Page 100 - Gas Purification 5E

90 Gas PuniJication

The Kent-Eisenberg model assumes all activity coefficients and fugacity coefficients to
be 1.0 (i.e., ideal solutions and ideal gases), and forces a fit between experimental and pre-
dicted values by treating two of the reaction equilibrium constants as variables. The reac-
tions so treated are the amine dissociation reaction (equation 2-6) and the carbamate forma-
tion reaction (equation 2-7). Since tertiary amines do not form carbamates, a modified
approach is required in developing a generalized correlation for these amines. Jou et al.
(1982) describe such an approach for the correlation of HzS and COz solubilities in aqueous
MDEA solutions.
According to Weiland et al. (1993), the Kent-Eisenberg correlation provides a good fit
between experimental and predicted values only in the loading range of 0.2 to 0.7 moles acid
gas per mole of amine, and gives inaccurate results for mixed acid gases. However, it has the
important advantage of computational simplicity, and has been incorporated into several
computer models used for treating plant design. A comparison of VLE data predicted by the
Kent-Eisenberg correlation and experimental data for the system MEA-HzO-COz is given in
Figure 2-17.
A more rigorous, and therefore more generally applicable model was proposed by Desh-
mukh and Mather (1981). It uses the same chemical reactions in solution as the Kent-Eisen-
berg correlation, but, instead of assuming activity and fugacity coefficients to be unity, val-
ues for these coefficients are estimated and used in the calculation of liquid phase
equilibrium constants and in the application of Henry’s law to the gas-liquid equilibrium.
The basic elements required for the Deshmukh-Mather model are

1. equilibrium constants for the chemical reactions
2. Henry’s law constants for COz and HzS in water
3. fugacity coefficients for gas phase components
4. activity coefficients for all species in the solution

Sufficient data and methods exist to permit reasonable estimates to be made for items 1,2,
and 3. The approach has, therefore, been taken to accept these estimates, and to adjust the
interaction parameters used in estimating the activity coefficients so that the final calculated
equilibrium values match experimental data. In Deshmukh and Mather’s original publica-
tion, a rather cumbersome method was used to solve the system of equations. Chakravarty
(1985) proposed a simpler technique, which greatly reduces computation times. The
improved model has been used for estimating VLE data in an absorption system simulation
model developed by Sardar and Weiland (1985) and by Weiland et al. (1993) for evaluating
and condensing a large amount of published data for MEA, DEA, DGA, and MDEA. The
interaction parameters developed by Weiland et al. provide a sound basis for estimating VLE
data for the most important commercial amines over a wide range of conditions.
In a related study, Li and Mather (1994) used Pitzer’s excess Gibbs energy equations
(Pitzer, 1991) to predict VLE data for the MDEA-MEA-H20-CO2 system using interaction
parameters determined from experimental data for MDEA-H20-C02 and MEA-H20-C02
systems. Li and Mather’s presentation provides insights into concentrations of various ionic
and molecular species in the liquid phase when an acid gas is dissolved into a mixed amine
solution. Figure 2-55, for example, shows how the concentrations of all key species vary
with increasing COz/amine mole ratio in a solution containing 10 wt% MEA and 20 wt8
MDEA at 40°C.
The most sophisticated, and probably the most accurate model available at this time was
proposed by Austgen et al. (1991). This model is based on the electrolyte-NRTL model of