Preparation of fluorinated polyimides with bulky structure and their gas separation performance correlated with microstructure

Article abstract of DOI:10.1016/j.polymer.2015.05.045

The fluorinated polyimides with high fractional free volume were prepared from 6FDA and aromatic diamines with bulky triptycene and pendent phenyl moieties. These polymers showed excellent solubility, high thermal stabilities and outstanding mechanical properties. The correlation of gas separation performance with the microstructure of these polyimide membranes was investigated. The results indicated that the GSPI-P membranes based on the diamines with pendent phenyl moieties exhibited higher fractional free volumes than GSPI-T membranes derived from diamines with triptycene moieties; as a result, the former gave the much higher permeability coefficients. The gas permeability of these membranes is strongly depended on their free volume and also affected by fluorine content. The GSPI-P membranes also provided good selectivity for CO₂/CH₄ and CO₂/N₂ gas pairs because their appropriate cavity size is favorable to separate the CO₂ from the other gases.

Full text of DOI:10.1016/j.polymer.2015.05.045

Polymer 69 (2015) 138e147
Contents lists available at ScienceDirect
Polymer
journal homepage: www.elsevier.com/locate/polymer
Preparation of fluorinated polyimides with bulky structure and their
gas separation performance correlated with microstructure
,
, *
Hui Tong a, Chenchen Hu b, Shiyong Yang a, Yanping Ma b, Hongxia Guo b *, Lin Fan a
a Laboratory of Advanced Polymer Materials, Institute of Chemistry, Chinese Academy of Sciences, Zhongguancun, Beijing 100190, China
b Laboratory of Polymer Physics and Chemistry, Institute of Chemistry, Chinese Academy of Sciences, Zhongguancun, Beijing 100190, China
a r t i c l e i n f o
a b s t r a c t
Article history:
The fluorinated polyimides with high fractional free volume were prepared from 6FDA and aromatic
diamines with bulky triptycene and pendent phenyl moieties. These polymers showed excellent solu-
bility, high thermal stabilities and outstanding mechanical properties. The correlation of gas separation
performance with the microstructure of these polyimide membranes was investigated. The results
indicated that the GSPI-P membranes based on the diamines with pendent phenyl moieties exhibited
higher fractional free volumes than GSPI-T membranes derived from diamines with triptycene moieties;
as a result, the former gave the much higher permeability coefficients. The gas permeability of these
membranes is strongly depended on their free volume and also affected by fluorine content. The GSPI-P
membranes also provided good selectivity for CO2/CH4 and CO2/N2 gas pairs because their appropriate
cavity size is favorable to separate the CO2 from the other gases.
Received 20 April 2015
Received in revised form
26 May 2015
Accepted 28 May 2015
Available online 30 May 2015
Keywords:
Polyimide membranes
Gas separation
Microstructure
© 2015 Elsevier Ltd. All rights reserved.
1. Introduction
mobility of molecular chains restrain the diffusion of small mole-
cules [27].
Polymeric membranes have been received more and more
attention because they play an important role in gas separation
applications, such as hydrogen recovery [1,2], nitrogen generation
[3] and carbon dioxide removal [4e9]. Gas separation by polymeric
membranes has potential advantages of cost-effective, energy-
efficient and environmentally benign processes in many industries
[10]. In the past decades, many different polymers have been
investigated as gas separation materials, including rubbery poly-
mers, such as poly(dimethylsiloxane) [11,12], and glassy polymers,
such as polysulfone [13e16], poly(phenyl oxide) [17,18], cellulose
acetate [19,20] and polyimides [21e23]. Glassy Aromatic poly-
imides are among the most attractive and promising gas separation
materials owing to their admirable gas selectivity, excellent ther-
mal stability, good mechanical properties and superior chemical
resistance [24e26]. Moreover, their gas separation properties are
easily tuned by numerous combinations between dianhydrides and
diamines. However, most of aromatic polyimide membranes have
low to moderate permeability because their high packing and low
An ideal gas separation membrane should have both high
permeability and high selectivity. Many efforts have been devoted
to develop polyimide membranes with increasing permeability to
specific gases without greatly impairing their inherent good
selectivity. The introduction of packing-disrupting units into the
polyimide backbone is the most effective way to reduce the chain
packing, which can result in a high fractional free volume (FFV) in
membrane to enhance the gas permeation properties [28]. Oka-
moto and his coworkers reported the permeability and selectivity
of gases in fluorinated and non-fluorinated polyimides [29]. They
found that polyimides with eC(CF3)2ꢀ linkages had lower chain
packing density and revealed higher permeability with relatively
higher selectivity. The polyimides based on 4,40-(hexafluor-
oisopropylidene)diphthalic (6FDA) have been extensively investi-
gated to fabricate high performance membranes [30,31]. The
eC(CF3)2ꢀ linkages in 6FDA is believed to serve as a molecular
spacer and a chain stiffener, which reduces the intra-segmental
mobility and limits the chain packing thereby increasing the FFV
[32]. Many other researches on development of high performance
polyimide membranes are mainly focused on designing the aro-
matic diamines with bulky structure. The polyimides derived from
methyl-substituted phenylenediamines and different dianhydrides
have been developed by Tanaka et al. [33]. They claimed that the
* Corresponding authors.
E-mail addresses: hxguo@iccas.ac.cn (H. Guo), fanlin@iccas.ac.cn (L. Fan).
http://dx.doi.org/10.1016/j.polymer.2015.05.045
0032-3861/© 2015 Elsevier Ltd. All rights reserved.

H. Tong et al. / Polymer 69 (2015) 138e147
139
methyl substituents in diamines restricted the internal rotation
around the bonds between the phenyl rings and the imide rings;
therefore, the rigidity and nonplanar polymer chain together with
the bulky methyl groups made chain packing insufficient, resulting
in high fraction of free space and high permeability. However, they
also found that these polyimide membranes displayed very low
selectivity. The bulky triptycene-based polyimides have been
investigated by many researchers [34]. Triptycene is a symmetric
three-dimensional, bulky and rigid structure composed of three
phenyl rings bound together by a single hinge, which can disrupt
chain packing and create void spaces. It is confirmed that the
triptycene-based polyimides possessed high internal free volume,
allowing fast molecular diffusion in membrane and resulting in
high gas permeability. Calle et al. developed polyimide membranes
from aromatic diamine containing bulky di-tert-butyl side groups,
i.e., 1,4-bis(4-aminophenoxy)2,5-di-tert-butylbenzene (TBAPB) and
commercial dianhydrides [35]. They proved that the bulky di-tert-
butyl side groups combined with rigid main chain could yield
polyimides with fractional free volume as high as 0.199. The results
indicated that the polyimide membranes derived from TBAPB dis-
played good permeability. However, like other high free volume
glassy polymers, the significant improvement in gas permeability
of polyimide membranes is not always accompanied by the similar
improvement in selectivity [36]. The performance of polyimide
membranes is still limited by the trade-off relationship between
permeability and selectivity. Therefore, recent studies on polyimide
membranes for gas separation have been done to develop new
macromolecular structures with high fractional free volume but
narrow cavity size distribution, which is favorable to gain the
membranes with high permeability and maintained high selec-
tivity [37]. However, to our knowledge, very few attempts have
been carried out on the correlation between the microstructure and
gas separation performance of polyimide membranes.
