You are on page 1of 15

Chemical Engineering and Processing 43 (2004) 701715

Modelling selective H
2
S absorption and desorption
in an aqueous MDEA-solution using a rate-based
non-equilibrium approach
Markus Bolhr-Nordenkampf
a,
, Anton Friedl
a,1
, Ulrich Koss
b,

, Thomas Tork
b,2
a
Institute of Chemical Engineering, Vienna University of Technology, Getreidemarkt 9/166, A-1060 Vienna, Austria
b
Lurgi Oel Gas Chemie, Lurgiallee 5, D-60295 Frankfurt am Main, Germany
Received 27 December 2002; received in revised form 12 February 2003; accepted 12 February 2003
Abstract
A rate-based algorithm was used to yield a predictive tool for MDEA gas scrubbing processes. The model adopts the two-lm theory,
assuming that thermodynamic equilibrium exists only at the gasliquid interphase, but not in the boundary layers, where temperature and
concentration gradients are present. Correspondingly chemical equilibrium among the reacting species in the liquid phase is assumed for the
bulk phase, but not for the liquid boundary layer. Mass transfer is modelled using calculated mass transfer coefcients in combination with an
enhancement model to account for the chemical reactions. Correlations for geometric data, like hold-up and interfacial area, and for reaction
rates are provided to give reliable results. The latter correlations are also used to describe the desorption process, which is calculated with an
equilibrium approach, considering the kinetics of CO
2
desorption. The so obtained tool is tested against measurements done recently by Lurgi
GmbH at a commercially operated selective MDEA plant in Germany. A closed absorption and desorption loop was build up using Aspen
RATEFRAC
TM
, capable of modelling the whole process with all necessary equipment.
2003 Elsevier B.V. All rights reserved.
Keywords: Non-equilibrium stage model; Mass-transfer; Alkanolamines; Carbon dioxide; Hydrogen sulphide; Absorption of acid gases
1. Introduction
Removal of acid gas components from gas streams
containing CO
2
and H
2
S by aqueous alkanolamines has
become a well-established process. With the increase in
environmental awareness, the exploitation of poorer quality
oil and natural gases, precise modelling of the gas ab-
sorption process has become important for industrial plant
design.
For example, H
2
S removal from natural gas must be max-
imised to meet with pipeline specications while CO
2
ab-

Corresponding authors. Tel.: +43-1-58801-15933;


fax: +43-1-58801-15999 (M.B.-N.); tel.: +49-69-5808-3740;
fax: +49-69-5808-2645 (U.K.).
E-mail addresses: bolhar@mail.zserv.tuwien.ac.at
(M. Bolh` ar-Nordenkampf), afriedl@fbch.tuwien.ac.at (A. Friedl),
ulrich koss@lurgi.de (U. Koss), thomas tork@lurgi.de (T. Tork).
1
Tel.: +43-1-58801-15920; fax: +43-1-58801-15999.
2
Tel.: +49-69-5808-2825; fax: +49-69-5808-2645.
sorption is often best kept minimal. Or maximum CO
2
re-
moval for use in enhanced oil recovery is desired. To meet
with the pollution standards, tail gas specications are con-
stantly undergoing restrictions requiring stringent scrubbing
processes.
As it is shown in Fig. 1 a typical industrial plant con-
sists of an absorption and a desorption column, a solution
interchanger for heat recovery, a solution cooler, a solu-
tion pump, and a reboiler as well as a reux-system for
the desorber. The absorber operates from ambient pressure
up to 70 bar and from 25 to 70

C. The energy consum-


ing desorption of the acid gases is carried out at around
130

C and at pressures from ambient up to 3 bar. Desorp-


tion pressure must not necessarily be lower than the absorp-
tion ones (e.g. tail gas treatment), this depends further on the
requirements of the connected Claus-Plant. MDEA-plants
can process up to 400 000 Nm
3
/h feed gas in one single
train.
For accurate plant design it is of great importance to be
able to predict the mass transfer behaviour in the absorption
0255-2701/$ see front matter 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0255-2701(03)00011-4
702 M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715
Fig. 1. Simplied ow sheet of a MDEA-acid gas removal plant.
and desorption column. Desorption can be calculated using
an equilibrium approach, but it has to be taken into account
that the CO
2
desorption is kinetically driven. An equilibrium
approach for the absorption is not suitable, if predictive capa-
bilities of the model are necessary, as it is the case for selec-
tive H
2
S and/or CO
2
removal in alkanolamine-systems. This
can only be achieved using a rate-based non-equilibrium
model as it is done in this work. It is based on the mass
and heat transfer between the liquid and the vapour phase
occurring on a height-increment of the structured and ran-
dom packing, respectively. Mass and energy balances are
connected by rate-equations across the interface using the
two-lm theory to calculate the transfer rates. Thermody-
namic equilibrium is assumed at the gasliquid interface.
In the liquid bulk phase additional chemical equilibrium is
assumed.
The objective of this work is to adapt a rate-based algo-
rithmimplemented in Aspen (RATEFRAC
TM
) to yield a pre-
dictive tool for MDEA gas scrubbing processes. Therefore,
the mass transfer coefcient of the liquid phase is calculated
adjusting a formulation from Brunazzi [1] to experimental
results, while for the gas phase the Onda [2] formulation is
used, which is already integrated in RATEFRAC
TM
. A new
enhancement model is developed to account for the chemical
reactions in the liquid phase. New correlations for geomet-
ric data, like hold-up and interfacial area, and for reaction
rates are provided to give reliable results. The latter correla-
tions are also used to describe the desorption process, which
is calculated with an equilibrium approach, considering the
kinetics of CO
2
desorption.
The so obtained system is tested against measurements
done recently by Lurgi
3
at a commercially operated selective
MDEA plant in Germany. A closed absorption and desorp-
tion loop was build up, capable of modelling the whole pro-
cess with all necessary equipments. The developed model is
used for designing a large natural gas purication and con-
ditioning project built by Lurgi.
3
Lurgi Oel Gas Chemie GmbH, Frankfurt, Germany.
2. Model theory
2.1. Mass and energy balance
The absorption can be treated as a kinetically determined
mass transfer process in which the degree of separation is
determined by the mass and energy transfer rates between
the phases being contacted on each tray or within sections of
a packed column. This approach allows dealing with real
trays and real packing right from the outset and it results in
a physically more realistic model based on the fundamental
chemistry and physics of the process.
Calculations are made of actual tray-by-tray or section-
by-section transfer rates, being determined by mass and heat
transfer coefcients with concentration and temperature dif-
ferences as driving forces.
The rate approach has been described in detail for non-
reactive separation processes by Krishnamurthy and Tay-
lor [3], Cornelissen [4], and Weiland [5]. The presence of
chemical reactions, due to calculation of the reaction rates,
increases the complexity of the mathematical system, con-
cerning vapourliquid equilibrium as well as mass transfer
process calculations. Detailed treatment of these two aspects
can be found at Chakravarty and Weiland [6], and at Sadar
and Weiland [7].
A schematic diagram of a non-equilibrium stage is shown
in Fig. 2, packed and trayed towers consist of a number of
such stages. Vapour and liquid streams from adjacent stages
are brought into contact on the stage and are allowed to
exchange mass and energy across their common interface,
represented in the diagram by the vertical wavy line. The
ux over the interphase is calculated using the two lm the-
ory, assuming that the mass transfer resistance is located in
the boundary layer on the gas side and on the liquid side,
respectively. The stage is assumed to be at mechanical equi-
librium and steady state operation. The transferring gases
react with the amine in the liquid phase, yielding reaction
products and liberating heat.
The overall mass and energy transfer rates through the
interfacial area on stage k of the column are given by the
M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715 703
Fig. 2. Non-equilibrium stage.
following equations [3]:

N
i,k
= y
i,k+1
V
k+1
y
i,k
V
k
= a
int,k
N
V
i,k
= a
int,k
_
_
_
_
_
_
_
_
_
n1

j=1
j=i
k
V
ij,k
(y
V
j,k
y
I
j,k
)
. ,, .
Diffusion
+ y
V
i,k
N
V
t,k
. ,, .
Convection
_
_
_
_
_
_
_
_
_
(1)
= a
int,k
N
L
i,k
= a
int,k
_
_
_
_
n1

j=1
j=i
k
L
ij,k
(x
I
j,k
x
L
j,k
)+x
L
i,k
N
L
t,k
_
_
_
_
,
with i = 1, 2, . . . , n 1; k = 1, 2, . . . , m
E
k
=H
k+1
V
k+1
H
k
V
k
=a
int,k
_
_
_
_
_
_
n

i=1
N
V
i,k

H
V
i,k
. ,, .
Convection
+h
V
k
(T
V
k
T
I
k
)
. ,, .
Conduction
_
_
_
_
_
_
(2)
=a
int,k
_
n

i=1
N
L
i,k

H
L
i,k
+h
L
k
(T
I
k
T
L
k
)
_
,
with k = 1, 2, . . . , m
Henrys law is used to calculate the mole fractions x
i,k
and y
i,k
at the vapourliquid interface:
H
i,k
=
y
I
i,k
x
I
i,k
(3)
The correlations for the Henry coefcients used in this work
are given in Table 1.
In contrast to the rate-based absorption model, the desorp-
tion process is calculated using an equilibrium model (As-
pen Plus, RADFRAC
TM
). The equilibrium model is chosen
for this process due to the reason that the desorption process
is, because of the higher temperatures present and, there-
fore, faster reaction rates, more equilibrium like. Further
on, for industrial plant design it is necessary to incorporate
the leading effects to get a reliable model capable to predict
the acid gas loadings of the lean solvent. Therefore, in this
work the same reaction system in the liquid phase as for the
absorption process is considered (see Section 2.3).
2.2. Mass transfer coefcient for the gas and
for the liquid phase
The mass transfer is calculated using the above described
two lm theory [3] in combination with the generalised
MaxwellStefan approach to multicomponent mass transfer.
In comparison to Ficks Law the ux J
i
is not linear depen-
dent with respect to the molecular average mixture veloc-
ity and its composition gradient x
i
. The MaxwellStefan
approach takes into account the chemical potential as the
Table 1
Henry parameters
Component i Source
CO
2
[17,36]
H
2
S [17,36,2]
N
2
[37]
704 M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715
main driving force, therefore, this approach is also able to
describe highly non-ideal systems [8,9]. Assuming that the
driving force is completely determined by the gradient of
the chemical potential and by neglection of the Soret effect
the following simplied equation can be obtained:
(
i
)
p,T
RT
=
j=n

j=1
j=i
x
i
N
j
x
j
N
i
c
T
K
ij
(4)
Eq. (4) gives the relation between the thermodynamic
properties of the system and the ux over the gasliquid in-
terphase. It is used to obtain the diffusion coefcients nec-
essary for the following mass transfer coefcients: vapour
mass transfer coefcients are calculated using the Onda
model [2]. Although primary developed for random pack-
ings, the model for the vapour side yields good results with
structured packings too [2].
k
V
ij
=5.23
_
G
V
a
rp

V
_
0.7
Sc
V
1/3
ij
(a
rp
d
rp
)
2
(a
rp
K
V
ij
)
1
R T
(5)
where G is the gas supercial mass velocity (gas density
times gas velocity), a
rp
the specic surface area of the pack-
ing, the dynamic viscosity, Sc the Schmidt number, d
rp
the
nominal diameter of the packing, K
ij
the MaxwellStefan
diffusion coefcient, R the gas constant, and T the temper-
ature.
For calculation of the liquid mass transfer coefcients for
structured packings some relations can be found in the lit-
erature [1012], with the decit of neglecting the depen-
dence on the gas velocity and the packing height. Nawrocki
and Chuang [13] showed in their work, that the ow dis-
tance on an inclined plate is considerably important in the
mass-transfer process. Brunazzi [1] developed a Sherwood
correlation for the mass transfer on an inclined plate, Ponter
and Yeung [14] introduced a mixing factor to account for the
mixing in the junctions between the planes of the packing,
resulting in the following equation:
Sh = A
Gz
B
Ka
C
(6)
Sh, Gz and Ka are the dimensionless Sherwood, Grtz, and
Kapitsa number (see Appendix A), while A, B, and C are
adjustable parameters. Using Eq. (6) the mass transfer co-
efcient in the liquid phase can be expressed as
k
L
0
ij
=
K
L
ij
d
A

