N6

MD

a

ARRAA

KTCNO

1

urser(

fpc

0d

Journal of Membrane Science 390– 391 (2012) 141– 151

Contents lists available at SciVerse ScienceDirect

Journal of Membrane Science

jo u rn al hom epa ge: www.elsev ier .com/ locate /memsci

atural gas purification and olefin/paraffin separation using cross-linkableFDA-Durene/DABA co-polyimides grafted with �, �, and �-cyclodextrin

ohammad Askari, Youchang Xiao, Pei Li, Tai-Shung Chung ∗

epartment of Chemical and Biomolecular Engineering, National University of Singapore, 10 Kent Ridge Crescent 4 Engineering Drive 4, Singapore 117576, Singapore

r t i c l e i n f o

rticle history:eceived 8 October 2011eceived in revised form 7 November 2011ccepted 15 November 2011vailable online 22 November 2011

eywords:hermal cross-linkable co-polyimideyclodextrinatural gas purificationlefin/paraffin separation

a b s t r a c t

Using a cross-linkable co-polyimide (6FDA-Durene/DABA (9/1)), we have developed new flexible andhigh-performance gas separation membranes that can enhance both membrane permeability and plasti-cization resistance simultaneously by grafting various sizes of cyclodextrin to the polyimide matrix andthen decomposing them at elevated temperatures. The gas permeability of thermally treated pristinepolyimide (referred as the original PI) and CD grafted co-polyimide (referred as PI-g-CDs for 200 and300 ◦C and partially pyrolyzed membranes (PPM)-CDs for 350, 400, and 425 ◦C) has been determinedusing O2, N2, CO2, CH4, C3H6, and C3H8 at 35 ◦C. The permeability of all gases increases with an increasein thermal treatment temperature from 200 to 425 ◦C. However, permeability increases more for thosegrafted with bigger size CD. Permeability of the original PI thermally treated at 425 ◦C is about 4–6 timeshigher than that treated at 200 ◦C. The permeability increase jumps to 8–10 times for PPM-�-CD and15–17 times for PPM-�-CD due to CD decomposition at high temperatures and bigger CD creating biggermicro-pores. Interestingly, the permeability ratios of PPM-�-CD to PPM-�-CD and PPM-�-CD to PPM-�-CD at 400 and 425 ◦C are around 0.6 and 0.8, respectively. These numbers are almost the same asthe cavity diameter ratios of �-CD to �-CD and �-CD to �-CD. Clearly, the bigger CD creates the big-ger micro-pores. Permselectivity decreases first with an increase in thermal treatment temperature upto 350 ◦C and then increases. Permselectivity of thermally treated CD grafted co-polyimide membranesis also slightly higher than that of the original PI due to higher degrees of cross-linking in CD graftedco-polyimide membranes. In addition, for co-polyimide membranes grafted by CDs, the higher thermal

treatment temperature results in membranes with the better plasticization resistance to CO2 and the bet-ter separation performance for 50:50 CO2/CH4 mixed gases. The best result for pure gas tests is achievedfor PPM-�-CD-425. This membrane has a CO2 permeability of 4211 Barrers with a CO2/CH4 ideal selec-tivity of 22.44 and a C3H6 permeability of 521 Barrers with a C3H6/C3H8 ideal selectivity of 18.09. It canalso resist against CO2 plasticization until 30 atm. The CO2 permeability drops slightly to 3976 Barrers

2/CH4

with almost the same CO

. Introduction

Separation of gases is one of practical but very importantnit operations in chemical and petrochemical industries such asecovery of hydrogen from product streams of ammonia plants,eparation of methane from the other components of biogas,nrichment of air by oxygen for medical or metallurgical purposes,emoval of hydrogen, water vapor, CO2, and H2S from natural gasnatural gas purification) and olefin/paraffin separation [1].

The demand of natural gas is continuously growing due to the

act that not only is natural gas a clean and efficient fuel, but also arinciple feedstock for the manufacture of many important chemi-als. Removal of acid gases (CO2) is one of the major steps in natural

∗ Corresponding author. Tel.: +65 6516 6645; fax: +65 6779 1936.E-mail address: chencts@nus.edu.sg (T.-S. Chung).

376-7388/$ – see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.memsci.2011.11.030

selectivity of 22.84 in mixed gas tests.© 2011 Elsevier B.V. All rights reserved.

gas purification because it can (i) increase the heating value of natu-ral gas, (ii) decrease the volume of gas to be transported in pipelinesand cylinders, (iii) prevent corrosion of pipeline during gas trans-port and distribution and (iv) reduce atmospheric pollution [2–4].Therefore, CO2/CH4 separation is very important for natural gassweetening. Olefins and other saturated hydrocarbons are existedin various petrochemical streams that mostly come from steamcracking units, catalytic cracking units, or the dehydrogenation ofparaffins [5]. Olefins are used for the production of esters, alco-hol, polymers, and other synthesis intermediates. Therefore, oneof the most important processes in petrochemical industries andpetroleum refining is the separation of olefins from the correspond-ing paraffins [6].

Currently, amine absorption and pressure swing adsorption aremajor methods for natural gas purification and the separationof olefin and paraffin mixtures is typically performed using rec-tification, adsorption, and cryogenic distillation [2–4,7–9]. These

142 M. Askari et al. / Journal of Membrane Science 390– 391 (2012) 141– 151

f three

mctbainsoekeohrgmr

(hmtaaMmHtdpiswbapigacp

btissipc

posed and formed free volumes in the polymer matrix and theresultant membrane showed extremely enhanced permeability. Inaddition, carboxylic groups of the polyimide backbone underwent

Fig. 1. Chemical structure o

ethods are expensive and energy intensive. The increase in energyost due to the resources depletion have arisen the demands forhe development of low energy separation technologies. Mem-rane separation, compared to amine absorption, pressure swingdsorption, rectification and cryogenic distillation, has advantagesncluding low energy consumption, easy operation and mainte-ance, environmental benign and small footprint [10–12]. Manytudies have been done on membrane separation for the purposef replacing traditional separation technologies [3]. Membranes,specially polymeric membranes have been explored in variousind of gas separation applications such as natural gas sweet-ning [2–4,11,12], and olefin/paraffin separation [5,6,13–19]. Inrder to have a good gas separation performance, membranes mustave high permeability and permselectivity, excellent chemicalesistance (for resistance against corrosive materials such as H2S),reat thermal stability (for high temperature applications), goodechanical properties (for high pressure applications), and supe-

ior plasticization resistance [11,20,21].Aromatic polyimides, especially the ones containing 4,4′-

hexafluoroisopropylidene) diphthalic anhydride (6FDA) moiety,ave shown good mechanical and thermal properties, high gas per-eability as well as good permselectivity. It has been found that

he –(CF3)– group restricts the torsional motion of phenyl ringsnd chain packing in 6FDA dianhydrides so that both permeabilitynd permselectivity may be improved simultaneously [16,17,22].any 6FDA-based polyimides exhibit good separation perfor-ance for CO2/CH4 and C3H6/C3H8 gas pairs [4,5,15–17,19,20,22].owever, similar to other polymers, 6FDA polyimides still face

wo major challenges. First, there is a limitation in achieving theesired performance of a high permeability combined with a highermselectivity. There is a trade-off relation between permeabil-

ty and permselectivity for most gas pairs [10]. Secondly, theeparation performance of 6FDA polyimides generally degradeshen separating highly soluble gases such as CO2 or hydrocar-

ons at relatively high pressures. This phenomenon is referreds plasticization [19–24]. In the occurrence of plasticization, theenetrant species such as CO2 swell up the polymer matrix, facil-

tate segmental motions, and accelerates diffusion processes of allas species [25,26]. The gas pair selectivity is therefore reducednd the ideal selectivity measured by means of pure gas testsan no longer be used to estimate the mixed gas membraneerformance [27].

