Cm

La

b

a

ARR1AA

KPCMG

1

sthdiPamTiteimHnhbg

0d

Journal of Membrane Science 326 (2009) 285–292

Contents lists available at ScienceDirect

Journal of Membrane Science

journa l homepage: www.e lsev ier .com/ locate /memsci

rosslinking and stabilization of nanoparticle filled PMP nanocompositeembranes for gas separations

ei Shao a, Jon Samseth b, May-Britt Hägg a,∗

Department of Chemical Engineering, Faculty of Natural Sciences and Technology, Norwegian University of Science and Technology, N-7491 Trondheim, NorwaySINTEF Materials and Chemistry, N-7465 Trondheim/Akershus University College, N-2001 Lillestrøm, Norway

r t i c l e i n f o

rticle history:eceived 21 April 2008eceived in revised form5 September 2008ccepted 29 September 2008

a b s t r a c t

Poly(4-methyl-2-pentyne) (PMP) has been crosslinked using 4,4′-(hexafluoroisopropylidene) diphenylazide (HFBAA) to improve its chemical and physical stability over time. Crosslinking PMP renders it insol-uble in good solvents for the uncrosslinked polymer. Gas permeability and fractional free volume (FFV)decreased as crosslinker content increased, while gas sorption was unaffected by crosslinking. Therefore,

vailable online 17 October 2008

eywords:oly(4-methyl-2-pentyne)rosslinking

the reduction in permeability upon crosslinking PMP was due to decrease in diffusion coefficient. Com-pared to the pure PMP membrane, the permeability of the crosslinked membrane is initially reducedfor all gases tested due to the crosslinking. By adding nanoparticles (FS, TiO2), the permeability is againincreased; permeability reductions due to crosslinking could be offset by adding nanoparticles to themembranes. Increased selectivity is documented for the gas pairs O2/N2, H2/N2, CO2/N2, CO2/CH4 andH /CH using crosslinking and addition of nanoparticles. Crosslinking is successful in maintaining the

ty of

c2lsm

s(ttiisdvb1

t

embraneas separation

2 4

permeability and selectivi

. Introduction

The use of polymeric materials for membrane gas and vaporeparation has been on the market for more than 25 years, andhe potential for use is steadily increasing. High permeability andigh selectivity in favor of the permeating gas as well as goodurability and high mechanical strength of the material are the

mportant properties for a commercial gas separating membrane.oly(4-methyl-2-pentyne) (PMP) is an amorphous, disubstitutedcetylene-based high free volume glassy polymer. It is one of theost permeable purely hydrocarbon-based polymer known [1].

he high permeation property and hence also low gas selectiv-ty of PMP, results from very poor polymer chain packing due tohe stiffness of the polymer chain as can be understood with ref-rence to its chemical structure, having a repeated double bondn the main chain and bulky side groups (Fig. 1). The unique per-

eation property has been documented by several authors [2–7].owever, it has also been documented that the gas permeability is

ot stable over time, and that seems to be sensitive to processingistory. PMP undergoes significant physical aging over time causedy the gradual relaxation of non-equilibrium excess free volume inlassy polymers [4,5]. For example, nitrogen permeability coeffi-

∗ Corresponding author. Tel.: +47 7359 4033; fax: +47 7359 4080.E-mail address: may-britt.hagg@chemeng.ntnu.no (M.-B. Hägg).

itttsaadP

376-7388/$ – see front matter © 2008 Elsevier B.V. All rights reserved.oi:10.1016/j.memsci.2008.09.053

PMP membranes and PMP/filler nanocomposites over time.© 2008 Elsevier B.V. All rights reserved.

ient in PMP has been reported to decrease by 25% over a period of9 days [4]. PMP is also soluble in some common organic solvents,

eading to potential dissolution of the membrane if used in processtreams where such solvents could be present. These phenomenaay compromise the practical utility of PMP.

Recently it was reported that the addition of nonporous fumedilica nanoparticles to PMP and poly(1-trimethysilyl-1-propyne)PTMSP) increased their permeabilities with increasing filler con-ent up to a level as high as 30 vol.% [2,4,8]. This behavior is contraryo the observed fact that permeability typically decreases withncreasing filler loading in traditionally filled polymer systems. Thencreased permeability has been ascribed to increased free volumeizes in the polymer caused by the nanosized fumed silica particlesisrupting the packing of the polymer chains. This enhanced freeolume of the filled superglassy polymers has been confirmed byoth positron annihilation lifetime spectroscopy (PALS) [2,4,8] and29Xe NMR [9]. Merkel et al. also observed an increase of both selec-ivity and permeability when fumed silica was added to PMP, whichs remarkable with respect to the generally observed inverse rela-ionship between permeability and selectivity [2]. It was also notedhat, in addition to filler concentration, filler particle size appearso affect permeability in filled superglassy systems as well. In their

tudy [2], five fillers with different particle diameters (7–500 nm)nd different surface chemistries (precipitated silica, hydrophilicnd hydrophobic fumed silicas, carbon black, and �-alumina pow-er) were added to PMP. At a fixed filler volume fraction, differentMP/filler nanocomposites showed significant differences in gas

286 L. Shao et al. / Journal of Membrane

p(a

acope

mpsoseamaibpct[Prmabtdnctbvnr

2

2

mws

P

t˛

a

wtbrtov

S

D

P

woia

oiria

F

wdci

A number of techniques may be used to study free volume inpolymers, including spin probe methods [27], molecular modeling[28,29], inverse gas chromatography [30], small-angle X-ray scat-tering [31], 129Xe NMR [9,32], and positron annihilation lifetimespectroscopy (PALS) [2,4,8,33–35]. PALS is one of the most widely

Fig. 1. Chemical structure of PMP.

ermeability. In general, it appeared that smaller filler particlessmaller than 50 nm) produce larger increases in permeability atconstant filler volume fraction [2].

