Vapour-Liquid Equilibrium of CO2 in Aqueous Solutions

of 2-Amino-2-Methyl-1-Propanol
Paper # 703
Madhusree Kundu; Bishnu P. Mandal and Syamalendu S. Bandyopadhyay
Separation Science Laboratory
Cryogenic Engineering Centre
Indian Institute of Technology
Kharagpur 721302, India

ABSTRACT

The solubility of CO2 in aqueous solutions, 18 wt%, 25 wt% and 30 wt% (2.0, 2.8, 3.4 M) of
2-amino-2-methyl-1-propanol (AMP) has been measured over a temperature range of 303-
323 K and partial pressure range of 1-100 kPa. In this work, the modified Clegg-Pitzer
equation is used to correlate and predict the vapour-liquid equilibria of CO2-AMP-H2O
system. The model predicted results have been found to be in good agreement with the
experimental results of VLE measurement of this work. The nontraditional optimization
algorithm, simulated annealing, has been used for the parameter estimation.

INTRODUCTION

Removal of acid gas impurities such as CO2 and H2S from gas mixtures is very important in
natural gas processing, hydrogen purification and treating refinery off gases and synthesis gas
for ammonia making. Regenerative chemical absorption of the acid gases, H2S and CO2, into
solutions of alkanolamines e.g., monoethanolamine (MEA), diethanolamine (DEA), di-2-
propanolamine (DIPA), N-methyldiethanolamine (MDEA), is widely used for gas treating
processes. Recently a new class of amines, sterically hindered amines, has been introduced as
commercially attractive new absorbent over conventional amines.1 Sharma2 observed that
steric effect influences the stability of the carbamates formed due to the amine-CO2 reaction,
and proposed the use of highly branched amines such as 2-amino-2-methyl-1-propanol
(AMP) for higher cyclic absorption capacity for CO2. AMP is considered today as one of the
most important hindered amines for natural gas treatment processes. As in MDEA, the CO2
loading in AMP approaches a value of 1.0 mole of CO2 per mole of amine, while the reaction
rate constant for CO2-AMP is much higher than that for CO2-MDEA.3 Since sterically
hindered amine, AMP, does not form stable carbamate,3 bicarbonate and carbonate ions may
be present in the solution in larger amounts than the carbamate ions. Hence, the regeneration
energy costs when aqueous solutions of AMP are used to absorb CO2 may be lower as in the
case of using aqueous MDEA solutions.

Design of gas treating absorption/stripping systems by the traditional equilibrium stage

approach require the knowledge of the vapour-liquid equilibrium (VLE) behaviour of the
aqueous acid gas-alkanolamine systems. Besides, equilibrium solubility of the acid gases in
aqueous alkanolamine solutions determines the minimum recirculation rate of the solution to
treat a specific sour gas stream and it determines the maximum concentration of acid gases
which can be left in the regenerated solution in order to meet the product gas specification. In
rate based models, physical and chemical equilibria plays an important role by defining the
boundary conditions for the partial differential equations, describing mass transfer coupled
with chemical reactions. Representation of the experimental data with thermodynamically

1
rigorous model is required, so that one can systematically correlate and predict the vapour-
liquid equilibria (VLE) of these systems. A few results of CO2 solubility in aqueous AMP
solutions has been reported earlier (Roberts and Mather,4 Teng and Mather,5 Silkenbumer et
al.6). Among the recent literature on modeling VLE of CO2 and H2S in single and mixed
amine solvents, the work of Kent and Eisenberg,7 Li and Shen,8 Deshmukh and Mather,9
Austgen and Rochelle,10 Li and Mather11 are the major ones.

In this work new experimental results of CO2 solubility in aqueous AMP solutions of three
different amine compositions (2.0, 2.8, 3.4 M), are reported over a wide range of temperature
(303-323 K) and CO2 partial pressure (1-100 kPa). The Clegg-Pitzer equations are used to
correlate and predict the vapour-liquid equilibria of AMP-CO2-H2O system. The
nontraditional optimization algorithm like simulated annealing12 used for parameter
estimation, has been introduced here for the first time in the numerical part of the model
solution for predicting VLE of acid-gas-alkanolamine-water system.