(99%, J&K Chemical), 3,4,5-trifluorobenzaldehyde, potassium car-
bonate (Beijing Chemical Works, China) and concentrated hydro-
chloric acid (37 wt.%) were used as received. 4,40-(Hexafluoroiso
propylidene)diphthalic anhydride (6FDA) was dried in a vacuum
oven at 150 ꢁC for 12 h prior to use. Triptycene-based diamines, 1,4-
bis(4-aminophenoxy)triptycene (BAT) and 1,4-bis(4-amino-2-
trifluoromethylphenoxy)triptycene (6FBAT), were synthesized in
our laboratory according to the literature [38]. Commercially
available N-methyl-2-pyrrolidinone (NMP) and N,N-dimethylace-
tamide (DMAc) were purified by distillation under reduced pres-
sure and dehydrated with 4 Å molecular sieves prior to use. Other
solvents and regents were used as received.
2.2. Monomer synthesis
In a typical synthesized procedure of
a,a-bis(4-amino-3,5-
dimethyphenyl)-1-phenylmethane (BAPM), 2,6-dimethylaniline
(48.47 g, 0.4 mol) and 100 mL distilled water was placed into a
three-necked round-bottom flask equipped with a mechanical
stirrer and nitrogen inlet and outlet. The mixture was stirring at
room temperature for 1 h until a cream yellow emulsion was ob-
tained. Concentrated hydrochloric acid (37 wt.%, 47 mL) was then
added dropwise over an hour and the reaction solution was
maintained at 30e40 ꢁC. After that, benzaldehyde (22.28 g,
0.21 mol) was added in batches followed by heated to 100 ꢁC and
kept refluxing for 12 h. The resulting nattier blue mixture was
cooled to room temperature, and then, potassium carbonate (60 g)
was added to neutralization. The BAPM was obtained by water
vapor distillation to remove the by-product, which was then puri-
fied by recrystallization from ethanol to yield white powder (59.2 g,
93%). mp: 183e185 ꢁC. 1H NMR (CDCl3,
d, ppm): 7.17e7.19 (t, 3H),
7.10e7.12 (d, 2H), 6.69 (s, 4H), 5.27 (s, 1H), 3.71 (s, 4H), 2.08 (s, 12H).
Elemental analysis: Calculated for C23H26N2 (330.47): C, 83.59%; H,
7.93%; N, 8.48%. Found: C, 83.14%; H, 7.96%; N, 8.28%.
In this study, the novel aromatic diamines with pendent phenyl
structures, i.e.,
ane (BAPM),
yl)methane (BAFM), and
a
,
a
-bis(4-amino-3,5-dimethyphenyl)-1-phenylmeth
-Bis(4-amino-3,5- dimethyphenyl)-1-(40-fluorophen
-Bis(4-amino-3,5- dimethyphenyl)-1-
The other two aromatic diamines, i.e.,
dimethyphenyl)-1-(40- fluorophenyl)methane (BAFM) and
bis(4-amino-3,5-dimethyphenyl)-1-(30,40,50-
trifluorophenyl)
a,a-bis(4-amino-3,5-
a,a
a,a-
a,a
(30,40,50-trifluorophenyl)methane (BATFM), were synthesized. These
diamines and the diamines with triptycene moieties, i.e., 1,4-bis(4-
aminophenoxy)triptycene (BAT) and 1,4-bis(4-amino-2-trifluorome
thylphenoxy)triptycene (6FBAT) were polymerized with commer-
cial available dianhydride 6FDA, respectively. The incorporation of
the bulky triptycene and pendent phenyl moieties is expected to
endow the polyimides with high fractional free volume and result in
high gas permeability. The trifluoromethyl and methyl groups were
also incorporated to the polymer backbone, in order to produce more
free spaces and provide rigid main chain, which will thereby improve
the gas permeability on the premise of maintaining high selectivity.
The solubility, thermal and mechanical properties as well as gas
separation performance of these fluorinated polyimides were eval-
uated. In order to clarify the relationship between gas separation
performance and microstructure of membranes, the microstructure
parameters, such as average interspacing distance and fractional free
volume were determined by X-ray diffraction measurements and
molecular dynamics simulation, respectively. These results were
compared with the fractional free volume and cavity size obtained
from experimental approach of positron annihilation measurements
and discussed in detail.
methane (BATFM), were synthesized by the similar procedure,
except the benzaldehyde was replaced by 4-fluorobenzaldehyde
and 3,4,5-trifluorobenzaldehyde, respectively.
For BAFM, yield: 64.5 g, 93%. mp: 170e172 ꢁC. 1H NMR (CDCl3,
d,
ppm): 7.04e7.07 (t, 2H), 6.91e6.95 (t, 2H), 6.66 (s, 4H), 5.24 (s, 1H),
3.73 (s, 4H), 2.12 (s, 12H). Calculated for C23H25FN2 (348.46): C,
79.28%; H, 7.23%; N, 8.04%. Found: C, 79.28%; H, 7.42%; N, 7.92%.
For BATFM, yield: 70.2 g, 91%. mp: 151e152 ꢁC, 1H NMR (CDCl3,
d,
ppm): 6.69e6.72 (t, 2H), 6.63 (s, 4H), 5.16 (s, 1H), 3.73 (s, 4H), 2.13
(s, 12H). Calculated for C23H23F3N2 (384.44): C, 71.86%; H, 6.03%; N,
7.29%. Found: C, 72.06%; H, 6.09%; N, 7.04%.
2.3. Polymer synthesis
A series of fluorinated polyimides were prepared from dianhy-
dride 6FDA and aromatic diamines, i.e., BAT, 6FBAT, BAPM, BAFM
and BATFM, respectively. In a typical experiment, GSPI-P3, which
derived from 6FDA and BATFM, was prepared according to the
following procedure.
BATFM (19.22 g, 0.05 mol) and anhydrous NMP (120 mL) were
added to a completely dried 500 mL three-necked flask, which was
equipped with a mechanical stirrer, a nitrogen inlet, a thermom-
eter, and a DeaneStark trap. The mixture was stirred at ambient
temperature under nitrogen flow until BATFM was completely
dissolved to give a homogeneous solution. Then, 6FDA (22.21 g,
0.05 mol), isoquinoline (0.25 g) and toluene (20 mL) were added
and the solution was stirred in nitrogen at room temperature for
6 h. The reaction solution was gradually heated to 180 ꢁC and
2. Experimental
2.1. Materials
2,6-Dimethylaniline (>99%, Alfa Aesar), benzaldehyde (98.5%,
Sinopharm Chemical Reagent Co., China), 4-fluorobenzaldehyde

140
H. Tong et al. / Polymer 69 (2015) 138e147
maintained at the temperature for 12 h, in which the water evolved
in the procedure of imidization was removed simultaneously by
azeotropic distillation. After the solution was cooled down to
80e100 ꢁC, the viscous solution was poured slowly into excess
ethanol to give pale-yellow, fiber-like precipitate, which was then
collected by filtration, washed thoroughly with ethanol and dried
in vacuum at 120 ꢁC for 8 h.
