Journal of Membrane Science 428 (2013) 251–259

Contents lists available at SciVerse ScienceDirect

Journal of Membrane Science

journal homepage: www.elsevier.com/locate/memsci

High-performance ester-crosslinked hollow fiber membranes

for natural gas separations
Canghai Ma, William J. Koros n
School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, 311 Ferst Drive, Atlanta, GA-30332, United States

a r t i c l e i n f o abstract

Article history: A significantly improved defect-free ester-crosslinkable hollow fiber membrane was developed through
Received 16 August 2012 simultaneous optimization of the spinning solution and dry-jet/wet-quench spinning process variables.
Received in revised form The resultant ester-crosslinked hollow fibers show extremely high separation productivity and efficacy
10 October 2012
with CO2 plasticization resistance under aggressive feed conditions. The CO2 permeance was improved
Accepted 14 October 2012
Available online 26 October 2012
from 50 GPU to 117 GPU with a CO2/CH4 selectivity of 37 in testing at 200 psi with 50/50 CO2/CH4 feed,
35 1C. Moreover, the crosslinked hollow fibers maintain a high CO2 permeance under highly aggressive
Keywords: feed pressures up to 800 psi of 50/50 CO2/CH4 feed without CO2 plasticization. A lower operating
Hollow fiber temperature can improve the CO2/CH4 selectivity significantly without apparent loss of CO2 permeance.
Crosslink
The mixed gas permeation with high CO2 feed pressure demonstrates that ester-crosslinked hollow
Plasticization
fiber membranes provide a significant advance in the state of the art for CO2/CH4 separations.
Natural gas
Separation & 2012 Elsevier B.V. All rights reserved.

1. Introduction cost of membranes in gas separation technology due to the high

As one of the most important fuel sources worldwide, raw
natural gas often contains an unacceptably high level of carbon
dioxide (CO2), which must be reduced to a level of o2% to avoid
problems caused by CO2 induced pipeline corrosion, compression
cost and a reduction of heating value [1]. The main technologies
currently used to separate CO2 from natural gas include cryogenic
distillation, amine absorption and membranes [1–3]. Cryogenic
distillation is highly energy intensive because gases must be
cooled down to realize the CO2/CH4 separation [4]. While amine
absorption removes CO2 almost completely, high capital cost,
complex operation, expensive maintenance and corrosion makes
this technology problematic [3]. Membrane separation technology
described here can overcome the key operational problems
associated with cryogenic distillation and amine absorption pro-
cesses [5,6].
Many materials can be used to produce membranes, including
polymers, carbon molecular sieves (CMS), zeolites and ceramics
[3]; however, polymers are the dominant membrane materials
used for natural gas separations. Polymeric hollow fiber mem-
brane devices were developed as an alternative to flat sheet and
spiral wound membrane devices and were first commercialized
by Monsanto in 1977[7]. Asymmetric hollow fiber membranes are
industrially preferred to improve the productivity and reduce the
n
Corresponding author. Tel.: þ1 404 385 2845; fax: þ1 404 385 2683.
E-mail address: bill.koros@chbe.gatech.edu (W.J. Koros).
surface area to volume ratios up to 10,000 m2/m3[8]. The cylind-
rical structure of the hollow fibers can withstand high transmem-
brane pressure difference up to 1000 psi [9]. Despite its excellent
separation performance, high CO2 partial pressure induced plas-
ticization tends to cause polymeric membranes to show lower
CO2/CH4 separation efficiency and loss of CH4 into the permeate.
To mitigate this problem, covalent ester crosslinking has been
investigated and shown to stabilize polymer membranes against
CO2 plasticization by suppressing the degree of swelling and
segmental chain mobility in the polymer [10–13]. This study
focuses upon extending the success of crosslinking to even more
productive hollow fibers with enhanced CO2 plasticization resis-
tance to meet the industrial application requirements.
2. Background and theory
2.1. Gas transport mechanism in membranes
Several mechanisms have been proposed for gas transport in
polymeric membranes [5,6,9]; however, the so-called sorption–
diffusion is the most preferred. The sorption–diffusion model first
qualitatively described by Graham in 1831 is the most widely
accepted description of gas molecule transport in nonporous
polymeric membranes [5,14]. In this model, the permeants first
sorb in the membrane material and then diffuse through the
membrane under a partial pressure difference. The differences in
the solubility of gas in the membrane and the rate of permeants

0376-7388/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.memsci.2012.10.024
252 C. Ma, W.J. Koros / Journal of Membrane Science 428 (2013) 251–259
diffusion through the membrane cause the different gases to be
separated. In this case, the permeability coefficient, P, of a
polymer membrane can be described by the product of the
diffusion coefficient, D, and sorption coefficients, S, as shown in
the following equation [5]:
P¼DS
2.2. Characterization of membrane performance
To characterize the performance of a membrane, two key
factors are commonly studied: permeability and selectivity. The
permeability represents the intrinsic productivity of a membrane
and is defined by the flux of penetrant i, normalized by the
membrane thickness and the partial pressure difference across
the membrane, as shown in the following equation [5]:
ni  l
Pi ¼
Dpi
In Eq. (2), ni represents the flux of penetrant i through the
membrane; l refers to the membrane thickness, and Dp describes
the partial pressure or fugacity difference of each penetrant
across the membrane. The common unit of membrane perme-
ability is the Barrer, shown as
 
