2080 Ind. Eng. Chem. Res.

1999, 38, 2080-2090

Representing Vapor-Liquid Equilibrium for an Aqueous MEA-CO2

System Using the Electrolyte Nonrandom-Two-Liquid Model
Yunda Liu,* Luzheng Zhang, and Suphat Watanasiri
Aspen Technology, Inc., Ten Canal Park, Cambridge, Massachusetts 02141

Following the work of Austgen et al., the electrolyte nonrandom-two-liquid (NRTL) model was
applied in a thermodynamically consistent manner to represent the vapor-liquid equilibrium
(VLE) of the aqueous monoethanolamine (MEA)-CO2 system with rigorous chemical equilibrium
consideration. Special attention was given to the accurate VLE description of the system at
both absorbing and stripping conditions relevant to most aqueous MEA absorption/stripping
processes for CO2 removal. The influence from chemical equilibrium constants, Henry’s constant,
experimental data, and data regression on the representation of the VLE of the system was
discussed in detail. The equilibrium constant of the carbamate reversion reaction as well as
important interaction parameters of the electrolyte NRTL model were carefully fitted to
experimental data. A good agreement between the calculated values and the experimental data
was achieved. Moreover, the model with newly fitted parameters was successfully applied to
simulate three industrial cases for CO2 removal using a rate-based approach. The results from
this work were compared with those using the model by Austgen et al.

Introduction CO2 removal process and for achieving a business

competitiveness.
Aqueous monoethanolamine (MEA) is widely used for
the removal of CO2 from gas streams. It has several A thermodynamic property model capable of accurate
advantages over other commercial alkanolamines: representation of the vapor-liquid equilibrium (VLE)
(a) It is the strongest base with the quickest reaction of the aqueous MEA-CO2 system is essential for a
rate, yielding the highest purity, which make it most successful computer simulation of the process. In the
suitable for processing to-be-liquefied natural gas, syn- last few decades, considerable progress has been made
thesis gas, and hydrogen stream, which require low CO2 in modeling VLE of the acid gas (CO2 and H2S) in
leakage. aqueous alkanolamine systems, including the aqueous
(b) It has the lowest molecular weight and thus the MEA-CO2 system.
highest absorbing capacity on a weight basis. Further- A practical and widely used model was proposed by
more, its cost is low. Kent and Eisenberg.3 This model is essentially the
(c) It is very thermally stable and less likely to Klyamer-Kolesnikova-Rodin model,4 assuming all ac-
undergo thermal degradation. tivity coefficients and fugacity coefficients as unity (ideal
(d) It has a relatively low solubility for hydrocarbon liquid and vapor phases). The chemical reaction equi-
which reduces the hydrocarbon loss when processing the libria are defined in the liquid phase. The equilibrium
natural gas and refinery gas streams. constants of the amine dissociation reaction and car-
The disadvantages of aqueous MEA include the bamate formation reaction are regressed based on the
following: experimental acid gas solubility data in the aqueous
(a) It has a high reaction heat with CO2 that leads to amine systems. The Kent-Eisenberg model has some
higher stripping energy consumption. deficiencies: the fit is good only in a limited loading
(b) It is unable to remove mercaptans. range, and its extension to the mixed acid gas is less
(c) It forms degradation products with COS, CS2, and successful. However, the model has been popular among
oxygen-bearing gases. practitioners because it correlates the experimental data
(d) It suffers more vaporization loss than other al- reasonably well while retaining extreme computational
kanolamines because of its higher vapor pressure. simplicity.
(e) It is appreciably more corrosive than many other A more rigorous and therefore more generally ap-
alkanolamines. plicable model was proposed by Deshmukh and Mather.5
In 1990, aqueous MEA comprised 40% of the market It uses the same chemical reactions in the liquid phase
(Dupart et al.2). Although its market share has de- as the Kent-Eisenberg model does, but instead of
creased in recent years, it remains as one of the assuming activity coefficients to be unity, these activity
principal acid gas absorbents with significant com- coefficients, except for water, are calculated by an
mercial importance. There are still strong commercial extended expression of the Debye-Hückel theory pro-
interests in minimizing equipment and operation costs posed by Guggenheim and Stokes.6 The activity coef-
for aqueous MEA absorption/stripping processes for CO2 ficient of water is still set as unity. The interaction
removal. Computer process simulation is an effective parameters in the Guggenheim-Stokes expression were
means for optimizing the design and operation of the obtained from regression of experimental VLE data.
Weiland and co-workers7 provided the values of the
* Corresponding author. Tel: 617-949 1201. Fax: 617-949 Guggenheim-Stokes interaction parameters for most
1030. E-mail: yunda_liu@aspentech.com. of the commercially important amine systems over a
10.1021/ie980600v CCC: $18.00 © 1999 American Chemical Society
Published on Web 03/26/1999
 Ind. Eng. Chem. Res., Vol. 38, No. 5, 1999 2081

wide range of conditions. Later, a similar model was tion forces among all species. The electrolyte NRTL
proposed by Li and Mather.8 The Guggenheim-Stokes expression was developed based on the NRTL local-
expression was replaced by the well-known Pitzer model composition concept (Renon and Prausnitz12), the like-
to calculate the activity coefficients. The Pitzer interac- ion repulsion assumption, and the local electroneutrality
tion parameters were determined from experimental assumption. The like-ion repulsion assumption stipu-
data for the MEA-CO2-H2O and methyldiethanol- lates that in the first coordination shell of a cation
amine (MDEA)-CO2-H2O systems. On the basis of (anion) the local composition of all other cations (anions)
these parameters, good VLE prediction for the MEA- is zero. The local electroneutrality assumption imposes
MDEA-CO2-H2O system was achieved. a condition that in the first coordination shell of a
Probably the most sophisticated and thermodynami- molecular species the composition of cations and anions
cally rigorous model available at this time was proposed is such that the local electric charge is zero.
by Austgen et al.1 This model was based on the The Pitzer-Debye-Hückel expression for excess Gibbs
electrolyte nonrandom-two-liquid (NRTL) model of Chen energy, normalized to a mole fraction of unity for the
and co-workers.9,10 Not only water but also alkanol- solvent and zero mole fraction for ions, is given as
amines were regarded as solvents. So, the solution was follows:
treated as a mixed-solvent system. The model accounted
for the rigorous chemical equilibria and mass balance

( )( )
and thus provided the insight of various ionic and gex*,PDH 1000 1/2 4AΦIx
molecular species in the liquid phase. The activity
coefficients in the liquid phase were represented with RT
) -( ∑k xk) Ms F
ln(1 + FIx1/2) (1)
the electrolyte NRTL model treating both long-range
ion-ion interactions and short-range interactions be-
tween all true species in the liquid phase. The fugacity where
coefficients in the vapor phase were represented using

( )( )( )
the Redlich-Kwong equation of state modified by
Soave.11 Under the absorber column condition of the 1 2πN0d 1/2
e2 2/3

alkanolamine-CO2 removal process, the vapor fugacity

AΦ )
3 1000 DkT
coefficients should not be neglected. The model is
1
particularly suitable for the column simulation based
on the kinetics and mass- and heat-transfer rates
Ix ) ∑
2 i
Zi2xi
because it provides vital information such as interface
VLE, bulk interface driving forces, and true liquid
species. The model by Austgen et al.1 has been used to The Born expression for excess Gibbs energy is given
simulate the aqueous MEA absorption/stripping pro- as follows:
cesses for CO2 removal using the rate-based approach.

