Available online at www.sciencedirect.

com

Separation and Purification Technology 59 (2008) 85–90

CO2 removal from a gas stream by membrane contactor

Aldo Bottino a , Gustavo Capannelli a , Antonio Comite a , Renzo Di Felice b,∗ , Raffaella Firpo a
a Dipartimento di Chimica e Chimica Industriale, Università degli Studi di Genova, Via Dodecaneso 31, 16146 Genova, Italy
b Dipartimento di Ingegneria Chimica e di Processo “G. Bonino”, Università degli Studi di Genova, Via Opera Pia 15, 16145 Genova, Italy
Received 1 February 2007; received in revised form 28 May 2007; accepted 28 May 2007

Abstract
Gas–liquid membrane contactors were applied to remove carbon dioxide from a gas stream by using an aqueous monoethanolamine (MEA)
solution as absorbent. Various modules composed by different numbers of commercial polypropylene capillary membranes were constructed and
tested in a laboratory-scale plant fed with a N2 –CO2 gas mixture. Attention was especially focused on the CO2 removal efficiency of the different
membrane modules when gas flow rate was increased from 5 up to 360 L/h. A mathematical model was developed to simulate the absorption
process in order to predict gas removal efficiency from the knowledge of the system physical parameters. The overall membrane mass transfer
coefficient kM was determined and used to compare experimental and predicted removal efficiencies. A good agreement between the developed
model and experimental results was found.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Carbon dioxide; Membrane contactor; Mathematical modelling

1. Introduction allurgical process streams [8–11], recovery of sulfur aroma

compounds from food industry wastewaters [12–14], and gas
Gas/liquid and liquid/liquid contacting operations are tradi- absorption [15–17].
tionally carried out by using towers or columns. The main effort More specifically, typical absorbers for removing carbon
in designing and operating these absorbers is to maximize the dioxide include bubble columns, packed towers, venturi scrub-
mass transfer rate by producing as much interfacial area as pos- bers and sieve trays. An aqueous solution of sodium hydroxide,
sible. In the case of packed columns, this is achieved by proper sodium carbonate, monoethanolamine or diethanolamine is
selection of packing material and uniform distribution of fluids often employed as absorbent. When a suitable membrane config-
fed to the packed bed [1,2]. Conventional packed bed absorbers uration such as the capillary one is used, the fluids to be contacted
present several disadvantages such as flooding at high flow rates, flow on the opposite side of the capillary (lumen and shell), and
unloading at low flow rates, and channeling and foaming, that the gas–liquid interface forms at the mouth of each membrane
hinder the mass transfer between gas and liquid. pore. The available contact area remains undisturbed even at
An alternative technology that overcomes these disadvan- a high or low flow rate because the two fluid flows are inde-
tages is a non-dispersive contactor through a microporous pendent from each other. This type of membrane represents a
membrane, combining the advantages of selective absorption highly efficient alternative to conventional packed towers [18],
with membrane modularity, compact equipment and no need of because of the following reasons: no flooding at high flow rates,
a washing section [3]. Membrane contactors have been inves- no unloading at low flow rates, no emulsions, no need for a
tigated extensively since the mid-1980s for a wide range of density difference between fluids.
applications, including racemic leucine separation [4], removal In addition, scaling-up is more straightforward with the mem-
of ethanol from fermentation broth [5,6], extraction of mevino- brane contactor. Membrane operations usually scale linearly, so
linic acid from fermentation broth for pharmaceutical use [7], that a predictable increase in capacity is achieved simply by
extraction of metal ions from industrial waste and hydromet- adding new membrane modules. A modular design also allows
a membrane plant to operate over a wide range of capacities. The
interfacial area is known and constant, allowing the performance
∗ Corresponding author. Tel.: +39 0103532924; fax: +39 0103532586. to be predicted more easily than the conventional absorption
E-mail address: [email protected] (R. Di Felice). system like packed columns, where the interfacial area per unit