EXPERIMENTAL

For VLE measurements a stirred glass equilibrium cell was used. The experimental setup is
shown in Figure 1.

Figure 1 Stirred equilibrium cell, 1. CO2 cylinder, 2. N2 cylinder, 3. mass flow controller, 4.
saturator, 5. Pt sensor, 6. G.C, 7. eudiometric device, 8. DC drive, 9. mercury seal, 10. gas
phase stirrer, 11. magnetic stirrer, 12. controller, 13. water circulator, 14. recorder

A magnetic stirrer was used for the liquid phase and two four bladed impeller mounted on a
shaft, passing through a mercury seal and driven by a D.C. motor, was used for the gas phase.
The temperature of the equilibrium cell was controlled within 0.2 K of the desired level
with a circulator temperature controller operated on external control mode (JULABO, FP 55,
FRG). Precalibrated platinum sensors (Pt-100, JULABO, FRG) were used for measurement

2
of temperatures in the cell as well as in the thermostated bath. All VLE measurements were
done at atmospheric pressure. A precise manomatric device was employed to maintain
atmospheric pressure in the cell throughout the period of equilibrium experiment. Precision
mass flow controllers (Sierra Instruments, USA) were used to control the flow rates of CO2
and N2.

2-amino-2-methyl-1-propanol (AMP) was supplied by E. Merck Germany, and had a purity

of 98%. Amine concentration was determined by titration with standard HCl using methyl
orange indicator. Double distilled water degassed by boiling was used to prepare aqueous
amine solutions. CO2 cylinder with certified purity of 99.995% of Hydrogas India was used.
N2 cylinder was procured from BOC, India. The purity of N2 used was better than 99%.

For each run the equilibrium cell was allowed to reach thermal equilibrium with respect to
the desired temperature for VLE measurement. The cell was then purged with the desired
mixture of CO2 and N2. The N2 and CO2 gas streams at the outlet of the respective mass flow
controllers were mixed and passed through water vapour saturators, maintained at the
measurement temperature, before being introduced into the cell. Complete purging of the cell
with dilute CO2 of desired partial pressure was ensured by gas analysis at the inlet and outlet
of the cell using a gas chromatograph (SHIMADZU GC-16A, JAPAN). The gas phase stirrer
was kept on at 70 rpm during purging to ensure uniform gas phase concentration throughout
the cell. After completion of purging 10 ml of freshly prepared amine solution of desired
concentration was quickly transferred in the cell and the cell was fully sealed. The liquid
phase and gas phase stirrers were turned on to commence absorption. As absorption
proceeded, CO2 make up from the pure CO2 cylinder was carefully introduced into the cell
through a fine control valve, precisely maintaining atmospheric pressure in the cell
throughout the period of VLE measurements. The attainment of equilibrium was ensured
when there was no absorption for at least one hour while the temperature was maintained
constant at the desired level. It took about 6-8 hours to reach equilibrium for each run. After
equilibrium was reached, gas phase samples were analyzed by gas chromatography (GC-16A,
SHIMADZU, JAPAN), using porapaq-Q and porapaq-R column connected in series, TCD
detector, and hydrogen carrier gas. The equilibrium CO2 loading in the liquid phase was
determined by acidulating known volumes of the loaded liquid samples with 6 (M) HCl and
measuring the volume of the evolved gas. The experimental results of VLE measurement are
presented in Tables 1-3.

