a r t i c l e i n f o a b s t r a c t
Article history: The absorption of CO2 from the gas mixture (CO2 CH4 ) by polyvinylideneuoride (PVDF) hollow ber
Received 4 October 2011 membrane contactor using monoethanolamine (MEA) as the absorbent was performed. The mathemat-
Received in revised form 6 January 2012 ical model has been developed to predict the absorption performance. The model is validated with the
Accepted 31 January 2012
experimental results for estimating the wetting ratio (x*) as the function of liquid velocity and MEA con-
Available online 9 February 2012
centration. The suitable hollow ber membrane module with effective ber length of 50 cm is selected
for the design of multistage membrane contactors. The absorption ux of multistage membrane contac-
Keywords:
tor is simulated based on the value of x* obtained from the experiments. The three-stage cascade design
Carbon dioxide
Cascade design
is selected to compare the system performance with different gas and liquid ow patterns. The results
Membrane contactor of the simulation show that the individual gas ow (G-ID) gives higher performance compared to the
Membrane wetting gas ow in series (G-IS) for all operating conditions studied. The three different ow patterns of liquid
including (i) liquid ow in series (L-IS), (ii) liquid ow in series with splitting (L-ISS) and (iii) liquid ow in
series with recycle (L-ISR) are compared. At low MEA concentration (0.25 M), the L-ISR can improve the
system performance at low liquid velocities, while L-ISS shows the highest performance at high liquid
velocities. For the system with high MEA concentration (1.0 M), L-ISR can improve the performance at
low to moderate liquid velocities, whereas L-ISS does not improve the system performance at any liquid
velocity.
2012 Elsevier B.V. All rights reserved.
0376-7388/$ see front matter 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2012.01.048
176 S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189
where dry is the dry thickness of the membrane, M is membrane FA,i = FA,i+1 JA,i Ai (13)
tortuosity and M is membrane porosity. The gas effective diffusiv-
ity (DG,eff ) is determined by the interactions between the molecules where FA and fA are the molar ow rate of CO2 in gas phase and
(molecular diffusion) as well as the interactions of the molecules liquid phase, respectively. JA,i is absorption ux of CO2 determined
with the pore wall (Knudsen diffusion). The gas effective diffusivity by the following;
is estimated using the following equation;
JA,i = KL,i (CA,i CAL,i ) (14)
1 1 1
= + (9) where KL,i is local overall mass transfer coefcient, CAL,i is the bulk
DG.eff DG,M DG,Kn concentration of CO2 in liquid phase and C* A,i is the CO2 concentra-
tion in liquid phase in equilibrium with the bulk concentration of
where DG,M and DG,Kn are the Fickians molecular diffusion coef-
CO2 in gas phase given by the following equation;
cient and the Knudsen diffusion coefcient of gas, respectively.
The molecular diffusion coefcient of gas (DG,M ) is calculated PAG,i
from the kinetic gas theory; CA,i = (15)
H
0.001858T 3/2 where PAG,i is the partial pressure of CO2 in gas phase estimated as
DG,M = (10) the following;
1/2 2
PMA AB D
PAG,i = yA,i Ptot (16)
where MA is the gas molecular weight, P and T are the gas pressure
and temperature, respectively. AB is characteristic length, the col- where yA,i and Ptot are mole fraction of CO2 in gas phase and the
lision integral D is dimensionless function of temperature which total pressure of gas phase, respectively. The mole fraction of CO2
is calculated by empirical equations (see Appendix A). The Knud- can be estimated from the molar ow rate as;
sen diffusion coefcient (Dg,Kn ) can be calculated by the following
FA,i
equation; yA,i = (17)
FA,i + FI,i
T where FI,i is molar ow rate of inert gas not absorbed by the liquid
DG,Kn = 4850dpore (11)
MA absorbent. Therefore, FI is constant along the ber length and is
equal to the molar ow rate at inlet feed gas (FI0 ).
For the wetted part of the membrane pores, the membrane mass The contact area (Ai ) for each small element is determined
transfer coefcient can be expressed as [15]; from the following:
Di,L M Ai =
(di + x M ) z (18)
kM = (12)
M wetted
where x* is wetting ratio representing the degree of membrane
where wetted is the wetted thickness of the membrane. wetting.
178 S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189
Fig. 2. The schematic diagram showing mass balance of a small element in gasliquid membrane contacting process.
4. Experimental where QG,in and QG,out are the gas ow rates at inlet and outlet,
respectively. CAG,in and CAG,out are the concentrations of CO2 at inlet
4.1. Materials and outlet, respectively. A is mass transfer area. In addition, the
removal efciency dened as % removal was also reported. It is
The polyvinylideneuoride (PVDF) hollow ber membranes estimated by the following equation:
were purchased from Altrateck (China). The specications of the
membrane and membrane module are shown in Table 1. Carbon
QG,in CAG,in QG,out CAG,out
dioxide (CO2 , 99.8 vol.%), methane (CH4 , 99.9 vol.%) were obtained % removal = 100% (22)
QG,in CAG,in
from Thai Industrial Gases PLC. The absorbent used was prepared
S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189 179
5. Results and discussions and bulk liquid is enhanced with increasing liquid velocity, while
increasing MEA concentration resulted in enhanced chemical reac-
5.1. Experimental results tion rate between CO2 and MEA. The similar results were reported
in the literature [3,5,13].