1383-5866/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.seppur.2007.05.030
86 A. Bottino et al. / Separation and Purification Technology 59 (2008) 85–90

2. Mathematical modelling of the membrane contactor

Nomenclature
By considering a small cylindrical tube with porous wall, for
C concentration (kmol/m3 )
fully developed laminar flow and in steady-state condition, when
d mean pore size (m)
no reaction takes place in the fluid phase the mass balance for
Da Damkholer number, defined by Eq. (3)
the generic component A can be written in dimensionless term:
DAB molecular diffusion coefficient (m2 /s)
Knudsen diffusion coefficient (m2 /s)  2 ∗ ∗
DK ∂C∗ ∂ CA 1 ∂CA
k numerical parameter, defined by Eq. (8) ϕ ∗A = α + (1)
∂z ∂r ∗2 r∗ ∂r ∗
kG gas phase mass transfer coefficient (m/s)
kM membrane mass transfer coefficient (m/s) where
L capillary length (m) CA z r πR2 u
∗
n number of capillaries CA = 0
, z∗ = , r∗ = , ϕ= and
CA L R v0
R capillary radius (m)
s membrane thickness (m) πDAB L
Sh Sherwood number α=
v0
v0 volumetric flow rate (m3 /s) The boundary condition at the wall can be expressed by
z axial distance (m) equating the flux of the component A at the fluid–membrane
interface to that through the membrane. When concentration of
Greek letters component A outside the membrane is negligible, its flux in the
α dimensionless parameter membrane can be expressed, introducing an overall mass trans-
ε dimensionless membrane porosity fer coefficient kM , by the product between this coefficient and
η removal efficiency, defined by Eq. (10) gas concentration at the gas–membrane interface, so that we
τ dimensionless membrane tortuosity have (again in dimensionless terms)
ϕ dimensionless velocity
∂CA∗
∗
= −DaCA,w at r∗ = 1 (2)
∂r ∗
volume may be known, but it is often difficult to determine the where Da is the Damkholer number, defined, for our particular
loading, which represents the fraction of the available interfacial system:
area actually used [19]. kM R
The successful use of the membrane contactor process Da = (3)
DAB
over the conventional absorption processes will largely depend
on the gas–liquid system, the types of membranes used and Eq. (1) represents a partial differential equation that can be
the operating conditions. Many authors developed mathemat- solved numerically to yield concentration distribution function
ical models using the resistance in series concept to predict of axial and radial position.
overall membrane contactor performance, combining process However, it is much easier to tackle the same problem by
conditions, membrane properties and module geometric char- using a one-dimensional approach as suggested, for example,
acteristics. Among them, Mavroudi et al. [20] simulated the by Tronconi et al. [23]. By introducing the average component
CO2 removal in various absorbents, focusing on the possibility concentration over the tube section as

of partial membrane-wetting; Wang et al. [21] investigated the
∗ ϕC∗ dS
CO2 absorption flux by employing three typical amine solutions, CA,m =  A (4)
ϕ dS
flowing through the membrane hollow fiber lumen.
In a previous work [22] we reported the results of screening we can write the mass balance as
tests aiming at selecting the more appropriate membrane mate- ∗
dCA,m ∗ ∗
rial and configuration for an efficient carbon dioxide removal = −2α Sh(CA,m − CA,w ) (5)
dz∗
from flue gases.
In the present study the results aimed at assessing the per- with
formance of capillary membrane previously selected for carbon kG R
dioxide removal from a CO2 –N2 gas mixture, as a function of Sh =
DAB
module design (i.e., number of capillaries assembled in a given
module volume) and operating parameters (i.e., gas flow rate) are The condition at the boundary wall is this time:
presented. A mathematical one-dimensional model was devel- ∗ ∗ ∗
Sh(CA,m − CA,w ) = DaCA,w (6)
oped to relate the membrane contactor efficiency to the mass
transfer coefficient. The experimental parameters obtained for As demonstrated by Tronconi et al. [23], this simpler one-
each membrane module were compared with the mathematical dimensional approach is practically numerically equivalent to
model, so that an agreement between theory and experiments the more complex two-dimensional approach if a correct value
could be found. for the Sh is utilised: they suggested that for the whole tube
 A. Bottino et al. / Separation and Purification Technology 59 (2008) 85–90 87

