Journal of Membrane Science 204 (2002) 247–256

Preparation and characterization of high-flux polysulfone

hollow fibre gas separation membranes
Dongliang Wang a,1 , W.K. Teo a , K. Li b,∗
aDepartment of Chemical and Environmental Engineering, National University of Singapore,
10 Kent Ridge Crescent, Singapore 119260, Singapore
b Department of Chemical Engineering, University of Bath, Claverton Down, Bath BA2 7AY, UK
Received 21 June 2001; received in revised form 9 January 2002; accepted 30 January 2002

Abstract
Polysufone (PSf) hollow fibre membranes with high gas separation performance were prepared from N-methyl-2-pyrrolidone
(NMP)/H2 O and NMP/ethanol solvent systems. Water was used as the external coagulant. The internal coagulants used in-
clude water, ethanol, 2-propanol, the mixture of water/ethanol and water/2-propanol. The effect of air gaps, the polymer
concentration and the coagulation bath temperature on the membrane structure and gas permeation properties was investi-
gated. The membranes were characterized by measuring their gas permeation properties for He, CO2 , O2 and N2 and by
examining cross-sectional structures using scanning electron microscopy (SEM). A dilute silicone coating method was used
to seal defects on the outer surface of the hollow fibres. The silicone-coated hollow fibres exhibit O2 permeance of 20–30
GPU and O2 /N2 selectivity of 5–6.5 at 25 ◦ C. Apparent separating layer thickness determined by the gas permeation is in the
range of 370–500 Å. © 2002 Elsevier Science B.V. All rights reserved.
Keywords: Polysulfone; Hollow fibre membranes; Gas separation; N-Methyl-2-pyrrolidone; Water; Ethanol; Nonsolvent-additive

1. Introduction membranes in both academia and industry. Develop-

Gas separation using polymeric membranes has
achieved important commercial success in some in-
dustrial processes since the first commercial-scale
membrane gas separation system was produced in
the late of 1970s. In order to extend its application
and compete successfully with traditional gas sepa-
ration processes such as cryogenics, pressure swing
adsorption and absorption, great attention has been
made in fabricating high separation performance
∗ Corresponding author. Tel.: +44-1225-826372;
fax: +44-1225-826894.
E-mail address: kli@bath.ac.uk (K. Li).
1 Present address: Institute of Material Science and Engineering,
Singapore 119260, Singapore.
ment of ultrathin-skinned asymmetric hollow fibre
membranes from highly selective polymers has been
a challenge in gas separation membrane processes
[1]. Although thousands of polymers exhibit permse-
lective properties for gas mixtures, only a few glassy
polymers are useful in making asymmetric gas sep-
aration membranes, particularly, polysufone (PSf),
polyethersulfone (PESf), polyetherimide (PEI) and
polyimide (PI). The PSf is the most promising poly-
mer for separating oxygen and nitrogen with respect
to its permeability and selectivity. For a given poly-
mer, how to make a high separation performance
membrane by studying the membrane formation pro-
tocol, is an important topic in the membrane research.
Formation of asymmetric hollow fibre contains many
variables such as the nature of the solvent, the additive