Fig. 4(i) shows of the effect of liquid velocity and MEA concen- The effects of gas velocity and gas composition on absorption
tration on the absorption ux. It was found that the absorption ux are depicted in Fig. 4(ii). It was found that increasing in gas
ux increased moderately with liquid velocity, but the increase velocity did not affect the absorption ux as observed from the con-
was more signicant with increasing MEA concentration. From liq- stant absorption ux. Atchariyawut at el. [16] reported the same
uid velocity 0.10.4 m s1 , the absorption ux increased roughly 21 trend of results for absorption of CO2 from CO2 CH4 gas mixture
and 25%, for MEA concentrations of 0.5 M and 1.0 M, respectively. using PVDF hollow ber, the CO2 ux remained constant when the
The increase in liquid velocity could increase the liquid phase mass gas velocities were varied from 1.0 to 8.0 m s1 . Khaisri et al. [5]
transfer coefcient (Eq. (7)) resulting in the increase of overall mass also found that for physical absorption of CO2 using PTFE mem-
transfer coefcient (KL ). In addition, the driving force of the system brane, the CO2 ux remained constant while the gas velocity was
which is the concentration difference of CO2 at gasliquid interface increased. Meanwhile, the absorption ux increased approximately
(i) 2.5
(ii)
1.2
1.0 M
CO2 flux (x10 mol/m .s)
2 1
2
0.5 M
2
0.25 M 0.8
1.5
-3
-3
0.125 M
0.6
1
0.4
40 %vol. CO2
0.5
0.2 50 %vol. CO2
0 0
0.05 0.15 0.25 0.35 0.45 0.55 0.06 0.07 0.08 0.09 0.1 0.11
Liquid velocity, vl (m/s) Gas velocity, v g (m/s)
Fig. 4. (i) Effect of liquid velocity on absorption ux of CO2 at various MEA concentrations (yCO2 = 0.4, QG = 0.5 l/min) (ii) effect of gas velocity on absorption ux of CO2 at
different CO2 concentrations in gas phase (CMEA = 0.125 M, vl = 0.3 m s1 ).
180 S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189
14.3% when CO2 concentration in gas phase was increased from 5.3. Resistance analysis
40 to 50 vol.% as a result of the enhanced driving force for mass
transfer. In this work, the experiments were performed at the liquid
velocities in the range of 0.10.4 m s1 . For the process design, it
is assumed that the wetting ratio (x*) correlation given by Eq. (23)
can be used to predict the wetting ratio with the liquid velocities
5.2. Membrane wetting estimation
up to 1.0 m s1 , which is still in the laminar ow region. By using
Eq. (23), at MEA concentration of 1.0 M, the values of x* for liquid
The data obtained from the experiments were validated with the
velocities of 0.6, 0.8 and 1.0 m s1 are 0.1025, 0.1045 and 0.1062,
proposed model for estimating the membrane wetting. The degree
respectively, or the increase is not signicant. The total resistance of
of membrane wetting is expressed in term of wetting ratio (x*),
the partially wetted membrane can be determined by Eq. (1) or (5).
dened as the ratio of the length of the liquid lled pores to the
The gas phase mass transfer coefcient (kG ) is calculated from Eq.
membrane thickness (x* = wetted /M ). The values of x* were deter-
(6), the liquid phase mass transfer coefcient (kL ) is determined
mined by adjusting the value (x*) in the model until the absorption ) membrane
from Eq. (7), the non-wetted (kM ) and wetted (kM
ux obtained from the simulation was very close to or equal to
mass transfer coefcients are determined from Eqs. (8) and (12),
the experimental data. The difference allowed between these two
respectively. The enhancement factor (E) accounting the chemical
values did not exceed 0.1%. In Fig. 5(i), the dots represent the wet- is estimated by Eq. (2).
reaction expressed in the terms of kL and kM
ting ratio obtained from validation of the experiments with the
The resistance contributions of gas, non-wetted membrane, wet-
model, while the trend lines are used to obtain the wetting ratio
ted membrane and liquid phases are dened as RG /Rtot , RM /Rtot ,
correlation as a function of the operating conditions. It can be seen /R
RM tot and RL /Rtot , respectively and are shown in Fig. 6. For MEA
that x* increases with liquid velocity and MEA concentration. The
concentration of 0.25 M (see Fig. 6(i)), the gas phase resistance is
increase in MEA concentration leads to signicant reduced surface
around 3% and the non-wetted membrane resistance is less than 1%
tension and contact angle between the membrane surface and solu-
for all liquid velocities. At low liquid velocities (0.10.2 m s1 ), the
tion. These could increase the membrane wetting. Lu et al. [11]
liquid phase resistance dominates the overall resistance (8060%),
reported that the membrane wetting observed from ux decline
while the estimated resistances for wetted membrane are in a
for absorption with 1.0 M MEA was higher than that of 0.5 M MEA.
range of 1636%. For moderate liquid velocities (0.30.4 m s1 ), the
They discussed that the increase in aqueous organic solution con-
wetted membrane resistances become comparable to the liquid
centration would affect the pore wetness on account of alternation
phase resistances and the wetted membrane resistance contribu-
of solution properties. It is also observed that at low MEA concen-
tions are around 4657%. At high liquid velocities (0.61.0 m s1 ),
tration (0.125 and 0.25 M), the increase of x* with liquid velocity
the overall resistance is controlled by the wetted membrane resis-
is more signicant compared to that of high concentrations. Due
tance (6169%). It can be observed that the contribution of liquid
to the distribution of membrane pore sizes, the large pores may
phase resistance greatly decreases with liquid velocity. In contrast,
be penetrated by liquid absorbent [11,17]. With high MEA concen-
the wetted membrane resistance signicantly increases with liquid
trations, the liquid absorbent can easier penetrate the large pores
velocity due to the increase in x*. Fig. 6(ii) shows the contribution
compared to the low MEA concentration. However, there are lim-
of individual resistance at MEA concentration 1.0 M. The gas phase
ited numbers of large pores which can be wetted. Therefore, the
and non-wetted membrane phase resistances are about 6.5 and
wetting ratio is limited to the typical value. It can be observed from
1.0%, respectively, for all liquid velocities. For low to moderate liq-
Fig. 5(i) that the values of x* for all MEA concentrations become
uid velocities, the resistances of wetted membrane are comparable
close at high liquid velocity. Boributh et al. [17] studied the effect
to liquid phase resistances. At high velocities, the wetted mem-
of pore size distribution on membrane wetting. It was found that at
brane resistance dominates mass transfer of system. By comparing
any liquidgas pressure difference, the membrane wetting would
the liquid phase resistance of the system with MEA concentrations
occur in the pores with size equal to and larger than the criti-
of 0.25 and 1.0 M, it is found that the resistance contribution of
cal radius. Fig. 5(ii) reveals that gas velocity and gas composition
liquid phase of 0.25 M is clearly higher than that of 1.0 M. It is
do not affect the membrane wetting in range of the experiments
because higher MEA concentration could signicantly improve the
(gasliquid ow rate ratios (QG /QL ) in range of 0.452.0). The sim-
rate of reaction which is expressed in term of enhancement factor
ilar result was reported by El-Naas et al. [18]. They found that at
(E) leading to reduced liquid phase resistances [19].
