Journal of Membrane Science 294 (2007) 50–59

Carbon dioxide recovery from post-combustion processes:

Can gas permeation membranes compete with absorption?
Eric Favre ∗
LSGC (CNRS UPR 6811) ENSIC-INPL, 1 rue Grandville, 54001 Nancy, France
Received 10 March 2006; received in revised form 26 January 2007; accepted 4 February 2007
Available online 9 February 2007

Abstract
The design of efficient capture processes to recover carbon dioxide from combustion processes, to restrict greenhouse gas emissions, has received
considerable attention. Gas permeation processes based on dense polymeric membranes are almost routinely rejected for this application. Several
engineering studies conclude that absorption processes (using packed towers or membrane contactors) are the best available technology. This paper
aims to provide a critical comparison of dense polymeric membrane capture processes versus amine absorption when applied in a post-combustion
(i.e. flue gas treatment) situation. Technological and scientific challenges, as well as prospects for future developments are discussed. The potential
of dense polymeric membranes to solve the flue gas treatment problem may have been underestimated.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Membranes; Carbon dioxide; Absorption; Energy; Design

1. Introduction Schematically, carbon dioxide capture in a fossil fuel com-

bustion process can be achieved according to three different
The recovery of carbon dioxide from large emission sources strategies [3]:
is a formidable technological and scientific challenge which has
received considerable attention for several years [1,2]. A great (i) A first option consists to perform the oxygen/nitrogen sep-
number of engineering studies has been dedicated to the achieve- aration on the feed stream, so that a CO2 /H2 O mixture is
ment of this goal in order to restrict greenhouse gases emissions, produced through the combustion process. This so-called
which remains the number one challenge of global warming oxycombustion alternative is appealing since it makes use
scenario. In that context, the identification of a capture process of a proven gas separation technology, namely the separa-
which would fit the needs of target separation performances, tion of oxygen from air, in order to solve the problem.
together with a minimal energy penalty, is a key issue. In fact, the (ii) Another option consists to achieve an in situ separation
capture step accounts for 60–80% of the overall costs of the com- and/or CO2 recovery, such as the chemical looping or pre-
plete sequence, including capture, transportation and storage combustion capture alternatives (the latter one being more
[3,4]. The attainment of a technico-economically viable carbon specific of clean power plants).
capture and storage strategy thus demands an efficient capture (iii) Finally, carbon dioxide recovery can also naturally be per-
process to be developed, given the stringent targets of projects in formed at the end of pipe (i.e. from the fume exhaust), in
this field. For instance, U.S. DOE targets for CO2 sequestration a so-called post-combustion situation. This third possibil-
are 90% carbon dioxide sequestration at a cost increase of not ity is by far the most challenging one since a diluted, low
more than 30% of the cost of power plant electricity and not pressure, hot and wet CO2 /N2 mixture has to be treated.
more than 10% by 2012. Nevertheless, it corresponds also to the most widely appli-
cable option in terms of industrial sectors (cement and steel
production for instance). Moreover, it shows the essential
advantage to be compatible to a retrofit strategy (i.e. an
∗ Tel.: +33 383 17 53 90; fax: +33 383 32 29 75. already existing installation can be, in principle, subject to
E-mail address: [email protected]. this type of adaptation).

0376-7388/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2007.02.007
 E. Favre / Journal of Membrane Science 294 (2007) 50–59 51