Cross-linking has been advised as a tactic to improve the mem-rane performance by decreasing plasticization effects, improvinghe permselectivity, as well as improving the thermal and chem-cal stability of the membranes [4,6,20,21,28–32]. Kita et al. [28]tudied the effects of photo cross-linking on the permeability and

electivity of BTDA-containing polyimides. They reported that withncreasing UV-irradiation time, considerable improvement in gasair selectivity was achieved but gas permeability was signifi-antly reduced due to the densification of membrane structure and

kinds of cyclodextrin [35].

reduction of polymer chains mobility. Staudt-Bickel and Koros [29]studied the chemical cross-linking between the free carboxylic acidgroup of 6FDA-mPD/DABA and ethylene glycol. They found that theCO2/CH4 selectivity increased with an increment in the degree ofcross-linking. Interestingly, the CO2 permeability did not signifi-cantly decrease because the ethylene glycol group not only createdmore free volume but also reacted with the carboxylic acid groupand inhibited the hydrogen bonding. Shao et al. [30] employed dif-ferent generations of diaminobutane (DAB) dendrimers to inducecross-linking reactions. The membrane selectivity increased butthe permeability decreased. Xiao et al. [21] cross-linked 6FDA-Durene polyimide containing internal acetylene units for thermalcross-linking. They reported that increasing amounts of cross-linking sites led to a decrease in chain mobility and an incrementin gas selectivity. However, gas permeability decreased afterthermal cross-linking and went lower with increased molar ratioof cross-linking sites.

Since cross-linking normally causes a severe decrease inpolymer permeability by reducing FFV and chain flexibility, inves-tigations have been carried out to design high free volumepolyimides to alleviate the decrease in permeability. One of themethods to prepare high free volume membrane is the decompo-sition of labile components in block or graft co-polymers as lowk materials used in semi-conductor industries [32,33]. Recently inour group [20], �-cyclodextrin (�-CD), a thermal labile moleculewith a large molecular size [34], has been grafted on a polyimide.After thermal treatment, the thermal labile molecules decom-

Fig. 2. Thermo gravimetric analyses (TGA) of three kinds of cyclodextrin.

rane S

caag3Ctrtaa

6Crbotsfvldstoifevm

2

2

spbdtf1amia(w9

2

atdamathsw

M. Askari et al. / Journal of Memb

ross-linking reactions at high temperatures, tightened d-space,nd increased the gas selectivity and plasticization resistance. Inddition, the free volume radius (mean cavity radius) of �-CDrafted co-polyimides after heat treatment at 425 ◦C was around.8 A (cavity diameter is around 7.6 A) which is higher than CO2,H4, C3H6, and C3H8 kinetic diameters. Since the kinetic diame-ers of CO2, CH4, C3H6, and C3H8 are 3.3 A, 3.8 A, 4.5 A, and 4.3 A,espectively, better selectivity for CO2/CH4 and C3H6/C3H8 separa-ion may be achievable if one can manipulate the free volume radiusnd increase the total free volume by changing the cyclodextrin sizend optimizing the heat-treatment conditions.

In this work, three cyclodextrins (CDs) were grafted on theFDA-Durene/DABA co-polyimide by esterification [20]. The threeDs, namely, �-, �-, and �-CD, are formed by six to eight glucopy-anoside units. Their chemical structure and basic properties haveeen presented in Figs. 1 and 2 and Table 1 [35]. Thermal treatmentf the co-polyimide was conducted from 300 to 425 ◦C to studyhe effects of different annealing temperatures on the membraneeparation performance. The functions of thermal treatment are asollows (1) decomposing the structure of CD to create more poreolumes in the membrane matrix, and (2) inducing thermally cross-inked rigid-chain structure that avoids pore collapsing [20,36]. Theesired cross-linked structure has been shown in Fig. 3. As can beeen in Fig. 1, the sizes of these cyclodextrins are different. Afterhermal treatment, the decomposed CD would leave different sizesf microvoids in the membrane. Therefore, the purpose of this works to explore the science and engineering if one can incorporate dif-erent sizes of CD and manipulate the free volume and its size withnhanced permeability and selectivity of the resultant membraneia by thermal degradation and partial cross-linking of the polymeratrix.

. Experimental

.1. Materials

4,4′-(Hexafluoroisopropylidene) diphthalic anhydride (6FDA),upplied by Clariant, and 3,5-diaminobenzoic acid (DABA) sup-lied by Aldrich (Singapore) were purified by vacuum sublimationefore usage. 2,3,5,6-Tetramethyl-1,4-phenylenediamine (Dureneiamine) supplied by Aldrich (Singapore), was recrystallized twoimes from methanol. 1-Methyl-2-pyrrolidone (NMP) purchasedrom Merck Chemicals (Germany), was distilled at 42 ◦C under

mbar and was dried with molecular sieve before usage. Aceticnhydride was received from Aldrich (Singapore) and dried witholecular sieve before usage. Other chemicals and solvents includ-

ng triethylamine, methanol, p-toluenesulfonic acid, and �, �,nd �-cyclodextrin were all reagent grade or better from AldrichSingapore) and were used without further purification. The gasesere supplied by SOXAL (Singapore) with purities higher than

9.95%.

.2. Co-polyimide synthesis

The 6FDA-Durene/DABA (9:1) co-polyimide was synthesized via two-step chemical imidization in a NMP solution. For the prepara-ion of polyamic acid, a designated molar ratio of Durene and DABAiamine was dissolved in NMP under nitrogen atmosphere and thenn equimolar of solid 6FDA was gradually added to the solution in aoisture free flask with mega-stirring under nitrogen atmosphere

t room temperature. The reason of adding solid 6FDA gradually is

o suppress the reaction between water and 6FDA. Aromatic dian-ydrides can react with water and other impurities in the amideolvents, but have a slower reaction rate relative to their reactingith diamines. Therefore, gradually adding the solid 6FDA reduces

cience 390– 391 (2012) 141– 151 143

its availability for competing reactions with water or other impu-rities, thus leads to a higher conversion and molecular weight [37].After reacted for 24 h, a high molecular weight polyamic acid wasformed. Then for the chemical imidization step, a mixture of aceticanhydride and triethylamine at 4:1 molar ratio were slowly addedto the polyamic acid solution to perform imidization for 24 h undernitrogen atmosphere. Finally, the co-polyimide was precipitated inmethanol, filtered, washed and dried at 120 ◦C in vacuum for 24 h.The schemes of co-polyimide synthesis and chemical structure ofthis co-polyimide are shown in Fig. 4.