The physical aging of PMP/silica nanocomposite membrane waslso reported by Merkel et al. [4], the addition of silica nanoparti-les does not arrest physical aging in PMP; the gas permeabilityf nanocomposites is still unstable over time, but exhibits higherermeability than the base polymer throughout the aging periodxamined [4].

Polymer modification by crosslinking has attracted interest inembrane-based separation of gases or vapors. Today crosslinked

oly(dimethylsiloxane) (PDMS) is used commercially as a vaporeparation membrane material [10]. Crosslinked polymer filmsffer obvious advantages as membranes, particularly in terms oftability. Without crosslinking the material will tend to swell whenxposed to certain gas mixtures, and hence separation propertiesre affected. There is a lot of published literature on crosslinkedembranes for gas and liquid separation [11–19], with the most

ttractive processes being those initiated thermally or photochem-cally. Jia and Baker [20] reported that crosslinking PTMSP withis(aryl azide)s has been shown to increase the chemical andhysical stability. Crosslinked PTMSP membranes are insoluble inommon PTMSP solvents such as toluene, and the permeability ofhe crosslinked membranes is reported to be constant over time20]. These results are very interesting for the current work, asTMSP and PMP are both high free volume polymers. All theseesults encouraged us to adapt the technique of crosslinking PMP

embranes in order to increase its stability. A plausible mech-nism for the crosslinking reactions is shown in Fig. 2 where ais(aryl azide) is being used as a crosslinker [20]. Under pho-ochemical irradiation or thermal treatment the bis (aryl azide)ecomposes to nitrogen gas and reactive nitrenes, the resultingitrenes can add to double bonds to form aziridines or insert intoarbon-hydrogen bonds in PMP to form substituted amines. Inhe current study, the chemical stability of PMP crosslinked withisazide is reported. The permeability, solubility, and diffusivity ofarious gases in crosslinked PMP are presented, and the effect ofanoparticle addition on crosslinked PMP transport properties iseported.

. Background

.1. Gas transport and free volume in polymers

The permeation of gases and vapors through a dense polymeric

embrane is generally described as a solution–diffusion processhere permeability, P, is the product of the gas diffusivity, D, and

olubility, S, i.e.,

= DS (1)

Science 326 (2009) 285–292

The ability of a membrane to separate two molecules, A and B, ishe ratio of their permeabilities; here expressed as ideal selectivityA/B,

A/B = PA

PB=

(DA

DB

)(SA

SB

)(2)

here the first term on the right-hand side is the diffusivity selec-ivity and the second is the solubility selectivity. The balanceetween the solubility selectivity and the diffusivity selectivityelates strongly to the material properties as well as the proper-ies of gases A and B to be separated. The temperature dependencyf solubility, diffusivity and permeability may be expressed as thean’t Hoff–Arrhenius relationships [21–23].

= S0 exp(−�HS

RT

)(3)

= D0 exp(−Ed

RT

)(4)

= P0 exp(−Ep

RT

)(5)

here S0, D0 and P0 are pre-exponential factors, �HS is the enthalpyf sorption, Ed is the activation energy of diffusion, Ep (=Ed + �HS)

s the activation energy of permeation, R is the ideal gas constant,nd T is the absolute temperature.

Molecular diffusion through a dense polymer depends stronglyn the amount of free volume that a material possesses; i.e. depend-

ng on the structure as well as the state of the polymer (glassy orubbery). The quantity used to compare the amount of free volumen polymers is the fractional free volume (FFV); usually estimatedccording to Bondi’s method:

FV = (V − 1.3VW)V

(6)

here V is the polymer specific volume (i.e., reciprocal of geometricensity) and VW is the specific van der Waals volume, which can bealculated by group contribution methods [24–26]. The FFV of PMPs 0.28, which is one of the highest values of any known polymer.

Fig. 2. Illustration of crosslinking reaction of PMP.

brane Science 326 (2009) 285–292 287

amt

3

3

crNfiCn(mfib

3

mchafigcmtmuroNs

3

s((dr

danrs1p

Fc

avoao

3

Cppncs0tdrrfrau4itare

sH2

3

L. Shao et al. / Journal of Mem

pplied techniques and provides the most direct and detailed infor-ation on the size and concentration of free volume elements in

he materials.

