Intermediate polymer to carbon gas separation membranes based on Matrimid PI

Journal of Membrane Science 238 (2004) 93–102

Intermediate polymer to carbon gas separation membranesbased on Matrimid PI

J.N. Barsemaa, S.D. Klijnstraa, J.H. Balstera, N.F.A. van der Vegtb,G.H. Koopsa,∗, M. Wesslinga

a Membrane Technology Group, Faculty of Science and Technology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlandsb Max-Planck-Institute for Polymer Research, P.O. Box 3148, 55021 Mainz, Germany

Received 24 September 2003; received in revised form 18 February 2004; accepted 15 March 2004

Abstract

Matrimid polyimide gas separation membranes were exposed to different heat treatments between 300 and 525◦C to investigate theintermediate structures that evolve between annealing and carbonization. The exposure time was either 5 or 30 min, while the atmospherewas N2. It was found, using TGA and FTIR, that below 425◦C no thermal decomposition takes place. At higher temperatures thermaldecomposition leads to the formation of intermediate structures as the process of carbonization starts. Gas permeation measurements withHe, N2, O2, CO2, and C3H6 show that the structure becomes more dense at treatments below theTg of the polymer, whereas above theTg aconcurrent formation of charge transfer complexes takes place. A steep increase of permeability was found at heat treatments above 475◦C,attributed to the thermal decomposition and the transition to carbon membranes. All heat-treated membranes showed a good resistance toplasticization by C3H6, although the permeability decreased, when compared to untreated membranes. Membranes treated at 475◦C for 30 minshowed no plasticization combined with sustained permeation rates.© 2004 Elsevier B.V. All rights reserved.

Keywords: Gas separation; Heat treatment; Plasticization; Intermediate carbon structures; Carbon membranes

1. Introduction

Over the last decades, polymeric membranes have provento operate successfully in industrial gas separations[1].To obtain membranes that combine high permeabilityand high selectivity together with thermal stability, newpolymers, so-called high-performance polymers, were de-veloped like polyimide (PI), poly(phenyl oxide) (PPO),poly(trimethylsilylpropyne) (PTMSP), and polytriazole.The commercially produced polyimide Matrimid 5218 hasshown promising properties when separating gas mixtureslike O2 and N2, CO2 and CH4, CO2 and N2, and C3H6 andC3H8 [2–4]. Combined with the ability to prepare asymmet-ric integrally skinned hollow fiber membranes with ultra-thintop layers, these membranes allow the preparation of gasseparation membrane modules for large-scale industrial

∗ Corresponding author. Tel.:+31-53-4894185; fax:+31-53-4894611.E-mail address: [email protected] (G.H. Koops).

applications[2,4]. Lab scale experiments with dense flatsheet membranes, however, have shown a major side effect.When the feed gas mixture contains condensable gases like,e.g. CO2 or C3H6 [4–7], a minimum is observed in thepermeability versus feed pressure curve followed by a steepincrease in permeability combined with a decrease in selec-tivity when exceeding a certain feed pressure. This effecthas been explained by the occurrence of plasticization of thepolymer above the plasticization pressure. The membranestructure swells and permeability of all components in thefeed increases significantly. Wessling et al.[8] have shownfor dense flat sheet membranes prepared from Matrimid5218 that the plasticization pressure decreases with filmthickness. This was recently confirmed by Barsema et al.[9]showing that asymmetrical integrally skinned membranesof P84 co-polyimide with thin top layers showed severeplasticization, whereas for their relatively thick dense flatsheet counterparts no plasticization effects were observed.

In literature, many articles can be found reporting dif-ferent procedures to reduce the extent of plasticization

0376-7388/$ – see front matter © 2004 Elsevier B.V. All rights reserved.doi:10.1016/j.memsci.2004.03.024

94 J.N. Barsema et al. / Journal of Membrane Science 238 (2004) 93–102

phenomena in glassy polymer membranes. Methods usedare: (1) heat treatment (e.g. annealing)[4], (2) thermally in-duced cross-linking[3], (3) chemical cross-linking[10], (4)inter-penetrating network formation[11], and (5) polymerblending[12,13].

A relatively recent development in gas membrane separa-tion is the preparation of carbon molecular sieve membranes.These membranes are generally produced by pyrolysis ofpolymeric precursors and can potentially achieve high selec-tivity while sustaining the permeability and more importantthey do not show any plasticization phenomena. However,these membranes have two distinct disadvantages. Firstly,they are expensive as compared to their polymeric counter-parts, but secondly and more important it is quite difficult toproduce large defect-free surface areas with sufficient me-chanical strength for commercial application.

