling (q) and where
van der Waals volume
ψ represent the modified (r).groupFor the volume mod. UNIFAC-Larsen
fraction. For the mod. UNIFAC-Gmehling 12
et al. contributions. In the ASOG9 model
and residual the combinatorial
θ represents the molecular Eq. is
part 9), F is area
obtained
surface the fraction
molecular (seevolume Eq. 9), Φfraction is the molecular (see Eq. volume 9) fraction
Differences in the models arise from the interpretation (see given Eq.in9)each onepure-component
about the l, is function of van der Waals surfa
byInusing
the thegroup In contribution
the groupequation
Flory-Huggins
contribution
models the models activity the activityof
coefficient co-the andmodel,the
liquid andphase the pure-component
areln γlattice
Comb
lattice parameter,
 ψ i  lattice
  + 1 − l,i is function
ψ i parameter, l, is func-
9 (see Eq. 9) and the pure-component Comb i
= ln parameter, Φi 
(15)of van 10 der Waals surface
ombinatorial and residual efficientcontributions.
of the liquidInphase the ASOG model thewith
are calculated area (q)
theand van part
combinatorial der
tion Waals
isofobtained
van volume
ln der
γi ln(r).
Waals
= δ i + 1For δi −xthe
− surface qi i  mod.
 Z
 lnarea
  xiUNIFAC-Larsen
Φ i  +(q)1 − and  van der
(16) model,
Waals
calculated with thefollowing
followingequation: equation:   area (q) and van der Waals volume volume (r). (r). For
For the the
mod. mod.  2 
UNIFAC-Larsen


UNIFAC-Larsen
 η i  η i 
 10
10 model,
model,
y using the Flory-Huggins equation    ψ iarea  fraction ψ ii
c c where ψ represent themodified
modified ln γ group
Comb
lnvolume  + 1 − fraction. and δ For the mod. UNIFAC-Gmehling
Comb Comb ϑ i Re sid
lnlnγγi i = ln (11)(10)
= 1γ+i ln + ln γ i c
ϑi
−
(10)Comb
i
being η the
 ψ i   xi ψ ii x
molecular =surface the(15)modified molecular volume

∑ ∑ = ln the
ln γ i models,  +residual
c 1 − part ofi the activity
(15) coefficient (Eq.(15)
 model,
 x ϑ fraction. In the UNIFAC
 x jϑ j 12 being ϑki
 from  c j j j  cj  xi  xi
Differences in the  models

Comb
arise ϑgiven
where
 thein
ψ each
representoneinterpretation
the
 ϑ
the number about
of groups the
modified
of type kgroup
in molecule volume fraction.
i,) is replaced by theFor the mod.
solution-of-groups UNIFAC-Gmehling
concept. The
ln γ i = 1 + ln i
− i
where(11)
ψ represent the modified Comb
group  Z   coefficient,
Φ  Φ  12
c  ln γ i is =used i + 1volume
ln δfor i  fraction.
δ i − qactivity  ln i For
+ 1 − thei  mod. UNIFAC-Gmehling
following equation the−group (16)
∑ ∑
9 c
being ϑjcand
combinatorial theresidual
number contributions.
of atoms (non Inx the
hydrogen
ϑ
j j 
ASOGatoms)
x model
ϑ
j j
in the
the combinatorial
molecule
model, j. The part is
residualobtained
part is 2
    i  η ηi 
 j  model,  APRIL-  JUNE 2018 | 101
as follow, equation   j
determined
by using the Flory-Huggins  
k 1 − ln ∑ Θ m
Comb ln Γk = Qsurface 

∑
Θ ξ
ξZmk−fraction

 Φ 
 Φ i m mkand (17)
being η the modified molecular
ln γ i area
= ln δ i + 1 − δ i −m qi  lnm ∑ Θ n+ξ mn 1 − iδ the modified (16) molecular volum
eing ϑjc the number of atoms (non hydrogen atoms) in the molecule j. The residual
Comb part is   Z    2Φ i   ηni Φ i   ηi 
component i). The ASOG model
i ∑ ki  the
considers k ln γ
group
k = ϑ  ln Γ −12ln Γ 
 activity coefficient is given by the Wilson
(12) 21

1 group k in molecule i, Γ∑the

n
n (F −F )2
F ) = in ∑
exp cal E
being ϑki as
equation thefollows,
number of atoms (non hydrogen
MAD(atoms) Fexp − Fcal ; SD( F ) = k 1 group ; F ≡ y1 , T , p, γ1 , γ 2 , G /RT
n−2 1
n−2
activity coefficient of group k and Γki the
 activity coefficient
X a
of group k in a standard state (pure
ln Γ = 1 − ln
l Table 4 Results ∑X a
of predictions using
−
group ∑ l lk
contribution models and PR-EOS (13)
∑
k  lk 
component i). The ASOG model considers
 l the group
 l activity
X m acoefficient
lm
is given by the Wilson
mod. UNIFAC- 8,11 mod. UNIFAC-
equation as follows, UNIFAC-1991
m ASOG-19799
Lyngby10 Dortmund12 PR13 PRSV13,15 PRWS13,14 PRSVWS13-15
OH/CCOO OH/COO
where Xl represent the group OH/COOC
fraction OH/CCOO
of group l in the liquid solution. In the above expression
 
Table 4XResults
l a lk of predictions using group contribution models and PR-EOS
ln Γk = 1 − ln ∑ X l alk  − ∑ methyl
(13) acetate (1) + 1-butanol (2) at 0.6 MPa
(Eq. 13) the summations
MAD(y )
extend
1
over

 all
0.011
l
groups

 l ∑ X maalklm, alm are the group interaction parameters.
and 0.028 0.023
m
0.046 0.007 0.014 0.012 0.013
8,10-12 mod. UNIFAC-Lyngby UNIFAC-1991 mod. UNIFAC-Dortmund
The UNIFAC models are generally based on the equations of2.76 modelASOG-1979
the UNIQUAC 5.93
10 8,11 12 9
MAD(T)/K 0.93 3.32 0.82 PR13 0.4713,15
PRSV 0.37 PRSVWS13-150.45
PRWS13,14
where Xl represent the group fractionOH/COOC
MAD(p)/MPa
OH/CCOO
of group l in the liquid
0.011
OH/CCOO
0.044solution. In the 0.036
above expression 0.078 OH/COO
for the combinatorial part, being for the clasical UNIFAC8 model,
methyl acetate (1) + 1-butanol (2) at 0.6 MPa
MPD(γ 1) extend
(Eq. 13) the summations
MAD(y 1
) over all4.12
groups and alk, alm17.81
0.011 are
0.028 12.78 parameters.
the group interaction 0.023 32.96 0.046 0.007 0.014 0.012 0.013
MAD(T)/K
MPD(γ )  0.93
 12.75    3.32
4.13 3.70 2.76 4.31 5.93 0.82 0.47 0.37 0.45
Φi θi
 Z  based Φi
∑
8,10-12
2 Comb
The UNIFAC ln γ i models
MAD(p)/MPa = ln are generally
 + qi   ln
0.011 on the
 + li − 
equations
 x j l j
 0.044 of the UNIQUAC
(14) 0.036 model 0.078
MPD(γ1)  x 4.12
 2  Φ x
17.81 ethyl acetate (1) + 1-butanol
12.78 (2) at 0.6 MPa
32.96
 i   i  i  j
for the combinatorial part,
MPD(γ
MAD(y being
) for the clasical UNIFAC80.020
12.75 model,
4.13 3.70 4.31
1) 0.012 0.012
(1) + 1-butanol (2) 0.023 0.008 0.008 0.012 0.006
2
ethyl acetate at 0.6 MPa
θ represents the molecular
MAD(T)/K
MAD(y ) surface
 Φi  area
1.17
 0.012 fraction
Z   θi  (see Eq.
 Φ i 2.16
 9),
0.020 Φ is the molecular
0.70 volume
0.012 fraction
2.39 0.023 0.74 0.008 0.38
0.008 0.0120.55 0.006 0.26
∑
Comb 1
ln γ i
MAD(T)/K = ln  + qi  1.17
 ln  + li −   2.16
x jl j (14) 0.70 2.39 0.74 0.38 0.55 0.26
MAD(p)/MPa x  0.011 2 Φ x 0.028 0.009 0.031
(see Eq. 9) and the pure-component
MAD(p)/MPa  i   0.011
  i parameter,
lattice  jl, is function of van 0.009
 i  0.028 der Waals surface 0.031
MPD(γ 1)
MPD(γ 1
) 5.33
5.33 17.45
17.45 6.16 6.16 21.18 21.18
10
area (q) and the
θ represents vanMPD(γ
derMPD(γ
Waals
molecular
2) 2
) volume
surface area(r). For the
fraction
10.74
10.74 (seemod.Eq. UNIFAC-Larsen
9),
4.13Φ is the molecular
4.13 model,
volume
3.51 3.51 fraction 4.2 4.2