_
_
u
max
d
L

L
_

_

L

L
K
L
ij
_


H
_
B
_

3

L

L
4
g
_
C
(7)
where d is a characteristic length of a thin liquid lm, ob-
viously related to , the thickness of the liquid lm on the
inclined plate of the structured packing. For d the character-
istic length of four times is used [1], a typical ow length
after which the inuence of local perturbations is thought to
be faded. His a characteristic dimension of the column pack-
ing. For the structured packing H is chosen as the distance
from one junction point to the next junction point of the
metal structure and, therefore, it can be easily obtained. u
max
is the maximum velocity of the liquid lm (see Section 2.5).
A, B and C are constants, which have to be determined by
general mass transfer experiments on structured packings.
The original equation of Brunazzi [1] was adopted, using
the following values for the constants: A: 3 and B: 0.8. For
C a value of 0.09 was chosen according to literature [15].
2.3. Chemical reactions
Reactions which take place in the liquid phase can be di-
vided in principle into two groups. Reactions equilibrium
controlled and reactions kinetically determined. The chemi-
cal reactions determine the composition of the different ion
species in the liquid phase and, therefore, the enhancement
of the mass transfer.
Equilibrium reactions are fast enough to assume chem-
ical equilibrium throughout the entire liquid phase. This
assumption is fullled if reaction kinetics is signicantly
faster than mass transport in the phase. These reactions
can be modelled using equilibrium constants. A certain
number of equilibrium reactions occur within the system
CO
2
H
2
SAlkanolamines [16]. An overview over these re-
actions and correlations are given in Table 2 (Reaction IVI).
Kinetic reactions must be modelled differently. The as-
sumption that reaction kinetics is much faster than mass
transfer can not be applied, therefore, reaction kinetics has
to be included in the calculations.
The rst reaction to be considered is the hydration of CO
2
:
CO
2
+H
2
O H
+
+HCO

3
(VII)
This reaction is very slow [17] and may be neglected. The
second reaction is the bicarbonate formation:
CO
2
+OH

HCO

3
(VIII)
This reaction is fast and can enhance mass transfer even
when the concentration of the hydroxyl ions is low and may
Table 2
Equilibrium reactions and parameters
Reaction Source
I H
2
S +MDEAMDEAH
+
+HS

[34]
II HCO
3

+OHH
2
O+CO
2
3
[34]
III MDEA+H
2
OOH

+MDEAH
+
[34]
IV 2H
2
OH
3
O
+
+OH

[38]
V MDEA
+
+H
2
OMDEA+H
3
O
+
[36]
VI H
2
S +H
2
OHS

+H
3
O
+
[38]
M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715 705
have signicant contribution to the observed reaction rate.
The reaction kinetics was measured by Pinsent [18]. Augsten
[19] proved in his work the dominant role of this reaction of
CO
2
with OH

at pH-values greater than 8. This condition


applies for MDEA-solution. The correlations for reaction
VIII which are used in this work can be found at Pinsent [18].
Tertiary amine acts as a base catalyst for the hydrolysis
of CO
2
to bicarbonate [20]:
CO
2
+H
2
O+MDEA MDEAH
+
+HCO

3
(IX)
This mechanism implies that tertiary alkanolamines, such
as MDEA, do not react directly with CO
2
. This thesis is
proved by the work of Versteeg and van Swaaij [16]. It would
not be necessary to implement reaction IX into the Aspen
system were it not for the fact that for the implementation of
the enhancement model an overall reaction is necessary (see
enhancement factor). Measurements on the kinetic of this
reaction have been made by various authors showing quite a
wide scattering [2,3,2129]. In this model the measurements
done by Rinker [25] are used to model the overall reaction
rate of reaction IX.
In the desorber the same reactive system is assumed as
for the absorber.
2.4. Enhancement model
When a transferring component undergoes reaction after
dissolving physically into the liquid, mass transfer rates of-
ten are increased dramatically. A several-thousand-fold im-
provement is not uncommon. This is reected in a much
higher value of the liquid-side mass transfer coefcient, de-
noted as above, for the chemically reactive case by k
L
ij
. The
physical and reactive transfer coefcients are related through
a so-called enhancement factor E
i
by the expression:
k
L
ij
= E
i
k
L
0
ij
(8)
where k
L
0
ij
is the mass transfer coefcient for the same pro-
cess taking place without present reaction. The enhancement
factor accounts quantitatively for the effect of reaction on
mass transfer and it depends, among other things, on the ki-
netic details of the particular reaction taking place. Among
all reactions occurring, the kinetically slowest one deter-
mines the enhancement factor and liquid-phase chemical re-
action does not inuence gas-side mass transfer coefcients.
To determine E
i
, it is necessary to consider only chemical
reactions taking place in the liquid phase.
Generally, the enhancement factor is a function of the
transport properties like diffusion coefcients and kinetic
parameters like the reaction rate and order of the reactants.
Therefore, it varies quite widely from stage to stage. This
is also the reason, why a single overall packing efciency
cannot equalise the inaccurate equilibrium-stage approach.
Each gas-solvent pair must be treated individually. Conse-
quently, the enhancement factor is unique to the system and
the operating conditions.
For H
2
S the resistance of the mass transfer is on the
gaseous side, therefore, the enhancement factor of H
2
S is
not important for determining absorption rates. For CO
2
the
absorption rate is dominated by the liquid side mass trans-
fer, requiring proper CO
2
enhancement factor calculations
[15]. In this work the Aspen enhancement model for CO
2
is replaced by the enhancement factor model, due to sev-
eral problematic assumptions the model makes. The Aspen
model calculates the enhancement using an average diffusion
coefcient for all components and an average mass transfer
coefcient. Due to this assumption every component has the
same concentration boundary layer. This is not a realistic
approach, because of the fact that the concentration prole
builds up in order of the reaction rates.
The enhancement factor model in this work does not use
averaged values of the diffusion coefcient and the mass
transfer coefcient. Thus a unique enhancement factor is
calculated for each component and then multiplied with the
corresponding mass transfer coefcient, as shown in Eq. (9).
This equation was developed by Danckwerts [30] and takes
a residual concentration in the bulk phase into account:
E

