Embed Size (px)
Citation preview
Author's Accepted Manuscript
Mass transfer and kinetics of carbon dioxideabsorption into loaded aqueous monoethano-lamine solutions
Xiao Luo, Ardi Hartono, Saddam Hussain,Hallvard Svendsen
PII: S0009-2509(14)00581-8DOI: http://dx.doi.org/10.1016/j.ces.2014.10.013Reference: CES11928
To appear in: Chemical Engineering Science
Received date: 9 June 2014Revised date: 3 October 2014Accepted date: 10 October 2014
Cite this article as: Xiao Luo, Ardi Hartono, Saddam Hussain, HallvardSvendsen, Mass transfer and kinetics of carbon dioxide absorption intoloaded aqueous monoethanolamine solutions, Chemical Engineering Science,http://dx.doi.org/10.1016/j.ces.2014.10.013
This is a PDF file of an unedited manuscript that has been accepted forpublication. As a service to our customers we are providing this early version ofthe manuscript. The manuscript will undergo copyediting, typesetting, andreview of the resulting galley proof before it is published in its final citable form.Please note that during the production process errors may be discovered whichcould affect the content, and all legal disclaimers that apply to the journalpertain.
www.elsevier.com/locate/ces
1
Mass transfer and kinetics of carbon dioxide absorption
into loaded aqueous monoethanolamine solutions
Xiao Luo, Ardi Hartono, Saddam Hussain, Hallvard Svendsen*
*Department of Chemical Engineering, Norwegian University of Science and Technology, Sem Sælands
vei 4, N-7491 Trondheim, Norway
ABSTRACT
The kinetics of the reaction between carbon dioxide and aqueous solutions of 1 and 5 mole/L
monoethanolamine (MEA) pre-loaded with CO2 were investigated over the temperature range 298 to
343 K and for CO2 loadings from 0 to 0.4 mole CO2/ mole MEA in a wetted wall column reactor
(WWC) and a string of discs contactor (SDC). A total of 227 new data points are provided for loaded
solutions including all underlying data necessary for other researchers to develop own models.
Comparisons are made between recent literature data and this study and they are found to be consistent
with each other. Three different kinetic models, a simplified soft model, a concentration-based model
and an activity-based model were developed and validated against the experimental data and by a
penetration type mass transfer model in order to analyze the absorption rate and understand the reaction
process. Results show good agreement between the models at low loadings and kinetic parameters are
provided for all models. Above a loadings of 0.3 mole CO2/mole MEA it is recommended to use the
activity based model as systematic deviations occurred in the soft and concentration based models. The
effect of depletion of free amine at the gas-liquid interface on the kinetic and mass transfer calculations
2
was investigated and it was found insignificant at high amine concentrations, low CO2 loadings, low
CO2 driving forces and temperatures. However, the effect does become significant when either reducing
the amine concentration or increasing the CO2 driving force, CO2 loading or temperature. Furthermore,
there is an upper limit for the CO2 driving force for each amine concentration below which the chemical
reaction can be assumed to be in the pseudo 1st order regime.
KEYWORDS:
CO2; MEA; Absorption; Kinetics; Mass transfer; Activity-based.
1. Introduction
Aqueous monoethanolamine (MEA) is the most commonly used solvent for separating carbon dioxide
(CO2) from flue gases. Proper and accurate chemical reaction data between aqueous MEA and CO2 play
an important role in the simulation and modeling of the absorption process and in the design and scale
up of the absorber column. Post combustion processes normally work in the loading range between 0.2
– 0.5 mole CO2/mole amine and it is thus important to validate kinetic models also for loaded solutions.
Because of uncertainty and experimental difficulty when studying CO2 reacting with loaded aqueous
MEA, and because an accurate vapor-liquid equilibrium (VLE) model is needed, only few investigators
have published results in this area. Littel (Littel et al. 1992) studied the kinetics of CO2 reacting with
aqueous MEA in a stirred cell reactor in the temperature range 318-333K, in the MEA concentration
range 0-3.2 M, and with CO2 loadings from 0.01 to 0.1 mole CO2/mole amine. Because of low loadings
the CO2 back pressure of the loaded solution was not taken into consideration. Aboudheir (Aboudheir et
al. 2003) reported kinetic data of the reaction between CO2 and aqueous MEA from a laminar jet
absorber over the temperature range 293-333 K, over the MEA concentration range 3-9 M with CO2
loadings from 0.1 to 0.5 mole CO2/mole amine. Dang (Dang and Rochelle 2003) reported absorption
3
measurements in 2.5 and 5 M MEA solutions from a wetted wall column for 313 and 333K and in the
loading range 0.01 – 0.5 mole CO2/mole amine. Puxty (Puxty et al. 2010) performed absorption
experiments in 5 M MEA solution at 313 and 333K in a wetted wall column reactor and varied the CO2
loading from 0 to 0.5 mole CO2/mole amine. Dugas (Dugas and Rochelle 2011) used a wetted wall
column reactor to measure the absorption rate constant into the CO2-MEA-H2O system within the MEA
concentration range 2.5 – 7.3 M, temperature range from 313 to 373 K and CO2 loading range from 0 to
0.5 mole CO2/mole amine. Comparisons between data from this study with literature data were made as
shown in Figure 4 and Figure 5.
A numerical analysis is needed in order to describe the mass transfer process. Versteeg (Versteeg et al.
1989) developed numerical methods using both the film and penetration theories (Higbie 1935). For
CO2/H2S reacting with aqueous MEA it showed a maximum of 40% deviation due to uncertainties in
the underlying physical properties and experimental procedures. Aboudheir (Aboudheir et al. 2003)
proposed a penetration type absorption-rate/kinetics model which considered all the possible reactions
and mass balances in the liquid phase. It was used to numerically analyze both their own data and some
literature data available within a maximum 33.5% deviation. Akanksha (Akanksha et al. 2007) used a
two-dimensional numerical approach to describe the mass transfer process in a continuous film
contactor and the results gave maximum 15% deviation with their own experimental data but no other
literature data were used for comparison.
In the present work 227 new experimental CO2 absorption data points from a wetted wall column and
a string of discs contactor for 1 and 5 M monoethanolamine (MEA) pre-loaded with CO2 are presented
over a range in temperature from 298 to 343 K and CO2 loadings from 0 to 0.4 mole CO2/mole amine.
For the interpretation of kinetic parameters, three approaches were used. First a simplified equilibrium
model relating CO2 partial pressure to loading and temperature was used with concentration based
kinetics. Next a concentration based equilibrium model combined with concentration based kinetics was
tested, and finally an extended UNIQUAC based equilibrium model with activity based kinetics was
4
used. All models used kinetic data extracted from the mass transfer measurement using the pseudo first
order assumption. Finally, a penetration model approach was used to validate the other models.
2. Mass transfer and reaction mechanisms
The chemical reactions describing CO2 absorption into the CO2-MEA-H2O system based on
carbamate formation with dissociation of MEA can be written as:
+2 2 2 3CO MEA H O MEACO H O−⎯⎯→+ + +←⎯⎯ (1)
+3 2MEA H O MEAH H O+⎯⎯→+ +←⎯⎯ (2)
When combining these two reactions, the general chemical reaction can be written as:
2 22CO MEA MEACO MEAH− +⎯⎯→+ +←⎯⎯ (3)
In aqueous amine solutions, there are other reactions than carbamate formation that take place in
parallel, mainly the hydration of CO2 and the direct reaction with hydroxyl ions (Eq. 4). Here the rate
constant of the hydration of CO2 while compared to carbamate formation its contribution to the overall
reaction rate can be negligible since it is a very slow reaction.
2 3CO OH HCO− −+ ↔ (4)
In the wetted wall column and string of discs apparatuses used in this study, the method of
interpretation of the mass transfer data is almost the same as used for CO2 reacting with unloaded MEA
solutions (Luo et al. 2012). The main difference is that for loaded solutions an equilibrium model has to
be implemented to obtain back-pressures for CO2.
Using a film model approach the CO2 mass transfer rate can be expressed as the product of an overall
mass transfer coefficient and the logarithmic mean pressure difference between the inlet and outlet of
the reactor.
2=CO GN K LMPD⋅ (5)
w
o
t
D
r
V
o
a
c
t
1
t
where
Based on t
overall mass
When con
the gas pha
Depending o
reaction rate
Versteeg (Ve
order regime
amine conce
checked by e
the infinite r
1999). In the
the film mod
Therefore,
the two-film
s transfer coe
sidering a g
se but is so
on the regim
can be simp
ersteeg et al
e. In general
entration is h
evaluating th
reaction rate
e fast irreve
del by the fol
, E can be su
m theory, and
efficient can
as containin
oluble in th
me under wh
plified. For d
l. 1996) sugg
l terms, the
high enough
he Hatta num
enhancemen
ersible reacti
llowing equa
ubstituted by
2 1CON
k
=
d assuming
be expresse
1 1=kG gK
+
ng CO2 in co
he liquid, wh
hich the mas
deriving CO
gested that t
pseudo first
h not to chan
mber ( Ha ),
nt factor COE
ion regime,
ation.