. Experimental

.1. Instrumental characterization

NMR spectra were recorded on Bruker Avance DPX 400 withhemical shifts referenced to tetramethylsilane for deuteriochlo-oform. FT-IR spectra were recorded on a Thermo Nicolet FT-IRexus spectrometer. NMR and FT-IR spectra were used to con-rm chemical structure. UV–vis spectra were recorded on a Varianary 50 UV–vis spectrophotometer. Morphology of cross sectionalanocomposites was examined by scanning electron microscopyLV FESEM, SUPRATM 55, Zeiss) operating in variable pressure (VP)

ode. In such a VP mode, the SEM analysis could be directly per-ormed on non-conducting polymers without coating. The SEMmages were acquired collecting the secondary electrons inducedy using 5 keV electron beam and 8 mbar partial pressure.

.2. Polymer synthesis

PMP was synthesized as described in literature [1,36,37]. Theonomer, 4-methyl-2-pentyne (Lancaster, Inc.) was dried over cal-

ium hydride for 24 h, and was then distilled in an atmosphere ofigh-purity nitrogen. The catalysts, niobium pentachloride (NbCl5)nd triphenyl bismuth (Ph3Bi) (Aldrich Chemicals) used withouturther purification. A solution of 0.33 g of NbCl5, and 0.54 g Ph3Bin 47 ml cyclohexane was stirred at 80 ◦C for 10 min under dry nitro-en. Then the monomer solution of 5 g 4-methyl-2-pentyne in 7 mlyclohexane was added dropwise to the catalyst solution, and theixture was reacted at 80 ◦C for 4 h. The viscosity of the solu-

ion increased very rapidly. The resulting gel was precipitated inethanol, filtered to recover the precipitated polymer, and dried

nder vacuum. The polymer was dissolved in cyclohexane andeprecipitated twice from methanol to remove excess monomer,ligomers and catalysts. The polymer yield was 85%. 1H and 13CMR and FT-IR analyses of the polymer confirmed the chemical

tructure of PMP (see Fig. 1).

.3. Synthesis of crosslinking agent

The synthesis and structure of bis azide crosslinking agent ishown in Fig. 3. 4,4′-(hexafluoroisopropylidene) diphenyl azideHFBAA) was obtained by diazotization of the corresponding amineAldrich Chemicals) followed by nucleophilic displacement of theiazonium salt with NaN3. The synthetic procedure follows thateported in the literature [38–41].

4,4′-(Hexafluoroisopropylidene)dianiline (0.70 g, 2 mmol) wasissolved in 2 ml of water containing 1.1 ml of concentrated HC1,nd cooled to 0 ◦C, then treated dropwise with a solution of sodium

itrite (0.34 g, 5 mmol) in 1.2 ml of water. After the addition, theeaction was maintained at 0–5 ◦C for 1.5 h. To the resultant clearolution was added dropwise 0.31 g (5 mmol) of sodium azide in.2 ml of water. The solution was stirred for 15 min, as a whiterecipitate formed. The solid was collected, washed with water,

tat

Fig. 3. Synthesis of 4,4′-(hexafluoroisopro

ig. 4. UV–vis spectra of the bis(aryl azide) HFBAA crosslinking agent used forrosslinking PMP membranes.

llowed to dry, dissolved in dichloromethane, and heated with acti-ated charcoal. Filtration and solvent evaporation gave 0.63 g (82%)f 4,4′-(hexafluoroisopropylidene) diphenyl azide. 1H and 13C NMRnd FT-IR analyses of the product confirmed the chemical structuref HFBAA (see Fig. 3).

.4. Membrane preparation and modification

Dense membranes of pure PMP, and PMP/HFBAA were cast fromCl4 solution containing 1.2 wt.% polymer. Fumed silica (FS) with arimary particle size of 7 nm and titanium dioxide (TiO2) with arimary particle size of 21 nm were obtained from Degussa. Theanocomposite membranes were prepared by a three-step solventasting procedure. Nanoparticles (FS or TiO2) were initially dis-olved in CCl4 at room temperature with ultrasonic treatment at◦C for 30 min followed by stirring with a magnetic stirrer at room

emperature for 5 h. The PMP (or PMP with HFBAA added) was thenissolved in the nanoparticles/CCl4 dispersion with magnetic stir-ing for 3 days, after which the solution was poured into a castinging placed on a glass plate and covered with a funnel to allowor slow solvent evaporation. The casted membranes were dried atoom temperature for 5 days and then placed in a vacuum ovent room temperature for at least 24 h to completely remove resid-al solvent. The final as-cast membrane thicknesses varied from0 to 50 �m. All membranes were crosslinked by UV irradiation

n a vacuum oven at room temperature for 30 min. After 30 min,he peak associated with the azide group in the FT-IR spectrumt 2110 cm−1 disappeared, indicating that the azide molecules hadeacted to form nitrenes. Crosslinking time is kept constant so thatach sample has the same photochemical crosslinking history.

The irradiating wavelength for photochemical crosslinking waset to correspond to the peak of the absorption band for the bisazideFBAA, the peak of the absorption is near 254 nm (see Fig. 4). Thus54 nm light was used for crosslinking membranes.

.5. Permeability measurements

The membranes were masked using an impermeable aluminumape, leaving open a defined permeation area. Epoxy was thenpplied along the interface of the tape and the membrane. A sin-ered metal disc covered with a filter paper was used as support for

pylidene) diphenyl azide (HFBAA).