low QG /QL range (approximately 1.02.0), the membrane wetting
was quite constant. While, at high QG /QL range (2.33.5), the wet-
5.4. Suitable module length
ting decreased with increasing QG /QL . This is especially true when
the membrane has signicant large pores. Therefore, the effects of
Previous experiments on CO2 absorption by membrane contact-
gas phase velocity and composition on wetting ratio were ruled out
ing process have been carried out using the hollow ber modules
and the correlation expressing x* as a function of liquid velocity and
with effective lengths of 1030 cm [5,11,20,21]. These laboratory
MEA concentration was developed:
modules provide enough mass transfer area for observing the effect
of operating conditions on the absorption performance. However,
0.1787
x = (0.0725CMEA
2
0.1456CMEA + 0.0803)ln vL + 0.1062CMEA these modules are not suitable for the pilot or industrial scales since
they provide too low contact areas to achieve high removal ef-
(23)
ciency. Therefore, in the design of membrane contacting process for
the real application, the suitable membrane module specications
In the simulation, if the value of x* estimated from above equation is (especially module length) should be carefully considered.
negative (at very low liquid velocity and low MEA concentration), The hollow ber modules have received wide attention due to
x* is adjusted to be zero. The above correlation is assumed to be providing very high mass transfer area. Although, most hollow ber
able to predict the value of x* for higher liquid velocity, but is in the modules are designed for pressure-driven membrane processes
range of laminar ow. The system performance of the multistage (rather than concentration-driven processes) such as MF and UF
cascade membrane contacting process (presented later in Section processes, they can also be employed as membrane contactors.
5.5) is simulated based on the membrane wetting obtained from One of the most well-known transverse ow hollow ber modules
this correlation. is the Liqui-Cel Exra-Flow modules commercialized by CELGARD
S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189 181
0.1 0.08
Wetting ratio ( x* )
Wetting ratio ( x* )
0.08
0.07
0.06
1.00 M 0.06
0.04 0.50 M
0.25 M 40 %vol CO2
0.05 50 %vol. CO2
0.02 0.125 M
0 0.04
0.05 0.15 0.25 0.35 0.45 0.06 0.07 0.08 0.09 0.1 0.11
Liquid velocity, v l (m/s) Gas velocity, v g (m/s)
Fig. 5. Simulation results of (i) effect of liquid velocity and MEA concentration on wetting ratio (yCO2 = 0.4, QG = 0.5 l/min), (ii) effect of gas velocity and CO2 concentration in
gas phase on wetting ratio (CMEA = 0.125 M, vl = 0.3 m s1 ).
Fig. 6. Contribution of individual resistance to overall resistance (i) MEA concentration = 0.25 M, (ii) MEA concentration = 1.0 M, (QG = 0.5 l/min, yCO2 = 0.40).
LLC (Charlotte, USA). These module are 20.3118.6 cm length. Pall determining the wetting ratio is obtained (Eq. (23)). The suitable of
Corporation provides the modules of hydrophobic hollow ber module length and membrane characteristics used in the designed
membranes (PVDF, PP and PTFE) called Microza module. The avail- are as shown in Table 1. The important parameters affecting the
able commercial module lengths are in range of 13222.7 cm. performance of the multistage cascade membrane contactor are
Gabelman and Hwang [1] recommended the commercially avail- investigated, i.e., number of module, MEA consumption, gas ow
able parallel ow hollow ber modules provided by various sources pattern and liquid ow pattern. The simulation was performed for
which would be suitable for membrane contactors and the module the Reynolds number 89890 and 414.23, for the liquid side and
lengths are 17.8304.9 cm. gas side, respectively. The system performance is evaluated based
From the above information, the module with an effective ber on the total feed gas which is the indicator that the system is able
length of 50 cm is selected for the design of multistage membrane to handle high feed gas capacity.
contacting process. The selected length is also in the same range
as the membrane module used by Yeon et al. [22] for absorption 5.5.1. Number of modules
of CO2 from ue gas using pilot-scale membrane contactor (PVDF The module arrangement of the multistage cascade membrane
membrane) with module length of 52 cm. To maintain the dynamic contacting process is shown in Fig. 8. The liquid absorbent is fed in
behavior of gas ow in the shell side, the packing density of ber the lumen of the hollow bers, counter currently to the gas ow
in the design module is xed to be the same as the experimen- in the shell side. Fig. 9 presents the total feed gas for achieving
tal module. The specications of designed module are summarized 90% CO2 removal for the system using different number of mod-
in Table 1. It is assumed that the wetting behavior occurring in ules. It is found that the total feed gas increases with number of
this module can be predicted by Eq. (23) because the membrane module because the contact area is increased leading to improved
specications are the same as those of in the experimental module. absorption capacity.