E H=
y Ha in Eq. 7
20
1 CO
g l
LMPDH
k k E
=+
fast reaction
ed as:
2
0
1k
CO
l
HE k
+ =⋅
ontact with a
hile the am
ss transfer is
O2 reaction ki
the experime
t order regim
nge through
which indica
2 ,O ∞ (Danckw
when 3 H<
0obs
l
k DHa
k=
7, leading to
1 C
g obs
LMPDH
k k D
=+
n only taking
g
1 1k k'g
+
a liquid amin
mine is main
s carried out
inetics from
ents should b
me is achiev
the reaction
ates whether
werts 1970;
2 ,COHa E ∞ ,
2COD
the pseudo f
2
2
GCO
CO
D K
D
= ⋅
g place in th
ne solution, t
nly present
t, the genera
amine absor
be carried ou
ved if the rea
n layer. The
r the reaction
Versteeg et
, then E can
first order ra
LMPD
he reaction l
the CO2 is p
in the liqui
al expression
rption measu
ut in the pse
action is fast
se condition
n is fast or s
al. 1996; Le
n be calcula
ate expressio
5
(6)
layer, the
(7)
present in
id phase.
n for the
urements,
eudo first
t and the
ns can be
slow, and
evenspiel
ated from
(8)
on:
(9)
6
The calculation of the physical liquid and gas mass transfer coefficients 0lk and gk in the wetted wall
and string of discs columns is based on characterization of the two apparatuses done in previous works
(Hartono et al. 2009; Luo et al. 2012).
The observed kinetic rate constant kobs can then be calculated from Eq. 10. When also taking the
direct reaction between CO2 and OH- into account, an apparent rate constant relating to the reaction
between CO2 and MEA can be expressed as:
app obs OHk k k OH−
−⎡ ⎤= − ⎢ ⎥⎣ ⎦ (10)
For solutions with CO2 loading, the effect of ionic strength should be taken into account. For the
hydroxide reaction this was done by Eq. 11 (Pohorecki and Moniuk 1988):
OH OHk OH k OH bI− −
− ∞ −⎡ ⎤ ⎡ ⎤= +⎢ ⎥ ⎢ ⎥⎣ ⎦ ⎣ ⎦ (11)
The ionic strength contribution I can be calculated from the speciation of the system, which can be
obtained e.g. from an extended UNIQUAC model (Aronu et al. 2011). The contribution of the van
Krevelen coefficient b can be obtained from Browning (Browning and Weiland 1994). It was found that
the contribution of hydroxyl ion in Eq. 10 in CO2 loaded aqueous MEA solutions was maximum 10% of
the observed kinetic rate constant for the highly CO2 loaded solutions, and does not affect the rate
constant determination significantly in the low CO2 loading range.
In order to interpret the appk into kinetic rate constants for MEA, a kinetic model is needed. Two
models have been used in the literature, the zwitterion model proposed by (Caplow 1968; Danckwerts
1979), as shown in Eq. 12. And the direct, or termolecular, model proposed by Crooks and Donnellan
(Crooks and Donnellan 1989) and da Silva and Svendsen (da Silva and Svendsen 2004), as shown in Eq.
14.
[ ][ ]
[ ]2
2 2
11CO
b
k MEA COr k
k B−
− =+∑
(12)
7
In Eq. 12, [ ]bk B∑ indicates the contribution to proton transfer in the reaction Eq. 2 by all bases
present in the solution. Blauwhoff (Blauwhoff et al. 1984) suggested that water, hydroxide ions and the
amines can act bases in reaction Eq. 2. When the zwitterion formation, Eq. 1, is the rate determining
step, then [ ]11 bk k B−≥ ∑ and Eq. 12 can be simplified as:
[ ] [ ][ ]2 2 2 2CO appr k CO k MEA CO− = = (13)
When water, hydroxide ions and the amines become the dominating bases, then [ ]11 bk k B−≤ ∑ and
the forward reaction rate of the zwitterion mechanism, with deprotonation of the zwitterion being the
rate determining step can be seen to become similar to the termolecular or direct mechanism (da Silva
and Svendsen 2004). It is common to lump the effects of hydroxide ions and water on the deprotonation
together, here denoted as water, and the expression for the forward reaction becomes:
[ ] [ ] [ ]{ }[ ][ ]2 22 2 2
T TCO app MEA H Or k CO k MEA k H O MEA CO− = = + (14)
Here superscript T denotes the Termolecular mechanism, but using the zwitterion mechanism would
result in the same expression.
3. Experimental method and procedure
3.1 Chemicals and procedures
Gases used, CO2, purity > 99.999 mol% and N2, purity > 99.999 mol%, were supplied by YARA
PRAXAIR. The purity of monoethanolamine (MEA), supplied by SIGMA-ALDRICH, was higher than
99.0% and was used without further purification.
Only absorption rate measurements were carried out both in a wetted wall column apparatus and a
string of discs contactor. During the experiments a blend of N2 and CO2 was used as gas phase with
varying volume fraction of CO2 in the range of 1% – 15% in the WWC and between 0.25% - 5% in the
SDC. The reason for the CO2 volume fraction differences is the difference in contact time in the two
contactors. A long contact time and high CO2 volume fraction will cause significant CO2 loading
changes in the liquid phase, and affect the data analysis. 1M and 5M aqueous MEA (Monoethanolamine)
8
solutions were used as liquid phase in the WWC and only 5M aqueous MEA solution was used in the
SDC. The CO2 loadings varied in the range 0 – 0.4 mole CO2/mole MEA, and the temperature varied
from 298 K to 343 K.
The wetted wall column experimental setup is shown in Figure 1. The reactor is operated counter
currently with liquid flowing from top to bottom and gas flowing upwards. The liquid is pumped into
the apparatus by a gear pump with a defined flowrate, normally 8.4×10−7m3 s−1. Liquid samples were
taken from both the inlet and outlet of the reactor. The main gas circulation is provided by a channel
blower, normally set at 4×10−4 Nm3 s−1. The gas flow rate was set as high as possible to have a small
gas phase resistance, but limited to ascertain insignificant impact on the liquid phase flow structure. The
gas makeup stream was a mixture of CO2 and N2 containing 1 – 15 vol% CO2. The concentrations were
adjusted by digital mass flow controllers. The absorption rate was measured (accuracy ±0.01%) by a
CO2 balance between the inlet CO2 through the mass flow controller, and the CO2 leaving through the
vent (water lock bleed, shown as “Bleeding” in Figure 1). The gas sampling line, yellow in Figure 1,
exits the main gas circulation loop, goes to the analyzer, and then is returned to the main circuit via the
constant pressure bleed (water lock) in order to keep constant atmospheric pressure in the system. The
whole wetted wall column setup was placed inside a closed thermostated cupboard which could be
heated to temperatures up to 353 K. Temperatures (accuracy ± 0.1 ºC) and main circuit pressure
(accuracy ±0.25%) were measured at a rate of 1 Hz. More detailed information about the experiment
procedure can be found in Luo et al. (Luo et al. 2012).
Figure 1 Diagram of wetted wall column experiment setup
The string of discs column, shown in Figure 2, was also operated in counter current mode and the
principle of operation is the same as for the WWC. The liquid enters at the top and the gas at the bottom.
The liquid flowrate was set at 8.4×10−7m3 s−1 which is higher than the minimum flowrate needed for the
CO2 absorption flux to be flowrate independent. Mixtures of CO2 and N2 containing 0.25 – 5 vol% CO2
9
as a make-up gas were produced as for the WWC using flow controllers for CO2 and N2, and mixed
with the main loop gas and then fed into the SDC with CO2 flowrate 1.0×10−3 Nm3 s−1. Gas for analysis
exited the main loop before the circulation blower, was analyzed in a Rosemount Binos 100 analyzer
(accuracy ±0.01%) and returned to the main loop before the blower. A more detailed explanation of the
apparatus and experimental procedure can be found in Hartono (Hartono 2009).
Figure 2 Diagram of string of discs contactor experiment setup
The total liquid phase CO2 concentration was analyzed by a BaCl2 based precipitation titration
method, see Ma’mun (Ma'mun et al. 2005) for details. The total amine analyses were done by acid-base
titration, see Hartono et al. (2014). The CO2 loading accuracy was ±0.01 mole CO2/mole amine. A
Labview program was developed for the operation and control of the apparatus through a 12 bit
Fieldpoint I/O system.