ccðSTPÞcm
Barrer ¼ 1010 ð3Þ
cm2 s cm Hg
In an asymmetric hollow fiber membrane, the actual mem-
brane thickness is difficult to define unambiguously; therefore,
the productivity of a hollow fiber membrane is usually described
by the permeance, Pi/l, which is the flux of penetrant i normalized
by partial pressure or fugacity difference, as shown in the
following equation:
Pi n be explained by Fig. 1.
¼ i
l Dpi
The common unit of permeance is the GPU [15], which is
defined as
 
ccðSTPÞ
GPU ¼ 106 ð5Þ
cm2 s cm Hg
The ideal selectivity measures the intrinsic membrane separa-
tion efficacy. For a given pure gas pair, if the total upstream
pressure is much larger than the total downstream pressure, the
ideal selectivity, aij, is defined by the ratio of the fast gas (i)
permeability or permeance to the slow gas (j), as shown in the
following equation:
Pi P =l
aij ¼ ¼ i
Pj Pj =l
Practically, complications including plasticization, non-infinite
pressure ratio and competition among gas molecules may occur
for mixed gas permeation. Therefore, the separation factor, (S.F.)ij,
is commonly used to evaluate the separation performance, as
shown in the following equation [15]:
yi =yj
ðS:F:Þij ¼
xi =xj
Fig. 1. Ester-crosslinking mechanism showing the formation of new ester bonds between polymer chains.
In Eq. 7, x and y represent the mole fraction of the penetrants
in the upstream and downstream of the membrane, respectively;
and i and j means the different penetrants in the mixed gases. The
aij and (S.F.)ij become equal when the ratio of upstream pressure
to downstream pressure is very high. The relative permeance
ð1Þ ratio is preferred to measure a membrane’s intrinsic separation
performance, since the (S.F.)ij is affected by the ratio of feed to
permeate pressure, which complicates polymer separation
properties.
2.3. Plasticization and crosslinking
The plasticization of a polymeric membrane is often observed
when a high CO2 partial pressure is fed on the upstream, which
increases the permeance and reduces selectivity significantly. In
the presence of plasticization, the penetrant–polymer interaction
increases the segmental mobility of polymer chains and causes an
ð2Þ
increase of diffusion coefficients of penetrants, thereby increasing
the permeance [16–18]. When the plasticization pressure is
reached, the selectivity decreases with the pressure. The loss of
selectivity is due to the relatively larger increase of permeance for
the slow CH4 than the fast CO2. To develop a practical membrane
with sufficient separation performance, the plasticization must be
suppressed to achieve a high permeance without a loss of
selectivity.
Different strategies can be utilized to stabilize polymer mem-
branes against CO2 induced plasticization, including polymer
thermal annealing [19,20], blending [21,22] and crosslinking
[11,17,23–25]. A highly effective approach, covalent ester-cross-
linking, has been shown to improve the CO2 plasticization
resistance by reducing the degree of swelling and segmental
chain mobility in the polymer. The crosslinking mechanism can
ð4Þ Notwithstanding the advantages in stabilizing polymeric
membrane against plasticization, crosslinking can cause a loss
of CO2 permeance, which undermines the separation productivity
for crosslinked hollow fibers [15]. In such a case, crosslinking is
believed to also cause a densification of transition layer of hollow
fibers, thereby increasing the transport resistance and essentially
reducing the permeance. Therefore, this research is focusing on
development of the most productive asymmetric hollow fiber
membranes with the thinnest skin layer thickness to achieve the
desirable highest separation performance potential of crosslinked
hollow fibers.
2.4. Asymmetric hollow fiber membranes
ð6Þ
Typically, asymmetric hollow fiber membranes can be spun
using a dry-jet/wet-quench hollow fiber formation process [6].
The homogeneous polymer solution (called dope) is co-extruded
with bore fluid through an annular die, called a spinneret, into an
aqueous quench bath. During the dry-jet period, the volatile
components evaporate from the dope and increase the polymer
concentration in the outer layer of nascent fibers. When the fiber
ð7Þ enters the quench bath, phase separation occurs rapidly and the
dope demixes into a polymer rich phase and a polymer lean
 C. Ma, W.J. Koros / Journal of Membrane Science 428 (2013) 251–259
Polymer
Vitrified
region
B’ Binodal
Spinodal
Two-phase
A B”
One-
phase
Solvent Non-Solvent
Fig. 2. Ternary phase diagram showing the skin formation. The solid line A-B0
represents the trajectory of the skin layer formation during the dry-jet step and
the dashed line A-B00 describes the trajectory of the substructure formation.
phase. The solvent diffuses from fibers into the quench bath while
the non-solvent from the quench bath diffuses into the fibers.
A solid skin and a porous structure are formed during this phase
separation. The formation of the outermost skin layer can be
depicted by the ternary phase diagram shown in Fig. 2 [26].
In Fig. 2, Point A represents the initial polymer concentration
of the dope. The increase of polymer concentration in the outer
surface of the fiber is represented by point B0 , which can be
caused by the evaporation of solvent and/or non-solvent, and
moisture adsorption from the air. B00 describes the lower local
solvent concentration within the support layer due to the influx of
non-solvent in the quench bath. The solid line from A-B0