Much research focuses on separate parts of the tempera-ture range: annealing around theTg, to prevent plasticizationor carbonization at temperatures well above 500◦C. Littlesystematic information exists for the intermediate region ofthe polymer-to-carbon transition. This paper discusses theintermediate structures that evolve in between annealing andcarbonization. These structures combine properties of bothpolymers and carbons. This work will focus on the effectsof heat treatment, extended considerably over theTg of thematerial, on the chemical composition, the molecular struc-ture, the permeation properties, and the plasticization phe-nomena using Matrimid 5218 polyimide membranes.

2. Background

Polymeric membranes are heat-treated by allowing themembrane to be exposed to a heated atmosphere. This at-mosphere can be either reacting (e.g. O2, air, or CO2), in-ert (e.g. N2, Ar), or vacuum. When a reacting atmosphereis applied the membrane is annealed, but thermochemicalcross-linking and decomposition can take place[3], depend-ing on the thermal resistance of the polymer at relativelylow temperatures. In inert or vacuum atmospheres the heattreatment of polymers can be classified into three processes:(1) annealing 100–400◦C [14], (2) pyrolysis forming carbon500–1000◦C [15], and (3) an intermediate state 400–500◦C,where decomposition starts but no carbon structure is formedyet.

During the annealing step the polymer membrane under-goes an accelerated physical ageing reducing permeabilityby relaxation of the frozen glassy structure. Concurrently, inthe case of polyimides, the formation of charge transfer com-plexes (CTC) can occur[4,14,16]. CTCs are weak, intra- andinter-molecular, bonds between the electron-rich aromaticring and the electron-deficient imide ring of the polyimide,formed by the donation of�-electrons. As the temperatureof the annealing process increases (most significantly abovethe Tg of the polymer), a higher mobility of the moleculesallows for a more dense (energetic more favorable) chain

packing and ultimately to a higher density of the polymer[17].

As the temperature is raised above the glass transitiontemperature also rejuvenation can occur counteracting theeffects of physical ageing, increasing the free volume of thepolymer.

If a polyimide membrane is subjected to temperaturesover 500◦C in an inert atmosphere, decomposition to car-bon takes place[18]. The weight of the polymer decreasesrapidly, while pyrolysis gases, like CO2, CO, H2, CH4, andN2, evolve from the polymer. The resulting material consistsof aromatic carbon with a turbostratic graphite or vitreouscarbon structure. These materials combine high permeabil-ity with molecular sieve properties and have shown to bepromising membrane materials[15].

In the temperature range 400–500◦C the polymer struc-ture is damaged and an intermediate material is formed.

3. Experimental section

3.1. Materials

As membrane material, a BTDA-AAPTMI polyimide(Matrimid 5218, Ciba Specialty Chemicals Corp.) was used.The chemical structure is shown inFig. 1. This polyimidehas aTg of 322◦C.

N-methyl-pyrrolidone (NMP), Merck 99%, was used asthe solvent. All gases used for heat treatment and gas per-meation experiments had a purity of at least 99.5%.

3.1.1. Membrane preparationAll membranes were prepared based on a solution of

13 wt.% Matrimid in NMP. The solutions were allowed todissolve overnight and were subsequently filtered over a25�m metal filter to remove non-dissolved residual mate-rial.

The solutions were cast on glass plates using a 250�mcasting knife and placed in a dry N2 box for 3 days, thenplaced in a N2 oven for a further 2 days at 150◦C to removemost of the solvent. During the drying step at 150◦C someannealing of the precursor membranes will take place. Theextent of annealing depends on the exact procedure followed.To allow for removal of any thermal and physical history,all membranes used for permeation and plasticization ex-periments were treated at 350◦C for 120 min. By increasingthe temperature above theTg of the polymer the increasedchain mobility in the rubber phase can lead to rejuvenation

CH3

O

O

N

C

C

O

C

CH 3

C

C

N

O

O H3C n

Fig. 1. Structure of Matrimid polyimide.

J.N. Barsema et al. / Journal of Membrane Science 238 (2004) 93–102 95

stainless steel grid sledge matrimid films

N2, in

oven

650571

heating element heating element

quartz glass

gases out

temperature controller

Fig. 2. A schematic overview of the furnace setup.

of the membrane structure. The obtained precursor film hada thickness of approximately 25�m.

3.1.2. Heat treatment procedureThe heat treatment is preformed using a Carbolite® TZF

12/100 High Temperature Tube Furnace, mounted with aEurotherm 2408 CP temperature controller.Fig. 2 gives aschematic overview of the furnace setup. The membrane isplaced in a quartz glass tube, using a stainless steel grid or aquartz glass as sledge. The atmosphere in the quartz tube dur-ing pyrolysis was N2, with a flow rate of 10 cm3/min. Heattreatment procedures allowed the temperature to rise formroom temperature to 150◦C with a heating rate of 50◦C/min.Subsequently this temperature was held for 15 min to re-move any adsorbed water. From 150 until 350◦C the tem-perature was raised with a rate of 5◦C/min, thereafter thetemperature increase was controlled at 1◦C/min.