0376-7388/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 6 - 7 3 8 8 ( 0 2 ) 0 0 0 4 7 - 9
248 D. Wang et al. / Journal of Membrane Science 204 (2002) 247–256
and the coagulant, the spinning conditions etc. all of
which may affect the structure of the hollow fibre
membranes and separation performance. Reviewing
advances in preparing asymmetric gas separation
membranes, the selection of a suitable solvent, addi-
tive and coagulant is the most important. From the
view point of economic and environmental basis,
the solvents, additives and coagulants used must be
commercially available and low toxicity. In addition,
the manufacture method must be easy, economic and
reliable. However, in practice, the selection of solvent
and coagulant is quite limited. Water is the best choice
as the coagulant to remove solvent from hollow fibre
membrane. For water soluble solvents used, the most
common solvents are N-methyl-2-pyrrolidone (NMP)
and dimethylacetamide (DMAc). Therefore, the use
of a suitable additive became a most efficient tool to
prepare desired membranes. In fact, all high separa-
tion performance membranes have been prepared by
the use of right additives [2–5].
Development of a silicone-coated PSf hollow fibre
membrane was made by the researchers in Monsanto
and later, Perm Co. (Air Products). In the production
of the first generation PSf hollow fibre, the solvent
and the additive used are formylpiperidine (FP) and
formamide, respectively. The membrane prepared
from this solvent system with FP/FA mass ratio of
87/13 exhibited O2 flux of 10 GPU and O2 /N2 se-
lectivity of 5.2 at 50 ◦ C. Later on, comprehensive
studies towards improving membrane gas separa-
tion performance were done with different solvents
and different nonsolvent additives. The solvents used
include NMP, DMAc, triethylphosphate, FP and
dimethylformamide. The nonsolvent-additives (NSA)
used include formamide, glycerol, ethylacetate, tri-
ethyl phosphate, ethylene carbonate, acetic acid,
propionic acid, butyric acid, isobutyric acid, glutaric
acid and sulfuric acid. Of these solvent/NSA solvent
systems, only NMP/propionic acid solvent mixture
with the PA/NMP mass ratio of 43/57 was the most
efficient in fabricating high-flux PSf hollow fibre
membranes and the membrane prepared exhibited O2
pressure-normalized flux (or permeance) of 43 GPU
and O2 /N2 selectivity of 5–5.2 at 50 ◦ C. In preparing
this kind of membrane, the polymer dope with high
polymer concentration (37 wt.%) and high viscosity
(∼105 cP at 70 ◦ C) is required. It is generally difficult
to prepare, degas and spin the polymer dope with high
polymer concentration and high viscosity. In addition,
the propionic acid is odorous particularly when the
spinning solution is at high temperature. Recently,
extensive studies in our laboratory have been made
towards fabricating high-flux PESf [3,5] and PEI [4]
asymmetric hollow fibre membranes by adding suit-
able NSA to the polymer dope with moderate high
polymer concentrations. It has been shown that good
NSA should have great micibility and diffusivity
with the coagulant. The concentration of NSA in the
polymer dope is quite different for different NSA. A
simple equation has been developed to determine the
mass ratio in the polymer dope [6]. In spinning PESf
hollow fibre membranes, the membranes with the
best combination of permeability and selectivity were
prepared using water and ethanol [3,5] as a NSA. It
is, therefore, interesting to know whether high sepa-
ration performance PSf hollow fibre membranes can
also be prepared using water or ethanol as the NSA.
In this study, PSf hollow fibre membranes have been
prepared from both NMP/H2 O and NMP/EtOH sol-
vent systems. The effect of polymer concentration, air
gap length, internal coagulant and coagulation bath
temperature on the membrane structure and gas per-
meation properties were investigated.
2. Experimental
2.1. Materials
PSf (Udel P-1800) was purchased from Amoco
Performance Ltd (USA). The polymer was dried
at 100 ◦ C for 10 h. NMP (99%) was supplied by
MERCK and used without further purification. Wa-
ter and ethanol (99.8v%, MERCK) were used as the
NSA. Water, ethanol, 2-propanol (99.7v%, MERCK),
and the mixture of 50 wt.% ethanol and 50 wt.%
2-propanol aqueous solution were used as the internal
coagulant. Tap water was also used as the external
coagulant and washing medium. The coating solution
was prepared by mixing silicone elastomer and curing
agent (Sylgard-184) in n-pentane.
