Poly(amide-imide)/TiO2 nano-composite gas separation membranes: Fabrication and characterization

jou

of

MEMBRANE SCIENCE ELSEVIER

Journal of Membrane Science 135 (1997) 65-79

Poly(amide-imide)/TiO2 nano-composite gas separation membranes" Fabrication and characterization a*

Q. Hu u, E. Marand ' , S. Dhingra a, D. Fritsch b, J. Wen a, G.

Wilkes a

a Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061, USA b GKSS Research Center, Max-Planck Strasse, D-21502 Geesthacht, Germany

Received 17 March 1997; receivedin revised form 25 April 1997; accepted9 May 1997

Abstract

Nano-composite membranes based on a fluorinated poly(amide-imide) and ' r i o 2 w e r e fabricated by a sol-gel method. Permeability data for gases such as 02, N2, CO2, H2 and CH4 were collected as a function of pressure and temperature. With the exception of CO2 and H2, all other gases exhibited higher activation energies for the nano-composite membrane when compared with the pure poly(amide-mide), consistent with the picture of a more rigid or denser structure as suggested by the physical characterization data. The decrease in the activation energy for permeation in the case of CO2 and H2 has been attributed to specific interactions of these gases with the TiO2 domains. Significant improvements in permselectivies of the poly(amide-imide) membrane have been observed in view of the volume percentage of the TiO2 incorporated into the polymer matrix.

Keywords." Composite membranes; Polymer/ceramic membranes; Sol-gel; Polymer/TiO2; Gas separations; Hybrid materials

1. Introduction

Due to the wide range of physical properties and structural characteristics, organic-inorganic hybrid materials may offer new possibilities to those gas separation applications, where purely organic polymers have found limitations. Hybrid materials can be easily fabricated by sol-gel methods which entail a low temperature inorganic polymerization process, permitting tailored modification of polymer matrix properties [1-11] or the synthesis of entirely new organic-inorganic network materials [l,2,10-13]. In their final form these materials can take on the desired *Corresponding author. 0376-7388/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved. PII S0376-7388(97)00 120-8

geometries of membranes such as free standing films, thin films cast onto porous supports and potentially hollow fibers. Hybrid organic-inorganic materials have become an area of intense research, focusing on applications such as optical and electronic materials [14-16], biochemical sensors [11], catalysts [17], adsorbents [11] and abrasive-resistant coatings [18,19l. Only recently, have permeability studies of certain hybrid material membranes appeared in the literature [12,13,20]. For example, polyimide-silica hybrid films, investigated by Schrotter et el. [13], have exhibited superior permeabilities and selectivity to He/CO2 gases, compared to pure polyimides, although similar hybrid materials studied by Kita et al. [20] have demonstrated no significant improvement in

66

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

the separation properties. Recent characterization studies by Smaihi et al. [12] of hybrid organicinorganic materials, prepared by co-hydrolysis of phenyltrimethoxysilane or diphenyldimethoxysilane with tetramethoxysilane, have shown promising selectivities for CO2/N2 and He/N2 gas pairs especially at high organoalkoxide proportions, although the low Tg values associated with some of the materials would preclude their use in practical applications. Hybrid organic-inorganic materials can be formed utilizing a sol-gel process via a number of approaches. One method involves treating an orthosilicate directly with the organic polymer or an oligomer which contains functional groups capable of a cross-reaction with the inorganic oxide, thus providing connectivity between the organic phase and the inorganic network [1,2,10-13]. Alternatively, alkoxysilane monomers containing a polymerizable organic moiety covalently attached to the silicon can be precipitated, followed by polymerization of the organic moiety [11 ]. Finally, in situ polymerization of the alkoxide within a swollen polymer network can be used to form nano- or microcomposite materials without covalent crosslinks [4-8]. What is interesting is that a high level of mixing can be achieved in these latter systems when there are specific interactions between the polymer and the inorganic oxide. Polymers, which contain functional groups such as carbonyls, hydroxyls or ether oxygens, can form hydrogen bonds with the inorganic network [4,5,8]. It is this last method that we are particularly interested in, because it can be employed to directly modify commercially available polymer systems. Furthermore, the synergistic incorporation of an inorganic component within a polymer matrix is known to reduce segmental mobility of the polymer chains and to inhibit chain packing [5]. In a recent work, Moaddeb and Koros [21] have indeed shown that the incorporation of colloidal silica into thin polyimide, polycarbonate and polysulfone films can significantly enhance the gas separation properties of these materials. Since it has been well documented that high performance polymeric membranes should have molecular structural features such as hindered segmental and subsegmental mobility and inhibited segmental packing [22-24] the hybrid organic-inorganic materials may represent an alternative to synthetic modification.