( ) (∑ )
The simulation result of the absorber column was good.
However, the result of the stripper column was not gex*,Born e2 1 1
1/2
xiZi2
satisfactory; especially the reboiler duty was overpre- ) - × 10-2 (2)
dicted. Further improvement is, therefore, necessary. RT 2kT D Dw i ri
The objective of this work is to investigate the cause
which makes the model by Austgen et al.1 predict a high
reboiler heat duty of the stripper column and to improve where D stands for the dielectric constant of the solvent
the model so it is capable of accurately describing the mixture with the same solvent ratio as that in the
VLE of the system at both absorbing and stripping electrolyte solution. Dw is the dielectric constant of
conditions relevant to the common aqueous MEA ab- water.
sorption/stripping processes for CO2 removal. The local-composition electrolyte NRTL expression for
excess Gibbs energy is given as follows:
Electrolyte NRTL Model
The electrolyte NRTL model used by Austgen and co-
workers1 and in this work consists of three contribu- gex,lc
∑j XjGjmτjm
tions. The first contribution is the long-range contribu- ) ∑ Xm +
∑k XkGkm
tion represented by the Pitzer-Debye-Hückel RT m
expression, which accounts for the contribution due to

( )
the electrostatic forces among all ions. The second
contribution is an ion-reference-state-transfer contribu-
Xa′
∑j XjGjc,a′cτjc,a′c
∑c Xc∑
tion represented by the Born expression. In the elec-
trolyte NRTL model, the reference state for ionic species +
is always infinitely dilute state in water even when
a′
∑ Xa′′ ∑XkGkc,a′c
there are mixed solvents. The Born expression accounts

( )
a′′ k
for the change of the Gibbs energy associated with
moving ionic species from a mixed-solvent infinitely Xc′
∑j XjGja,c′aτja,c′a
dilute state to an aqueous infinitely dilute state. The ∑a Xa∑ (3)
Born expression drops out if water is the sole solvent
in the electrolyte system. The third contribution is a
c′
∑ c′′
Xc′′ ∑XkGka,c′a
k
short-range contribution represented by the local-
composition electrolyte NRTL expression, which ac-
counts for the contribution due to short-range interac- where
2082 Ind. Eng. Chem. Res., Vol. 38, No. 5, 1999

Gjm ) exp(-Rjmτjm) Ionization of water

Gjc,ac ) exp(-Rjc,acτjc,ac) 2H2O S H3O+ + OH- (6a)

Gja,ca ) exp(-Rja,caτja,ca) Hydrolysis and ionization of dissolved CO2

∑a XaGca,m CO2 + 2H2O S H3O+ + HCO3- (6b)

Gcm ) Dissociation of bicarbonate

∑
a′
Xa′ HCO3- + H2O S H3O+ + CO32- (6c)

Dissociation of MEAH+ (protonated MEA)

∑c XcGca,m
C2H8NO+ (MEAH+) + H2O S
Gam )
H3O+ + C2H7NO (MEA) (6d)
∑ Xc′
MEACOO- (carbamate) reversion to bicarbonate
c′

∑a XaRca,m C3H6NO3- (MEACOO-) + H2O S

Rcm ) C2H7NO (MEA) + HCO3- (6e)
∑
a′
Xa′
Additional reactions other than the above ones may
occur to produce nonregenerable heterocyclic compounds
∑c XcRca,m (Kohl and Nielsen13). We chose not to include these
degradation reactions because they are not considered
Ram ) important in the basic aqueous MEA absorption/strip-
∑
c′
Xc′ ping operation for CO2 removal. MEA may also react
with COS, CS2, and oxygen-bearing gases to form
degradation products. These reactions were not taken
τmc,ac ) τcm - τca,m + τm,ca into account in this work.
τma,ca ) τam - τca,m + τm,ca
Problem Analysis and Resolution
The variables τcm and τam are computed accordingly from The model by Austgen et al.1 has been used to
Gcm and Gam. It is worth mentioning that the first term simulate the aqueous MEA absorption/stripping pro-
on the right-hand side of eq 3 represents the short-range cesses for CO2 removal using the rate-based approach.
interaction contribution where the molecular species are The simulation result of the absorber column was good.
the local center and the second and third terms account However, the result of the stripper column was not
for the short-range interaction contributions where satisfactory; especially the predicted reboiler heat duty
cations and anions are the local center, respectively. was too large. The first objective of this work is to
After proper consideration of unsymmetrical conven- investigate the cause that makes the model predict a
tion for the solutes and ionic species, the complete excess high reboiler heat duty of the stripper. In an initial
Gibbs energy expression of the electrolyte NRTL model study, we carefully analyzed the heat consumed in the
is given as follows: stripper column.
In a conventional stripping operation, heat is supplied
to the column by steam or by other heat mediums within
gex* ) gex*,PDH + gex*,Born + gex*,lc (4) tubes in the reboiler. The heat is consumed mainly for
three purposes: (1) to raise the temperature of the rich
The activity coefficient for any species i, ionic or mo- solution feeding the column to the temperature of the
lecular, solute or solvent, is derived from the partial lean solution leaving the reboiler, (2) to reverse the
derivative of the excess Gibbs energy with respect to MEA-CO2 reactions and dissociate the MEA-CO2
the mole number of species i: compounds, and (3) to vaporize water into steam which
serves as the stripping vapor. We separated the amount

[ ]
ex* of heat consumed for each purpose based on the simula-
1 ∂(nt g ) tion result in the stripper condition using the model by
ln γi ) (5)
RT ∂ni T,P,nj*i Austgen et al.1 We found that the quantities of heat
consumed for the first two purposes were reasonable.
However, the heat consumed for the third purpose was
where nt is the total mole number for all species in the
too high because too much water was being vaporized.
system.
This indicates that the model does not accurately
describe the VLE under the stripper condition. As
Basic Solution Chemistry pointed out by Kohl and Nielsen,13 the inability to make
an accurate VLE description for the stripper column is
MEA is a primary amine that has two hydrogen a common problem for many models proposed for
atoms directly attached to a nitrogen atom. Following designing and simulating the amine acid-gas processes.
the work of Austgen and co-workers,1 the principal Although the following analysis is aimed at the model
reactions occurring in the aqueous MEA-CO2 system by Austgen et al.,1 it might be beneficial to other models
are written as as well.
 Ind. Eng. Chem. Res., Vol. 38, No. 5, 1999 2083