The other fluorinated polyimides, i.e., GSPI-T1 (6FDA/BAT), GSPI-
T2 (6FDA/6FBAT), GSPI-P1 (6FDA/BAPM) and GSPI-P2 (6FDA/BAFM)
were synthesized by the similar procedure, except BATFM was
replaced by the other aromatic diamines.
The molecular models for molecular dynamics simulation were
built by Cerius2 package from Accelrys. In model building, three
molecular chains composed of dianhydride and diamine with 30
repeating units for each were set in a cubic simulation cell and
optimize the trade-off between cell dimension and calculation
time. All of the molecular models were treated with an energy
minimization process over 1000 iterations to obtain a stable
structure. Then, the MD calculations were performed under an NVT
ensemble (fixed number of atoms, cell volume, and temperature)
for duration of 1 ns at 308 K. The Newton second law of motion was
employed to calculate the dynamic behaviors of molecules. The
COMPASS force field was adopted for the theoretical calculations.
The FFV of a membrane can be calculated by the following
equation:
2.4. Membrane preparation
The polyimide resin was dissolved in DMAc with stirring to give
a homogeneous polyimide solution with a solid content of 20 wt.%.
The polyimide solution was filtered by a 0.2-mm Teflon syringe
V ꢀ Vo V ꢀ 1:3Vw
FFV ¼
¼
(2)
V
V
filter and then spread onto a glass plate with a knife gap of 300 mm
where V, Vo and Vw are the specific volume, occupied volume and
the van der Waals volume of the polymer, respectively.
on the Elcometer 4340 automatic film applicator. After that, the wet
film was thermally baked in an oven at 80 ꢁC for 2 h, 120 ꢁC for 1 h,
at 150 ꢁC for 1 h and 180 ꢁC for 1 h, successively. The free-standing
Positron annihilation measurements were performed with a
fast-slow coincidence ORTEC system with a time resolution of
195 ps for the full width at half-maximum. The two identical pieces
of samples were placed on either side of the 22Na positron source,
and then the sample-source-sample sandwich was placed between
the two detectors to acquire the lifetime spectra. A total of 2 ꢂ 106
counts were accumulated for each spectrum to reduce the statis-
tical error in the calculation of lifetimes. The positron annihilation
spectra were de-convoluted using the LT-9 software. The raw data
were resolved into three lifetime components, which assume a
Gaussian distribution for the logarithm of the lifetime for each
component. The third lifetime component of t3 (>0.5 ns) is
assigned to the decay of orthopositronium (o-Ps) in the polymer
free volume. The average cavity radius (R) in the polymer can be
computed using the following semiempirical equation:
polyimide membrane with the thickness of 40e50
mm was ob-
tained by immersion the glass plate in water followed by drying in
an oven at 100 ꢁC for 12 h.
2.5. Characterization
2.5.1. Measurements
1H NMR spectra were performed on a Bruker Avance 400
Spectrometer operating at 400 MHz in CDCl3. FTIR spectra were
recorded on a PerkineElmer 782 Fourier transform spectropho-
tometer. The number average molecular weights (Mn) and poly-
dispersities (Mw/Mn) were determined by gel permeation
chromatography (GPC) on a Waters GPC system equipped with a
Waters 1515 HPLC pump, a Waters 2414 differential refractometer,
and three Styragel columns of HT-3, HT-4 and HT-5 using NMP
containing 0.02 M LiBr as eluent at a flow rate of 1.0 mL/min at
40 ꢁC. The system was calibrated using a series of monodisperse
linear polystyrene standards. Solubility of polyimides was deter-
mined by dissolving polymer resins in different solvents with
15 wt.% of solid content at room temperature. The absolute vis-
cosities of PI solutions in DMAc were measured on a Brookfield DV-
IIþ programmable viscometer at 25 ꢁC. Differential scanning calo-
rimetry (DSC) and thermal gravimetric analysis (TGA) were carried
out using a TA Q100 and a TA Q50 instruments in nitrogen,
respectively, at a heating rate of 10 ꢁC/min. Mechanical properties
were measured on an Instron-3365 tensile apparatus with
120 ꢂ 10 ꢂ 0.05 mm specimens at 25 ꢁC in accordance with Chinese
national standard of GB/T 1040.3-2006 at a drawing rate of 5.0 mm/
min. Wide-angle X-ray diffraction (WAXRD) measurements were
conducted on a Rigaku D/max-2500 X-ray diffractometer with Cu/
ꢀ
ꢁ
ꢂꢃꢀ1
1
2
R
1
2pR
R þ DR
t3 ¼
1 ꢀ
þ
sin
(3)
R þ DR 2p
where t3 is the o-Ps lifetime, and
DR is an empirical constant of
1.66 Å. Since the cavities are assumed to be spherical, the mean free
volume of a cavity (V) and the relative fractional free volume (FFV)
are given as the following equations (4) and (5), respectively.
4
3
V ¼ pR3
(4)
(5)
FFV ¼ cVI3
where c is a material-dependent constant (defined as 10 in this
paper), and I3 is the corresponding intensity of t3
.
Ka radiation, operated at 40 kV and 200 mA. The average inter-
spacing distance (d-spacing) was determined based on the Bragg
equation:
2.5.3. Permeability measurements
A barometric permeation method was employed to determine
steady state pure gas permeability at 23 ꢁC according to interna-
tional standard of ISO 15105-1. The membranes were degassed in
the permeation apparatus on both sides under high vacuum before
testing. The downstream pressure was kept below 10ꢀ2 mbar, while
the upstream pressure was maintained at 1 bar. For the permeation
experiments, the gases of CO2, O2, N2 and CH4 with purity higher
than 99.99% were used. Under conditions of steady state of
permeation, the permeability coefficient (P) is determined from the
slope of the downstream pressure versus the time plot (dp(t)/dt)
according to the following equation:
nl ¼ 2dsinq
(1)
where d is the d-spacing,
q
is the scattering angle and n is an integer
number (1, 2, 3 …) related to the Bragg order.
2.5.2. Fractional free volume
The fractional free volume (FFV) of polyimide membranes was
calculated by molecular dynamics (MD) simulation and measured
by positron annihilation lifetime spectroscopy (PALS).

H. Tong et al. / Polymer 69 (2015) 138e147
141
273 Vl dpðtÞ
76 ATp0 dt
P ¼
(6)
where V is the calibrated permeate volume, l and A are the effective
thickness and area of the tested membrane, respectively, T is the
measurement temperature in K and p0 is the pressure of the feed
gas in the upstream chamber. The permeation measurement results
were determined by the average value of three specimens for each
test condition.