After reaching the end temperature, the membranes werekept at this temperature for either 5 or 30 min, before beingquenched to room temperature in an external stainless steeldouble-hulled cooler containing a nitrogen atmosphere. Themembranes were cooled by tap water flowing through theouter hull of the cooler.

3.2. Analysis

3.2.1. Weight lossThe weight loss during heat treatment was determined

using thermogravimetric analysis (TGA). The TGA experi-ments were preformed using a Perkin-Elmer TGA 7 with aN2 atmosphere and flow rates of 20 cm3/min. Heating routeswere based on heat treatment procedures.

3.2.2. DensityFor the density measurements, an AccuPyc 1330 Pyc-

nometer with a 0.1 cm3 sample insert was used. The Accu-Pyc measures the amount of displaced gas. The pressuresobserved upon filling the sample chamber and then discharg-ing it into a second empty chamber allow computation ofthe sample solid-phase volume. Gas molecules rapidly fill

the tiniest pores of the sample; only the truly solid phase ofthe sample displaces the gas.

SF6 was chosen as pressure gas, because of its largemolecule size and its inert nature, in order to prevent adsorp-tion of the pressure gas inside (the pores of) the membranesas much as possible.

3.2.3. Chemical structureThe changes in the chemical structure during the heat

treatments were followed using Fourier transformed infraredspectroscopy (FTIR). Heat-treated Matrimid powders weremixed with KBr powder at low concentrations and thereaftercompressed into discs. These discs were placed into a Bio-rad Digilab FTS60 FTIR Spectrometer. The IR absorptionspectra were measured at room temperature from 4000 to500 cm−1 with a spectral resolution of 8 cm−1 and averagedover 64 scans.

X-ray photoelectron spectroscopy (XPS) was performed,using a Quantum 2000 Scanning Esca Probe of PhysicalElectronics, on samples at different stages in the heat treat-ment process. The experiments were carried out on theetched (500 eV Ar) surface of the heat-treated samples toremove surface contamination, if present.

3.2.4. Gas permeationTo elucidate the effect of the heat treatment on the per-

meability and selectivity of the membranes, the pure gasHe, CO2, O2, N2, and C3H6 permeability was determined.If the membranes were brittle, they were glued into metaldiscs, using an Araldit two-component adhesive, to preventcracking of the membranes by the rubber ring seals. Thegas permeation experiments were done at the temperatureof 25◦C, using a variable permeate pressure setup. In theexperimental setup, a vacuum was applied on the permeateside, whereas the other side was brought into contact withthe feed gases. All measurements were taken after the fluxesbecame constant in time. A more extensive description ofthe determination of both permeability and selectivity canbe found elsewhere[9]. The reported results are based onmeasurements performed on two or more membranes.


3.2.5. PlasticizationBesides permeation experiments using C3H6 as plasti-

cizing agent, permeation experiments were conducted onpre-plasticized membranes. The membranes were first plas-ticized, after heat treatment, at 56 bar of CO2 in a pressurechamber at room temperature for 120 min and subsequentlythe He permeability was followed in the time.

3.2.6. CO2 sorptionThe CO2 solubility in a number of heat-treated samples,

at 25◦C, was determined using a Rubotherm magnetic sus-pension balance operated by MessPro software. This balancehas an accuracy of 1�g and a reproducibility of 2�g. Thesamples were evacuated to a pressure below 10 Pa, beforethe CO2 was applied. The obtained data were corrected forbuoyancy effects, which were experienced because of theincreased gas density at higher pressures. Besides the solu-bility of CO2, the average diffusion coefficient of CO2 wasdetermined. In case of CO2 in a glassy polymer, both coeffi-cients are concentration-dependent. When the permeability(Pi) is defined as the flux (Ji) normalized for the pressuredifference across the membrane (pfeed,i − ppermeate,i) andmembrane thickness (l), one can write down an equation forthe permeability:

Pi = Ji

(pfeed,i − ppermeate,i)/ l= −D(ci)∂ci/∂x

(pfeed,i − ppermeate,i)/ l,

(1a)

and

Ji = −D(ci)∂ci

∂x, (1b)

wherePi is the permeability (barrer),Ji the flux (cm3 (STP)cm−2 s−1), andD(ci) a local concentration-dependent diffu-sion coefficient of a penetrant at any arbitrary point betweenthe feed and permeate side of the membrane (cm2 s−1) with∂ci/∂x the local concentration gradient at the same point inthe membrane[19].