2.2. Hollow fibre preparation and characterization
In formualting spinning dopes, the mass ratios of
H2 O to NMP and EtOH to NMP were first determined
 D. Wang et al. / Journal of Membrane Science 204 (2002) 247–256
Table 1
Composition (wt.%) and properties of polymer dopes
Solution Composition (wt.%) Gv (g)
PSf NMP H2 O EtOH
Ho1 26.0 69.66 4.34 – 2.0
Ho2 28.0 67.78 4.22 – 2.0
Ho3 30.0 65.90 4.10 – 2.0
Et1 28.0 52.32 – 19.68 1.5
Gv : coagulation value, η: absolute viscosity.
based on the equation reported [6]. The precipitation
values of water and ethanol in PSf/NMP system are
8.4 and 50 g at 25 ◦ C, respectively [7]. For a required
coagulation value (defined as the quantity of a coag-
ulant in grams required to make 100 g polymer so-
lution containing 2 g polymer turbid), the mass ratio
of H2 O/NMP and EtOH/NMP was determined. About
600 g mixture of NMP/H2 O or NMP/EtOH with the
calculated mass ratio were prepared in 1000 ml bot-
tle and then the polymer was added into the bottle for
the desired polymer concentration. Compositions, co-
agulation values and absolute viscosities of the dopes
formulated in this study were listed in Table 1.
The polymer dopes were degassed and filtered.
The nascent hollow fibre was formed through a spin-
neret with orifice diameter/inner diameter of the tube
of 0.6/0.15 mm. The internal coagulant was injected
through the tube during the spinning. The nascent
fibre was solidified after passing coagulation bath
and wash bath and then rolled into a store box. In
spinning these hollow fibres, the take-up velocity is
4–6 m/min. The details of the hollow fibre spinning
procedure have been described elsewhere [3].
The fibres were kept in water for at least 3 days and
then dried at the ambient condition (25 ◦ C and relative
humidity (RH) = 60–65%). Two test modules were
made for each sample. Each module contains 20–30
pieces of fibres with the length of 20–25 cm. The
silicone-coated membranes were prepared by a dilute
solution immersion coating method under the vac-
uum. The coating solution was prepared with 3 wt.%
Sylgard-184 in n-pentane. The permeation fluxes of
silicone-coated asymmetric membranes for various
gases were measured at 25 ◦ C and pressure differ-
ence of 5 bar. Examination of the membrane mor-
phology using scanning electron microscopy (SEM)
has also been performed and the detailed procedures
249
of the examination has been described elsewhere
[3].
η (cP)
3. Results and discussion
21862
38100
54430
3.1. Effect of polymer concentration and air gap
12246
The PSf hollow fibre membranes were first prepared
from the PSf/NMP/H2 O spinning dopes with differ-
ent polymer concentrations (30, 28 and 26 wt.%) and
at different air gaps. Tap water at the temperature of
26 ± 1 ◦ C was used as the internal and the external
coagulants. After washing using tap water for 3 days,
the hollow fibre membranes were dried at the ambi-
ent condition (T = 25 ± 1 ◦ C and RH = 60–65%)
and then coated using silicone rubber (Sylgard-184).
The pressure-normalized fluxes (or permeances) of
He, CO2 , O2 and N2 were determined at the pressure
difference of 5 bar and the temperature of 25 ◦ C. The
ideal selectivities of He/N2 , CO2 /N2 and O2 /N2 were
then calculated based on the gas pressure-normalized
fluxes measured. The results are shown in Table 2. As
can be seen, the membranes prepared from the dopes
of Ho3 (30 wt.% of PSf) at the air gaps 5 or 10 cm show
high selectivities, but low permeabilities. Although the
hollow fibres prepared from the dopes, Ho2 (28 wt.%)
and Ho1 (26 wt.%) exhibit high permeabilities, their
selectiviiees are generally low, especially for the mem-
branes prepared from the dope Ho1 (26%). The PSf
hollow fibre with the best combination of permeabil-
ity and selectivity for O2 /N2 was prepared from the
28 wt.% polymer dope and at the air gap of 2 cm.
For a silicone-coated asymmetric membrane, if the
selectivity of the membrane is essentially equal to that
of the asymmetric membrane material, measured at
the same temperature, an apparent skin layer thickness
can be estimated using the following equation [8].
Pi
L= (1)
(P /L)i
where Pi is the permeability coefficient of gas
i (cm3 (STP) cm/cm2 cmHg s) in membrane mate-
rial. The permeability coefficient of O2 through
PSf film is 1.06 × 10−10 cm3 (STP)/cm2 cmHg s
and the O2 /N2 selectivity is 6.2 at 25 ◦ C [9].
(P/L)i is the pressure-normalized flux of gas i,
(cm3 (STP)/cm2 cmHg s) measured. The calculated
250 D. Wang et al. / Journal of Membrane Science 204 (2002) 247–256