The ideal selectivity for a pair of gases A and B is calculated from
the following equation:
aA=B ¼ PA=PB
(7)
where PA and PB are the permeability coefficients of A and B,
respectively.
3. Results and discussion
3.1. Monomer synthesis
The aromatic diamines with pendent phenyl group, i.e., BAPM,
BAFM and BATFM, were synthesized by a typical electrophilic
substitution reaction as shown in Scheme 1. The 2,6-dimethylani-
line was reacted with benzaldehyde or fluorobenzaldehyde under
the catalysis of concentrated hydrochloric acid, followed by
neutralization with potassium carbonate. In this procedure, the
excess benzaldehyde and fluorobenzaldehyde, which performed
not only as reactants but also as good solvents, were used to ensure
a sufficient reaction and to afford the resultants in high yield
(91e93%).
The chemical structures of the synthesized diamines were
confirmed by 1H NMR, FTIR and elemental analysis. Fig. 1 shows the
1H NMR spectra of the compounds BAPM, BAFM and BATFM, in
which all the protons in the structures could be assigned clearly
according to the integral values of the intensity. The signals
assigned to the protons in the methyl groups, H1, were detected at
2.08e2.13 ppm. The protons in amino groups resonated in the
range of 3.71e3.73 ppm. The signals assigned to proton in the
methylidyne group, H2, shifted to downfield region of
Fig. 1. 1H NMR spectra of BAPM (a), BAFM (b) and BATFM (c) in CDCl3.
5.16e5.27 ppm on account of the conjugative effect from the phenyl
group. The aromatic protons, H3~H6, resonated in the region of
6.63e7.19 ppm. The aromatic protons adjacent to methyl group (H3)
were observed at 6.63e6.69 ppm with the relatively lower chem-
ical shift due to the electron-donating effect of adjacent methyl
groups. It is noticed that the fluorine atom affects the chemical shift
of protons in the pendent phenyl group significantly. The signal
assigned to H4 in BAPM appeared at 7.10e7.12 ppm, which shifted
to 6.91e6.95 ppm for BAFM and 6.69e6.72 ppm for BATFM because
of the strong electron-withdrawing effect of the fluorine atoms.
The structures of these aromatic diamines were further
confirmed by means of FTIR spectra. As shown in Fig. 2, the
Scheme 1. Synthesis of diamines with pendent phenyl structure.

142
H. Tong et al. / Polymer 69 (2015) 138e147
acted as a catalyst, to give poly(amic acid). After that, the inter-
mediate without isolation was thermally cyclodehydrated at
elevated temperature to give the corresponding polymer. The water
by-product was removed by azeotropic distillation with toluene to
ensure the thermal cyclization being processed completely. The
fiber-like precipitates in pale-yellow were obtained by pouring the
viscous polymer solution slowly into excess of ethanol, indicating
that the polymers have relatively high molecular weight.
Table 1 summarized the physical properties and GPC data of the
fluorinated polyimides. It can be seen that all the polyimides were
produced in high yields of 96e98%. The fluorine contents of GSPI-
T1~GSPI-P3 ranged from 13.0 to 22.5%. These polymers showed
reasonable high molecular weight and narrow molecular weight
distribution with Mn and Mw in the range of 3.64e5.60 ꢂ 104 g/mol
and 8.27e11.12 ꢂ 104 g/mol, as well as Mw/Mn of 1.93e2.27,
respectively.
Fig. 3 shows the FTIR spectra of the fluorinated polyimides, in
which the characteristicꢀ1absorptions of imide groups were
observed at about 1780 cm (C]O, asymmetric),1720 cmꢀ1 (C]O,
symmetric) and 1370 cmꢀ1 (CeN, asymmetric), respectively.
Meanwhile, the characteristic NeH stretching vibration of animo
group in poly(amic acid) at around 3300e3500 cmꢀ1 were not
detected, indicating that the polymers were imidized completely.
Moreover, the absorptions assigned to stretching vibration of CeF
at 1250 and 1140 cmꢀ1 were also observed in the spectra of these
polymers. The results demonstrated that the fluorinated poly-
imides had the expected chemical structures.
The solubility of the fluorinated polyimides was measured by
dissolving the polyimide resins in different organic solvents with
15 wt.% of solid content at room temperature. It is known that ar-
omatic polyimides generally have poor solubility in common
organic solvents because of their strong chain interaction of imide
rings. However, the fluorinated polyimides synthesized in this
study could be easily dissolved in many organic solvents except in
ethanol as shown in Table 2. The good solubility of these polyimides
is due to the synergetic effects of the bulky triptycene and pendent
phenyl moieties as well as the fluorine groups in the polymer
structure, which disrupt the regularity of the molecular chains and
hinder the dense chain stacking, consequently, permitting solvent
to dissolve the polymers more easily [39].
Fig. 2. FTIR spectra of BAPM, BAFM and BATFM.
characteristic absorptions due to the NeH stretching vibration of
amino groups and the CeH stretching vibration of methyl groups in
the region of 3300e3500 cmꢀ1 and 2800e3000 cmꢀ1, respectively,
were detected for all of diamines. Whereas, the characteristic ab-
sorptions around 1620 cmꢀ1 were assigned to the C]C vibration of
phenyl groups. Meanwhile, the absorptions assigned to CeF
stretching vibration at 1228 and 1151 cmꢀ1 were detected for
fluorinated diamines BAFM and BATFM. The latter gave the rela-
tively stronger signal due to the higher fluorine content. In addi-
tion, the elementary analysis values of these monomers were in
good agreement with the calculated ones. The characterization
results demonstrated that the aromatic diamines BAPM, BAFM and
BATFM were successfully synthesized and the obtained monomers
were pure enough to be employed for preparation of polyimides.
3.2. Polymer synthesis and characterization
The solubility of these polyimides was also quantitatively eval-
uated. Fig. 4 depicts the dependence of the absolute viscosities of
polyimide solutions in DMAc on the solid contents. The absolute
viscosities of these polyimides in DMAc with 20 wt.% of solid
content were no more than 1250 mPa$s at 25 ꢁC, which exhibited a
gradually increasing for GSPI-T polyimides and an abruptly
enhancing for GSPI-P polyimides as the solid content in excess of
25 wt.%. Moreover, the absolute viscosity of these polyimide solu-
tions with same concentration increased in the order of GSPI-
A series of fluorinated polyimides were synthesized via one-pot
solution polycondensation from dianhydride 6FDA and aromatic
diamines with bulky triptycene and pendent phenyl moieties in
NMP as shown in Scheme 2. It is considered that the one-pot
method is suitable for low reactive monomers to synthesize solu-
ble PIs, however, the polycondensation should be implemented at
elevated temperature in order to get the polymer with high mo-
lecular weight. Initially, 6FDA and aromatic diamine were reacted at
the ambient temperature in the presence of isoquinoline, which
Scheme 2. Synthesis of fluorinated polyimides.