For the proposed model, the effect of pressure drop on mem-
5.5. Design of multistage cascade membrane contacting process brane wetting as well as system performance is not included. From
the previous results, the membrane wetting is only a function of liq-
The design approach is schematically shown in Fig. 7. From uid velocity and MEA concentration. Therefore, to lower the effect
the experimental and modeling results, the correlation for of pressure drop which is proportional to the module length on the
182 S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189
Experiments
Wetting ratio (x*) Suitable membrane module
Modeling
Number of module
Fig. 7. The design approach for multistage cascade gasliquid membrane contacting process.
membrane wetting, the appropriate number of modules should be is depicted in Fig. 10. It is found that pressure difference linearly
considered. The pressure drop ( P) of liquid ow in the lumen of increases with number of module for all liquid velocities. Theo-
the ber is determined using the HagenPoiseuille equation; retically, the liquid can penetrate the membrane pores when the
pressure difference between liquid and gas sides is higher than
32vL L
P = (24) the wetting pressure. In contrast, the gas bubble formation would
di2 occur if the pressure difference between gas and liquid phases
where is liquid viscosity, L is ber length and di is inner diame- ( PGL = PG PL ) is greater than bubbling pressure [23]. The wet-
ter of the ber. For the gas phase, the pressure drop is much lower ting pressure or/and bubbling pressure ( PC ) can be determined
compared to that of liquid phase [11], thus, the pressure of the gas by Laplace equation;
side is assumed constant over the ber length. In gasliquid mem- 2 cos
PC = (25)
brane contacting process, the pressure difference between liquid rP
and gas phases signicantly affects the operation mode. The cal-
where is surface tension of the liquid absorbent (see Ref. [24]),
culated maximum liquidgas pressure difference ( PLG = PL PG )
is contact angle and rP is average pore radius. The membrane
vl = 1.0 m/s
difference, PL-G (kPa)
12 120
Total feed gas (l/min)
8 80
6 60
1.0 M
4 0.5 M 40
2 0.25 M 20
0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6
Number of module Number of module
Fig. 9. The simulation results of the effect of number of module on total feed gas to Fig. 10. The calculated maximum liquidgas pressure difference, PLG at different
achieve 90%CO2 removal at different MEA concentrations (vl = 0.5 m s1 , yCO2 = 0.4). number of modules and liquid velocities.
S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189 183
(i) 25 (ii) 20
Non-wetted mode 18 Non-wetted mode
20 16 Partially wetted mode
Partially wetted mode
% MEA consumption
% MEA consumption
14
Wetted mode Wetted mode
15 12
10
10 8
6
5 4
2
0 0
0.1 0.3 0.5 0.7 0.9 1.1 0.1 0.3 0.5 0.7 0.9 1.1
Concentration of MEA in feed liquid (mol/l) Liquid velocity, v L (m/s)
(iii) 20 (iv) 35
Non-wetted mode
18
30
16 Partially wetted mode
% MEA consumption
% MEA consumption
Non-wetted mode 25
14 Wetted mode
12 Partially wetted mode 20
10
Wetted mode 15
8
6 10
4
5
2
0 0
0.15 0.25 0.35 0.45 0.55 0.65 0 20 40 60 80 100
Gas velocity, v G (m/s) Concentration of CO 2 in feed gas (%Vol)
Fig. 11. The simulation results of total %MEA consumption of threemodule cascade for three different operation modes including non-wetted, partially wetted (wetting
ratio estimated using Eq. (23)) and completely wetted modes: (i) At different MEA concentrations (vl = 0.2 m s1 , vg = 0.5 m s1 and yCO2 = 0.4); (ii) at different liquid veloc-
ities (vg = 0.5 m s1 , yCO2 = 0.4 and CMEA = 0.25 M); (iii) at different gas velocities (vl = 0.2 m s1 , yCO2 = 0.4 and CMEA = 0.25 M); (iv) at different gas compositions (vl = 0.2 m s1 ,
vg = 0.5 m s1 and CMEA = 0.25 M).
wetting results in performance deterioration, while the bubble for- 5.5.2. MEA consumptions
mation causes gas loss during the operation [13,23]. Therefore, the Fig. 11 presents the simulation results of the total %MEA
pressures of gas and liquid phases must be carefully controlled. consumption for three-module cascade at different operating
Generally, the pressure of liquid side is adjusted to be slightly conditions of the three cases: non-wetted mode (x* = 0), par-
higher than that of gas side. The value of PLG must be kept pos- tial wetted mode (x* estimated using Eq. (23)) and wetted
itive or equal to zero over the ber length. If PLG is xed to mode (x* = 1.0). The consumption of MEA is dened as the
be zero at liquid outlet, the pressure of liquid at the inlet must ratio of moles of MEA consumed to the moles of MEA in feed
be adjusted to be greater than or equal to the value of pressure ((CMEA,feed CMEA,out )/CMEA,feed 100%). It can be obviously seen
drop of liquid ow over the ber length. For the hollow ber that the system operated in non-wetted mode shows highest value
membranes used in this work (rP = 0.08 m, = 93.4 ), MEA con- of %MEA consumption followed by partially wetted mode and
centration of 0.25 M, the wetting pressure calculated from Eq. (25) completely wetted mode, respectively. These reveal that the non-
is 104.93 kPa. Therefore, the membrane wetting should not be wetted mode offers highest absorption performance resulted in
observed from our experiments although the operation was car- highest MEA consumption. For completely wetted mode, variation
ried out at the conditions of highest wetting ratio (0.4 m s1 and of MEA concentration, gas velocity, liquid velocity as well as gas
CMEA 1.0 M). However, due to the distribution of membrane pore composition slightly affects MEA consumption since the overall
size, the large pores may be wetted even the pressure difference resistance is dominated by the membrane phase.