Table 1. Solubility of CO2 in 2 M AMP solutions at 303 K

Partial pressure Loading
kPa ( CO 2 )
4.406 0.674
7.936 0.7917
10.25 0.82
27.145 0.907
39.4 0.929
52.22 0.9462
90.13 0.9656

3
Table 2. Solubility of CO2 in 2.8 M AMP solutions.
303 K 313 K 323 K
Partial Loading Partial Loading Partial Loading
pressure ( CO 2 ) pressure ( CO 2 ) pressure ( CO 2 )
kPa kPa kPa
4.05 0.689 3.25 0.527 4.621 0.43
8.7091 0.74 4.49 0.5655 15.184 0.605
10.92 0.784 8.393 0.656 29.21 0.713
16.2 0.82 16.78 0.743 55.808 0.8023
18.409 0.8422 26.19 0.8035 71.42 0.825
24.06 0.85 57.75 0.8973 91.08 0.841
48.10 0.901 71.42 0.911
58.0 0.911 91.5 0.924
64.77 0.92
76.22 0.932
87.47 0.938

Table 3. Solubility of CO2 in 3.4 M AMP solutions.

303 K 313 K 323 K

Partial Loading Partial Loading Partial Loading

pressure ( CO 2 ) pressure ( CO 2 ) pressure ( CO 2 )
kPa kPa kPa
3.86 0.599 3.2 0.475 5.247 0.412
6.78 0.665 4.8 0.525 9.89 0.532
8.373 0.693 12.8 0.673 22.32 0.623
16.167 0.757 24.6 0.756 52.2 0.746
30.515 0.8125 47.6 0.801 88.2 0.7813
53.03 0.848 64.77 0.851
93.96 0.889 93.0 0.876

MODELLING

Chemical equilibria

In the aqueous phase for the AMP-CO2-H2O system, the following chemical equilibria are
involved.

K
CO2 + H2O 11
H+ + HCO3- (1)

4
 K
AMP + H2O 22 OH- + AMPH+ (2)

K
H2O 33 H+ + OH- (3)

Henrys law relates the acid gas partial pressure to the physically dissolved acid gas
concentration in the solvent according to the following relation

pCO2 = H CO2 [CO2] (4)

The expression for thermodynamic equilibrium constants is as follows.

K 11 = (m H + m HCO / mCO2 m H 2O )( Hm + HCO

m
/ CO H O )
m
2
m
2
(5)
3 3

Where mi are the molality of the species, and im s are the activity coefficient based on
molality scale.

K 22 = (C OH C AMPH + / C AMP C H 2O )( OH
c

c
AMPH +
/ AMP
c
Hc 2O ) (6)

K 33 = (C H + C OH / C H 2O )( Hc + OH
c
/ H O)
c
2
(7)

Where Ci are the molarity of the species and ic s are the activity coefficient based on
molarity scale.

Thermodynamic equilibrium constant K11 is based on molality scale and K22, K33 are given on
molarity scale. All are considered to be function of temperature. The Henrys constant (H)
has the unit of pascals, and considered to be a function of temperature. The functional form of
temperature dependence and temperature coefficients for both equilibrium constants and
Henrys constant are taken from different literature sources as shown in Table 4.

Table 4. Temperature dependence of the equilibrium constants for reaction 1-3 and Henrys
constant for CO2
ln K11 = A / T + B ln T + C T + D
ln K22 = A / T + B ln T + C T + D
ln K33 = A + B / T + C / T2 + D / T3 + E / T4
ln H = A + B / T + C ln T + D T
Reaction A B C D E Source
a
1 -7742.6 -14.506 -2.8104e-2 102.28
a
2 -7261.78 -22.4773 0 142.58612
b
3 39.5554 -9.879e4 0.568827e8 -0.14645e11 0.136145e13
c
Henrys 170.7126 -8477.711 -21.95743 0.005781
constant
a
Silkenbumer et al;6 bLi and Shen8 cLi and Mather11

5
From Eqs. 2 and 3 we can write

AMP + H
K '
AMPH+ 2 +
(8)

K2' = K33 / K22 (9)

where K2' is the thermodynamic equilibrium constant for Eq. 8 and based on molarity scale.
To use those thermodynamic equilibrium constants in the present model, the equilibrium
constants, K11 and K2' are to be converted to mol fraction based equilibrium constants in the
following way.