H. Tong et al. / Polymer 69 (2015) 138e147
143
Table 1
Physical properties and GPC data of fluorinated polyimides.
PI
Yield (%)
Fa (%)
GPC data
Mn ꢂ 10ꢀ4
Mw ꢂ 10ꢀ4
Mw/Mn
GSPI-T1
GSPI-T2
GSPI-P1
GSPI-P2
GSPI-P3
98
97
97
97
96
13.0
22.5
15.4
17.6
21.6
5.60
5.20
4.46
5.04
3.64
11.12
10.28
9.01
9.75
8.27
1.98
1.98
2.02
1.93
2.27
a
F: fluorine content.
Fig. 4. Dependence of the absolute viscosity of polyimide solution in DMAc on solid
content.
ortho position to the imide ring in polyimide structure restricted
the rotation of nitrogen atom along the phenyl ring, which in turn
increase the Tg of polyimides [40]. Moreover, TGA results revealed
that these polymers were thermally stable up to 496 ꢁC without
obvious weight loss in nitrogen. In comparing these polyimides, it
can be seen that the GSPI-P polyimides gave the decomposition
temperatures at 5% and 10% weight loss (T5 and T10) 15e22 ꢁC lower
than GSPI-T polyimides. This may correspond to the loss of methyl
substituents in GSPI-P polyimides [41], although they have more
rigid backbone than GSPI-T polyimides.
The unoriented fluorinated polyimide membranes could be
prepared by solution casting and solvent evaporation. All the
membranes presented excellent mechanical property with the
tensile strength of 90e115 MPa, the Young's modulus of
2.1e2.9 GPa, and the elongation at break of 4.3e7.4%, respectively.
The strong and tough property of these polyimide membranes are
owing to their strong chemical bond in the main chains [25].
Fig. 3. FTIR spectra of fluorinated polyimides.
T2<GSPI-T1<GSPI-P3<GSPI-P2<GSPI-P1. It is noted that the solu-
bility depended on their backbone structure. The GSPI-T polyimides
showed better solubility than GSPI-P polyimides due to the pres-
ence of bulky triptycene structure and flexible polymer backbone. It
is also found that the polyimides with relative higher fluorine
content gave the lower absolute viscosity although they have the
similar backbone. For instance, the absolute viscosity of GSPI-P3
(21.6% of fluorine content) at the solid content of 30 wt.% was
3.42 ꢂ 104 mPa s, which was only half of that for GSPI-P1 (15.4% of
fluorine content). The bulky fluorinated substituents both in the
polymer main chain and side chain could inhibit the close chain
packing, which leading to the improvement of polymer solubility.
The polymer solutions were very stable and no phase separation or
precipitation was observed after storage for several weeks.
The thermal and mechanical properties of these polyimides
were evaluated and the results are summarized in Table 3. These
polyimides exhibited excellent thermal stability with glass transi-
tion temperatures (Tg) for GSPI-T and GSPI-P polyimides around
300 ꢁC and 330 ꢁC, respectively. The relatively higher Tg values of
GSPI-P polyimides are attributed to their more rigid polymer main
chains. It has been reported that the methyl substituents in the
3.3. Gas transport properties and their correlation with
microstructure
The gas transport properties of these fluorinated polyimide
membranes were investigated using a constant-volume, pressure-
variable gas permeation method. The gas permeability coefficients
of CO2, O2, N2 and CH4 as well as the ideal selectivity for gas pairs
are listed in Table 4. It can be observed that, in all cases, the order of
permeability was P(CO2) > P(O2) > P(N2) > P(CH4). This order co-
incides with the kinetic diameter of gas molecules (CO2, 3.30 Å; O2,
3.46 Å; N2, 3.64 Å; CH4, 3.80 Å), indicating that the size-sieving is a
dominant factor for gas separation in these membranes. As
comparing the permeability of these polyimide membranes, it is
noted that the GSPI-P membranes revealed much higher perme-
ability coefficients than the GSPI-T membranes combined with an
Table 2
Solubility of fluorinated polyimides.a
PI
NMP
m-Cresol
DMAc
THF
Dioxane
Chloroform
Acetone
Ethanol
GSPI-T1
GSPI-T2
GSPI-P1
GSPI-P2
GSPI-P3
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þþ
þ
ꢀ
ꢀ
ꢀ
ꢀ
ꢀ
þþ
þþ
þþ
þþ
a
þþ: soluble at room temperature; þ: partially soluble; ꢀ: insoluble even on heating.

144
H. Tong et al. / Polymer 69 (2015) 138e147
Table 3
Thermal and mechanical properties of fluorinated polyimides.a
PI
Tg (ꢁC) Td (ꢁC) T5 (ꢁC) T10 (ꢁC) TS (MPa) TM (GPa) Eb (%)
GSPI-T1 300
GSPI-T2 295
GSPI-P1 332
GSPI-P2 331
GSPI-P3 336
507
504
496
501
496
523
523
503
508
501
540
540
520
524
520
100
115
100
90
2.8
2.9
2.6
2.5
2.1
4.6
4.3
5.5
5.0
7.4
96
a
Tg: determined by DSC; Td: onset decomposition temperature; T5, T10: the
decomposition temperature at 5% and 10% weight loss, respectively; TS: tensile
strength; TM: Young's modulus; Eb: elongation at break.
improvement in selectivity. For example, P(CO2) and P(CH4) for the
GSPI-P1 membrane are 10.3 and 5.9 times for the GSPI-T1 mem-
brane, respectively, meanwhile, the a(CO2/CH4) for the former is 1.8
times of the latter. It is worth to mention that the CO2 and CH4
permeability of GSPI-P membranes is superior to the commercially
available Matrimid membrane, which has P(CO2) of 10 barrer and
P(CH4) of 0.28 barrer [42,43], without significant loss of selectivity.
It is well known that the permeability and selectivity of dense
polymer membranes are strongly depended on their microstruc-
ture. The average interspacing distance (d-spacing) is considered to
represent the distance between segments of different chains and
correlate with the permeability coefficient of membranes [44]. The
X-ray diffraction patterns of fluorinated polyimides were investi-
gated and the d-spacing values of these polymers were determined
from the angle of maximal peaks by Bragg's equation. A broad
amorphous halo is observed for all the polyimides as shown in
Fig. 5. The amorphous nature of these polyimides is attributed to
the presence of bulky eC(CF3)2- group in the dianhydride com-
bined with bulky triptycene moiety or pendent phenyl group in the
diamine, resulting in loose chain packing and aggregation. It can be
seen that GSPI-T polyimides revealed the relatively broader
diffraction peaks than GSPI-P ones, which suggested the bulky
triptycene moiety is more efficient to separate molecular chains
from dense arrangement. The GSPI-T and GSPI-P polyimides gave
the maximal peaks at 13.3ꢁe13.5ꢁ and 15.0e15.2ꢁ that correspond
to the d-spacing of 6.57e6.64 Å and 5.83e5.97 Å, respectively. From
the XRD pattern of GSPI-T2, a shoulder diffraction peak on the high
Fig. 5. XRD patterns of fluorinated polyimides.
not only correlate with the average distance of chain segments but
also rely on the free spaces in microstructure. In addition, GSPI-P
polyimides displayed relatively sharper XRD curves, suggesting a
certain degree of regularity in the interchain distance. This result is
probably related to their better gas selectivity especially for gas pair
of CO2/CH4 and CO2/N2.