288 L. Shao et al. / Journal of Membrane

tCmsTsa

tTlo(fsc

4

4

ahaaoasso2s(tm

oPafHmiti

picb

attiabaaCcp

Tt3pHnt

gmcutit

odtCcrosslinked PMP is more size selective to gases than uncrosslinked

TG

C

A

NH

Fig. 5. FT-IR spectra of PMP/3 wt.% HFBAA composite membranes.

he membrane in the test cell. Single gases (N2, O2, H2, CH4 andO2) were measured using a constant volume/variable-pressureethod in a standard pressure–rise setup (MKS Baratron® pres-

ure transducer, 0–134 mbar range) with LabView® data logging.he temperature was varied between 35 and 70 ◦C; the feed pres-ure was varied between 2 and 10 bar. The experimental methodnd equipment is described elsewhere [42,43].

The membrane thickness was measured by an electronic Mitu-oyo 2109F thickness gauge (Mitutoyo Corp., Kanagawa, Japan).he gauge was a non-destructive drop-down type with a reso-

ution of 1 �m. Flat sheet membrane was scanned at a scalingf 100% (uncompressed tiff-format) and analyzed by Scion ImageScion Corp., MD, USA) software. This image tool is availablerom http://www.scioncorp.com at no cost. The effective area wasketched with the draw-by-hand tool both clockwise and counter-lockwise several times.

. Results and discussion

.1. Uncrosslinked and crosslinked PMP membranes

The bis azide HFBAA was used as crosslinking agent. This biszide dissolved easily in PMP to form homogeneous mixtures. Atigh loadings (>4.5 wt.% HFBAA) the membranes became cloudynd optical microscopy confirmed phase separation of crosslinkernd polymer. All crosslinking studies reported here were performedn clear membranes which showed no apparent signs of phase sep-ration. The dried membranes were clear, and UV–vis and FT-IRpectra show that the spectra of the as-prepared membranes areimply the linear combination of the spectrum of PMP and thatf the azide crosslinker. The stretching vibration for the azide at110 cm−1 is easily monitored in the FT-IR, and the loss in its inten-

ity can be correlated with the progress of the crosslinking reactionFig. 5). Double bonds on the PMP backbone and methyl groups onhe side chains are two possible crosslinking sites, but the latter is

ore likely since access to the double bonds is sterically hindered.

Pico

able 1as permeabilities [Barrer]a of uncrosslinked and photochemically crosslinked PMP mem

rosslinking agent Before crosslinking

zide wt.% Azide N2 O2 H2 CH4

oneb 0 950 1780 3970 1790FBAA 1.1 790 1580 3660 1500

2.0 730 1540 3610 14303.0 690 1420 3490 1070

a Permeability is in unit of Barrer (1 Barrer = 10−10 cm3(STP) cm cm−2 s−1 cm Hg−1).b Irradiated at 254 nm for 30 min.

Science 326 (2009) 285–292

A preliminary indication of significant crosslinking is the lackf solubility of crosslinked membranes in standard solvents forMP. In this regard, crosslinked PMP was insoluble in cyclohex-ne and carbon tetrachloride, which are known to be good solventsor uncrosslinked PMP [1]. It was found that when the bisazideFBAA concentration was about HFBAA 1.0 wt.% or higher, PMPembranes were insoluble in carbon tetrachloride. The insolubil-

ty of crosslinked PMP membranes is defined when there is lesshan 0.5% weight loss in a dry membrane before and after soakingn carbon tetrachloride for 24 h.

Photochemical crosslinking of membranes with HFBAA, waserformed at 254 nm using a lamp at room temperature for 30 min

n a vacuum oven. The irradiating wavelength for photochemicalrosslinking was set to correspond to the peak of the absorptionand for the azide (254 nm).

Permeability data in Table 1 show membranes with differentmounts of crosslinker. For all gases considered with the excep-ion of that for CH4 (largest molecule), the addition of bisazideo PMP decreased the permeability only slightly before photo-rradiation, with hardly any improvement of selectivity (Table 2);gain with the exception of that for CO2/CH4. The crosslinked mem-ranes show a significant decrease in permeabilities (Table 1) forll gases except for H2 (the smallest molecule), steadily decreasings crosslinker content increase. The selectivities of O2/N2, H2/N2,O2/N2, CO2/CH4 and H2/CH4 increased with increasing amount ofrosslinker. Higher degrees of crosslinking resulted in a lower gasermeability and higher selectivity.

Blanks were run for photochemical reactions (line 1 inables 1 and 2), to study the effects of irradiation on the proper-ies of pure PMP. When membranes were irradiated at 254 nm for0 min (the condition used for the irradiation of PMP/HFBAA com-osites), the selectivities of O2/N2, H2/N2, CO2/N2, CO2/CH4 and2/CH4 improved and the permeabilities decreased. There is a sig-ificantly larger reduction in permeation for N2 and CH4 than forhe other gases when HFBAA is added.

For photochemical crosslinking, the permeability changes inoing from pure PMP, to PMP with azide additive, to the crosslinkedembrane were predictable. For example, the azide additive in

omposite membranes is expected to occupy much of the free vol-me in the polymer and thus the permeability is lower comparedo pure PMP membranes. Crosslinking connects adjacent chains,ncreases the local segment density, and causes a further decline inhe permeability.