2.3. Cyclodextrin grafting

Cyclodextrin (CD) was grafted to the co-polyimide via esterifi-cation. For this purpose, co-polyimide and �, �, and �-CD were firstdried at 120 ◦C in vacuum for 12 h. After that co-polyimide was dis-solved in NMP and then a large excess amount of CD (3–5 timesmore than stoichiometric balance to the carboxylic acid group)was added to the NMP solution. The esterification was carriedout by adding a catalytic amount (5 mg per gram of polymer) ofp-toluenesulfonic acid and heating to 120 ◦C under nitrogen atmo-sphere for 18 h. Finally, the resultant co-polyimide grafted with �,�, or �-CD (referred as PI-g-�-CD, PI-g-�-CD, or PI-g-�-CD) wasprecipitated in methanol, washed to remove unreacted CD anddried under 120 ◦C in vacuum for 24 h. The chemical structure ofco-polyimide after grafting CD is shown in Fig. 4.

2.4. Dense flat sheet film formation

Dense films were prepared using the solution casting method.Prior to use, the original co-polyimide (PI) and PI-g-CDs were driedovernight at 120 ◦C under vacuum. A 2% (w/w) polymer solutionwas prepared by dissolving the original PI and PI-g-CDs together inNMP with a weight ratio of 1:1. The polymer solution was magneti-cally stirred overnight and filtered using a Whatman’s filter (1 �m)to remove dust and insoluble particles before it was cast onto asilicon wafer covered with a steel ring. The polymer casting solu-tion was heated at 40 ◦C for 12 h and vacuum heated at 70 ◦C for24 h, respectively. Then, the dense films were formed and pealedoff from the silicon wafer and placed into a vacuum oven to be fur-ther vacuum heated at 200 ◦C for 24 h in purpose of removing a traceamount of residual solvent. The absent of residual solvent was con-firmed by TGA. Finally, films with a thickness of 50 ± 10 �m wereprepared for testing and characterization in the following studies.

2.5. Thermal cross-linking

The thermal treatment was performed using a CenturionTM

Neytech Qex vacuum furnace (Yucaipa, CA, USA). The films wereheated under vacuum from room temperature to specific temper-atures such as 300, 350, 400, and 425 ◦C. The heating rate was10 ◦C/min and the films were held for 120 min at the final tempera-tures. After completing the thermal treatment process, membraneswere cooled naturally in the vacuum furnace to room temperatureand used for permeation tests. All thermal treated membranes wereflexible and mechanically strong enough for gas permeation tests.For sample classification, the treated samples at 200 and 300 ◦Cwere named as ‘polyimide-grafted-�, �, or �-CD-final treated tem-perature’, for example “PI-g-�-CD-300” and from 350 to 425 ◦Cwere name as ‘partially pyrolyzed membrane-�, �, or �-CD-finaltreated temperature’, for example “PPM-�-CD-400”.

2.6. Characterizations

Gel permeation chromatography (GPC) was used to measurethe molecular weights of the co-polyimides before and after

144 M. Askari et al. / Journal of Membrane Science 390– 391 (2012) 141– 151

Table 1Properties of the three types of cyclodextrin.

CD type No. of glucose unit Mw [35] Cavity diameter (A) [35] Cavity height (A) [35] Decomposition temp. (◦C)a

�-CD 6 972 4.7–5.7 7.8 321 ± 1�-CD 7 1135 6.0–7.8 7.8 328 ± 1

50 ◦C.

gHltwscp

ulfifsoVc

C

�-CD 8 1297 7.5–9.5

a Td (5%), at which the mass of the sample is 5.0% less than its mass measured at

rafting CD. GPC measurements were carried out on a HP 1100PLC system equipped with the HP 1047A RI detector and the Agi-

ent 79911GP-MXC columns. Tetrahydrofuran (THF) was used ashe solvent, with a controlled flow rate of 1.0 ml/min. The columnas calibrated using a standard polystyrene sample and testing

ample was 0.005 wt%. The molecular weights were estimated byomparing the retention times in the column to those of standardolystyrene.

Density was measured according to the Archimedean principlesing a Mettler Toledo balance (Singapore) with a density kit. At

east three density measures were recorded for each membranelm. Then, the fractional free volume (FFV) is calculated by the

ollowing equation [38]: FFV = (V−V0)/V, where V is the observedpecific volume calculated from the measured density and V0 is theccupied volume calculated from the correlation, V0 = 1.3 Vw where

is the van der Waals volume. V is calculated from the group

w w

ontribution method of Bondi [39].A TGA 2050 Thermo gravimetric Analyzer (TA Instruments, New

astle, DE, USA) was employed to characterize the weight loss

Fig. 3. A hypothesis of the cross-linked structure (a) low temperat

7.8 322 ± 1

of the polyimide. The analysis was carried out with a heatingramp of 10 ◦C/min from room temperature to 800 ◦C in nitro-gen atmosphere, using ∼10 mg sample. Fourier Transform InfraredSpectroscopy (FT-IR) was applied to investigate the evolution ofchemical structure changes. The FTIR measurements were per-formed using an attenuated total reflection mode (FTIR-ATR) witha Perkin-Elmer Spectrum 2000 FTIR spectrometer (Cambridge, MA,USA).

A Brukel wide-angle X-ray diffractometer (Bruker D8 advanceddiffractometer) was employed to determine the d-space value thatindicates the average inter-segmental distance of polymer chains.The measurement was completed with a step increment of 0.02◦

and Ni-filtered Cu-K� radiation at a wavelength � = 1.54 A was used.The d-space was determined based on the Bragg’s law [38] as fol-lows: n� = 2d sin �, where d, �, and � are the dimension spacing,diffraction angle, and X-ray wavelength, respectively, and n is an

integral number (1, 2, 3, . . .). A differential scanning calorimeterDSC 822e (Mettler Toledo, Singapore) was used to monitor thethermal properties of polymers and measure their glass transition

ure (below 350 ◦C) and (b) high temperature (above 350 ◦C).

M. Askari et al. / Journal of Membrane Science 390– 391 (2012) 141– 151 145

d chemical structure of co-polyimide.

ttira

2

svOTcmtt

P

wro

Fig. 4. The synthetic scheme an

emperatures (Tg) before and after grafting CD. For this purpose,esting samples were heated to 450 ◦C at a heating rate of 30 ◦C/minn nitrogen environments during the first heating, quenched at aate of 30 ◦C/min, and reheated at 30 ◦C/min. All Tg were measuredt the second heating cycle.

.7. Gas permeation measurements

The dense flat films were tested in both pure gas and mixed gasystems. The pure gas permeation properties were evaluated by aariable-pressure constant-volume method at 35 ◦C and 10 atm for2, N2, CH4, and CO2 and 3.5 atm for C3H8, and C3H6, by sequence.he film of dense membrane was mounted onto the permeationell and vacuumed at 35 ◦C for more than 12 h before the gas per-eation test was carried out. The slope of downstream pressure vs.

ime (dp/dt) at the steady state condition was used for the calcula-ion of gas permeability with accordance to the below equation:

= 273.5 × 1010 × 760 × V × L × (dp/dt)(1)

A × T × P2 × (76/14.7)

here P refers to the gas permeability of a membrane in Bar-er (1 Barrer = 1 × 10−10 cm3 (STP) cm/cm2 s cmHg), V is the volumef downstream chamber (cm3), L is the membrane thickness

Fig. 5. FTIR-ATR spectra of co-polyimide membranes at different heat treatmenttemperatures.

146 M. Askari et al. / Journal of Membrane Science 390– 391 (2012) 141– 151

Table 2Properties of the synthesized co-polyimides before and after grafting different kindsof CD.