Separation processes such as cryogeny, absorption, adsorp- eters for that situation. While amine absorption (especially with
tion, and membranes are most often listed as potential candidates monoethanolamine aqueous solutions) is by far the separation
for the three above situations [4,5,46]. Mineral membranes process which is finally chosen, membranes gradually play an
are considered as particularly promising for oxycombustion increased role towards this application. This topic is largely
(O2 /N2 ) and precombustion (CO2 /H2 ) applications [3,6–8]. For described in the membrane literature and a classical (and proba-
post-combustion applications however, the situation is com- bly oversimplified) statement consists to claim that membranes
pletely different: a large consensus has arisen, based on several can potentially compete with absorption as soon as the CO2
engineering projects (such as the carbon capture project [5]), content in natural gas exceeds 10% [14].
that chemical absorption in a liquid solvent is by far the best The major reason which governs the choice of absorption for
available technology. Even though numerous challenges remain natural gas treatment is clear: given the fact that a valuable prod-
for this type of application (too high energy requirement, sol- uct (namely methane) has to be recovered, methane losses have
vent losses by flooding, amine deactivation, secondary CO2 to be minimized. Consequently, process selectivity plays a key
production associated to solvent regeneration by steam, . . .), role and amine washing, which shows an extremely high selec-
a review of the engineering context of post-combustion shows tivity towards CO2 , is decisively advantaged. The main interest
that membrane processes such as gas permeation through a dense of membranes in thus to provide, for streams showing a high CO2
polymeric material are discarded in the early stages of the pro- content, an abatement in a first stage, while absorption provides,
cess selection project. For instance, in a recent review published in a second step, a polishing operation.
by the International Panel on Climate Change and dedicated The last column of Table 1 summarizes the context of carbon
to carbon capture and storage (CCS), gas permeation through capture and sequestration (CCS) for a post-combustion situation.
membranes is considered as inappropriate [3]. It is the opinion At first glance, two major arguments can be exposed in order to
of the author that the above statement is based on either incom- maintain the superiority of amine absorption:
plete or unclear arguments. Various explanations are exposed
in order to justify dense membranes rejection. The high spe- (i) Feed pressure being very low, the driving force of mem-
cific energy requirement associated to these processes is most brane units will be too low in order to achieve the target
often mentioned [1,9,10]. Additionally, the limited resistance performances.
of polymers to high temperature, the inadequacy of membrane (ii) Temperature being much higher than for natural gas feeds,
processes for high feed flowrates, the sensitivity to clogging by the limited temperature resistance of polymers excludes
dust, the absence of economy of scale or the too low selectiv- processes making use of such materials.
ity of membranes towards the CO2 /N2 mixture have been also
proposed. It is important to note however that several problems can also
This analysis is surprising when one compares a somehow be identified when one considers amine absorption for CCS:
similar situation, namely the removal of acid gases for natural
gas treatment, for which membranes have continuously gained (i) The high selectivity argument which prevails for amine
competitiveness towards amine absorption [11–15]. selection in natural gas treatment, so that hydrocarbon
It is the intention of this paper to provide a tentatively lucid, losses are minimized, has to be reconsidered. The CO2
critical and somehow provocative reassessment of the pros and target purity is by no means governed by a maximal selec-
cons of chemical absorption versus membrane separation when tivity constraint, but has to be balanced against the overall
applied in a post-combustion situation. Given the importance energy requirement of the capture, transport and storage
of this application, both from the environmental and economic chain. This key issue has not been explored in detail up to
point of views (annual worldwide CO2 emissions are estimated now and will be developed later.
around 20 × 109 tonnes and target capture costs are roughly (ii) The fact that the feed stream contains oxidizing com-
around $20 per tonne of CO2 recovered [3]), it appears crucial to pounds (oxygen or sulfur dioxide) induces major problems
fill the gap between the positions of membranologists [6,16,17] of amine deactivation and promotes corrosion [18].
and reports from energy or engineering companies [3–5]. In a (iii) Amine losses and degradation, either on the absorption
first step, the similarities and differences between natural gas side (i.e. release in the atmosphere) or on the stripping
and post-combustion treatment will be discussed. The selectiv- side (i.e. with the CO2 stream to be transported in a pipe)
ity, energy and design challenges will then be reviewed. In a are not negligible (typically, losses in the range of 0.1%
last part, prospects in terms of processes and unexplored top- on a weight basis, i.e. 1 kg of solvent losses per ton of
ics, which could play a role in the overall framework of carbon CO2 absorbed, are often mentioned) and are very prob-
dioxide capture and storage will be proposed. lematic. The increased risk of corrosion of columns or
pipes when oxygen and amine are simultaneously present
2. Overall framework and comparison with natural gas is documented [19] but remains rarely mentioned.
treatment (iv) A wet stream of CO2 is produced, which demands a drying
step to be implemented in order to fit transport specifica-
Acid gas removal from an inert diluting feed stream tions [20].
has been practiced for decades in the natural gas industry. (v) Last but not least, the steam regeneration operated in the
Table 1 summarizes the main variables and operating param- stripper unit requires a large amount of energy (between 4
52 E. Favre / Journal of Membrane Science 294 (2007) 50–59

Table 1
Synopsis of the natural gas vs. post-combustion capture framework (from [3,14,31,42], adapted)
Natural gas treatment Post-combustion capture and storage

Feed composition
Major compound CH4 N2
CO2 5–30% 4–30%
Other compounds H2 S, hydrocarbons O2 (∼5%)
H2 O (∼5–12%)
Feed conditions
Pressure (bar) 50–150 ∼1
Temperature (◦ C) 20–60 100–250

Target

CH4 purification CO2 bulk removal

Key issues CO2 < 2%, H2 O < 120 ppm, H2 S < 4 ppm, total CO2 > 80%, 0.5 < H2 O < 500 ppm, H2 S < 1500 ppm,
inerts < 4% O2 < 10 ppm, recovery (R) > 80%
Challenges Minimize CH4 losses Minimize energy requirement and equipment size
Best available technology Amine absorption (or hybrid with membranes if CO2 in Absorption (chemical or hybrid), possibly in membrane
feed > 10%) contactors

and 6 MJ/kg of CO2 recovered) which is still considered A sketch of an absorption and membrane unit for post-
as too high [3]. Simultaneously, a considerable amount of combustion capture is shown in Fig. 1. It can be seen that the
CO2 can be produced, if steam is obtained thanks to a first drawback listed above for gas permeation processes (namely
fossil fuel combustion process. This so called secondary temperature sensitivity) can be quickly removed since the feed
CO2 production can be close to the quantity of CO2 to be mixture has also to be cooled down to 40–60 ◦ C for gas absorp-
recovered in the feed stream. tion, in order to promote CO2 solubility. Polymeric membranes

Fig. 1. Post-combustion carbon dioxide capture: (a) absorption in chemical solvent and (b) membrane permeation (feed compression).
 E. Favre / Journal of Membrane Science 294 (2007) 50–59 53

can obviously withstand this temperature range, which is similar

to currently running units for natural gas treatment [15].
Thus, it appears that the major drawback of membrane opera-
tion, which remains at this stage, concerns the too low selectivity
of membrane materials, which induces a redhibitory energy
requirement. In fact, natural gas is usually delivered to the pro-
cessing plant at a high pressure and it is desirable for the methane
to remain at pressure and the carbon dioxide to be vented at low
pressure. Membranes can do this and so are favored for this
application. In a carbon capture and sequestration plant, the flue
gas arrives at low pressure and must be compressed. It would
be desirable to maintain the carbon dioxide at pressure and vent
the nitrogen at low pressure; thus, membranes are not a priori
favored. A minimal CO2 /N2 selectivity of 100–200 is often men-
tioned as an essential prerequisite [10]. This statement explains
the large number of studies devoted to the search of selective
CO2 membranes which have been reported recently. This objec-
tive is often considered as a major target in membrane science
prospects [6,16,17]. A state of the art analysis towards this key Fig. 3. A non-exhaustive collection of experimental selectivity/permeability
data for CO2 /N2 mixture reported in the literature [22–28,47] and corresponding
issue is detailed hereafter. tradeoff curves for glassy (dotted line) and rubbery (bold dotted line) polymers
(computations performed according to the equations and data set reported in
3. The selectivity challenge [21]).