Now we can rephrase the objective of this work more sis and ionization of dissolved CO2), and 6c (dissociation
precisely as improving the VLE representation for the of bicarbonate) are well-grounded because these equi-
MEA-CO2-H2O system in the stripper condition using librium constants were fitted based on extensive ex-
the model by Austgen et al.1 In the work by Austgen et perimental data and had been used successfully in VLE
al., both water and MEA are regarded as solvents. So, representation of many other aqueous electrolyte sys-
the MEA-CO2-H2O system was treated as a mixed- tems containing CO2. We continue using these equilib-
solvent electrolyte system. The vapor-liquid equilibria rium constants without change.
for the solvent species are given by
For reaction 6d (dissociation of protonated MEA),

( )
vs(P - P0s ) Austgen and co-workers obtained the equilibrium con-
ysφvs P ) xsγsφ0s P0s exp (7) stant based on the experimental data of Bates and
RT Pinching.14 The temperature range of the Bates-Pinch-
ing data is only from 0 to 50 °C. The common temper-
The vapor-liquid equilibria for the supercritical com-
ature range in the stripper column is around 120 °C.
ponents are given by
The temperature extrapolation might cause improper

( ) ( )
∞ dissociation of protonated MEA at the stripper column
Hm
i vj i-m (P - P0m)
yiφvi P ) xiγi exp (8) which ultimately affects the accuracy of VLE calcula-
γ∞i RT tion. To ascertain whether the temperature extrapola-
tion by Austgen et al. is proper, we perturbed the
where equilibrium constant at the stripper condition around

() ( )
the extrapolation value of Austgen et al. and used the
Hm Hi-s perturbed value for the simulation of three industrial
∑s ws ln
i
ln ) (8a) cases. The simulation results suggested that the tem-
γ∞i ∞
γi-s perature extrapolation by Austgen et al. was in the
reasonable range. Because the equilibrium constant at
Hi-s is Henry’s constant for solute i, in the pure solvent the stripper condition is still unknown, we accepted the
s, at the system temperature and saturation pressure temperature extrapolation by Austgen et al. However,
∞
of the solvent s; γi-s is the infinite dilution activity we point out that there are many uncertain factors
coefficient for solute i, in the pure solvent s, at the impeding the accurate VLE description for the MEA-
system temperature; ws is the weight fraction for the CO2-H2O system. It would be well worth the effort to
solvent s, based on the true solvent composition ratio eliminate this uncertainty by directly measuring the
in the MEA-CO2-H2O system; vs is the molar volume equilibrium constant at the stripper condition.
of the pure solvent at the system temperature and As for reaction 6e (carbamate reversion to bicarbon-
∞
saturation pressure P0s ; vj i-m is the partial molar vol- ate), one section in the paper by Austgen and co-
ume for solute i, at infinite dilution in the solvent workers1 has been specifically dedicated to the discus-
mixture at the system temperature and saturation sion of the reaction. The discussion basically states that
pressure of the solvent mixture P0m; x is the true mole the experimental equilibrium constant reported in the
fraction in the liquid phase which is based on all species, literature was very limited for the carbamate reaction.
molecular and ionic, after consideration of solution
Furthermore, the quality of the literature data was not
chemistry.
well justified for the direct use. Therefore, Austgen and
It should be noted that, in the work by Austgen et
al., a simplification was made that Henry’s constants co-workers chose to treat the carbamate equilibrium
in pure MEA are replaced by those in pure water. The constant as an adjustable parameter. It was fitted,
Poynting correction of eq 8, i.e., the exponential term, simultaneously with the appropriate interaction param-
was changed accordingly. eters of the electrolyte NRTL model, on the VLE data
After closely examining the vapor-liquid equilibria of the MEA-CO2-H2O system. On the basis of our
of the MEA-CO2-H2O system at the stripper condition experiences, the simultaneous fit of both the equilibrium
combining the all-around analysis of eqs 7 and 8, we constant and the interaction parameters might result
decided to pursue the improvement by considering (1) in overfitting of the data. It has been noticed that the
equilibrium constants for the reactions in the liquid VLE measurement for the MEA-CO2-H2O system may
phase, (2) Henry’s constant of CO2 in the MEA-H2O have errors up to and even beyond 40% (detail can be
solvent mixture, (3) evaluation of experimental data, seen in the subsequent section entitled Evaluation of
and (4) equilibrium constant and interaction parameters Experimental Data). Overfitting of these data is likely
from data regression. to have a negative effect on the actual process simula-
It is worth mentioning that, because the pressure of tion calculation. Weiland and co-workers7 proposed a
the stripper column is low, the Poynting correction in two-stage procedure for fitting the VLE data for the
eqs 7 and 8 is not a key influential factor on VLE of the alkanolamine (including MEA)-CO2-H2O systems. First,
system under the stripper condition. the equilibrium constants were fitted based on the VLE
Equilibrium Constants for the Reactions in the data. Second, based on the same VLE data, the interac-
Liquid Phase. Equilibrium constants govern the extent tion parameters were adjusted to achieve the best
of reactions 6a-e and determine the true composition
possible results. In this work, we adopted the two-stage
in the liquid phase, thereby greatly influencing the VLE
procedure by Weiland et al. Therefore, the equilibrium
of the MEA-CO2-H2O system.
The equilibrium constants of reactions 6a-e in the constant for reaction 6e was obtained from the first-
model by Austgen et al.1 were carefully examined. We stage fitting.
found that the equilibrium constants used by Austgen In the model by Austgen et al.1 and in this work, the
et al. for reactions 6a (ionization of water), 6b (hydroly- temperature function of equilibrium constants for reac-
2084 Ind. Eng. Chem. Res., Vol. 38, No. 5, 1999

Table 1. Temperature Dependence of the Equilibrium Table 2. Temperature Dependence of Henry’s Constants
Constants for Reactions 6a-e for Carbon Dioxide

ln K ) Ck1 + Ck2/T + Ck3 ln T + Ck4T ln Hi-s ) CH H H H

1 + C2 /T + C3 ln T + C4 T

reaction Ck1 Ck2 Ck3 Ck4 ref CH

1 CH
2 CH
3 CH
4 ref
6a 132.899 -13445.9 -22.4773 0.0 a CO2-H2O 170.7126 -8477.711 -21.95743 0.005 780 748 a
6b 231.465 -12092.1 -36.7816 0.0 a CO2-MEA 89.452 -2934.6 -11.592 0.016 44 b
6c 216.049 -12431.7 -35.4819 0.0 a a Austgen et al.1 b Equation 10 using data from Austgen et al.,1
6d 0.79960 -8094.81 0.0 -0.007484 b
6e 1.282562 -3456.179 0.0 0.0 c Wilhelm et al.,17 and Wang et al.16
a Austgen et al.1 b Austgen et al.,1 with MEA infinitely dilute

activity coefficient correction. c Fitted on the VLE data.

tion 6a-e is represented as

Ck2
ln K ) Ck1 + + Ck3 ln T + Ck4T (9)
T

K is the equilibrium constant which is dimensionless.