CO
2
=
Ha
_
E

CO
2
E

1
tanh
_
Ha
_
E

CO
2
E

1
_

_
_
_
_
_
_
_
_
1
c
B
CO
2
c
B
CO
2
equilib
c
I
CO
2
cosh
_
_
Ha
_
E

CO
2
E

1
_
_
_

_
(9)
with
Ha =
_
k
MDEA
D
L
CO
2
c
B
MDEA
k
L
0
CO
2
(10)
E

= 1 +
D
L
MDEA
D
L
CO
2

c
B
MDEA
c
I
CO
2
(11)
where k
MDEA
is the reaction rate corresponding to Rinker
[25] and k
L
0
CO
2
is the mass transfer coefcient for physical
absorption of CO
2
in MDEA solution.
Eq. (9) can be simplied for large values of E

or for
Hatta numbers lower than 2:
E

CO
2
=
Ha
tanh(Ha)
(12)
Last [15] developed in his work a simplied approxima-
tion to Eq. (9) for calculating the enhancement factor for
Hatta numbers large than 2 and E

values lower than 100:


E

CO
2
=
1
_
1
1
E

Ha
3/2
+
1
E
3/2

_
2/3
(13)
706 M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715
Table 3
Parameters for packing
Liquid load w (m
3
/m
2
h) Packing
<40 Mellapak X (60

) 0.0169 0.37 0.25


Mellapak Y (45

) 0.02 0.37 0.25


Rhombopak 9M 0.021 0.37 0.42
>40 Mellapak X (60

) 0.0075 0.59 0.25


Mellapak Y (45

) 0.0089 0.59 0.25


Rhombopak 9M
In this work Eqs. (12) and (13), respectively, are used for
calculation of the enhancement factor for CO
2
.
2.5. Geometric data input: hold-up, interfacial area and
liquid lm thickness
2.5.1. Hold-up in the absorber (structured packing)
Last [15] showed that the inuence of segment wise
calculation becomes most important at high pressures,
where among other effects the enhanced exothermic CO
2
-
absorption has a great inuence on the viscosity of the
liquid, which again has a retroaction on the hold-up.
Sulzer Chemtech Ltd. provides a correlation for their
packing which is based on empirical data.
h
L
= a
0.83
geo
w

_

L

+
H
2
O
_

(14)
The parameters , , of the different packings are given
in Table 3.
2.5.2. Hold-up desorber (random packing)
Equilibrium calculations normally do not require a
hold-up calculation, because of the assumed VapourLiquid-
equilibrium between the two phases. Due to the fact that
kinetic reactions requiring the hold-up to calculate the reac-
tion rate, will be used in the desorber too, a hold-up model
had to be implemented into the system.
Billet and Schultes [31] studied in 1999 a large number of
random packing and retained the following correlation for
calculating the hold-up:
h
L
=
_
12

L
g
L
w a
2
rp
_
1/3

_
d
h
a
rp
_
2/3
(15)
with
Re
L
=
w
L
a
rp

L
< 5 :
d
h
a
rp
= C
h

_
w
L
a
rp

L
_
0.15

_
w
2
a
rp
g
_
0.1
(16)
Re
L
=
w
L
a
rp

L
5 :
d
h
a
rp
= 0.85 C
h

_
w
L
a
rp

L
_
0.25

_
w
2
a
rp
g
_
0.1
(17)
where w is the liquid load, a
rp
the specic surface area of
the random packing, d
h
the hydraulic diameter and C
h
is the
hydraulic constant for the packing.
2.5.3. The liquid lm thickness (absorber)
The liquid lm thickness (absorber) on the inclined plate
of the structured packing can be obtained by setting the
potential and dissipation energy equal [32]:
=
3
_
3 w
L
a
int

L
g sin
2

(18)
with
w = a
int
u sin (19)
where a
int
is the interfacial area, u the average liquid lm
velocity and the angel of the inclined plane in reference
to the horizontal plane.
2.5.4. Interfacial area (absorber)
The correct calculation of the interfacial area is a difcult
aspect itself. Last [15] compared in his work interfacial cal-
culation routines form various authors, nding a large scat-
tering in the obtained values for the same packing (Mellapak
500 Y) from 36.4 to 255 m
2
/m
3
.
The interfacial model used in this work is based on a corre-
lation developed by Brunazzi [1]. Her approach is modied
in this work by introducing a weight factor , accounting
for non-perfectness, i.e. that not all of the hold-up takes part
in the lm ow, e.g. through channelling or accumulation
of liquid on the edges of the structure. In this approach only
h
L
of the hold-up takes part in mass transfer. The inter-
facial area can now be calculated with the adapted Brunazzi
equation:
h
L
= a
int
(20)
where h
L
is the liquid hold, a
int
the interfacial area and
the liquid lm thickness. If assumed that the weight fac-
tor is a function of the liquid load, can be correlated
to the ratio of the actual liquid loading to the liquid load-
ing at the loading point: =(w/w
0
)

. It shall be pointed
out that at the loading point should equal unity, thereby
assuming that the total hold-up is owing as liquid lm
down the packing. By solving the set of Eqs. (18) and
(20) the following correlation for the interfacial area can be
retrieved:
a
int
=
_
_
w
w
0
_