A classical starting point in the analysis of the feasibility

of a dense membrane gas separation process is based on so- other cases, especially if a combination of highly permeable
called tradeoff curves (ideal selectivity versus fast component and selective enough polymers is needed. More specifically, the
permeability) [11,12]. CO2 /N2 theoretical tradeoff curve, com- fact that polymeric membranes have been able to compete with
puted according to a predictive approach developed by Freeman, amine absorption for natural gas treatment based on a trade-off
is shown in Fig. 2 [21]. Two other mixtures, corresponding to curve which is globally less favorable than the CO2 /N2 mixture
already exiting industrial applications have been added in order could be seen as an encouraging indication.
to enable a comparison: O2 /N2 which is also typical of the oxy- In a second step, a series of experimental permeability data
combustion approach for carbon capture, and CO2 /CH4 for acid obtained with CO2 and N2 has been collected and the results are
gas removal in natural gas treatment. One point is worth men- shown in Fig. 3. The corresponding trade-off curves at 298 K,
tioning according to this preliminary approach: CO2 /N2 mixture obtained for rubbery and glassy polymers, have been added.
appears globally as an easier separation compared to the two Several points are worth mentioning:

(i) First, the highest ideal selectivity which has been reported
up to now is close to the trade-off limit for highly perme-
able glassy membranes and remains below 100. It could
be concluded at this stage that significant efforts have to
be made in order to reach the 200 value mentioned in the
introductory part.
(ii) Trade-off predictions suggest that rubbery polymers will
exhibit higher selectivity compared to glassy ones. This
statement, which remains to be experimentally proven,
could have positive implications: rubbery membranes are
in principle less sensitive to contaminants such as heavy
compounds and less subject to plasticization effects. These
complex and undesired characteristics are often considered
as a major limitation of the glassy membranes used for
natural gas treatment [12].
(iii) Finally, it is important to note that both the trade-off pre-
dictions and the experimental data shown in Fig. 3 concern
a non-reactive permeation process, taking place in homo-
Fig. 2. Comparative tradeoff curves for O2 /N2 , CO2 /CH4 and CO2 /N2 mix- geneous media. For an acid gas such as CO2 , significant
tures through glassy polymers at 293 K (computations performed according to improvements of selectivity could be likely achieved based
equations and data set reported in [21]). on ionic or chemically reactive membranes (similarly to the
54 E. Favre / Journal of Membrane Science 294 (2007) 50–59

amine absorption process) [29,30,48]. The potentialities of

heterogeneous media such as mixed matrix membranes,
which have given promising improvements for natural gas
treatment [12], or membrane processes using carriers [54]
should also be explored for this application. Nevertheless,
these various tracks of selectivity improvement are beyond
the scope of this paper and will not be considered further.

A key question at this stage is to evaluate how the best poly-

meric membranes, working on a “simple” physical process,
could compete with amine absorption for CO2 capture. Accord-
ing to Fig. 3, a CO2 /N2 ideal selectivity (α*) of 60 associated to
a CO2 permeability of 500 Barrer can be taken as the upper limit
today. It is important to stress that these data differ significantly
from those which are classically used in engineering studies.
For instance, an analysis of the performances of gas permeation
membranes in a CCS framework has been recently reported with Fig. 4. Permeate composition yp vs. carbon dioxide recovery ratio R for 10%
(filled symbols) and 20% (open symbols) carbon dioxide in the feed mixture and
a CO2 /N2 selectivity of 20 and a CO2 permeability of 72 Bar-
a membrane selectivity α of 50 (circles) and 200 (squares). To draw this graph,
rer [45]. Specific energy requirement for CO2 capture (usually solutions obtained from the numerical resolution of the cross plug flow model
expressed in MJ/kg of CO2 recovered and ranging between 4 with a pressure ratio (ψ) of 0.01 have been taken. Any increase of the pressure
and 6 for amine absorption in post-combustion) being the pri- ratio will induce a decrease in permeate composition.
marily variable to be taken into account, an energy analysis has
to be undertaken. A systematic analysis of the set of solutions of the problem
for a single stage membrane module, based on these two key
4. The energy challenge variables (R and y), has been achieved. A sample curve is shown
in Fig. 4. It can be seen that for a 10% mole fraction of CO2 in the
The basic framework of a single stage gas permeation pro- feed mixture, a 0.9 permeate mole fraction and a 90% CO2 recov-
cess, such as it could be tentatively applied in a post-combustion ery cannot be simultaneously fulfilled, even with a membrane
situation, is sketched in Fig. 1b. In order to estimate the energy showing an ideal selectivity (α) of 200. The achievement of this
requirement E associated to a given set of target conditions, the goal demands a 20% mole fraction of CO2 in the feed mixture.
driving force, expressed as the pressure ratio (Ψ = P /P ) needed This first analysis shows that, within the range of constraints
to achieve the performances, has first to be determined. The fol- classically imposed for carbon capture (R > 0.8 and y > 0.8),
lowing expression, assuming a single stage adiabatic process for gas permeation performances seem to be extremely sensitive
feed compression operation, can then be computed in order to to selectivity and pressure ratio. Consequently, the associated
evaluate the energy requirement: energy requirement, which depends on the pressure ratio, is also
  expected to vary largely.