The reaction between the bisazide crosslinker and PMP wasbserved using FT-IR analysis. The initial permeability of PMPecreased with increasing crosslinking due to the loss in frac-ional free volume (FFV). The selectivities of O2/N2, H2/N2, CO2/N2,O2/CH4 and H2/CH4 increased as the FFV decreased, showing that

MP. The permeability stability of crosslinked PMP for the gasess clearly improved. The increased stability may be caused byrosslink constraining the PMP chains and not allowing relaxationf the excess, non-equilibrium FFV that is inherent in PMP.

branes at 35 ◦C and feed pressure of 2.0 bar.

After crosslinking

CO2 N2 O2 H2 CH4 CO2

6700 490 1490 3640 920 48606250 380 1190 3420 810 47106100 290 1080 3380 630 43205710 230 990 3300 510 3950

L. Shao et al. / Journal of Membrane Science 326 (2009) 285–292 289

Table 2Selectivitiesa of various gas pairs in uncrosslinked and photochemically crosslinked PMP membranes.

Crosslinking agent Selectivity (before crosslinking) Selectivity (after crosslinking)

Azide wt.% Azide O2/N2 H2/N2 CO2/N2 CO2/CH4 H2/CH4 O2/N2 H2/N2 CO2/N2 CO2/CH4 H2/CH4

Noneb 0 1.9 4.2 7.1 3.7 2.2 3.0 7.4 9.9 5.3 4.0HFBAA 1.1 2.0 4.6 7.9 4.2 2.4 3.1 9.0 12.4 5.8 4.2

2.0 2.1 5.0 8.4 4.3 2.5 3.7 11.7 14.9 6.9 5.4

4m

mTofifiPCt(mctCondtraps

ept

nsarflit

scp

CctcctwdtsdPa

TG

F

(

23

TG

T

(

23

3.0 2.1 5.1 8.3 5.3

a Selectivity is the ratio of the permeabilities for the pure gases.b Irradiated at 254 nm for 30 min.

.2. Uncrosslinked and crosslinked PMP nanocompositeembranes

Fumed silica (FS) and TiO2 were used as fillers, and theembranes prepared as described in Section 3.4. The data in

ables 3 and 4 show that the permeation and selectivity resultsbtained with the uncrosslinked filled PMP membranes increasedor all gases as the filler content increased. The permeabilityncreased about 62–102% for all the gases in the PMP membranelled with 35 wt.% FS, and about 58–92% for all the gases in theMP membrane filled with 35 wt.% TiO2. The selectivities of H2/N2,O2/N2, CO2/CH4 and H2/CH4 decreased slightly as the filler con-ent increased. For these gas pairs at the temperature measured35 ◦C) in this high free volume polymer, the selectivity is deter-

ined primarily by differences in the sizes and hence the diffusionoefficients of the gases. The non-ideal properties of CO2 also con-ributes to the relatively higher selectivity of CO2/N2 compared toO2/CH4. The slight loss of selectivity indicates that the additionf the filler to the PMP matrix created larger gaps in the intercon-ected free volume of the filled membranes, thereby increasing theiffusion coefficient of the larger molecule proportionately morehan the smaller molecule. The selectivities of O2/N2 and CH4/N2emained about the same as loading increased, because the CH4nd N2 molecules are similar in size and not to different in physicalroperties, so both diffusion coefficients increases by roughly theame factor. The same holds for the gas pair O2–N2.

As mentioned above, the particle sizes of FS and TiO2 are differ-nt (FS with a primary particle size of 7 nm and TiO2 with a primaryarticle size of 21 nm), the data obtained in Tables 3 and 4 shownhat at a fixed filler volume fraction, these two different PMP/filler

ppPfi

able 3as permeabilities [Barrer]a and selectivitiesb of pure PMP and PMP/FS nanocomposite m

umed silica Permeability

FS) (wt.%) N2 O2 H2 CH4 CO2

0 (pure PMP) 950 1780 3970 1790 670015 1010 2150 4120 2010 7130

5 1240 2490 4870 2510 87205 1780 3380 6430 3620 11250

a Permeability is in unit of Barrer (1 Barrer = 10−10 cm3(STP) cm cm−2 s−1 cm Hg−1).b Selectivity is the ratio of the permeabilities for the pure gases.

able 4as permeabilities [Barrer]a and selectivitiesb of pure PMP and PMP/TiO2 nanocomposite

iO2 Permeability S

wt.%) N2 O2 H2 CH4 CO2 O

0 (pure PMP) 950 1780 3970 1790 6700 115 980 1980 4030 1980 6980 2

5 1210 2320 4760 2460 8430 25 1680 3110 6270 3420 10970 1

a Permeability is in unit of Barrer (1 Barrer = 10−10 cm3(STP) cm cm−2 s−1 cm Hg−1).b Selectivity is the ratio of the permeabilities for the pure gases.

3.3 4.3 14.4 17.2 7.8 6.5

anocomposites showed a small difference in gas permeability; themaller FS particles seem to produce a larger increases in perme-bility at a constant filler volume fraction. This result is most likelyelated to smaller particles yielding larger polymer/particle inter-acial area, since at a fixed volume fraction of particles there are aarge number of small particles per unit volume of nanocompos-te, which gives them a greater capacity to disrupt chain packing,hereby affecting transport property.