Polymer Mn (g/mol) PDI (Mw/Mn) Inherent viscosity (dL/g)(0.5% in NMP 25 ◦C)

Original PI 82,345 3.25 1.01PI-g-�-CD 20,359 2.96 0.54

(os

vspiwgpmdaot

3

3

(iT(pael

mtc1seusasettoa

sbtbolti

Fig. 6. Residual weight and decomposition rate vs. temperature for original co-

PI-g-�-CD 17,149 2.92 0.55PI-g-�-CD 19,299 2.80 0.47

cm), A refers to the effective area of membrane (cm2), T is theperating temperature (K) and P2 is the upstream operating pres-ure (psi).

The permeability of each gas was obtained from the averagealue of at least 3 tests with deviation of smaller than 1%. The idealelectivity is defined as follows: ˛A/B = PA/PB, where PA and PB are theermeability of gases A and B, respectively. To investigate the CO2-

nduced plasticization behavior of the films, the testing pressureas intermittently ramped from 1 to 30 atm. Binary CO2/CH4 (1:1)

as permeation tests were obtained by a modified constant volumeermeation system in which an addition valve at the upstream seg-ent is included to adjust the stage cut and another valve at the

ownstream port is installed to introduce the accumulated perme-te gas to an Agilent 7890 gas chromatography (GC) for the analysisf gas composition. For easy comparison with pure gas tests, theesting pressure of mixed gas CO2/CH4 permeation tests is 20 atm.

. Result and discussion

.1. Characterization

The inherent viscosities of original PI (6FDA-Durene/DABA9:1)), PI-g-�-CD, PI-g-�-CD, and PI-g-�-CD are given in Table 2. Thenherent viscosities decrease after CD grafting in all co-polyimides.he number average molecular weight (Mn) and polydispersityPDI) of original PI, PI-g-�-CD, PI-g-�-CD, and PI-g-�-CD co-olyimides are given in Table 2. Both Mn and PDI of co-polyimidesfter grafting CDs decrease possibly due to chain scission duringsterification because of high reaction temperature and acid cata-yst.

In this study it was found that the heat treatment changesembrane color, flexibility, density and permeation characteris-

ics. The FTIR-ATR spectrum of 6FDA-Durene-DABA (imide group)an be characterized by bands around 1351 cm−1 (C–N stretch),716 cm−1 (assymmetric C O stretch), 1786 cm−1 (symmetric C Otretch) and 741 cm−1 (bending of C O). As shown in Fig. 5, in gen-ral the imide groups of these co-polyimides are thermally stablentil 400 ◦C and the intensities of these peaks below 400 ◦C are con-tant. This implies the decomposition of CD structure may not have

significant effect on the imide group until 400 ◦C. These results areimilar to the results obtained by Wieneke and Staudt [40], Shaot al. [41], and Xiao et al. [20]. Wieneke and Staudt [40] showed thathere was a strong relationship between the intensity of peaks inhe region of 3200–3400 cm−1 which represented the OH vibrationf the acid group and the amount of DABA in the 6FDA-copolyimidend these peaks decreased after treatment at 400 ◦C.

The intensities of peaks at 1580–1600 cm−1, representing C–Ctretching vibrations in the aromatic ring, increase when the mem-ranes are treated at 425 ◦C. This phenomenon may arise fromhe formation of new covalent bonds between benzene rings [20]ecause radical DABA groups are formed due to decarboxylation

◦

ver 400 C [36]. In other words, two radical DABA groups may forminkages as biphenyl [42]. For the membranes treated at 425 ◦C,he intensities of imide groups are reduced visibly because themide ring starts to decompose at this temperature. On the other

polyimide and PI-g-CDs.

hand, the intensities of peaks at 1580–1600 cm−1, representing C–Cstretching vibrations in the aromatic ring increase, indicating thestronger cross-linking reaction between aromatic rings. Comparedwith the original co-polyimide, the co-polyimide grafted by CDsshows a stronger intensity of the characteristic peaks of OH groupsat 1620 and 3650 cm−1, indicating the success of esterificationbetween the co-polyimide and CD. Heat treatment at 200 ◦C doesnot decompose the grafted CD because the characteristic peaksof OH groups at 1620 and 3650 cm−1 are still clear. By increas-ing temperature, CD decomposes and the intensities of these peaksdecrease.

The intensities of the characteristic peaks of OH groups at1620 cm−1, decrease by decreasing the number of OH groups of CD.As a result, PI-g-�-CD has a stronger intensity than that of PI-g-�-CDand PI-g-�-CD has a stronger intensity than that of PI-g-�-CD. Com-pared to the original PI, the co-polyimides grafted by CDs at 200 ◦Cshow a weaker intensity of the characteristic peaks of COOH groupsat 3200–3400 cm−1 because some COOH groups have reacted withCD during the esterification reaction.

The weight losses and derivatives of weight losses (decomposi-tion rates) of the original co-polyimide and PI-g-CDs as a function oftemperature in nitrogen atmosphere are shown in Fig. 6. The origi-nal co-polyimide shows two stages of decomposition. The first stageof weight loss of around 6% is detected from 400 to 500 ◦C, whichcomes from decarboxylation of DABA moiety of the co-polyimide[34]. In the second stage, additional 40% weight loss occurs from500 to 700 ◦C mainly because of the decomposition of imide and6FDA groups in the polymer backbone. Co-polyimides grafted CDshave one more stage of decomposition. In this stage, around 10%weight loss takes place in the range of 200–400 ◦C mainly due tothe decomposition of CDs’ structure [20,34]. The weight losses forthe second and third stages are around 4% and 35%, respectively.The weight losses of PI-g-CDs polymers due to the decarboxylationof DABA at 400–500 ◦C are smaller than that of the original PI. Thisis because the former has less COOH groups than the latter. The-oretically, if all COOH groups react with CD, the weight losses ofPI-g-�-CD, PI-g-�-CD, and PI-g-�-CD owing to CD decompositionat 200–400 ◦C should be around 14.5, 16.5, and 18.5%, respectively.However, the real weigh losses are around 9, 10, and 11.5% for PI-g-�-CD, PI-g-�-CD, and PI-g-�-CD, respectively because only around60% of COOH groups have reacted with CDs.

The XRD patterns of original co-polyimide and co-polyimidemembranes grafted by CDs in different treatment temperaturesare shown in Fig. 7. All of the wide angle X-ray diffraction (WAXD)

M. Askari et al. / Journal of Membrane Science 390– 391 (2012) 141– 151 147

Fa

papdiaob[miciatdCiottr

bHflvpbissaeirsssFC�C�

atures are shown in Table 4. As can be seen, the permeability ofall gases increases with heat treatment temperature. At low tem-peratures, hydrogen bonds among COOH groups of DABA tightenthe polymer segmental packing and reduce the free volume within

Table 3Physical properties of co-polyimide membranes in different heat treatmenttemperatures.