Coefficients, Ck1-Ck4, used in this work are summarized
in Table 1 for reactions 6a-e. Note that the coefficients
of reaction 6d are somewhat different from those in the
model by Austgen et al.1 That is because in this work
the MEA-H2O interaction parameter values have been
changed, which produce the different values of the
infinitely dilute activity coefficient for MEA in water,
and the coefficients of reaction 6d in Table 1 reflect this Figure 1. Temperature dependence of Henry’s constants of CO2
change accordingly. in water and in MEA.
Henry’s Constant of CO2 in the MEA-H2O Sol-
vent Mixture. Henry’s constant of CO2 has a direct The CO2-N2O analogy method is based on the
effect on the VLE of the MEA-CO2-H2O system. In the proposal by Clarke15 that if CO2 reacts with a solvent,
model by Austgen et al.,1 both H2O and MEA are treated its true solubility can be determined by the proportion-
as solvents. According to eq 8a, we need to know Henry’s ality to the solubility of a similar gas that is unreactive
constants of CO2 in the two pure solvents: water and to the solvent. N2O is usually used as the similar gas
MEA. Henry’s constant of CO2 in H2O used by Austgen for CO2 with regard to pure alkanolamines because it
et al. is well-established and well-tested. Therefore, it has molecular weight, configuration, volume, electronic
was kept in this work without modification. Because of structure, and Lennard-Jones potential very similar to
the chemical reactions, Henry’s constant of CO2 in MEA those of CO2 and it is unreactive to alkanolamines. The
cannot be directly measured. Austgen and co-workers CO2-N2O analogy method in the Henry’s constant form
assumed that its value would be the same as that in is expressed as
water. The assumption simplified the calculation. How-
ever, it might introduce errors in the final simulation. HN2O-MEA
Because the available experimental VLE data of the HCO2-MEA ) HCO2-H2O (10)
MEA-CO2-H2O system are the CO2 solubility data, HN2O-H2O
adjustable parameters are fitted according to eq 8 only.
The electrolyte NRTL model for activity coefficient is a In this work, the temperature dependence of Henry’s
thermodynamically rigorous model and satisfies the constants has the same functional form as that of eq 9.
Gibbs-Duhem relationship. Thus, the activity coef- The units are Pascals. Table 2 gives the coefficients of
ficients for the solvents, i.e., water and MEA, in eq 7 Henry’s constants of CO2 in water and in MEA. The
are calculated using the fitted parameters based on eq former is just the same as that in the work by Austgen
8. When Henry’s constant for CO2 in pure MEA is et al.1 The latter is obtained using eq 10, where the
assumed as that in water, it will influence the fitted required Henry’s constant for N2O in MEA is from Wang
parameter values according to eq 8. Consequently, it will et al.16 and that for N2O in water is from Wilhelm et
affect the activity coefficient of water and MEA, which al.17 Figure 1 shows the temperature dependency of
as a result changes the partial vapor pressure of water Henry’s constants of CO2 in water and in MEA. Noted
and MEA. Evidently, the effect increases when the MEA that the two Henry’s constant lines intersect at about
concentration increases. Also, when the CO2 loading 3.8 °C (276.95 K), which is much lower than the
decreases, more MEA stays in molecular form and the intersecting point of 40 °C for CO2 in water and in
assumption becomes less suitable. In the stripper MDEA reported by Posey.18
column, the partial vapor pressure of water becomes To verify Henry’s constant of CO2 in MEA obtained
significant and the assumption might cause significant using eq 10, we designed a plant-data-optimization
errors. method to obtain Henry’s constant of CO2 in MEA
In this work, we employed two indirect methods to independently. In essence, Henry’s constant of CO2 in
obtain Henry’s constants of CO2 in pure MEA. One is MEA was treated as an adjustable parameter to opti-
the CO2-N2O analogy method, and the other is the mize the results of three industrial cases for CO2
plant-data-optimization method. removal. It was a surprise to find that Henry’s constants
 Ind. Eng. Chem. Res., Vol. 38, No. 5, 1999 2085

Figure 2. Partial pressure and relative deviation of CO2 in a MEA Figure 3. Comparison of the sum of partial pressures for MEA
(30 wt %)-H2O system at 120 °C with normal CO2 loading as well and H2O in a MEA (30 wt %)-H2O system with the CO2 loading
as those perturbed -0.04. The model with the fitted parameter from 0 to 1 at the temperature of 120 °C.
from this work was used in the calculation. The relative deviation
is calculated by (Pptb - Pnormal)/Pnormal %.
standard for comparison, the lower part in Figure 2
illustrates that an error of -0.04 CO2 loading can
of CO2 in MEA obtained from the two independent produce around and beyond 40% deviation on the partial
methods match so closely with each other. The relative pressure of CO2. We also found one factor that may be
differences are smaller than 0.8%. The verification overlooked when measuring the VLE data for the
indicates that Henry’s constants of CO2 in MEA ob- MEA-CO2-H2O system. Using a common static ap-
tained from eq 10 can be used with confidence. paratus to measure the solubility of CO2 in a MEA-
Evaluation of Experimental Data. Because MEA H2O system, Jou et al.24 used Raoult’s law to calculate
is one of the first alkanolamines commercially used for the sum of partial pressures for MEA and H2O and then
treating acid gas and is still widely used, a large amount subtracted the value from the total pressure to obtain
of experimental VLE data for the MEA-CO2-H2O the sum of partial pressures for CO2 and an inert gas,
system have been published. A detailed list for the most which was added to maintain the apparatus pressure.
published papers presenting the experimental data on The nonideality in the MEA-H2O system was neglected
the subject can be found in the book by Kohl and and so was the effect of the dissolved CO2 on the partial
Nielsen.13 pressures of MEA and water. These negligences might
Austgen and co-workers used experimental VLE data cause noticeable errors in the final results, especially
of the MEA-CO2-H2O system from five different in the high-temperature region. Figure 3 shows a
sources in the data regression to determine the adjust- comparison of the sum of partial pressures for MEA and
able parameters of the model. The five data sources are H2O in a MEA (30 wt %)-H2O system with a CO2
Lee et al.,19 Isaacs et al.,20 Lawson and Garst,21 Jones loading from 0 to 1 at the temperature of 120 °C using
et al.,22 and Muhlbauer and Monaghan.23 There is three different models. The dashed line was calculated
significant inconsistency of the experimental data among using the model of Austgen et al.1 The dot-dashed line
the five sources. Because of lack of the information was obtained based on Raoult’s law, which was used
regarding the data quality of these sources, Austgen and by Jou et al.24 to deduce the partial pressure of CO2 from
co-workers thought that the best approach was to treat the raw measuring data. The solid line was calculated
the five data sources equally with a weighting factor of using the model with a fitted parameter from this work.
1 for each of them. It is clear that over 0.5 CO2 loading the nonideality of
In this work, we studied the quality of the experi- the MEA-H2O system as well as the effect of the
mental VLE data for the MEA-CO2-H2O system. We dissolved CO2 should not be neglected. What we found
found that a key factor that is critical to the data quality explains, in part, why there is significant inconsistency
is the accuracy of the measured CO2 concentration in of the experimental VLE data for the MEA-CO2-H2O
the liquid phase. In the normal CO2 loading (moles of system among various sources. A quality evaluation of
CO2 per moles of MEA) range, the partial pressure of the various data sources is highly necessary. The
CO2 arises so fast that it spans about 8 orders of accurate and reliable VLE measurement on the MEA-
magnitude. Therefore, any small measurement error of CO2-H2O system still remains as one of the most
the CO2 concentration will cause a large deviation in challenging tasks for the chemical engineering experi-
the partial pressure of CO2. For example, Jou et al.24 mentalists.
pointed out that there was a systematic error of about In this work, we followed the work of Weiland and
-0.04 mol of CO2/mol of MEA in the VLE measurement co-workers7 in a quality evaluation of the various data
for the MEA-CO2-H2O system by Lee et al.25 The sources. The electrolyte NRTL model was used to do a
upper part in Figure 2 shows a partial pressure of CO2 preliminary two-stage fit of the VLE data for the MEA-
in a MEA (30 wt %)-H2O system at 120 °C with normal CO2-H2O system from six sources: five used by Aust-
CO2 loading as well as with the CO2 loading shifted gen et al.1 and one from Jou et al.24 The fitting error of
-0.04. The model with a fitted parameter from this each data source was analyzed, and the data sources
work was used in the calculation. Suppose the calcu- that appeared to deviate significantly were excluded in
lated partial pressure of CO2 corresponding to the the final data regression. The excluded data sources
normal CO2 loading is correct and is used as the were those from Isaacs et al.,20 Jones et al.,22 and
2086 Ind. Eng. Chem. Res., Vol. 38, No. 5, 1999