Because the steady-state permeability is a constant forfixed feed and permeate conditions, the product ofD(ci) and∂ci/∂x is constant at each point in a membrane. Integrationof Eq. (1a)over the membrane thickness (x = 0 → x = l)gives

(pfeed,i − ppermeate,i)Pi

l

∫ l

0dx = −

∫ cfeed,i

cpermeate,i

D(ci) dci, (2)

which results inEq. (3) for negligible permeate pressure

Pi = 1

pfeed,i

∫ cfeed,i

0D(ci) dci. (3)

Multiplying the numerator and denominator ofEq. (3) bycfeed,i gives for the permeability (Pi):

Pi =(∫ cfeed,i

0

D(ci)

cfeed,idci

) (cfeed,i

pfeed,i

)= D̄iSi (4)

whereD̄i is theaverage diffusion. If the permeability coef-ficient and solubility coefficient are determined at the samepressure and temperature an average diffusion coefficientcan be calculated from the ratio of the permeability and av-erage solubility coefficients.

4. Results and discussion

4.1. Thermal degradation followed by TGA

The membranes and Matrimid samples prepared in thisstudy are labeled by their end temperature and residencetime, so the membrane treated at 425◦C for 30 min is labeled425/30.

Fig. 3 shows the thermally induced weight loss ofMatrimid versus time and temperature as measured by TGA.Also the first derivative, the decomposition rate (ratedec,wt.%/min) is depicted.

Until 350◦C the weight loss is minimal and should beattributed to residual solvent and adsorbed water in thesample. When looking atFig. 3 the sample weight remainsalmost constant (ratedec ≈ 0 wt.%/min) from 350◦C up toapproximately 450◦C. At this temperature, ratedec startsto increase, until a maximum is reached at approximately520◦C (ratedec = −0.4 wt.%/min). A second maximum oc-curs at approximately 610◦C (ratedec = −0.175 wt.%/min).Above 610◦C ratedec decreases, although at 650◦C theweight is still decreasing significantly. According to Fig. 4from Inagaki et al.[18]:

• up to 450◦C, relative low weight loss caused by evapora-tion of residual water and solvents;

• 450–650◦C, rapid decrease of weight caused by degrada-tion of the polymer, while H2, CO, CO2, and CH4 evolvefrom the sample;

0 50 100 150 200 250 300 35060

65

70

75

80

85

90

95

100

Temperature [0C]650600550500450400350150

Dec

ompo

sitio

n ra

te [w

t%/m

in]

Rel

eativ

e sa

mpl

e w

eigh

t [w

t%]

Time [min]

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

Fig. 3. Relative sample weight vs. time and temperature and decompositionrate vs. time.


untr

eate

d

300/

5

300/

30

350/

5

350/

30

425/

5

425/

30

475/

5

475/

30

525/

5

525/

30

575/

5

625/

5

650/

560

65

70

75

80

85

90

95

100

105

Fin

al w

eigh

t yie

ld [w

t%]

Sample id.

Fig. 4. Final weight yield after heat treatment.

• 650◦C and higher, slow decrease in weight, caused by theevolving of residual non-elementary carbon components,primarily N2.

FromFig. 3, it seems that up to 450◦C no decompositiontakes place, however different TGA measurements suggestthat the first decomposition starts at 425◦C. The weight de-crease of a sample held for 3200 min at 425◦C was 10%.This weight loss cannot be explained by water and solventevaporation only. Already at 425◦C, decomposition reac-tions take place. However, the reaction rate is extremely lowat 425◦C. From approximately 450◦C, the rate of the firstdecomposition reaction is high enough to be detected forsamples heated at a rate of 1◦C/min.

In Fig. 4, the weight loss of the Matrimid samples afterdifferent heat treatments is followed for samples heat-treatedup to 650◦C. There is a gradual decrease in final weightfor samples treated up to 475◦C for 30 min. Although thesample 475/30 clearly shows a lower final weight than thesamples treated at lower temperatures, the weight loss is rel-atively small. Samples treated at temperatures of 525◦C andhigher show an increasing weight loss as they are carbonized.

4.2. Thermal degradation followed by FTIR

The chemical changes taking place going through thetransformation from polymer to carbon can be observedby FTIR. Fig. 5 shows a FTIR spectrum of the pretreatedMatrimid (350/120).Table 1 gives an overview of bandassignments and wave numbers for the IR spectrum ofMatrimid powder, heat-treated for 120 min at 350◦C.

Fig. 6 shows the evolution of the chemical structure, vi-sualized by FTIR spectra.

It becomes obvious fromFig. 6 that the absorptionspectrum changes with the intensity of the heat treatment

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

500150025003500

Wavenumber (cm-1)

Ab

sorb

anc

e

CH3

O

O

NC

COC

3

CH3

CO

2 4,4’

Nz

5,6,7 O H3C

1

2b

5

63 744’2a1

Fig. 5. IR spectrum of Matrimid powder, heat-treated for 120 min at350◦C.

(350/120 to 475/780). Generally, the absorption intensitydecreases with a more intense heat treatment, as the concen-tration of the functional groups, responsible for a specificpeak, decreases.