3.2 Physical properties
Aqueous solutions of MEA were prepared on a mass basis (Precision balance model MS6002S with
accuracy ± 10-5 kg). MEA was dissolved in deionized water and partially loaded by blowing pure CO2
through the aqueous MEA solutions. The CO2 loadings were estimated from the weight change and
determined accurately by the methods given in section 3.1 (Ma'mun et al. 2005, Hartono et al. 2014).
More detailed information of the physical properties measurements can be found in (Hartono et al.
2014).
Densities of aqueous 1M and 5M MEA solutions with CO2 loading were measured with an Anton
Paar DMA 4500 Densitometer with the XSample 452 for automatic filling, rinsing and drying by
Hartono (Hartono et al. 2014). The data were regressed with a relative total standard deviation of 0.06%
and 0.17%, respectively and are given by the following equations:
2
269006.2 415.286531 51.42446 exp 0.01 1387 1000: MEA T TM
T Tα α
ρ⎛ ⎞⎛ ⎞ − ⎟⎜⎟⎜ ⎟= + + ⋅ + ⋅⎜⎟⎜ ⎟⎜⎟⎜ ⎟⎟⎝ ⎠ ⎜⎝ ⎠
(15)
10
2
249573 394.98881 222.25267 e5 xp 0.04254 10 0: 0MEA T T T TM α
ρα⎛ ⎞⎛ ⎞ − ⎟⎜⎟⎜ ⎟= + + ⋅ + ⋅⎜⎟⎜ ⎟⎜⎟⎜ ⎟⎟⎝ ⎠ ⎜⎝ ⎠
(16)
The viscosities of aqueous 1M and 5M MEA with CO2 loading were measured with an Anton Paar
MCR-100 Viscometer with Double gap measuring cell (DG-26.7) by Hartono (Hartono et al. 2014). The
data were regressed with a relative total standard deviation of 1.3% and 0.9%, respectively and given by
the equations below:
2
2 3
2060.04 6214791 : 1 67.1354 363188 exp 8143970 0.001MEAMT T T T Tα α α
μ⎡ ⎤⎛ ⎞ ⎛ ⎞−⎢ ⎥⎟ ⎟⎜ ⎜= + + ⋅ + − ⋅⎟ ⎟⎜ ⎜⎢ ⎥⎟ ⎟⎜ ⎜⎝ ⎠ ⎝ ⎠⎢ ⎥⎣ ⎦
(17)
2
2 3
2128.24 7141745 : 1 430.952 157492 exp 194757 0.001MEAMT T T T Tα α α
μ⎡ ⎤⎛ ⎞ ⎛ ⎞−⎢ ⎥⎟ ⎟⎜ ⎜= + + ⋅ + − ⋅⎟ ⎟⎜ ⎜⎢ ⎥⎟ ⎟⎜ ⎜⎝ ⎠ ⎝ ⎠⎢ ⎥⎣ ⎦
(18)
For the CO2 loaded aqueous 1M and 5M MEA solutions CO2 solubilites were obtained by Hartono
(Hartono et al. 2014). The solubility measurements were regressed with a relative total standard
deviation of 1.7% and 2.1%, respectively:
2
22
1733.99361054.83 121659 exp 0.17932 128338: 1 01 00CO MEA T TH
TM α α
α−
⎛ ⎞ ⎛ ⎞⎟ ⎟⎜ ⎜+ ⋅ − ⋅ − − ⋅⎟ ⎟⎜ ⎜⎟ ⎟⎝ ⎠ ⎝ ⎠=⎜ ⎜ (19)
2
22
1472.25496.563 341697 exp 1.69131 128338 1: 005 0CO MEA TM H
T Tα α
α−
⎛ ⎞ ⎛ ⎞⎟ ⎟⎜ ⎜+ ⋅ ⋅ − − ⋅⎟ ⎟⎜ ⎜⎟ ⎟⎜ ⎜⎝ ⎠ ⎝ ⎠= (20)
The diffusivity of CO2 in MEA can be determined by the Stokes-Einstein equation as reported by
Versteeg (Versteeg and Van Swaaij 1988), which can be seen in Eqs. 21 and 22. The diffusivity of
MEA in the aqueous MEA solutions was determined from the correlation developed by Snijder (Snijder
et al. 1993), which can be seen in Eqs. 23 and 24.
From (Versteeg et al. 1996), the diffusivity of CO2 in H2O can be calculated by:
2 2
6 21192.35 10 expCO H ODT−
− −⎛ ⎞⋅ ⋅ ⎜⎝
= ⎟⎠
(21)
And diffusivity of CO2 in aqueous MEA solution (Versteeg et al. 1988) can be obtained as:
11
2
2 2 2
0.8H O
CO CO H Oamine
D Dμμ−
⎛ ⎞= ⋅⎜ ⎟
⎝ ⎠ (22)
The diffusivity of the alkanolamine in aqueous alkanolamine solutions can be correlated with a
Stokes-Einstein relation (Versteeg et al. 1988):
2
2
0.6H O
MEA MEA H OMEA
D Dμμ−
⎛ ⎞= ⋅⎜ ⎟
⎝ ⎠ (23)
The calculation of diffusivity of amines in water was further investigated by Snijder (Snijder et al.
1993) and for MEA a correlation was suggested as:
[ ]2
0.82198.3exp 13.275 0.07814MEA H OD MEAT−
⎛ ⎞= − − − ⋅⎜ ⎟⎝ ⎠
(24)
It was mentioned by Snijder (Snijder et al. 1993) that the Stokes-Einstein correlation of alkanolamine
diffusivity was evaluated from the measurements of infinite dilution and extended to concentrated
solutions, which can give up to 25% deviation. Hence for the MEA diffusivity calculation, the Snijder’s
correlation is preferable and was used.
A soft model for calculating the back pressure of 5 M and 1M MEA solution with CO2 loading can be
fitted with vapor-liquid equilibrium experiments (Aronu et al. 2011) as following equations showed
respectively, which can be seen in Figure 3 (a) and (b):
( )( )
2
* -10337.51 105M MEA: P = exp 1.8 log + +31.16 +4177.13 4516.981+exp +9.59 exp +8.69 log
CO TT T
αα
⎛ ⎞⎟⎜ ⎟⎜ ⎟⎜ ⎛ ⎞ ⎟⎜ ⎟ ⎟⎜⋅⎜ ⎟ ⎟⎜ ⎟⎜ ⎜ ⎟⎛ ⎞ ⎛ ⎞⎝ ⎠⎜ ⎟⎟ ⎟⎜ ⎜ ⎟⎜ − ⋅ − ⋅⎟ ⎟⎜ ⎜ ⎟⎜ ⎟ ⎟⎜ ⎜ ⎟⎜ ⎝ ⎠ ⎝ ⎠⎝ ⎠
(25)
( )( )
2
* 9710.48 101M MEA: P = exp 1.8 log + +28.62 +1520.63 252.031+exp +5.15 exp 0.63 log
CO TT T
αα
⎛ ⎞⎟⎜ ⎟⎜ ⎟⎜ ⎛ ⎞ ⎟⎜ ⎟ ⎟⎜⋅ −⎜ ⎟ ⎟⎜ ⎟⎜ ⎜ ⎟⎛ ⎞ ⎛ ⎞⎝ ⎠⎜ ⎟⎟ ⎟⎜ ⎜ ⎟⎜ − ⋅ − − ⋅⎟ ⎟⎜ ⎜ ⎟⎜ ⎟ ⎟⎜ ⎜ ⎟⎜ ⎝ ⎠ ⎝ ⎠⎝ ⎠
(26)
Figure 3 The soft equilibrium model fitting with experimental VLE data in (a) 5M MEA and (b) 1M MEA solution
4. Results and discussion
12
4.1 Mass transfer coefficients
Because most literature data on mass transfer coefficients are reported only in 5M MEA solutions, the
comparisons are made in 5M MEA system as an example in this study. By introducing the soft
equilibrium model, Eqs. 25 and 26 into Eqs. 5 – 7, GK and gk ′ from the 5M MEA experiments can be
calculated and compared with measurements from literature, see the points in Figure 4 and Figure 5.
Figure 4 Prediction of kg’ by soft model with experimental data and literature data
Figure 5 Prediction of KG by soft model with experimental data and literature data
Figure 4 and Figure 5 show the overall mass transfer coefficient and liquid film mass transfer
coefficient for the 5M MEA systems from wetted wall column and string of discs contactor experiments
respectively. Also included are data from available literature for laminar jet absorber, (Aboudheir 2002),
and wetted wall column, (Dang and Rochelle 2003; Puxty et al. 2010; Dugas and Rochelle 2011). One
problem with many literature sources is that the mass transfer data are not reported and thus they cannot
be used for reinterpretation. The gk ′ and GK decrease with increasing CO2 loading 0.5 and a significant
temperature dependency of the mass transfer coefficient for each CO2 loading can be seen. Furthermore,
the mass transfer coefficients systematically decrease when the CO2 driving forces decreasing as
indicated by the downward trend in the series of data points at certain temperature, in particular for the
SDC.