represents the path leading to the formation of the outer skin
layer of fibers, while the dashed line A-B00 represents the
formation of an open and porous substructure of fibers.
Nucleation and growth is a possible phase separation mechan-
ism for the formation of a defect-free skin layer [27]. In this
scenario, the nascent skin of hollow fiber remains as a single
phase material without undergoing phase separation, while
polymer lean domains are nucleated and grow in the substrate
region to intersect and create a low substrate resistance; how-
ever, if the intersection and rupture of the polymer lean domains
fails to occur, closed cell that supports with undesirable high
transport resistance are formed. Ideally, the formation of support
structure can sometimes involve so-called ‘‘spinodal decomposi-
tion’’, which leads to desirable naturally bicontinuous polymer
lean and polymer rich phases. The loss of solvent and non-solvent
occurs in the region underneath the skin layer to produce an
interconnected open support with low substrate resistance [5].
3. Experiments
3.1. Materials
The starting material used in this study is a polyimide
copolymer, called 6FDA/DAM: DABA (3:2), as shown in Fig. 3.
This copolymer structure and monomer ratio are chosen based on
its excellent CO2/CH4 separation performance and high CO2
induced plasticization resistance, which has been investigated
by previous researches [11–13]. The 6FDA/DAM: DABA (3:2)
polymer is synthesized via two step polycondensation and che-
mical imidization from 4, 40 -(hexafluoroisopropylidene) diphtha-
lic anhydride (6FDA), 2, 4, 6-trimethyl-1, 3-diaminobenzene
(DAM) and 3, 5-diaminobenzoic acid (DABA), as described in
reference [28]. The acid pendant group of DABA units in backbone
253
of the copolymer is additionally partially monoesterified with 1,
3-propanediol to form the propane-diol monoesterified cross-
linking polyimide, called PDMC polyimide. The PDMC contains
both carboxylic acid groups and ester groups, which have the
potential to form new ester bonds; thereby crosslinking the
polymer chains and stabilizing the polymer against plasticization.
This two-step monoesterification reaction is shown in Fig. 3.
3.2. Dope development
A typical spinning dope consists of polymer, solvent and non-
solvent. The solvent dissolves the polymer to form a homoge-
neous single phase solution. N-methylpyrrolidone (NMP) is
selected in this work as the solvent due to its low volatility, low
toxicity and high boiling point. In addition, tetrahydrofuran (THF)
is added in the dope to promote the formation of the skin layer
during spinning [30]. Non-solvent is used to make the dope
composition near to the binodal to allow a rapid phase separation.
Ethanol and lithium nitrate (LiNO3) are chosen in this work
as non-solvents. Ethanol provides more flexibility to prepare
a spinnable dope and also affects the skin layer formation
due to its volatility. Lithium nitrate can modify the dope viscosity
and accelerate phase separation since it forms a complex with
NMP [29].
To develop and optimize the dope composition, the binodal of
the ternary phase diagram must be determined to ensure that the
reformulated dope composition locates in the one-phase region.
A straightforward technique to determine the binodal is through
the experimental method, called ‘‘cloud point determination’’ [8].
During this process, small dopes with different compositions
were prepared and the status of each dope (one-phase, two-
phase or cloudy) was determined by visual observation. The
binodal line lies in between the one-phase and two-phase
regions. A spinnable dope composition should be in the one-
phase region and close to the binodal line.
3.3. Asymmetric hollow fiber spinning
Typically, asymmetric hollow fibers are spun through a hollow
fiber spinning system, as shown in Fig. 4.
As mentioned above, the dope is co-extruded with bore fluid
through a spinneret. The nascent fibers enter the aqueous quench
bath and phase separation occurs. The hollow fibers are wound on
a rotating drum and soaked in water bath for  3 days. Solvent
and non-solvent must be removed from hollow fibers to avoid
subsequent fiber collapse. This is particularly important for
preserving the transition porous layer just beneath the skin layer.
The fibers from the quench bath contain water and must not be
dried directly since the capillary force in the fiber induced by
dehydration will collapse the pores and damage the fibers.
Therefore, solvent exchange must be conducted as the final
spinning step. Typically, alcohol (e.g. ethanol) is used first to
replace water in the fibers. Then the alcohol is washed out by
using a volatile non-solvent (e.g. hexane). Through solvent
exchange, the fiber surface tension is reduced significantly and
the fibers can be dried under vacuum and heating without
collapsing the fiber structure. Again, this is particularly important
for preserving the transition layer of hollow fibers.
3.4. Ester-crosslinking
After solvent exchange, the crosslinking reaction was carried
out by heating fibers in a heated vacuum for 2 h at 200 1C.
Dynamic vacuum serves as a driving force to promote the
completion of reaction by removing the byproduct, 1, 3-propane
diol, from the system. After crosslinking, the fibers can be potted
254 C. Ma, W.J. Koros / Journal of Membrane Science 428 (2013) 251–259

Fig. 3. Materials and monoesterification reaction to form PDMC polyimide [13].