Table 3. Fitted Interaction Parameter Values for the Table 3a,b were regarded as unimportant, which include
MEA-CO2-H2O and CO2-H2O Systems Using the two molecule-molecule, most of the molecule-ion pair
Electrolyte NRTL Modela (ion pair-molecule), and all of the ion pair-ion pair
a. MEA-CO2-H2O System interaction parameters. These unimportant parameters
molecule-molecule were assigned a default value of zero. From the theo-
parameter Amm′ Bmm′ ref retical point of view, the default value of zero essentially
reduces the electrolyte NRTL model to the Pitzer-
MEA-H2O -1.609 693 -265.1962 b
H2O-MEA 1.648 945 125.0832 b Debye-Hückel equation for the relevant ionic species
with an insignificant amount and to Raoult’s law for
molecule-ion pair the relevant molecular species. Note that there is a
(ion pair-molecule) difference in the parameter defaulting (except the
parameter τm,ca(τca,m) parameters listed in Table 3b) between this work and
H2 O-(MEA+,HCO3 -) 0.129 079 6 c the work by Austgen et al.1 Austgen and co-workers
(MEA+, HCO3-)-H2O -0.424 708 7 c used 8 and -4 as default values for the water-ion pair
H2O-(MEA , MEACOO-)
+ 4.352 250 c and ion pair-water interaction parameters, respec-
(MEA+, MEACOO-)-H2O -3.340 842 c
tively. They used 15 and -8 as default values for MEA-
b. CO2-H2O Systemd ion pair and ion pair-MEA as well as CO2-ion pair and
molecule-molecule ion pair-CO2 interaction parameters, respectively.
parameter Amm′ Bmm′ Rmm′ In this work, the temperature dependence of the
molecule-molecule interaction parameters is given as
CO2-H2O 10.064 -3268.135 0.20
H2O-CO2 10.064 -3268.135 0.20
τmm′ ) Amm′ + Bmm′/T (11)
molecule-ion pair
(ion pair-molecule) The molecule-ion pair (ion pair-molecule) interaction
parameter τm,ca(τca,m) Rm,ca(Rca,m) parameters are assumed to be temperature indepen-
H2O-(H3O+, HCO3-) 8.0 0.2 dent. The NRTL nonrandomness factor is fixed at 0.2
(H3O+, HCO3-)-H2O -4.0 0.2 for molecule-molecule, water-ion pair, and ion pair-
H2O-(H3O+, OH-) 8.0 0.2 water interaction parameters.
(H3O+, OH-)-H2O -4.0 0.2
H2O-(H3O+, CO32-) 8.0 0.2
Data regression for determining the parameter values
(H3O+, CO32-)-H2O -4.0 0.2 was carried out using an algorithm derived by Britt and
CO2-(H3O+, HCO3-) 15.0 0.1 Luecke.26 The objective function is formulated based on
(H3O+, HCO3-)-CO2 -8.0 0.1 the maximum likelihood principle and is given as
CO2-(H3O+, OH-)

[ ]
15.0 0.1
(H3O+, OH-)-CO2 -8.0 0.1 ND NV
(Zcal - Zexp) 2

∑l ∑l′
CO2-(H3O+, CO32-) 15.0 0.1
(H3O+, CO32-)-CO2 -8.0 0.1 F) (12)
σ l,l′
a NRTL nonrandomness factors for these fitted parameters were

fixed at 0.2 except CO2-ion pair (ion pair-CO2) parameters. where the summation is over all of the measured
b Nath and Bender,27 Touhara et al.,28 and Kohl and Nielsen.13
c Lee et al.,19 Lawson and Garst,21 and Jou et al.24 d The parameter variables, NV, for all of the data points, ND. Zcal and Zexp
values were obtained from Austgen et al.1 are calculated and experimental values, respectively. σ
is the standard deviation of measurement for the lth
Muhlbauer and Monaghan.23 Weiland and co-workers7 data point of the l′th measured variable. To make the
also observed that these three data sources were less comparison easier between the data regression results
accurate. For the data source of Jou et al.,24 we from the work of Austgen et al.1 and those from this
eliminated the data points over 0.5 CO2 loading for the work, we used the same standard deviation values
reason mentioned in the above paragraph. assigned by Austgen et al.1
Equilibrium Constant and Interaction Param- The MEA-H2O and H2O-MEA interaction param-
eters from Data Regression. Equilibrium constant eters were fitted based on the experimental data for the
of carbamate reversion reaction (reaction 6e) and some MEA-H2O system. The experimental data used by
of the interaction parameters of the electrolyte NRTL Austgen et al.1 were the VLE data only. Posey18 used
model were obtained from the regression of the relevant the VLE data, the heat-mixing data, and the freezing-
experimental data. point data altogether for the fitting and obtained more
In the MEA-CO2-H2O system, there are three mo- reliable parameter values. In this work, the VLE data,
lecular species and six ionic species created by reactions the heat-mixing data, the heat capacity data, and the
6a-e. The corresponding interaction parameters in the pH data were fitted simultaneously and the fitted values
electrolyte NRTL model will be 6 molecule-molecule, of the MEA-H2O and H2O-MEA interaction param-
24 molecule-ion pair (ion pair-molecule), and 54 ion eters are given in Table 3a.
pair-ion pair. Because about half of the species are In the work of Austgen et al.,1 the equilibrium
present in the liquid phase at a low or negligible constant of the carbamate reversion reaction (reaction
amount, the parameters associated with them do not 6e) and the four important water-ion pair (ion pair-
significantly affect the VLE representation for the water) interaction parameters of the electrolyte NRTL
system. In this work, only the interaction parameters model were simultaneously fitted based on the VLE data
that significantly affect the VLE representation were for the MEA-CO2-H2O system. As discussed in the
fitted to the experimental data. They are listed in Table section Equilibrium Constants for the Reactions in the
3a. Table 3b shows the interaction parameters related Liquid Phase, we adopted the two-stage procedure
to the binary CO2-H2O system. The values of those proposed by Weiland et al.7 in the data regression. In
parameters were obtained directly from the work of the first stage, the equilibrium constant of reaction 6e
Austgen et al.1 All other parameters not reported in was fitted with the values of the four important water-
 Ind. Eng. Chem. Res., Vol. 38, No. 5, 1999 2087