QIN γRT P  (γ−1)/γ This point has been investigated and the most relevant
E= −1 (1) results are shown in Fig. 5, where the membrane selectivity
η γ −1 PIN
α ≡ P CO2 /P N2 , discussed in the preceding section, is put on
Compressor efficiency (η) is typically assumed to range the x-axis, as a function of the energy requirement (E) and the
between 0.85 and 0.90 in engineering studies [10,49,50]. The corresponding pressure ratio (Ψ ). Within the range of mem-
pressure ratio can be determined providing that a series of brane selectivity data used (from 40 to 200), no solution can be
assumptions is made concerning module characteristics. For obtained if a recovery ratio of 90% and a permeate composi-
gas permeation operation, the so-called cross plug flow model, tion above 0.8 are simultaneously imposed, in accordance with
which assumes plug flow on the upstream side, is classically pro- Fig. 4. Multistage operation should be applied if these objec-
posed in order to achieve that purpose [31]. Details of the system tives are to be fulfilled. This limitation of membrane processes,
of equations and resolution method which has been used in this which could be seen as a decisive penalty, has to be taken into
paper, based on DASSL software, can be found elsewhere [32]. account in post-combustion analysis and might be part of the
Cross plug flow model has been used in order to estimate the explanation of their (too) early rejection.
energy requirement for different target separation performances, Nevertheless, for a recovery of 80% and a permeate composi-
expressed through the CO2 recovery ratio R (typically 0.8 or 0.9) tion of 80% in CO2 , which remain within the range of acceptable
and the permeate CO2 mole fraction y (typically 0.8 and 0.9). target performances [3], Fig. 5a shows the set of solutions which
Based on the variables detailed in Fig. 1b, R can be computed can be obtained in terms of operating pressure ratio (Ψ = P /P ,
through: right y-axis) and corresponding energy requirement (E, left y-
axis). For a 10% mole fraction of CO2 in the feed mixture, a
QP · y value which is typically taken in capture studies [2–5], a mem-
R= (2)
QIN · xIN brane selectivity above 120 is required, and, even under these
 E. Favre / Journal of Membrane Science 294 (2007) 50–59 55

moderately decreases the required energy, suggesting that more

selective membrane materials would not decisively change the
picture in that case.
In Fig. 5b, a different operation process has been investigated
which is based on vacuum pumping in place of feed compression
in order to achieve the target pressure ratio. This strategy is
obviously more interesting from the energy point of view, since
the corresponding energy requirement E will now apply only
on the permeate flowrate QP , that is
 (γ−1)/γ 
 γ · RT PIN
E = θQIN −1 (3)
γ −1 P 

with θ = QP /QIN the so-called stage cut. This possibility has been
already pointed out, for instance for industrial applications such
as O2 /N2 separation, when separation energy has to be min-
imized [33]. According to this alternative mode of operation,
actual membranes remain out of the scope of applicability for
a 10% mole fraction in the feed. For a 20% mole fraction how-
ever, existing membrane materials (with a selectivity of 60),
would generate an energy requirement E of only 0.7 MJ/kg,
with a downstream pressure of 30 mbar, which is technically
realistic and achievable by classical vacuum pumps (such as
those used for pervaporation operation, for instance). Ambi-
tious international research programs aim at decrease the energy
requirement of absorption processes by a factor 2 [34,51,52],
that is around 2 MJ/kg. Very promising experimental results,
obtained with membrane contactors, in the range of 0.76 MJ/kg
(electrical energy consumption) have been also recently reported
[53]. Furthermore, in a recent study performed with an integrated
Fig. 5. Energy requirement (circles, in MJ/kg of recovered CO2 ) and corre- system based on membrane contactors, an energy requirement
sponding pressure ratio (squares, P /P ) vs. membrane ideal selectivity (α*)
for a CO2 recovery ratio of 80% and a CO2 permeate mole fraction of 0.8
down to 0.26 MJ/kg was estimated [55].
obtained by an isothermal module fed by a binary CO2 /N2 mixture and operated These results call for a more systematic exploration of
under cross plug flow mode. Filled symbols correspond to 10% CO2 in feed membrane permeation versus absorption processes for post-
mixture, open symbols to 20% CO2 in feed mixture. (a) Module operated with combustion applications. Nevertheless, they show that, for some
feed compression (P = Pin = 1 atm). (b) Module operated with vacuum pumping situations of industrial interest, gas permeation performances
(P = Pin = 1 atm).
can be close to those of absorption processes from the energy
requirement point of view.
Several additional comments should be made at this point:
conditions, the overall energy requirement of the membrane pro-
cess will be far above the 4 MJ/kg of the absorption data. A value
around 8 MJ/kg is obtained. For sake of comparison, it has to be (i) At first glance, the fact that a vacuum operation is chosen
noted that about 1.3 kWh of power (4.5 MJ) are produced in a may seem unrealistic, since carbon dioxide has to be further
power plant for every kg of CO2 emitted. compressed to 100–150 bar for transportation and seques-
The previous statement, which corroborates the results of tration. Nevertheless, an optimal pressure ratio can be
previously reported analysis in the field [9,10], is clearly an easily identified for a given set of conditions and suggests
argument which has justified membrane permeation processes a vacuum operation combined to a moderate compression,
to be discarded. Nevertheless, Fig. 5a also shows that this figure in order to minimize the overall energy requirement (cap-
has to be completely reassessed as soon as the CO2 mole fraction ture + transportation) [32].
in the feed mixture increases. For xin = 0.2, a value which cor- (ii) More generally, vacuum operation remains seldom chosen
responds to numerous situations in post-combustion processes in industry, due to the large footprint of pumps compared to
such a cement or steel factories [3], the target performances can compressors, and also because of the low energy efficiency
be achieved providing that a selectivity above 40 is achieved. For of vacuum machines [55,58], especially when a high level
a selectivity of 60, which has been shown to correspond to the of vacuum is needed. These technological limitations have
actual threshold in the previous section, the associated energy major incidences in the capital and operating costs of a
requirement corresponds to the amine absorption specificities. carbon capture plant and should be taken into account since
It is important to note that increasing membrane selectivity only they could be decisive penalties.
56 E. Favre / Journal of Membrane Science 294 (2007) 50–59