The in situ formation of an alkoxide network within a polymer matrix is governed by several parameters. The composite system is prepared by co-dissolving its precursor, a tetra-functional orthosilicate ester or an orthotitanate ester with the polymer in a common solvent. A small amount of catalyst is added to the solution to catalyze the sol-gel reactions, which consists of a primary hydrolysis reaction of the alkoxide to an alcohol and two condensation reactions involving either two alcohol groups or an ester reacting with an alcohol functional group. These reactions are concurrent and their relative rates are governed by pH, solvent, water-to-alkoxide ratio, concentration, type of catalyst and temperature [11]. In the case of polymer matrices, whose glass transition temperature is above ambient temperature, the final structure of-the inorganic network and morphology of the composite is further influenced by the relative rates of vitrification and the rate of inorganic network formation. In particular, pronounced differences in structure can occur, depending on whether the processing is carried out under acidic or basic conditions [11,25]. Base-catalyzed systems with a high water content, tend to consist of highly branched non-interpenetrating clusters, while acid-catalyzed systems with low water content, have linear or random branches. This is because under acidic conditions, the hydrolysis reaction is more rapid than either of the two condensation reactions [25]. For reasons given below, we have chosen a poly(amide-imide) as the polymer matrix material. Aromatic poly(amide-imide)s bring together both superior mechanical properties typically associated with polyamides and the high thermal stability, solvent resistance and high gas permeability and permselectivity characteristic of polyimides [22-24,26]. In particular, the polyamide unit can facilitate hydrogen bonding to other components having either a proton donor or a proton acceptor group. Comparative experiments with tetraethoxy silane (TEOS) and tetraethoxy titanate, both incorporated in a poly(amide-imide) matrix, have shown that the titanate ester is much more reactive than TEOS and leads to more highly dispersed, homogeneous structures [27]. In this paper we explore the permeability and corresponding structural characteristics of fluorinated poly(amide-imide)/Ti02 nano-composite systems.

67

Q. Hu et al.IJournal of Membrane Science 135 (1997) 65-79

F

CF3

0

H~C

CH~

0

CF

0 H

H3C

NH

CH.~ 0

Fig. 1. The molecularstructure of the 6FPAIpoly(amide-imide). 2. Experimental 2.1. Materials

The molecular structure of the fluorinated poly(amide-imide) used in this study is shown in Fig. 1. The synthesis and polymer properties of this polymer and similar materials are reported elsewhere [28]. The tetraethyl titanate (TET) was obtained from Aldrich Chemical Company and was used without further purification. The titanium content in the received TET was approximately 20% by weight. 2.2. Fabrication o f composite membranes

The Pyrex glass plate, onto which the polymer solutions were cast, was pre-treated to facilitate easy removal of the membrane films. This treatment consisted of soaking the plate in an aqueous solution containing 2% KOH by weight for 24 h, followed by washing with a sulfuric acid solution that contained ,-~5% of K2Cr207 and 1% of water by weight. Finally, the plate was rinsed with double-distilled water and dried in an oven at 120°C for one to two hours and then cooled down to room temperature. The plate was sprayed with acetone before the casting of the solution. The hexafluorinated poly(amide-imide) was dissolved in tetrahydrofuran (THF) and filtered to yield approximately 5% polymer in solution by weight. Tetraethyl titanate (TET) sol was obtained by dissolving up to 0.5 mg of TET (depending on the desired T i t 2 concentration) separately in 2 ml of THF under fast agitation. Both, the 6FPAI-THF and the TET sol also contained approximately 4% HC1 by weight and double distilled water, the actual proportion depend-