Fig. 4. Dry-jet/wet-quench spinning to form asymmetric hollow fiber membranes.

into modules and characterized by Scanning Electron Microscope

(SEM) and gas permeation.

Fig. 5. SEM of a hollow fiber membrane from PDMC.

3.5. Hollow fiber membrane characterization

feed is generally not a factor for gas permeation, and in this study
3.5.1. Scanning Electron Microscope (SEM)
it has been totally avoided by maintaining very low ( o0.01)
To observe the substructure of asymmetric hollow fibers,
mixed gas stage cut [13]. The stage cut is determined by the ratio
Scanning Electron Microscope (SEM) is needed to provide a higher
of the permeate rate to feed flow rate, as shown in the following
magnification over 40,000x. The sample for SEM test can be
equation:
prepared by first soaking hollow fibers in hexane and then
cryogenically fracturing fibers in liquid nitrogen to preserve their permeate molar flow rate
Stage cut ¼ ð8Þ
inner structure. SEM images can show the macrovoid, porous feed molar flow rate
structure and the thin skin layer of fibers. However, the effective
Mixed CO2/CH4 gas permeation under high feed pressures is
skin layer thickness is often estimated by gas permeation. A SEM
performed to study the gas separation performance under more
image of a hollow fiber membrane spun from PDMC in this work
realistic feed conditions in order to explore plasticization and
is shown in Fig. 5.
competition effects for natural gas.

3.5.2. Gas permeation

Gas permeation is used to determine the gas separation
performance of hollow fiber membranes. Under higher feed
pressure, the feed is often introduced into the shell side of a
hollow fiber module. Shell-fed system is generally preferred for
mixed gas permeation when elevated feed pressure is required
since only a loss in productivity, as opposed to an open ‘‘jet’’,
occurs due to a fiber failure. Concentration polarization in the
4. Results and discussions
4.1. Phase separation and dope development
With the identified polymer, solvents and non-solvents, a
ternary phase diagram must be constructed to prepare a spin-
nable dope. The ternary phase diagram shows the one-phase,
 C. Ma, W.J. Koros / Journal of Membrane Science 428 (2013) 251–259 255

PDMC Table 1
0 Hollow fiber spinning conditions (1st spin).

10 Spinning conditions
80

20 Dope extrusion rate 180 ml/h

70
Bore fluid composition NMP/H2O 80/20 wt%
30 Bore fluid rate 60 ml/h
60
Spinneret temperature 70 1C
40 Air gap 1–33 cm
50
Quench bath 50 1C
50 40 Take-up rate 50 m/min

60 30

70 20 Uncrosslinked
80 10

0
Solvent 20 30 40 50 60 70 80 90 100 Non-solvent

Fig. 6. Ternary phase diagram showing the binodal (black solid line) of PDMC
polymer/solvent/non-solvent system. Solid points and open circles represent one-
phase dope and two-phase dope, respectively.

two-phase region and the binodal. Due to the complicated

properties of polymer solution, a common and effective method
to determine the binodal is through the cloud point technique, as
described above. The results of cloud point experiments are
summarized and plotted in a ternary phase diagram, shown in
Fig. 6.
Fig. 7. Scanning electron micrographs of the cross section of hollow fibers,
In Fig. 6, the solid points represent the one-phase samples and showing the dense skin layer and the porous substructure for the uncrosslinked
open circles represent the two-phase samples. The binodal hollow fibers.
(shown as the black solid line) lies between the one-phase region
and two-phase region. Once the ternary phase diagram is plotted,
the dope composition for the hollow fiber spinning is chosen near Crosslinked
the binodal. Macroscopically, the dope must be in one-phase and
have a viscosity closer to thick honey (about 10,000 cP) [8]. It also
must have sufficient volatile solvent and non-solvent to form the
skin layer of hollow fibers. To modify the dope composition, a
common approach is tuning the ratio of two components in the
dope while fixing the composition of the third component. The
complex nature of polymer solutions and multiple requirements
that must be met require that the hollow fiber spinning includes
considerable trial and error in dope formulation to provide the
most productive hollow fiber membranes.

