Preparation and gas separation performance of flexible pyrolytic membranes by low-temperature pyrolysis of sulfonated polyimides

Journal of Membrane Science 261 (2005) 17–26

Preparation and gas separation performance of flexible pyrolyticmembranes by low-temperature pyrolysis of sulfonated polyimides

Md. Nurul Islam, Weiliang Zhou, Tatsuaki Honda, Kazuhiro Tanaka,Hidetoshi Kita, Ken-ichi Okamoto∗

Department of Advanced Materials Science and Engineering, Faculty of Engineering, Yamaguchi University, Ube, Yamaguchi 755-8611, Japan

Received 23 June 2004; received in revised form 11 February 2005; accepted 14 February 2005

Abstract

Dense flat membranes of sulfonated polyimides (SPIs) were prepared from 1,4,5,8-naphthalene tetracarboxylic dianhydride (NTDA)and aromatic diamines of 2,2-bis[4-(4-aminophenoxy)phenyl]hexafluoropropane disulfonic acid, benzidine-2,2′-disulfonic acid, 4,4′-diaminodiphenylether-3,3′-disulfonic acid and 9,9′-bis(4-aminophenyl)fluorene. The SPI membranes were pyrolyzed at a relatively low-t ◦ f thep n polymerm eability fort e polymerc s,w ffusivity togt©

K

1

mclhepiasco

riersedtorspre-aneo

d byd to

ofm-s of

y-us

0d

emperature of 450C for 1.5 h at a nitrogen flow. During the pyrolysis, theSO3H groups decomposed without substantial cleavage oolyimide backbone. The pyrolytic membranes had the interesting intermixed properties of toughness and good flexibility as iembranes and high gas permeability with reasonably high selectivity as in carbon molecular sieve membranes. The gas perm

he pyrolytic membranes seems to be controlled by the sulfonic acid group content in the precursor polyimides, the relaxation of thhains during the pyrolysis and the gas permeation level of the precursor membranes. Decomposition ofSO3H groups induced microvoidhich are considered to remain to some extent as larger-size free volume holes in the polymer matrix, resulting in the higher dias with larger diameter. The pyrolytic membranes showed the higher permeability of C3H6 and/or the higher selectivity of C3H6/C3H8 than

he 6FDA-polyimide membranes which have defined the upper bound line.2005 Elsevier B.V. All rights reserved.

eywords: Sulfonated polyimides; Template effect; Pyrolytic membrane; Gas permeation; Propylene/propane separation

. Introduction

Separation of olefin and paraffin gases is one of theost important processes in petrochemical industries and

urrently performed by low-temperature distillation witharge energy consumption[1]. Membrane-based technologiesave advantages of both low capital cost and high-energyfficiency compared to conventional separation methods inrinciple. Facilitated transport membranes containing silver

ons as a complex agent have very high olefin/paraffin sep-ration selectivity and high olefin fluxes[2–4]. However,ome problems, especially poor chemical stability due to thearrier poisoning, currently limit the practical applicationsf the facilitated transport membranes. The olefin/paraffin

∗ Corresponding author. Tel.: +81 836 85 9660; fax: +81 836 85 9601.E-mail address:[email protected] (K. Okamoto).

separation through polymeric membranes without carhas also been studied[5–13]. There have been reportsome polyimides having relatively high separation facof propylene over propane. Most of the polyimides werepared from 2,2-bis(3,4-dicarboxyphenyl) hexafluoropropdianhydride (6FDA)[7–13]. However, their permeability tpropylene is not high enough for practical uses.

Carbon molecular sieve (CMS) membranes preparepyrolysis of polymeric membranes have been reportedisplay high olefin/paraffin separation performance[14–17].Hayashi et al. reported the excellent performanceC2H4/C2H6 and C3H6/C3H8 separation for carbonized mebranes prepared by pyrolyzing composite membranepolyimide from 3,3′,4,4′-biphenyltetracarboxylic dianhdride (BPDA) and oxydianiline (ODA) coated on poroalumina tubes at temperature of 600–900◦C in a nitro-gen stream. The permeance to propyleneRC3H6 was 9 GPU

376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserved.oi:10.1016/j.memsci.2005.02.019

18 Md.N. Islam et al. / Journal of Membrane Science 261 (2005) 17–26

(1 GPU = 10−6 cm3 (STP)/(cm2 s cmHg)) and separation fac-tor of propylene over propaneα(C3H6/C3H8) was 33 at100◦C for the membrane pyrolyzed at 700◦C[14]. Okamotoet al. prepared carbonized asymmetric hollow fiber mem-branes by pyrolyzing the precursor asymmetric hollow fiberof BPDA-based polyimide at temperature of 500–700◦Cunder a nitrogen stream. The membranes showed fairlyhigh C3H6/C3H8 separation performance such asRC3H6 of32 GPU andα(C3H6/C3H8) of 15 in a mixed gas at 100◦Cand 1 atm[17].

Although CMS membranes display better gas separationselectivity and better thermal stability than polymeric mem-branes, the poor operation property as the result of theirinherent brittleness restricts their practical application. Com-posite CMS membranes composed of a carbonized layer onan inorganic support have an advantage of excellent mechan-ical strength but have a disadvantage of rather complicatedpreparation procedure. The self-standing CMS membranesderived from capillary or asymmetric hollow fiber mem-branes have advantages of simple preparation procedure andlarge effective membrane area per volume, but in general havepoor mechanical strength, especially poor resistance againstshock[18,19].