Repeated permeability measurements were performed with theame fillers (FS, TiO2) at the same concentrations, however, nowrosslinked using 2 wt.% of the crosslinker HFBAA. The results areresented in Fig. 6(a–e).

The effect of increasing the filler content on CO2 permeability,O2/N2 and CO2/CH4 selectivities of crosslinked PMP membranesontaining 2 wt.% HFBAA is shown in Fig. 6(a and b). Compared tohe pure PMP membrane, the CO2, N2 and CH4 permeabilities of therosslinked membrane are now initially being reduced due to therosslinking with a fairly large increase in selectivities. By addinghe nanoparticles (FS, TiO2), the permeabilities are again increased,hile the CO2/N2 and CO2/CH4 selectivities show only a small

ecrease. However, with the filler content increased, the selec-ivities of crosslinked PMP/filler nanocomposite membranes areignificantly increased compared to the pure PMP membrane. Theata obtained show that at a fixed filler volume fraction, crosslinkedMP/FS nanocomposites have a little higher CO2 permeability andtrend of a little lower selectivities of CO2/N2 and CO2/CH4, com-

ared to crosslinked PMP/TiO2 nanocomposites. The increase inermeability upon filling, as already seen for the uncrosslinkedMP, was attributed to increased free volume in the nanoparticle-lled PMP nanocomposites as compared to the unfilled polymer.

embranes at 35 ◦C and feed pressure of 2.0 bar.

Selectivity

O2/N2 H2/N2 CO2/N2 CH4/N2 CO2/CH4 H2/CH4

1.9 4.2 7.1 1.9 3.7 2.22.1 4.1 7.1 2.0 3.5 2.02.0 3.9 7.0 2.0 3.5 1.91.9 3.6 6.3 2.0 3.1 1.8

membranes at 35 ◦C and feed pressure of 2.0 bar.

electivity

2/N2 H2/N2 CO2/N2 CH4/N2 CO2/CH4 H2/CH4

.9 4.2 7.1 1.9 3.7 2.2.1 4.1 7.1 2.0 3.5 2.0.0 3.9 7.0 2.0 3.5 1.9.9 3.7 6.5 2.0 3.2 1.8

290 L. Shao et al. / Journal of Membrane Science 326 (2009) 285–292

F and st

IgeItaCcstmttlommtom

HcttmadtncpAcc

ig. 6. (a–e) The effect of nanoparticles (FS, TiO2) content on gas permeabilitiesemperature: 35 ◦C, feed pressure: 2.0 bar.

n the uncrosslinked PMP it was stated that the selectivity for theas pairs CO2–N2 and CO2–CH4 is determined primarily by differ-nces in the sizes and hence the diffusion coefficients of the gases.n the crosslinked PMP it was documented that with a filler con-ent of around 25 wt.% FS or TiO2, the permeability of the CO2 isbout the same as that of pure PMP, but the selectivity of bothO2/CH4 and CO2/N2 in the crosslinked PMP is now doubled. Theomparable permeability indicates that there is approximately theame FFV available in the uncrosslinked PMP without filler as inhe crosslinked PMP with 25 wt.% filler. The increased selectivity

ust therefore also be attributed to the differences in transport ofhese gases through the restricted crosslinked matrix. For CO2, ashe smallest, but also a highly non-ideal gas, the sorption will mostikely also play an important role compared to CH4 and N2. The lossf selectivity with increased filler content also for the crosslinked

embranes, indicates that the addition of the filler to the PMPatrix creates larger gaps in the interconnected free volume of

he filled membranes, thereby increasing the diffusion coefficientf the larger molecule proportionately more than the smallerolecule.

iCoPa

electivities of crosslinked PMP membranes containing 2 wt.% HFBAA crosslinker;

The effect of increasing the filler content on H2 permeability,2/N2 and H2/CH4 selectivities of crosslinked PMP membranes

ontaining 2 wt.% HFBAA is shown in Fig. 6(c and d). Similarly,he H2 permeability of the crosslinked membrane is also ini-ially reduced because of crosslinking compared to the pure PMP

embrane. After adding nanoparticles (FS, TiO2), the H2 perme-bility is increased again, while the H2/N2 and H2/CH4 selectivitiesecreased as the filler content increased. However, compared tohe pure PMP membrane, the selectivities of crosslinked PMP/filleranocomposites are also increased significantly. Already at a fillerontent of ∼15 wt.%, the H2 permeability is restored as for theure PMP, while the selectivities are increased about three times.lthough the FFV should be approximately the same in both therosslinked and uncrosslinked PMP for the same permeability, therosslinking results most likely in a more restricted diffusion path

n the disrupted chain packing for the larger molecules (here N2 andH4) than for H2; hence an increased selectivity results. The databtained also show that at a fixed filler volume fraction, crosslinkedMP/FS nanocomposites have a little higher H2 permeability andlittle lower selectivities of H2/N2 and H2/CH4, compared to

brane Science 326 (2009) 285–292 291

ciftftfst

aPmvpndcmtitfiti

P(rfisp

4

Ptcomdai

FH

Fb2

5

ibPiaiCopn

tatnei

L. Shao et al. / Journal of Mem

rosslinked PMP/TiO2 nanocomposites. As already mentioned, thencrease in H2 permeability upon filling was attributed to increasedree volume in the filler-filled PMP nanocomposites as comparedo the unfilled PMP. The decrease in selectivity with increased fillerraction indicates that the addition of the nanoparticles (FS, TiO2) tohe PMP matrix polymer created larger gaps in the interconnectedree volume of the filled membranes, thereby increasing the diffu-ion coefficient of the larger molecule proportionately more thanhe smaller molecule.