The free volume is the sum of the static voids which are created
by chain packing or transient gaps generated by thermally induced
chain rearrangement and provides the diffusing molecules a low-
resistance path for their transport. The fractional free volumes of
these polymers were investigated by both molecular dynamics
simulation and experimental approach in order to clarify the rela-
tionship between gas permeability and microstructure of mem-
brane. Fig. 6 is a schematic representation of a simulated molecular
cell of GSPI-P3, in which three molecular chains composed of 6FDA
and BATFM with 30 repeating units for each were set in the cubic
cell. The blue and the gray areas in the simulation cell are related to
the free volume and the occupied volume of GSPI-P3, respectively.
The fractional free volume of fluorinated polyimides were calcu-
lated from the specific volume and the occupied volume of the
polymer based on their optimized model structure.
Table 5 presents the density, cell length, specific volume and
occupied volume obtained by molecular dynamics simulation as
well as the fractional free volume (FFVsim) for fluorinated polyimide
membranes. The density of these polymers in stable state in the
simulation cell varied from 1.180 to 1.346 g/cm3, which is relatively
low compared to other 6FDA based polyimides [45]. The cell
lengths for GSPI-P membranes were around 45 Å, whereas that for
GSPI-T membranes increased to 46.96e48.52 Å, as a result, the
latter gave the higher specific volume than the former. However,
the GSPI-P membranes exhibited the lower occupied volume than
GSPI-T ones, which finally lead to the larger FFVsim values. It is noted
that the FFVsim results for these polyimide membranes are incon-
sistent with the finding from the corresponding d-spacing. We
consider that the bulky triptycene groups in GSPI-T polyimides may
separate molecular chains more effectively than pendent phenyl
groups in GSPI-P polyimides and result in larger d-spacing and
specific volume. However, the triptycene groups in the polymer
chains also occupy more volume, as a consequence, providing less
free volume. It is worth to mention that GSPI-P membranes
exhibited the FFVsim of 0.2274e0.2517, which is higher than most of
6FDA based polyimide [46]. The larger FFV of GSPI-P membranes is
angle side (2
q
¼ 16.7ꢁ) with d-spacing of 5.31 Å is also observed.
This peak could be assigned to interlocking of phenyl rings of
triptycene perpendicular to polymer chain, and then
teractions between phenyl rings [34].
pep in-
Generally, the permeability of penetrant gases in a polymer
membrane enhances with the d-spacing increasing. The perme-
ability for both GSPI-T and GSPI-P membranes showed an
increasing trend accompanied with the d-spacing increasing due to
incorporation of bulky trifluoromethyl or fluorine substituents.
However, as comparing the GSPI-P and GSPI-T membranes, it is
unexpected that the former exhibited significantly higher perme-
ability coefficients, although they gave the smaller d-spacing than
the latter. We suspect that the gas permeation in membrane may
Table 4
Permeability coefficient and selectivity of fluorinated polyimide membranes at 1 bar
and 23 ꢁC.
PI
P (barrer)a
CO2 O2
11.65
a
(PA/PB)
N2
CH4 CO2/CH4 CO2/N2 O2/N2 N2/CH4
GSPI-T1
GSPI-T2
4.10 1.57 0.78 14.9
7.22 2.07 1.06 26.7
7.4
13.6
24.1
23.7
20.1
2.6
3.5
4.2
4.3
3.8
2.0
2.0
1.1
1.1
1.1
28.26
GSPI-P1 120.20 21.20 4.99 4.57 26.3
GSPI-P2 119.10 21.59 5.03 4.41 27.0
GSPI-P3 142.61 27.07 7.11 6.62 21.5
a
Permeability in barrers, where 1 barrer ¼ 10ꢀ10 cm3(STP)cm/cm2 s cmHg.

H. Tong et al. / Polymer 69 (2015) 138e147
145
Fig. 7. Relationship between CO2 permeability and simulated fractional free volume of
fluorinated polyimide membranes.
Fig. 6. Schematic representation of a simulated molecular cell of GSPI-P3.
interaction effect of fluorine substituents for GSPI-P3 is superior to
the permeability decreasing caused by the reduction of FFV, as a
result, it revealed the relatively higher permeability. From this
point of view, we can say that the opposite effects on the perme-
ability achieved a good balance for GSPI-P2 because of the less
fluorine atoms in the side chain, which gave the permeability co-
efficient similar to GSPI-P1.
The free volume in a polymeric membrane generally consists of
many isolated spaces with variable volume and size, which pro-
vides variation in the membrane permeability and selectivity [49].
The positron annihilation lifetime spectroscopy, an experimental
approach to determine the size and concentration of free volume
cavities in a polymeric membrane, was adapted to verify the FFV
results derived from theoretical simulating and furthermore un-
derstand the gas separation performance of these fluorinated pol-
yimide membranes. The positronium lifetimes, intensities, average
directly associated with their higher gas permeability as illustrated
in Table 4, even though they gave narrower d-spacing than GSPI-T
membranes.
The relationships between CO2 permeability and their simulated
fractional free volume for these polyimide membranes are depicted
in Fig. 7. For GSPI-T membranes, FFVsim values enhanced with the
incorporation of large trifluoromethyl substituents in the backbone
due to their steric hinder effect, which coincides with their
changing trend of d-spacing. Therefore, the GSPI-T2 revealed
improved CO2 permeability than GSPI-T1. On the other hand, for
GSPI-P membranes, an unanticipated decreasing in FFVsim values
accompanied with the introducing of fluorine substituents in the
pendent phenyl group was observed. According to the simulation
results listed in Table 5, it is found that the cell length and specific
volume for GSPI-P membranes showed a slight decreasing com-
bined with an increasing in occupied volume with the enhancing of
fluorine content. The decreasing in FFVsim of GSPI-P membranes
can be explained by the filling effect of fluorine atoms. As shown
schematically in Fig. 8, the specific volume for all the GSPI-P
membranes is dominated by the bulky pendent phenyl moieties.