Okamoto and coworkers reported highly permeable CMSmembranes prepared by pyrolyzing a thermosetting phenolicr ◦T wedt rredb merm -c have“ es.H ndc m eta es ofsap -p em-b .

ndfl tedp es e didn tion,c roper-t d ind

2

2

ride( ro-

propane (BAHF), oxydianiline (ODA), BDSA and 9,9′-bis(4-aminophenyl)fluorene (BAPF) were purchasedfrom Tokyo Kasei Co. NTDA and BAPF were purifiedby vacuum sublimation before use. 2,2-Bis[4-(4-aminopheoxy)phenyl]hexafluoropropane disulfonic acid(BAHFDS) and 4,4′-diaminodiphenylether-3,3′-disulfonicacid (ODADS) were prepared by direct sulfonation ofBAHF and ODA, respectively, as previously reported[24].m-Cresol was used as received.

2.2. Preparation of sulfonated polyimides

The chemical structures of homo and copolyimides usedas precursors in this study are shown inFig. 1. NTDA-BAHFDS/BAPF(4/1) copolyimide was prepared by a one-step method as follows[24]. To a 100 ml of completely driedfour-necked flask with stirrer was added 1.63 g (2.4 mmol)BAHFDS, 0.7 ml (5 mmol) triethylamine, 0.21 g (0.6 mmol)BAPF and 11 mlm-cresol under N2 flow. After BAHFDSwas completely dissolved, 0.80 g (3 mmol) NTDA and 0.54 g(4.5 mmol) benzoic acid were added into this flask. Themixture was heated at 80◦C for 4 h and 185◦C for 20 hin order to carry out condensation polymerization and ther-mal immidization. After cooling to room temperature, addi-tional 40 ml of m-cresol was added to dilute the highlyv ides itate

Fig. 1. Chemical structures of NTDA-based and 6FDA-based polyimides.

esin with a pendant sulfonic acid group at 500C [20,21].hermogravimetry–mass spectroscopy (TG–MS) sho

hat the decomposition of sulfonic acid groups occuut no vigorous decomposition of the skeleton of polyatrix occurred below 450◦C under a N2 stream. This indi

ates that sulfonic acid groups bonded to phenolic resintemplate-like effect” in preparation of CMS membranowever, the phenolic resin had poor film-forming ability aould be used only in form of composite membranes. Kil. prepared CMS membranes by pyrolyzing membranulfonated polyimides (SPIs) from benzidine-2,2′-disulfoniccid (BDSA) with or without metal cations[22]. Theyyrolyzed the SPI membranes at 590◦C, where final decomosition of the precursors occurred. The resulting CMS mranes displayed rather poor gas permeation properties

In a previous paper[23], we briefly reported dense aexible pyrolytic flat membranes by pyrolyzing sulfonaolyimides at a low temperature of 450◦C, where most of thulfonic groups decomposed but the polyimide backbonot substantially decompose. In this paper, the preparaharacterization and gas permeation and separation pies of the flexible pyrolytic flat membranes are reporteetail.

. Experimental

.1. Materials

1,4,5,8-Naphthalene tetracarboxylic dianhydNTDA), 2,2-bis[4-(4-aminophenoxy)phenyl]hexafluo

iscous solution, and the resulting dark-brown polyimolution was poured into acetone. The fiber-like precip

Md.N. Islam et al. / Journal of Membrane Science 261 (2005) 17–26 19

was collected by filtration, washed with acetone and dried invacuo.

The other NTDA-based polyimides, namely, NTDA-BAHFDS, NTDA-BDSA and NTDA-ODADS homopoly-imides and NTDA-BDSA/BAPF(4/1) copolyimide, wereprepared by the similar procedure. For NTDA-BAHFDS/BAPF(4/1) and NTDA-BDSA/BAPF(4/1) copolyimides, themolar ratio of the sulfonated/non-sulfonated was 4/1.

2.3. Membrane preparation

The polyimides in triethylammonium salt form were dis-solved inm-cresol to prepare 2–5 wt.% polyimide solutions.The solutions were cast onto Petri dish and dried at 120◦Cfor 10 h. The as-cast membranes were soaked in methanol at60◦C for 1 h and in 1.0N hydrochloric acid at room tempera-ture for 24 h successively. The proton-exchanged membraneswere washed with de-ionized water, and then dried in vac-uum at 150◦C for 10 h. The membrane thickness was in therange of 25–45�m. The proton-exchanged membranes werepyrolyzed at 370◦C for 1.5 h or 450◦C for 1.5 h at a heatingrate of 5◦C/min and a nitrogen flow rate of 100 ml/min.

2.4. Measurements

Fourier transform infrared spectrometry (FTIR) was car-r outo MS-Q :1 asc e of1

entsw suret df en 5a was4 usinga lec-tc ilib-r db

D

Io t ofr nds

α

M werec ato-g r

α(A/B) was calculated by the following equation:

α

(A

B

)= yA/yB

xA/xB(3)

wherexi andyi are mole fractions of componenti in feedand permeate gases, respectively. These measurements werecarried out at 35 or 50◦C and up to 10 atm.

3. Results and discussion

3.1. Characterization

Fig. 2 shows the TG curve and the simultaneous massspectra of evolved gases of NTDA-BDSA/BAPF(4/1) mem-brane during the pyrolysis. Two stages of decomposition wereobserved. In the first stage, the 15% weight loss occurring inthe range of 250–450◦C was mainly due to the evolutionof SO2 and H2O. Then a short plateau in the TG curve wasobserved around 450◦C. This temperature corresponded tothe valleys of the mass spectra of SO2 and H2O. On the otherhand, the very small amounts of CO2 and CO were observedbelow 450◦C. In the second stage, further 25% weight lossoccurred in the range of around 500–700◦C accompanied bysignificant evolution of CO2, CO and H2O. The other SPIss TIRa firsts dt conds ideb

het yzeda te:

F sis ofN

ied out on a Jasco FT/IR-610. TG–MS was carriedn a Rigaku TG-8120 connected to a Shimadzu GCP 5050 at a heating rate of 5◦C/min in helium (flow rate00 cm3/min). Differential scanning calorimetry (DSC) warried out on a Rigaku DSC-8230 at a heating rat0◦C/min.