4.2 Soft model
Using the pseudo first order assumption, 2 2g obs CO COk k D H′ = ⋅ , and the experimentally determined
gk ′ values, kobs and kapp can be determined, and subsequently a second order kinetic rate constant can be
obtained from Eq. 12. The second order rate constant is what is given in most literature sources.
However, is seen that the second order rate constant will depend on the amine concentration and with
the soft model approach, kinetic expressions must then be developed individually for each MEA
concentration. The same conclusions were drawn by Aboudheir et al., 2003 and Luo et al., 2012. This
13
was done for the two MEA concentrations using non-linear regression for the parameter fitting. For the
1M and 5M aqueous MEA with CO2 loading the k2 values can be expressed as:
52
34581 : 8.87 10 expM kT
⎛ ⎞⎟⎜= × ⋅ − ⎟⎜ ⎟⎜⎝ ⎠ (27)
62
36935 : 4.396 10 expM kT
⎛ ⎞⎟⎜= × ⋅ − ⎟⎜ ⎟⎜⎝ ⎠ (28)
The k2 values from Eqs. 27 and 28 were inserted in the expressions for gk ′ and GK and the lines given
in Figure 4 and Figure 5 are based on this model. All the data points fit nicely with the soft model trends
of CO2 loading dependency in the lower loading region. However, more deviation in the high loading
range can be seen, and the very low CO2 driving forces partly associated with the high loadings, resulted
in GK and gk ′ values relatively far away from the trend lines. This might be caused by higher
experimental uncertainty in the mass transfer experiments at low driving forces, but also by
uncertainties in the solubility measurements when CO2 loadings are high. The low driving force
experiments will be particularly sensitive to this.
The second-order kinetic rate constant for loaded solutions as a function of loading and for different
temperatures in the 1M and 5M MEA solutions can be seen in Figure 6 and Figure 7, respectively. The
points are the experimental data from the wetted wall column and string of discs apparatus and the solid
lines are based on the fitted soft models (Eqs. 27 and 28). It is seen that there is no effect of CO2 loading
on the second order kinetic rate constant for 1M MEA over the whole loading range 0 to 0.4 mole
CO2/mole MEA. Also for the 5M MEA solution data the correspondence between model and data is
good for the WWC experiments. However, the SDC data show quite a bit of scatter and for high
loadings there is a growing discrepancy between the experimental data and the model. The reason is the
same as for the GK and gk ′ values as the low driving forces at high CO2 loading will give low mass
transfer rate and higher experimental uncertainty in the mass transfer rate determination. Also, when
d
t
w
t
d
s
D
a
s
o
f
b
a
determining
the Henry’s
Figu
Figur
4.3 Conc
According
when presen
to CO2 and t
developed b
simple appro
Donnellan 1
apparent rate
It should b
soft kinetic
objective fun
fitted by line
both CO2 un
as:
k2 from gk ′
law constant
re 6 Prediction o
re 7 Prediction o
centration-ba
g to the dire
nt in associat
transfer of a
ased on the
oach where t
989; Luo et
e constant.
be noted that
model, by u
nction is Eq
ear regressio
nloaded (Luo
, the CO2 so
t and k2, and
of second order k
of second order k
ased model b
ect mechanis
tion with ano
a proton. A c
free concen
the free MEA
al. 2012) ca
t the observe
using the sa
q. 30, and th
on. In this ca
o et al. 2012
olubility is n
d thus increa
kinetic rate cons
kinetic rate const
dot: W
by the direct
sm proposed
other base su
concentration
ntration of M
A is calculat
an then be fi
ed kinetic rat
ame CO2 eq
e optimal va
ase the correl
2) and loade
2.00TMEAk =
needed again
ased uncertai
stant by soft mod
tant by soft mod
WWC; square: SD
t mechanism
d by Crooks
uch as water
n-based mod
MEA and w
ted by Eq. 29
fitted to the e
te constant, k
quilibrium m
alues for all
lations for th
d MEA data
1003 10 exp× ⋅
n, leading to
inty
del with experim
del with experim
DC
m
s (Crooks a
r and amine,
del, Eq. 30,
water. Such a
9. The direct
experimenta
kobs, was obt
model (Eqs
parameters
he kinetic rat
a with tempe
4742T
⎛ ⎞⎟⎜− ⎟⎜ ⎟⎜⎝ ⎠
a quadratic d
mental data in 1M
mental data in 5M
and Donnella
, reacts by si
for the CO2
a model can
t mechanism
al data using
tained in the
25 and 26)
shown in T
te constants
erature depe
dependency
M MEA solution
M MEA solution.
an 1989), th
imultaneous
2 loaded syst
n be obtained
m Eq. 30 (Cro
Eq. 31 to o
e same way a
. However,
Table 1 can b
of MEA and
endency can
14
between
.
he MEA,
bonding
tems was
d from a
ooks and
btain the
(29)
(30)
(31)
as for the
here the
be easily
d H2O in
be fitted
(32)
a
b
c
m
a
r
k
An advant
and thus the
between the
concentration
4.4 Activ
The activit
model. In th
al. (Aronu et
rate is define
Still, using
kinetic const
Table 1 T
T
K
297.1
306.7
314.2
326.9
338.4
tage of the c
model can b
soft model
n approach.
vity-based m
ty based mo
his case the e
t al. 2011) to
ed as in (Knu
g the pseudo
tant can be e
The parameters f
concentratio
be used rega
l and conce
model by the
del needs bo
equilibrium m
o calculate a
uutila 2009)
o first order
expressed as:
24.14T
H Ok =
fitted to the conc
[TMEAk M
6m k⋅
2325
3854
5799
9294
17121
on based mo
ardless of M
ntration app
direct mech
oth an activit
model used
ll concentrat
:
assumption
:
E Hγ
647 10 exp× ⋅
centration based
] TMEAk
MEA σ±
2 1kmol s− −⋅
5.00±363.26
4.21±605.72
9.52±842.97
4.66±1410.02
1.15±1504.3
odel is that t
EA concentr
proaches, is
hanism
ty based equ
was the ext
tions and act
the enhance
0obs
l
kHa
kγ
⋅=
3110T
⎛ ⎞− ⎟⎜ ⎟⎜ ⎟⎜⎝ ⎠
d kinetic model b
A
6
2
7
2
32
the MEA co
ration. Rega
very small,
uilibrium mo
tended UNIQ
tivity coeffic
ement factor
2
0COD
by direct mechan
[2 2
TH Ok H O
6m kmo⋅
122.9
156.2
203.7
315.2
423.0
oncentration
arding compl
, but the us
odel and an a
QUAC mode
cients. The a
r and the act
nism
]2
TH Ok
O σ±
2 1ol s− −⋅
7±24.87
4±35.91
2±50.21
7±81.33
6±91.02
is inside th
lexity, the d
se is easier
activity base
el given by A
activity base
tivity based o
15
(33)
he model,
difference
with the
ed kinetic
Aronu et
ed kinetic
(34)
observed
(35)
b
M
(
d
I
O
Then, base
Here the o
be easily fitt
MEA and H
(Luo et al. 20
4.5 Comp
Figure 8a
data in this w
It gave 27.5%
One has to
ed on the dir
objective fun
ed by linear
2O with tem
012) and loa
Table
T
K
297.1
306.7
314.2
326.9
338.4
parison of m
show a com
work, by usi
%, 27.1% an
note that se
rect mechani
nction is Eq.
regression.
mperature dep
aded data as:
2 The paramete
models
mparison of
ing the soft m
nd 23.8% av
everal under
k
ism,
37, and the
So that the c
pendency in
:
1.84MEAk γ =
22.06H Ok γ =
ers fitted to the a
[MEAk Mγ
6m k⋅
17197
31042
36847
59852
10056
the predicte
model, the c
verage absolu
lying model
2
obsobs
CO
kk γ
γ=
optimal val
correlation o
n loaded solu
1044 10 exp× ⋅
564 10 exp× ⋅
activity based kin
]MEAk
MEA γσ±
2 1kmol s− −⋅
7.83±1970.0
2.52±3313.6
7.86±3544.8
2.34±4788.0
68.17±8872.9
ed apparent
concentration
ute deviation
ls were used
lues for all p
of the activity
ution can be
4112T
⎛ ⎞− ⎟⎜ ⎟⎜ ⎟⎜⎝ ⎠
1766T
⎛ ⎞− ⎟⎜ ⎟⎜ ⎟⎜⎝ ⎠
netic model by d
A
03
61
82
09
94
kinetic rate
n-based mod
ns (AAD) fo
d in this fitt
parameters s
y based kine
fitted to the
direct mechanism
[2 2H Ok H Oγ
6m kmo⋅
545.9
638.8
759.9
928.17
1118.1
e constant w
del and the a
or the three
ting, as give
hown in Tab
etic rate cons
e combined u
m
]2H Ok
O γσ±
2 1ol s− −⋅
1±42.79
9±73.08
0±78.59
7±108.08
2±125.68
with the expe
activity-base
models resp
en in section
16
(36)
(37)
ble 2 can
stants for
unloaded
(38)
(39)
erimental
d model.
pectively.
n 4.2. In
17
particular the solubility model is important as the Henry’s law coefficient enters in the 2nd power in the
equations. However, the AAD values are higher than for unloaded solutions alone, see Luo et al. (Luo et
al. 2012). Comparing the three models no large difference between the predictions can be seen based on
the AAD’s.