K1 = K11 (Ms / 1000) (10)

ln K1 = ln K11 - ln (1000 / Ms) (11)

K2 = K2' (Ms / 1000) (12)

ln K2 = ln K2' - ln (1000 / Ms) (13)

where K1 and K2 are the mol fraction based equilibrium constant for Eqs. 1 and 8
respectively. Here Ms is the molecular weight of the pure solvent, is the density of the pure
solvent. In general, to convert the logarithm of an equilibrium constant for the dissociation of
an electrolyte in water from molality scale to the mol fraction scale, it is necessary to subtract
ln (1000/ Ms) for each non-water component on the right hand side of a stoichiometric
expression and to add ln (1000/ Ms) for each non-water component on the left hand side of a
stoichiometric expression, as shown in Eq.11. In case of conversion from molarity scale to
mol fraction scale the subtraction and addition factor is ln (1000 / Ms), as shown in Eq.13.
There are only four main species, two neutral solvents, AMP and H2O, and two ionic species,
AMPH+ and HCO3-, in the equilibrated liquid phase. In order to reduce complexity, the
existence of the molecular species of CO2 and the ionic species of CO32- in the aqueous phase
is neglected, because their concentrations are very low compared to the other species in the
equilibrated liquid phase. It is thus assumed that all the dissolved CO2 is converted into
HCO3- ions. The vapour phase has been assumed to be ideal, as the present results limited to
low to moderate pressures. From Eqs.1, 4 and 8 we get the expression for the equilibrium
partial pressure of CO2 over aqueous AMP solution.
PCO2 = ( H CO2 K 2 x AMPH + AMPH + x HCO HCO ) /( K 1 x H 2O x AMP H 2O AMP ) (14)
3 3

xns are the liquid phase mole fractions, based on true molecular or ionic species. Activity
coefficients I s (based on mol fraction scale) for different species present in the liquid phase
are calculated from modified Clegg-Pitzer equation according to the method proposed by Li
and Mather.11

Thermodynamic expression

In the modified Pitzer equation, all the species in a system are considered as interacting
particles. The excess Gibbs energy arises from inequalities in interparticle forces. The excess

6
Gibbs energy is assumed to consist of short-range forces and long range Debye-Huckel
forces.

gE = gS + gDH (15)

For a short-range term, only two and three suffix Margules expansions are used.

gS/ RT = xI xn F F W
c a
c a nca + x x
n n'
n n' (An'n xn+Ann' xn') (16)

where, c, a, n, n' represent cation, anion and neutral species and gS is the short-range
contribution to the excess Gibbs energy. The total mol fraction of ions (xI) is given by

xI = Total mole fraction of ions = 1 - x

n
n (17)

The cation and anion fraction Fc and Fa are defined for fully symmetrical systems by

Fc = 2xc / xI (18)

Fa = 2 xa / xI (19)

An'n, Ann' are solvent-solvent interaction parameters. Wnca are the ion-solvent interaction
parameters. For long-range contribution to the excess Gibbs energy, the following expression
is used.

gDH/RT = (-4Ax Ix/ ln(1+ Ix1/2 ) + x x

c a
c a Bca g(1Ix1/2) (20)

where, Bca is ion-ion interaction parameter and (Ix) is mol fraction ionic strength.

Ix = 0.5 zi2 xi (21)

The function g(x) is expressed by
g(x) = 2[1-(1+x)exp(-x)]/x2 (22)

where

x = (1Ix1/2) = 2 I1/2 (23)

I is the ionic strength in molar concentration.

Ax = A ( Cn )1/2 (24)
n

where, x is the Debye-Huckel parameter on mol fraction basis, Cn is the molar concentration
of the solvent n, A is original Debye-Huckel parameter and a function of temperature,
density and relative permitivity of the mixed solvent. The parameter is related to the hard-
core collision diameter, and given by the relation,

7
 = 2150{(18.02 m Cn ) / (1000 mT)}1/2 (25)
n

where m is the density of the mixed solvent.