Pure (single-component) gas permeation experimere carried out using a vacuum time-lag method to mea

he gas permeability coefficientP, which was determinerom a steady-state permeation flux in a period betwend 10 times the time lag. The effective membrane area.2 cm2. Pure gas sorption measurement was carried outsorption cell equipped with a Sartorius S3D-P model e

ronic microbalance. Equilibrium solubility coefficientSwasalculated as the ratio of the sorption amount to the equium pressure. Average diffusion coefficientD was evaluatey the following equation:

= P

S(1)

deal separation factor,αid(A/B), was calculated as theP ratiof pure gases A and B, and is expressed as the producDatio andSratio of pure gases, i.e. diffusivity selectivity aolubility selectivity:

id

(A

B

)= PA

PB=

(DA

DB

) (SA

SB

)(2)

ixed (binary-component) gas permeation experimentsarried out by a vacuum method followed by gas chromraphic analysis of the permeate gases[7]. Separation facto

howed the similar decomposition behavior. Taking the Fnalysis result to be mentioned below into account, thetage decomposition occurring below 450◦C was attributeo the decomposition of sulfonic acid groups and the setage one was attributed to the decomposition of polyimackbone.

Thin films of about 5�m in thickness were used for transmission FTIR measurements. The films were pyrolt 450◦C for 1.5 h under a nitrogen flow (heating ra

ig. 2. TG curve and mass spectra of evolved gases during pyrolyTDA-BDSA/BAPF(4/1) polyimide.


Fig. 3. FTIR spectra of precursor and pyrolytic (450◦C, 1.5 h) membranesof NTDA-BDSA polyimide.

5◦C/min). The FTIR spectra of NTDA-BDSA precursor andits pyrolyzed films are shown inFig. 3. The strong absorp-tion bands at 1717 and 1671 cm−1 are assigned to the stretchvibration of carbonyl groups of imide rings. The peaks at1099 and 1031 cm−1 assigned to SO stretch of sulfonic acidgroups, which were observed for NTDA-BDSA precursor,disappeared in the spectrum of the pyrolytic film, indicat-ing the decomposition of sulfonic acid groups. Most of theother peaks hardly changed after the pyrolysis at 450◦C for1.5 h, indicating that the substantial cleavage of the polyimidebackbone did not yet occur.

Non-sulfonated polyimide, NTDA-BAHF showed a glasstransition temperature of about 280◦C, which was higherthan that (244◦C) of BPDA-BAHF from 3,3′,4,4′-biphenyl-tetracarboxylic acid dianhydride (BPDA)[25]. The precur-sor SPIs did not show any DSC signal assigned to theglass transition up to 300◦C. Above 300–330◦C, they dis-played a small and broad endothermic DSC signal, of whichthe peak appeared around 330–350◦C. The similar type ofendothermic DSC signal was also observed in the tempera-ture range of 260–350◦C for sulfopropoxylated polyimidessuch as NTDA-2,2′-BSPB derived from 2,2′-bis(3-sulfo-propoxy)benzidine (2,2′-BSPB). It is noted that the onsettemperature of the signal (300 or 260◦C) corresponds tothe onset decomposition temperature of sulfonic acid groupsor sulfopropoxy groups (300 or 250◦C, respectively). Afterh sol-

uble in any solvent. Therefore, the signal for the present SPIswas attributed to the decomposition of sulfonic acid groupsfollowed by some kind of crosslinking reaction.

Table 1lists the SO3H group contents of the precursorSPIs and the values of weight loss (WL(450◦C, TG)) in therange of 250–450◦C determined from their TG curves andthe values of weight loss (WL(450◦C, 1.5 h)) of membranespyrolyzed at 450◦C for 1.5 h. The WL(450◦C, TG) valueswere smaller than theSO3H contents by 11–30%, whereasthe WL(450◦C, 1.5 h) values were larger than theSO3Hcontents by 36–60%. These results mentioned above indicatethat the pyrolysis at 450◦C for 1.5 h made the decompositionof SO3H groups more complete followed by slight furtherdecomposition but did not cause the substantial decomposi-tion of the polyimide backbone.

The membranes (25–45�m in thickness) pyrolyzed at450◦C for 1.5 h were flexible and tough enough for thegas permeation measurements, whereas the membranespyrolyzed above 500◦C were brittle and their gas perme-ation measurement could not be carried out. The flexibilityof a membrane was evaluated by means of a bending test. Aslip of a sample (5 mm in width and 15 mm in length) wasbended in U-shape by holding both ends (3 mm in length)of the slip between fingers. When the slip was bent with-out break, the membrane was evaluated to be flexible. Allthe pyrolytic (450◦C, 1.5 h) membranes listed inTable 1w lyticm n thec is ath

3

/1)pT heg re ctiv-i lysiswml fonica aftert ered

TS brane orm

S G)ra

NNNNN )

tic/pre

eating up to 350◦C, both types of membranes became in

able 1ulfonic acid group content and weight loss (WL) of precursor memembranes

PIs SO3H content(wt.%)

WL (450◦C, T(wt.%)

TDA-BAHFDS 18 16TDA-BAHFDS/BAPF(4/1) 15 –TDA-BDSA/BAPF(4/1) 23 16TDA-BDSA 28 22TDA-ODADS 27 –a At 35◦C, 1 atm. The figure in parenthesis refers to theP ratio for pyroly

ere evaluated to be flexible. Thus, the present pyroembranes had much better mechanical properties tha

onventional CMS membranes prepared by the pyrolysigher temperatures.