Table 2
Gas permselective properties of silicone-coated asymmetric PSf hollow fibre membranes spun from three polymer dopes and at different
air gaps
Membrane No. o.d./i.d. (mm) Air gap (cm) Permeancea (GPU)b Ideal selectivity, α i/j L (Å)

He CO2 O2 He/N2 CO2 /N2 O2 /N2

Ho1-1 512/350 5 225.8 151.4 25.5 48.7 32.6 5.5 415

Ho1-2 525/400 2 219.0 167.5 29.6 39.6 30.3 5.4 358
Ho2-1 612/425 5 155.2 116.5 20.2 46.6 35.0 6.1 527
Ho2-2 550/287 2 253.3 161.7 28.0 52.2 34.0 6.2 378
Ho2-3 600/350 0 132.0 101.5 38.9 4.5 3.5 1.3 –
Ho3-1 675/325 20 145.3 86.1 45.4 3.2 1.9 1.0 –
Ho3-2 700/400 15 116.6 54.7 12.9 20.4 9.6 2.3 –
Ho3-3 750/462 10 152.0 75.6 14.9 72.4 36.0 7.1 711
Ho3-4 675/412 5 187.2 97.2 17.1 73.1 38.0 6.7 620
Data measured at 5 bar and 25 ◦ C, o.d.: outer diameter of hollow fibre.
a Also expressed as pressure-normalized flux.
b 1 GPU = 10−6 cm3 (STP)/cm2 cmHg s.

apparent skin layer thicknesses of PSf hollow fibre
membranes prepared in this study is ranged between
360 and 700 Å.
Similar to PESf hollow fibre gas separation mem-
branes prepared from solvent/NSA systems [3,5],
the PSf hollow fibre membrane prepared by the wet
method (air-gap = 0) also exhibited poor selectivity
as shown in Table 2. However, unlike the preparation
of PESf hollow fibre, the height of air gap seriously
affects the gas permeability and selectivity of PSf
membranes. It has been shown that the dry process (air
gap) is crucial for preparing high performance hollow
fibre membranes from solvent/NSA systems [3]. The
outer dense skin layer formation in the dry process
is not caused by the solvent evaporation, rather the
coalescence of polymer aggregates and rearrange of
polymer molecules due to the surface tension. The
coalescence depth depends on the interaction of poly-
mer/solvent/NSA and surface tension. In this study,
the take-up velocity was about 5 m/min. The drying
period at the air gaps of 2–20 cm is 0.2–2 s. The good
overall permeation/separation property obtained at
the air gap of 2 cm for O2 /N2 means that the poly-
mer aggregate coalescence and polymer molecule
orientation are completed quickly after the polymer
dope is leaving the spinneret. The longer drying time
(higher air gap) may cause coalesced polymer aggre-
gates become big due to strong interaction between
water and PSf, resulting in a decrease in gas perme-
ability. In the case of the wet process (0 air gap),
the coalescence of these aggregates on the outmost
region of the membrane is restricted due to the rapid
solidification in the water coagulation process [3] and
the skin layer with big pores may be easily formed in
the final membrane. These pores are too big to seal
by the dilute solution coating method. As a result, the
hollow fibre membrane prepared by the wet phase
inversion method shows poor separation properties.
Experimental results also show that the membranes
prepared from the other long air gaps (above 10 cm)
exhibited poor selectivity. Because of strong inter-
action between PSf and water, the internal coagu-
lant (water) could quickly diffuse from the internal to
the external walls. In addition, presence of water in
the polymer dope accelerates the water diffusion rate
along the membrane wall. At long air gaps, the inter-
nal water may diffuse through and damage the dense
skin layer formed during the outer dry process, result-
ing in the poor selectivity of the membranes. Such a
phenomenon has also been observed in the prepara-
tion of PEI hollow fibre membrane [10].
The cross-sectional structures of the PSf hollow fi-
bres spun from the 30 and 28 wt.% polymer dopes
by the dry/wet process were examined by SEM. The
samples used for SEM analysis were prepared at the
air gap of 5 and 2 cm. The SEM photographs are
shown in Figs. 1 and 2, respectively. As can be seen,
the cross-sectional structures of the membranes spun
from the both polymer dopes are similar. The fibre
wall consists of many long and regular finger-like
 D. Wang et al. / Journal of Membrane Science 204 (2002) 247–256 251