( PLG ) is lower than the wetting pressure estimated based on the Fig. 11(i) shows that at lower MEA concentrations, the percent-
average pore size. Hence, the pressure drop over the length must ages of MEA consumption are higher compared to those of higher
not exceed the wetting pressure. From Fig. 10 it can be seen that concentrations since the amount of MEA consumed in the system
at highest liquid velocity (1.0 m s1 ), the systems with the num- with higher MEA concentration is higher than that of lower MEA
ber of module up to four modules show the maximum liquidgas concentration due to higher rate of reaction. However, because of
pressure difference lower than critical wetting pressure. For four- the very high amount of MEA in the system with high MEA concen-
module cascade, the pressure difference is about 92 kPa which is tration, the MEA consumption of high MEA concentration is still
quite close to the critical wetting pressure. To ensure that the lower than that of lower MEA concentration. The effect of liquid
pressure drop would not signicantly affect the membrane wet- velocity on MEA consumption is depicted in Fig. 11(ii). At lower
ting, the pressure difference should be moderately lower than liquid velocities, the percentages of MEA used for the reaction with
the critical wetting pressure. Therefore, the three-module cascade CO2 are higher compared to those of higher liquid velocities due
which shows the pressure drop around 66.9 kPa is selected for our to longer contact time. Fig. 11(iii) reveals the %MEA consumption
design. for the system with different gas velocity. It is found that the %MEA
184 S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189
0.8
0.4
(i) 20 (ii) 9
18 8
Total feed gas (l/min)
16 7
Total feed gas (l/min)
14 6
12
5
10
4
8
3
6
4 G-ID 2 G-ID
2 G-IS 1 G-IS
0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1
Concentration of MEA in feed liquid (mol/l) Liquid velocity, v l (m/s)
(iii) 35
25
(iv)
G-ID G-ID
30
G-IS 20 G-IS
Total feed gas (l/min)
25
15
20
10
15
10 5
5 0
10 30 50 70 90 110 10 30 50 70 90 110
%Removal of CO 2 Concentration of CO 2 in feed gas (%Vol.)
Fig. 14. Comparison of total feed gas of two different gas ow patterns: (i) at different MEA concentrations (vl = 0.5 m s1 , yCO2 = 0.4 and CO2 removal = 90%) (ii) at various
liquid velocities (CMEA = 0.25 M, yCO2 = 0.4 and CO2 removal = 90%); (iii) at different %CO2 removal (vl = 0.5 m s1 , CMEA = 0.25 M and yCO2 = 0.4); (iv) at different gas compositions
(vl = 0.5 m s1 , CMEA = 0.25 M and CO2 removal = 90%).
S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189 185
5
G-IS patterns at various MEA concentrations. The simulation result
reveals that G-ID gives higher performance compared to G-IS for
4 all MEA concentrations. The total feed gas of G-ID is higher than
that of G-IS roughly 17 and 33% for MEA concentrations of 0.125
3 and 1.0 M, respectively. This trend can be explained that at low
MEA concentrations, the percentage of MEA consumption in the
Module 1 is higher than that of high concentration. The absorption
2 rate signicantly decreases in the next module. This results in the
slight reduction of CO2 concentration in gas phase for Module 3,
1 but the decrease is more signicant for Module 1. Therefore, the
driving force of two gas ow patterns at lower MEA concentration
is closer compared to at high concentration.
0
Fig. 14(ii) illustrates that G-ID has higher system performance
0 1 2 3 4 5 6 7
than G-IS for all liquid velocities. For G-IS, the total feed gas
Total feed gas (l/min) increases with liquid velocity. In case of G-ID, the total feed gas
sharply increases at low liquid velocities (0.10.2 m s1 ), then,
Fig. 15. Comparison of required MEA solution of two different gas ow patterns for
achieving 90% CO2 removal at various total feed gases. there is a slight increase at moderate velocities (0.30.5 m s1 ),
and a gradual decrease at high velocities (0.61 m s1 ). These can
be explained by referring to the resistance analysis (Section 5.3).
Therefore, the CO2 concentration decreases continuously along the At low liquid velocities, the mass transfer resistance is dominated
length from z/L = 3 to z/L = 0 which is the same as that observed for by liquid phase, thus, the system performance is considerably
one module with module length of 3L (150 cm). When the gas mix- enhanced with increasing velocity. For the moderate liquid veloc-
ture at high CO2 concentration is fed into Module 3, CO2 is partially ities, the wetted membrane resistance becomes comparable to
absorbed and the CO2 concentration is greatly reduced in this mod- the liquid phase resistance leading to a slight improve in absorp-
ule. The reductions in CO2 concentration for Modules 2 and 1 are tion ux, whereas at high velocities, the system is controlled by
less signicant because the driving force for G-IS can be ranked as the wetted membrane resistance. Increasing in liquid velocity
Module 3 > Module 2 > Module 1. In case of G-ID, the gas ow direc- could increase membrane wetting rather than improving the liquid
tion is the same as G-IS. The gas is individually fed to one module phase mass transfer, resulting in gradual absorption performance
to achieve 90% removal at the outlet of each module. The values deterioration. The performance comparison of two different gas
of CCO2,i /CCO2,feed at the inlet and outlet for each module are the ow patterns at various % CO2 removals and gas compositions is
Fig. 16. Schematic diagrams of three-module cascade membrane contactor with different liquid ow patterns (i) liquid ow in series (L-IS), (ii) liquid ow in series with
splitting (L-ISS), (iii) liquid ow in series with recycle (L-ISR).
186 S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189
(i) (ii) 9
9
8.5 8.5
8 8
7.5 7.5
7 7
6.5 6.5
6 6
5.5 5.5
5 5
4.5 4.5
4 4
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Fig. 17. Simulation results of (i) effect of liquid velocity on total feed gas at different splitting ratios, (ii) effect of liquid velocity on total feed gas at different recycle ratios
(CO2 removal = 90%, CMEA = 0.25 M, vl = 0.5 m s1 and yCO2 = 0.4).