.2. Permeation and separation property

Pure gas permeation results of NTDA-BDSA/BAPF(4recursor and its pyrolytic membranes are listed inTable 2.he pyrolysis at 370◦C for 1.5 h significantly enhanced tas permeability. The pyrolysis at 450◦C for 1.5 h furthenhanced the gas permeability without loss of the sele

ty. These increases in gas permeability with the pyroere attributed to the decomposition ofSO3H groups. Asentioned above, the membrane pyrolyzed at 450◦C for 1.5 h

ost the sulfonic acid groups. The spaces occupied by sulcid groups might be kept as microholes or microvoids

he pyrolysis. Such microvoids are reasonably consid

s and CO2 and O2 gas permeability for pyrolytic (450◦C, 1.5 h) and precurs

WL (450◦C, 1.5 h)(wt.%)

PCO2 (Barrer),pyrolytic/precursora

PO2 (Barrer),pyrolytic/precurso

26 920/13 (71) 121/3.0 (40)24 400/36 (11) 69/7.0 (10)34 380/2.0 (190) 54/0.43 (126)38 158/0.5 (320) 24/0.1 (240)37 52/0.2 (260) 8.1/0.03 (270

cursor membranes.


Table 2Gas permeability and permeability ratio of precursor and pyrolytic membranes of NTDA-BDSA/BAPF(4/1) polyimide at 35◦C and 1 atm

Membrane PH2 PCO2 PO2 PC2H4 PC3H6 PO2/PN2 PCO2/PN2 PC2H4/PC2H6

Precursor 14.4 2.0 0.43 – – 8.6 40 –370◦C, 1.5 h 165 145 23 7.9 – 4.9 31 4.3450◦C, 1.5 h 310 380 54 31 15 4.1 29 4.2

to contribute to the high gas permeability for the pyrolyticmembranes. The ideal separation factors hardly changed aspyrolysis temperature was elevated from 370 to 450◦C. Weselected the pyrolysis condition of 450◦C for 1.5 h to preparepyrolytic membranes from various SPIs. The membranespyrolyzed at 450◦C for 1.5 h were used as pyrolytic oneshere after unless otherwise noted.

Figs. 4 and 5show pure gas permeability for the precur-sor and pyrolytic membranes, respectively, of NTDA-basedpolyimides as a function of effective diameter of gases. Weadopted the collision diameter of Lennard–Jones potentialσLJ as the effective diameter because of good correlation be-tween the diffusion coefficient and the diameter[7,8]. Table 1also lists the permeability values to CO2 and O2 for theprecursor and pyrolytic membranes. The gas permeabilityfor the precursor membranes was in the following order ofdiamine: ODADS < BDSA < BDSA/BAPF(4/1) < BAHFDS< BAHFDS/BAPF(4/1). The gas permeability of NTDA-ODADS and NTDA-BDSA precursor membranes was toolow to be measured exactly, probably because the pla-nar structure of NTDA and BDSA, flexible structure ofODADS and the strong polarSO3H groups resulted in dense

F A-bb

chain packing of these polyimides. Bulky groups such asC(CF3)2 in BAHFDS and Cardo-structure in BAPF effec-

tively inhibited the chain packing, resulting in the higher gaspermeability[25–27].

The gas permeability for the pyrolytic membraneswas in the following order of diamine of precursor poly-imide: ODADS < BDSA < BDSA/BAPF≤ BAHFDS/BAPF< BAHFDS. The gas permeability for these pyrolyticmembranes seems to be controlled by the sulfonic acidgroup content in the precursor polyimides and the molecularrelaxation of the polymer chains during the pyrolysis aswell as the gas permeation level of the precursor mem-branes. As shown inTable 1, the rate of increase in the gaspermeability by the pyrolysis was in the following order:NTDA-BAHFDS/BAPF(4/1) < NTDA-BAHFDS < NTDA-BDSA/BAPF(4/1) < NTDA-BDSA∼ NTDA-ODADS. Thisis the same as the increasing order ofSO3H content in theprecursor polyimides. This fact strongly suggests that thehigher SO3H content caused the more microvoids by thepyrolysis, resulting in the larger rate of increase in the gaspermeability. The NTDA-BAHFDS/BAPF(4/1) pyrolytic

Fig. 5. Plots of logPvs.σ of gases for the pyrolytic membranes of NTDA-
ig. 4. Plots of logP vs.σLJ of gases for precursor membranes of NTD
◦
ased polyimides at 35C and 1 atm and for 6FDA-based polyimide mem-ranes at 50◦C and 2 atm. b
LJ

ased polyimides at 35◦C and 1 atm.


membrane was less permeable than the NTDA-BAHFDSpyrolytic membrane in spite of the higher permeabilityof the former precursor membrane. This was due to thesmaller rate of increase in the gas permeability as a resultof the lower SO3H content. The pyrolytic membranesof NTDA-ODADS and NTDA-BDSA polyimides showedthe larger rate of increases in the gas permeability by thepyrolysis than the other pyrolytic membranes because of thehigher SO3H contents. However, their gas permeabilitywas still in the lower level, probably because of two reasons.One is the lower level of gas permeability of their precursormembranes, as mentioned above. The other is the ease ofmolecular relaxation of polymer chains during the pyrolysis,as discussed below.

The following is a phenomenon that could be occurringin these membranes during the pyrolysis. Assuming that thepyrolysis temperature of 450◦C is above the glass transition,as the SO3H groups cleave to form microvoids, the polymerchains could relax to repack the resultant microvoids. On theother hand, the cleavage of sulfonic acid groups could simul-taneously induce the crosslinking of polymer chains, whichcould increase the glass transition temperature of the polymerand could reduce the molecular chain relaxation. As a result,the microvoids formed could remain to some extent after thepyrolysis. The bulky groups such asC(CF3)2 and Cardo-structure in BAHFDS and BAPF moieties, respectively, coulde thefl ulde

alsos ares edt oly-it erc meh Sp DA-T asb rel-aC theN er-m gherp

cec A-b ti . Thep dp -r r form lyticmo s asa r.

Fig. 6. Plots of logSvs.ε/k of gases for the pyrolytic membranes at 35◦Cand 1 atm and for the 6FDA-based polyimide membranes at 50◦C and 2 atm.