Tg (◦C) Density (g/cm3) V (cm3/g) FFV

Original PI-200 399 1.332 0.751 0.184Original PI-300 400 1.327 0.754 0.187Original PI-350 402 1.329 0.753 0.187Original PI-400 408 1.325 0.755 0.189Original PI-425 – 1.304 0.767 0.201PI-g-�-CD-200 400 1.325 0.755 0.203PI-g-�-CD-300 402 1.315 0.760 –PPM-�-CD-350 404 1.304 0.767 –PPM-�-CD-400 410 1.291 0.774 –PPM-�-CD-425 – 1.284 0.779 –PI-g-�-CD-200 401 1.327 0.754 0.205PI-g-�-CD-300 402 1.316 0.760 –PPM-�-CD-350 405 1.305 0.767 –PPM-�-CD-400 410 1.287 0.777 –PPM-�-CD-425 – 1.278 0.783 –PI-g-�-CD-200 401 1.324 0.755 0.208

ig. 7. WAXD patterns of co-polyimide membranes at different treatment temper-tures.

atterns of these co-polyimides are broad indicating that they haven amorphous structure. The WAXD peak in the amorphous glassyolymer is often used to estimate the average inter-chain spacingistance (d-spacing). As shown in this figure, the d-spacing value

ncreases with increasing post-treatment temperature up to 350 ◦Cnd a further heat treatment at 425 ◦C reduces the d-space. For theriginal co-polyimide, the initial increment in d-spacing is possi-ly due to an increase in chain mobility by increasing temperature41]. However, the decomposition of cyclodextrin structure, which

akes a lot of microvoids and channels between polyimide chains,s the drive mechanism for the initial increment in d-spacing ino-polyimides grafted by CDs [20]. The possible cause of decreas-ng d-spacing above 350 ◦C arises from the cross-linking reactiont this high temperature that leads to higher chain packing; hence,he distance between chains becomes smaller. Interestingly, theecreasing trend in d-spacing is severer in co-polyimides grafted byDs (decreasing from 6.40 to 5.84) than in the original PI (decreas-

ng from 6.30 to 6.10). As illustrated in Fig. 7, the d-spacing valuesf PPM-�-CD-425, PPM-�-CD-425 and PPM-�-CD-425 are lowerhan that of the original PI. This phenomenon may be attributed tohe factor that the decomposition of CD may promote cross-linkingeactions due to the presence of more free radicals.

Fig. 8 shows the evolution of color change of co-polyimide mem-rane grafted �-CD from colorless to black due to heat treatment.owever, the membrane thermally treated 425 ◦C still has very highexibility and mechanical strength that are much better than con-entional carbon molecular sieve membranes. Additional physicalroperties of the original PI and co-polyimide membranes graftedy CDs after heat treatment at different temperatures are provided

n Table 3. The values for the glass transition temperature (Tg), den-ity, and fractional free volume (FFV), which refers to the ratio of theo-called “expansion volume” to the observed volume are shown. Inll cases, Tg increases with increasing heat treatment temperaturespecially at 400 ◦C. Clearly, an increase in Tg means an increasen polymer chain rigidity due to occurrence of the cross-linkingeaction at higher temperatures. All co-polyimides at 425 ◦C do nothow a Tg by means of running DSC up to 450 ◦C, though there areome thermal events. Moreover, compared to �-CD and �-CD, den-ity of co-polyimide grafted �-CD decreases more. Also, the orderFV of these co-polyimides follows PI-g-�-CD > PI-g-�-CD > PI-g-�-D due to the size effect of various CD. Since �-CD is bigger than

-CD and �-CD is bigger than �-CD, the micro-pores created by �-D after decomposition are bigger than those created by �-CD and-CD. In case of density, it can be seen that it decreases by increasing

Fig. 8. Flexibility of the co-polyimide grafted with �-CD membrane thermallytreated at 425 ◦C.

heat treatment temperature and as a result, the FFV value increases.This decrease in density and increase in FFV, with the aid of heattreatment, is mainly attributed to the decomposition of CD struc-ture. According to TGA, these polymers lose CD at the temperaturerange from 300 to 400 ◦C and if polymer chains are adequatelyrigid, the spaces occupied by CD might be kept as micro-poresafter heat treatment [20]. Since it is still unclear that how muchCD has decomposed and been lost between 300 and 425 ◦C, cal-culating FFV for the co-polyimide grafted CDs in this temperaturerange is difficult.

3.2. Pure gas permeation experiments

The pure gas permeability and ideal selectivity of co-polyimidesbefore and after CD graft and then heat treated at different temper-

PI-g-�-CD-300 403 1.315 0.761 –PPM-�-CD-350 405 1.298 0.770 –PPM-�-CD-400 412 1.281 0.780 –PPM-�-CD-425 – 1.268 0.788 –

148 M. Askari et al. / Journal of Membrane Science 390– 391 (2012) 141– 151

Table 4Pure gas permeability and selectivity of co-polyimide membranes (10 atm and 35 ◦C).

Membranes Permeability (Barrer) Selectivity

O2 N2 CO2 CH4 C3H6 C3H8 O2/N2 CO2/CH4 CO2/N2 C3H6/C3H8

Original PI-200 53.86 13.40 235.28 10.42 – – 4.02 22.58 17.55 –Original PI-300 88.24 23.16 326.82 17.04 – – 3.81 19.18 14.11 –Original PI-350 112.63 31.24 392.86 22.80 11.06 0.81 3.61 17.23 12.57 13.65Original PI-400 144.63 38.01 519.68 28.85 30.13 2.09 3.81 18.01 13.67 14.42Original PI-425 212.59 53.51 1302.21 62.52 73.15 4.07 3.98 20.83 24.33 17.97PI-g-�-CD-200 57.73 14.52 241.23 11.98 – – 3.98 20.14 16.61 –PI-g-�-CD-300 112.25 31.58 373.61 22.25 – – 3.55 16.79 11.83 –PPM-�-CD-350 141.19 40.03 416.64 25.95 16.52 1.24 3.53 16.06 10.40 13.32PPM-�-CD-400 184.91 48.51 568.65 30.72 61.44 4.64 3.81 18.51 11.72 13.24PPM-�-CD-425 572.77 127.67 2423.03 111.67 295.67 16.64 4.49 21.70 18.97 17.77PI-g-�-CD-200 57.56 14.35 238.75 11.55 – – 4.01 20.67 16.63 –PI-g-�-CD-300 120.11 34.23 403.46 25.14 – – 3.51 16.05 11.78 –PPM-�-CD-350 165.70 47.65 593.25 37.29 17.47 1.24 3.48 15.91 12.27 14.09PPM-�-CD-400 243.22 62.91 772.33 41.42 81.36 5.86 3.87 18.65 12.45 13.88PPM-�-CD-425 754.35 166.1 3112.26 140.25 396.55 22.20 4.54 22.19 18.73 17.86PI-g-�-CD-200 61.04 15.46 251.38 11.85 – – 3.95 21.21 16.26 –PI-g-�-CD-300 125.22 34.89 416.38 23.85 – – 3.59 17.46 11.93 –

21102521

tpchmwotbplapwbmXfTstacr

PPM-�-CD-350 184.44 51.22 663.35 40.59

PPM-�-CD-400 307.75 77.79 929.97 48.43

PPM-�-CD-425 1024.35 231.23 4211.12 187.66

he polymer matrix [43]. Hydrogen bonding is broken and chainacking becomes loser with heat treatment temperature and thisan be a reason that the permeability of original PI increases witheat treatment temperature. Other reason may be due to the ther-al expansion of the polymer matrix. The chain mobility increasesith heat treatment temperature due to the thermal expansion

f the polymer matrix and permeability increases [41]. As men-ioned before, after heat treatment, the spaces occupied by CD cane kept as micro-pores that increase fractional free volume and gasermeability. The degree of increase in permeability generally fol-

ows the order of �-CD > �-CD > �-CD due to different CD structuresnd thermal decomposition behaviors. The selectivity of these co-olyimides decreases and then increase and, this trend is consistentith the evolution of d-spacing of the co-polyimides. As mentioned

efore, cross-linking reaction takes place at high temperatures andakes chain packing denser. According to Steel and Koros [44] andiao and Chung [20], ultramicropores and micro-pores are possibly

ormed in the polymeric membrane treated at high temperatures.he ultramicropores have dimensions equivalent or close to the d-pacing of the co-polyimide and determine gas selectivity, whereas

he micro-pores have bigger sizes created by the CD decompositionnd enhance gas permeability. Since the dimension of ultrami-ropores decreases at high temperatures because of cross-linkingeaction, the gas pair selectivity is increased. According to WAXD