(see Eq. 9) and the pure-componentComb ψi  ψ l,

lattice parameter, is function of van der Waals
100 surface
n F
exp − Fcal
∑
ii
x  ln γ i 10
= ln  +1− (15) MPD( F ) =
area (q) and van der Waals volume (r). For the
 mod.
i  xi
UNIFAC-Larsen model, n−2 1
Fexp

where ψ represent the modified ψ 

Comb group
ln γ i = ln i volume
ψi
 + 1 − i fraction. For the mod. UNIFAC-Gmehling12
(15)
where y represent thexi  modified
xi group volume frac- 470

model, 26
450

tion. For the mod. UNIFAC-Gmehling12 model,

where ψ represent the modified group volume fraction. For the mod. UNIFAC-Gmehling12
460

Comb  Z   Φ  Φ 
440

T/K
model, ln γ i = ln δ i + 1 − δ i − qi   ln i  + 1 − i  (16)
(16) 450
 2    ηi  ηi  430

Comb  Z    Φ i  Φi  440 420

being η the modified ln γ i = ln δ i + 1 −surface

molecular δ i − qi  area
 ln  + 1surface
− and  the(16)
being h the modified 2    ηfraction
molecular i  ηi  δarea modified
fraction molecular volume 0 0.2 0.4 0.6 0.8 1

T/K
x1 , y1

fraction. In the and d the

UNIFAC modified
models, molecular
the residual part of volume fraction. In (Eq. 12 being ϑ 430

being η the modified molecular

the UNIFAC surface
models, area
thefraction
residual andthe activity
δ the
part
coefficient
ofmodified molecular
the activity volume ki
420
he number
fraction. of groups
In the coefficient
UNIFAC k inthe
of models,
type(Eq. beingi,)
molecule
12residual Jiski of
part replaced
thethenumberby the
activity ofsolution-of-groups
groups(Eq.
coefficient concept.
of 12 being ϑki The
following type
equation is k in
used molecule
for the group i,) is replaced
activity by
coefficient, the solution-of-
the number of groups of type k in molecule i,) is replaced by the solution-of-groups concept. The
410

groups concept. The following equation is used for the 400

following equation is used for thegroup activity coefficient, 
group activity coefficient,
  Θ m ξ mk 
ln Γk = Q k 1 − ln ∑
Θ m ξ mk  −   ∑ (17) 390

    m
ln Γk = Qk 1 − ln
  Θ
∑
Θ m ξ mk  −
m m ξ mk Θn ξ mn 
 ∑

∑ (17) (17)
0 0.2 0.4
x1 , y1
0.6 0.8 1

∑Θ ξ
  n
  m  m n mn  Fig. 5 Representation of experimental T-x1-y1 data for the binary systems MA1B ( , )

    and EA1B ( , ) at 0.6 MPa together with (green color lines) mod. UNIFAC-Larsen10
where Qk is the van der Waals surface area of ngroup k, Θ represents the group surface areaand (dark brown color lines) mod. UNIFAC-Gmehling predictions.
12

Fig. 5 Representation of experimental T-x1-y1 data for the

where Qk is the van der Waals surface area of group k, Θ represents the group surface area
fraction and definedQby
ξ iswhere k
is
an the van
equation der
that Waals
includes surface
the area
group of group
contribution parameters. Both systems
binary Θ MA1B ( , ) and EA1B ( , ) at 0.6 MPa
fraction and ξ is defined
k, Q by an equation
represents thethat includes
group the group
surface areacontribution
fraction parameters.
and Bothtogether
Θ with (green color lines) mod. UNIFAC-Larsen10 and
and ξ have different expressions in the UNIFAC versions.
x isexpressions
and ξ have different defined in bytheanUNIFAC
equation that includes the group
versions. (dark brown color lines) mod. UNIFAC-Gmehling12 predictions.
contribution parameters. Both Q and x have differ- 20

ent expressions in the UNIFAC versions. In general terms, the best prediction for gi was ob-
8
Table 4 shows prediction results from the group tained
8 with the mod. UNIFAC model proposed by
contribution models. Figs. 4 and 5 show the exper- Gmehling et al.12 However, most accurate results in
imental data and the fitting curves of predictions composition were found using the mod. UNIFAC
when using these group contribution models. model proposed by Larsen et al.10 The Gmehling et
0 .4 al.12 version also returns a good prediction for EA1B.
The UNIFAC model8 with the Hansen et al.11 parame-
ters and the ASOG model9 yield poor predictions and
0 .3 higher deviations in pressure and temperature, being
the mean proportional deviation (MPD) for the pre-
diction of the vapor phase mole fraction, respectively:
y1 - x1

0 .2
0 .1
0 al.12 are more successful.
0 0 .2
Fig. 4 Plot of experimental VLE data for MA1B ( ) and
EA1B ( ) at 0.6 MPa together with predictions of mod.
UNIFAC-Larsen10 (¾ ¾, green color line); mod. UNI-
FAC-Gmehling12 (¾ ¾, dark brown color line); PR-EOS13
(¾ ¾ , blue color line); PRSVWS-EOS13-15 (¾, red color
line); PRWS-EOS13,14 (¾ · ¾ ·, magenta color line) and
PRSV-EOS13,15 (¾ · ¾ ·, yellow color line).
9% and 16% for MA1B and 12% and 14% for EA1B.
Results indicate that with the current parameters,
some of the group contribution models do not repro-
duce adequately the VLE at moderate pressures; how-
ever, it is often observed that predictions given by the
models proposed by Larsen et al.10 and Gmehling et
0 .4 0 .6 0 .8 1
x1
Modelling with PR-EOS
The reliability in VLE modelling for mixtures that
have hydrogen bonding via proton donor and proton
acceptor is low if proper binary interaction param-
eters are not employed, especially at high pressure.
Consequently, the prediction of phase equilibrium
can be done by using EOS; if necessary, using the ap-
propriate mixing rules. This is why in this work the