Fig. 7shows plots of logDversusσLJ of gases for pyrolyticmembranes and 6FDA-based polyimide membranes[7,8].The effective diameter of CO2 used is noted to be 0.35 nmrather than 0.40 nm ofσLJ because of a better correlation

Fand 1 atm and for 6FDA-based polyimide membranes at 50C and 2 atm.

ffectively reduce the repacking of microvoids, whereasexible ether bonds in ODADS and BAHFDS moieties conhance the repacking.

Gas permeability data for 6FDA-based polyimides arehown inFig. 4for comparison. Their chemical structureshown inFig. 1. 6FDA-TrMPD polyimide has been reporto display the highest level of gas permeability among pmides [7,28]. The methyl groups on bothortho positionso an imide ring greatly inhibit dense packing of polymhains, resulting in high fraction of large-size free voluoles[28,29]. The gas permeability for the NTDA-BAHFDyrolytic membrane was comparable to that for the 6FrMPD polyimide membrane. 6FDA-DDBT polyimide heen reported to display the high permselectivity with thetively low permeability for olefin/paraffin separations[7,8].ompared to the 6FDA-DDBT polyimide membrane,TDA-BDSA pyrolytic membrane showed the similar peability to the gases with smaller diameter but the hiermeability to the gases with larger diameter.

Fig. 6 shows plots of logS versus Lennard–Jones foronstantε/k of gases for pyrolytic membranes and 6FDased polyimide membranes[7,8]. There is a tendency thaS

s larger for more condensable gas for every membraneyrolytic membranes showed higherSthan the 6FDA-baseolyimide membranes. The difference inSbetween the pyolytic and 6FDA-based polyimide membranes was largeore condensable gases. The NTDA-BDSA/BAPF pyroembrane showed the higherS than the NTDA-BAHFDSne. This may be due to the larger amount of microvoidresult of the higherSO3H group content of its precurso

ig. 7. Plots of logD vs.σLJ of gases for the pyrolytic membranes at 35◦C◦


Table 3Pure gas permeability and permselectivity of pyrolytic membranes at 35◦C and 1 atm and 6FDA-based polyimide membranes at 50◦C and 2 atma

Membranes PO2 PO2/PN2 PCO2 PCO2/PN2 PC2H4 PC2H4/PC2H6

NTDA-BAHFDS pyrolytic 121 4.0 920 31 66 3.5NTDA-BAHFDS/BAPF(4/1) pyrolytic 69 4.0 400 23 30 3.4NTDA-BDSA/BAPF (4/1) pyrolytic 54 4.1 380 29 31 4.2NTDA-BDSA pyrolytic 28 4.8 167 29 14 4.8NTDA-ODADS pyrolytic 8.1 5.6 52 36 2.3 3.16FDA-TrMPD polyimideb 126 3.4 570 15 58 2.96FDA-DDBT polyimideb 24 4.8 116 23 5.4 5.0

a P is in 10−10 cm3 (STP) cm/(cm2 s cmHg).b Cited from Refs.[7,8].

[7,8,30]. The 6FDA-based polyimide membranes had theroughly linear correlation between logD and the diame-ter [7,8]. 6FDA-TrMPD polyimide membrane is known tohave relatively high fraction of larger-size free volume holesand the high gas diffusivity. The NTDA-BAHFDS pyrolyticmembrane displayed the high gas diffusivity comparable tothat for 6FDA-TrMPD polyimide membrane. The microvoidsformed by decomposition ofSO3H groups are considered toremain as large-size free volume holes in the polymer matrix,resulting in the high gas diffusivity. Although having thesimilar diffusivity to gas with smaller diameter, the NTDA-BDSA/BAPF pyrolytic membrane had the higher diffusivityto gas with larger diameter compared to 6FDA-DDBT poly-imide membrane. The pyrolytic membranes might have ahigher fraction of the larger free volume holes compared withthe polyimide membrane with the relatively dense packing ofpolymer chains such as 6FDA-DDBT.

Pure gas permeability and selectivity for some gas pairsare listed inTables 3 and 4. The clear superiority of thepyrolytic membranes to the 6FDA-based polyimide mem-branes was observed only for C3H6/C3H8 separation. Thepyrolytic membranes showed the higher permeability ofC3H6 and/or the higher selectivity of C3H6/C3H8 than thepolyimide membranes. The higher permeability was due tothe higher solubility of C3H6 and the higher or compara-ble diffusivity of C H for the pyrolytic membranes. Theh ityo itys ort Asc e

and ethylene, being planar molecules, tend to be larger thanthose evaluated from the correlation lines between logD andσLJ. This tendency is clearer for propylene and the pyrolyticmembranes. The pyrolytic membranes might have the shape-selectivity that planar molecules can diffuse through thematrix more easily than less-planar molecules.

3.3. Sorption isotherm

In general, sorption isotherms of gases in glassy polymersare well described by the dual-mode sorption model[31].According to this model, the sorption amountC is representedby the following equation:

C = kDp + C′Hbp

1 + bp(4)

wherekD is the Henry’s law solubility constant,C′H the Lang-

muir capacity constant,b the Langmuir affinity constant andp the equilibrium gas pressure.