(iii) A comparison based on a strict energy requirement criterion in order to claim the inadequacy of membrane permeation in
has to be taken with care since absorption makes use of a post-combustion carbon capture.
low value energy source (heat provided by steam), while Given the equipment size associated to the level of feed
membrane modules require electricity. Nevertheless, even rates, membrane contactors have been suggested and are inten-
if a factor 3 is incorporated in the above figure in order to sively investigated in order to achieve process intensification and
take into account the difference between heat and electricity decrease the size of the absorption unit. A factor around 2 can
[35] (that is 2.1 MJ/kg in place of 0.7), membranes remain be obtained based on this strategy [1,9]. Table 2 summarizes the
very interesting. main characteristics of packed columns and membrane contac-
(iv) A single stage adiabatic compression has been assumed tors for CO2 capture. It can be seen that the overall CO2 transfer
in the power estimation (Eqs. (1) and (2)). A significant rates are 0.5 and 1 mol m−3 s−1 respectively for these two types
decrease in energy requirement can be practically achieved of processes [41]. The fact that two towers or membrane contac-
with multistage compressors [32]. This strategy is often tors (absorption and stripping) are needed for this type of process
chosen for industrial applications. Additionally, for indus- suggests to divide the former data by a factor of 2, in order to
trial compression design, turbine expanders can also be take into account the total working volume of the unit. Thus,
considered to recover up to 80% of the theoretical energy 0.5 mol m−3 s−1 can be seen as the actual upper intensification
of gas compression. threshold of CO2 capture units in post-combustion.
A rough comparison has been achieved towards this indicator
It appears at this stage that already existing membrane mate- for membrane permeation processes. Assuming a permeabil-
rials may offer promising performances for post-combustion ity of 500 Barrer, an active layer thickness of 0.1 ␮m and an
carbon dioxide recovery, as soon as the inlet mole fraction average driving force of 0.5 bar (which would correspond for
exceeds 0.2. The key specific energy consumption factor hav- instance to a compression slightly above 5 bar on the upstream
ing been solely taken into account, it might be that the practical side with xin = 0.2 and moderate vacuum pumping on the down-
utilization of membranes remain prohibited due to an excessive stream side), one ends up with an overall transfer rate around
size (or number of membrane modules) for this application. This 0.1 mol CO2 m−3 s−1 . According to this rough computation,
point is explored hereafter. dense membrane recovery units would compete with membrane
contactors in terms of the size of the required unit, providing that
the permeability of the most selective polymers is increased by
5. The productivity/design challenge a factor of 4–6. A thinner active membrane layer could also be
simultaneously aimed in order to achieve that purpose. Coming
The energy requirement of the capture process is a major back to the trade-off curves (Fig. 3), this target may not be so
issue since it will drive to a large extent the corresponding oper- unrealistic. Increasing membrane permeability is indeed usually
ating costs. Nevertheless, similarly to any design project, the considered as easier to achieve than increasing selectivity [11].
investment (or capital) costs have obviously to be also carefully More generally, the energy requirement/membrane area
estimated. Again, the author notices that such a comparative trade-off will be primarily governed by the compression strat-
study between absorption and gas permeation membranes has egy. Compression will require a much smaller membrane area
not been reported up to now for the post-combustion situation. than vacuum, the difference increasing with a smaller pressure
Basically, amine absorption is classically achieved in a ratio. Very cheap membrane materials would thus be required to
packed column, which, for large emission sources, requires such make vacuum operation possible.
large units that multicolumn systems have to be considered. Typ-
ical inlet flowrates to be treated in post-combustion concern 6. Forthcoming issues and prospects
indeed emission sources that classically range between 1000
and 10,000 tonnes of gas mixture per day. Even though gas per- In the above sections, a binary CO2 /N2 mixture has been pos-
meation membrane units are claimed to be inadequate for high tulated, similarly to a large majority of studies devoted to carbon
flowrate applications, it is important to note that applications in capture [3]. Nevertheless, the feed mixture shown in Table 1 is
the range of 700 MMSCFD, that is close to 1000 tonnes/day, definitely much more complex. Further computations and exper-
have been recently reported for natural gas treatment [15]. Con- iments have obviously to be performed in order to approach what
sequently, the argument of scale cannot be so directly used could be a real CCS situation.

Table 2
Comparative specific surface area and overall transfer capacity of absorption (packed column and membrane contactor) and membrane gas permeation processes
(from [11,36–41], adapted)
Packed column Membrane contactor Gas permeation module

(m−1 )
Specific surface area 30–300 ∼3000 5000–10,000
Overall CO2 transfer capacity (mol m−3 s−1 ) ∼0.5 ∼1 ∼0.1a
a Computation performed assuming a 0.1 ␮m thick membrane showing a permeability of 500 Barrer and inserted in a 10,000 m−1 specific surface area module

with an average driving force (P) of 0.5 bar.

 E. Favre / Journal of Membrane Science 294 (2007) 50–59 57

Among the secondary compounds existing in post-combus-

tion streams, water plays unambiguously a key role. It remains
in fact a major issue in CCS, particularly in the transportation
framework: while dry carbon dioxide does not corrode carbon
manganese steels, the presence of free water in the pipe promotes
corrosion and may induce hydrate formation [3]. Thus, it is of
importance to define the maximum acceptable water content of
the CO2 stream which exits the capture process. A broad range
of data can be found in the literature: 0.5 ppm [3], a few ppm
[42–44] up to 0.5% [20]. These various data show by themselves
how difficult it can be to achieve a clear picture of the water
removal specifications.
Similarly to the energy and capacity comparison performed
above, it is tempting to estimate to what extent membrane per-
meation could accommodate for water specifications. To that
respect, a first strategy consists to analyze the behavior of water Fig. 7. Retentate (x) and permeate (y) composition vs. stage cut (θ) for sin-
in a single stage membrane module, together with CO2 and gle stage membrane permeation of a ternary CO2 /H2 O/N2 gaseous mixture.
nitrogen. Apart from two studies [56,57], such a multicompo- Feed mixture (vol.%)—N2 : 0.85, CO2 : 0.10, H2 O: 0.05. Cross plug flow model,
pressure ratio ψ = 0.01. Computing method is detailed in [32]. Permeabilities
nent problem has not been addressed up to now and is highly (in Barrers) are supposed to be representative of a cellulose acetate membrane
prospective. First, a few comments and prospects can be made: ([22])—CO2 : 4.8, H2 0: 6000, N2 : 0.15.