When comparing 425/30 with 350/120, the ratio of peakintensity of the anti-symmetric stretch peak of the imidecarbonyl group (1725 cm−1) to the stretch peak of the ben-zophenone carbonyl group (1673 cm−1) decreases from 1.8to 1.4. The peaks are pointed out inFig. 6 for 350/120 and425/30 by arrows. This is an indication that the imide car-bonyl group concentration is decreasing compared to thebenzophenone carbonyl group, pointing towards imide ringcleavage, resulting in destruction of the imide C=O groups.A further indication for this destruction of the imide C=Ogroups is obtained by XPS.Table 2shows the atomic com-position of the membranes treated at 350/5, 425/5, 475/5,and 475/30. As the membranes are exposed to temperaturesof 425◦C and higher we see a significant reduction of theatomic oxygen content, from 12.9 to 9.3 at.%. Whereas theIR spectra inFig. 6for the Matrimid sample treated for 5 minat 475◦C still shows comparable absorption intensities when

Table 1Overview of band assignments and wave numbers for IR spectrum ofMatrimid powder, heat-treated for 120 min at 350◦C

Number ν (cm−1) Band assignment

1 2960–2860 ν (C–H) stretch of methyl groups2a 1779 ν (C=O) symmetric stretch2b 1725 ν (C=O) anti-symmetric stretch3 1673 ν (C=O) stretch of benzophenone

carbonyl4 and 4′ 1512 and 1488 ν (C=C) aromatic stretching5 1374 ν (CNC) axial stretch6 1096 ν (CNC) transverse stretch7 717 ν (CNC) out-of-plane bending


350/120

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

500150025003500

Wavenumber (cm-1)

Ab

sorb

ance

425/5

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

500150025003500

Wavenumber (cm-1)

Ab

sorb

ance

425/30

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

500150025003500

Wavenumber (cm-1)

Ab

sorb

ance

475/5

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

500150025003500

Wavenumber (cm-1)

Ab

sorb

ance

475/30

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

500150025003500

Wavenumber (cm-1)

Ab

sorb

ance

475/780

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

500150025003500

Wavenumber (cm-1)

Ab

sorb

ance

Fig. 6. IR spectra of Matrimid powder after different heat treatments (350–475◦C).

Table 2Chemical composition as determined by XPS, for membranes 350/5,425/5, 475/5, and 475/30

Membrane C (at.%) O (at.%) N (at.%)

Theoretical 82.9 12.2 4.9350/5 82.6 12.9 4.5425/5 84.4 11.1 4.5475/5 85.2 9.4 5.4475/30 85.8 9.3 4.9

compared to membranes treated at 425◦C, the membranetreated for 30 min at 475◦C shows lower absorption inten-sities. The sample treated for 780 min at 475◦C shows evenlower absorption intensities. After 475/780 still some imiderings are present in the Matrimid sample, the reaction hasnot ended entirely.

4.3. Density measurements

The density was determined of Matrimid membranes (pre-treated for 120 min at 350◦C) heat-treated between 350 and


350/

120

425/

5

425/

30

475/

5

475/

30

500/

30

525/

30

1.22

1.24

1.26

1.28

1.30

1.32

Den

sity

[g/c

m3 ]

Sample id.

Fig. 7. Density of Matrimid membranes after different heat treatments.

525◦C. The results are shown inFig. 7; the density increasesslightly for membranes treated at 425◦C, as compared tothe pretreated membranes. Thermal annealing results in ahigher concentration of CTCs, which in turn results in amore dense chain packing and, although minimal, a higherdensity. As the Matrimid membranes are treated at temper-atures of 475◦C and higher, the density increases signifi-cantly, due to the onset of carbonization.

The membranes treated up to 425◦C are dense poly-meric membranes; they have lower densities and possesspolymeric properties. The membranes treated at 500◦C andhigher are carbon membranes, they have a higher density(1.5–2.3 g cm−3 [20]) and possess carbon properties. Themembranes treated at 475◦C show densities in between thepolymeric and the carbon membranes. They cannot be de-scribed only in terms of polymer membrane or carbon mem-brane.

4.4. Permeation properties

The effect of heat treatment on the permeation proper-ties is shown inFigs. 8–10, displaying the permeability ofN2, O2, and CO2 for different heat treatments at 2, 3, and5 bar feed pressure. The lines are to guide the eye. All mem-branes have been pretreated at 350◦C for 120 min to obtainmembranes with the same thermal history. It is clear thatalthough the heat treatment has a large effect on the perme-ability, the effect is comparable for all feed pressures, mostclearly shown inFig. 10, indicating that the shape of the sol-ubility curve is not significantly altered by the applied heattreatment. We will discuss this in a later section.

For N2 and O2 (Figs. 8 and 9, respectively) the effect ofthe heat treatment is similar. The 300/30 membrane showsa depression of the permeability caused by annealing of thestructure when treated at a temperature below theTg of thepolymer. Prolonged treatments (30 min versus 5 min) at 350and 425◦C lead to a depression of the diffusion coefficient,

300/

30

350/

5

350/

30

425/

5

425/

30

475/

5

475/

30

0.20

0.25

0.30 2 bar 3 bar 5 bar

PN

2 [bar

rer]

Sample id.