Fig. 9. (a) Dimensions of �, �, and �-CD [45] and (b) permeability ratios of PP

.61 1.52 3.60 16.34 12.95 14.22

.43 7.32 3.96 19.28 11.95 13.98

.32 28.82 4.43 22.44 18.21 18.09

results, the degrees of cross-linking reaction are higher for co-polyimide membranes grafted by CDs than that for the original PI,thus the former has a higher selectivity than the later after thermaltreatment. One observation from Table 4 is that the permeabilityof the membranes heat treated at 400 and 425 ◦C are related tothe cavity diameter of the CDs. Fig. 9 shows the dimensions of CDs[45] and permeability ratios of PPM-�-CD to PPM-�-CD and PPM-�-CD to PPM-�-CD at 400 and 425 ◦C. As can be seen in this figure,the cavity diameters of �-CD, �-CD, and �-CD are around 5.7, 7.8and 9.5 A, respectively, and the cavity diameter ratios of �-CD to�-CD and �-CD to �-CD are around 0.6 and 0.8, respectively. Theseratios are almost the same as the permeability ratios of PPM-�-CDto PPM-�-CD and PPM-�-CD to PPM-�-CD at 400 and 425 ◦C. Thus,the micro-pores size is strongly related to the CD cavity size, whilethe permeability is strongly associated with the micro-pores size.

Permeability of the original PI thermally treated at 425 ◦C is 4–6times higher than that treated at 200 ◦C. This increase is 8–10 timesfor PPM-�-CD-425 and 15–17 times for PPM-�-CD-425. These rapidpermeability jumps are resulted from the fact that CD decomposesat high temperatures and also bigger CD creates bigger micro-

pores. Other observation is that all membranes heat treated at200 ◦C have almost the same permeability. This trend is due tothe fact that CD does not decompose in this low temperatureand all membranes may have almost the same size of micro-

M-�-CD to PPM-�-CD and PPM-�-CD to PPM-�-CD at 400 and 425 ◦C.

M. Askari et al. / Journal of Membrane Science 390– 391 (2012) 141– 151 149

c) PI-g

pt(ttfoiCil

tcpsgttabamctpah

3

oitf

Fig. 10. CO2 plasticization behavior of (a) original PI, (b) PI-g-�-CD, (

ores. The order of gas permeability for all membranes followshis sequence: C3H8 (kinetic diameter: 4.30 A) < CH4 (3.80 A) < C3H64.5 A) < N2 (3.64 A) < O2 (3.46 A) < CO2 (3.30 A). Except for C3H6 gas,he increasing order is consistent with decreasing kinetic diame-er of gas molecules that shows the separation mechanism mayollow the solution diffusion mechanism. The higher permeabilityf C3H6 over CH4 and C3H8 may be due to the effect of solubil-ty. In other words, C3H6 has a higher solubility than C3H8 andH4. In addition, CO2 has the highest permeability because of

ts high condensability, sorption capability and unique molecularinearity [41].

As aforementioned, plasticization is not desirable in gas separa-ion processes. Plasticization is a pressure dependent phenomenonaused by the dissolution of certain components within theolymer matrix which disrupts chain packing and enhances inter-egmental mobility [1–5]. The pressure at which the correspondingas permeability exhibits a minimum is known as the plasticiza-ion pressure. Fig. 10 illustrates the CO2 plasticization behavior ofhe newly developed co-polyimide membranes thermally treatedt different temperatures. As shown in this figure, all mem-ranes before heat treatment appear to undergo plasticizationt around 5 atm CO2 feed pressure and by increasing the ther-al treatment temperature, plasticization pressure increases. It’s

lear that cross-linking reaction at high-temperatures confineshe movement of polymer chains and enhances anti-plasticizationerformance. PPM-�-CD-425 and PPM-�-CD-425 show superiornti-plasticization characteristics up to 30 atm possibly becauseigher degrees of cross-link reaction.

.3. Mixed gas permeation experiments

Since the crucial applications of membranes are in the presence

f mixed gas streams, it is important to achieve a high selectiv-ty for mixed gas experiments to show commercial feasibility. Inhis study, mixed gas experiments were conducted using a 50:50eed mixture of CO2/CH4 with a total feed pressure of 20 atm

-�-CD, and (d) PI-g-�-CD at different heat treatment temperatures.

at 35 ◦C. Table 5 summarizes the gas permeability and selectiv-ity of co-polyimides thermally treated at 425 ◦C under pure andmixed gas tests. For the original PI-425, the mixed gas selectiv-ity decreases tremendously when comparing with pure gas resultsdue to the effect of plasticization. Clearly, the thermal treatmentat 425 ◦C does not enhance its anti-plasticization characteris-tics. Consistent with Fig. 10, the permeability of PPM-�-CD-425and PPM-�-CD-425 in mixed gas tests shows slightly decreases,while their selectivity is almost the same as those in pure gastests.

The differences in pure gas and mixed gas systems havebeen studied by Chern et al. [46]. Assuming the permeabilityof a pure gas in glassy polymers can be described by the dualmode or partial immobilization model, the permeability of gas Amay be expressed by Eq. (2) when the downstream pressure iszero.

P = KDA · DDA

(1 + FAKA

1 + bApA

)(2)

where KDA, DDA, bA and pA refer to the Henry’s law constant, Henry’slaw diffusion coefficient of the penetrant, Langmuir affinity con-stant, and upstream pressure of pure gas, respectively. FA is the ratioof the diffusion coefficient of the penetrant in the Langmuir sites(i.e., microvoids) to that in the Henry sites for gas A (FA = DHA/DDA),and K = CHA

′ bA/k, is a useful collection of the sorption parame-ters introduced above when CHA

′ refers to the Langmuir capacityconstant for pure gas A.

The permeability of gas A in mixed gas tests is given by Eq. (3).For simplicity, gas A is to the component of primary interest, whilegas B is the second competing component. PA and PB refer to thepartial pressures of components A and B, respectively.

P = KDA · DDA

(1 + FAKA

)(3)

1 + bApA + bBpB

As can be seen, Eq. (3) has one extra term in the denominatorcompared to Eq. (2), which decreases the permeability amount forgas A. Therefore, gas permeability obtained under mixed gas test is

150 M. Askari et al. / Journal of Membrane Science 390– 391 (2012) 141– 151

Table 5Mixed gas permeability and selectivity of co-polyimide membranes treated in 425 ◦C.