Table 4. Comparison between the Model of Austgen et al.1 and Its Improved Version of This Work on the
Root-Mean-Square Relative Deviation of Partial Pressure for CO2, σP,CO2, as Well as That of the Liquid-Phase Mole
Fraction for CO2, σx,CO2, at the Temperature Range of 25-120 °C and the MEA Concentration Range of 1.0-5.0 Ma

temp, MEA concn 1.0 M MEA concn 2.5 M MEA concn 3.75 M MEA concn 5.0 M
°C σP,CO2 σx,CO2 σP,CO2 σx,CO2 σP,CO2 σx,CO2 σP,CO2 σx,CO2
This Work
25 0.0545 0.0616 0.0178 0.0280 0.0071 0.0197 0.0163 0.0260
40 0.0546 0.0762 0.1020 0.0489 0.0189 0.0384 0.0113 0.0223
60 0.0314 0.0477 0.0139 0.0237 0.0082 0.0145 0.0088 0.0154
80 0.0551 0.0642 0.0355 0.0306 0.0318 0.0293 0.0354 0.0403
100 0.0886 0.0482 0.0422 0.0318 0.0160 0.0191 0.0257 0.0379
120 0.0531 0.0427 0.0964 0.0728 0.1050 0.0839 0.1130 0.0987
average 0.0588 0.0584 0.0610 0.0415 0.0429 0.0394 0.0468 0.0463
Austgen et al.
25 0.0407 0.0768 0.0255 0.0641 0.0283 0.0694 0.0312 0.0661
40 0.0677 0.1080 0.1730 0.1100 0.0688 0.1217 0.0645 0.1120
60 0.0586 0.0821 0.0642 0.0858 0.0668 0.0889 0.0561 0.0778
80 0.0692 0.0720 0.0642 0.0740 0.0660 0.0704 0.0764 0.0686
100 0.1070 0.0699 0.1170 0.0874 0.1260 0.0842 0.1490 0.0920
120 0.0977 0.0833 0.1220 0.1020 0.1160 0.1040 0.0975 0.1020
average 0.0762 0.0831 0.1050 0.0882 0.0842 0.0912 0.0871 0.0875
a The results were based on the experimental data from Lee et al.19

ion pair (ion pair-water) interaction parameters de-

faulted to zero. In the second stage, the values of the
equilibrium constant for reaction 6e obtained in the first
stage were kept unchanged and the four relevant
interaction parameters were fitted to achieve the best
possible results. In both stages, the same VLE data for
the MEA-CO2-H2O system were used. The fitted
values from the first stage can be found in Table 1 and
those from the second stage in Table 3a. Because both
water and MEA were treated as solvents, when proper
MEA-H2O and H2O-MEA interaction parameters were
provided, a theoretically more sound way to obtain the
values of the equilibrium constant for reaction 6e is to
use only the low loading VLE data in the first-stage
fitting, and then in the second stage the four relevant
interaction parameters are fitted based on theVLE data
of the whole loading range. However, when the model
with parameter values fitted using this approach was
applied to simulate three industrial cases for CO2
removal, the results showed a general trend that the
reboiler heat duties were somewhat underpredicted. We
think that it might be caused by the poor accuracy of
the low loading VLE data for the MEA-CO2-H2O
system.
Table 4 gives a comparison between the model of
Austgen et al.1 and its improved version of this work
on the root-mean-square relative deviation of the partial
pressure for CO2 as well as that of the liquid-phase mole
fraction for CO2 at the temperature range of 25-120
°C and the MEA concentration range of 1.0-5.0 M. The
comparison is based on the result from fitting the VLE
data of Lee et al.,19 which is the main data source in
this work. It is apparent that the root-mean-square
relative deviations from this work are smaller than
those from Austgen et al.1 at all of the temperature and
concentration ranges. The main reason for that is that Figure 4. Liquid-phase speciation and concentration in a 2.5 M
MEA solution with CO2 loadings from 0 to 1 at temperatures of
we used fewer but more consistent data sources in the (a) 40 and (b) 120 °C. The speciation and concentration were
data regression. predicted using the liquid-phase reactions of 6a-e as well as the
Parts a and b of Figure 4 are the speciation- electrolyte NRTL model with the parameter values fitted in this
concentration plots which show how the concentration work.
of each key species varies with the CO2 loading from 0
to 1 in a 2.5 M MEA solution at temperatures of 40 and Note that the speciation-concentration plots from this
120 °C, respectively. The speciation and concentration work are very similar to the corresponding speciation-
of each key species were predicted using the liquid- concentration plots from the work by Austgen et al.1 The
phase reactions of 6a-e as well as the electrolyte NRTL speciation-concentration plot is a useful tool in analyz-
model with the parameter values fitted in this work. ing simulation results of the aqueous MEA absorber
2088 Ind. Eng. Chem. Res., Vol. 38, No. 5, 1999

Table 5. Comparison against Plant Data between the Model of Austgen et al.1 and This Work on the Simulation of
Three Industrial Cases for CO2 Removal Using the Aqueous MEA Absorption/Stripping Processesa
case 1 case 2 case 3
Austgen et al. this work Austgen et al. this work Austgen et al. this work
Absorber
gas inlet temp, F 0.00 0.00 0.00 0.00 0.00 0.00
gas inlet pressure, psi 0.00 0.00 0.00 0.00 0.00 0.00
gas outlet temp, F 0.02 0.02 b b b b
gas outlet pressure, psi -0.46 -0.46 0.00 0.00 -0.02 -0.02
gas outlet CO2, ppmv 211.11 0.00 206.67 26.67 -70.37 -14.81
lean inlet temp, F 0.02 0.02 0.48 0.48 0.03 0.03
rich outlet temp, F -3.76 -3.43 b b 0.58 -0.51
Stripper
rich inlet temp, F 0.25 0.11 b b b b
lean outlet temp, F b b -0.16 0.12 0.07 0.26
lean outlet pressure, psi b b 0.00 0.00 0.00 0.00
reboiler duty, mmbtu/h 58.43 8.99 13.73 -5.49 35.11 4.58
condenser temp, F 0.00 0.00 b b b b
loading, mol/mol
lean solution 0.00 0.00 0.00 0.00 0.00 0.00
rich solution -9.85c -9.85c b b 0.00 0.00
a The results are in percent and expressed as (Zpred - Zplant)/Zplant %. b Plant data are not available. c Plant data does not satisfy mass

balance.