Similarly, the O2 permeability of the crosslinked membrane islso initially reduced due to crosslinking compared to the pureMP membrane. After adding nanoparticles (FS, TiO2), the O2 per-eability is increased again, while the O2/N2 selectivity showed

ery little decrease with the filler content increased. Compared toure PMP membrane, the O2/N2 selectivity of crosslinked PMP/filleranocomposite is significantly increased (Fig. 6e). Similarly, theata obtained also show that at a fixed filler volume fraction,rosslinked PMP/FS nanocomposites have a slightly higher O2 per-eability, while the O2/N2 selectivity is almost same compared to

he crosslinked PMP/TiO2 nanocomposites. O2, as the most spher-cal molecule of the two, will move more easily than N2 throughhe crosslinked matrix. The maintained selectivity with increasedller content can be understood by relatively similar increase in

he diffusion coefficients for the two gases, as sorption will have nonfluence.

The distribution of fumed silica nanoparticles in crosslinkedMP/FS membrane is shown in the scanning electron microscopySEM) image presented in Fig. 7. It shows that these particles areelatively well dispersed in PMP, consistent with weak polymer-ller interactions. In addition, the SEM images also indicate thatome nanoparticles aggregate into clusters. The concentration ofarticles increases with nanoparticles loading.

.3. Membrane stability

The stability of the uncrosslinked and crosslinked PMP,MP/filler membranes stored in air were checked over a fairly longime. The results are shown in Fig. 8. The nitrogen permeabilities ofrosslinked PMP and PMP/filler membranes were almost constantver a period of 6 months. The uncrosslinked PMP and PMP/filler

embranes showed a large decrease in the nitrogen permeability

uring the same time. The permeability stability of crosslinked PMPnd crosslinked PMP/filler nanocomposite membranes for the gass clearly improved.

ig. 7. SEM photomicrograph of cross-section of crosslinked PMP + 2 wt.%FBAA + 25 wt.% FS membrane.

rmpuc

A

tDh

R

ig. 8. Temporal stability of uncrosslinked and crosslinked PMP, PMP/filler mem-ranes measured as nitrogen permeability; temperature: 35 ◦C, feed pressure:.0 bar.

. Conclusions

Crosslinking PMP with bis(aryl azide) has been shown toncrease the chemical and physical stability. Crosslinked PMP mem-rane is insoluble in good solvents of PMP. Compared to the pureMP membrane, the permeability of the crosslinked membrane

s initially reduced for all gases tested due to the crosslinking. Bydding nanoparticles (FS, TiO2), the permeability is again increased;ncreased selectivity is documented for the gas pairs O2/N2, H2/N2,O2/N2, CO2/CH4 and H2/CH4 using crosslinking and additionf nanoparticles. Crosslinking is successful in maintaining theermeability and selectivity of PMP membranes and PMP/filleranocomposites over time.

Permeability coefficients in PMP increase systematically withhe addition of nanoscale nonporous inorganic filler particles, suchs FS and TiO2. The increased permeability has been ascribedo increased free volume in the polymer matrix caused by theanoparticles disrupting the packing of the polymer chains. Thenhancement in permeability obtained for a given filler loadingncreases as the primary particle size of the filler decreases. Thisesult is likely related to smaller particles yielding larger poly-

er/particle interfacial area, since at a fixed volume fraction ofarticles there are a large number of small particles per unit vol-me of nanocomposite, which gives them a greater capacity disrupthain packing, thereby affecting transport property.

cknowledgements

The authors want to thank the Norwegian Research Council forhe financial support to the work. We also gratefully acknowledger. Keith Redford and Siren M. Neset from SINTEF Oslo for valuableelp with synthesis and analysis work.

eferences

[1] A. Morisato, I. Pinnau, J. Membr. Sci. 121 (1996) 243.[2] T.C. Merkel, B.D. Freeman, R.J. Spontak, Z. He, I. Pinnau, P. Meakin, A.J. Hill,

Science 296 (2002) 519.[3] Z. He, I. Pinnau, A. Morisato, Desalination 146 (2002) 11.[4] T.C. Merkel, B.D. Freeman, R.J. Spontak, Z. He, I. Pinnau, P. Meakin, A.J. Hill, Chem.

Mater. 15 (2003) 109.[5] K. Nagai, A. Sugawara, S. Kazama, B.D. Freeman, J. Polym. Sci. B: Polym. Phys.

42 (2004) 2407.[6] K. Nagai, S. Kanehashi, S. Tabei, T. Nakagawa, J. Membr. Sci. 251 (2005) 101.[7] W. Yave, S. Shishatskiy, V. Abetz, S. Matson, E. Litvinova, V. Khotimskiy, K.-V.