∗ /C ∗
Fig. 1. CA,w A,m as a function of Damkholer number Da. Fig. 2. Expected removal efficiency at various geometrical and transport prop-
erties of the membrane.

length, with the exception of a small portion at the entrance, Sh removal efficiency of the membrane as
is constant and equal to 1.85.
∗
CA,in ∗
− CA,out
The magnitude of the Damkholer number Da (relative to the 
∗ : η= ∗ = 1 − e−k (10)
Sherwood number, Sh = 1.85) will determine the value of CA,w CA,in
∗
it will approach CA,w for small values of Da (i.e., all the resis-
tance to the mass transfer is concentrated on the membrane) and In practice, once the system has been determined, only gas
will approach zero for very large values of Da (i.e., the resistance flow rates can be adjusted in order to optimise performances.
is concentrated on the gas phase) as depicted also in Fig. 1. The efficiency of the system can then be obtained from:
 /ν
The knowledge of Da is therefore of fundamental impor- η = 1 − e−k 0
tance and requires the estimation of the overall membrane mass
transfer coefficient kM . With the simplifying assumption of a where
homogenous porous membrane kM is given by the ratio between
k = 2πLRkM
the effective component diffusivity in the membrane and the
membrane thickness. As it will be quantified later, this leads to Fig. 2 depicts, for operating values typical of this work (max-
a value of the Da number for our case of the order of 10−1 : imum gas flow rate in a channel equal to 0.0001 m3 /s), expected
from Fig. 1 we can estimate the wall concentration CA,w ∗ to be removal efficiency at various geometrical and transport prop-
∗
practically coincident with the average concentration CA,m and erties of the membrane. As expected, large values of k (i.e.,
Eq. (5) therefore reduces to: good membrane transport properties) favour large efficiencies;
∗
at the same time increasing gas flow rate efficiency declines.
dCA,m ∗ This decline, however, is very sharp in a very specific range of
= −2α Da CA,m = −k CA,m
∗
(7)
dz∗ flow rate which should, consequently, be avoided.
with
3. Experimental
2πLRkM
k = (8) Table 1 lists the main characteristics of the capillary mem-
v0
branes Accurel S6/2 (Membrana, Germany), used in this study.
Simple integration of Eq. (7), between inlet (where CA∗ =1 Fig. 3 shows the scanning electron micrographs of the mem-
by definition) and outlet conditions, yields expression for the brane cross-section (Fig. 3A), external surface (Fig. 3B) and
dimensionless concentration at the exit of the membrane module internal surface (Fig. 3C). The external surface represents the
function of the system physical characteristics: contact area between the gas stream and the absorbent solution.
Four modules were prepared using different numbers of cap-
∗ 
CA,out = e−k (9) illary membranes (1, 3, 10, 18) with the same length of 17 cm, in
order to obtain different contact areas. The capillary membrane
As in this work our specific interest is in the removal of the ends were inserted into small plastic tubes and sealed using an
component A from the gaseous stream, it is useful to define a epoxy resin to form a module.
88 A. Bottino et al. / Separation and Purification Technology 59 (2008) 85–90

Table 1
Main characteristics of Accurel S6/2 capillary membrane supplied by the
manufacturer
Material Polypropylene
Structure Asymmetric
Water wettability Hydrophobic membrane
Nominal pore size (␮m) 0.2
Thickness (␮m)a 400
Inner radius (␮m) 900
Porosity (%) 60
a Measured by SEM micrograph shown in Fig. 3A.