h
L
_
3/2
_

L
g sin
2

3w
L
(21)
where w
0
is the manufacture dened liquid load at the load-
ing point, and a adjustable parameter. It could be shown
in experiments with KOH-solution that the interfacial area
a
int
is proportional to the liquid loading: a
int
w
0.2
. Us-
ing this correlation and Eq. (21) one can determine by
M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715 707
exponent comparison. It follows =0.09667 according to
[15].
2.6. Thermodynamic models
Non-idealities of the gas phase are calculated using the
SRKmodel. For the liquid phase the Electrolyte-NRTL-model
is used to calculate the chemical potential and the activity
coefcients, respectively. The Electrolyte-NRTL model was
originally proposed by Chen et al. [33], for aqueous elec-
trolyte systems. It was later extended to mixed solvent elec-
trolyte systems [34]. For modelling the MDEAH
2
SCO
2
system the non-randomness NRTL parameters measured by
Austgen [19] are used.
The correlations and parameters used in Aspen were com-
pared with measured correlations [15]. Besides surface ten-
sion and viscosity only slight discrepancy between them
were found. Therefore, and to lower complexity of the sim-
ulation system, surface tension and viscosity only, were in-
cluded in user subroutines. An overview is given in Table 4.
2.7. Modelling approach
2.7.1. Absorber
The absorber as aforementioned is modelled with the
RATEFRAC
TM
model. This model is linked to the Aspen
Plus simulation engine, which provides the user interface,
the models to calculate the different apparatus in the ow di-
agrams and the physicochemical parameters. The user sub-
routines for calculation of viscosity, surface tension and the
kinetic routines are, therefore, linked directly to Aspen Plus,
whereas the mass transfer routine and the routine for calcu-
lation of the interfacial area are linked to RATEFRAC
TM
.
Fig. 3 shows the linkage and transfer variables of the differ-
ent user routines, adopted from Pacheco [35].
Fig. 3. Simulation system absorber.
Table 4
Overview of used correlations
Parameter Phase Correlation applied
Density Liquid Clark
Vapour Soave-Redlich-Kwong
Viscosity Liquid Last [15]
Vapour Chapman-Enskog-Brokaw
Diffusion-coefcients Liquid Wilke-Chang, Nerst-Hartley
Vapour Chapman-Enskog, Wilke-Lee
Fugacitiy Liquid Electrolyte-NRTL
Vapour Soave-Redlich-Kwong
Surface tension Liquid Last [15]
Aspen Plus allows the usage of user routines for nearly
all physiochemical properties, RATEFRAC
TM
provides
data interfaces to link user routines for binary mass transfer
coefcients, heat transfer coefcients, for interfacial area
and pressure drop. For the linkage of the hold-up model
and the enhancement factor model no data interfaces are
provided by the RATEFRAC
TM
model so they had to be
created.
2.7.2. Desorber
As the desorber is based on the RADFRAC
TM
module in-
cluded in Aspen Plus, where all routines are linked too. The
viscosity and surface tension subroutines provide, as in the
absorber simulation, the necessary physiochemical proper-
ties. Additionally a subroutine for calculating the reaction
rates for the kinetically controlled CO
2
reaction is linked to
the RADFRAC
TM
module. The hold-up, necessary for the
kinetic reaction, is obtained by a subroutine, which resolves
the necessary packing parameters out of the packing data
subroutine.
Fig. 4 shows the linkage and transfer variables of the
different user routines.
708 M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715
Fig. 4. Simulation system desorber.
3. Results and discussion
The developed model was tested on experimental data of
an industrial plant in Germany built by Lurgi. Representa-
tive input conditions for the acid gas absorber in a plant in
Germany are summarised in Tables 5 and 6. These condi-
tions were used for the simulation. A summary of the results
is given in Table 7.
In Figs. 5 and 6 a comparison between the measured
and calculated values for the H
2
S and CO
2
concentration
in the clean gas over the different measurements (D-1D-6)
Table 5
Characteristics of absorber and desorber
Characteristics Renery Germany
Absorber Desorber
Column diameter (m) 0.9 1.7
Packing height (m) 3.789/5.486 12
Backwash trays (stages) 13
Packing Mellapak 250 X Pall metal 35 mm,
Stage: 419
Number of segments/stages 20 20
Table 6
Typical operating data of amine-acid gas absorber/desorber renery Ger-
many
Parameter Renery Germany
Absorber Desorber
Inlet gas ow rate (kmol/h) 110140
Inlet liquid ow rate (kmol/h) 17182170
Inlet L/G ratio () 3.52
Inlet gas temperature (

C) 40.641.3
Inlet liquid temperature (

C) 25 112.9115.1
Inlet gas loading H
2
S (vol ppm) 800010 000
Inlet gas loading CO
2
(vol.%) 2.83.8
MDEA concentration (wt.%) 50 50
Inlet liquid H
2
S loading
(mol/mol amine)
0.0054 0.08950.179
Inlet liquid CO
2
loading
(mol/mol amine)
0.00016 0.00060.01522
Pressure (bar) 1.1 2.242.26
Table 7
Design results renery Germany
Parameter Renery Germany
Absorber Desorber
Outlet gas ow rate (kmol/h) 108142 13.7529.82
Outlet liquid ow rate (kmol/h) 351528 17632208
Outlet gas temperature (