Fig. 8 shows sorption isotherms of C3H6 and C3H8for the pyrolytic membranes. The measurement for eachgas was carried out in the order of increasing experimen-tal pressure. These isotherms were analyzed by Eq.(4)using a non-linear least squares regression technique toobtain the values of the three sorption parameters. The val-ues thus obtained are listed inTable 5 together with thed olidaE e ingw oids

TP tios of3H6 -T

M SC

N 4N –NN –6 26 1

n 10−2 c

/50 mol

3 6igher selectivity was due to the higher diffusivity selectivf C3H6/C3H8 for the pyrolytic membranes. The solubilelectivity of C3H6/C3H8 was in the range of 1.2–1.3 fhe pyrolytic and 6FDA-based polyimide membranes.an be seen inFig. 7, the diffusivity values of propylen

able 4ure gas permeability, diffusion and solubility coefficients and their rarMPD and 6FDA-DDBT polyimide membranes at 50◦C and 2 atma

embranes PC3H6c DC3H6

TDA-BAHFDS pyrolytic 41 (30) 0.97TDA-BAHFDS/BAPF(4/1) pyrolytic 15 –TDA-BDSA/BAPF(4/1) pyrolytic 15 (9.3) 0.27TDA-BDSA pyrolytic 6.4 –FDA-TrMPD polyimideb 30 (20) 1.3FDA-DDBT polyimideb 1.8 0.10

a P is in 10−10 cm3 (STP) cm/(cm2 s cmHg),D is in 10−8 cm2/s andS is ib Cited from Refs.[7,8].c The data in parenthesis were obtained for mixed components (50

ata for the 6FDA-TrMPD polyimide membrane. The snd dotted lines shown inFig. 8 were calculated fromq. (4) using the parameter values. These lines werood agreement with the experimental data. TheC′

H value,hich is considered as a measure of amount of microv

C/C3H8 system for pyrolytic membranes at 35◦C and 1 atm and for 6FDA

3H6 PC3H6/PC3H8c DC3H6/DC3H8 SC3H6/SC3H8

2 26 (11) 20 1.321 – –

57 34 (19) 28 1.229 – –

3 10 (6.0) 8.8 1.29 20 16 1.3

m3 (STP)/(cm3 mmHg).

%) system.


Fig. 8. Sorption isotherms of C3H6 and C3H8 of the pyrolytic membranesfrom NTDA-BAHFDS and NTDA-BDSA/BAPF(4/1) polyimides at 50◦C.Lines are calculated from Eq.(4).

formed in the pyrolysis for the pyrolytic membranes, wasin the following order: 6FDA-TrMPD polyimide < NTDA-BAHFDS pyrolytic < NTDA-BDSA/BAPF pyrolytic. Thissuggests that the higherSO3H content of precursor led tothe larger amount of microvoids. The Langmuir populationdominated to the total sorption amount for these membranesespecially at a lower pressure. For example, fractions of theLangmuir population were 84, 90 and 96% of C3H6 sorptionfor 6FDA-TrMPD polyimide, the NTDA-BAHFDS and theNTDA-BDSA/BAPF pyrolytic membranes, respectively, at1 atm and 50◦C.

3.4. Permeation and separation properties to mixedgases

Fig. 9 shows time course ofPC3H6, PC3H8 andα(C3H6/C3H8) in mixed gas (50/50 mol%) permeation for theNTDA-BDSA/BAPF pyrolytic membrane. For comparison,time course of pure gas permeation at the same partial pres-

Table 5Sorption parameters of C3H6 and C3H8 for pyrolytic and 6FDA-TrMPDpolyimide membranes at 50◦C

Membranes C3H6 C3H8

kD C′H b kD C′

H b

6 .0N .5N 6

k

Fig. 9. Time course ofPC3H6, PC3H8 andα(C3H6/C3H8) in mixed gas per-meation for the NTDA-BDSA/BAPF(4/1) pyrolytic membrane at a feedcomposition of 50/50 mol% and total pressure of 2 atm at 35◦C. For com-parison, time course in single gas permeation at 35◦C and 1 atm are alsoshown.

sure difference as that of mixed gas permeation is also shownin this figure. ThePC3H6 in pure gas permeation reached asteady state within 30 min or three times the time lag. Onthe other hand, thePC3H6 in mixed gas permeation turned todecrease slowly after its initial great rise and reached a steadystate in a much longer time. The steady statePC3H6 value inmixed gas permeation was lower than the value in pure gaspermeation by 40%. Although it took 720 min for thePC3H8

in pure gas permeation to reach a steady state, thePC3H8 inmixed gas permeation reached within 300 min. The steadystatePC3H8 value in mixed gas permeation was larger thanthe value in pure gas permeation by 10%. Permeability of themore permeable gas decreased and that of the less perme-able gas increased. As a result, the selectivity (19) decreaseddown to 55% of the ideal selectivity. Similar time course wasobserved for the NTDA-BAHFDS pyrolytic membrane. Thistime course is similar to the result observed for the permeationof 1,3-butadiene/n-butane mixture through a 6FDA-DDBTpolyimide membrane[8]. The increase inPC3H8 was due tothe increase in diffusivity, judging from the shorter time lagcompared with pure gas permeation. This is due to a kindof plasticization caused by sorbed gas molecules. Solubil-ity coefficients of both gases in the mixed gas system areconsidered to be lower than those in the pure gas systembecause of competitive sorption to the microvoids. Therefore,the decrease inP was due to the competitive sorption in

FDA-TrMPD polyimide 3.9 26 3.8 5.1 17 7TDA-BAHFDS pyrolytic 2.7 35 2.7 1.8 31 2TDA-BDSA/BAPF(4/1)pyrolytic

1.8 46 5.7 1.7 37 5.

D is in cm3 (STP) cm−3 atm−1, C′H is in cm3 (STP) cm−3 andb is in atm−1.
C3H6


Langmuir sites whereas a little increase inPC3H8 was dueto the result of the synergetic effects of competitive sorptionand plasticization.