Fig. 4. Glass housing for removable membrane module: (A) glass housing; (B)
O-ring; (C) metal ring; (D) union joint.

Fig. 4 shows the photo of the glass housing (A). The hydraulic
seal between the module and the glass housing was provided
by the O-ring (B), which was installed on the plastic tube outer
surface and squeezed by the metal ring (C) by tightening a nipple
in the female half of the union joint (D). An important feature
of the glass housing was its reusable nature, since at the end
of a given test the module could be simply removed from the
housing by unscrewing the nipple connected to the union joint.
Tests were performed using the laboratory-scale plant, whose
scheme is shown in Fig. 5. A gas mixture (supplied by SIAD,
Italy) containing 15% (v/v) of CO2 and 85% of N2, which is the
typical composition of a flue gas from a coal combustion plant,
was fed to the membrane lumen. The gas flow rate was carefully
regulated by a fine metering valve (5) and measured at the inlet
and outlet of the modules (G1 and G2) by soap bubble flow
meters (F). A monoethanolamine (MEA) 3 M aqueous solution
(V = 800 mL) was used as CO2 absorbent. Its temperature was

Fig. 5. Scheme of the lab-scale plant used for CO2 absorption experiments. (1)
Cryostatic bath; (2) recirculation pump; (3) absorption solution tank; (4) glass
housing and membrane module; (5) fine metering gas valve; F: soap bubble
Fig. 3. SEM micrographs of Accurel S6/2 capillary membrane: (A) cross- flow meter; M: manometer; T: thermometer; L1: absorption solution inlet; L2:
section; (B) external surface; (C) internal surface. absorption solution outlet; G1: gas mixture inlet; G2: gas mixture outlet.
 A. Bottino et al. / Separation and Purification Technology 59 (2008) 85–90 89

kept at 20 ◦ C by a stainless steel coil immersed in the absorbent

solution tank (3) and run through by water cooled by the bath
(1). The liquid phase was fed on the shell side of the membrane
module with a recirculation pump (2) at a flow rate of about
100 L/h. Its pressure was kept at a proper value, in order to avoid
any possible penetration of the solution into the membrane pores
as well as any gas bubble formation in the absorbent solution.
The gas mixture flow rate was increased from 0.14 × 10−5
to 10.0 × 10−5 m3 /s. CO2 removal was estimated calculating
the difference in CO2 concentration in the gas phase between
membrane module inlet and outlet.
The CO2 concentration was measured employing a gas chro-
matograph (mod. Autosystem, Perkin Elmer) equipped with a
packed Porapak Q column and a TCD.

4. Results and discussion

As Fig. 6 shows the percentage of CO2 at the outlet of each

type of membrane module increases by increasing the gas flow
rate, ranging from 0 up to 15% (v/v), i.e. the whole amount Fig. 7. Experimental CO2 removal efficiency as a function of gas flow rate.
of CO2 in gas mixture. The rate of CO2 percentage increase
depends on the number of capillaries: the higher the number of evaluated by a scanning electron micrograph of the membrane
capillaries the lower the CO2 percentage. cross-section, shown in Fig. 3A) is reported in Table 1 along
Fig. 7 shows the removal efficiency as a function of gas flow with the values of porosity (60%) and the nominal pore size
rate for two specific runs carried out with 1 and 10 capillaries (0.2 ␮m) supplied by the manufacturer. In the SEM micrographs
modules, respectively. of Fig. 3B and C the contrast between the external and internal
The behaviour shown in Fig. 7 is in qualitative agreement surfaces is striking. On the external surface pores some tens of
with the theoretical prediction depicted in Fig. 2. nanometers large can be clearly observed, while on the internal
Obviously, it would be of great interest if we could predict gas one much smaller pores are present, whose size is often lower
removal efficiency from the knowledge of the system physical than the one (0.2 ␮m) declared by the manufacturer. Inspection
parameters. The parameter to be determined is the membrane of the cross-section micrograph (Fig. 3A) also reveals a tortuous
overall mass transfer coefficient kM , which is represented by pore structure.
the ratio between the effective diffusion coefficient (Deff ) and On the basis of these experimental evidences and keeping
the membrane thickness (s). This latter value (experimentally into consideration the mean free path of CO2 molecules, it is
possible to consider CO2 diffusion as mainly controlled by a
combined bulk-Knudsen diffusion mechanism.
Knudsen diffusion coefficient DK was determined by using
the well-known correlation [24]:

T
DK = 48.5d (12)
M
where d is the mean pore size, T the operating temperature and
M is the CO2 molecular mass.
Calculations lead to a Knudsen diffusion coefficient of
25.2 × 10−6 m2 /s. A molecular diffusion coefficient taken from
the literature [25] for CO2 bulk diffusion (DAB ), equal to
16.9 × 10−6 m2 /s was also considered. Both the coefficients
were corrected for the effect of porosity, ε = 0.6 (Table 1), and
tortuosity, τ, by using the relations:
εDK εDAB
DK,eff = and DAB,eff = (13)
τ τ
where the tortuosity was calculated through the correlation:
1
Fig. 6. Percentage of CO2 in gas samples taken at the membrane modules outlet, τ= (14)
as a function of gas flow rate. ε2
90 A. Bottino et al. / Separation and Purification Technology 59 (2008) 85–90

References
[1] J.T. Yeh, H.W. Pennline, K.P. Resnik, Study on CO2 absorption and des-
orption in a packed column, Energy Fuels 15 (2001) 274–278.
[2] B. Benadda, K. Kafoufi, P. Monkam, M. Otterbein, Hydrodynamics and
mass transfer phenomena in counter-current packed column at elevated
pressures, Chem. Eng. Sci. 55 (2000) 6251–6257.
[3] P.H.M. Feron, A.E. Jansen, CO2 separation with polyolefin membrane con-
tactors and dedicated absorption liquids: performances and prospects, Sep.
Pur. Technol. 27 (2002) 231–242.
[4] H.B. Ding, P.W. Carr, E.L. Cussler, Racemic leucine separation by hollow-
fiber extraction, AIChE J. 38 (1992) 1493–1498.
[5] M. Matsumura, H. Märkl, Elimination of ethanol inhibition by perstraction,
Biotechnol. Bioeng. 28 (1986) 534–541.
[6] G. Vatai, M.N. Tekic, Membrane-based ethanol extraction with hollow-
fiber module, Sep. Sci. Technol. 26 (1991) 1005–1011.
[7] R. Prasad, K.K. Sirkar, Hollow fiber solvent extraction of pharmaceutical
products: a case study, J. Membr. Sci. 47 (1989) 235–259.
[8] R.-S. Juang, H.-L. Huang, Mechanistic analysis of solvent extraction of
heavy metals in membrane contactors, J. Membr. Sci. 213 (2003) 125–135.
[9] Z.-F. Yang, A.K. Guha, K.K. Sirkar, Novel membrane-based synergistic
metal extraction and recovery processes, Ind. Eng. Chem. Res. 35 (1996)
Fig. 8. ln(1 − ηexp )] as a function of n/v0 compared with theoretical predictions. 1383–1394.
[10] K. Yoshizuka, K. Kondo, F. Nakashio, Effect of interfacial reaction on rates
of extraction and stripping in membrane extractor using a hollow fiber, J.
Chem. Eng. Jpn. 19 (1986) 312–318.
Finally, using the Bonsaquet equation [24]: [11] C.H. Yun, R. Prasad, A.K. Guha, K.K. Sirkar, Hollow fiber solvent extrac-
1 1 1 tion removal of toxic heavy metals from aqueous waste streams, Ind. Eng.
= + (15) Chem. Res. 32 (1993) 1186–1195.
Deff DAB,eff DK,eff [12] F.-X. Pierre, I. Souchon, M. Marin, Recovery of sulfur aroma compounds
using membrane-based solvent extraction, J. Membr. Sci. 187 (2001)
an effective diffusion coefficient in the membrane of
239–253.
1.61 × 10−6 m2 /s and, consequently, an overall mass transfer [13] F.-X. Pierre, I. Souchon, V. Athès-Dutour, M. Marin, Membrane-based