C) 25.1 40.0
Outlet liquid temperature (

C) 23.424.5 126.9128.3
Outlet gas loading H
2
S
(mol ppm/mol%))
46165 9596
Outlet gas loading CO
2
(mol%) 2.73.5 0.311.05
Outlet liquid H
2
S loading
(mol/mol amine)
0.02310.0475 0.00560.010
Outlet liquid CO
2
loading
(mol/mol amine)
0.00640.0045 2.9 E-50.00014
is given. The two different packing heights in the measure-
ments result from two different inlet nozzles chosen for the
lean amine. Calculated values were obtained using the de-
veloped rate based model and TSWEET
4
.
TSWEET is a commercial equilibrium approach based
simulation tool, developed especially for gas sweetening pur-
poses. The usage of this program requires the estimation of
residence times on assumed trays in the absorber for the in-
ternal kinetic reaction model calculations. The disadvantage
of TSWEET is that the estimated residence time corresponds
only for very few operating conditions to a real residence
time calculated using real geometric parameters. In most
cases a virtual residence time has to be chosen, requiring a
broad experience of the user on the absorption behaviour of
the system at the specic operating conditions in advance.
It can be seen, that the non-equilibrium model, although
developed and tested on a laboratory column, gives excellent
results on the industrial absorber. It predicts the measured
H
2
S values far better (21 ppm average deviation) than the
TSWEET (88 ppm average deviation) simulation and gives
comparable results for the CO
2
values (Non-equilibrium:
0.22 vol.%; TSWEET: 0.28 vol.% average deviation). How-
ever, the most important aspect is, that the scaling up of the
non-equilibrium model was done without tting any param-
eter, of the measured values, whereas for the TSWEET cal-
culations, as aforementioned the residence time proles and
efciencies were selected with respect to optimised simula-
tion results.
In Fig. 7 the calculated vapour mole fractions concentra-
tion proles of CO
2
and H
2
S over the packing height for 3.9
and 5.62 m are given for two representative measurements,
respectively. The difference between the absorption of H
2
S
and CO
2
can be seen clearly. CO
2
which has to dissolve rst
before reacting with MDEA, has a nearly constant absorp-
tion gradient over the whole packing, whereas H
2
S shows
the typically bend chemical reactive absorption prole. It
can be generally stated that all simulation programs predict
a better cleaned gas in case of H
2
S for all measurements.
4
Brian Research Inc.
M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715 709
Fig. 5. Comparison of calculated and measured H
2
S concentrations in the clean gas.
For CO
2
this general trend can not be seen, but the relative
deviation of the calculated values is much lower. In Fig. 8
the liquid and vapour temperature prole of the absorber
are given. The difference in the temperature prole of the
vapour and the liquid phase is obvious. In equilibrium cal-
culations, this difference is neglected and calculations are
carried out with an average stage temperature.
The desorber model was tested against 13 measurements
on different days on an industrial desorber in Germany with
the parameters given in Table 6. These H
2
S and CO
2
load-
ings are compared with equilibrium calculations performed
with Aspen and TSWEET, which can be found in Figs. 9
and 10. Thereby Aspen-kinetic represents calculations done
with the developed regeneration model including the kinetic
Fig. 6. Comparison of calculated and measured CO
2
concentrations in the clean gas.
reaction of CO
2
, whereas Aspen-equilibrium represents sim-
ulations done with equilibrium based reactions only.
By looking at Fig. 10 the unsuitability of modelling the
CO
2
remaining loading with TSWEET or Aspen totally
based on vapourliquid equilibrium calculations is seen.
TSWEET obtains a regenerated amine solution, with only
innite remaining CO
2
loadings. The same applies to the
Aspen-equilibrium approach. Because much of the reboiler
duty is needed to strip CO
2
to such low values, less duty is
left to strip H
2
S of. Therefore, the H
2
S remaining loadings
calculated by TSWEET and Aspen-equilibrium are much
too high. TSWEET: predicts values which are more than
twice as high as the measured once, Aspen-equilibrium ap-
proach predicts the values four times higher.
710 M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715
Fig. 7. Comparison H
2
S and CO
2
concentration prole.
By variation of equilibrium constant of the CO
2
reaction
in the Aspen-equilibrium model an interaction between des-
orption of CO
2
and H
2
S was found: the lower the CO
2
lean
amine loading, the higher the H
2
S loading gets. This is due
to the fact that the reboiler duty is primarily used for strip-
ping of CO
2
, therefore, less duty is left for the stripping of
H
2
S. The reason for this behaviour is that the equilibrium
Fig. 8. Comparison vapour and liquid temperature prole.
reaction for CO
2
at stripping temperature is very fast, nearly
innite.
In the kinetic approach of CO
2
desorption the equilibrium
reaction of CO
2
is replaced with a kinetic reaction. This
model gives very good results for CO
2
(average deviation:
0.04 mmol/mol
Amin
) and H
2
S loadings (average deviation:
2 mmol/mol
Amin
) in comparison to the other two simulation
M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715 711
Fig. 9. H
2
S loading of regenerated MDEA.
Fig. 10. CO
2
loading of regenerated MDEA.
tools. Due to the interaction between the stripping of H
2
S
and CO
2
the higher remaining load for CO
2
results in a lower
remaining load for H
2
S. Thus the reboiler duty which is not
used for stripping of CO
2
can now enhance H
2
S desorption.
Fig. 11 shows the effect on the loading prole for H
2
S and
CO
2
over the packing height for different reboiler duties. The
desorption prole of CO
2
shows a much stronger inuence
on a variation of the reboiler duty as the H
2
S prole.
3.1. Application of both models on a currently running
project
Finally the rate based absorber and equilibrium based des-
orber model were used in a currently running project de-
signing a gas sweetening plant. A raw gas stream (methane,
ethane and trace components) of 8300 kmol/h with 3 mol%
H
2
S and 4 mol% CO
2
should be cleaned. The challenge
modelling this system was due to the high pressure of 56
bar and stringent clean gas conditions of 3 ppm for H
2
S and
0.71.5 for CO
2
.
Closed loop calculations were performed. As input data
for the absorber the design specications for the plant were
taken, whereas for the desorber the composition for the
rich amine stream calculated with the absorption model was
taken. The calculated stream value for the regenerated amine
solution was then used again as lean amine input stream to
the absorber.
Conditions for the acid gas absorber and desorber are
summarised in Table 8. In Fig. 12 the vapour mole fraction
prole of H
2
S and CO
2
in the absorber obtained from the
calculation is given over the packing height: the H
2
S con-
centration of the clean gas is achieved after 4 m of pack-
ing, whereas for achieving the desired CO
2
specication the
complete packing height is needed. It can be clearly seen
Table 8
Characteristics of absorber and desorber (current project)
Characteristics Current Project
Absorber Desorber
Column diameter (m) 2.6 4.2
Packing height (m) 11 12
Backwash trays (stages) 13
Packing Mellapak 250 Y Pall metal 35 mm,
Stages: 419
Number of segments/stages 30 19
712 M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715
Fig. 11. CO
2
and H
2
S loading prole at different reboiler duties.
Fig. 12. Vapour mole fraction prole absorber currently running project.
that the CO
2
concentration in the clean gas can be greatly
inuenced by the packing height, whereas for the H
2
S con-
centration in the clean gas the packing height plays a minor
role. Low H
2
S concentrations are achieved only by low re-
maining loading of the regenerated amine solution. Lower
H
2
S concentration in the clean gas can be obtained only
by adjusting the reboiler duty, in order to obtain lower H
2
S
loadings in the regenerated MDEA solution, whereas the
CO
2
concentration in the clean gas stays nearly unaffected.
4. Conclusion
In this work a rate-based algorithm implemented in As-
pen (RATEFRAC
TM
) was used to yield a predictive tool for
MDEA acid gases scrubbing processes. Mass transfer coef-
cients of the liquid phase are calculated adjusting a formu-
lation from Brunazzi [1] to experimental results, while for
the gas phase the Onda [2] formulation is used. A new en-
hancement model is developed to account for the chemical
M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715 713
reactions in the liquid phase. New correlations for geomet-
ric data, like hold-up and interfacial area, and for reaction
rates are provided to give reliable results. The tool which
was developed on laboratory measurements was applied on
an industrial scale absorber with excellent results. It allows
predicting acid gas concentrations in the clean gas of the
absorber, with only the input of geometrical data needed.
No estimations of parameters having an empirical charac-
ter rather than a physical background like the virtual resi-
dence time in TSWEET, have to be made which makes this
non-equilibrium model a powerful and straightforward tool.
For the desorber an equilibrium based model (RAD-
FRAC
TM
) with an implementation of the kinetic CO
2
re-
action was used. This tool was applied on industrial scale
as well, yielding excellent results, though not completely
rate-based. Together with the absorber model closed loop
calculations can be performed.
In the near future further measurements are planned at
industrial selective MDEA plants, which would enable to
test the developed models further and ne tune them.
Acknowledgements
We have to thank Matthias Linicus and Gerhard Schnei-
der, Lurgi Oel Gas Chemie/Frankfurt as well as to Fred
Gobin, Aspen Technology Inc. for their support and assis-
tance on computational matters.
Appendix A. Nomenclature
a
int
interfacial area (m
2
/m
3
)
a
geo
specic surface area of structured
packing (m
2
/m
3
)
a
h
hydraulic diameter (m
2
/m
3
)
a
rp
specic surface area of random
packing (m
2
/m
3
)
c
i
concentration of component i in the
solution (mol/kg
solution
)
c
t
total concentration per volume
(kmol/m
3
)
C
h
hydraulic constant for dumped
packing ()
d characteristic dimension (m)
d
h
hydraulic diameter (m)
d
rp
nominal diameter of packing or
packing size (m)
D
i
Fick diffusion coefcient (m
2
/s)
e point energy ux (J/(m
2
s))
E
CO
2
enhancement factor of CO
2
()
E