Fig. 8. N2 permeabilities at different pressures for different membranestreated up to 475◦C: (�) 2 bar; (�) 3 bar; (�) 5 bar.T = 25◦C.

since the solubility is not effected by the heat treatment(seeFig. 11), and therefore the permeability decreases, mostlikely caused by the formation of CTCs, through the in-creased chain mobility. As the membranes are exposed to atemperature of 475◦C we observe an initial decrease of thepermeability at short treatment times (475/5), but as decom-position reactions commence the permeability shows a steepincrease (475/30). Further exposure to temperatures higherthan 475◦C lead to a continuous increase in permeability, aswe will show later. The permeability of CO2 (Fig. 10) showssignificantly different behavior when compared to N2 andO2. The membrane 300/30 shows a decrease of permeabil-ity, but not to the extent seen inFigs. 8 and 9. More surpris-ing is the increase in CO2 permeability of the 425◦C (425/5and 425/30)-treated membranes compared to the membrane350/30.Fig. 11 shows that the absolute solubility and the

300/

30

350/

5

350/

30

425/

5

425/

30

475/

5

475/

30

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

2.1

2.2

2 bar 3 bar 5 bar

PO

2 [bar

rer]

Sample id.

Fig. 9. O2 permeabilities at different pressures for different membranestreated up to 475◦C: (�) 2 bar; (�) 3 bar; (�) 5 bar.T = 25◦C.


300/

30

350/

5

350/

30

425/

5

425/

30

475/

5

475/

30

5

6

7

8

9

10

11

2 bar 3 bar 5 bar

PC

O2 [b

arre

r]

Sample id.

Fig. 10. CO2 permeabilities at different pressures for different membranestreated up to 475◦C: (�) 2 bar; (�) 3 bar; (�) 5 bar.T = 25◦C.

shape of the sorption curve of CO2 are hardly affected bythe heat treatment. However, comparing the average diffu-sion coefficients for samples treated at 350/5, 350/30, and425/5 determined fromEq. (4), Fig. 12 clearly shows thatthe diffusion coefficient increases after reaching a minimumat 350/30, for all feed pressures, resulting in a relative highpermeability of CO2 through the 425/5 membrane as com-pared to the membrane treated at 350/30. A further indica-tion for the different polymer–penetrant interaction of CO2as compared to O2 and N2 is seen inFig. 13. Here we splitthe permeation results, obtained at a feed pressure of 2 bar,for the different gasses for membranes treated at the endtemperature for 5 and 30 min. We can clearly see two dif-ferent trends. The membranes treated for 5 min show a con-stant decrease of the permeability for all gasses. The curvesfor the membranes treated for 30 min, however, do not showsuch an unambiguous trend. All membranes show a sharp

0 1 2 3 4 5 6 7 8 9 10 110

5

10

20

30

40

CO2

N2

O2

350/5 350/30 425/5

c [c

m3 (S

TP

)cm

-3 p

olym

er]

Pressure [bar]

Fig. 11. CO2 concentration in heat-treated Matrimid membranes ((�)350/5, (�) 350/30, and (�) 425/5) vs. the pressure.T = 25◦C.

2 3 4 5 60

2

4

6

8

10

12

14

350/5

425/5

350/30

DC

O2*1

09 [cm

2 s-1]

Pressure [bar]

Fig. 12. Average diffusion coefficients of CO2 in heat-treated Matrimidmembranes vs. the pressure ((�) 350/5, (�) 350/30, and (�) 425/5).T = 25◦C.

increase of their N2, O2, and CO2 permeability when treatedat 475◦C. This is caused by the onset of thermal decomposi-tion. The trend in the CO2 permeability differs substantiallyfrom the trend seen for O2 and N2. The CO2 permeabilitydecreases going from heat treatment temperatures of 300◦Cto 350◦C, opposite to the trend seen for O2 and N2. Sec-ondly the trends for O2 and N2 show a distinct minimum inthe permeability at a heat treatment temperature of 425◦C,absent in the permeability trend of CO2.

Table 3shows the permeability and the selectivity of anumber of membranes heat-treated at temperatures up to525◦C. It is clear that with increasing temperature of treat-ment the permeability that increases exponentially, however,leads to a decrease in selectivity. This can be attributed tothe increased thermal decomposition as the first stages ofcarbonization have commenced, also visibly observed by theblackening of the membranes during exposure. A side effectof the heat treatment at 475◦C and higher is that the mem-branes become increasingly brittle and difficult to handle.The O2 permeability reported inTable 3for the membranestreated between 300 and 350◦C is relatively high, althoughsimilar values have been reported[4]; this may be attributedto the rejuvenation taking place during the pretreatment.