The relative permitivity (m) of the mixed solvents

(m) = n n (26)
n

where n is the volume fraction of solvent n.

n = (CnVn / CnVn ) (27)

n
The temperature dependence of density of the pure solvents is listed in Table 5.

Table 5. Temperature dependence of density for pure solvents

Solvent MW Expression Temp range o C
t
H2O 18.02 b
= 0.999382+0.00007208t-7.2849110-6t2 20-80
+ 2.6517710-8t3
AMP 89.14 = 1.1851-8.5314310-4t
a
20-80
a b 11
derived empirical relation from experimental data, Li and Mather

As AMP is methyl substituted MEA, the relative permitivity of AMP is assumed here as that
of MEA. The temperature dependence of relative permitivity for pure solvents is presented in
Table 6.

Table 6. Temperature dependence of relative permitivity for pure solvents

Solvent Expression Temp range o C
t
a
H2O =78.54[1-4.57910-3(t-25) + 1.1910-5(t-25) 0-100
2
2.8 10-8(t-25) 3
a
MEA =35.76 + 14836(1/(t+273)-1/273.15) -70 to 60
a
Li and Mather11

Differentiation of Eqs. 16 and 20 yields the expressions for the short-range and long range
force contribution to the activity coefficients of ions and neutral species.

Data Regression: Estimation of Interaction Parameter

In this work the experimental solubility data comprising acid gas loadings in the range 0.03 to
1.0 mol/mol and the partial pressures of acid gas in the range 0.300-1000 kPa, have been used
to regress the interaction parameters. The objective function for regression is as follows.

8
 = {( P )
CO2 cal ( PCO2 )exp
}/(P
CO2 )exp (28)

The VLE data used for regression along with their average correlation deviation have been
summarized in Table 7.

Table 7. Experimental solubility data used to regress the interaction parameters and their
correlation deviation for AMP - CO2 - H2O ternary system
Reference [AMP] Temp Data points Average
(K) Correlation
Deviation %
Teng and 2.0 M 313 15 16.0
Mather 5
Li and Chang 16 3.4 M 333,353,373 36 9.34

This Study 2.8 M 303 14 13.0

Among the interaction parameters, one is due to ion-ion interaction, two ion-solvent
interaction and two solvent-solvent interaction parameters. As no reliable experimental
information for the binary system AMP-H2O is available, the solvent-solvent interaction
parameters were also regressed from the ternary system (AMP-CO2-H2O) solubility data. The
parameters thus obtained, with their temperature coefficients, are listed in Table 8.

Table 8. Fitted values of interaction parameters for AMP-CO2-H2O system

B (or W or A) = a + b T + c T 2
B or W a b c 303 K 313 K 353 K 373 K
or A
W1MX 163.65422 -.98456 .0015 1.0 2.399017 3.0 5.0

W2MX -20.28054 0.11078 0 14.045685 14.358887 17.49957 22.25622

BMX - 0.36105 0 -31.12158 -29.79285 -14.80304 -6.107605

140.46609
A12 -43.0574 0.23011 0 27.923676 29.094055 35.697327 45.017731

A21 23.50638 -0.115 0 -10.0 -14.15012 -18.30112 -18.60907

Subscripts: 1= AMP, 2 = H2O, M= AMPH+, X = HCO3-

Method of solution

The problem of parameter estimation in the model predicting VLE of an aqueous acid gas
alkanolamine system involves regression of experimental data wherein optimum values of the
parameters are found. In this paper, the regression function has been considered to be an
objective function that is to be minimized. Simulated annealing (SA) is a computational
stochastic technique for finding near global minimum solutions to optimization problems.
The method mimics the thermodynamic process of cooling of molten metals for achieving the
minimum free energy state. The main driving force behind SA is to occasionally allow for

9
wrong-way movement (uphill moves for minimization) simultaneously providing
convergence to the global optimum. The SA method works with a point in the variable search
space of the vector according to the Boltzmann probability distribution viz.