102 | AFINIDAD LXXV, 582

 the Modelling
mean proportional
Resultswith deviation
PR-EOS
indicate (MPD)
that with the forparameters,
current the prediction
someofofthe
thevapor
groupphase mole fraction,
contribution models
respectively:
The9%reliability
do not reproduce and 16% inforVLE
adequately MA1B
the and 12%for
modelling
VLE andmixtures
at moderate 14% for that
pressures;EA1B.
have hydrogen
however, bonding
it is often via that
observed proton
donorResults
and
predictions indicate
proton
given that
by the withisthe
acceptor
models lowcurrent parameters,
if proper
proposed by some
al.10 of
binaryetinteraction
Larsen andtheGmehling
group contribution
parameters al.12notaremodels
etare employed,
more
do not reproduce
especially
successful. adequately
at high theConsequently,
pressure. VLE at moderate pressures; of
the prediction however, it is often observed
phase equilibrium that by
can be done
10 12
predictions
using EOS;given by the models
if necessary, usingproposed by Larsen
the appropriate et al.
mixing rules.and Gmehling
This is why in et this
al. work
are more
the PR-
successful.
EOS13 using
Modelling PR-EOSmixing
quadratic
with PR-EOS
13
using quadratic
rules mixing
and the WS 14 rulesrules
mixing andhave
the been
WS14used. The
being aij=aji as well as tij and tji adjustable parameters
adjustable
15 mixing rules have been used. The adjustable param- 13 to obtain agreement between theory and experiment
parameter employed
The reliability in by
VLESVmodelling
in the attractive termthat
for mixtures washave
also hydrogen
applied. The PR-EOS
bonding has the
via proton
eter employed by SV in the attractive term was also
15
for VLE. The non-randomness parameter (a) of the
Modelling
donor and with
following proton PR-EOS
equation: acceptor The
applied. is low PR-EOS if proper 13
has binary the following interaction equation: parameters are not employed, NRTL model23 was fixed to 0.47 in this study.
especially Theatreliability
high pressure. in VLE Consequently,modelling RT for mixtures a( T ) that have hydrogen (18)
bonding viadoneVapor
proton bypressures reported in literature
1,20,21
were used
p= − the prediction of phase (18) equilibrium can be
donor EOS; and proton acceptor is low
v − b v( v + b ) + b( v − b )
if proper binary interaction parameters
to obtain the k1 adjustable parameter of SV 15
using the
using if necessary, is why in are thisnot employed,
where using p, T the andappropriate R represent mixing
pressure, rules. This
temperature work
PR the PR-
equation. The critical properties employed (see Ta-
where
especially p,
at Tquadratic
high andpressure. R represent Consequently, pressure, temperature and thephase ideal equilibrium gas constant.can Furthermore, bya is taken from Daubert and Danner19. SD was
EOS 13
using and mixing the ideal rules gasand constant. thethe WSprediction 14
mixing of
Furthermore, rules have a is the beentem- used. Thebeadjustable bledone 2) were
theEOS; temperature-dependant perature-dependant 15 attractive parameter
attractive and rules. b is the temperature-independent repulsive
using
parameter if necessary,
employed by SV using inthetheappropriate attractive term parameter
mixing was alsoThis applied. and is why b The is in the this work
PR-EOS 13usedthe
has PR- for
the minimization parameters.
parameter,
13 which temperature-independent
are determined by the following repulsive
14 equations: parameter, which The Matlab program from Martin et al.27 is an ed-
EOS
following using quadratic mixing rules and the WS mixing rules have been used. The adjustable
equation: are determined by the following 2 2
equations: ucational program with a library of different EOS’s
15
parameter employed by SV p =in RT the attractive
a( T− ) = 0.457235 a( T )term R Tc was
β( T )
also applied. (19)
The PR-EOS13including has the PR13, PRSV13,15, PRWS13,14 and PRSVWS13-15.
(18)
v − b v ( v + b ) + b ( v − Pbc )
following equation: (19) We developed a FORTRAN code, based on a simpli-
where p, T and R represent pressure, temperature RTc the ideal gas constant. Furthermore,
and fied version a is of the above-mentioned Matlab program,
RT a( T )
p= − b = 0.077796 P (20) (18) (20) but related by separate to PR-EOS13, PRSV-EOS13,15, PR-
the temperature-dependant attractive v − b parameter v( v + b ) + b(and v − bcb) is the temperature-independent repulsive
WS-EOS13,14 and PRSVWS-EOS13-15.
being p, T Tcand and Pcrepresent respectively, the critical temperature and pressure of each pure component. a isThe
where
parameter, whichR are being determined Tc pressure,
and by Pc the temperature
respectively, following equations: and thethe critical ideal gas
temperature constant. Furthermore, Martin’s Matlab program27 uses inner functions, such
the classical correlation
temperature-dependant and pressure for temperature-dependant
attractive 2 of each parameter  pure 0.52and
2β function
component.
1TT−c02.T50.2500b ..255is
is: The classical
2 2the temperature-independent as fzero, fsolve or fminsearch. These inner math func-
repulsive
β( T ) = correlation 1 + m1 − Tr β(aT
0.5
)β=( Tβ1()+T=(21) m )1=1+−1mT+r1m R1−0− r(
2 2
)(21) (21) (21)
 determined   for
β (βT
( T β )temperature-dependant
(=
β T ( 0T) .
 457235
= ) 
+  = 1  +1  m
+ m  .150T .5rT
r−
 β
  (21)
 T   (19)
(21) b
(21) function is: tions of Matlab are robust and reliable but they imply
( T) =)the =1 +1following m  m 1−1 T − rTPr equations: r
parameter, which are by 2r0.5  
2 (21)
β( T )= 1+ m1 −c T  (21) long times of computer use. We substitute them for our
0.5    2
mω=−0m 0.βm
37464
=m
.(26992
T )=.037464 =.037464+.37464
1+
2
.54226 m++11.1+
 −1ωT.54226 −
r Rω
0RT .226992 −ω 20− .0 0ω.26992 (21)
2 2
ω 2 2(22) (22) (22)
m = 0.37464 + 1.54226 m=0= 0ω .b37464  + 54226
. 54226
1 (22)
. 54226  ωT 2− −
cω
26992
. 026992.26992 2ω 2ωω (22)(22) (21) own FORTRAN subroutines that lead us to similar ac-
mm= =0.037464 a(.(37464 T) )==+1=01++.01457235 ..077796
54226 54226 ω0ω .− − 
0.026992c
.26992 β( Tω)(21) ω(20) 2 (19) (22) (22) 9
m+11.−54226
5
When
V15 equation Whenthe
When
is SV
the
employed,
15
the
SV SV15 15
equation15equation
15 in equation is
PRSV-EOS =Tis
mβemployed, 0.37464 is  employed,
employed, m  is inTmodified r P
PRSV-EOS in
Pω
ccin − 0PRSV-EOS
PRSV-EOS .26992 according m ω ism modified m
is to (22)
is
modified
Eq. modified 23, (22)
according according
according to Eq.curacy.
to 23,
to
Eq. Eq. At
23, 23,the same time, we change Matlab programs
WhenWhen When
Whenthe thethe SV the
SV SV
15 SV
15
equation
equation
m equation
= equation
0 . 37464 is is is
employed,
employed,
+ is
1 employed,
. employed,
54226 ω in
− in 0 in
PRSV-EOS
. 26992 inPRSV-EOS
PRSV-EOS PRSV-EOS
ω
2
m m is
(22) m
is mis
modified ismodified
modified modified according according
according according to to Eq.to toEq.
Eq. Eq.
23,23,23,23,
When the SV 15
equation is employed, in PRSV-EOS structure, avoiding unnecessary loops, and eliminating
being
here
tionwhere
ofκ0T
where the κand
cis When
aκis00isPais
function
0acentric respectively,
the
aafunction
c function SV
function
factor,of15the equation
ofand
m of
the
=
the
acentric
0ω .the
critical
acentric
37464 and isb+factor,
acentric employed,
κ
1=.1
temperature
0.factor,
54226 are and
factor,
ω − 0RT
adjustable .andω incand
26992 PRSV-EOS
and ωωω and2
ω κparameters
and 1(20)pressure
and areκκ1(22) 11m
κκadjustable
are are is of each
modified
adjustable
adjustable
for each pure
parameters component.
according
parameters
parameters for each to
for Eq.The
for
each 23,
each
where
where
κ0κ0isthe κ κ
0 isa 15m functionis modified of of the the acentric acentric
according 077796factor, factor, to and
Eq. and 23, ω
andκm and and
where
1κ1are 1 are are
k0 adjustable
adjustable
is a func- parameters
parameters for
redundant for each
each assignment sentences, improving our FOR-
where
where When isa afunction
SV function equationofofthe theis acentric employed,
acentric factor,factor, in PRSV-EOS and and Pcωωand isare adjustable
modified adjustable parameters
according parameters to Eq. forfor23,eacheach
ureclassical
component,
where
purepure
follows: correlation
κ0 isthe
component,
component,
When asa follows:function
SV tion
as 15for
as temperature-dependant
of
follows:
follows:
equation of the
the acentric
is acentricfactor,
employed, factor, in PRSV-EOS βand function
and ω ω and and m is: is κmodified
k1 1 are are adjustable adjustableaccording parameters
to Eq. TRAN
23, for each code in speed of execution.
pure
pure
here pure 0Tcomponent,
pure
κcomponent,
component,
is component,
a function asasfollows: as offollows:
as
follows: follows:
the acentric factor, and ω andand κ1 pressure are adjustable parameters for eachThe
being c and Pc respectively, parameters thefor critical each temperature
0pure.5  0.5 component, as follows: of each pure component.
pure component,
where κ0mis= aκ 0function as follows:
of 0.5the acentric factor,
T(23) 1+0and ( 0r .T70.ω 5
0.5and
5T.
0(.0 ( ) κ−1)Tare ) adjustable parameters for eachUsing these28new programs, our results are k1=0.0234
+ κ1 1 + Tr (m0.=
 7 −κm0Tm= +m)κκ=10κ+
=κ0κ +
10+κ+ 1 rκ
κ1κ1+1 .1T
150T
+ .5 T (0  − 
 r7 0 − .
(.07.−7 T− 7T T) )
(23) (23) (23)
(23)
(23) (k1=0.05791 ) with SD=3.27 and MPD=1.6 for methyl
reclassical
component, as follows:  mm= =κ 0κ=+ rm 0+ + (23)
r r
.7−T
r
pure component, correlation  for temperature-dependant
as follows: 0 +κ1κ11+1 +  rT r (r00β(..r507function
T −rT)r ) r ris:(23) (23)
m = κ 0+κ11 + Tr (02.7 − Tr ) (23) acetate; 9k1=0.0692 (k1=0.0646428) with SD=5.40 and
κ 0 = κ 00ω.κ=378893 .5  3  ω + 0.20196554 ω  ω 3 ω
3
=  0.378893 + 1.4897153 − 378893
00=0..17131848
0.m + 1=.4897153
378893 κ+ 0 1ω +.
2 
1+1.4897153
+4897153
κ  ω
01 . + −T0ω
0196554 .017131848
0− .5ω  (
0 0−
. ω . 7 0.−17131848
17131848 T ) ω (24) ω2+ 20
2
+
(23). 0 .
0196554 0196554 3
3
3 (24) (24) (24)
 κ0κ00= 
= 0.0378893 .378893 +κ 01+.+14897153 .κ4897153 Tr0ω.17131848 ω−(− 0.0717131848 Tr )ω 2 +
.−17131848 2ωω(23) ++ 0.00196554
.0196554 3 3ωω   (24)(24) (24) MPD=1.2 for ethyl acetate and k1=0.4874 (k1=0.3343115)
ωω  3   (24)
r r
 κ 0κ 0= = .0378893
.378893 +m1+=.14897153 .4897153 11ω+ω
 −− 0.17131848 ω +02.00196554 .0196554 (24)
κ0=  0.378893 + 1.4897153ω − 0.17131848ω + 0.0196554 ω   (24) with SD=4.43 and MPD=1.7 for 1-butanol.
The The classical
The  0mixing
classical  mixing rules, with
rules, the binary 2interaction parameter  k1ij k1 and ijk1 ijk2 wereijk2
κclassical mixing rules, with 0with the the binary binary
2 + 0.interaction k1interaction ω3  k2 parameter and
ij k2 were
al mixing rules, with the binary interaction parameter 3 parameter and were
The
The 0classical
classical mixingmixing rules, rules, −with with the the binary binary ij and
interaction
interaction were
parameter
parameter k1 k1 and and k2 k2 9
ijwere
= .378893 + 1+.4897153 ωω .17131848 ω 0196554 (24) ij were
The ij
The Theclassical
classical κ 0 mixing
= 0 .
mixingrules,
378893 classical 1 .
rules,with
4897153 mixing withthe − 0 . thebinary
17131848 rules, binaryinteraction ω with + 0 . the
interactionparameter
0196554 binary
ω parameterk1k1 interac-
(24) ij ijand andk2k2
ij ij When
ij ijwere
ij
were comparing the results with those cited in the
mployed. employed.
a
e, employed.
the Therefore,
andThe b classical
Therefore,
Therefore, parameters the athe
tion mmixing
parameterand
the a
were: a bmmand
and parameters
rules,b b
k1 parameters with
parameters
and were:
the
k2 binary
were:were were: interaction Therefore,
employed. parameter k1ij andliterature k2ij were 15,28
, a good match for the ethyl acetate and
employed.
employed.employed.
m Therefore,
m
Therefore, Therefore, the thethe amthe aand
m
and mamand band bparameters
m bmparameters
m
ijparameters ij were: were:
mparameters
m were:
employed.
TheThe Therefore,
classical
classical mixing mixing amrules, rules, mbwith with the the binary
binary were: interaction
interaction parameter
parameter k1 k1 and and k2 k2 were were
employed. Therefore, the a the and
m am and 0. 5 bN m b N parameters
Nm Nparameters N N
0.5 were: 0 were:
. 5 0 . 5
ij ij ij ij
discrepancies in the other two substances is observed.
x j (xaii xa(25) ) 0(.1 )
N N
∑∑ ( ) ( )
0.50.5
mployed.
employed. = ∑∑ the
Therefore,
a m Therefore, xi xthe ja(ami aand mj )m (b
a and =1am
ma m=j∑∑
−bmamkparameters N)=
aN=1mparameters NxN
=∑∑ ∑∑
N NN N