(i) Given the high permeability of water in polymers, a drying Even though the above considerations call for a complex
step could be proposed before the CO2 capture module. modeling strategy for wet feed mixtures computations, a pre-
Unfortunately, water and CO2 permeability are systemati- liminary picture has been attempted and a sample of results is
cally linked, water being systematically faster than CO2 shown in Fig. 7. For sake of simplicity, the assumptions used
(Fig. 6). Thus, a membrane drying step is expected to in the above sections have been maintained: permeability has
simultaneously induce significant CO2 losses. been assumed to be constant, no coupling effect is taken into
(ii) If a single membrane module strategy is chosen, the strong account and isothermal conditions are postulated. Even though
transport coupling between CO2 and water, which has been these stringent hypotheses call for a careful conclusion, it is inter-
occasionally mentioned but remains to be properly mod- esting to notice that water could be removed in a very small part
eled, could play a key role [27]. A significant increase in of the entrance of the module. This specificity could enable an in
membrane selectivity could potentially result, according to situ drying step. The remaining part of the module could thus per-
this phenomenon, which could be of interest. form the CO2 /N2 separation. This rough computation recalls that
(iii) The influence of a large water permeation flux on the reten- multicomponent gas permeation studies with selectivity among
tate temperature should also be considered. The assumption gas pairs which differ largely have not been investigated in detail
of an isothermal module could possibly have to be recon- up to now, both from the experimental and theoretical point of
sidered based on water heat of condensation and Joule view. Given the challenge addressed by post-combustion CO2
Thomson coefficient. capture, it is hoped that this gap will be filled in the near future.
Finally, it is important to note that the specifications men-
tioned in Table 1, which have governed to a large extent the
analysis developed in this paper, deal with an enhanced oil
recovery (EOR) situation. This type of application is indeed
more stringent, concerning for instance the presence of inert
gases or the water content. These specifications have however
to be reconsidered for sequestration, for which CO2 purity, inert
gases or water content are not so significant [3]. A modification
on water or CO2 purity constraints would call for a complete
reassessment of the analysis performed in this paper. Neverthe-
less, it can be anticipated that a less stringent target on CO2
purity, in order to minimize the capture costs, would play in
favor of recovery operation by dense membranes.

7. Conclusion

Fig. 6. H2 O/CO2 vs. CO2 /N2 selectivity for a series of polymers (data from The major objective of this note was to evaluate the pros
[22,23]). and cons of membrane operation versus absorption for post-
58 E. Favre / Journal of Membrane Science 294 (2007) 50–59

combustion CCS. The following key messages can be proposed

as a conclusion: Nomenclature
E overall membrane unit power consumption (W or
(i) For 10% CO2 feed mixtures, the energy requirement asso-
MJ/kg CO2 recovered)
ciated to membranes for post-combustion carbon capture
P permeability coefficient (mol m−1 s−1 Pa−1 )
is much larger than that of absorption. Even very selective
P membrane module upstream side pressure (bar)
membranes, showing selectivity above 120, require much
P membrane module downstream side pressure
more energy than absorption.
(bar)
(ii) Membranes potentially compete with absorption in terms
Pin feed mixture inlet pressure (bar)
of energy requirement as soon as CO2 content in the feed
Qin inlet total flow rate (mol/s)
mixture exceeds 20%. This situation corresponds to a large
QP permeate total flow rate (mol/s)
number of situations of industrial interest (cement factories,
R carbon dioxide recovery ratio
steel production, . . .).
R perfect gas constant (8.314 J mol−1 K−1 )
(iii) The target membrane selectivity required for the latter case
T temperature (K)
is around 60. This has been already achieved by the most
xin mole fraction in the feed mixture
promising membrane materials. Nevertheless, increasing
yp mole fraction in the permeate side
membrane selectivity does not change significantly energy
requirements in the range of variables which has been cov- Greek letters
ered. α* ideal separation factor
(iv) At the contrary, applying vacuum pumping in place of γ adiabatic gas expansion coefficient
upstream compression induces a drastic decrease of the η isentropic efficiency
energy requirements of membrane units. This option has ψ pressure ratio (P /P )
to be combined to the further compression step needed for θ stage cut
transportation but a global analysis call for a compromise
which make use of moderate vacuum in the membrane mod-
ule [32]. For a 20% CO2 content, the energy requirement
decreases down to 0.75 MJ/kg of CO2 recovered, which is References
close to the best absorption performances.
[1] H. Herzog, What future for carbon capture and sequestration? Environ. Sci.
(v) In terms of unit size, membrane units in post-combustion
Technol. 35 (7) (2001) 148–153.
applications offer an overall CO2 transfer rate which [2] C.M. White, Separation and capture of CO2 from large stationary sources
is approximately four to six times smaller than that of and sequestration in geological formations, J. Air Waste Manage. Assoc.
membrane contactors. An improvement in membrane per- 53 (2003) 645–715.
meability of the most selective polymers, together with the [3] O. Davidson, B., Metz, Special Report on Carbon Dioxide Capture and Stor-
age, International Panel on Climate Change, Geneva, Switzerland, 2005.
production of thin active layers, would be of utmost interest
www.ipcc.ch.
in order to achieve intensified operation. [4] J. Davison, K. Thambimuthu, Technologies for capture of carbon dioxide,
(vi) The role of water, which has been seldom taken into account in: Proceedings of the Seventh Greenhouse Gas Technology Conference,
up to now, should be systematically explored in order to Vancouver, Canada. International Energy Association (IEA), Greenhouse
identify the best type of operation which can fit the target Gas R&D Progamme, 2004, www.ghght7.ca.
[5] M. Simmonds, P. Hurst, M.B. Wilkinson, C. Watt, C.A. Roberts, A study of
water residual concentration.
very large scale post combustion CO2 capture at a refining & petrochemical
complex, 2003. www.co2captureproject.org.
Given the absence of pilot scale experiments with membranes [6] E. Drioli, M. Romano, Progress and new perspectives on integrated mem-
for post-combustion carbon capture, this note is necessarily brane operations for sustainable industrial growth, Ind. Eng. Chem. Res.
highly prospective. Additionally, solvent processes have been 40 (2001) 1277–1300.
used for decades in industry; they can deliver high quality CO2 [7] S.P. Kaldis, G. Skrodas, G.P. Sakellaropoulos, Energy and capital cost anal-
ysis of CO2 capture in coal IGCC processes via gas separation membranes,
and have economic value. Nevertheless, it is hoped that this study Fuel Process. Technol. 85 (2004) 337.
will stimulate efforts of membranologists towards a challenging [8] A. Corti, D. Fiaschi, L. Lombardi, Carbon dioxide removal in power gen-
application and convince them that selectivity is not necessarily eration using membrane technology, Energy 29 (2004) 2025.
the only issue in this field. [9] P.H.M. Feron, A.E. Jansen, CO2 separation with polyolefin membrane con-
tactors and dedicated absorption liquids: performances and prospects, Sep.
Purif. Technol. 27 (2002) 231–242.
Acknowledgements [10] J.P. Van der Sluis, C.A. Hendriks, K. Blok, Feasability of polymer mem-
branes for carbon dioxide recovery from flue gases, Energy Convers.
This work has been achieved thanks to a financial support Manage. 33 (5-8) (1992) 429.
from CNRS (ACI Carbomem grant) and Région Lorraine. The [11] R.W. Baker, Membrane Technology and Applications, 2nd ed., John Wiley
decisive contributions of Roda Bounaceur, Denis Roizard and & Sons, Chichester, England, 2004.
[12] W.J. Koros, G.K. Fleming, Membrane based gas separations, J. Membr.
Nancy Lape in frame of this project are gratefully acknowl- Sci. 83 (1993) 1–80.
edged. The author also thanks the reviewers for their valuable [13] R.L. McKee, M.K. Changela, G.J. Reading, CO2 removal: membrane plus
comments. amine, Hydrocarbon Process. 4 (1991) 63.
 E. Favre / Journal of Membrane Science 294 (2007) 50–59 59