The apparent rate constants can be used to back-calculate the rate of mass transfer for the individual
experiments through just reversing the equations. Figure 8b shows the comparison of the predicted CO2
mass transfer rates with experimental data by using all three models. The back calculation of CO2 mass
transfer rate follows the interpretations shown in earlier equations. The parity plot gives 11.4%, 11.1%
and 10% average absolute deviations (AAD) for the three models. As is seen from Figure 8b, now all
three models behave equally well and no tendency of under- or over-prediction is visible. This point to
the apparent under-predictions in the apparent rate constant by the soft model being due to uncertainties
in the underlying models and would not affect the capability to predict mass transfer fluxes. The
differences in AAD between the models is viewed insignificant difference indication that they all three
seem to be equal in predictive capability.
Figure 8 a: Parity plot between experimental data and model predicted apparent kinetic rate constant for Soft model, Concentration-based
model and Activity-based model.
b: Parity plot between experimental data and back calculated model predictions for CO2 mass transfer rate for Soft model, Concentration-
based model and Activity-based model.
In Figure 9 a, b and c are shown the ratios between predicted and measured mass transfer flux as
function of CO2 loading, temperature and CO2 driving force respectively, for the three models. From
Figure 9a it is clearly seen that the difference between the models becomes larger when the CO2 loading
increases. At the two lowest loadings very little difference is seen, whereas above a loading of 0.3 the
soft model starts over-predicting the flux and the concentration based model tends to under-predict the
flux. The activity based model shows a larger scatter at the higher loadings but no systematic under- or
over-prediction can be seen. This can be taken as an indication of a better reliability of the activity based
model at higher loadings. In Figure 9b, the same ratio is given as function of temperature. Here the
18
concentration based model shows a tendency for under prediction, and when Figure 9a is taken into
account, the under-prediction is associated mainly with the high loadings, and then over the whole
temperature range. No specific temperature trend is visible for either of the models. In Figure 9c it can
be seen that the same is true for driving force.
Figure 9 The ratio of predicted and measured mass transfer flux as function of (a) CO2 loading; (b) temperature; (c) CO2 driving force
It can be concluded that the three tested models have almost the same capacity to predict the kinetic
rate constant and CO2 mass transfer rates in a reaction process at low loadings. However, at loadings of
industrial importance it is clear that the activity based model is superior; in spite of the three models
showing very similar AAD values. The soft model presents a very simple approach which only has
temperature dependency. It is limited to the two concentrations 1M and 5M MEA and one need to use
the soft equilibrium model predicting CO2 partial pressure as function of loading and temperature. It can
thus be used in absorber calculations, but the variations in MEA concentration that are bound to appear
in operation are not taken into account and some over-predictions at high loadings are expected.
The concentration based direct mechanism model contains the theoretical direct reaction mechanism
to describe the reaction process and can be used for the whole range of MEA concentrations, CO2
loadings and temperatures. It should be used together with the simplified approach for calculating free
amine concentrations given by Eq. 25 and the soft model for equilibrium pressure of CO2, as given by
Eqs. 25 and 26. Variations in MEA concentration are accounted for but under-predictions at high
loadings are expected.
The activity-based direct mechanism model will give predictions similar to the “soft” and
concentration based models at low loadings, but is better at higher loadings. The activity-based model is
in line with experimental speciation data and is able to describe the real kinetics based on the definition
of activities of species in the solution mixture and should be more accurate when extrapolating beyond
the experimental data.
e
m
e
d
a
c
t
m
e
n
u
b
C
All the ra
experiments
materials.
5. Penet
In order t
evaluation o
developed a
absorption r
concentration
According
transfer com
The sourc
model (Eq. 1
Transport
equilibrium
numerically
uniquely, on
boundary co
CO2 loading
aw data of C
in wetted w
tration mode
to have a p
f the pseudo
and validated
rates, can be
n profiles in
g to the pene
mbined with c
e term rCO2
14) or the act
equations fo
and electro-
because an
ne initial co
ondition in li
and tempera
CO2 absorbe
wall column
el for mass t
possibility to
o first order
d by the ob
e the basis f
n the liquid p
etration theo
chemical rea
can be calc
tivity-based
for [OH-] an
-neutrality. T
analytical so
ondition and
iquid bulk it
ature:
ed by 1M a
n and string
transfer
o describe
assumption,
tained expe
for an enha
phase reactio
ory (Higbie
actions yield
culated by e
model (Eq.
d [H3O+] w
The four non
olution meth
d two bound
is assumed
and 5M CO
of discs ap
the mass tr
a numerica
rimental dat
ancement fac
on film.
1935) the m
the followin
either the so
37) in forme
ere not form
n-linear part
hod is not av
dary conditio
that the sys
O2 loaded aq
pparatus can
ransfer more
l model base
ta. The pen
ctor model,
material bala
ng set of equ
oft model (E
er sections.
mulated as th
tial different
vailable. Bef
ons are nece
tem conside
queous MEA
be found in
e accurately
ed on the pe
etration mod
and can sh
ances for ea
uations for th
Eq. 13), the
hese will fo
tial equation
fore solving
essary. As i
ered is in equ
A solution f
n the supple
y, and to en
enetration the
del can pred
ow how the
ach species
he MEA case
concentratio
ollow from t
ns need to b
this set of e
initial condi
uilibrium for
19
from the
ementary
nable an
eory was
dict CO2
e species
for mass
e:
(40)
on-based
the water
be solved
equations
ition and
r a given
v
b
n
k
p
N
n
o
t
T
It is assum
volatile. It i
boundary co
When the
needs to be c
kinetic analy
pressures an
NRTL mode
necessary to
one of the m
In wetted w
the following
The respectiv
med that all
is also assu
nditions at th
amine soluti
considered. T
ysis. In the
d activity co
el (Hessen
point out th
mentioned mo
wall column
g equations r
ve contact ti
species but
umed that th
he interphas
ion is alread
Therefore, th
e CO2-MEA
oefficients c
et al. 2010
hat no signifi
odels.
n and string o
respectively
WWCδ
imes can be
t water and
he CO2 mas
se are as follo
dy pre-loaded
he liquid bul
A-H2O system
an be obtain
0), or the ex
icant differen
of discs cont
y (Bird et al.
33 L
WCL
QW gμρ
=
calculated a
CO2, includ
ss transfer r
ow:
d with CO2,
lk concentra
m, the conc
n by from an
xtended UN
nce in the fin
tactor, the th
2001):
SDandg
δ
s:
ding the sol
rate is only
the bulk equ
ations of all c
centrations
n activity ba
NIQUAC mo
nal predicted
hickness of th
332
LDC
L
Qd gμρ
=
lute and ion
governed b
uilibrium of
chemical spe
of all the s
sed model s
odel (Aronu
d results was
he liquid film
Qg
nic species,
by diffusion
f all chemica
ecies are req
species, CO
uch as the r
u et al. 201
s found by u
m can be obt
20
(41)
are non-
n, so the
(42)
al species
quired for
O2 partial
efined e-
11). It is
using any
tained by
(43)
21
1 12 23 33 33 33 3
2l l l l
WWC SDCl l
g Q g Qh and dW d
ρ ρτ τ
μ μ
− −− −⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞⎟ ⎟⎜ ⎜⎟ ⎟⎜ ⎜⎟ ⎟= =⎜ ⎜⎟ ⎟⎜ ⎜⎟ ⎟⎟ ⎟⎜ ⎜⎜ ⎜⎟ ⎟⎜ ⎜⎝ ⎠ ⎝ ⎠⎝ ⎠ ⎝ ⎠ (44)
This is a stiff problem as the CO2 concentrations fall rapidly close to the gas liquid interphase and it
was found that the finite differences (FDM) scheme, in general, is not a proper algorithm in the case of
high reaction rates as an excessive number of discretization points were needed. Both a non-uniform
adaptive grid routine (Schiesser 1991), and a method based on orthogonal collocation (Villadsen and
Stewart 1967) were found accurate and robust, and give relatively fast solution. In this work the
orthogonal collocation method was used in MATLAB in order to combine fast and accurate
computations (Luo et al. 2011).