column for CO2 removal using mass- and heat-transfer considers (1) material and energy balance, (2) mass and
rate-based simulation models. energy transfer rate, (3) vapor-liquid equilibrium, and
(4) liquid-phase reactions and reaction rate on each
Process Simulation actual tray (or section of packing). Generally speaking,
this approach describes what actually occurs on a tray
The ultimate test of the capability and accuracy of (or section of packing) rather than the phase equilibrium
the model improved in this work was to apply it in the that does not exist anywhere in the column. Therefore,
simulation of the aqueous MEA absorption/stripping the concept of the tray (or section of packing) efficiency
processes for CO2 removal. The availability of the is avoided entirely. It is worth mentioning that the
operating data of the process, obtained under stable state-of-the-art rate-based approach with accurate ther-
conditions, as well as the relevant equipment data was modynamic and transport property models is a very
indispensable for such simulations. Although the com- useful tool to develop and operate more energy-efficient
plete and consistent sets of the operating and equipment alkanolamine acid-gas absorption/stripping processes
data for the industrial processes using aqueous MEA and to achieve competitive edges in the sophisticated
for CO2 removal were not so easy to find, we had the column design, especially in the design of the acid-gas
opportunity to obtain the plant data for three industrial selective absorption where too many trays (or sections
cases. of packing) can destroy the selectivity while too few can
All three cases used the basic process flow scheme in cause outlet gas off specification.
conventional alkanolamine acid-gas absorption/strip- For simulation of the stripping column, the traditional
ping processes. Gas to be purified is passed upward phase-equilibrium approach is applied. According to the
through the absorber column, countercurrent to the discussion on stripping by Kohl and Riesenfeld,29 we
CO2-lean aqueous MEA solution. The CO2-rich aqueous specified about 10 theoretical trays below the CO2-rich
MEA solution from the bottom of the absorber column solution feeding point as the stripping section of the
is heated by the heat exchange with the CO2-lean column and two theoretical trays above the point as the
aqueous MEA solution from the bottom of the stripping MEA recovering section.
column and is then fed to the stripping column at some Table 5 shows the simulation results for the three
point near the top. After CO2 is stripped, the CO2-lean industrial cases using the model improved in this work.
aqueous MEA solution from the bottom of the stripping For comparison, the model of Austgen et al.1 was also
column is first partially cooled in the lean-to-rich applied to simulate the three industrial cases, and the
solution exchanger and is then further cooled by heat results are included in Table 5. Both simulation results
exchange with water or air and thereafter is fed to the were expressed as
top of the absorber column to complete the cycle. Despite
the fact that all three cases use a similar process flow
(Zpred - Zplant)
scheme, the operating conditions and equipment data %
of the three cases vary widely. The absorber pressure Zplant
ranges from about 180 to 950 psi. The lean CO2 loading
(mole of CO2 per mole of MEA) ranges from about 0.1 where Zpred is the predicted values and Zplant is the
to 0.2 and the rich CO2 loading from about 0.3 to 0.55. corresponding plant data. In case 1, the reaction rate
The MEA concentration ranges from about 20 to 30 wt (kinetics) provided by the Aspen Plus process simulator
%. The column type can be tray column or packed was applied without modification in the absorber simu-
column, and the column diameter ranges from about 1 lation using the rate-based approach, and the simula-
to 2 m. tion result was very satisfactory. However, when we
The rate-based approach (RATEFRAC) in Aspen Plus applied the same reaction rate in case 2, we found that
process simulator was used to simulate the aqueous some modification on the reaction rate was necessary.
MEA absorber column for CO2 removal. This approach Note that the reaction rate is one of the key issues for
 Ind. Eng. Chem. Res., Vol. 38, No. 5, 1999 2089

a successful simulation of alkanolamine acid-gas ab- P0s ) saturation pressure of solvent s

sorption/stripping processes using the rate-based ap- R ) gas constant
proach and currently is an active research area. We T ) temperature, K
intend to discuss this topic in a subsequent paper. In X ) effective mole fraction
case 3, because of the lack of equipment data, the Z ) absolute value of the ionic charge
absorber column was simulated using the traditional Zcal ) calculated value in eq 12
phase-equilibrium approach with proper consideration Zexp ) experimental value in eq 12
of tray efficiencies. From Table 5, one can see without Zpred ) predicted value in simulation
a doubt that the model improved in this work gives Zplant ) plant data
better simulation results upon the three industrial cases e ) electron charge
than the model by Austgen et al. The reboiler heat d ) solvent density
duties predicted by the model improved in this work g ) Gibbs energy
were within the uncertainties of the plant data for all k ) Boltzmann constant
three industrial cases. nt ) total mole number for all species in the system
ni ) mole number of species i
Conclusion r ) Born radius
v ) molar volume
In comparison to the work of Austgen et al.,1 the vj ) partial molar volume
electrolyte NRTL model has been applied in a thermo- ws ) weight fraction
dynamically consistent manner to provide a better x ) true liquid-phase mole fraction based on all species:
representation of the vapor-liquid equilibrium (VLE) molecular and ionic
for the aqueous MEA-CO2 system. Rigorous chemical y ) vapor-phase mole fraction
equilibria in the liquid phase have been considered. It
has been found that the accurate VLE description of the Greek Letters
system at a temperature of around 120 °C is critical to R ) NRTL nonrandomness factor
a successful simulation of the stripper column of the γ ) activity coefficient
aqueous MEA absorption/stripping processes for CO2 F ) closest approach parameter of the Pitzer-Debye-
removal. The influence from chemical equilibrium con- Hückel equation
stants, Henry’s constant, experimental data, and data σ ) measured standard deviation in eq 12 or root-mean-
regression on the VLE of the system has been discussed squares relative deviation in Table 4
in detail. The equilibrium constant of the carbamate τ ) NRTL interaction parameter
reversion reaction as well as important interaction φ0 ) pure solvent vapor-phase fugacity coefficient at
parameters of the electrolyte NRTL model have been saturation pressure
fitted to experimental data using a two-stage procedure, φv ) vapor-phase fugacity coefficient
and good agreement between the calculated values and
the experimental data has been achieved. The model Superscripts
with newly fitted parameters has been successfully * ) unsymmetrical convention
applied to simulate three industrial cases for CO2 ∞ ) infinite dilution
removal using the rate-based approach. For comparison, Born ) Born contribution
the model of Austgen et al.1 has also been applied to ex ) excess property
simulate the same cases. The simulation results are in H ) Henry’s constant
favor of the model from this work. Currently, efforts are k ) equilibrium constant
being made on the accurate VLE representation for the lc ) local-composition NRTL equation
aqueous DEA acid-gas system, the aqueous MDEA nD ) number of data points
acid-gas system, and the mixed alkanolamine acid- nV ) number of measured variables
gas systems. 0 ) saturation
PDH ) Pitzer-Debye-Hückel equation
Acknowledgment w ) water