[14] J. Hao, P.A. Rice, S.A. Stern, Upgrading low-quality natural gas with H2 S [38] S. Nii, H. Takeuchi, Gas absorption with membrane permeation acid gas
and CO2 selective membrane, J. Membr. Sci. 209 (2002) 177–206. removal from flue gases by a permabsorption method, Inst. Chem. Eng. 72
[15] D. Dortmund, K. Doshi, CO2 removal membrane technology: recent devel- (1994) 21–26.
opments, Chem. Eng. World (2003) 55–66, September Issue. [39] K. Li, W.K. Teo, Use of permeation and absorption methods for CO2
[16] R.D. Noble, R. Agrawal, Separation research needs for the 21st century, removal in hollow fibre membrane modules, Sep. Purif. Technol. 13 (1998)
Ind. Eng. Chem. Res. 44 (2005) 2887. 79–88.
[17] R.W. Baker, Future directions of membrane gas separation technology, Ind. [40] H.A. Rangwala, Absorption of carbon dioxide into aqueous solutions using
Eng. Chem. Res. 41 (2002) 1393. hollow fiber membrane contactors, J. Membr. Sci. 112 (1996) 229–240.
[18] A. Veawab, P. Tontiwachwuthikul, A. Chakma, Corrosion behaviour of [41] S.H. Yeon, B. Sea, Y.I. Park, K.H. Lee, Determination of mass transfer rates
carbon steel in the CO2 absorption process using aqueous amine solutions, in PVDF and PTFE hollow fiber membranas for CO2 absorption, Sep. Sci.
Ind. Eng. Chem. Res. 38 (1999) 3917. Technol. 38 (2) (2003) 271.
[19] K. Yamamoto, K. Hosoya, Material problems in amine processes for CO2 [42] J.C.M. Farla, C.A. Hendriks, K. Blok, Carbon dioxide recovery from indus-
recovery, Trans. Mater. Res. Soc. Jpn. 18A (1994) 321. trial processes, Energy Convers. Manage. 36 (6-9) (1995) 827–830.
[20] S. Bachu, W.D. Gunter, Acid gas injection: A study of existing operations, [43] P. Jaud, R. Gros-Bonnivard, M. Kanniche, E. Amantini, T. Manai, C. Boual-
Report Number PH4/15, International Energy Agency, 2003. lou, C. Descamps, Technico-economic feasability study of CO2 capture,
[21] B.D. Freeman, Basis of permeability/selectivity tradeoff relations in poly- transport and geo-sequestration: a case study for France, in: Proceedings of
meric gas separation membranes, Macromolecules 32 (1999) 375–380. the Seventh Greenhouse Gas Technology Conference, Vancouver, Canada,
[22] C.E. Rogers, Solubility and diffusivity, in: D. Fox, M.M. Labes, A. Weiss- International Energy Association (IEA), Greenhouse Gas R&D Progamme,
berger (Eds.), Physics and Chemistry of the Organic Solid State, vol. 2, 2005, www.ghght7.ca.
Interscience Publishers, 1965. [44] O. Skovholt, CO2 transportation system, Energy Convers. Manage. 34
[23] J. Bandrup, E.H. Immergut, Polymer Handbook, John Wiley & Sons, New (9–11) (1993) 1095.
York, 1988. [45] T.M. Ho, G. Leamon, G.W. Allinson, D.E. Wiley, Economics of CO2 and
[24] H. Lin, B.D. Freeman, Gas solubility, diffusivity and permeability in mixed gas geosequestration of flue gas using gas separation membranes,
poly(ethylene oxide), J. Membr. Sci. 239 (2004) 105. Ind. Eng. Chem. Res. 45 (2006) 2546–2552.
[25] C. Charmette, J. Sanchez, P. Gramain, A. Rudatsikira, Gas transport proper- [46] D. Aaron, C. Tsouris, Separation of CO2 from flue gas: a review, Sep. Sci.
ties of poly(ethylene oxide-co-epichlorhydrin) membranes, J. Membr. Sci. Technol. 40 (2005) 321–348.
230 (2004) 161. [47] C.E. Powell, G.G. Qiao, Polymeric CO2 /N2 gas separation membranes for
[26] T. Nakagawa, T. Nishimura, A. Higuchi, Morphology and gas permeability the capture of carbon dioxide from power plant flue gases, J. Membr. Sci.
in copolyimides containing polydimethylsiloxane block, J. Membr. Sci. 206 279 (2006) 1–49.
(2002) 149. [48] J. Zou, W. Ho, CO2 selective polymeric membranes containing amines in