In the development of kinetic rate constants, the pseudo 1st order assumption was used. This
assumption can be tested with the penetration model. Figure 10 and Figure 11 show dimensionless
concentration profiles for four main components (free CO2, free MEA, MEAH+ and MEACOO-)
respectively in the liquid phase for different conditions, but with the same CO2 driving force of 1kPa.
The cases are 5M and 1M MEA at CO2 loadings of 0 and 0.4 mole CO2/mole amine, respectively. It can
be seen that the free MEA concentration drops in the liquid phase reaction zone are less than 5% for all
these cases with a low CO2 driving force. This indicates that the reactions are still in the pseudo first
order reaction regime so that the kinetic models based on the pseudo first order assumption are
reasonable in these cases, and that the kinetic model predictions will not be significantly affected by the
small decrease in MEA concentration.
Figure 10 Dimensionless concentration profiles in liquid phase, 5M MEA, α=0, ∆P = 1kPa, 298 K
Figure 11 Dimensionless concentration profiles in liquid phase, 1M MEA, α=0.4, ∆P = 1kPa, 298 K
However, when the CO2 driving force in the systems above is increased to 15 kPa, the situation
becomes more complicated. As shown in Figure 12, for zero loading in 5M MEA, still the pseudo 1st
order assumption holds. In Figure 13, for 1M MEA and loading 0.4 mole CO2/mole amine, the MEA
concentration drops about 40%. This indicates that the reaction is not in the pseudo first order reaction
regime any more. The developed kinetic rate constants will thus probably be affected by the changes of
22
MEA concentrations at the interface for the low MEA concentrations and high loadings. A new fit of
kinetic parameters based on the penetration theory might be necessary for this case.
Figure 12 Dimensionless concentration profiles in liquid phase, 5M MEA, α=0, ∆P = 15kPa
Figure 13 Dimensionless concentration profiles in liquid phase, 1M MEA, α=0.4, ∆P = 15kPa
Figure 14 shows a comparison between fluxes calculated by the penetration model and the
experimental data. Deviations between experimental values and penetration model predictions, using
parameters from the soft model, concentration-based model and the activity-based models respectively
were found to be 12.5%, 13.7% and 12.1%. This can be compared with the AADs found from the direct
back-calculation using the respective models of 11.4%, 11.1% and 10% average absolute deviations. It
can thus be seen that although the reactions are not strictly in the pseudo first order reaction regime, the
kinetic models based on this assumption can still be used in the penetration model to predict the CO2
mass transfer rate with good accuracy.
Figure 14 Comparison of penetration model predicted CO2 mass transfer rate with experimental data by using three different kinetics
models.
However, the penetration model is still required in order to give more accurate predictions in the cases
where the reactions are far from the pseudo first order reaction region as when the free MEA
concentration is very low and the CO2 driving force is high.
Figure 15 Dimensionless concentration profiles in liquid phase, 0.1M MEA, α=0, ∆P = 100kPa in CO2-MEA-H2O system
The problems with using the pseudo first order assumption become very clear when regarding Figure
15 where an assumed case at low MEA concentration (e.g. 0.1M) and high CO2 driving force (e.g. 100
kPa) is tested. It can be seen that there is an almost total depletion of free MEA when approaching the
interphase. There is also a steep gradient in carbamate as well as free CO2 under these conditions
showing that much of the CO2 is moving across the liquid reaction film in the form of carbamate. The
reaction zone is in fact reduced in width and the penetration model is required to give a more accurate
prediction in these cases where the reactions are not in the pseudo first order reaction region. This fact
can explain most of the literature data showing a k2 concentration dependency for low MEA
23
concentrations like 0.02-0.1 M. In these experiments the concentrations of MEA fall toward the
interface, and the departure from pseudo 1st order conditions increases with decreasing MEA
concentration (Luo et al. 2012). In the papers using such low MEA concentrations the reactions were
still assumed to be in the pseudo first order reaction regime during the interpretation of experimental
data and these data should therefore be disregarded when evaluating MEA kinetics.
Figure 16 Comparison of predicted CO2 mass transfer rate via CO2 driving force by using penetration-concentration based model and
pseudo first order-concentration based model
Figure 16 shows the effect of CO2 driving force on the CO2 mass transfer rates predicted respectively
by the pseudo 1st order model and the penetration model. All parameters in the two mass transfer
models are the same, and the case is for CO2 absorbed into unloaded 5M MEA solution at 25ºC, and
with varying CO2 driving force from 1 – 100 kPa. The predicted CO2 flux is clearly proportional to the
CO2 driving force when using the pseudo 1st order assumption, the blue line in the Figure 16, while the
penetration model predicts a decrease with increasing driving force. The difference between the two
models is insignificant when the CO2 driving force is lower than approximately 15 kPa, but becomes
larger with LMPD increasing. It indicates that the depletion of free MEA at the interface is not
significant, as seen in Figure 10, Figure 11 and Figure 12, and that the reactions are still within the
pseudo 1st order region. With LMPD increasing, the depletion of free MEA at the interface becomes
larger and does effect the CO2 mass transfer calculation. It can be concluded that there exists a
limitation for the CO2 driving force for each concentration of amine in order to control the chemical
reaction being within the pseudo 1st order regime. Otherwise, the interpretation which is based on the
pseudo 1st order assumption will not be valid.
Figure 17 Comparison of predicted CO2 mass transfer rate via CO2 loading by using penetration-concentration based model and pseudo
first order-concentration based model
Figure 17 shows the model predicted CO2 flux in the whole CO2 loading range by using the
penetration and pseudo first order models, respectively. All the parameters in the two mass transfer
models are the same, and the cases are CO2 absorbed into 5M MEA and 1M MEA solutions at 25ºC and
with a CO2 driving force of 15 kPa. The kinetic model used was the concentration-based direct kinetic
24
model (Eq. 29 and Eq. 30). The results show the reasonable trend that the CO2 mass transfer rate
decreases when the CO2 loading increasing, since the free MEA in the solution is lowered and the CO2
capture ability becomes smaller and smaller. It can be seen that the two model predictions give
insignificant differences in the 5M MEA case and this behavior can be used to explain the reason of the
two parity plots (Figure 8b and Figure 14) were the models are shown to be similar to each other.
However, with decreasing MEA concentration, e.g. in 1M MEA case, the difference becomes larger
between the two model predictions. It is reasonable that the limitation in CO2 driving force should be
reduced with amine concentration decreasing.
Figure 18 Effect of temperature on the dimensionless interfacial free MEA concentration in a 5M MEA solution as function of temperature,
loading and CO2 driving force
dot: LMPD=1 kPa; square: LMPD = 5 kPa; circle: LMPD = 10 kPa; triangle: LMPD = 15 kPa
Figure 18 shows the penetration model predictions for the dimensionless concentration of MEA at
gas-liquid interface for different CO2 loadings, temperatures and driving forces. Only 0 and 0.4 CO2
loadings points were used in this figure in order to make it clear. It can be seen that there are several
nearly linear trend lines at given liquid phase CO2 loading and CO2 driving force. This is in line with the
conclusions of Figure 16 above. This indicates that the effect of amine depletion at the interphase will
be more severe at higher than at lower temperatures. The reason for this is the stronger temperature
effects on kinetics that on diffusivity. However, for 5M MEA this depletion is still below 10% of 5M
MEA solution under the operation conditions in this study. However, considering the conclusions above,
it can be predicted that when the CO2 driving force, CO2 loading and temperature are high enough, the
depletion of MEA will become so significant that the pseudo 1st order assumption and the simplified
kinetic model will not be valid anymore.
6. Conclusions
25
Carbon dioxide absorption into aqueous solutions of 1 and 5 M monoethanolamine (MEA) with CO2
loading were experimentally studied over the temperature range from 298 to 343 K, a CO2 loading
range from 0 to 0.4 mole CO2/ mole MEA in a wetted wall column reactor (WWC) and string of discs
contactor (SDC). 227 new data points were produced and are used for interpretation. Comparisons of
liquid film mass transfer coefficient and overall mass transfer coefficient were made between recent
literature data and data from this study and show them to be consistent with each other.