N∑∑
i x j (
were:
i jx(a
were:
x
a
j (ia i ja)j1
a a50−
).a5 k1
(
0.)5 (1 (−
1 ij − 1k 1k1k1) )
− k )1 (25) (25)(25) However, while in the case of 1-butanol the tempera-
i xi ji xxi jxj(a j i ai ja)5j ) (1 (− −k1kij1)ij ) ij ij (25)
N xi (xa 1− (25) (25)
i j i j ij
a ma m=i=∑∑ ∑∑
ij j i j ij
(25)
i i j j x x 00.(5.a ) ( )
j ture range used in this study for correlations is wider,
N mi Ni ∑∑
i
a N = N a 1 − k 1 (25) (25)
xNi xj (jjaN(bai ai a+j j)b) (1(1b.−− +kk11bijij. )).  ..
j j i j i j ij
m .=N ∑∑
a ma= ∑∑
 bi +b b j= i jxN Nx NN i  bjib+b+(ib1.+
N N xiixN (25)
(25) in all cases the number of data used in this paper to ob-
bijbij −jkj(21.(ij1−()(−11k−
N N
bm = ∑∑ x  m b∑∑ (1i b− =m=kN∑∑ =Nij∑∑ i) j xixb xj(26)
i x+j+ bj .b j 2 2ijkk)22)ij(26))) (26) tain the SV parameter, is also much higher; this, along
2 bmbm=i=m mj i∑∑ N∑∑
ixj (26)
N j2 N
b (26)
2 b2(1(−1−k2kij2)ij )
 b b = x x  x k
− (26) (26)
∑∑N ∑∑
2
m
x xNx x i  i j j i2. +
i i 2 ij ij (26) (26)
 N i ii ji j i jjjj    b
∑∑  bxiix+jb2j2.   j  (1 − k 2 ij ) with the considerable differences in both the critical
i j
Nbmi N= (26)
bm = ∑∑ xi x j bi + bj  (12− k 2 ij )
j
( )
i j (26)
However,
r non-ideal However,
However,
mixtures
However,
for non-ideal
for However,
like
for fornon-ideal non-ideal
b
those
non-ideal
= ∑∑
mixtures
this forimixtures paper
mixtures non-ideal
mixtures
x x
j i j   the
i like  j those
like 2
2like