Fig. 1. SEM graphs of cross section of PSf hollow fibre membrane spun from 30 wt.% polymer dope and at air gap of 5 cm: (a) overall,
and (b) outer edge.

macrovoids from the inner surface towards the inte-
rior of the fibre (Figs. 1a and 2a). Examination of the
outer edge of these membranes reveals that there ex-
ist a thin dense skin layer. Under the thin dense skin
layer there are many circular macrovoids as shown in
Figs. 1b and 2b.
The cross-sectional structures of the membrane pre-
pared from the wet method were also examined. Fig. 3
shows the SEM graphs of the membrane spun from
the 28 wt.% dope at the 0 air gap. As can be seen, the
structures of the membrane are quite different from
those prepared from the dry/wet method. There are
no macrovoids in the outer region of the membrane
whereas the inner finger-like macrovoids are similar
compared to the membranes spun from the dry/wet
method. The cross-sectional structure difference be-
tween inner and outer area also demonstrated that the
coagulation behavior in the lumen is different from
that in the outerside although the coagulation occurs at
the same time. In the lumen side, the internal bore fluid
soon becomes a mixture of solvent and water due to
dissolution of the solvent in a small quantity of water

Fig. 2. SEM graphs of cross section of PSf hollow fibre membrane spun from 28 wt.% polymer dope and at air gap of 2 cm: (a) overall,
and (b) outer edge.
252 D. Wang et al. / Journal of Membrane Science 204 (2002) 247–256
Fig. 3. SEM graphs of cross section of PSf hollow fibre membrane spun from 28 wt.% polymer dope and at air gap of 0: (a) overall, and
(b) outer edge.
used as the internal coagulant. In this case, the coag-
ulation strength is not uniform in different positions.
The extended phase separation time prior to solidifi-
cation promotes the polymer-lean phase growth and
big macrovoid formation. Whereas in the outer coag-
ulation, the fast and nearly uniform coagulation oc-
curs and tends to form small and uniform macrovoids
along the membrane wall. This situation is similar to
the formation of microporous membranes by thermally
induced phase separation [11]. Fast phase inversion
along membrane wall tends to form the membrane
with small macrovoids. This also explains the com-
mon phenomenon that the big macrovoids are often
observed near the inner edge in preparing hollow fibre
membranes and macrovoids is reduced with a decrease
of the membrane wall. The results also show that the
skin layer structure and its separation performance of
the membrane are not related to the substrate structure.
It is of interest to note that the cross-sectional struc-
tures of PSf membranes are similar to those of PESf
membranes prepared from NMP/H2 O solvent system
[3]. The results indicate that the nonsolvent-aditive
and the dry process are very important in control-
ling membrane structure and separation performance.
The results further demonstrated that the influence of
the addition of coagulant into the polymer solution
on the membrane structure is quite different from that
of the influx of coagulant in the coagulation process.
The property of the polymer solution is completely
modified due to the interaction of PSf/NMP/H2 O.
Therefore, equilibrated two phases are easily formed
along the nascent membrane wall and grow fast until
the polymer-rich phase solidified through coales-
cence of polymer aggregates. After the flow stage,
the coalescing rate accelerates further due to density
difference of the two coexisting phases. For poly-
mer/solvent system, large amount of water quickly
diffuses into the polymer solution to occur fast phase
separation at the surface. Under the surface layer,
the phase separation rate is quite delayed. The influx
of water has a little effect on the polymer solution
property before the polymer-rich phase is solidified.
3.2. Effect of internal coagulant
It has been shown that the internal coagulant con-
trols the circular lumen formation and greatly in-
fluences the cross-sectional structure of the hollow
fibres [10]. The use of internal coagulant with suitable
nonsolvent strength for polymer and good miscibility
with the solvent can greatly reduce the big macrovoids
in the substrate, whereas the separation property can
be negligibly affected. For PSf, water is a strong
nonsolvent, while alcohol is the weak one [7]. The