The authors are very grateful to Dr. Navin Patel for Subscripts
helpful discussions and to Dr. Pao-Chen Wu for his a, a′, a′′ ) anion
input on building one of the simulation files and for his c, c′, c′′ ) cation
help on the evaluation of the MEA data package in the i, j, k ) any species
Aspen Plus process simulator. i-m ) solute i in solvent mixture
i-s ) solute i in solvent s
Nomenclature l ) lth data point
l′ ) l′th measured variable
A, B ) constants in eq 11 m, m′ ) molecular species
C ) temperature coefficient P ) pressure
D ) dielectric constant s ) solvent
F ) objective function w ) water
H ) Henry’s constant x ) true mole fraction
K ) equilibrium constant
M ) molarity Literature Cited
Ms ) solvent molecular weight
(1) Austgen, D. M.; Rochelle, G. T.; Peng, X.; Chen, C.-C. Model
N0 ) Avogadro’s number
of Vapor-Liquid Equilibria for Aqueous Acid Gas-Alkanolamine
P ) pressure Systems Using the Electrolyte-NRTL Equation. Ind. Eng. Chem.
P0m ) saturation pressure of the solvent mixture Res. 1989, 28, 1060.
2090 Ind. Eng. Chem. Res., Vol. 38, No. 5, 1999

(2) Dupart, M. S.; Bacon, T. R.; Edwards, D. J. Understanding (17) Wilhelm, E.; Battino, R.; Wilcock, R. J. Low-pressure
Corrosion in Alkanolamine Gas Treating Plants. Hydrocarbon Solubility of Gas in Liquid Water. Chem. Rev. 1977, 77, 219.
Process. 1993, 72, 89. (18) Posey, M. L. Thermodynamic Models for Acid Gas Loaded
(3) Kent, R. L.; Eisenberg, B. Better Data for Amine Treating. Aqueous Alkanolamine Solutions. Ph.D. Dissertation, The Uni-
Hydrocarbon Process. 1976, 2, 87. versity of Texas at Austin, Austin, TX, 1996.
(4) Klyamer, S. D.; Kolesnikova, T. L.; Rodin, Y. A. Equilibrium (19) Lee, J. I.; Otto, F. D.; Mather, A. E. Equilibrium between
in Aqueous Solutions of Ethanolamines during the Simultaneous Carbon Dioxide and Aqueous Monoethanolamine Solutions. J.
Absorption of Hydrogen Sulfide and Carbon Dioxide from Gases. Appl. Chem. Biotechnol. 1976, 26, 541.
Gazov. Promst. 1973, 18, 44. (20) Isaacs, E. E.; Otto, F. D.; Mather, A. E. Solubility of
(5) Deshmukh, R. D.; Mather, A. E. A Mathematical Model for Mixtures of H2S and CO2 in a Monoethanolamine Solution at Low
Equilibrium Solubility of Hydrogen Sulfate and Carbon Dioxide Partial Pressures. J. Chem. Eng. Data 1980, 25, 118.
in Aqueous Alkanolamine Solutions. Chem. Eng. Sci. 1981, 36, (21) Lawson, J. D.; Garst, A. W. Gas Sweetening Data: Equi-
355. librium Solubility of Hydrogen Sulfide and Carbon Dioxide in
(6) Guggenheim, E. A.; Stokes, R. H. Activity Coefficients of Aqueous Monoethanolamine and Aqueous Diethanolamine Solu-
2:1 and 1:2 Electrolytes in Aqueous Solution from Isopiestic Data. tions. J. Chem. Eng. Data 1976, 21, 20.
Trans. Faraday Soc. 1958, 54, 1646. (22) Jones, J. H.; Froning, H. R.; Claytor, E. E. Solubility of
(7) Weiland, R. H.; Chakravarty, T.; Mather, A. E. Solublity of Acidic Gases in Aqueous Monoethanolamine J. Chem. Eng. Data
Carbon Dioxide and Hydrogen Sulfide in Aqueous Alkanolamines. 1959, 4, 85.
Ind. Eng. Chem. Res. 1993, 32, 1419.
(23) Muhlbauer, H. G.; Monaghan, P. R. Sweetening Natural
(8) Li, Y.-G.; Mather, A. E. Correlation and Prediction of the
Gas with Ethanolamine Solutions. Oil Gas J. 1957, 55, 139.
Solubility of Carbon Dioxide in a Mixed Ethanolamine Solution.
Ind. Eng. Chem. Res. 1994, 33, 2006. (24) Jou, F.-Y.; Mather, A. E.; Otto, F. D. The Solubility of CO2
(9) Chen, C.-C.; Britt, H. I.; Boston, J. F.; Evans, L. B. Local in a 30 Mass Percent Monoethanolamine Solution. Can. J. Chem.
Composition Model for Excess Gibbs Energy of Electrolyte Sys- Eng. 1995, 73, 140.
tems. AIChE J. 1982, 25, 588. (25) Lee, J. I.; Otto, F. D.; Mather, A. E. The Solubility of H2S
(10) Chen, C.-C.; Evans, L. B. A Local Composition Model for and CO2 in Aqueous Monoethanolamine Solutions. Can. J. Chem.
the Excess Gibbs Energy of Aqueous Electrolyte Systems. AIChE Eng. 1974, 52, 803.
J. 1986, 32, 444. (26) Britt, H. I.; Luecke, R. H. The Estimation of Parameters
(11) Soave, G. Equilibrium Constants from a Modified Redlish- in Nonlinear, Implicit Models. Technometrics 1973, 15, 233.
Kwong Equation of State. Chem. Eng. Sci. 1972, 27, 1197. (27) Nath, A.; Bender, E. Isothermal Vapor-Liquid Equilibria
(12) Renon, H.; Prausnitz, J. M. Local Compositions in Ther- of Binary and Ternary Mixtures Containing Alcohol, Alkanol-
modynamic Excess Functions for Liquid Mixtures. AIChE J. 1968, amine, and Water with a New Static Device. J. Chem. Eng. Data
14, 135. 1983, 28, 370.
(13) Kohl, A. L.; Nielsen, R. B. Gas Purification, 5th ed.; Gulf (28) Touhara, H.; Okazaki, S.; Okino, F.; Tanaka, H.; Ikari, K.;
Publishing Co.: Houston, TX, 1997. Nakanishi, K. Thermodynamic Properties of Aqueous Mixtures of
(14) Bates, R. G.; Pinching, G. D. Acidic Dissociation Constant Hydrophilic Compounds. 2. Aminoethanol and its Methyl Deriva-
and Related Thermodynamic Quantities for Monoenthanolammo- tives. J. Chem. Thermodyn. 1982, 14, 145.
nium Ion in Water from 0 to 50 °C. J. Res. Natl. Bur. Stand. 1951, (29) Kohl, A. L.; Riesenfeld, F. C. Gas Purification, 4th ed.; Gulf
46, 349. Publishing Co.: Houston, TX, 1985.
(15) Clarke, J. K. A. Kinetics of Absorption of Carbon Dioxide
in Monoethanolamine Solutions at Short Contact Times. Ind. Eng. Received for review September 17, 1998
Chem. Fundam. 1964, 3, 239. Revised manuscript received February 16, 1999
(16) Wang, Y. W.; Xu, S.; Otto, F. D.; Mather, A. E. Solubility Accepted February 18, 1999
of N2O in Alkanolamines and in Mixed Solvents. Chem. Eng. J.
1992, 48, 31. IE980600V