like
quadratic 
those
those this
mixtures
1 those
− k 
2 paper
this this
ij mixing
this paper the
like
paper paper (26) quadratic
thethose
rules the the quadratic
can
quadraticthis
quadratic mixing
not mixingpa- mixing
rules
mixing rules can
rules
not
properties
rules can cancan not
notnot and acentric factor used in the case of me-
However, for non-ideal mixtures like  those this paper the quadratic mixing rules notnot
can
m
However,
However, for forper non-ideal
non-ideal
the quadratic mixtures
i mixtures
14 j 14 mixing
like like those those rules this this can paper paper not the the
to quadratic
quadratic
be adequate. mixing
mixing rules
rules can
thylcan not
acetate and 1-butanol, can explain the differences.
betoadequate.
erefore, to
bebebeadequate.
the WSTherefore,
However,
adequate.
However,
14 for
Therefore,
Therefore,
mixing
for non-ideal the
rulesWS
non-ideal the the
were
mixtures WS mixing
mixtures
WS also
14
14mixing
14 applied. mixing ruleslikerules were
those rules
The were also
thiswere
WS alsoapplied.
paper also
mixing applied.the applied. The
quadratic
rules WS
Theare The WSmixing
mixing
WS mixing rules
mixing rules
notrules arecanare
rules not
are
toto be adequate.
adequate. Therefore,
Therefore, the the WS
14WS 14 like mixing
14mixing those this
rules rules paperwere were the also quadratic
also applied. applied. mixing The The rules
WS WS can
mixing
mixing rulesrules are
are interaction parameters for the systems of this
totobe be adequate.
adequate.
However, Therefore,
for Therefore,
Therefore,
non-ideal the the WS
mixtures the
WS WS
mixing mixing
like
14
mixing
those rules rules this were wererules
paper also also were
the applied.
applied.
quadratic also The applied.
The WS
mixing WS mixing
mixing
rules rules
can The
rulesnot arebinary
are
ven totoby:beadequate.
given
given
be by: adequate.
by: Therefore, the 14WS mixing are rules were also applied. The WSThe WS rules mixing rules aretreated as fitting parameters of VLE data. The
given
givenby:by: by: Therefore, The WS themixing WS mixing rules rules were also
given by:applied. mixing are
work were
given
given
be by:
adequate. Therefore, the WS14N mixing rules were also applied. The WS mixing rules are
given
given by:by: N N N N N (27) physical properties from literature19 were used for data
i x j (x )ij)RT
N N
xi x j (bij − aij /(∑∑ RTN )N∑∑ ) ∑∑ biijxx−j i(xbaijijj (b−/(ijaRT ) )
∑∑
NNNN
NxN
x x x(b(b− − RTRT)) ))
− a /( RT )
N N∑∑
j i N∑∑
ven by: /( ) processing by PR-EOS13 considering the classical b(T)
xi ji xNxi jxj(bji (ijbi jij− j−aijijaijij/( /(RT RT)) ))
aijaij/( /(
i ∑∑
∑∑
ij

∑∑ ( ( ) )
b N = m m Ei  j∑∑
bm =
i j bm =bm b=m N= i x i x j bj x −x a (27) b /( RT
− a ) /( RT ) (27) (27) (27) function for the attractive interaction and also by PRSV-
b b = = i Ni j j  i j ij iN j ijN N ij ij
 (27) (27)
E x E (jb − a /(NRT
a)xxia)i/(aai/(iRTb
    (27)
1 − A∞E /( CRT ) −bmmbm=xb1=i−
∑i  ma1i∑∑
(27)
=A/(1∞1RTb−E/(−11AECRT−∞xi iiA
−A
E)/(∞E/(
A
)/(−ijCRT
∞ ) −)i− x∑
jCRT
CRT
/( ∑
CRT N) −
Nx
)
ijN
− i)∑
)
a−i ∑
− N∑
/(xNiRTb
x
)/( RTb

RTb
/( i ) )i(27)
RTb
) 
) (27) EOS13,15 using the SV modification. In both cases, the

 bm = 1−−1 A
i− A
 EAj /(CRT
∞1∞−/(∞ A
CRT
/(∞ CRT 
E ) −
∑ ∑ ∑
xi aiii xai/(iiaiiRTb /(i i /(RTb ii RTb   i i 
) i)i)    (27)
i 
numerical values of k1ij, k2ij with quadratic mixing rules
∞ /( CRT N i) − i ∑ xi ai /( RTb )
i
   E   i 
) − ∑ x(i ai /(
i
1 (b − a /( RT )) 1=−1A[(∞b /( i RTbi ) 
](1/(RT−RT were obtained. In addition, the numerical values of k1ij,
[(biij− a(bi ijij/((b−RT −1)=)a1))=i))j= ii [/( b)))j +−)()ab+jj (− −j a/())jRT )k))](ij)1]))()1−)]]((−11k− k )(28) (28)(28)
1−1CRT 
j − aij /( RT )) =
2 (b(b(ijbijij−(−b−aij(abijaijij−ij/(/(−
ija− a)ija/(  )(/(
a) ij+RT /(b RTj))1 =/( [(ab2[1(RT
1
(bRT−]ia(1/(aa)−/(i)iRT
b[(i−b−i)ia)−
+k/( RT
)) )+) + ( /(
b jaRT
b(jb(28) k ij)k)ijij )
ij
/(RT
RT
/(
RT
ij )
RT
)=)
RT
2 = [ b [
[(b − a /( RT )) + (b − a /( RT ))](1 − k )
( 2 ( b − −a a /( RT
/( i RT i )
RT
/(
)
ij RT
)+ ) +( b (b − −a a
−j −
/(
a ja/(j /(
RT
/( RT ) ) ]
)()
RT
1 ] (
−1 −k k ) )
−
ij
(28) (28)(28)
(28)
(28) (28) k2 were also calculated when the WS14 mixing rules and
(bij − aij /( RT1E 22))2=i i212i2[(Nbi ii −i ai /(N RTEj ))j+j (jEb jj E−j a j /( RTij ))](1ij −ij kij ) (28)
ij
kij, tij, tji parameters were employed in PRWS-EOS13,14 and
(bij −aNij /( RT ai )) =Aa∞ [(=bib− ai /(NNRT  Na)Ni) + N (a bAi∞jE a−Ei aAE∞j /( E E
A RT  ))](1 − k ) (28)
+ 2m a ammaa=m∑
 
b mbxi∑ iNi∑
 
 ∑ i bi
a m = bm 
x + a a A 
∞A∞  ∞ (29) ij
PRSVWS-EOS13-15. All calculated values were obtained by
xCi b∞i+i ∞ (29)
=m b(29) xi∞Ai+A + (29)
mixib∑
x =b= ∑
ax+iiaxibA (29)
Nm a

m∑
+ EC   (29)(29) (29) (29)
i∑
 i
Caamma m==b=bmmmbm

a m = N ibimi ∑
∑
  biibiibxi CEiC
i x i i + +
ib
a b  C
i iC ∞  
+  
A C C
  (29)
(29) using the FORTRAN programs above mentioned.
  x ai i + constant bi   C
( ( ( (() ) ) )) Ai
where C
( is a
) numerical ∑ equal to and k = k To systematize data analysis of the MA1B and EA1B
∞
a = b (29)
here
where
where
erical Cconstant
where
where
where
where
where CC
isCaC
CisCisC isnumerical
is
aequal
isaisaanumerical
is aanumerical
anumerical
numerical
numerical
toconstant
numerical
numerical
is theconstant
ln constantconstant
binary
m
constant
constant
− 1equal
2constant
constant / equal
m
equal  equal
interaction
equal equal
2i to equal
(( ( )) )
=
equal
toto
i 
−bln 0to
tolnln
i to (lntotoln22Cln2parameter
.623 −and
−ln ln 21−)−
12 /−2k
/12 2/−21−
1 /−