Pure gasa Mixed gasb

Permeability (Barrer) Selectivity Permeabilbility (Barrer) Selectivity

CO2 CH4 CO2/CH4 CO2 CH4 CO2/CH4

Original PI-425 1302 62 20.83 1221 91 13.41PPM-�-CD-425 2423 112 21.70 2392 117 20.44PPM-�-CD-425 3112 140 22.19 2911 133 21.85PPM-�-CD-425 4211 187 22.44 3976 174 22.85

a Measured at 10 atm and 35 ◦C.b Measured at 20 atm and 35 ◦C.

pty s

umo

aatPsifF4p4

4

wbrotm

(

(

(

Fig. 11. Trade off lines for CO2/CH4 and C3H6/C3H8 separation (em

sually lower than that under pure gas test. However, selectivityay stay almost constant but still somewhat depends on many

ther factors.A comparison of gas separation performance of the original PI

nd co-polyimide membranes grafted by CDs thermally treatedt 425 ◦C is shown in Fig. 11. The separation performance of thehermally treated original PI is much inferior to those CD graftedI. For CO2/CH4 separation, the entire PI grafted CDs membranesurpass the trade-off line, while the thermally treated original PIs still within the trade-off line. Clearly, the CD decompositionacilitates micro-pore formation and enhances the permeability.or C3H6/C3H8 separation, all membranes thermally treated at25 ◦C can surpass the trade-off line. They have almost the sameermselectivity, but permeability follows the order of PPM-�-CD-25 > PPM-�-CD-425 > PPM-�-CD-425 > Original PI-425.

. Conclusion

Using a cross-linkable co-polyimide (6FDA-Durene/DABA (9/1)),e have developed new high-performance gas separation mem-

ranes that can enhance both membrane permeability andesistance to plasticization simultaneously by grafting various sizesf cyclodextrin to the polyimide matrix and then decomposinghem at elevated temperatures. The following conclusions can be

ade from this work:

1) The decomposition of CD may convert the spaces originallyoccupied by CD to micro-pores and thus increase fractional freevolume and gas permeability.

2) Permeability increase is strongly related to the cavity size of CD;the bigger CD size, the higher permeability jumps after thermal

treatments at elevated temperatures.

3) Permselectivity increases if the thermal treatment is conductedover 400 ◦C due to the occurrence of cross-linking reaction inthe polyimide matrix that tightens the d-space.

ymbols are pure gas results, solid symbols are mixed gas results).

(4) The separation performance of thermally treated polyimidesthat were originally grafted with CD is much better than thatwithout CD. For CO2/CH4 separation, all the co-polyimide mem-branes grafted by CDs surpass the trade-off line, while thethermally treated original polyimide cannot.

(5) For C3H6/C3H8 separation, all membranes thermally treatedat 425 ◦C can surpass the trade-off line with almost the samepermselectivity. However, their permeability follows the orderof PPM-�-CD-425 > PPM-�-CD-425 > PPM-�-CD-425 > OriginalPI-425.

(6) The thermally treated polyimides that were originally graftedwith CD show much better anti-plasticization characteris-tics than that without CD. The plasticization pressure of thePPM-�-CD-425 membrane is over 30 atm, while the originalpolyimide is less than 15 atm.

(7) The best result was achieved from the PPM-�-CD-425 mem-brane. It has a CO2 permeability of 4211 Barrers with a CO2/CH4ideal selectivity of 22.44 and a C3H6 permeability of 521 Barrerswith a C3H6/C3H8 ideal selectivity of 18.09. The CO2 permeabil-ity drops slightly to 3976 Barrers with almost the same CO2/CH4selectivity of 22.84 in mixed gas tests.

(8) The newly developed PPM-�-CD-425, PPM-�-CD-425, andPPM-�-CD-425 membranes are highly flexible with goodmechanical strength.

Acknowledgements

The authors would like to thank the Singapore National ResearchFoundation (NRF) for the financial support on the CompetitiveResearch Program with the project entitled, “Molecular Engineer-

ing of Membrane Materials: Research and Technology for EnergyDevelopment of Hydrogen, Natural Gas and Syngas” (grant num-ber: R-279-000-261-281). The authors also thank Ms. M.L. Chua,Ms. H. Wang and Mr. F.Y. Li for giving assistance to this work.

M. Askari et al. / Journal of Membrane S

Nomenclature

A effective area of membrane (cm2)bA affinity constant of component A for the polymer

(atm−1)CHA

′ Langmuir capacity coefficient for component “A”((cm3 gas (STP))/(cm3 polymer))

d dimension spacing ( ´A)DDA Henry’s law diffusion coefficient of the penetrant

“A” (cm2/s)DHA diffusion coefficient of the penetrant “A” in the

microvoid environment (cm2/s)FFV fractional free volumeKDA Henry’s law constant for component “A” ((cm3 gas

(STP))/(cm3 polymer-atm))L membrane thickness (cm)Mn number average molecular weightpA partial pressure of gas “A” (atm)pB partial pressure of gas “B” (atm)P gas permeability (Barrer)PA permeability of gas “A” (Barrer)PB permeability of gas “B” (Barrer)PDI polydispersityP2 upstream operating pressure (psia)T temperature (K)v specific volume (cm3/g)v0 occupied volume (cm3/g)vw van der Waals volume (cm3/g)V volume of downstream chamber (cm3)˛A/B ideal selectivity� diffraction angle (◦)

R

[[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[[

[

[

[

[

[

[

� X-ray wavelength ( ´A)

eferences

[1] F.G. Kerry, Industrial Gas Handbook: Gas Separation and Purification, Taylorand Franc Group, New York, 2006.

[2] B.D. Bhide, A. Voskericyan, S.A. Stern, Hybrid processes for the removal of acidgases from natural gas, J. Membr. Sci. 140 (1998) 27–49.

[3] R.W. Baker, Reviews future directions of membrane gas separation technology,Ind. Eng. Chem. Res. 41 (2002) 1393–1411.

[4] Y.C. Xiao, B.T. Low, S.S. Hosseini, T.S. Chung, D.R. Paul, The strategies ofmolecular architecture and modification of polyimide based membranes forCO2 removal from natural gas – a review, Prog. Polym. Sci. 34 (2009) 561–580.

[5] R.L. Burns, W.J. Koros, Defining the challenges for C3H6/C3H8 separation usingpolymeric membranes, J. Membr. Sci. 211 (2003) 299–309.

[6] M.L. Chng, Y.C. Xiao, T.S. Chung, M. Toriida, S. Tamai, Enhanced propy-lene/propane separation by carbonaceous membrane derived from poly (arylether ketone)/2,6-bis(4-azidobenzylidene)-4-methyl-cyclohexanone inter-penetrating network, Carbon 47 (2009) 1857–1866.

[7] R.B. Eldridge, Olefin paraffin separation technology: a review, Ind. Eng. Chem.Res. 32 (1993) 2208–2212.

[8] S.U. Rege, J. Padin, R.T. Yang, Olefin/paraffin separations by adsorption: pi-complexation vs. kinetic separation, AIChE J. 44 (1998) 799–809.

[9] M. Azhin, T. Kaghazchi, M. Rahmani, A review on olefin/paraffin separationusing reversible chemical complexation technology, J. Ind. Eng. Chem. 14 (2008)622–638.

10] L.M. Robson, Upper-bound revisited, J. Membr. Sci. 320 (2008) 390–400.11] S.S. Hosseini, Y. Li, T.S. Chung, Y. Liu, Enhanced gas separation performance of

nanocomposite membranes using MgO nanoparticles, J. Membr. Sci. 302 (2007)207–217.

12] P.M. Budd, N.B. Mckeown, Highly permeable polymers for gas separation mem-branes, Polym. Chem. 1 (2010) 63–68.