In order to analyze numerically the absorption rate, three types of kinetic models (Soft model;
Concentration-based model; Activity-based model) and a penetration model were developed and
compared. Results showed reasonable agreement between the kinetic models at low loadings. At
loadings above 0.3 mole CO2/mole MEA the activity based model is significantly better than the other
two, but the latter are easier to implement.
A penetration type mass transfer model was developed for the CO2-MEA-H2O system in this study,
and gives not only concentration profiles in the liquid phase, but also was able to predict the mass
transfer rates with a statistical uncertainty less than 12% average absolute deviation to experimental data
even when kinetic constants based on the pseudo first order assumption were used. It was found that the
effect of depletion of free amine at the gas-liquid interface on the kinetics and mass transfer calculations
is insignificant at high amine concentrations, low CO2 loadings, low CO2 driving forces and low
temperature. It is thus still proper to use the pseudo 1st order assumption in these cases. However, the
effect does become significant when either reducing amine concentration or increasing CO2 driving
force or loading. There is an upper limit for the CO2 driving force for each amine concentrations below
which it can control the chemical reaction under pseudo 1st order regime.
Notation
d diameter of discs in String of discs contactor, m
D diffusivity, m2·s-1
26
E enhancement, -
h height of cylinder in wetted wall column, m
H Henry’s law constant, kmol-1·m3·kPa
k2 second order kinetic rate constant, m3·kmol-1·s-1
kg gas film mass transfer coefficient, kmol·m-2·s-1·kPa-1
kg’ liquid film mass transfer coefficient, kmol·m-2·s-1·kPa-1
kl liquid film mass transfer coefficient, m·s-1
kapp apparent kinetic rate constant, s-1
kH2O third order kinetic rate constant of H2O, m6·kmol-2·s-2
kMEA third order kinetic rate constant of MEA, m6·kmol-2·s-2
kobs observed kinetic rate constant, s-1
kOH- kinetic rate constant of CO2 reacts with OH-, m3·kmol-1·s-1
K dissociation constants, -
KG overall mass transfer coefficient, m·s-1 or kmol·m-2·s-1·kPa-1
M molarity, kmol·m-3
N mole flux, kmol·m-2·s-1
P partial pressure, kPa
Q volumetric flowrate, m3·h-1
r reaction rate, kmol·m-3·s-1
t temperature, ̊C
T temperature, K
u linear velocity of liquid, m·s-1
W circumference of cylinder in wetted wall column, m
Greek letter
27
α CO2 loading, mole CO2/ mole MEA
μ viscosity, kg·m-1·s-1
ρ density, kg·m-3
τ contact time, sec
δ liquid thickness, m
γ activity coefficient
Subscripts
0 initial
app apparent
bulk in the bulk
CO2 carbon dioxide
g gas phase
i at interface
H2O water
l liquid phase
MEA monoethanolamine
obs observed
Superscripts
∞ infinite
0 physical absorption
* in equilibrium
in inlet of wetted wall column
28
out outlet of wetted wall column
Solution in solution
T termolecular mechanism
29
References
Aboudheir, A. (2002). Kinetics, modeling, and simulation of carbon dioxide absorption into highly concentrated and loaded monoethanolamine solutions. Regina, Canada, University of Regina. Ph.D. Thesis. Aboudheir, A., P. Tontiwachwuthikul, et al. (2003). "Kinetics of the reactive absorption of carbon dioxide in high CO2-loaded, concentrated aqueous monoethanolamine solutions." Chemical Engineering Science 58(23-24): 5195-5210. Akanksha, K. K. Pant, et al. (2007). "Carbon dioxide absorption into monoethanolamine in a continuous film contactor." Chemical Engineering Journal 133(1–3): 229-237. Aronu, U. E., S. Gondal, et al. (2011). "Solubility of CO2 in 15, 30, 45 and 60 mass% MEA from 40 to 120°C and model representation using the extended UNIQUAC framework." Chemical Engineering Science 66(24): 6393-6406. Bird, R. B., W. E. Stewart, et al. (2001). Transport Phenomena, Wiley International edition. Blauwhoff, P. M. M., G. F. Versteeg, et al. (1984). "A study on the reaction between CO2 and alkanolamines in aqueous solutions." Chemical Engineering Science 39(2): 207-225. Browning, G. J. and R. H. Weiland (1994). "Physical Solubility of Carbon Dioxide in Aqueous Alkanolamines via Nitrous Oxide Analogy." Journal of Chemical & Engineering Data 39(4): 817-822. Caplow, M. (1968). "Kinetics of carbamate formation and breakdown." Journal of the American Chemical Society 90(24): 6795-6803. Crooks, J. E. and J. P. Donnellan (1989). "Kinetics and mechanism of the reaction between carbon dioxide and amines in aqueous solution." Journal of the Chemical Society, Perkin Transactions 2(4): 331-333. da Silva, E. F. and H. F. Svendsen (2004). "Ab Initio Study of the Reaction of Carbamate Formation from CO2 and Alkanolamines." Industrial & Engineering Chemistry Research 43(13): 3413-3418. Danckwerts, P. V. (1970). Gas Liquid Reactions. New York, McGraw-Hill. Danckwerts, P. V. (1979). "The reaction of CO2 with ethanolamines." Chemical Engineering Science 34(4): 443-446. Dang, H. and G. T. Rochelle (2003). "CO2 Absorption Rate and Solubility in Monoethanolamine/Piperazine/Water." Separation Science and Technology 38(2): 337-357. Dugas, R. E. and G. T. Rochelle (2011). "CO2 Absorption Rate into Concentrated Aqueous Monoethanolamine and Piperazine." Journal of Chemical & Engineering Data 56(5): 2187-2195. Hartono, A. (2009). Characterization of diethylenetriamine (DETA) as absorbent for Carbon Dioxide. Department of Chemical Engineering. Trondheim, Norwegian University of Science and Technology. Ph.D. Thesis.
30
Hartono, A., E. F. da Silva, et al. (2009). "Kinetics of carbon dioxide absorption in aqueous solution of diethylenetriamine (DETA)." Chemical Engineering Science 64(14): 3205-3213. Hartono, A., O. M. Emmanuel and Svendsen H.F., (2014). "Physical properties of partially CO2 loaded Monoethanolamine (MEA)." Journal Chemical and Engineering Data , Journal of chemical and Engineering data, 59(6), 1808-1816. Hessen, E. T., T. Haug-Warberg, et al. (2010). "The refined e-NRTL model applied to CO2–H2O–alkanolamine systems." Chemical Engineering Science 65(11): 3638-3648. Higbie, R. (1935). "The rate of absorption of a pure gas into a still liquid." Trans. Amer. Inst. Chem. Engrs. 35: 36-60. Knuutila, H. (2009). Carbon dioxide capture with carbonate systems. Department of Chemical Engineering. Trondheim, Norway, Norwegian University of Science and Technology. Ph.D Thesis. Levenspiel, O. (1999). Chemical reaction engineering. New York, Wiley. Littel, R. J., G. F. Versteeg, et al. (1992). "Kinetics of CO2 with primary and secondary amines in aqueous solutions—II. Influence of temperature on zwitterion formation and deprotonation rates." Chemical Engineering Science 47(8): 2037-2045. Luo, X., A. Hartono, et al. (2011). "The study of numerical methods and validation of a heat and mass transfer model in CO2 -MEA system." Energy Procedia 4(0): 1435-1442. Luo, X., A. Hartono, et al. (2012). "Comparative kinetics of carbon dioxide absorption in unloaded aqueous monoethanolamine solutions using wetted wall and string of discs columns." Chemical Engineering Science 82: 31-43. Ma'mun, S., J. P. Jakobsen, et al. (2005). "Experimental and Modeling Study of the Solubility of Carbon Dioxide in Aqueous 30 Mass % 2-((2-Aminoethyl)amino)ethanol Solution." Industrial & Engineering Chemistry Research 45(8): 2505-2512. Pohorecki, R. and W. d. w. Moniuk (1988). "Kinetics of reaction between carbon dioxide and hydroxyl ions in aqueous electrolyte solutions." Chemical Engineering Science 43(7): 1677-1684. Puxty, G., R. Rowland, et al. (2010). "Comparison of the rate of CO2 absorption into aqueous ammonia and monoethanolamine." Chemical Engineering Science 65(2): 915-922. Schiesser, W. E. (1991). The Numerical Method of Lines. San Diego, Academic Press, Inc. Snijder, E. D., M. J. M. te Riele, et al. (1993). "Diffusion coefficients of several aqueous alkanolamine solutions." Journal of Chemical & Engineering Data 38(3): 475-480. Versteeg, G. F., J. A. M. Kuipers, et al. (1989). "Mass transfer with complex reversible chemical reactions—I. Single reversible chemical reaction." Chemical Engineering Science 44(10): 2295-2310.