E
probability, P = exp (29)
k BT

where, kB is the Boltzmann Constant and T is the current annealing temperature.

The energy function becomes the objective function. The particle configurations are the
variable values. Finding a low-energy configuration becomes seeking a near optimal solution
for the function where temperature becomes a control parameter for the solution. This leads
to the generation of a Markov Chain of points that are given by distribution Eq. 29 and are
independent of the initial point. The acceptance criterion given by Eq. 29 is referred to as the
Metropolis criterion for SA. This expression suggests that a system at a high temperature has
almost uniform probability of being at high energy state, but at low temperature it has a small
probability of being at a high energy state. Therefore, by controlling the temperature T and
assuming the search process follows the Boltzmann probability distribution, the convergence
of an algorithm can be controlled. In the course of simulations, the temperature is
systematically annealed using a schedule like Kirkpatricks exponential schedule13,

Ti +1 = * Ti (30)

where is the cooling factor and 0.9 < < 1.0. The move schedule for consideration of a
given number of configurations at each temperature is done using a random schedule. The
choice of such schedules varies along with the problem being considered. Simulated
annealing is a point-by-point method. The algorithm begins with an initial point and a high
temperature T. A second point is created at random in the vicinity of the initial point and the
difference in the function values ( E ) at these two points is calculated. If the second point
has a smaller function value, the point is accepted; otherwise the point is accepted with a
probability (- E /T). This completes one iteration of the Simulated annealing procedure. In
order to simulate the thermal equilibrium at every temperature, a number of points (n) is
usually tested at a particular temperature, before reducing the temperature. The algorithm is
terminated when a sufficiently small temperature is obtained or a small enough change in
function values is found. The initial temperature T and the number of iterations performed at
a particular temperature are two important parameters, which govern the successful working
of the Simulated annealing procedure.

10
 Start

Choose an initial configuration (old) and

find the energy, old.E

Counter, i =0.Set T to a high value. Fix number of iterations at

each temperature = iter. Fix lowest temperature, Terr

Is No State the best obtained

T>Terr? solution(best) as optima

Yes
Stop
Tnew = Told Yes
Is
where,0.9 < < 1.0 i > iter?

No
Generate new configuration as per move
schedule (new) and calculate new.E
Calculate
del = new.E-old.E
Calculate
prob = exp(-del/T)
Generate 0<rnd<1, No Yes
Is
where new.E<old.E Assign, old=new
rnd is a random number

No No Is
Is Update counter, i=i+1
rnd<prob? new.E<best.E?

Yes
Yes
Assign best=new
Assign, old=new
and best remains as
before

Figure 2 Flowchart for Simulated annealing (SA)

11
RESULTS AND DISCUSSION

The regressed interaction parameters and the correlation presented in Table 8 for AMP-CO2-
H2O ternary system on the basis of experimental solubility data set, presented in Table 7,
have been used here to predict some solubility data for a range of temperature and solvent
composition which were not used for the regression. These predictions are summarized in
Table 9 and Figures 3-7 and represented in the figures as model predicted curves. However,
some Figures (3, 4 and 7) also include correlated solubility curves to show the correlation
deviation with the experimental solubility data, which were used for the regression. In
respective Figures (3, 4 and 7), these results are shown by correlation curves.

Table 9. Average deviation for prediction of solubility in AMP-CO2-H2O ternary system

Reference [AMP] Temp Data Error
(K) Points AAD%
Teng and Mather5 2.0 M 343 9 22.25
Jane and Li15 2.0 M 313 7 20.0
4
Roberts 2.0,3.0 M 313 15 17.58
17
Seo and Hong 3.4 M 313,333,353 19 15.72
6
Silkenbumer et al. 5.96,5.64 M 313,333 16 18.99
This Study 2.0, 2.8, 3.4 M 303,313,323 56 14.3

For modelling of vapour liquid equilibrium of CO2-alkanolamine-water systems, average

deviations in the range 15-30% have been reported by previous workers (Silkenbumer et al,6
Li and Shen,8 Deshmukh and Mather,9 Austgen and Rochelle,10 Li and Mather,11 Li and
Mather14) who used deterministic techniques for model solution. In this work, simulated
annealing technique was employed in an effort to predict the VLE with better accuracy.