internal coagulant with proper nonsolvent strength
can be prepared by the use of the mixture of water
and the alcohol with a desired ratio. In this study, the
influence of ethanol (EtOH), 2-propanol (2-PrOH)
 D. Wang et al. / Journal of Membrane Science 204 (2002) 247–256
Table 3
Gas permselective properties of silicone-coated asymmetric PSf hollow fibre membranes spun from different internal coagulants and at the
air gap of 2 cm
Bore fluids Permeancea (GPU)b
He CO2
EtOH 146.9 82.1
2-PrOH 33.8 16.9
50 wt.% EtOH/H2 O 189.0 134.5
50 wt.% PrOH/H2 O 175.4 119.1
Data measured at 5 bar and 25 ◦ C.
a Also expressed as pressure-normalized flux.
b 1 GPU = 10−6 cm3 (STP)/cm2 cmHg s.
and the mixtures of 50 wt.% EtOH/H2 O and 50 wt.%
2-PrOH/H2 O as the internal coagulants on separation
performance and the cross-sectional structure of PSf
hollow fibre membranes were investigated.
In spinning these PSf hollow fibre membranes, the
28 wt.% polymer dope was used. The air gap was ad-
justed at 2 cm. The gas permeation properties of the
silicone-coated PSf hollow fibres prepared using dif-
ferent internal coagulants were determined at the pres-
sure difference of 5 bar and temperature of 25 ◦ C. The
results are listed in Table 3. The membranes prepared
using ethanol and 2-propanol alone show low per-
meability and selectivity, particularly for 2-propanol.
The similar results were also reported in spinning
PESf hollow fibre membrane [10]. Because ethanol
and 2-propanol are very weak nonsolvents for PSf,
the internal coagulation rate and coagulant diffusion
rate are very slow. This tends to increase the sub-
strate resistance, and thus to reduce the membrane per-
meability and selectivity. The PSf hollow fibres with
good gas separation properties have been prepared us-
ing 50 wt.% EtOH/H2 O or 50 wt.% 2-PrOH/H2 O mix-
tures. The pressure-normalized fluxes and selectivities
are comparable to those of the membranes prepared
using water as the internal coagulant. This means that
the internal coagulation has a little influence on the
outer skin layer formation when suitable internal co-
agulant is used. The low permeabilities for the fast gas
(He) imply that the substrate and the inner skin may
give a small resistance for the gas permeation. The
SEM graphs of the cross section of the fibres prepared
from the four internal coagulants are shown in Fig. 4.
As can be seen, the cross-sectional structures are quite
different from that of the fibre spun using water as the
253
Ideal selectivity, α i/j L (Å)
O2 He/N2 CO2 /N2 O2 /N2
17.3 41.2 23.0 4.9 613
5.60 21.7 11.0 3.6 –
29.2 35.0 24.9 5.40 363
23.2 48.6 33.0 6.44 457
internal coagulant. The finger-like macrovoids near the
inner substrate of the fibres are quite reduced, but the
outer finger-like voids become longer. The slow coag-
ulation in the internal results in the further growth of
the outer macrovoids.
3.3. Effect of coagulation bath temperature
The influence of the coagulation bath temperature
on the hollow fibre membrane separation perfor-
mances and structures were also investigated. Table 4
shows gas pressure-normalized fluxes and ideal selec-
tivities for silicone-coated membranes measured at the
pressure difference of 5 bar and temperature of 25 ◦ C.
Comparison with the membrane prepared at the co-
agulation temperature of 26–27 ◦ C, the permeability
slightly increases and selectivity decreases when the
coagulation bath temperature was reduced to about
20 ◦ C. Further decreasing temperature, the selectivi-
ties of the membranes were considerably reduced. This
suggests that the membranes prepared at the coagu-
lation bath temperatures of 10–15 ◦ C exhibited large
surface porosity, which cannot be repaired with the
silicone coating. The results indicate that initial wet
phase separation rate is the most important to control
outer skin layer formation of hollow fibre membranes.
The temperature of coagulation bath is an important
factor to control membrane formation.
Figs. 5 and 6 show SEM of the cross section of the
membranes prepared at 20–21 and 10–11 ◦ C, respec-
tively. At 20 ◦ C, the structure is similar to that of the
membrane prepared at 26 ◦ C. But the macrvoids un-
der the skin layer were reduced. This suggests that the
decrease in temperature results in an increase in phase
254 D. Wang et al. / Journal of Membrane Science 204 (2002) 247–256