2ij/1==/2−k0ji=2.623
1=21/2−=/0= 2.−623 =is
20−
for
−= −0.each
.0=623 0−
623
the
.623
−
and
and
.00623
..623
623
and
kijand
and
binary and
ij
=kand
kijsystem.
and
=kijkkji=ijisk=jithe
kji kis
ji
ijkij=
the =kjikkbinary
ijkij==kjikjiisisthe
isjijiisis
the
binary
the
is the
the
binarythe
binary
binary
binary
binary
binary
binary
systems, it was decided to obtain the acentric
where C is parametera numerical constant equal toEE excess ln an 2excess − 1energy / 2Helmholtz = −free 0.623 and kij = kmodel at isinfinite the binary
( )
E
teraction
er eachparameter
interaction
interaction
for
interaction parameter
system.
parameter for
A∞Efor each
is
isfor anfor
an
each eachsystem.
excess
each
excess
system. system. system. AHelmholtz
Helmholtz Ato ∞E is AEA∞Ean AE is is
an
free free
excessenergy Helmholtz Helmholtz
model model atenergy energy
free
infinite free at
energy model
in-
energy model ji factor,
atatat infinite since the models are highly dependent on this
infinite
interaction
interaction parameter
parameter for for each each system. system. ∞ E is lnisan A ∞ is excess −an 1an/excess Helmholtz −Helmholtz free free model
energy isatmodel
infinite atinfinite
here C
interaction is a
interaction parameter numerical
parameter finitefor constant
each
for pressure each system. system. equal
that A A∞can
∞
∞is ∞2
an anis excess
E be equated to a low-pressure
excess excess 2Helmholtz
=Helmholtz 0Helmholtz
.623 and
free free kfree
ij =
energy
energy energy
kjimodel model themodel atatbinary
infinite
infinite
parameter.
infinite
Vapor pressures from literature1,20,21 and the
eessureinteraction
pressure
equated
pressure thatto
pressure can
that
that a parameter
that
can becan
can
low-pressure be equated
be be
equated
equated forequatedto each
excess
to a to a system.
low-pressure
to
a
low-pressure a
low-pressure
Gibbs low-pressure A∞ excess
energy is an
excess excess
model excessGibbs
excess
Gibbs
26 Helmholtz
Gibbs
. In energy
Gibbs
energy this energy energy
study,
model model freemodel
.
26
26 energy
model
theIn . In26 this
this 26.
26 model
In
26 . In
study, study,
this this atstudy,
study,
the theinfinite
the the
pressure
teractionpressure that
parameter cancan
that can be
excess
for be
eachequatedequatedGibbstotoato
system. to a aA
energy low-pressure
low-pressure
E model 26 excess
.excess In this Gibbs Gibbsstudy, energy energy the model 2626 . In
model
NRTL atthis
.this
In this study,
study,
critical thethethe
properties of Daubert and Danner19 were used.
pressure
pressure that 23 can
that be be equated
equated alow-pressure
low-pressure ∞ is an excess excess excess Helmholtz
Gibbs Gibbs energy
energy free energy
modelmodel .model
.In26Inthis infinite
study,
study, the
RTL NRTL
NRTL
used, model
pressure
NRTL
as modelmodel that
model
follows:
2323 23
was can
was
23was used, was
23 model be
used, as
used, equated
used, follows:
as as
23 follows:
was as
follows: to
follows: a
used, as follows: low-pressure excess Gibbs energy model . In this study, the
The Nelder and Mead method25 was employed for data
NRTL NRTL
NRTLmodel model
model
2323 was was used,used, as as follows:follows: 26
NRTL
essure model
that can was bewas
23 equatedused,
used, asas tofollows:
follows:
aN low-pressure excess Gibbs energy model . In this study, the
correlation. Results were: ω=0.328 (ω=0.325319) with
NRTL model N was used, as follows: N N N

RTLE model 23
∑ x
wasj used, G τ ji as Nfollows:
E ∑ ∑ Nj ∑
NxNxj G
x∑
j G NN
∑xj τjxGτjijj Gji τjijijijiτ jiji
N∑
ji x τ
ji τG jixji G τ τ SD=0.0004; ω=0.362 (ω=0.361119) with SD=0.0012
∞xNxi ∑
ji E
∑
NE
A∞
N A∞A∞A∞=E A
∑
E E N Nj N
∑ j jxGj jG
(N−j Gα ijji τijji ) with ( −(−α= ) ()
ij(−ijα ij τ ijij) ij )
RT ∑ E i =∑ =N∑
with G ==exp τexp (30)
xNijjix=jx∑ G exp
(−(−α ijατijijτ)ij ) (30)
j = A ji ji ji ji with α τ (30)
RTART E A = x with G − α τ (30)
(30) and ω=0.599 (ω=0.594519) with SD=0.0003, for methyl
xi∑ ∑
xi N E ∞ i NG GG =ij=exp
= with ∞= iexpx ij (30) (30)
N N j j ij ij ij
ij (= −(− ij τijijτ)ij )
ART with with ij G exp exp (30) (30)
∞ ∞ RT
∑ xr GriERTRT N i =i ∑i∑i xrji r∑ji ri∑
=RT
A i=∑
RT ∑ i xi i∑ x
i xi NG
G
x Grj τ ri x G with
NN with GG ij ij= =
with Gij = exp(− α ij τ ij )
exp
ij exp αα (30)
acetate, ethyl acetate and 1-butanol, respectively.
∑∑∑x∑xG∑G
i ∞ N N xjirxGxG r Gri
ri (30)
r
A∞
RT
= RTxi ∑ j r
i
N r r ri ri
rr rrri ri
∑ x Gwith
N r r r r r ri
Gij = exp − α ij τ ij ( ) (30)
i
∑ x G r rir
10
r
10 10 10 APRIL- JUNE 2018 | 103
10
1010
1010
10
btain the acentric factor, since the models are highly dependent on this parameter. Vapor
essures from literature1,20,21 and the critical properties of Daubert and Danner19 were used. The
elder and Mead method25 was employed for data correlation. Results were: ω=0.328
ω=0.325319) with SD=0.0004; ω=0.362 (ω=0.361119) with SD=0.0012 and ω=0.599
ω=0.594519) with SD=0.0003, for methyl acetate, ethyl acetate and 1-butanol, respectively.
To compositions
To predict the predict the compositions of the
of the MA1B and MA1B
EA1B and
binary EA1Bat 0.6 MPa it was
systems 470
450
binary systems at 0.6 MPa it was considered the p-x
onsidered the p-x1 bubble point scheme. The simplex method25 was used,1 and the objective 460
bubble point scheme. The simplex method was used,
25 440

nction (OF) wasand

as follows:

T/K
the objective function (OF) was as follows: 450
430

1
n
1
n

OF = U (T ) ∑ T − T ∑
exp calc exp calc
+ y1 − y1 (31)(31) 440 420

1
U ( y1 ) 1 0 0.2 0.4 0.6 0.8 1

T/K
430 x1 , y1

Often for the Eq. for

Often 31 the
themaximum-likelihood principle is considered.
Eq. 31 the maximum-likelihood prin- Therefore, the
ciple is considered. Therefore, the standard deviations
andard deviations (SD) are employed. However, in this work uncertainties U(T); U(y1) have
420

(SD) are employed. However, in this work uncertain-

een used in Eq. 31. This amendment arises when considering that SD is a standard uncertainty, 410
ties U(T); U(y1) have been used in Eq. 31. This amend-
nd therefore, itsment
use mayarisesbe when
sufficient when the experimental
considering that SD is adata standardare correlated with the 400

uncertainty, and therefore,models

quations of the thermodynamic-mathematical its use thatmay be the
enable sufficient
production of adjustable 390
when the experimental data are correlated with the 0 0.2 0.4 0.6 0.8 1
arameters. In addition, when the models can predict VLE data, it seems more reasonable to x1 , y1
equations of the thermodynamic-mathematical mod- Fig. 7 Representation of experimental T-x -y data for the binary systems MA1B ( , )
1 1
and EA1B ( , ) at 0.6 MPa. Predictions represented by green color lines and dark
onsider a higher els
level of confidence.
that enable the That is why we
production of have selectedparame-
adjustable the expanded uncertainty,
Fig. 7 Representation of experimental
brown color lines using PRWS-EOS 13,14
and PRSV-EOS T-x1-y1 data for the
13,15
respectively.

ters. In addition, when the models

ecause in this way we ensure a higher quality prediction by the models. can predict VLE binary systems MA1B ( , ) and EA1B ( , ) at 0.6 MPa.
data, it seems more reasonable to consider a higher Predictions represented by green color lines and dark
Temperature and composition of vapor phase were compared with the theoretical
level of confidence. That is why we have selected the brown color lines using PRWS-EOS13,14 and PRSV-EOS13,15
13 13,15 13,14
edictions of VLE using the uncertainty,
expanded PR-EOS , PRSV-EOS because in ,this PRWS-EOS
way we ensure and PRSVWS-EOS13- respectively.
a higher
. Table 4 presents the globalquality
resultsprediction
obtained forbyMAD(ythe models.
1 ) and MAD(T). On the other side, the
Temperature and composition of vapor phase were Results
13
indicate that the best overall prediction of the 22

sults for binary compared

interaction with
parameters
the theoretical predictionsobtained
and objective function of VLE using us- PR-EOS were
experimental VLE data in both systems is obtained
ing theOF=2.9
=0.0345, k2=-0.0354, PR-EOS for13,MA1B
PRSV-EOSsystem,
13,15
,andPRWS-EOS
k1=0.0319, 13,14
and
k2=0.0109, with the
OF=2.8 for PRSVWS-EOS model13-15 as can be seen in
A1B system; using PRSVWS-EOS
PRSV-EOS13,15 the
13-15
. Table 4 were
results presents the global
k1=0.0231, resultsOF=2.0 for
k2=0.0234, Table
MA1B 3 and Figs. 4 and 6. The PR-EOS model13 re-
obtained for MAD(y1) and MAD(T). On the oth- 13,14 turns reasonably good predictions when considering
ystem, and k1=0.0099,
er side,k2=0.0901,
the results OF=1.6 for EA1B
for binary system; using
interaction PRWS-EOS
parameters the
theresults
composition of the vapor phase (see Fig. 4). The
ere kij=0.1519, and objective
α=0.47, function
τij=-0.7693, obtained
τji=1.4983, OF=1.6usingforPR-EOS 13
MA1B system, were and kij=0.0588,
prediction of temperature for both systems is poor
kl=0.0345, k2=-0.0354, OF=2.9 for MA1B system, and (see Fig. 6), when compared to the results obtained
=0.47, τij=-0.2311, τji=0.8075, OF=2.3 for EA1B system. and using PRSVWS-EOS13-15 the
k1=0.0319, k2=0.0109, OF=2.8 for EA1B system; using with models that use the SV equation15. This can be
sults were kij=0.1269,
PRSV-EOS α=0.47,
13,15
theτij=0.0183,
results wereτji=0.0258,
k1=0.0231, OF=1.9 for MA1B system,
k2=0.0234, observedand in Figs. 4, 6, 7. On the other hand, in Fig. 4
OF=2.0 for MA1B system, and
j=0.4896, α=0.47, τij=-0.4529, τji=-0.7123, OF=1.1 for EA1B system.
k1=0.0099, k2=0.0901, a similar
Figs. 4, 6, 7 show the
prediction of the composition of the vapor
OF=1.6 for EA1B system; using PRWS-EOS13,14 the re- phase is observed with the PR-EOS13 and PRSVWS-
xperimental datasults
and the correlation
were curves α=0.47,
kij=0.1519, obtained tby=-0.7693,
using theset EOS.
=1.4983, EOS13-15 models. Fig. 6 shows that best temperature
ij ji
OF=1.6that
Results indicate fortheMA1B system,prediction
best overall and kij=0.0588, α=0.47, tij=-VLE data
of the experimental predictions
in both are archived by using the SV equation . A
15

0.2311, t =0.8075, OF=2.3 for 13-15 EA1B system. and using similar behavior can be seen in Figs. 4 and 7.
ystems is obtained with theji PRSVWS-EOS model as can be seen in Table 3 and Figs. 4 and
PRSVWS-EOS13-15 the results were kij=0.1269, α=0.47, Therefore, in general, when the SV equation is em-
13
The PR-EOS model returns
tij=0.0183, reasonablyOF=1.9
tji=0.0258, good predictions
for MA1B when considering
system, and the composition
ployed, better predictions are obtained than when the
kij=0.4896, α=0.47, tij=-0.4529, tji=-0.7123, OF=1.1 for PR model is used. However, this does not seem enough
EA1B system. Figs. 4, 6, 7 show the experimental data to describe12 the thermodynamic behavior, nor for VLE
and the correlation curves obtained by using these EOS. calculations. Nevertheless, when using the WS mixing
rules, results show that the PRSVWS model can satisfac-
470
torily represent the VLE, for non-ideal polar organic sys-
tems, such as the ester + alcohol mixtures of this work.
450

460
440
T/K

450
430

440
CONCLUSIONS
420

Experimental data for the binary mixtures methyl

0 0.2 0.4 0.6 0.8 1
T/K

430 x1 , y1

acetate/1-butanol and ethyl acetate/1-butanol at 0.6 MPa

420
were obtained. The thermodynamic consistency was ver-
410 ified by using the point-to-point test. Results satisfy the
global criterion of this test. In addition, data obtained in
400
this work shows good agreement with bibliographic data.
390
Predictions by using the ASOG and UNIFAC group
0 0.2 0.4
x ,y
0.6 0.8 1
contribution models were obtained. Results show that
1 1
Fig. 6 Representation of experimental T-x1-y1 data for the binary systems MA1B ( , )
and EA1B ( , ) at 0.6 MPa. Predictions represented by green color lines and dark
the mod. UNIFAC-Lingby version returns satisfacto-
Fig. 6 Representation of experimental T-x1-y1 data for the
brown color lines using PR-EOS13 and PRSVWS-EOS13-15 respectively.
ry and best predictions, although the mod. UNIFAC-
binary systems MA1B ( , ) and EA1B ( , ) at 0.6 MPa. Dortmund model returns a good prediction of EA1B.
Predictions represented by green color lines and dark Experimental data were correlated with PR-EOS
brown color lines using PR-EOS13 and PRSVWS-EOS13-15 using quadratic mixing rules and the Wong-Sandler
respectively. mixing rules. The effect on the attractive term using
21