coefficient kM = Deff /s = 4.2 10−3 m/s was obtained. solvent extraction of sulfur aroma compounds: influence of operating
This value lead to a Damkholer number of 0.22 (Eq. conditions on mass transfer coefficients in a hollow fiber contactor, Desali-
(3)), thus confirming the hypothesis put forward in 2. sec- nation 148 (2002) 199–204.
[14] I. Souchon, I. Athès, F.-X. Pierre, M. Marin, Liquid–liquid extraction and
tion of the resistance to the mass transfer concentrated
air stripping in membrane contactor: application to aroma compounds
in the membrane. By using the kM value so determined, recovery, Desalination 163 (2004) 39–46.
removal efficiencies for all the cases investigated in the [15] S. Karoor, K.K. Sirkar, Gas absorption studies in microporous hollow fiber
present study were predicted. By plotting the ln(1 − ηexp ) as membrane modules, Ind. Eng. Chem. Res. 32 (1993) 674–684.
a function of n/v0 (s/m3 ), a good agreement between the pre- [16] Z. Qi, E.L. Cussler, Microporous hollow fibers for gas absorption. I. Mass
transfer in the liquid, J. Membr. Sci. 23 (1985) 321–332.
dicted and experimental data was found, as demonstrated in
[17] Z. Qi, E.L. Cussler, Microporous hollow fibers for gas absorption. II. Mass
Fig. 8. transfer across the membrane, J. Membr. Sci. 23 (1985) 333–345.
[18] A. Gabelman, S.-T. Hwang, Hollow fiber membrane contactors, J. Membr.
5. Conclusions Sci. 159 (1999) 61–106.
[19] R.D. Noble, S.A. Stern, Membrane Separations Technology Principles and
Applications, Elsevier, The Netherlands, 1995.
Contactors with different capillary membranes assembled in [20] M. Mavroudi, S.P. Kaldis, G.P. Sakellaropoulos, Reduction of CO2 emis-
the same module volume have been tested to remove CO2 from sions by a membrane contacting process, Fuel 82 (2003) 2153–2159.
a N2 –CO2 gas mixture. [21] R. Wang, D.F. Li, D.T. Liang, Modeling of CO2 capture by three typical
Laboratory tests showed that the CO2 removal efficiency amine solutions in hollow fiber membrane contactors, Chem. Eng. Proc.
43 (2004) 849–856.
increased increasing the number of capillaries, and decreasing [22] A. Bottino, G. Capannelli, A. Comite, R. Firpo, R. Di Felice, P. Pinacci,
the gas flow rate. Separation of carbon dioxide from flue gases using membrane contactors,
Gas removal efficiency was predicted by developing a Desalination 200 (2006) 609–611.
mathematical one-dimensional model from the knowledge of [23] E. Tronconi, P. Forzatti, J.P. Gomez Martin, S. Mallogi, Selective catalytic
membrane thickness and porosity, assuming permeation through removal of NOx : a mathematical model for design of catalyst and reactor,
Chem. Eng. Sci. 47 (1992) 2401–2406.
membrane pores controlled by a combined bulk-Knudsen diffu- [24] R.H. Perry, D.W. Green, Perry’s Chemical Engineers’ Handbook, McGraw-
sion mechanism. Hill International Editions, 1997.
A good agreement between predicted and experimental [25] R.C. Reid, J.M. Prausnitz, B.E. Poling, The Properties of Gases & Liquids,
results was found. McGraw-Hill International Editions, 1988.