Vapor-Liquid Equilibrium Calculations For Alcohol and Hydrocarbon Mixtures Using Cosmo-Sac, NRTL, and Uniquac Models
Vapor-Liquid Equilibrium Calculations For Alcohol and Hydrocarbon Mixtures Using Cosmo-Sac, NRTL, and Uniquac Models
Vapor-Liquid Equilibrium Calculations For Alcohol and Hydrocarbon Mixtures Using Cosmo-Sac, NRTL, and Uniquac Models
a
Universidade Federal da Bahia, Escola Politécnica, Salvador, BA, Brazil
b
Universidade Federal do Rio de Janeiro, Centro de Tecnologia, Escola de Química, Rio de Janeiro, RJ, Brazil
c
University of Alberta, Department of Agricultural, Food and Nutritional Science, Edmonton, AB, Canada
ABSTRACT
Vapor-Liquid equilibrium (VLE) data involving alcohols and hydrocarbons are relevant to the gas and oil
industry. Under certain circumstances, for instance, in the determination of the distribution of
components between phases in equilibrium, and physical properties of the fluids in reservoirs, the
predictions from the available models for hydrocarbon systems in the presence of flow-assurance
additives, such as alcohols, are, at times, insufficient. This study used the Conductor-like Screening Model-
Segment Activity Coefficient (COSMO-SAC) to predict liquid phase activity coefficients for the description
of VLE of seven binary systems containing alcohols and hydrocarbons, at low and high pressures (10-
2500kPa), and under temperatures ranging from 298-400K. The COSMO-SAC has a relatively simple
mathematical form and can be easily incorporated into a process simulation software. COSMO-SAC model
predictions in this work showed average absolute relative deviation (AARD) values ranging from 5.0% to
13.8%.
KEYWORDS
COSMO-SAC; VLE; alcohol; hydrocarbon
1
To whom all correspondence should be addressed.
Address: DEQ - Escola Politécnica - Universidade Federal da Bahia, Rua Aristides Novis, 2 – Federação - Salvador, BA, Brazil
Zip Code: 40210-630|Telephone & Fax: (55) 71 3283-9809 |e-mail: [email protected]
doi:10.5419/bjpg2014-0012
127
BRAZILIAN JOURNAL OF PETROLEUM AND GAS | v. 8 n. 4 | p. 127-137 | 2014 | ISSN 1982-0593
In recent years, the scientific community has i at vapor and liquid mixtures, respectively.
studied a class of thermodynamic models for phase For components in the vapor phase:
equilibrium prediction, the COSMO-based
methods. These methods include the COSMO-RS
fˆiV yi ˆiV P (2)
(Klamt & Schuurmann, 1993; Klamt, 1995; Klamt
et al., 1998) and its variants such as the COSMO-
Where yi is the molar fraction of component i in
SAC (Lin & Sandler, 2002; Lin et al., 2004; Wang et
vapor mixture, ˆi is fugacity coefficient and P is
V
al., 2007) and the COSMO-RS (Ol) (Grensemann &
Gmehling, 2005). Essentially, these methods pressure.
consider the liquid phase solution non-ideality
using molecular interactions derived from solvation For components in the liquid phase, the gamma
calculations based on quantum chemistry (activity coefficient) approach may be used
(ab initio). The COSMO-SAC model does not require (Equation 3):
estimation of parameters from experimental data,
therefore, it is a predictive model. It has been used P V L dP
fˆi L xi i Pi sat i sat exp
to provide acceptable predictions of vapor-liquid sat RT (3)
equilibrium (VLE) of mixtures. However, Hsieh and P
Lin (2010) recently reported poor accuracy of the
Where xi is molar fraction of component i in
COSMO-SAC for the description of liquid-liquid
equilibrium (LLE). liquid mixture, i is the activity coefficient of
sat
component i in liquid mixture, Pi is the vapor
128
BRAZILIAN JOURNAL OF PETROLEUM AND GAS | v. 8 n. 4 | p. 127-137 | 2014 | ISSN 1982-0593
pressure of pure component i , i sat is the fugacity First, the tridimensional geometry of each
molecule must be determined by minimization of
coefficient of pure component i at saturation
configurational energy. These calculations were
condition, and V L is pure component i molar performed using the quantum chemistry software
ˆ is defined as follows:
volume. ChemBioOffice (trial version). The quantum
i
mechanical calculations are necessary for the
ˆ ˆiV determination of the geometric configuration of
the atoms in the molecule that presents the lowest
V L dP
i P
(4)
i sat exp energy, knowing which atoms compose the
P sat RT molecules and how they are bonded. It is desired
to determine the angles and distances between the
The equilibrium pressure can, then, be atoms nuclei in the molecule in its most stable
calculated by the complete equilibrium equation configuration. This procedure is known as
(Equation 5): geometry optimization. The necessary calculations
for the determination of these data are available in
ˆ P x P sat
yi (5) quantum chemistry software packages such as
i i i i
GAUSSIAN, TURBOMOLE, MOPAC, DMol3, and
The value ˆ was approximated to unity, which GAMESS (Gerber & Soares, 2010).