4.2. Defect-free asymmetric hollow fiber spinning

Previous studies show the PDMC polymer can be spun into

defect-free hollow fibers with a solution containing 35% polymer,
8.5% ethanol, 35% NMP, 6.5% LiNO3, and 15% tetrahydrofuran
(THF), which is essentially far away from the binodal determined
in Section 4.1. As a result, the crosslinked hollow fibers gave a CO2
permeance only  50 GPU using 200 psi of 50/50 CO2/CH4 mixed
gas feed [15]. This relatively low separation productivity limited
its application in the realistic operation and necessitated a
reformulation of dope composition to further the productivity.
Therefore, the 1st spin was done by reducing the polymer
concentration to 32% and moving the dope composition closer
to the binodal. The dope consists of 32% polymer, 32% NMP, 15.8%
Ethanol, 13.7% THF and 6.5% LiNO3. The spinning conditions for
the 1st spin are summarized in Table 1.
The spinning conditions shown in Table 1 are primarily based
on previous studies and spinning experience. The main spinning
parameters adjusted during the spinning is the air gap, which has
been identified as a key factor in determining the skin layer
formation and intrinsic separation properties. The hollow fibers
spun from this spinning were crosslinked by annealing the fibers
Fig. 8. Scanning electron micrographs of the cross section of hollow fibers,
showing the dense skin layer and the porous substructure for the crosslinked
hollow fibers.
at 200 1C for 2 h under vacuum. The cross-sections of hollow
fibers before and after crosslinking are shown in Figs. 7 and 8. The
skin layer and porous substructure are quite apparent in the SEM
images. The integrated dense skin layer and transition of cross-
linked hollow fibers suggests the crosslinking does not cause a
significant collapse of fiber inner wall.
The effect of air gap on the separation performance of hollow
fibers was studied by testing hollow fibers spun at different air
gaps, which are translated into different residence time. The
permeation results are shown in Figs. 9 and 10.
The CO2 and CH4 permeance trends in Fig. 9 show that the air
gap residence time affects the separation productivity signifi-
cantly. As expected, the CO2 and CH4 permeance decreases
drastically when increasing the air gap/residence time. The higher
CO2 permeance is translated to a thinner skin layer of hollow
256 C. Ma, W.J. Koros / Journal of Membrane Science 428 (2013) 251–259

10
140
CO2 permeance
120 8

CH4 permeance (GPU)

CO2 permeance (GPU)

CH4 permeance
100
6
80

60 4

40
2
20

0 0
0.0 0.1 0.2 0.3 0.4 0.5
Air gap residence time/s

Fig. 9. CO2 and CH4 permeance at different air gap residence time for 200 1C, 2 h
crosslinked hollow fibers. Air gap residence time determined by the ratio of air gap
to take-up rate. Test conditions: 50/50 CO2/CH4 at 200 psi, 35 1C. Permeance
uncertainty: 77%.
Fig. 11. CO2 permeance of crosslinked hollow fiber membranes at elevated feed
pressures from this work and literature. Permeance calculated by using fugacity.
Test conditions: 50/50 CO2/CH4, at 35 1C.