Correlation and prediction results are shown in Figures 3-7. For 2 M AMP solutions, the
experimental data of CO2 solubility from Teng and Mather,5 and of this study have been
compared with the predicted (those solubility data were not included in regression of
interaction parameters14) results for the temperatures 303 and 343 K, as shown in Figure 3.
Figure 3 also presents a correlated solubility curve at 313 K, showing deviation with the
experimental solubility data of Teng and Mather,5 (for 2 M AMP solutions, at 313 K which is
inclusive of the data set used for regression) Roberts4 and Jane and Li.15

As shown in Figure 4, there is a reasonable agreement between correlated as well as model

predicted results and the experimental data for the solubility of CO2 in 30 wt% (3.4 M) AMP
solution over a temperature range of 313-373 K.

12
 1000
Correlated
Model predicted
5
Teng and Mather
15
Jane and Li
100 4
Roberts
pCO,kPa Teng and Mather
5 343 K
This study
313 K
2

10

303 K
1

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
CO ,kmol CO2/kmol amine
2

Figure 3 Equilibrium partial pressure of CO2 over 2 M AMP solutions

1000
Correlated
Model predicted
16
Li and Chang
17
Seo and Hong
This study 353 K
100 333 K

373 K
pCO,kPa

313 K
2

323 K
10

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

CO ,kmol CO2/kmol amine

2

Figure 4 Equilibrium partial pressure of CO2 over 3.4 M AMP solutions

The regressed interaction parameters have been used to predict the solubility of CO2 in 3 M
AMP at 313 K, in 5.96 M AMP at 333 K and in 5.64 M AMP at 313 K. As shown in Figures
5 and 6 the model predicted equilibria are found to be in good agreement with the
experimental results of Roberts4 and Silkenbumer et al.6 respectively.

13
 1000
Model Predicted
4
Roberts

100

pCO,kPa
313 K
2

10

1
0.4 0.5 0.6 0.7 0.8 0.9 1.0
CO ,kmol CO2/kmol amine
2

Figure 5 Equilibrium partial pressure of CO2 over 3.0 M AMP solutions

10000
Model predicted
Experimental results
6
Silkenbumer et al.
1000 313 K
333 K
pCO ,kPa
2

100

10

0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

CO ,kmol CO2/kmol amine
2

Figure 6 Equilibrium partial pressure of CO2 over 5.96 and 5.64 M AMP solutions

The experimental solubility data of CO2 in 2.8 M AMP at the temperatures 313 and 323 K
have been compared with the predicted CO2 solubility in Figure 7. Figure 7 also presents the
correlated solubility curve at 303 K along with the experimental data, which were used for
correlation (2.8 M AMP solutions, 303 K). The agreement, in all the cases is excellent.

14
 1000
Correlated
Model predicted
Experimental results
303 K
313 K
pCO,kPa 100 323 K
2

10

0.4 0.5 0.6 0.7 0.8 0.9 1.0

CO ,kmol CO2/kmol amine
2

Figure 7 Equilibrium partial pressure of CO2 over 2.8 M AMP solutions

CONCLUSIONS

The equilibrium solubility of CO2 in aqueous solutions of AMP were measured in the temperature
range 303-323 K for a CO2 partial pressure range of 1-100 kPa. The equilibrium solubility of
CO2 in aqueous AMP solutions has been correlated using a simplified Clegg-Pitzer equation. The
model predicted partial pressures of CO2 over aqueous AMP solutions have been found to be
in good agreement with the experimental results of this study and results available in the open
literature. This provides additional support for the validity of the experimental technique.
Simulated annealing (SA) has been successfully implemented for parameter estimation of the
VLE model. Better accuracy of the predicted results establishes the utility of the non-
traditional optimization algorithms, eg, simulated annealing in parameter estimation.