It has been found[4] that heat treatments beneficiallyaffect the plasticization resistance of polymeric films. To

Table 3Permeability and selectivity for heat-treated Matrimid membranes (2 bar,20◦C)

Membrane PN2

(barrer)PO2

(barrer)PCO2

(barrer)α (O2/N2)(–)

α (CO2/N2)(–)

300/30 0.20 1.64 7.99 8.20 39.93350/30 0.24 1.80 7.37 7.51 30.71425/30 0.20 1.60 7.87 8.01 39.36475/30 0.31 2.14 10.13 6.91 32.70500/30 1.18 6.67 32.90 5.65 27.88525/30 8.74 40.2 190 4.60 21.75


Fig. 13. N2, O2, and CO2 permeability for membranes heat-treated for5 min (closed symbols) and 30 min (open symbols) vs. the heat treatmentend temperature.

investigate the effects of the heat treatment on the plasti-cization behavior permeation experiments were conductedusing propylene as the feed gas.Fig. 14 shows the perme-ability of propylene versus the feed pressure for the subse-quent heat treatments. The untreated membrane inFig. 14was not pretreated at 350◦C. It is clear that the plasticiza-tion is successfully suppressed by applying a heat treatmentbetween 300 and 475◦C. An unfavorable effect is the re-duction of permeability over 50%. However, the membrane475/30 shows no plasticization effects combined with re-tained permeability. The thermally induced changes to themolecular structure provide a resistance to the swelling ofthe membrane by propylene.

To investigate the resistance of the membranes to plas-ticization by CO2, the membranes were exposed to severe

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.50.00

0.05

0.10

0.15

0.20

0.25

0.30

475/30

425/30

350/5

300/30

Untreated

P C

3H6

[bar

rer]

Pressure [bar]

Fig. 14. Propylene permeabilities at different pressures for different mem-branes treated up to 475◦C. T = 25◦C.

plasticization conditions, 56 bar CO2 for 2 h. Immediatelyafter conditioning, the membranes were transferred from thepressure cell to the permeation setup.Fig. 15 shows theHe permeability of the membranes immediately after con-ditioning with CO2 (P0) together with the He permeabilityafter the membranes have reached equilibrium permeability(Pend). The effect of the heat treatments on the He perme-ability is comparable to that of N2 except for the low Hepermeability of the 475/30 membrane. However, these arepre-swollen membranes and this low permeability can beexplained by high resistance to plasticization of the 475/30membrane. The ratio ofPend over P0 is a measure for theability of the membrane to resist swelling and is shownin Fig. 15. The resistance to swelling is stable for mem-branes treated till 350◦C, then goes through a significantminimum at 425/5, and finally shows a increasing resistance

not t

reat

ed

300/

5

300/

30

350/

5

350/

30

425/

5

425/

30

475/

5

475/

30

15

20

25

30

35 P0 Pend

PH

e [b

arre

r]

Sample id.

0.86

0.88

0.90

0.92

0.94

0.96

Pend /P

0 [-]

Pend/P0

Fig. 15. He permeability ((�) P0, (�) Pend) of pre-plasticized Matrimidmembranes and plasticization resistance (⊗). T = 25◦C.


to plasticization with increasing temperature of the heattreatment.

5. Conclusions

In this experimental study we show that intermediatepre-carbonization gas separation membranes based on poly-imides, possess challenging properties. We have shown thatheat treatment of Matrimid flat sheet dense membranes, inan inert atmosphere, can alter the membranes properties aswell as the molecular structure. Below 425◦C the effectof exposure to a heated atmosphere is restricted to anneal-ing and the formation of CTCs. Prolonged (30 min versus5 min) exposure to a temperature of 425◦C leads to thermaldecomposition of the polymer chains. However, the extentof this decomposition is relatively small. TGA experimentsshow a weight loss with a maximum of 3.6 wt.% (475◦C for30 min). FTIR measurements show clearly that the chemicalstructure is altered beyond a treatment of 425/30.

The permeability of non-condensable gases (N2, O2) weremeasured at different pressures (2, 3, and 5 bar) and found tobe depressed by heat treatments below theTg of the polymer.Above theTg of the polymer, increasing formation of CTCsand resulting densification of the polymer structure leads toa gradual decrease of the permeability. A significant increasein permeability was found if the membranes are exposed totemperatures above 475◦C, caused by the onset of thermaldecomposition.

The permeability curve of CO2 presented versus the heattreatment conditions is similar to that of N2 and O2 witha distinct deviation for the more dense membranes treatedat 300 and 425◦C. This is attributed to the higher diffusioncoefficient of CO2 in these membrane structures.