60 60

50 CO2 /CH4 permselectivity 50

CO2 /CH4 permselectivity

Separation Factor
Separation Factor

40 40

30 30

20 20

10 10

0 0
0.0 0.1 0.2 0.3 0.4 0.5
Air gap residence time/s

Fig. 10. CO2/CH4 separation efficacy at different air gaps for 200 1C, 2 h cross-
linked hollow fibers. Air gap residence time determined by the ratio of air gap to
take-up rate. Test conditions: 50/50 CO2/CH4 at 200 psi, 35 1C. Selectivity uncer-
tainty: 7 2%.
fibers, which is formed due to less evaporation of solvent/non-
solvent in a lower air gap. This similar phenomenon was also
demonstrated in reference [27]. On the other hand, the curves in
Fig. 10 reflect that a lower air gap does not necessarily cause an
apparent decline of the CO2/CH4 selectivity or separation factor. In
fact, the crosslinked hollow fibers display impressive separation
efficacy over a long range of air gaps. This flexibility suggests that
hollow fibers can be spun successfully with a thinner-skinned
layer at an air gap lower than 1 cm to achieve desirable superior
separation productivity without loss of selectivity.
4.3. Natural gas separation performance
The crosslinked hollow fibers with the highest separation
productivity obtained in the 1st spin were further evaluated by
using the 50/50 CO2/CH4 mixed gas at a feed pressure up to
800 psi. Figs. 11 and 12 show the mixed gas permeation results of
200 1C, 2 h crosslinked hollow fibers from the 1st spin.
The high-pressure permeation data in Fig. 11 shows that the
crosslinked hollow fibers demonstrated a stable CO2 permeance
and CO2/CH4 selectivity during pressurization and depressuriza-
tion. During pressurization, no upswing of permeance was found,
indicative of the absence of plasticization at the maximum test
pressure of 800 psi (400 psi of CO2 partial pressure). Instead, there
Fig. 12. CO2/CH4 selectivity of crosslinked hollow fiber membranes at elevated
feed pressures from this work and literature. Permeance calculated by using
fugacity. Test conditions: 50/50 CO2/CH4, at 35 1C.
is a slight drop of permeance, due to the dual mode sorption effect
at elevated feed pressures when the sorption advantage of CO2
over CH4 is reduced. During depressurization, the slight increase
of CO2 permeance indicates a certain amount of swelling of
hollow fibers; however, as indicated by Fig. 12, this small amount
of swelling does not cause a significant loss of separation efficacy
due to the interconnected matrix formed during crosslinking.
Compared to the literature fiber samples spun from 183,000 (Mw)
PDMC using the same dry-jet/wet-quench spinning process (CO2
permeance  50 GPU, as shown in Fig. 11) [15], the separation
productivity (CO2 permeance) of crosslinked hollow fibers was
enhanced over 100%. Details about literature spinning dope com-
position and spinning conditions are described in references
[13,15]. As shown in Fig. 12, the CO2/CH4 selectivity in this study
is slightly lower than the literature data (aCO2/CH4  40), which is
possibly due to different intrinsic separation properties of poly-
mers used. This difference is presumably due to the different
imidization and monoesterification processes, since chemical
imidization and two-step monoesterification was used in this
work vs. thermal imidization and one-pot monoesterification in
literature [15]. The side-reactions in chemical imidization and
depolymerization in thermal imidization may produce significantly
 C. Ma, W.J. Koros / Journal of Membrane Science 428 (2013) 251–259
different chemical structures of polymer backbones and essen-
tially affect the separation properties of polymers [30–32], which
should be studied in the future.
4.4. Dope optimization for hollow fibers
In the 1st spinning work, crosslinked hollow fibers with a CO2
permeance up to 100 GPU was achieved by using a solution
containing 32% polymer. To further improve the separation
productivity, a solution with even lower polymer and THF con-
centration were applied for the 2nd spinning, which consists of
30.5% polymer, 30.5% NMP, 19.5% Ethanol, 13.0% THF and 6.5%
LiNO3. The spinning conditions for this spin are summarized in
Table 2.
In the 2nd spin, more spinning parameters were studied to
probe the optimum spinning conditions, including the dope flow
rate, spinneret temperature, air gap and take-up rate. Again, the
hollow fibers spun from this spinning were crosslinked by
annealing the fibers at 200 1C for 2 h under vacuum. The cross-
linked hollow fibers with the highest separation productivity
were characterized by using the 50/50 CO2/CH4 mixed gas at a
feed pressure up to 800 psi. Figs. 13 and 14 show the mixed
permeation results of 200 1C, 2 h crosslinked hollow fibers from
the 1st spin, 2nd spin and literature [15].
As indicated in Fig. 13, the crosslinked hollow fibers from the
2nd spin demonstrate a CO2 permeance over 100 GPU at testing
pressures from 200 to 800 psi. Specifically, the fibers show a CO2
permeance of 117 GPU at 200 psi of 50/50 CO2/CH4, at 35 1C, which
was improved over 130% vs. literature data under the same permea-
tion conditions. Moreover, the CO2/CH4 selectivity is attractively
high over the long range of feed with a CO2/CH4 of 37 under 200 psi
Table 2
Hollow fiber spinning conditions (2nd spin).
Spinning conditions
Dope extrusion rate 120–180 ml/h
Bore fluid composition NMP/H2O 80/20 wt%
Bore fluid rate 40–60 ml/h
Spinneret temperature 50–70 1C
Air gap 1–33 cm
Quench bath 50 1C
Take-up rate 50–75 m/min
Fig. 13. CO2 permeance of crosslinked hollow fiber membranes at elevated feed
pressures from this work and literature [15]. Permeances calculated by using
fugacity. Test conditions: 50/50 CO2/CH4, at 35 1C. 2nd spin fiber permeance
uncertainty: 75%.
257
Fig. 14. CO2/CH4 selectivity of crosslinked hollow fiber membranes at elevated
feed pressures from this work and literature [15]. Selectivities calculated by using
fugacity. Test conditions: 50/50 CO2/CH4, at 35 1C. 2nd spin fiber selectivity
uncertainty: 7 7%.
of 50/50 CO2/CH4, 35 1C. The significant improvement of separation
productivity is primarily due to a much thinner skin layer formed
when a reformulated dope composition and spinning conditions
were applied in this work. It is also found that compared to the
slight increase of CO2 permeance at a pressure above 500 psi in the
literature, there is no upswing at the maximum test pressure of
800 psi for the crosslinked hollow fibers spun in this work,
suggesting an enhanced CO2 plasticization resistance was achieved
in this study. The difference of CO2 permeance depression from
200-800 psi for the 1st and 2nd spin fibers is only within 4%,
suggesting that the fiber samples spun from the two spins h
ave similar high CO2 plasticization resistance. Those impressive
permeation results reflect that the dope reformulation and spinning
conditions optimization can effectively enhance natural gas separa-
tion performance.
4.5. Effect of operating temperatures
The crosslinked hollow fibers from 2nd spin were further
characterized by using different operating temperatures from
room temperature (22 1C) to elevated temperature of 50 1C
to probe their performance under more aggressive conditions.
The CO2 permeances and selectivities are summarized in Figs. 15
and 16.
Fig. 15 shows that a higher operating temperature will
increase the CO2 permeance while decrease the CO2/CH4 selectiv-
ity. A lower operating temperature at 22 1C gave us a CO2
permeance of 110 GPU and a CO2/CH4 selectivity up to 44 at
200 psi. Moreover, this low testing temperature does not cause
any CO2 plasticization at the maximum feed pressure of 800 psi.
At elevated temperatures, the increased polymer chain mobility
increases diffusion coefficients for both CO2 and CH4. However,
the larger molecule, CH4, is affected to a larger extent over CO2; as
a result, the higher temperature decreases the diffusion advan-
tage of CO2 over CH4. Moreover, the sorption advantage of CO2
over CH4 is also reduced when increasing the temperature.
Therefore, a lower temperature is preferable for CO2/CH4
separation.
The temperature dependence of permeance can be explained
by the equation below [33]:
 