13] K. Tanaka, A. Taguchi, J. Hao, H. Kita, K. Okamoto, Permeation and separationproperties of polyimide membranes to olefins and paraffins, J. Membr. Sci. 121(1996) 197–207.

14] S. Sridhar, A.A. Khan, Simulation studies for the separation of propylene andpropane by ethylcellulose membrane, J. Membr. Sci. 159 (1999) 209–219.

[

[

cience 390– 391 (2012) 141– 151 151

15] M. Yoshino, S. Nakamura, H. Kita, K. Okamoto, N. Tanihara, Y. Kusuki,Olefin–paraffin separation performance of asymmetric hollow fiber membraneof 6FDA, BPDA–DDBT co-polyimide, J. Membr. Sci. 212 (2003) 13–27.

16] C. Staudt-Bickel, W.J. Koros, Olefin–paraffin gas separations with 6FDA-basedpolyimide membranes, J. Membr. Sci. 170 (2000) 205–214.

17] S.S. Chan, R. Wang, T.S. Chung, Y. Liu, C2 and C3 hydrocarbon separationsin poly(1,5-naphthalene-2,2′-bis(3,4-phthalic) hexafluoropropane) diimide(6FDA-1,5-NDA) dense membranes, J. Membr. Sci. 210 (2002) 55–64.

18] S. Basu, A.C. Odena, F.J. Vankelecom, Asymmetric Matrimid/[Cu3 (BTC) 2] mixedmatrix membranes for gas separations, J. Membr. Sci. 362 (2010) 478–487.

19] M. Das, W.J. Koros, Performance of 6FDA–6FpDA polyimide for propy-lene/propane separation, J. Membr. Sci. 365 (2010) 399–408.

20] Y.C. Xiao, T.S. Chung, Grafting thermally labile molecules on cross-linkablepolyimide to design membrane materials for natural gas purification and CO2

capture, Energy Environ. Sci. 4 (2011) 201–208.21] Y.C. Xiao, T.S. Chung, H.M. Guan, M.D. Guiver, Synthesis, crosslinking and

carbonization of co-polyimides containing internal acetylene units for gas sep-aration, J. Membr. Sci. 302 (2007) 254–264.

22] S. Neyertz, D. Brown, S. Pandian, N.F.A. van der Vegt, Carbon dioxide diffu-sion and plasticization in fluorinated polyimides, Macromolecules 43 (2010)7813–7827.

23] Y.C. Xiao, T.S. Chung, M.L. Chng, Surface characterization, modification chem-istry and separation performance of polyimide and PAMAM dendrimercomposites, Langmuir 20 (2004) 8230–8238.

24] T. Visser, M. Wessling, Auto and mutual plasticization in single and mixed gasC3 transport through Matrimid-based hollow fiber membranes, J. Membr. Sci.312 (2008) 84–96.

25] W.J. Koros, M.W. Hellums, Gas separation membrane material selection cri-teria: differences for weakly and strongly interacting feed components, FluidPhase Equilib. 53 (1989) 339–354.

26] A. Bos, I.G.M. Punt, M. Wessling, H. Strathmann, Plasticization-resistant glassypolyimide membranes for CO2/CO4 separations, Sep. Purif. Technol. 14 (1998)27–39.

27] L.S. White, T.A. Blinka, H.A. Kloczewski, I.F. Wang, Properties of a polyimide gasseparation membrane in natural gas streams, J. Membr. Sci. 103 (1995) 73–82.

28] H. Kita, T. Inada, K. Tanaka, K. Okamoto, Effect of photocrosslinking on per-meability and permoselectivity of gases through benzophenone-containingpolyimide, J. Membr. Sci. 87 (1994) 139–147.

29] C. Staudt-Bickel, W.J. Koros, Improvement of CO2/CH4 separation characteris-tics of polyimides by chemical crosslinking, J. Membr. Sci. 155 (1999) 145–154.

30] L. Shao, T.S. Chung, S.H. Goh, K.P. Pramoda, Transport properties of cross-linkedpolyimide membranes induced by different generations of diaminobutane(DAB) dendrimers, J. Membr. Sci. 238 (2004) 153–163.

31] C.E. Powell, X.J. Duthie, S.E. Kentish, G.G. Qiao, G.W. Stevens, Reversible diaminecrosslinking of polyimide membranes, J. Membr. Sci. 291 (2007) 199–209.

32] B. Cruden, K. Chu, K. Gleason, H. Sawin, Thermal decomposition of low dielectricconstant pulsed plasma fluorocarbon films. II. Effect of postdeposition anneal-ing and ambients, J. Electrochem. Soc. 146 (1999) 4597–4604.

33] B. Lee, Y.H. Park, Y.T. Hwang, O.H. Weontae, J. Yoon, M. Ree, Ultralow-knanoporous organosilicate dielectric films imprinted with dendritic spheres,Nat. Mater. 4 (2005) 147–151.

34] L.X. Song, P. Xu, A comparative study on the thermal decomposition behaviorsbetween �-cyclodextrin and its inclusion complexes of organic amines, J. Phys.Chem. A 112 (2008) 11341–11348.

35] Y. Wang, T.S. Chung, H. Wang, S.H. Goh, Butanol isomer separation usingpolyamide–imide/CD mixed matrix membranes via pervaporation, Chem. Eng.Sci. 64 (2009) 5198–5209.

36] A.M. Kratochvil, W.J. Koros, Decarboxylation-induced crosslinking of a poly-imide for enhanced CO2 plasticization resistance, Macromolecules 41 (2008)7920–7927.

37] C.E. Sroog, Polyimides, Prog. Polym. Sci. 16 (1991) 561–694.38] Y.C. Xiao, T.S. Chung, M.L. Chng, S. Tamai, A. Yamaguchi, Structure and

properties relationships for aromatic polyimides and their derived carbonmembranes: experimental and simulation approaches, J. Phys. Chem. B 109(2005) 18741.

39] A. Bondi, van der Waals volumes and radii, J. Phys. Chem. 68 (1964)441–451.

40] J.U. Wieneke, C. Staudt, Thermal stability of 6FDA-(co-)polyimides containingcarboxylic acid groups, Polym. Degrad. Stabil. 95 (2010) 684–693.

41] L. Shao, T.S. Chung, K.P. Pramoda, The evolution of physicochemical and trans-port properties of 6FDA-durene toward carbon membranes; from polymer,intermediate to carbon, Micropor. Mesopor. Mater. 84 (2005) 59–68.

42] T.P. Eskay, P.F. Britt, A.C. Buchanan, Does decarboxylation lead to crosslinkingin low-rank coals? Energy Fuels 10 (1996) 1257–1261.

43] W. Qiu, C.C. Chen, L. Xu, L. Cui, D.R. Paul, W.J. Koros, Sub-Tg cross-linking ofa polyimide membrane for enhanced CO2 plasticization resistance for naturalgas separation, Macromolecules 44 (2011) 6045–6056.

44] K.M. Steel, W.J. Koros, Investigation of porosity of carbon materials and related

effects on gas separation properties, Carbon 41 (2003) 253–266.

45] http://www.chromatography-online.org/Chrial-GC/Contemporary-Chiral-Stationary-Phases/Cyclodextrin.html.

46] R.T. Chern, W.J. Koros, E.S. Sanders, R. Yui, Second component effects in sorptionand permeation of gases in glassy polymers, J. Membr. Sci. 15 (1983) 157–169.