104 | AFINIDAD LXXV, 582

the Stryjek-Vera parameter was also verified. Global-
ly, results show the best predictions were obtained by
using the Stryjek-Vera parameter and Wong-Sandler
mixing rules simultaneously.
AUTHOR INFORMATION
Funding
This work was supported by the authors. This work
has not financial support of the Spanish government.
Notes
The authors declare no competing financial interest.
REFERENCES
1. Susial, P.; Sosa-Rosario, A.; Rodríguez-Henríquez,
J.J; Rios-Santana, R. Vapor Pressures and VLE
Data Measurements. Ethyl Acetate+Ethanol at
0.1, 0.5 and 0.7 MPa Binary System. J. Chem. Eng.
Jpn. 2011, 44, 155–163.
2. Susial, P.; García-Vera, D.; Susial, R.; Castillo, V.D.;
Estupiñan, E.J.; Apolinario, J.C.; Rodriguez-Hen-
riquez, J.J. Experimental Determination of Vapor–
Liquid Equilibria. Binary Systems of Methyl Acetate,
Ethyl Acetate and Propyl Acetate with 1-Propanol at
0.6 MPa. J. Chem. Eng. Data 2013, 58, 2861-2867.
3. Gmehling, J.; Onken, U. Vapor-Liquid Equilibri-
um Data Collection; Dechema: Frankfurt, 2008;
Vol. 1, Part. 2j; pp 41-51 and 104-109.
4. Gmehling, J.; Onken, U.; Rarey-Nies, J.R. Vapor-Liq-
uid Equilibrium Data Collection; Dechema: Frank-
furt, 1990; Vol. 1, Part. 2f; pp 137 and 145-146.
5. Susial, P.; Estupiñan, E.J.; Castillo, V.D.; Rodríguez-
Henríquez, J.J.; Apolinario, J.C. Experimental De-
termination of Densities and Isobaric Vapor–Liq-
uid Equilibria of Methyl Acetate and Ethyl Acetate
with Alcohols (C3 and C4) at 0.3 MPa. Int. J. Ther-
mophys 2013, 34, 1906-1917.
6. Gmehling, J.; Onken, U.; Arlt, W. Vapor-Liquid
Equilibrium Data Collection; Dechema: Frank-
furt, 1990; Vol. 1, Part. 2b; p 148.
7. Van Ness, H.C.; Byer, S.M.; Gibbs, R.E. Vapor-Liq-
uid Equilibrium: I. An Appraisal of Data Reduc-
tion Methods. AIChE J. 1973, 19, 238–244.
8. Fredenslund, A.; Gmehling, J.; Rasmussen, P. Va-
por–liquid Equilibria Using UNIFAC. A Group
Contribution Model; Elsevier: Amsterdam, 1977;
pp 11-15, 73 and 214-219.
9. Kojima, K.; Tochigi, K. Prediction of Vapor-Liquid
Equilibria by the ASOG Method; Kodansha Ltd.:
Tokyo, 1979; pp 11-50.
10. Larsen, B.L.; Rasmussen, P.; Fredenslund, A. A
Modified UNIFAC Group-Contribution Model
for Prediction of Phase Equilibria and Heats of
Mixing. Ind. Eng. Chem. Res. 1987, 26, 2274–2286.
11. Hansen, H.K.; Rasmussen, P.; Fredenslund, A.; Schil-
ler, M.; Gmehling, J. Vapor-Liquid Equilibria by
UNIFAC Group-Contribution. 5. Revision and Ex-
tension. Ind. Eng. Chem. Res. 1991, 30, 2352–2355.
12. Gmehling, J.; Li J.; Schiller, M. A Modified UNI
13. FAC Model. 2. Present Parameter Matrix and
Results for Different Thermodynamic Properties.
Ind. Eng. Chem. Res. 1993, 32, 178–193.
14. Peng, D.Y., Robinson, D.B. A new two-constant equa-
tion of state. Ind. Eng. Chem. Fundam. 1976, 15, 59-64.
15. Wong, D.S.H, Sandler, S.I. A Theoretically Correct
Mixing Rule for Cubic Equations of State. AIChE
J. 1992, 38, 671-680.
16. Stryjek,R.; Vera, J.H. PRSV: An improved Peng–
Robinson equation of state for pure compounds and
mixtures. Can. J. Chem. Eng. 1986, 64, 323-333.
17. Susial, P.; Rios-Santana, R.; Sosa-Rosario, A. VLE
Data of Methyl Acetate + Methanol at 1.0, 3.0 and
7.0 bar with a New Ebulliometer. J. Chem. Eng. Jpn.
2010, 43, 650-656.
18. Hayden, J.G.; O’Connell, J.P. A Generalised Meth-
od for Predicting Second Virial Coefficients. Ind.
Eng. Chem. Process Des. Dev. 1975, 14, 209–216.
19. Yen, L.C.; Woods, S.S. A generalized equation for
computer calculation of liquid densities. AIChE J.
1966, 12, 95–99.
20. Daubert, T.E.; Danner, R.P. Physical and Ther-
modynamic Properties of Pure Chemicals: Data
Compilation; Hemisphere Publishing Co.: New
York, 1989; Vol. 2.
21. Susial, P.; Susial, R.; Estupiñan, E.J.; Castillo, V.D.;
Rodríguez-Henríquez, J.J.; Apolinario, J.C. Deter-
mination and thermodynamic evaluation of iso-
baric VLE of methyl acetate or ethyl acetate with
2-propanol at 0.3 and 0.6 MPa. Fluid Phase Equi-
libria 2014, 375, 1-10.
22. Susial, P.; Rodríguez-Henríquez, J.J.; Castillo, V.D.;
Estupiñan, E.J.; Apolinario, J.C.; Susial, R. Isobaric
(vapor + liquid) equilibrium for n-propyl acetate with
1-butanol or 2-butanol. Binary mixtures at 0.15 and
0.6 M. Fluid Phase Equilib. 2015, 385, 196–204.
23. Wilson, G.M. Vapor-Liquid Equilibrium. XI. A
New Expression for the Excess Free Energy of
Mixing. J. Am. Chem. Soc. 1964, 86, 127-130.
24. Renon, H., Prausnitz, J.M. Local compositions in
thermodynamic excess functions for liquid mix-
tures. AIChE J. 1968, 14, 135-144.
25. Abrams, D.S.; Prausnitz, J.M. Statistical Thermo-
dynamics of Liquid Mixtures: A New Expression
for the Excess Gibbs Energy of Partly or Complete-
ly Miscible Systems. AIChE J. 1975, 21, 116-128.
26. Nelder, J.; Mead, R. A simplex method for func-
tion minimization. Comput. J. 1965, 7, 308–313.
27. Wong, D.S.; Orbey, H.; Sandler, S.I. Equation-of-
state Mixing Rule for Non-ideal Mixtures Using
Available Activity Coefficient Model Parameters
and That Allows Extrapolation over Large Ranges
of Temperature and Pressure. Ind. Eng. Chem.
Res. 1992, 31, 2033–2039.
28. Martín, A., Bermejo, M.D., Mato, F.A., Cocero,
M.J. Teaching advanced equations of state in ap-
plied thermodynamics courses using open source
programs. Educ. Chem. Eng. 2011, 6, e114-e121.
29. Proust, P.; Vera, J.H. PRSV: The Stryjek-Vera modi-
fication of the Peng-Robinson equation of state. Pa-
rameters for other pure compounds of industrial
interest. Can. J. Chem. Eng. 1989, 67, 170-173.
APRIL- JUNE 2018 | 105