As the fluorine substituents are incorporated to the pendent phenyl
rings, the fluorine atoms will fill in the internal space and result in
less free volume. A similar filling effect of trifluoromethyl and
methyl groups for pentiptycene-based polyimides has been re-
ported by Luo et al. [47]. However, it is difficult to understand that
the permeability coefficients for GSPI-P membranes are in the order
of GSPI-P1 z GSPI-P2<GSPI-P3, which is in conflict with the order
of FFVsim. Stern et al. pointed that the polar fluorine-containing
groups could induce specific interactions with the penetrant mol-
ecules, and such interactions can significantly enhance the solubi-
lity process, which was favorable to gas permeation [48]. We
speculate that the permeability increasing related with the specific
cavity radii and the corresponding fractional free volumes (FFVPALS
)
are summarized in Table 6. The FFVPALS values for GSPI-T and GSPI-P
polyimides were in the range of 0.1069e0.1590 and 0.2082e0.2224,
respectively. It can be seen that the changing trend of FFVPALS values
was completely consistent with that of FFVsim results, demon-
strating the reliable FFV results were obtained from molecular
simulation method. It is also found that GSPI-P membranes gave
the longer t3 combined with higher I3 values as comparing to GSPI-
T ones. Meanwhile, the average cavity radii for the former were in
the range of 3.42e3.45 Å, while that for the latter slightly decreased
to 3.33e3.40 Å. The lifetime (t3) and intensity (I3) of orthoposi-
tronium have been found to be correlated well with the local free
volume in polymers in which t3 is proportional to the size of the
free volume cavity and the I3 is proportional to their number con-
centration [50]. Therefore, it can be deduced that the superior gas
permeability of GSPI-P membranes is attributed to their larger
mean cavity size and the higher free volume. The intensity and the
average cavity radius for GSPI-T membranes showed an increasing
Table 5
The density, cell length, specific volume and occupied volume of simulation cells as well as calculated fraction free volume for fluorinated polyimide membranes.
PI
Density (g/cm3)
Cell length(Å)
Specific volume (Å3)
Occupied volume (Å3)
FFVsim
GSPI-T1
GSPI-T2
GSPI-P1
GSPI-P2
GSPI-P3
1.286
1.346
1.180
1.226
1.280
46.96
48.52
45.44
45.25
45.27
103558
114211
93805
92584
92781
81376
88574
70196
70260
71683
0.2142
0.2245
0.2517
0.2411
0.2274

146
H. Tong et al. / Polymer 69 (2015) 138e147
Fig. 8. Schematic of filling effect of fluorine substituent for GSPI-P membranes.
Table 6
positron annihilation measurements, it is found that the cavity size
of GSPI-P membranes is larger than the dynamic radius of CO2 and
smaller than that of the other gaseous molecules, which made
them favorable to separate the CO2 from the other gases. These
results suggested that the large quantity of free volume combined
with appropriate cavity size in polymeric membranes is beneficial
to improve their permeability and selectivity simultaneously.
The positronium lifetime, intensity, free volume radius and fractional free volume
for fluorinated polyimide membranes.
PI
t3 (ns)
I3 (%)
R (Å)
FFVPALS
GSPI-T1
GSPI-T2
GSPI-P1
GSPI-P2
GSPI-P3
2.56 0.03
2.66 0.02
2.73 0.02
2.72 0.02
2.68 0.02
6.93 0.15
9.66 0.19
12.91 0.19
12.90 0.20
12.46 0.19
3.33
3.40
3.45
3.44
3.42
0.1069
0.1590
0.2224
0.2197
0.2082
Acknowledgment
The authors gratefully acknowledge the financial support from
National Basic Research Program of China (2014CB643600).
trend along with the incorporation of fluorinated groups, in
contrast, those for GSPI-P membranes exhibited a decreasing trend
due to the filling effect of fluorine atoms. The corresponding FFVPALS
values of these polyimides revealed the similar changing. It is also
noted that the GSPI-P membranes have the appropriate cavity size
(R ¼ 3.42e3.45 Å), which is larger than the dynamic radius of CO2
(3.30 Å) and smaller than that of the other gaseous molecules
(3.46e3.80 Å). The specific cavity size of GSPI-P membranes made
them favorable to transport the CO2 and prevent the other larger
gaseous molecules, which providing them improved CO2/CH4 and
CO2/N2 selectivity. The GSPI-T2 membrane displayed selectivity
comparable to GSPI-P ones because of the similar cavity size,
whereas, gave much lower permeability coefficients due to the
lower FFVPALS. These results suggested that the gas transport
properties of fluorinated polyimide membranes were strongly
dependent on their microstructure. The large quantity of free vol-
ume combined with appropriate cavity size in polymeric mem-
branes is beneficial to improve their permeability and selectivity
simultaneously.
References
[1] G. García Maria, J. Marchese, A. Ochoa Nelio, Int. J. Hydrog. Energy 35 (17)
(2010) 8983e8992.
[2] Z.P. Smith, R.R. Tiwari, T.M. Murphy, D.F. Sanders, K.L. Gleason, D.R. Paul,
B.D. Freeman, Polymer 54 (12) (2013) 3026e3037.
[3] M. Cecopieri-Gomez, J. Palaciosalquisira, J. Dominguez, J. Membr. Sci. 297
(1e2) (2007) 53e65.
[4] Y.C. Xiao, B.T. Low, S.S. Hosseini, T.S. Chung, D.R. Paul, Prog. Polym. Sci. 34 (6)
(2009) 561e580.
[5] S. Basu, A.L. Khan, A. Cano-Odena, C.Q. Liu, I.F.J. Vankelecom, Chem. Soc. Rev.
39 (2) (2010) 750e768.
[6] I.C. Omole, S.J. Miller, W.J. Koros, Ind. Eng. Chem. Res. 49 (10) (2010)
4887e4896.
[7] C.A. Scholes, G.W. Stevens, S.E. Kentish, Fuel 96 (1) (2012) 15e28.
[8] R. Swaidan, B. Ghanem, M. Al-Saeedi, E. Litwiller, I. Pinnau, Macromolecules
47 (21) (2014) 7453e7462.
[9] R. Swaidan, B. Ghanem, E. Litwiller, I. Pinnau, J. Membr. Sci. 475 (2015)
571e581.
[10] L. Shao, B.T. Low, T.S. Chung, A.R. Greenberg, J. Memb. Sci. 327 (1e2) (2009)
18e31.
[11] F.W. Wu, L. Li, Z.H. Xu, S.J. Tan, Z.B. Zhang, Chem. Eng. J. 117 (1) (2006) 51e59.
[12] K. Berean, J.Z. Ou, M. Nour, K. Latham, C. McSweeney, D. Paull, A. Halim,
S. Kentish, C.M. Doherty, A.J. Hill, K. Kalantar-zadeh, Sep. Purif. Technol. 122
(2014) 96e104.
4. Conclusions
The novel aromatic diamines with pendent phenyl structure, i.e.,
BAPM, BAFM and BATFM, were synthesized and characterized.
These diamines and the diamines with triptycene moieties were
successfully polymerized with commercial available dianhydride
6FDA via one-pot solution polycondensation at high temperature to
gain the GSPI-P and GSPI-T polyimides, respectively. These poly-
imides exhibited excellent solubility in various organic solvents due
to the presence of bulky triptycene and pendent phenyl moieties.