Fig. 4. SEM graphs of overall cross sections of PSf hollow fibre membranes spun from 28 wt.% polymer solution and different internal
coagulants: (a) ethanol; (b) 2-propanol; (c) 50 wt.% ethanol/water, and (d) 50 wt.% 2-propanol/water.

inversion rate and decrease in the coarsening time un- membrane prepared by the wet method (Fig. 3) with
der the skin layer. It is interesting to note that the respect to their skin layer structure and substructure,
cross-sectional structure of the fibre prepared at the suggesting that the decrease in the coagulation bath
coagulation temperature of 10–11 ◦ C is similar to the temperature enhances the phase separation rate.

Table 4
Gas permselective properties of silicone-coated asymmetric PSf hollow fibre membranes spun at different coagulation bath temperatures
Temperature (◦ C) Permeancea (GPU)b Ideal selectivity, α i/j L (Å)

He CO2 O2 He/N2 CO2 /N2 O2 /N2

20–21 270.0 199.9 30.6 45.0 33.3 5.1 346

15–16 155.2 158.6 30.7 18.2 18.5 3.6 –
10–11 196.2 173.0 33.0 21.1 18.6 3.6 –
Data measured at 5 bar and 25 ◦ C.
a Also expressed as pressure-normalized flux.
b 1 GPU = 10−6 cm3 (STP)/cm2 cmHg s.
 D. Wang et al. / Journal of Membrane Science 204 (2002) 247–256 255

Fig. 5. SEM graphs of cross section of PSf hollow fibre membrane spun from 28 wt.% polymer dope and at the coagulation bath temperature
of 20–21 ◦ C.

3.4. Polysulfone hollow fibre spun from
NMP/EtOH solvent system
It has been reported that PESf and PEI hollow fi-
bre membranes with good gas separation properties
could be prepared from NMP/EtOH solvent system
[3,4]. In order to evaluate the usefulness of EtOH as
the NSA, PSf hollow fibre membranes were also pre-
pared from NMP/EtOH solvent systems. The data of
permeance and selectivity for silicone-coated PSf hol-
low fibres prepared from NMP/EtOH solvent system
were determined at 25 ◦ C and the pressure difference
of 5 bar. The results are given in Table 5. As can be
seen, the membrane with good selectivity was pre-
pared at longer air gaps. With decreasing air gaps, the
permeability is comparable to those of the membranes
prepared using water as the additive, however, the se-
lectivity is lower.