P P0 Ep
¼ exp ð9Þ
l l RT
258 C. Ma, W.J. Koros / Journal of Membrane Science 428 (2013) 251–259
Fig. 15. CO2 permeance of crosslinked hollow fibers from the 2nd spin at different
operating temperatures. Permeances calculated by using fugacity at corresponding
temperatures and pressures. Test conditions: 50/50 CO2/CH4, 22–50 1C.
Fig. 16. CO2/CH4 selectivity of crosslinked hollow fibers from the 2nd spin at
different operating temperatures. Selectivities calculated by using fugacity
at corresponding temperatures and pressures. Test conditions: 50/50 CO2/CH4,
22–50 1C.
Fig. 17. Temperature dependence of CO2 and CH4 permeance. Permeances
calculated by using fugacity. Test conditions: 50/50 CO2/CH4 at 800 psi, 22–50 1C.
In this equation, P/l refers to the CO2 permeance; Ep is the
activation energy; P0/l is pre-exponential factor. By plotting log
(P) vs. 1/T as shown in Fig. 17, the activation energy and P0 can be
calculated.
The curves in Fig. 17 show that there exists a good linearity
between log (P/l) and 1/T as the R2 approaches to 1 for both CO2
and CH4 permeation (R2 ¼0.9998 for CO2; R2 ¼0.9963 for CH4),
suggesting that the experimental permeation data agrees well
with the model shown in Eq. (9).
5. Conclusions
High-performance ester-crosslinkable hollow fibers were spun
successfully in this work through dope development and spinning
conditions optimization. The dope development and reformula-
tion shows that a dope composition with relatively lower poly-
mer% and THF% can effectively improve the CO2 permeance
without reducing CO2/CH4 selectivity significantly. Moreover,
the fibers were spun under a take-up rate of 50–75 m/min, which
is comparable to or even higher than commercial spinning rates.
Ester-crosslinking was utilized to stabilize the hollow fibers
against CO2 plasticization. Subsequent high-pressure mixed gas
permeation indicates that the crosslinked hollow fibers have
stable CO2 permeance and CO2/CH4 selectivity during pressuriza-
tion and depressurization, due to the well-controlled swelling
through ester crosslinking in the presence of high CO2 partial
pressure. Furthermore, the crosslinked hollow fibers were sub-
jected to operating temperatures ranging from room temperature
to 50 1C. By following the Arrhenius type equation, the CO2
permeance was increased under higher operating temperatures
with a decrease of CO2/CH4 selectivity. The experimental data
agrees well the model in term of both the CO2 and CH4 permeance
and suggests that a lower operating temperature is preferable to
maximize the separation performance.
Notwithstanding the excellent natural gas separation produc-
tivity shown in this paper, the crosslinked hollow fibers were
characterized by using a model natural gas feed consisting of only
CO2 and CH4; however, as a typical raw natural gas contains a
certain amount of hydrocarbon impurities, it will be desirable to
explore this high-performance crosslinked hollow fibers in the
presence of high level impurities to probe their properties under
more realistic feed streams, which will be studied in the future.
Acknowledgment
The authors acknowledge the financial support of Chevron
Energy Technology Company and The Roberto C Goizueta Chair
for Excellence at Georgia Institute of Technology.
References
[1] R.W. Baker, K. Lokhandwala, Natural gas processing with membranes: an
overview, Ind. Eng. Chem. Res. 47 (2008) 2109–2121.
[2] R.W. Baker, Future directions of membrane gas separation technology, Ind
Eng Chem Res 41 (2002) 1393–1411.
[3] P. Bernardo, E. Drioli, G. Golemme, Membrane gas separation: a review/state
of the art, Ind. Eng. Chem. Res. 48 (2009) 4638–4663.
[4] A.A. Olajire, CO2 capture and separation technologies for end-of-pipe
applications—a review, Energy 35 (2010) 2610–2628.
[5] W.J. Koros, G.K. Fleming, Membrane-based gas separation, J. Membr. Sci. 83
(1993) 1–80.
[6] M. Mulder, Basic Principles of Membrane Technology, 2nd ed., Kluwer
Academic, Dordrecht; Boston, 1996.
[7] C.A. Scholes, S.E. Kentish, G.W. Stevens, Carbon dioxide separation through
polymeric membrane systems for flue gas applications, Recent Pat. Chem.
Eng. 1 (2008) 52–66.
 C. Ma, W.J. Koros / Journal of Membrane Science 428 (2013) 251–259
[8] D.W. Wallace, Crosslinked Hollow Fiber Membranes for Natural Gas Purifica-
tion and Their Manufacture from Novel Polymers, Ph.D. Dissertation, Chemical
Engineering, The University of Texas at Austin, Austin, TX, 2004.
[9] R.W. Baker, Membrane Technology and Applications, 2nd ed., J. Wiley,