(i) (ii) 21
9
8.5 19
7.5
15
7
6.5 13
6 11
5.5
9
5
7
4.5
4 5
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Liquid velocity, v l (m/s) Liquid velocity, v l (m/s)
Fig. 18. Comparison of total feed gas for three different liquid ow cascades (i) MEA concentration = 0.25 M, (ii) MEA concentration = 1.0 M, (CO2 removal = 90% and yCO2 = 0.4).
represented in Fig. 14(iii) and (iv), respectively. G-ID shows higher with ow rate of QL,R is recycled and is combined with the original
system performance for all values of % CO2 removals and gas feed before entering Module 1.
compositions. The total feed gas into the system decreases with Fig. 17(i) depicts the effect of splitting ratio () on system
increasing % CO2 removal and CO2 concentration in feed gas for performance of L-ISS at different liquid velocities (with MEA con-
both gas ow patterns. centration of 0.25 M). It is found that L-ISS shows higher system
Fig. 15 depicts the comparison of MEA solution required for performance compared to L-IS ( = 0) at high liquid velocities,
achieving 90% CO2 removal at various total gas ow rates. It is whereas the splitting ratio of 0.3 gives the highest system perfor-
clearly seen that required MEA solution for G-IS is much higher mance. The total feed gas of L-ISS with splitting ratio of 0.3 is higher
compared to that of G-ID. At total gas feed of 5.92 l/min, the required than that of L-IS ( = 0) around 16.7% at liquid velocity 0.9 m s1 . The
MEA solution for G-IS is approximately 5.1 l/min, while that of G-ID effect of recycle ratio () on system performance for L-ISR at vari-
is only 1.5 l/min. These reveal that MEA solution is more effectively ous liquid velocities (MEA concentration 0.25 M) is represented in
used in case of G-ID. Therefore, the gas ow pattern of G-ID is Fig. 17(ii). It is found that L-ISR can moderately improve system
selected to study the effect of liquid ow pattern in the multistage performance at low liquid velocities. The Highest system perfor-
cascade contactors. mance is observed at recycle ratio of 1.0. The total feed gas of L-ISR
with recycle ratio of 1.0 is greater than that of L-IS ( = 0) roughly
5.5.4. Liquid ow patterns 7.7% at liquid velocity 0.1 m s1 .
The schematic diagrams of three-module cascade with differ- Fig. 18(i) and (ii) compares the system performance of three
ent liquid ow patterns are shown in Fig. 16. The rst liquid ow different liquid ow patterns. For MEA concentration of 0.25 M, L-
pattern is (i) Liquid ow in series (L-IS), which is similar to that in ISR can moderately improve the system performance at low liquid
Fig. 12(i). The second liquid ow pattern is (ii) Liquid ow in series velocities, whereas L-ISS considerably enhances the performance
with splitting (L-ISS), from which the total liquid feed is partially at high liquid velocities. For MEA concentration of 1.0 M, the sys-
fed into Module 1 and the split liquid is partly combined with the tem performance can be signicantly increased by L-ISR at low to
stream leaving Module 1 before entering Module 2. It is noted that moderate liquid velocities (0.10.5 m s1 ). The total feed gases are
before entering module 3, the remained liquid is combined with the enhanced approximately 44.6 and 7.3% for liquid velocities of 0.1
liquid leaving Module 2. The third liquid ow pattern is (iii) Liquid and 0.4 m s1 , respectively. The L-ISS does not improve the system
ow in series with recycle (L-ISR), from which the liquid feed ow performance at any liquid velocity. The reason for these differ-
rate to each module is the same. After leaving Module 3, the liquid ent results for typical MEA concentrations is due to the different
S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189 187
membrane wetting behaviors. At low MEA concentration, the val- application. The operation of multistage cascade membrane con-
ues of x* greatly increase with liquid velocity (Section 5.2). At tactor offers several advantages including the improved system
high liquid velocities, the wetted membrane resistance controls the performance and the use of MEA effectively. However, the design
overall resistance, and L-ISS can reduce membrane wetting ratio by may be different depending on the conditions of each system.
reducing the liquid ow rate in each module resulting in improv- Therefore, the experimental study is still necessary in order to
ing the performance. On the other hand, at low liquid velocities, obtain the specic results for being used in the process design.
the overall resistance is dominated by liquid phase resistance, and
L-ISR can enhance the mass transfer coefcient of liquid phase by
increasing the liquid velocity in each module leading to enhanc- Acknowledgments
ing absorption ux. In case of high MEA concentration, the wetting
ratio slightly increases with liquid velocity, and the liquid phase The authors gratefully acknowledge the nancial support from
resistance is comparable to wetted membrane. The reduction of the Royal Golden Jubilee program and Senior Research Scholar
liquid ow in the module by splitting (L-ISS) does not consider- Grant from Thailand Research Fund (TRF) and from King Mongkuts
ably reduce the membrane wetting; thus, the system performance University of Technology Thonburi (KMUTT).
is not enhanced for all liquid velocities. L-ISR can improve system
performance at low to moderate liquid velocities, because in this
range the liquid phase resistance still inuences the system. The Appendix A.
increase of liquid velocity can improve the mass transfer in liquid
phase rather than increasing the membrane wetting. The results of The diffusivity of CO2 in gas mixture (CO2 CH4 ), DCO2,G can be
multistage design based on wetting ratio obtained from the exper- calculated by the following [25]:
imental results reveal that different wetting behaviors at various
operating conditions (liquid velocity and MEA concentration) sig- 1/2
0.001858T 3/2 (1/MCO2 + 1/MCH4 )
nicantly affects the design of membrane contactors. DCO2 ,G = 2
(A.1)
PAB D
where MCO2 and MCH4 are the molecular weights of CO2 and CH4 ,
6. Conclusions
respectively. P is pressure in atm, AB is characteristic length (m),
D is Collision integral. These parameters can be estimated by the
The experiments of CO2 absorption by MEA using PVDF hol-
followings;
low ber membrane contactor were carried out. The mathematical
model has been developed to predict the system performance. A + B
The experimental results were validated with the proposed model AB = (A.2)
2
for estimating the wetting ratio (x*) as function of liquid velocity
and MEA concentration. At low MEA concentration, the value of where A and B are characteristic lengths of CO2 and CH4 , respec-
x* greatly increases with liquid velocity, while that of high con- tively, and it can be determined as;
centration slightly increases with velocity. The resistance analysis
demonstrates that for MEA concentration of 0.25 M, the liquid
1.585Vb,A
phase resistance dominates the overall resistance at low liquid A = (A.3)
1 + 1.3
2
velocities. The wetted membrane resistance becomes comparable
to that of liquid phase at moderate velocities, and controls the
system at high liquid velocities. For MEA concentration of 1.0 M, where Vb is liquid molar volume at normal boiling point (cm3 /mol).