31
Versteeg, G. F., L. A. J. Van Dijck, et al. (1996). "On the kinetics between CO2 and alkanolamines both in aqueous and non-aqueous solutions. An overview." Chemical Engineering Communications 144(1): 113-158. Versteeg, G. F. and W. Van Swaalj (1988). "Solubility and diffusivity of acid gases (carbon dioxide, nitrous oxide) in aqueous alkanolamine solutions." Journal of Chemical & Engineering Data 33(1): 29-34. Villadsen, J. V. and W. E. Stewart (1967). "Solution of boundary-value problems by orthogonal collocation." Chemical Engineering Science 22(11): 1483-1501.
32
Figure Caption
CO2 MFC (high)
CO2 MFC (low)
Condensate
Pump
CO2N2
F
Flowmeter
N2 MFC
Fan
F
Rotameter
WWC
CO2 Analyzer
Condensate trap
F
Orifice flowmeter
Circulation Fan
Circulation pump
Solvent tank
Solvent tank
Bleeding
P
T
T
T
TSaturator
Figure 1 Diagram of experimental set-up for wetted wall column
33
Figure 2 Diagram of experimental set-up for string of discs contactor
34
(a)
(b)
Figure 3 The soft equilibrium model fit to experimental VLE data in (a) 5M MEA and (b) 1M MEA solution
0 0.1 0.2 0.3 0.4 0.5 0.6 0.710
-8
10-6
10-4
10-2
100
102
104
106
CO2 loading, mole CO2/mole MEA
Par
tial p
ress
ure
of C
O 2, kP
a
313 K333 K353 K373 K393 K
0 0.2 0.4 0.6 0.8 110
-8
10-6
10-4
10-2
100
102
104
CO2 loading, mole CO2/mole MEA
Par
tial p
ress
ure
of C
O 2, kP
a
313 K333 K
35
Figure 4 kg’ from experimental data (this work) and literature data. Lines are soft model representation
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.510
-4
10-3
10-2
α, mol CO2/ mol MEA
kg, , m
ol⋅ k
Pa-1
⋅ m-2⋅ s
-1
23.8oC33.5oC41.0oC53.7oC65.1oCWWC, this workSDC, this workAboudheir et al.,2003Dang et al.,2003Dugas et al.,2011
36
Figure 5 KG from experimental data (this work) and literature data. Lines are soft model representation
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.510
-4
10-3
10-2
α, mol CO2/ mol MEA
KG
, mol⋅ k
Pa-1
⋅ m-2⋅ s
-1
23.8oC33.5oC41.0oC53.7oC65.1oCWWC, this workSDC,this workAboudheir et al.,2003Puxty et al.,2007
37
Figure 6 Predicted second order kinetic rate constant by soft model compared with experimental data in 1M MEA solution
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.410
0
101
102
α, mol/mol
k 2, m3 /m
ol s
25.5oC33.9oC42.3oC53.9oC64.6oC
38
Figure 7 Predicted second order kinetic rate constant by soft model compared with experimental data in 5M MEA solution.
dots: WWC; squares: SDC
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.4510
0
101
102
103
α, mol/mol
k 2, m3 /m
ol s
23.8oC33.5oC41.0oC53.7oC65.1oC
39
(a)
(b)
Figure 8 a: Parity plot between experimental data and model predicted apparent kinetic rate constant for Soft model, Concentration-based
model and Activity-based model.
b: Parity plot between experimental data and back calculated model predictions for CO2 mass transfer rate for Soft model, Concentration-
based model and Activity-based model.
0 1 2 3 4 5 6
x 105
0
1
2
3
4
5
6x 10
5
Experimental apparent kinetic rate constant, s-1
Pre
dict
ed a
ppar
ent k
inet
ic ra
te c
onst
ant,
s-1
Soft modelConcentration-based modelActivity-based model
10-4 10-3 10-2 10-110-4
10-3
10-2
10-1
Measured CO2 flux, mol.m-2 s-1
Pre
dict
ed C
O2 fl
ux, m
ol.m
-2 s
-1
Soft modelConcentration-based modelActivity-based model
40
(a)
(b)
0 0.1 0.2 0.3 0.4 0.50.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
CO2 loading, mole CO2/mole MEA
Rat
io o
f pre
dict
ed C
O2 fl
ux a
nd m
easu
rem
ents
Soft modelConcentration-based modelActivity-based model
290 300 310 320 330 3400.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
Temperature, K
Rat
io o
f pre
dict
ed C
O2 fl
ux a
nd m
easu
rem
ents
Soft modelConcentration-based modelActivity-based model
41
(c)
Figure 9 The ratio of predicted and measured CO2 flux as function of (a) CO2 loading; (b) temperature; (c) CO2 driving force
0 5 10 150.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
CO2 driving force, kPa
Rat
io o
f pre
dict
ed C
O2 fl
ux a
nd m
easu
rem
ents
Soft modelConcentration-based modelActivity-based model
42
Figure 10 Dimensionless concentration profiles in liquid phase, 5M MEA, α=0, ∆P = 1kPa, 298 K
0 0.2 0.4 0.6 0.8 1
x 10-4
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance from interface, m
Dim
ensi
onle
ss C
once
ntra
tion
CO2
MEA
MEAH+
MEACOO-
43
Figure 11 Dimensionless concentration profiles in liquid phase, 1M MEA, α=0.4, ∆P = 1kPa, 298 K
0 0.2 0.4 0.6 0.8 1
x 10-4
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance from interface, m
Dim
ensi
onle
ss C
once
ntra
tion
CO2
MEA
MEAH+
MEACOO-
44
Figure 12 Dimensionless concentration profiles in liquid phase, 5M MEA, α=0, ∆P = 15kPa, 298 K
0 0.2 0.4 0.6 0.8 1
x 10-4
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance from interface, m
Dim
ensi
onle
ss C
once
ntra
tion
CO2
MEA
MEAH+
MEACOO-
45
Figure 13 Dimensionless concentration profiles in liquid phase, 1M MEA, α=0.4, ∆P = 15kPa, 298 K
0 0.2 0.4 0.6 0.8 1
x 10-4
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance from interface, m
Dim
ensi
onle
ss C
once
ntra
tion
CO2
MEA
MEAH+
MEACOO-
46
Figure 14 Comparison of penetration model predicted CO2 mass transfer rates with experimental data by using three different kinetics
models.
10-4 10-3 10-2 10-110-4
10-3
10-2
10-1
Measured CO2 flux, mol.m-2 s-1
Pre
dict
ed C
O2 fl
ux, m
ol.m
-2 s
-1
Soft modelConcentration-based modelActivity-based model
47
Figure 15 Dimensionless concentration profiles in liquid phase, 0.1M MEA, α=0, ∆P = 100kPa in CO2-MEA-H2O system
0 1 2
x 10-4
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance from interface, m
Dim
ensi
onle
ss C
once
ntra
tion
CO2
MEA
MEAH+
MEACOO-
48
Figure 16 Comparison of predicted CO2 mass transfer rate as function of CO2 driving force by using respectively a penetration-
concentration based model and a pseudo first order-concentration based model
0 20 40 60 80 100 1200
0.05
0.1
0.15
0.2
0.25
0.3
0.35
LMPD, kPa
Pre
dict
ed C
O2 fl
ux, k
mol
/m2 /s
Predicted by pseudo 1st order assumptionPredicted by penetration model
49
Figure 17 Comparison of predicted CO2 mass transfer rate as function of CO2 loading by using respectively a penetration-concentration
based model and pseudo first order-concentration based model
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.410
-3
10-2
10-1
CO2 loading, mol CO2/mol MEA
Pre
dict
ed C
O2 fl
ux, k
mol
/m2 /s
Pseudo first 1st assumption, 5M MEAPenetration model, 5M MEAPseudo first 1st assumption, 1M MEAPenetration model, 1M MEA
50
Figure 18 Effect of temperature on the dimensionless interfacial free MEA concentration in a 5M MEA solution as function of temperature,
loading and CO2 driving force
dot: LMPD=1 kPa; square: LMPD = 5 kPa; circle: LMPD = 10 kPa; triangle: LMPD = 15 kPa
Highlights
1. Kinetics for CO2-MEA-H2O reported for 298-343 K, 1 and 5 M and 0 to 0.4 mole CO2/ mole
MEA.
2. A soft model for the VLE was proposed.
3. Three kinetic models developed and compared; simplified, concentration and activity based.
4. Model validation and analysis performed based on a penetration type mass transfer model.
290 300 310 320 330 3400.93
0.94
0.95
0.96
0.97
0.98
0.99
1
Temperature, K
Dim
ensi
onle
ss C
once
ntra
tion
of M
EA
at i
nter
face
α=0α=0.4