i
is often a good approximation for fluids at low After, the cavity volume, the total number of
pressure. This approximation is valid if the segments, and the sigma profile of each substance
compressibility factor for the mixture and for pure are calculated. Solvation calculations in a perfect
components, as well as the system’s pressure and conductor are performed, using the equilibrium
the pure components’ saturation pressures, are of geometry, to obtain the surface charge of the
comparable magnitudes (Rousseau, 1987). The desired substance. The ab initio parameters for the
activity coefficients were calculated with the compounds of the present work were obtained
COSMO-SAC model. following the recommendations provided by
Mullins et al. (2006) and Alvarez et al. (2011).
P xi i Pi sat (6) More details about the theory can be found in the
i
work of Lin and Sandler (2002).
Bubble point pressure calculations were
The activity coefficient of the present species,
performed for different composition and
which are necessary for the vapor-liquid
temperatures to generate pressure versus
equilibrium calculations, can be determined from
composition phase envelope diagrams at given
the COSMO-SAC model using the sum of
temperatures.
combinatorial and residual contributions.
Vapor pressure for the pure components were
obtained from Antoine correlations available in ln i ln ires ln icom (7)
NIST WebBook database (Lemmon et al., 2011).
The residual contribution is calculated from the
2.2 COSMO-SAC molecular solvation in a perfect conductor. The
charge distribution of the molecular surface, called
In the COSMO approach in modeling excess sigma profile, p( ) , is determined from quantum
Gibbs energy, the distribution of charges on the mechanics calculations. Molecular interactions in
surface of the cavity surrounding a molecule the liquid phase are assumed to be the sum of
immersed in a perfect conductor material is used contributions from the interactions of surface
as reference. The activity coefficient is obtained segment through selected charges.
based on the difference in energy for the molecule
from the state of immersion in the ideal conductor ln ires ni pi ( m ) ln[ S ( m ) i ( m )] (8)
and the state of immersion in the real solution m
condition. The calculation of this difference in
energy is performed based on the charge density Where ni is the total number of surface segments
probability profile. in molecule i , pi ( m ) is the sigma profile for a
129
BRAZILIAN JOURNAL OF PETROLEUM AND GAS | v. 8 n. 4 | p. 127-137 | 2014 | ISSN 1982-0593
130
BRAZILIAN JOURNAL OF PETROLEUM AND GAS | v. 8 n. 4 | p. 127-137 | 2014 | ISSN 1982-0593
The sigma profiles generated for the substances because they are all commonly used in literature,
studied in this work are presented in Figures 1 and but they have different interpretations and
2. The horizontal axis represents the charge density limitations, as explained in the following lines.
induced in the molecules’ surfaces, these values
usually range from -0.025 to +0.025 (e/Å2). The Average absolute relative deviation (AARD) is
vertical axis represents the probability of a surface calculated by arithmetic mean of absolute values of
segment having a specific charge density multiplied relative deviations. These relative deviations are
by the molecule surface area. calculated by the difference between experimental
and predicted values, and divided by the
Figures 1 and 2 show that the sigma profiles are experimental values as a reference, giving a
similar among the solvents in each group. percentage value in relation to the experimental
Nevertheless, the values are very different value. This kind of measure is adequate to variables
between the two groups, as one can observe by expressed in absolute units, so that the result is not
looking at the different scales used for the chart of affected by the use of different unit systems. This
each group. In the same way that interactions measure is frequently used in phase equilibrium
between two molecules of different hydrocarbons literature to evaluate goodness of fit of
or different alcohols are similar to interactions thermodynamic models (Bosse & Bart, 2005; Hsieh
between two molecules in pure hydrocarbon or & Lin, 2009; Jia et al., 2012).