the wetted membrane resistance is comparable to the liquid phase The value of can be estimated by the following;
resistance for low to moderate liquid velocities, and dominates the
overall resistance at high velocities. 1.94 103 p2
The suitable hollow ber membrane module with effective ber
= (A.4)
Vb Tb
length of 50 cm is selected for the design of multistage membrane
contactors. The three-module cascade is selected based on the pres- where p is dipole moment (Debye). Tb is boiling temperature (K).
sure drop. The system performance is evaluated as the total feed The Collision integral (D ) is in function of temperature deter-
gas and MEA consumption. By comparing three different modes of mined as;
operation, the non-wetted mode presents highest MEA consump-
tion followed by partially wetted mode and completely wetted a c e g
D = + + + (A.5)
mode, respectively. Different gas and liquid ow patterns are com- (T )b exp(dT ) exp(fT ) exp(hT )
pared. For gas ow pattern, G-ID shows higher system performance
compared to G-IS for all operating conditions studied. In case of liq- where a, b, c, d, e, f, g and h are constants. T* can be calculated as the
uid ow pattern, L-ISS achieves the highest performance at splitting following equation;
ratio () of 0.3, while that of L-ISR is obtained at recycle ratio () of
1.0. For low MEA concentration (0.25 M), the L-ISR can moderately T
improve the system performance at low liquid velocities, while T = (A.6)
AB
L-ISS considerably enhances the performance at high liquid veloc-
ities. For the system with high MEA concentration (1.0 M), L-ISR
where is Boltzmanns constant and AB is characteristic energy
can improve the performance at low to moderate liquid velocities,
determined by;
whereas the L-ISS does not improve the system performance at any
liquid velocity. 1/2
In conclusion, this work presents the design of multistage cas- AB =( A B) (A.7)
cade membrane contacting process for chemical absorption of
CO2 that has never been reported in the literature. The process where A and B are characteristic energy of CO2 and CH4 , respec-
design gives the insight and guideline for the scale up for real tively.
188 S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189
Nomenclature
Greek letters
A contact area (m2 ) D collision integral for molecular diffusion (dimen-
CA concentration of CO2 in bulk liquid (mol m3 ) sionless)
CAG,in inlet concentration of CO2 in gas phase (mol m3 ) Boltzmanns constant (J K1 )
CAGout , outlet concentration of CO2 in gas phase (mol m3 ) AB characteristic length (m)
CA,eq concentration of CO2 at gasliquid interface (mol density (kg m3 )
m3 ) dry dry membrane thickness (m)
CR,L concentration of absorbent (mol m3 ) wetted wetted membrane thickness (m)
dh hydraulic diameter (m), dh = ds2 n d02 /ds + n d0 split ratio (dimensionless)
di inner diameter of membrane (m) recycle ratio (dimensionless)
dint interfacial diameter (m2 ) M membrane porosity (dimensionless)
dln logarithmic mean diameter of non-wetted mem- kinematic viscosity (cm2 s1 )
brane (m), dln = d0 dint /ln(d0 /dint ) R stoichiometric coefcient of reaction
dln logarithmic mean diameter of wetted membrane M membrane tortuosity (dimensionless)
=d
(m), dln int di /ln(dint /di ) i characteristic energy of species i (J K1 )
d0 outer diameter of membrane (m)
dpore average pore size diameter of membrane (m)
ds module diameter (m)
DG,eff gas effective diffusivity (m2 s1 ) References
DG,Kn Knudsen diffusion coefcient (m2 s1 )
DG,M Fickians molecular diffusion coefcient (m2 s1 ) [1] A. Gabelman, S.-T. Hwang, Hollow ber membrane contactors, Journal of Mem-
Di,G diffusivity of component i in gas mixture (m2 s1 ) brane Science 159 (1999) 61106.
[2] P. Keshavarz, J. Fathikalajahi, S. Ayatollahi, Analysis of CO2 separation and
Di,L diffusivity of component i in liquid solution (m2 s1 ) simulation of a partially wetted hollow ber membrane contactor, Journal of
E enhancement factor (dimensionless) Hazardous Materials 152 (2008) 12371247.
E asymptotic innite enhancement factor [3] S. Atchariyawut, R. Jiraratananon, R. Wang, Separation of CO2 from CH4 by using
gasliquid membrane contacting process, Journal of Membrane Science 304
fA molar ow rate of CO2 in liquid phase (kgmol s1 ) (2007) 163172.
FA molar ow rate of CO2 in gas phase (kgmol s1 ) [4] A. Bottino, G. Capannelli, A. Comite, R. Di Felice, R. Firpo, CO2 removal from a
FI0 molar ow rate of inert component in gas phase gas stream by membrane contactor, Separation and Purication Technology 59
(2008) 8590.
(kgmol s1 ) [5] S. Khaisri, D. deMontigny, P. Tontiwachwuthikul, R. Jiraratananon, Comparing
Gz Graetz number (dimensionless) membrane resistance and absorption performance of three different mem-
Ha* Hatta number (dimensionless) branes in a gas absorption membrane contactor, Separation and Purication
Technology 65 (2009) 290297.
H Henrys constant (kPa m3 kgmol1 ) [6] M. Mavroudi, S.P. Kaldis, G.P. Sakellaropoulos, Reduction of CO2 emissions by
JA,i absorption ux of CO2 (mol m2 s1 ) a membrane contacting process[small star, lled], Fuel 82 (2003) 21532159.