They also revealed high thermal stabilities and outstanding me-
chanical properties, which made them suitable for further inves-
tigation of gas transport properties. The GSPI-P membranes
revealed much higher permeability coefficients than the GSPI-T
membranes combined with good selectivity. The larger fractional
free volume of GSPI-P membranes is directly associated with their
higher gas permeability, even though they gave narrower d-spacing
than GSPI-T membranes. The GSPI-P membranes also provided
good selectivity for CO2/CH4 and CO2/N2 gas pairs. According to the
[13] J.S. McHattie, W.J. Koros, D.R. Paul, Polymer 32 (14) (1991) 2618e2625.
[14] J.S. McHattie, W.J. Koros, D.R. Paul, Polymer 33 (8) (1992) 1701e1711.
[15] C. Camacho-Zuniga, F.A. Ruiz-Trevino, S. Hernandez-Lopez, M.G. Zolotukhin,
F.H.J. Maurer, A. Gonzalez-Montiel, J. Membr. Sci. 340 (1e2) (2009) 221e226.
[16] P. Moradihamedani, N.A. Ibrahim, W.M.Z.W. Yunus, N.A. Yusof, Polym. Eng.
Sci. 54 (7) (2014) 1686e1694.
[17] R.T. Chern, L. Jia, S. Shimoda, H.B. Hopfenberg, J. Membr. Sci. 48 (2e3) (1990)
333e341.
[18] G. Chowdhury, R. Vujosevic, T. Matsuura, B. Lavert, J. App. Polym. Sci. 77 (5)
(2000) 1137e1143.
[19] E. Sada, H. Kumazawa, J.S. Wang, M. Koizumi, J. App. Polym. Sci. 45 (12) (1992)
2181e2186.
[20] R. Morita, F.Z. Khan, S. Sakaguchi, M. Shiotsuki, Y. Nishio, T. Masuda, J. Membr.
Sci. 305 (1e2) (2007) 136e145.
[21] K. Tanaka, Y. Osaka, H. Kita, K.I. Okamoto, J. Polym. Sci. Part B Polym. Phys. 33
(13) (1995) 1907e1915.
[22] W.H. Lin, R.H. Vora, T.S. Chung, J. Polym. Sci. Part B Polym. Phys. 38 (21) (2000)
2703e2713.
[23] W.H. Lin, T.S. Chung, J. Membr. Sci. 186 (2) (2001) 183e193.
[24] G. Maier, Prog. Polym. Sci. 26 (1) (2001) 3e65.
[25] D.J. Liaw, K.L. Wang, Y.C. Huang, K.R. Lee, J.Y. Lai, C.S. Ha, Prog. Polym. Sci. 37
(7) (2012) 907e974.

H. Tong et al. / Polymer 69 (2015) 138e147
147
[26] W.L. Qiu, C.C. Chen, M.R. Kincer, W.J. Koros, Polymer 52 (18) (2011)
4073e4082.
[27] L.M. Robeson, J. Membr. Sci. 320 (1e2) (2008) 390e400.
[28] M. Calle, C. Garcia, A.E. Lozano, J.G. de la Campa, J. de Abajo, C. Alvarez,
J. Membr. Sci. 434 (2013) 121e129.
[39] F. Sun, T.P. Wang, S.Y. Yang, L. Fan, Polymer 51 (17) (2010) 3887e3898.
[40] X.J. Zhao, J.G. Liu, H.X. Yang, L. Fan, S.Y. Yang, Eur. Polym. J. 44 (2008)
808e820.
[41] Y.H. Kim, B.S. Moon, F.W. Harris, S.Z.D. Chen, J. Therm. Anal. 46 (1996)
921e933.
[29] K. Tanaka, H. Kita, M. Okano, K.I. Okamoto, Polymer 33 (3) (1992) 585e592.
[30] Y. Li, X.Q. Wang, M.X. Ding, J. Xu, J. App. Polym. Sci. 61 (5) (1996) 741e748.
[31] W.L. Qiu, L. Xu, C.C. Chen, D.R. Paul, W.J. Koros, Polymer 54 (22) (2013)
6226e6235.
[42] D.R. Paul, Y.P. Yampolskii, Polymer Gas Separation Membranes, CRC Press,
Boca Raton, FL, 1994.
[43] D.F. Sanders, Z.P. Smith, R.L. Guo, L.M. Robeson, J.E. McGrath, D.R. Paul,
B.D. Freeman, Polymer 54 (18) (2013) 4729e4761.
[32] N.Y. Du, H.B. Park, M.M. Dal-Cin, M.D. Guiver, Energy Environ. Sci. 5 (6) (2012)
7306e7322.
[33] K. Tanaka, M. Okano, H. Toshino, H. Kita, K.I. Okamoto, J. Polym. Sci. Part B
Polym. Phys. 30 (8) (1992) 907e914.
[44] K.C. Obrien, W.J. Koros, J. Membr. Sci. 35 (2) (1988) 217e230.
[45] H. Ohya, V.V. Kudryavsev, S.I. Semenova, Polyimide Membranes: Applications,
Fabrications and Properties, Gordon and Breach Publishers, Tokyo, 1996.
ꢀ
[46] C. Bas, R. Mercier, C. Dauwe, N.D. Alberola, J. Membr. Sci. 349 (1e2) (2010)
[34] Y.J. Cho, H.B. Park, Macromol. Rapid Commun. 32 (7) (2011) 579e586.
[35] M. Calle, A.E. Lozano, J. de Abajo, J.G. de la Campa, C. Alvarez, J. Membr. Sci.
25e34.
ꢀ
[47] S.J. Luo, Q. Liu, B.H. Zhang, J.R. Wiegand, B.D. Freeman, R.L. Guo, J. Membr. Sci.
480 (2015) 20e30.
365 (1e2) (2010) 145e153.
[36] M. Carta, R. Malpass-Evans, M. Croad, Y. Rogan, J.C. Jansen, P. Bernardo,
F. Bazzarelli, N.B. McKeown, Science 339 (6117) (2013) 303e307.
[37] H.B. Park, C.H. Jung, Y.M. Lee, A.J. Hill, S.J. Pas, S.T. Mudie, E. Van Wagner,
B.D. Freeman, D.J. Cookson, Science 318 (5848) (2007) 254e258.
[38] S.H. Hsiao, H.M. Wang, W.J. Chen, T.M. Lee, C.M. Leu, J. Polym. Sci. Part A
Polym. Chem. 49 (14) (2011) 3109e3120.
[48] S.A. Stern, Y. Mi, H. Yamamoto, J. Polym. Sci. Part B Polym. Phys. 27 (9) (1989)
1887e1909.
[49] T.M. Madkour, J.E. Mark, J. Membr. Sci. 431 (2013) 37e46.
[50] J. Zhang, A.Z.M.S. Rahman, Y.X. Li, J. Yang, B.Z. Zhao, E. Lu, P. Zhang, X.Z. Cao,
R. Yu, B.Y. Wang, Opt. Mater. 36 (2) (2013) 471e475.