Peinemann, Macromol. Chem. Phys. 208 (22) (2007) 2412.[8] T.C. Merkel, Z. He, I. Pinnau, B.D. Freeman, P. Meakin, A.J. Hill, Macromolecules

36 (18) (2003) 6844.

2 brane

[

[

[

[[[[[[[

[[[

[

[[

[[

[

[

[

[[

[[

[

[

[[

[

[

[41] A. Reiser, H.M. Wagner, R. Marley, G. Bowes, Trans. Faraday Soc. 63 (1967)

92 L. Shao et al. / Journal of Mem

[9] T.C. Merkel, L.G. Toy, A.L. Andrady, H. Gracz, E.O. Stejskal, Macromolecules 36(353) (2003).

10] R.W. Baker, Membrane Technology and Application, 2nd ed., J. Wiley, New York,2004.

11] D.R. Paul, Y.P. Yampolskii, Polymeric Gas Separation Membranes, CRC Press,Boca Raton, FL, 1994.

12] S. Lin, T. Kai, B.D. Freeman, S. Kalakkunnath, D.S. Kalika, Macromolecules 38(2005) 8381.

13] C. Staudt-Bickel, W.J. Koros, J. Membr. Sci. 155 (1999) 145.14] H. Kita, T. Inada, K. Tanaka, K. Okamoto, J. Membr. Sci. 87 (1994) 139.15] K.K. Hsu, S. Nataraj, R.M. Thorogood, P.S. Puri, J. Membr. Sci. 79 (1993) 1.16] C.T. Wright, D.R. Paul, J. Membr. Sci. 124 (1997) 161.17] T.-S. Chung, M.L. Chng, K.P. Pramoda, Y. Xiao, Langmuir 20 (2004) 2966.18] W. DeForset, Photoresist Materials and Processes, McGraw Hill, New York, 1975.19] C. Grant Willson, “Organic resist materials - theory and chemistry,” in Intro-

duction to Microlithography, L. F. Thompson, C. G. Willson, M. J. Bowden, eds.,ACS Symposium Series 219 (American Chemical Society, Washington DC, 1983)p.87.

20] J. Jia, G.L. Baker, J. Polym. Sci. B: Polym. Phys. 36 (1998) 959.21] N. Toshima, Polymers for Gas Separation, VCH, Weinheim, 1992.22] S.A. Stern, Polymers for gas separations: the next decade, J. Membr. Sci. 94

(1994) 1.23] M. Mulder, Basic Principles of Membrane Technology, 2nd ed., Kluwer Academic

Publishers, Amsterdam, 1996.

24] K. Ghosal, B.D. Freeman, Polym. Adv. Technol. 5 (1994) 673.25] A. Bondi, Physical Properties of Molecular Crystals Liquids and Glasses, Wiley,

New York, 1968.26] D.W. van Krevelen, Properties of Polymers, Elsevier, Amsterdam, 1997.27] Y.P. Yampolskii, M.V. Motyakin, A.M. Wasserman, T. Masuda, M. Teraguchi, V.S.

Khotimskii, B.D. Freeman, Polymer 40 (1999) 1745.

[

[

Science 326 (2009) 285–292

28] D. Hofmann, M. Entrialgo-Castano, A. Lerbret, M. Heuchel, Y. Yampolskii, Macro-molecules 36 (2003) 8528.

29] D. Hofmann, M. Heuchel, Y. Yampolskii, V. Khotimskii, V. Shantarovich, Macro-molecules 35 (2002) 2129.

30] Y.P. Yampolskii, N.E. Kaliuzhnyi, S.G. Durgar’yan, Macromolecules 19 (1986)846.

31] R.J. Roe, J.J. Curro, Macromolecules 16 (1983) 428.32] G. Golemme, J.B. Nagy, A. Fonseca, C. Algieri, Y. Yampolskii, Polymer 44 (2003)

5039.33] Y. Kobayashi, K. Haraya, S. Hattori, T. Sasuga, Polymer 35 (1994) 925.34] Y.P. Yampolskii, V.P. Shantorovich, F.P. Chernyakovskii, A.I. Kornilov, N.A. Plate,

J. Appl. Polym. Sci. 47 (1993) 85.35] G. Consolati, I. Genco, M. Pegoraro, L. Zanderighi, J. Polym. Sci. B: Polym. Phys.

34 (1996) 357.36] V.S. Khotimsky, S.M. Matson, E.G. Litvinova, G.N. Bondarenko, A.I. Rebrov,

Polym. Sci. A 45 (8) (2003) 740 (Russia).37] I. Pinnau, A. Morisato, US Patent 5,707,423 (1998).38] C. Ling, M. Minato, P.M. Lahti, H. van Willigen, J. Am. Chem. Soc. 114 (1992)

9959.39] Organic Syntheses, Collective vol. 5, John Wiley and Sons, New York,

1973.40] T.X. Neenan, U. Kumar, T.M. Miller, Polym. Prep., ACS Div. Polym. Chem. 35 (1994)

391.

3162.42] K.C. O’Brien, W.J. Koros, T.A. Barbari, E.S. Sanders, J. Membr. Sci. 29 (1986)

229.43] J.A. Lie, Ph.D. Dissertation, Norwegian University of Science and Technology,

Norway, 2005.