alcohol liquids. While interactions between
hydrocarbons and alcohols are different from the 100 n Pexp Pcalc
interactions in the pure hydrocarbon or alcohol AARD % P (14)
n i 1 exp
liquids. Therefore, noticeable deviations from ideal
solution behavior are expected along with the
This calculation is useful in comparisons
occurrence of vapor-liquid azeotropes.
between different works, and between a model
calculation accuracy and an application
2.4 Calculation Analysis
requirement.
The capability of the model was evaluated by
comparing calculated values and experimental data Correlation coefficient ( ) is a measure of the
for equilibrium pressures through three measures: degree of linear dependence between two
average absolute relative deviation, linear variables. In the context of evaluating goodness of
correlation, and coefficient of determination. The fit, it is defined as follows:
use of these three different measures is important
131
BRAZILIAN JOURNAL OF PETROLEUM AND GAS | v. 8 n. 4 | p. 127-137 | 2014 | ISSN 1982-0593
(P
i 1
exp Pexp ) 2
(P
i 1
calc Pcalc ) 2 & Pinto, 2007). However, if its value is low, it may
indicate poor quality of model, as well as excessive
experimental variance.
The correlation coefficient can be used to
measure the degree of linear dependence between The coefficient of determination, also referred
calculated and experimental values. If the expected to as captured variance, R 2 adj (Equation 16),
value for the residuals statistical distribution is represents the fraction of total variation among the
zero, i.e. the residuals are equally expected to be experiments that can be explained by the model. It
positive or negative, this measure can be used to is calculated by comparing model calculation
analyze the goodness of fit between model and variance to global variance of experimental values.
data. The correlation coefficient may assume
Calculation variance is the sum of square
residuals, SSE (Equation 17), divided by the degree
of freedom of the model evaluation problem.
Global variance is the sum of the squared
difference between each experimental values and
the mean of all experimental values, SST (Equation
18), divided by the number of experiments less
one. This approach naturally penalizes models with
more parameters (Montgomery & Runger, 2003).
SSE / (n p)
R 2 adj 1 (16)
SST / (n 1)
Figure 3 Pressure x composition phase diagram for SSE ( Pexp Pcalc )2 (17)
the system: butanol/decane at 358.2 K, 373.2 K,
388.2 K as function of butanol molar fraction.
Experimental data from Bernatová et al., (1992).
SST ( Pexp P )2
Model calculation from this work.
(18)
Figure 4. Pressure x composition phase diagram for Figures 3 to 9 describe the vapor-liquid
the system butanol/hexane at 298.15 K as function equilibrium behavior of seven binary mixtures of
of butanol molar fraction. Experimental data from alcohols and hydrocarbons, for different
Rodriguez et al. (1993). Model calculation from this temperatures, as plots of equilibrium pressures
work. versus phase composition, as predicted by the
COSMO-SAC model. COSMO-SAC is a predictive
132
BRAZILIAN JOURNAL OF PETROLEUM AND GAS | v. 8 n. 4 | p. 127-137 | 2014 | ISSN 1982-0593
133
BRAZILIAN JOURNAL OF PETROLEUM AND GAS | v. 8 n. 4 | p. 127-137 | 2014 | ISSN 1982-0593