Plasticization by propylene is successfully suppressedby heat treatments over 300◦C; however, a significant de-crease of the permeability is observed. Heat treatment at atemperature of 475◦C not only suppresses the plasticiza-tion, but shows sustained permeation rates compared tountreated membranes. The ability to resist the swelling by acondensable gas was also shown by measuring the He per-meability through pre-conditioned membranes (56 bar CO2,2 h), showing good resistance to swelling for membranestreated over 425◦C (30 min). From the different experi-ments conducted in this study it becomes obvious that themembranes treated at 425◦C for 5 min show a divergent be-havior as compared to differently treated membranes. BothTGA and FTIR show that the onset of thermal degradationis located at 425◦C. Although the density and CO2 sorp-tion remain unchanged, the CO2 permeation and swellingexperiments show an increased affinity for CO2; thismakes these membranes an interesting subject for furtherinvestigation.

References

[1] W.J. Koros, G.K. Fleming, Membrane-based gas separation, J.Membr. Sci. 83 (1993) 1–80.

[2] D.T. Clausi, W.J. Koros, Formation of defect-free polyimide hollowfiber membranes for gas separations, J. Membr. Sci. 167 (2000) 79–89.

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

[4] J.J. Krol, M. Boerrigter, G.H. Koops, Polyimide hollow fiber gasseparation membranes: preparation and the suppression of plasticiza-tion in propane/propylene environments, J. Membr. Sci. 184 (2001)275–286.

[5] A. Bos, I.G.M. Punt, M. Wessling, H. Strathmann, CO2-inducedplasticization phenomena in glassy polymers, J. Membr. Sci. 155(1999) 67–78.

[6] M. Wessling, I. Huisman, T. van der Boomgaard, C.A. Smolders,Dilation kinetics of glassy, aromatic polyimides induced by carbon-dioxide sorption, J. Polym. Sci. B Polym. Phys. 33 (1995) 1371–1384.

[7] W.J. Koros, M.R. Coleman, D.R.B. Walker, Controlled permeabilitypolymer membranes, Annu. Rev. Mater. Sci. 22 (1992) 47–89.

[8] M. Wessling, M.L. Lopez, H. Strathmann, Accelerated plasticizationof thin-film composite membranes used in gas separation, Sep. Purif.Technol. 24 (2001) 223–233.

[9] J.N. Barsema, G.C. Kapantaidakis, N.F.A. van der Vegt, G.H. Koops,M. Wessling, Preparation and characterization of highly selectivedense and hollow fiber asymmetric membranes based on BTDA-TDI/MDI co-polyimide, J. Membr. Sci. 216 (2003) 195–205.

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

[11] A. Bos, I.G.M. Punt, M. Wessling, H. Strathmann, Suppression ofCO2-plasticization by semiinterpenetrating polymer network forma-tion, J. Polym. Sci. B Polym. Phys. 36 (1998) 1547–1556.

[12] A. Bos, I. Punt, H. Strathmann, M. Wessling, Suppression of gasseparation membrane plasticization by homogeneous polymer blend-ing, AIChE J. 47 (2001) 1088–1093.

[13] G.C. Kapantaidakis, S.P. Kaldis, X.S. Dabou, G.P. Sakellaropou-los, Gas permeation through PSF-PI miscible blend membranes, J.Membr. Sci. 110 (1996) 239–247.

[14] H. Kawakami, M. Mikawa, S. Nagaoka, Gas transport properties inthermally cured aromatic polyimide membranes, J. Membr. Sci. 118(1996) 223–230.

[15] A.B. Fuertes, D.M. Nevskaia, T.A. Centeno, Carbon composite mem-branes from Matrimid (R) and Kapton (R) polyimides for gas sepa-ration, Micropor. Mesopor. Mater. 33 (1999) 115–125.

[16] F.J. Dinan, W.T. Schwartz, R.A. Wolfe, D.S. Hojnicki, T. Stclair, J.R.Pratt, Solid-state C-13-NMR spectral evidence for charge-transfercomplex-formation in aromatic diimides and dianhydrides, J. Polym.Sci. Pol. Chem. 30 (1992) 111–118.

[17] T. Hasegawa, K. Horie, Photophysics, photochemistry, and opticalproperties of polyimides, Prog. Polym. Sci. 26 (2001) 259–335.

[18] M. Inagaki, T. Ibuki, T. Takeichi, Carbonization behavior of poly-imide films with various chemical structures, J. Appl. Polym. Sci.44 (1992) 521–525.

[19] W.J. Koros, R.T. Chern, Separation of gaseous mixtures using poly-mer membranes, in: R.W. Rousseau (Ed.), Handbook of SeparationProcess Technology, Wiley, New York, 1987, pp. 862–953.

[20] H.O. Pierson, Handbook of Carbon, Graphite, Diamond, andFullerenes: Properties, Processing and Applications, 1st ed., NoyesPublications, Park Ridge, 1993.

Documents

Intermediate polymer to carbon gas separation membranes based on Matrimid PI