3.5. Comparison of membrane performance ofC3H6/C3H8 separation

Burns and Koros have conducted a comprehensive reviewby evaluating all of the available literature data to define theupper bound relationship for C3H6/C3H8 separation[12].Fig. 10 shows plots of the data for polymeric membranescompiled in the literature and the upper bound line[12].Two plots with higher permeability than 100 Barrer and muchlower selectivity less than 1.7 and one plot at 100◦C are omit-ted for convenience. The upper bound line was practicallyconstructed using the data obtained in pure gas permeation.The membranes defining the upper bound line are 6FDA-DDBT and 6FDA-TrMPD polyimides, which have high fac-tional free volume, larger average size of free volume holesand rigid polymer chains. The plot for 6FDA/BPDA(1/1)-DDBT polyimide membrane[32] in pure gas permeation liesa little above the upper bound line. The plots for the presentpyrolytic membranes in pure gas permeation lie above andbeyond the boundary. The pyrolytic membranes from SPIsare superior to normal polymeric membranes for C3H6/C3H8s

erel em-b levelo DA-

Fp -i int eso ata inm ition of5 nes.T s.

DDBT, 6FDA-TrMPD and 6FDA/BPDA(1/1)-DDBT poly-imide membranes[7,32]. Consequently, the pyrolytic mem-branes still displayed the higher permeability and selectivityfor mixed gas permeation than the polyimide membranes.

4. Conclusion

Dense and flat SPI membranes were pyrolyzed at 450◦Cfor 1.5 h to prepare flexible membranes having large amountof microvoids caused by “template-like effect” ofSO3Hgroups. The low-temperature pyrolysis increased gas perme-ability with keeping the membrane flexibility and toughness.The rate of increase in gas permeability was larger for the SPIwith the higher SO3H content. The gas permeability for thepyrolytic membranes seems to be controlled by theSO3Hcontent in SPI, the relaxation of the polymer chains duringthe pyrolysis and the gas permeation level of the precursormembranes. The microvoids formed by cleavage ofSO3Hgroups are considered to remain as larger-size free volumeholes in the polymer matrix, resulting in the higher diffusivityand solubility to gas with larger diameter.

The pyrolytic membranes showed the higher permeabil-ity of C3H6 and/or the higher selectivity of C3H6/C3H8 thanthe 6FDA-polyimide membranes. The higher permeabilityw ro ywT achp rme-a op-o sim-i ver,t heh torf nes.T ixedp merm highs

A

forD theM ch-n

R

Ind.

ns-Sci.

eparation.The separation factors for mixed gas permeation w

ower than the ideal separation factors for the pyrolytic mranes by 45–60% as mentioned above. The similarf reduction in the selectivity has been observed for 6F

ig. 10. Plots ofα(C3H6/C3H8) or αid(C3H6/C3H8) vs. PC3H6 for theyrolytic membranes from NTDA-based SPIs at 50◦C, 6FDA-based poly

mide membranes[7,32] at 35◦C, and polymeric membranes compiledhe literature[12] at temperatures from 26 to 50◦C and at feed pressurf 2–4 atm. The data in pure gas permeation are at 1 atm and the dixed gas permeation are at total pressure of 2 atm and feed compos0/50 mol% for the pyrolytic and the 6FDA-based polyimide membrahe pyrolytic membranes are indicated with the diamine abbreviation

as due to the higher solubility of C3H6 and the higher comparable diffusivity of C3H6. The higher selectivitas due to the higher diffusivity selectivity of C3H6/C3H8.he separation factor in mixed gas permeation for eyrolytic membrane was lower than that in pure gas petion. This was due to the competitive sorption in micrres and the plasticization effect. This reduction was

lar to that observed for polyimide membranes. Howehe NTDA-BAHFDS pyrolytic membrane still displayed tigher C3H6 permeability and the higher separation fac

or C3H6/C3H8 separation than other polymeric membrahus, the pyrolytic membranes had the interesting intermroperties of toughness and good flexibility as in polyembranes and high gas permeability with reasonably

electivity as in carbon molecular sieve membranes.

cknowledgement

This work was supported partly by a Grant-in-Aidevelopment Scientific Research (No. 13355031) froministry of Education, Culture, Sports, Science, and Teology of Japan.

eferences

[1] R.B. Eldridge, Olefin/paraffin separation technology: a review,Eng. Chem. Res. 32 (1993) 2208.

[2] J.C. Davis, R.J. Valus, R. Eshraghi, A.E. Velikoff, Facilitated traport membrane hybrid systems for olefin purification, Sep.Technol. 28 (1993) 463.


[3] I. Pinnau, L.G. Toy, Solid polymer electrolyte composite membranesfor olefin/paraffin separation, J. Membr. Sci. 184 (2001) 39.

[4] A. Morisato, Z. He, I. Pinnau, T. Merkel, Transport propertiesof PA12-PTMO/AgBF4 solid polymer electrolyte membranes forolefin/paraffin separation, Polym. Eng. Sci. 85 (2001) 96.

[5] A. Ito, S.-T. Hwang, Permeation of propane and propylene throughcellulosic polymer membranes, J. Appl. Polym. Sci. 38 (1989) 483.

[6] K.R. Lee, S.T. Hwang, Separation of propylene and propane bypolyimide hollow-fiber membrane module, J. Membr. Sci. 73 (1992)37.

[7] K. Tanaka, A. Taguchi, J. Hao, H. Kita, K. Okamoto, Permeationand separation properties of polyimide membranes to olefins andparaffins, J. Membr. Sci. 121 (1996) 197.

[8] K. Okamoto, K. Noborio, J. Hao, K. Tanaka, H. Kita, Permeationand separation properties of polyimide membranes to 1,3-butadieneandn-butane, J. Membr. Sci. 134 (1997) 171.

[9] A. Shimazu, T. Miyazaki, T. Matsushita, M. Maeda, K. Ikeda, Rela-tionships between chemical structures and solubility, diffusivity andpermselectivity of 1,3-butadiene andn-butane in 6FDA-based poly-imides, J. Polym. Sci., Part B: Polym. Phys. 37 (1999) 2941.

[10] A. Shimazu, T. Miyazaki, M. Maeda, K. Ikeda, Relationshipsbetween chemical structures and the solubility, diffusivity and perms-electivity of propylene and propane in 6FDA-based polyimides, J.Polym. Sci., Part B: Polym. Phys. 38 (2000) 2525.