Fig. 6. SEM graphs of cross section of PSf hollow fibre membrane spun from 28 wt.% polymer dope and at the coagulation bath temperature
of 10–11 ◦ C.
256 D. Wang et al. / Journal of Membrane Science 204 (2002) 247–256
Table 5
Gas permselective properties of silicone-coated asymmetric PSf hollow fibre membranes spun from NMP/EtOH and at different air gaps
Air gap (cm) Permeancea (GPU)b
He CO2 O2
25 176.3 132.4 23.0
20 185.7 155.0 24.7
10 163.4 158.0 27.8
5 143.8 157.4 29.5
2 138.3 136.6 24.2
Data measured at 5 bar and 25 ◦ C.
a Also expressed as pressure-normalized flux.
b 1 GPU = 10−6 cm3 (STP)/cm2 cmHg s.
4. Conclusions
PSf hollow fibre membranes with good permeabil-
ities and selectivities for important gas pairs were
fabricated from NMP/H2 O and NMP/EtOH solvent
systems with polymer concentration of 28 wt.%. The
separation performance of the membranes prepared
from the NMP/H2 O solvent system is better than
that of the membranes spun from NMP/EtOH solvent
system. The O2 pressure-normalized flux of the mem-
branes prepared is in the range of 20–30 GPU with
the O2 /N2 selectivity of 5–6.5 at 25 ◦ C. The apparent
skin layer thickness estimated by gas permeation is
about 360–500 Å. The air gap has significant influ-
ence on the hollow fibre separation performance. For
the PSf/NMP/H2 O spinning dope, the air gap should
be short, whereas, longer air gap is required for the
PSf/NMP/EtOH dope. The selectivity decreases with
a decrease of the coagulation bath temperature. Op-
timal coagulation bath temperature is in the range
from 20 to 30 ◦ C. Proper selection of internal coagu-
lant such as the mixture of water and C1 -C3 aliphatic
alcohols is an efficient method to alter membrane
structure with the maintenance of good permeabil-
ity and selectivity. The results further demonstrated
that the NSA plays an important role in controlling
membrane structure and separation performance.
References
[1] R.W. Baker, E.L. Cussler, W.J. Koros, H. Strathmann,
Membrane Separation Systems: Recent Development and
Future Directions, Noyes Data Corp. Park Ridge, NJ, 1991.
Ideal selectivity, α i/j L (Å)
He/N2 CO2 /N2 O2 /N2
42.8 32.1 5.59 461
37.0 32.0 4.92 429
27.5 26.6 4.68 –
16.1 19.2 3.59 –
24.7 24.4 4.32 –
[2] R.E. Kesting, A.K. Fritzsche, M.K. Murphy, A.C.
Handermann, C.A. Cruse, R.F. Malon, Process for forming
asymmetric gas separation membranes having graded density
skins, US Patent 4,871,494 (1989).
[3] D. Wang, K. Li, W.K. Teo, Polyethersulfone hollow fibre gas
separation membranes prepared from NMP/alcohol solvent
systems, J. Membr. Sci. 115 (1996) 85.
[4] D. Wang, K. Li, W.K. Teo, Preparation and characterization
of polyetherimide hollow fibre membranes for gas separation,
J. Membr. Sci. 138 (1998) 193–201.
[5] W.K. Teo, K. Li, D. Wang, Highly permeable polyethersulfone
hollow fibre membranes for gas separation, US Patent
6,017,474 (2000).
[6] D. Wang, K. Li, W.K. Teo, Relationship between mass
ratio of nonsolvent-additive to solvent in membrane casting
solution and its coagulation value, J. Membr. Sci. 98 (1995)
233.
[7] D. Wang, K. Li, S. Sourirajan, W.K. Teo, Phase separation
phenomena of polysulfone solvent organic, nonsolvent and
polyethersulfone/solvent/organic nonsolvent, systems, J. Appl.
Polym. Sci. 50 (1993) 1693–1700.
[8] P.H. Pfromm, I. Pinnau, W.J. Koros, Gas transport through
integral-asymmetric membranes: A comparison to isotropic
film transport properties, J. Appl. Polym. Sci. 48 (1993)
2161.
[9] I. Pinnau, W.J. Koros, Structures and gas separation properties
of asymmetric polysulfone membranes made by dry, wet,
and dry/wet phase inversion, J. Appl. Polym. Sci. 43 (1991)
1491.
[10] D. Wang, K. Li, W.K. Teo, Preparation of polyethersulfone
and polyetherimide hollow fibre membranes for gas
separation: effect of internal coagulant, in: I. Pinnau, B.
Freeman (Eds.), Membrane Formation and Modification,
United States ACS Symp. Ser., 744, (2000) Chapter 7,
pp. 96–109.
[11] S.-W. Song, J.M. Torkelson, Coarsening effects on the
formation of microporous membranes produced via thermally
induced phase separation of polystyrene-cyclohexanol
solutions, J. Membr. Sci. 98 (1995) 209.