Components T [K] Pressure range [bar] Pressure range [bar] AARD [%] R 2 adj
358.2 0.05 - 00.28 0.0509 - 0.2763 7.29 0.984 0.954
Butanol/Decane 373.2 0.09 - 00.52 0.0956 - 0.5215 6.34 0.991 0.968
388.2 0.16 - 00.92 0.168 - 0.9248 5.77 0.995 0.977
Butanol/Hexane 298.2 0.17 - 00.20 0.17875 - 0.20174 4.99 0.966 0.881
348.2 0.07 - 00.49 0.0732 - 0.4879 9.29 0.982 0.924
Heptane/1-Pentanol 358.2 0.12 - 00.68 0.1212 - 0.6807 7.31 0.989 0.958
368.2 0.19 - 00.93 0.1937 - 0.9308 6.07 0.991 0.957
Ethanol/Ciclohexane 298.2 0.07 - 00.19 0.0791 - 0.1864 12.23 0.972 0.558
Isobutane/Ethanol 363.5 3.38 - 16.71 3.38 - 16.71 9.07 0.991 0.937
372.7 2.24 - 06.80 2.241 - 6.843 12.73 0.973 0.733
Pentane/Ethanol 397.7 4.82 - 12.01 4.826 - 12.011 9.95 0.967 0.788
422.6 9.64 - 19.63 9.642 - 19.629 8.46 0.985 0.794
372.7 3.47 - 08.46 3.471 - 8.456 13.08 0.984 0.662
Pentane/Methanol 397.7 7.29 - 15.18 7.295 - 15.117 11.51 0.979 0.539
422.6 13.74 - 25.27 13.748 - 25.276 10.57 0.984 0.542
AARD is the average absolute relative deviation; is the linear correlation; R 2 adj is the adjusted coefficient of
determination
134
BRAZILIAN JOURNAL OF PETROLEUM AND GAS | v. 8 n. 4 | p. 127-137 | 2014 | ISSN 1982-0593
Figure 8 Pressure x composition phase diagram for Figure 9. Pressure x composition phase diagram for
the system pentane/ethanol at 372.7 K, 397.7 K, the system pentane/methanol at 372.7 K, 397.7 K,
422.6 K as function of pentane molar fraction. 422.6 K as function of pentane molar fraction.
Experimental data from Campbell et al. (1987). Experimental data from Wilsak et al., (1987). Model
Model calculation from this work. calculation from this work.
135
BRAZILIAN JOURNAL OF PETROLEUM AND GAS | v. 8 n. 4 | p. 127-137 | 2014 | ISSN 1982-0593
Alvarez, V. H.; Mattedi S.; Martin-Pastor M.; Hsieh, C.M.; Lin S.T. Prediction of liquid–liquid
Aznar M.; Iglesias M. Thermophysical properties of equilibrium from the Peng Robinson + COSMO-SAC
binary mixtures of {ionic liquid 2-hydroxy equation of state. Chemical Engineering Science v.
ethylammonium acetate+ (water. methanol. or 65, p. 1955–1963, 2010.
ethanol)}. The Journal of Chemical http://dx.doi.org/10.1016/j.ces.2009.11.036
Thermodynamics v. 43, p. 997-1010, 2011.
http://dx.doi.org/10.1016/j.jct.2011.01.014 Jia, Q.;Wang, Q.; Ma P.; Xia S.;Yan, F.; Tang, H.
Prediction of the Flash Point Temperature of
Bernatová, S.; Linek J.; Wichterle I. Vapour- Organic Compounds with the Positional
liquid equilibrium in the butyl alcohol -n-decane Distributive Contribution Method. Journal of
system at 85, 100 and 115 °C. Fluid Phase Chemical and Engineering Data v. 57, p. 3357–
Equilibria. v. 74, p. 127-132, 1992. 3367, 2012.
http://dx.doi.org/10.1016/0378-3812(92)85057-F http://dx.doi.org/10.1021/je301070f
136
BRAZILIAN JOURNAL OF PETROLEUM AND GAS | v. 8 n. 4 | p. 127-137 | 2014 | ISSN 1982-0593
Klamt, A.; Conductor-like Screening Model for Mullins, E.; Oldland R.; Liu Y.A.; Wang S.;
Real Solvents: A New Approach to the Quantitative Sandler S.I.; Chen C.C.; Zwolak M.; Seavey K.C.
Calculation of Solvation Phenomena. The Journal Sigma-profile database for using COSMO-based
of Physical Chemistry v. 99, p. 2224–2235, 1995. thermodynamic methods. Industrial and
http://dx.doi.org/10.1021/j100007a062 Engineering Chemistry Research v. 45, p. 4389–
4415, 2006.