[11] C.S. Bickel, W.J. Koros, Olefin/paraffin gas separations with 6FDA-based polyimide membranes, J. Membr. Sci. 170 (2000) 205.

[12] R.L. Burns, W.J. Koros, Defining the challenges for C3H6/C3H8

separation using polymeric membranes, J. Membr. Sci. 211 (2003)299.

[13] J.J. Krol, M. Boerrigter, G.H. Koops, Polyimide hollow fiber gas sep-aration membranes: preparation and the suppression of plasticization

275.[ oka,

e sys-.

[ rbon

[ H.their

[ N.izedber

[ aga,arbonllow

[19] N. Tanihara, H. Shimazaki, Y. Hirayama, S. Nakanishi, T. Yoshinaga,Y. Kusuki, Gas permeation properties of asymmetric carbon hollowfiber membranes prepared from asymmetric polyimide hollow fiber,J. Membr. Sci. 160 (1999) 179.

[20] W. Zhou, M. Yoshino, H. Kita, K. Okamoto, Carbon molecular sievemembranes derived from phenolic resin with a pendant sulfonic acidgroup, Ind. Eng. Chem. Res. 40 (2001) 4801.

[21] W. Zhou, M. Yoshino, H. Kita, K. Okamoto, Preparation andgas permeation properties of carbon molecular sieve membranesbased on sulfonated phenolic resin, J. Membr. Sci. 217 (2003)55.

[22] Y.K. Kim, H.B. Park, Y.M. Lee, Carbon molecular sieve membranesderived from metal-substituted sulfonated polyimide and their gasseparation properties, J. Membr. Sci. 226 (2003) 145.

[23] W. Zhou, T. Watari, H. Kita, K. Okamoto, Gas permeation propertiesof flexible pyrolytic membranes from sulfonated polyimides, Chem.Lett. (2002) 534.

[24] J. Fang, X. Guo, S. Harada, T. Watari, K. Tanaka, H. Kita, K.Okamoto, Novel sulfonated polyimides as polyelectrolytes for fuelcell application. 1. Synthesis, proton conductivity and water stabilityof polyimides from 4,4′-diaminodiphenyl ether-2,2′-disulfonic acid,Macromolecules 35 (2002) 9022.

[25] K. Tanaka, H. Kita, M. Okano, K. Okamoto, Permeability and perms-electivity of gases in fluorinated and non-fluorinated polyimides,Polymer 33 (1992) 585.

[26] T.H. Kim, W.J. Koros, G.R. Husk, K.C. O’Brien, Relationshipbetween gas separation properties and chemical structure in a seriesof aromatic polyimides, J. Membr. Sci. 37 (1988) 45.

[27] S.A. Stern, Y. Mi, H. Yamamoto, A.K.St. Clair, Struc-ture/permeability relationships of polyimide membranes. Applica-tions of the separation of gas mixtures, J. Polym. Sci., Part B: Polym.

[ t ofes ines, J.

[ Ito,rop-992)

[ ai,from

[ rt in

[ , Y.llowci.

in propane/propylene environments, J. Membr. Sci. 184 (2001)14] J. Hayashi, H. Mizuta, M. Yamamoto, K. Kusakabe, S. Moro

S.H. Suh, Separation of ethane/ethylene and propane/propylentems with a carbonized BPDA-p,p′-ODA polyimide membrane, IndEng. Chem. Res. 35 (1996) 4176.

15] H. Suda, K. Haraya, Alkene/alkane permselectivities of a camolecular sieve membrane, Chem. Commun. (1997) 93.

16] K. Okamoto, M. Yoshino, K. Noborio, H. Maeda, K. Tanaka,Kita, Preparation of carbon molecular sieve membranes andgas separation properties, ACS Symp. Ser. 744 (2000) 314.

17] K. Okamoto, S. Kawamura, M. Yoshino, H. Kita, Y. Hirayama,Tanihara, Y. Kusuki, Olefin/paraffin separation through carbonmembranes derived from an asymmetric polyimide hollow fimembrane, Ind. Eng. Chem. Res. 38 (1999) 4424.

18] Y. Kusuki, H. Shimazaki, N. Tanihara, S. Nakanishi, T. YoshinGas permeation properties and characterization of asymmetric cmembranes prepared by pyrolyzing asymmetric polyimide hofiber membrane, J. Membr. Sci. 134 (1997) 245.

Phys. 27 (1989) 1887.28] K. Tanaka, M. Okano, H. Toshino, H. Kita, K. Okamoto, Effec

methyl substituents on permeability and permselectivity of gaspolyimides prepared from methyl-substituted phenylenediaminPolym. Sci., Polym. Phys. Ed. 30 (1992) 907.

29] K. Tanaka, M. Katsube, K. Okamoto, H. Kita, O. Sueoka, Y.Correlation between positron annihilation and gas diffusion perties of a series of polyimides, Bull. Chem. Soc. Jpn. 65 (11891.

30] K. Okamoto, K. Tanaka, H. Kita, M. Ishida, M. Kakimoto, Y. ImGas permeability and permselectivity of polyimides prepared4,4′-diaminotriphenylamine, Polym. J. 24 (1992) 451.

31] D.R. Paul, W.J. Koros, Carbon dioxide sorption and transpopolycarbonate, J. Polym. Sci. 14 (1976) 675.

32] M. Yoshino, S. Nakamura, H. Kita, K. Okamotoa, N. TaniharaKusuki, Olefin/paraffin separation performance of asymmetric hofiber membrane of 6FDA/BPDA-DDBT copolyimide, J. Membr. S212 (2003) 13.

Documents

Preparation and gas separation performance of flexible pyrolytic membranes by low-temperature pyrolysis of sulfonated polyimides