[27] F. Piroux, E. Espuche, R. Mercier, The effects of humidity on gas transport crosslinked poly(vinyl alcohol), J. Membr. Sci. 286 (2006) 310–321.
properties of sulfonated copolyimides, J. Membr. Sci. 232 (2004) 115. [49] D.G. Jandel, Select the right compressor, Chem. Eng. Progr. 7 (2000)
[28] F. Piroux, E. Espuche, R. Mercier, M. Pinéri, G. Gebel, Gas transport mech- 15–29.
anism in sulfonated polyimides. Consequences on gas selectivity, J. Membr. [50] G. Göttlicher, The Energetics of Carbon Dioxide Capture in Power Plants,
Sci. 209 (2002) 241. US Department of Energy, National Energy Technology Laboratory, 2004.
[29] T. Yamaguchi, C.A. Koval, R.D. Noble, C.N. Bawman, Transport mech- [51] T. Mimura, H. Shimayoshi, T. Suda, M. Iijima, S. Mitsuoka, Development
anism of carbon dioxide through perfluorosulfonate ionomer membranes of energy saving technology for flue gas carbon dioxide recovery in power
containing an amine carrier, Chem. Eng. Sci. 54 (1996) 4781–4789. plant by chemical absorption method and steam system, Energy Convers.
[30] H. Matsuyama, A. Terada, T. Nakagawara, Y. Kitamura, M. Teramoto, Manage. 38 (1997) S57–S62.
Facilitated transport of CO2 through polyethylenimine/poly(vinyl alcohol) [52] T. Mimura, S. Satumi, Kagaku Kogaku Ronbunshu (Chem. Eng. in
blend membrane, J. Membr. Sci. 163 (1999) 221–227. Japanese) 24 (1998) 546–551.
[31] R. Zolandz, G.K. Fleming, Gas permeation applications, in: W.S. Ho, K.K. [53] S.-H. Yeon, K.-S. Lee, B. Sea, Y.-I. Park, K.-H. Lee, Application of pilot-
Sirkar (Eds.), Membrane Handbook, Van Nostrand Reinhold, New York, scale membrane contactor hybrid system for removal of carbon dioxide
1995. from flue gas, J. Membr. Sci. 257 (2005) 156–160.
[32] R. Bounaceur, N. Lape, D. Roizard, C. Vallières, E. Favre, Membrane [54] M. Teramoto, S. Kitada, N. Ohnishi, H. Matsuyama, N. Matsumiya, Sep-
processes for post-combustion carbon dioxide capture: a parametric study, aration and concentration of CO2 by capillary-type facilitated transport
Energy 31 (14) (2006) 2220–2234. membrane module with permeation of carrier solution, J. Membr. Sci. 234
[33] W.J. Koros, Gas permeation applications, in: R.W. Baker, E.L. Cussler, W. (2004) 83–94.
Eykamp, W.J. Koros, R.L. Riley, Strathmann (Eds.), Membrane Separation [55] N. Matsumiya, M. Teramoto, S. Kitada, H. Matsuyama, Evaluation of
Systems, Noyes Data Corporation, 1991. energy consumption for separation of CO2 in flue gas by hollow fiber
[34] P. Deschamps, P.A. Pilavachi, Research and development actions to reduce facilitated transport membrane module with permeation of amine solution,
CO2 emissions within the European Union, Oil Gas Sci. Technol. 59 (3) Sep. Purif. Technol. 46 (2005) 26–32.
(2004) 323–330. [56] K. Haraya, M. Nakaiwa, N. Itoh, C. Kamisawa, Feasibility study of the
[35] W.J. Koros, Evolving beyond the thermal age of separation processes: application of membrane separation in CO2 removal from flue gases,
membranes can lead the way, AIChE J. 50 (10) (2004) 2326. Kagaku Kogaku Ronbunshu (Chem. Eng. in Japanese) 19 (1993) 714–721.
[36] A. Gabelman, S.-T. Hwang, Hollow fiber membrane contactors, J. Membr. [57] N. Matsumiya, N. Inoue, H. Mano, K. Haraya, Effect of water vapor
Sci. 159 (1999) 61–106. on CO2 separation performance on membrane separator, Kagaku Kogaku
[37] H.B. Al-Saffar, J.S. Oklany, B. Ozturk, R. Hughes, Removal of carbon Ronbunshu (Chem. Eng. in Japanese) 25 (1999) 367–373.
dioxide from gas streams using a gas/liquid hollow fibre module, Trans. [58] H. Silla, Chemical Process Engineering, in: Compressors, Pumps and Tur-
Inst. Chem. Eng. 73 (1995) 144–150. bines, Marcel Dekker, Inc., New York, 2003 (Chapter 5).