enhancement factor for innite fast


reaction ()
f
i
component feed rate (kmol/s)
g gravitational constant (m/s
2
)
G gas supercial mass velocity (kg/(m
2
s))
h
L
liquid holdup (m
3
/m
3
)
H specic enthalpy (J/kmol)
H ow distance (m)
H
i
Henry-coefcient of component i (Pa)
Ha Hatta number ()
k
ij
binary mass transfer coefcient for
the binary pair i and j (kmol/
(Pa m
2
s))
k
OH
reaction rateforward reaction VIII
(m
3
/(kmol s))
k
1
OH

reaction ratebackward reaction


VIII (1/s)
k
MDEA
reaction rateIX (m
3
/(kmol s))
K
ij
MaxwellStefan binary diffusion
coefcient of component i in
system i and j (m
2
/s)
K
i
equilibrium constant of specic
component i (unit depends on
specic reaction)
L total liquid ow (kmol/s)
n moles ()
N
i
molar ux of component i
(kmol/(m
2
s))
p
i
partial pressure of component i (Pa)
R gas constant (J/(mol K))
Re Reynolds number ()
S side stream ow (kmol/s)
Sc Schmidt number ()
T temperature (K)
u(y) velocity of liquid dependent on
coordinate y (m/s)
v
i
component vapour ow (kmol/s)
V total vapour ow (kmol/s)
w liquid load (m
3
/(m
2
s))
x
i
liquid phase molar fraction of
component i ()
y
i
vapour phase molar fraction of
component i ()
z
i
absolute value of ionic charge
Greek letters
constant for Sulzer Eq. (4) ()
constant for Sulzer Eq. (4) ()
constant for Sulzer Eq. (4) ()
liquid lm thickness (m)
dynamic viscosity (Pa s)
part of holdup owing as lm ()

i
chemical potential of component i
(J/mol)
kinematic viscosity (m
2
/s)
density (kg/m
3
)

L
liquid surface tension (Nm)
angel of inclined plane in reference
to the horizontal plane (

)
714 M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715
Subscripts
i, j component number
k stage number
m molecular species
p variable at operating pressure
s solvent
t estimated true value
w water
Superscripts

modication of specic variable


+ standard conditions (10
5
Pa; 293.15 K)
I interface
B bulk phase
L liquid bulk
LF liquid feed
V vapour bulk
VF vapour feed
0 refers to variable without reaction
Dimensionless numbers
Galilei number Ga =
g d
3
h
v
2
Grtz number Gz = Re Sc