KL overall mass transfer coefcient (m s1 ) [7] R. Wang, D.F. Li, C. Zhou, M. Liu, D.T. Liang, Impact of DEA solutions with and
without CO2 loading on porous polypropylene membranes intended for use as
kG gas mass transfer coefcient (kgmol m2 kPa1 s1 )
contactors, Journal of Membrane Science 229 (2004) 147157.
kL liquid mass transfer coefcient (m s1 ) [8] Y. Lv, X. Yu, S.-T. Tu, J. Yan, E. Dahlquist, Wetting of polypropylene hollow ber
kM gas-lled pores membrane mass transfer coefcient membrane contactors, Journal of Membrane Science 362 (2010) 444452.
(m s1 ) [9] M. Mavroudi, S.P. Kaldis, G.P. Sakellaropoulos, A study of mass transfer resis-
tance in membrane gasliquid contacting processes, Journal of Membrane
kM liquid-lled pores membrane mass transfer coef- Science 272 (2006) 103115.
cient (m s1 ) [10] S. Khaisri, D. deMontigny, P. Tontiwachwuthikul, R. Jiraratananon, A mathe-
k2 second-order reaction rate constant (L mol1 s1 ) matical model for gas absorption membrane contactors that studies the effect
of partially wetted membranes, Journal of Membrane Science 347 (2010)
L length of hollow ber membrane (m) 228239.
Mi molecular weight of component i (kg kgmol1 ) [11] J.-G. Lu, Y.-F. Zheng, M.-D. Cheng, Wetting mechanism in mass transfer process
n number of bers of hydrophobic membrane gas absorption, Journal of Membrane Science 308
(2008) 180190.
N number of stages [12] R. Wang, H.Y. Zhang, P.H.M. Feron, D.T. Liang, Inuence of membrane wetting
P total system pressure (kPa) on CO2 capture in microporous hollow ber membrane contactors, Separation
PA partial pressure of CO2 (kPa) and Purication Technology 46 (2005) 3340.
[13] S. Boributh, S. Assabumrungrat, N. Laosiripojana, R. Jiraratananon, Effect of
QG total volumetric ow rate of gas phase (m3 s1 ) membrane module arrangement of gasliquid membrane contacting process
QL total volumetric ow rate of liquid phase (m3 s1 ) on CO2 absorption performance: a modeling study, Journal of Membrane Sci-
Re Reynolds number (dimensionless) ence 372 (2011) 5786.
[14] M.C. Yang, E.L. Cussler, Designing hollow-ber contactors, AIChE Journal 32
RG gas phase resistance (s m1 )
(1986) 19101916.
RL liquid phase resistance (s m1 ) [15] H. Kreulen, C.A. Smolders, G.F. Versteeg, W.P.M. Van Swaaij, Determination
RM gas-lled pores membrane resistance (s m1 ) of mass transfer rates in wetted and non-wetted microporous membranes,
RM liquid-lled pores membrane resistance (s m1 ) Chemical Engineering Science 48 (1993) 20932102.
[16] S. Atchariyawut, R. Jiraratananon, R. Wang, Mass transfer study and modeling
Rtot total resistance (s m1 ) of gasliquid membrane contacting process by multistage cascade model for
Sc Schmidt number (dimensionless) CO2 absorption, Separation and Purication Technology 63 (2008) 1522.
Sh Sherwood number (dimensionless) [17] S. Boributh, S. Assabumrungrat, N. Laosiripojana, R. Jiraratananon, A modeling
study on the effects of membrane characteristics and operating parameters
T temperature (K) on physical absorption of CO2 by hollow ber membrane contactor, Journal of
x* wetting ratio (dimensionless) Membrane Science 380 (2011) 2133.
yCO2 molar fraction of CO2 in gas mixture (dimensionless) [18] M.H. El-Naas, M. Al-Marzouqi, S.A. Marzouk, N. Abdullatif, Evaluation of the
removal of CO2 using membrane contactors: membrane wettability, Journal of
Membrane Science 350 (2011) 410416.
[19] P.S. Kumar, J.A. Hogendoorn, P.H.M. Feron, G.F. Versteeg, Approximate solution
to predict the enhancement factor for the reactive absorption of a gas in a liquid
owing through a microporous membrane hollow ber, Journal of Membrane
Science 213 (2003) 231245.
S. Boributh et al. / Journal of Membrane Science 401402 (2012) 175189 189
[20] S.-H. Lin, K.-L. Tung, W.-J. Chen, H.-W. Chang, Absorption of carbon dioxide by [23] V.Y. Dindore, D.W.F. Brilman, P.H.M. Feron, G.F. Versteeg, CO2 absorption at ele-
mixed piperazine-alkanolamine absorbent in a plasma-modied polypropy- vated pressures using a hollow ber membrane contactor, Journal of Membrane
lene hollow ber contactor, Journal of Membrane Science 333 (2009) 3037. Science 235 (2004) 99109.
[21] Y.-S. Kim, S.-M. Yang, Absorption of carbon dioxide through hollow ber [24] G. Vazquez, E. Alvarez, J.M. Navaza, R. Rendo, E. Romero, Surface tension of
membranes using various aqueous absorbents, Separation and Purication binary mixtures of water + monoethanolamine and water + 2-amino-2-methyl-
Technology 21 (2000) 101109. 1-propanol and tertiary mixtures of these amines with water from 25 C to
[22] S.-H. Yeon, K.-S. Lee, B. Sea, Y.-I. Park, K.-H. Lee, Application of pilot-scale mem- 50 C, Journal of Chemical & Engineering Data 42 (1997) 5759.
brane contactor hybrid system for removal of carbon dioxide from ue gas, [25] R.C. Reid, J.M. Prausnitz, B.E. Poling, The Properties of Gases and Liquids, 4th
Journal of Membrane Science 257 (2005) 156160. ed., McGraw-Hill, Inc, New York, 1987.