NOMENCLATURE

H CO2 Henry's law constant, Pa

K1, K2 thermodynamic chemical equilibrium constant, expressed in activity and in
mol fraction
K2',K22, K33 thermodynamic chemical equilibrium constant, expressed in activity and in
molarity
K11 thermodynamic chemical equilibrium constant, expressed in activity and in
molality
Zi valency of an ion
gE excess Gibbs free energy.
gS short range force contribution to the excess Gibbs free energy
gDH long range force contribution to the excess Gibbs free energy
pCO2 partial pressure of CO2, Pa or kPa
T absolute temperature, K
t temperature, oC

15
C molar concentration
density of pure solvent
Ms molecular weight of pure solvent
F ionic fraction
I ionic strength on molality basis
Ix ionic strength on mol fraction basis
R gas constant
relative permitivity
Ann' interaction parameter between neutral molecules
W interaction parameter between neutral and ionic species
Ax Debye-Huckel parameter on mol fraction basis
A Debye-Huckel parameter for osmotic coefficient

Greek letters

(1) Pitzer universal constant in eq. (20)

CO 2
liquid phase loading of CO2, kmol CO2 /kmol amine
activity coefficient
volume fraction
hard core collision diameter
objective function for regression

Superscripts

S short range term

DH long range term

Subscripts

a, X anion
c, M cation
n,n' neutral solvent species
cal calculated value
ex experimental value

ACKNOWLEDGEMENT

The financial support by the Centre for High Technology (CHT), New Delhi, India, is
gratefully acknowledged.

REFERENCES

1. Sartori, G., Savage, D. W. Ind. Engg Chem. Fundamentals, 1983, 22, 239-249.

2. Sharma, M. M. Ph. D thesis, University of Cambridge, 1964; Trans. Faraday. Soc., 1965,
61, 681-688.

16
3. Saha, A. K., Bandyopadhyay, S. S., Biswas, A. K. Chem. Eng. Sci., 1995, 50, 3587-
3598.

4. Roberts, B. E. M. Sc. Thesis, University of Alberta, Edmonton, Alberta, 1983.

5. Teng, T. T., Mather, A.E. Can. J. Chem. Eng., 1989, 67, 846-850.

6. Silkenbumer, D., Rumpf, B. and Lichtenthaler, R. N. Ind. Eng. Chem. Res., 1998, 37,
3133-3141.

7. Kent, R.L., and Eisenberg, B. Hydrocarbon processes., 1976, 55(2), 87-90.

8. Li, M. H., and Shen, K. P. Fluid Phase Equilibria., 1993, 85, 129-140.

9. Deshmukh, R. D., and Mather, A.E., Chem. Eng. Sci., 1981, 36, 355-362.

10. 10. Austgen, D. M., and Rochelle, G. T. Ind. Eng. Chem. Res., 1991, 30, 543-555.

11. Li, Y. G., and Mather, A. E. Ind. Eng. Chem. Res., 1994, 33, 2002-2005.

12. Aarts, E., and Korst, J. Simulated Annealing and Boltzmann mechanics: A Stochastic
Approach to Combinatorial Optimization and Neural Computing. Chichester: Wiley,
1989
13. Kirpatric, S., Gelatt, C.D. and Vecchi, M. P. Science, 1983, 220, 671-680.

14. Li, Y. G., and Mather, A. E. Ind. Eng. Chem. Res., 1997, 36, 2760-2765.

15. Jane, I. S., and Li, M. H. J. Chem. Eng. Data., 1997, 42, 98-105.

16. Li, M. H. and Chang, B. C. J. Chem. Eng. Data., 1994, 39, 448-452.

17. Seo, D. J., and Hong, W. H. J. Chem. Eng. Data., 1996, 41, 258-260.

17
18