H
Hatta number Ha =
_
k
MDEA
D
CO
2
c
B
MDEA
k
0
CO
2
Kapitsa
number
Ka =

3

4
g
Reynolds
number
Re =
u
max
d

Schmidt
number
Sc
ij
=

K
ij
Sherwood
number
Sh =
k
0
ij
d
K
ij
References
[1] E. Brunazzi, A. Paglianti, Liquid-lm mass-transfer coefcients in
a column equipped with structured packings, Ind. Eng. Chem. Res.
36 (1997) 37923799.
[2] K. Onda, H. Takeuchi, Y. Okumoto, Mass transfer coefcients be-
tween gas and liquid phases in packed columns, J. Chem. Eng. Jpn.
12 (1967) 5670.
[3] R. Krishnamurthy, R. Taylor, A non-equilibrium stage model of
multicomponent separation processes-part I: model description and
method of solution, AIChE J. 3 (1985) 449456.
[4] A.E. Cornelissen, Simulation of absorption of H
2
S and CO
2
into
aqueous alkanolamines in tray and packed columns, Trans. Inst.
Chem. Eng. 58 (1980) 242250.
[5] R.H. Weiland, M. Rawal, R.G. Rice, Stripping of carbon dioxide
from monoethanolamine solutions in a packed column, AIChE J. 28
(1982) 963973.
[6] R.H. Weiland, T. Chakravarty, A.E. Mather, Solubility of carbon
dioxide and hydrogen sulde in aqueous alkanolamines, Ind. Eng.
Chem. Res. 32 (1993) 14191430.
[7] H. Sardar, M.S. Sivasubramanian, R.H. Weiland, Simulation of com-
mercial amine treating units, Proceedings of the Gas Conditioning
Conference 35th, 1985.
[8] M.J. Frank, J.A. Kuipers, G.F. Versteeg, Modelling of simultaneous
mass and heat transfer with chemical reaction using the Maxwell
Stefan-Theory-I. Model development and isothermal study, Chem.
Eng. Sci. 10 (1995) 16451659.
[9] M.J. Frank, J.A. Kuipers, G.F. Versteeg, Modelling of simulta-
neous mass and heat transfer with chemical reaction using the
MaxwellStefan-Theory-II. Non isothermal study, Chem. Eng. Sci.
10 (1995) 16611671.
[10] R. Billet, M. Schultes, Predicting mass transfer in packed columns,
Chem. Eng. Technol. 16 (1993) 19.
[11] J.L. Bravo, J.A. Rocha, J.R. Fair, Comprehensive model for the
performance of columns containing structured packings, Inst. Chem.
Eng. Symp. Ser. 1 (1992) A489A507.
[12] M.H. Brito de, U. von Stockar, A.M. Bangerter, P. Bomio, Effective
mass-transfer area in a pilot plant column equipped with structured
packings and with ceramic rings, Ind. Eng. Chem. Res. 33 (1994)
647656.
[13] P.A. Nawrocki, K.T. Chuang, Carbon dioxide absorption into a stable
liquid rivulet, Can. J. Chem. Eng. 74 (1996) 247255.
[14] A.B. Ponter, P.H. Au-Yeung, Estimation of liquid lm mass transfer
coefcients for columns randomly packed with partially wetted rings,
Can. J. Chem. Eng. 60 (1980) 9499.
[15] Last, Absorption mit berlagerter chemischer Reaktion in Pack-
ungskolonnen bei Drcken bis 50 bar, Universitt Mnchen,
Mnchen, 1999.
[16] G.F. Versteeg, W.P.M. Van Swaaij, On the kinetics between CO
2
and alkanolamines both in aqueous and non-aqueous solutions-II.
Tertiary amines, Chem. Eng. Sci. 3 (1988) 587591.
[17] D.M. Austgen, G.T. Rochelle, Model of vapourliquid equilibria for
aqueous acid gasalkanolamine systems using the electrolyte-NRTL
equation, Ind. Eng. Chem. Res. 28 (1989) 10601073.
[18] B.R. Pinsent, L. Pearson, F.J. Roughton, The kinetics of combination
of carbon dioxide with hydrogen ions, Trans. Faraday Soc. 13 (1956)
1512.
[19] D.M. Austgen, A Model of VapourLiquid Equilibria for Acid
GasAlkanolamineWater Systems, University of Texas, Austin,
1989.
[20] J. Ko, M. Li, Kinetics of absorption of carbon dioxide into solutions
of N-methyldiethanolamine plus water, Chem. Eng. Sci. 19 (2000)
41394147.
[21] G. Astarita, D.W. Savage, Gas absorption and desorption with re-
versible instantaneous chemical reaction, Chem. Eng. Sci. 35 (1980)
17551764.
[22] N. Haimour, A. Bidarian, O.C. Sandall, Kinetics of the reaction
between carbon dioxide and methyldiethanolamine, Chem. Eng. Sci.
6 (1987) 13931398.
[23] R.J. Littel, W.P. van Swaaij, G.F. Versteeg, Kinetics of carbon dioxide
with tertiary amines in aqueous solution, AIChE J. 11 (1990) 1633
1640.
[24] G. Xu, C. Zhang, S. Qin, Y. Wang, Kinetics study on absorption
of carbon dioxide into solutions of activated methyldiethanolamine,
Ind. Eng. Chem. Res. 3 (1992) 921927.
[25] E.B. Rinker, S.S. Ashour, C.S. Orville, Kinetics and modelling
of carbon dioxide absorption into aqueous solutions of N-methyl-
diethanolamine, Chem. Eng. Sci. 5 (1995) 755768.
[26] W. Yu, G. Astarita, Kinetics of carbon dioxide absorption in solution
of methyldiethanolamine, Chem. Eng. Sci. 8 (1985) 15851590.
[27] D. Barth, C. Tondre, J. Delpuech, Kinetic and mechanism of the
reactions of carbon dioxide with alkanolamines: a discussion con-
cerning the case of MDEA and DEA, Chem. Eng. Sci. 12 (1984)
17531757.
M. Bolh` ar-Nordenkampf et al. / Chemical Engineering and Processing 43 (2004) 701715 715
[28] Y. Wang, C. Zhang, S. Qin, Kinetic study of absorption of carbon
dioxide in aqueous MDEA, J. Chem. Ind. Eng. 4 (1991) 466
474.
[29] D. Richon, F. Pani, A. Gaunand, R. Cadours, C. Bouallou, Ki-
netics of absorption of carbon dioxide in concentrated aqueous
methyldiethanolamine solutions in the range 296343 K, J. Chem.
Eng. Data 2 (1997) 353359.
[30] P.V. Danckwerts, Absorption by simultaneous diffusion and chemical
reaction, Trans. Faraday Soc. 46 (1950) 300304.
[31] R. Billet, M. Schultes, Prediction of mass transfer columns with
dumped and arranged packings, Trans. Inst. Chem. Eng. 38 (1999)
498.
[32] J.H. Spurk, Strmungslehre-Einfhrung in die Theorie der
Stromung, second ed, Springer, Berlin, 1987.
[33] C.C. Chen, H.I. Brit, J.F. Boston, L.B. Evans, A local composition
model for the excess Gibbs energy of aqueous electrolyte systems:
part I: single solvent, single completely dissociated electrolyte sys-
tem, AIChE J. 4 (1982) 588596.
[34] G. Valle, P. Mougine, S. Julian, W. Frst, Representation of CO
2
and H
2
S absorption by aqueous solutions of diethanolamine using
an electrolyte equation of state, Ind. Eng. Chem. Res. 38 (1999)
34733480.
[35] M.A. Pacheco, G.T. Rochelle, Rate-based modelling of reactive ab-
sorption of CO
2
and H
2
S into aqueous methyldiethanolamine, Ind.
Eng. Chem. Res. (1998) 41074117.
[36] D.M. Austgen, G.T. Rochelle, Model of vapourliquid equilibria for
aqueous acid gasalkanolamine systems, 2. Representation of H
2
S
and CO
2
solubility in aqueous MDEA and CO
2
solubility in aqueous
mixtures of MDEA with MEA or DEA, Ind. Eng. Chem. Res. 30
(1991) 543555.
[37] G.T. Rochelle, M. Posey, Modelling CO
2
and H
2
S solubility in
MDEA and DEA: design implications, Proceedings of the 1996 75th
Annual Convention of the Gas Processors Association (1996) 8091.
[38] T.J. Edwards, G. Maurer, J. Newman, J.M. Prausnitz, Vapourliquid
equilibria in multicomponent aqueous solutions of volatile weak
electrolytes, AIChE J. 24 (1978) 966976.

You might also like