Klamt, A.; Jonas V.; Burger T.; Lohrenz J.J.C.W. http://dx.doi.org/10.1021/ie060370h
Refinement and Parametrization of COSMO-RS.
The Journal of Physical Chemistry A v. 102, p. Pires, A.P.; Mohamed, R. S.; Mansoori, G. A.; An
5074–5085, 1998. equation of state for property prediction of
http://dx.doi.org/10.1021/jp980017s alcohol-hydrocarbon and water-hydrocarbon
systems, Journal of Petroleum Science and
Klamt, A.; Schuurmann G. COSMO: a new Engineering, v. 32 p. 103-114, 2001.
approach to dielectric screening in solvents with
explicit expressions for the screening energy and its Rodriguez, V.; Pardo. J.; Lopez. M. C.; Royo. F.
gradient. Journal of Chemical Society, Perkin M.; Urieta. J. S. Vapor pressures of binary mixtures
Transactions 2 v. 5 , p. 799–805, 1993. of hexane + 1-Butanol. + 2-Butanol. + 2-methyl-1-
http://dx.doi.org/10.1039/p29930000799 propanol. or + 2-methyl-2-propanol at 298.15 K.
Journal of Chemical and Engineering Data v. 38, p.
Lemmon, E. W.; McLinden M. O.; Friend, D. G.;
350-352, 1993.http://dx.doi.org/10.1021/je00011a003
Thermophysical Properties of Fluid Systems, NIST
Chemistry WebBook, NIST Standard Reference Rousseau, R. W.; Handbook of separation
Database Number 69, Eds. Linstrom, P.J.; Mallard, process technology, Chapter 1, Phase Equilibria,
W.G.; National Institute of Standards and Abbott, M. M.; Prausnitz, J. M.; John Wiley & Sons,
Technology, Gaithersburg MD, 20899, 2011, Inc, NY, 1987.
http://webbook.nist.gov.
Schwaab, M.; Pinto, J. C.; Análise de Dados
Lin, S. T.; Chang J.; Wang S.; Goddard W. A.; Experimentais I: Fundamentos de Estatística e
Sandler S. I. Prediction of Vapor Pressures and Estimação de Parâmetros, Rio de Janeiro,
Enthalpies of Vaporization Using a Cosmo Solvation e-papers, 2007.
Model. The Journal of Physical Chemistry A v. 108,
p. 7429-7439, 2004. Wang, S.; Sandler S. I.; Chen. C.C. Refinement of
http://dx.doi.org/10.1021/jp048813n COSMO-SAC and the Applications. Industrial and
Engineering Chemistry Research v. 46, p. 7275-
Lin, S. T.; Sandler S. I. A Priori Phase Equilibrium 7288, 2007.
Prediction from a Segment Contribution Solvation http://dx.doi.org/10.1021/ie070465z
Model. Industrial and Engineering Chemistry
Research v. 41, p. 899-913, 2002. Wilsak, R. A.; Campbell S.W; Thodos. G. K.
http://dx.doi.org/10.1021/ie001047w Vapor-liquid equilibrium measurements for the n-
pentane-methanol system at 372.7, 397.7 and
Machova, I.; Linek, J.; Wichterle, I. Vapour-liquid 422.6 k. Fluid Phase Equilibria. v. 33, p. 157-171,
equilibria in the heptane-1-pentanol and heptane- 1987. http://dx.doi.org/10.1016/0378-3812(87)87009-7
3-methyl-1-butanol systems at 75, 85 and 95 ° C.
Fluid Phase Equilibria. v. 41, p. 257-267, 1988. Zabaloy, M. S.; Gros. H. P.; Bottini. S. B.;
http://dx.doi.org/10.1016/0378-3812(88)80010-4 Brignole, E. A. Isothermal Vapor-Liquid Equilibrium
Data for the Binaries Isobutane-Ethanol. Isobutane-
Montgomery, D. C.; Runger, G. C.; Applied 1-Propanol and Propane-Ethanol. Journal of
Statistics & Probability for Engineers, John Wiley & Chemical and Engineering Data v. 39, p. 214-218,
Sons, Inc., Third Edition, 506-563, 2003. 1994.
http://dx.doi.org/10.1021/je00014a005
137