Chichester; New York, 2004.
[10] J.D. Wind, C. Staudt-Bickel, D.R. Paul, W.J. Koros, Solid-state covalent cross-
linking of polyimide membranes for carbon dioxide plasticization reduction,
Macromolecules 36 (2003) 1882–1888.
[11] A.M.W. Hillock, W.J. Koros, Cross-linkable polyimide membrane for natural
gas purification and carbon dioxide plasticization reduction, Macromolecules
40 (2007) 583–587.
[12] D.W. Wallace, C. Staudt-Bickel, W.J. Koros, Efficient development of effective
hollow fiber membranes for gas separations from novel polymers, J. Membr.
Sci. 278 (2006) 92–104.
[13] I.C. Omole, Crosslinked Polyimide Hollow Fiber Membranes for Aggressive
Natural Gas Feed Streams, Ph.D. Dissertation, Chemical and Biomolecular
Engineering, Georgia Institute of Technology, Atlanta, GA, 2008.
[14] T. Graham, On the Law of the Diffusion of Gases, J. Membr. Sci. 100 (1995)
17–21.
[15] I.C. Omole, R.T. Adams, S.J. Miller, W.J. Koros, Effects of CO2 on a high
performance hollow-fiber membrane for natural gas purification, Ind. Eng.
Chem. Res. 49 (2010) 4887–4896.
[16] C. Staudt-Bickel, W.J. Koros, Improvement of CO2/CH4 separation character-
istics of polyimides by chemical crosslinking, J. Membr. Sci. 155 (1999)
145–154.
[17] J.D. Wind, S.M. Sirard, D.R. Paul, P.F. Green, K.P. Johnston, W.J. Koros, Carbon
dioxide-induced plasticization of polyimide membranes: Pseudo-equilibrium
relationships of diffusion, sorption, and swelling, Macromolecules 36 (2003)
6433–6441.
[18] S. Kanehashi, T. Nakagawa, K. Nagai, X. Duthie, S. Kentish, G. Stevens, Effects
of carbon dioxide-induced plasticization on the gas transport properties of
glassy polyimide membranes, J. Membr. Sci. 298 (2007) 147–155.
[19] J.T. Vaughn, W.J. Koros, J.R. Johnson, O. Karvan, Effect of thermal annealing on
a novel polyamide-imide polymer membrane for aggressive acid gas separa-
tions, J. Membr. Sci. 401 (2012) 163–174.
[20] A.F. Ismail, N. Yaacob, Performance of treated and untreated asymmetric
polysulfone hollow fiber membrane in series and cascade module configura-
tions for CO2/CH4 gas separation system, J. Membr. Sci. 275 (2006) 151–165.
View publication stats
259
[21] A. Bos, I.G.M. Punt, M. Wessling, H. Strathmann, Suppression of CO2-
plasticization by semiinterpenetrating polymer network formation, J. Polym.
Sci. Pol. Phys. 36 (1998) 1547–1556.
[22] P.C. Raymond, W.J. Koros, D.R. Paul, Comparison of mixed and pure gas
permeation characteristics for CO2 and CH4 in copolymers and blends
containing methyl-methacrylate units, J. Membr. Sci. 77 (1993) 49–57.
[23] J.D. Wind, C. Staudt-Bickel, D.R. Paul, W.J. Koros, The effects of crosslinking
chemistry on CO2 plasticization of polyimide gas separation membranes, Ind.
Eng. Chem. Res. 41 (2002) 6139–6148.
[24] A. Taubert, J.D. Wind, D.R. Paul, W.J. Koros, K.I. Winey, Novel polyimide
ionomers: CO2 plasticization, morphology, and ion distribution, Polymer 44
(2003) 1881–1892.
[25] A.M.W. Hillock, S.J. Miller, W.J. Koros, Crosslinked mixed matrix membranes
for the purification of natural gas: effects of sieve surface modification,
J. Membr. Sci. 314 (2008) 193–199.
[26] S.B. Carruthers, Integral-Skin Formation in Hollow Fiber Membranes for Gas
Separations, Ph.D. Dissertation, Chemical Engineering, The University of
Texas at Austin, Austin, TX, 2001.
[27] D.T. Clausi, W.J. Koros, Formation of defect-free polyimide hollow fiber
membranes for gas separations, J. Membr. Sci. 167 (2000) 79–89.
[28] M. Das, W.J. Koros, Performance of 6FDA–6FpDA polyimide for propylene/
propane separations, J. Membr. Sci. 365 (2010) 399–408.
[29] J. Kurdi, A.Y. Tremblay, The influence of casting solution structure on the
microporosity of polyetherimide gas separation membranes prepared by the
coagulation post-leaching method, J. Membr. Sci. 184 (2001) 175–186.
[30] M.H. Kailani, C.S. Sung, S.J. Huang, Syntheses and characterization of model
imide compounds and chemical imidization study, Macromolecules 25
(1992) 3751–3757.
[31] M.H. Kailani, C.S.P. Sung, Chemical imidization study by spectroscopic
techniques. 1. Model amic acids, Macromolecules 31 (1998) 5771–5778.
[32] P.R. Dickinson, C.S.P. Sung, Kinetics and mechanisms of thermal imidization
studies by UV–visible and fluorescence spectroscopic techniques, Macromo-
lecules 25 (1992) 3758–3768.
[33] T.S. Chung, C. Cao, R. Wang, Pressure and temperature dependence of the gas-
transport properties of dense poly[2,6-toluene-2,2-bis(3,4dicarboxylphenyl)-
hexafluoropropane diimide] membranes, J. Polym. Sci. Pol. Phys. 42 (2004)
354–364.