ing on the desired pH of the solution. The TET sol was added to the PAI/THF solution under continuous agitation and stirred for 30 to 36 h. The homogeneous solution was then cast in a Teflon ring placed on the pre-treated Pyrex glass plate. The drying was carried out in a closed box with dry nitrogen or a dry air purge. The drying rate was controlled by a heat lamp. After approximately 12 h of drying, the free standing films automatically lifted off the glass plate. The film thicknesses achieved ranged between 15-17k tin. The resulting films were then subjected to a heat treatment in a vacuum for 12 h at 150°C, followed by additional 2 h at 200°C to promote the condensation reactions of the titanium alkoxide [11]. Any residual solvent was removed with subsequent solvent exchange with methanol up to 24 h. Methanolextracted films were heat treated again in vacuum at 150°C for 24 h, followed by 24 h at 220°C. It should be pointed out that although we label the reacted inorganic component as Tit2, the final product actually contains titanium-based domains possessing unreacted alkoxide groups, as well as hydroxyl groups after heating to 200°C. Densification, that is, complete condensation, is not achieved until well above 700°C [25]. 2.3. Physical characterization methods

Transmission Electron Microscopy (TEM) analysis was performed on thin cross-sections of the membrane, 800-1200.~, using a Philips EM-420 scanning transmission electron microscope (STEM) operated at the transmission mode at 100 kV. The samples for the TEM analysis were microtomed to obtain the cross-section through the thickness of the sample.

68

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

ATR and transmission FTIR spectra were obtained using a BIO-RAD FTS-40A spectrometer equipped with a MCTA detector. The spectral resolution was 4 cm -1. Samples for the IR transmission studies were prepared by directly casting the polymer/titanium alkoxide solution in THF onto KBr discs at room temperature. The samples were dried under a heat lamp, followed by annealing under vacuum at 100°C for 24 h. The film thickness of these films was less than 1 pm. Samples for ATR studies consisted of much thicker free standing films than those obtained by casting the THF solution onto the pre-treated Pyrex plate. Dried films were cured in a vacuum for 36 h, followed by methanol extraction for 24 h to remove any residual solvent and finally annealed again under vacuum at 100°C for additional 24 h. ATR spectra were collected by employing a zinc selenide hemisphere IRE and SEAGULL accessory from Harrick Scientific Corp. Thermal Gravimetric Analysis, TGA, was carried out with a TG/DTA 200 Seiko I instrument. The temperature was varied from 25°C to 750°C at a heating rate of 10°C/min for all experiments. Prior to each run, the 5-10 mg of the membrane material was placed in an aluminum pan and preheated at 100°C to eliminate any water absorbed from the atmosphere during sample loading. All TGA experiments were carried out in a N2 atmosphere. Wide Angle X-ray Diffraction, WAXD, analysis was carried out on a Nicolet diffractometer equipped with a STOE Bragg-Brentano type goniometer. Data were collected from 5 ° to 50 ° with increments of 0.05 °, employing CuKa radiation having a wavelength of 1.54 A after monochromatization. Differential Scanning Calorimetry (DSC) was carried out on a DSC220C Seiko II instrument. A 17 mg sample was placed in an aluminium pan and heated at the rate of 10°C/min in the DSC instrument from room temperature to 290°C and then quenched using liquid nitrogen to 70°C. The quenched sample was then heated from 70°C to 450°C at 10°C/rain. The experiment was carried out in a N2 atmosphere.

2.4. Permeability measurements Single gas permeation experiments were carried out by using the constant volume permeation technique which is described elsewhere [29]. The diffusion

coefficients were calculated by the well-known time-lag method [30]. The permeabilities of gases H2, 02, N2, CH4 and CO2 were studied at three different feed pressures; namely 2, 4 and 6 atmospheres and three different temperatures; 35, 55 and 75°C. After the annealing treatment described in Section 2.2, the membrane was cooled down to room temperature and immediately transferred to the membrane cell, where it was degassed for at least 24 h. After degassing, pressure as low as 0.01 mm Hg on the two sides of the membrane was maintained for 10 h. To minimize the effect of CO2 influence on the transport properties of the other gases, CO2 permeability was measured after all the other gases were analyzed. The permeation data were collected from zero to a time ranging from 15 to 50 times the time lag, depending on the specific membrane/gas system.

3. Results and discussion

3.1. Physical characterization Optical transparency can often be used as an initial criterion to discern homogeneous mixing of the organic and inorganic components. When the inorganic domains and the polymer matrix have different refractive indices, it is possible that an optically opaque membrane will contain ceramic domains larger than 200 nm. This is generally the lower bound detected by the light scattering technique [31]. The optical appearance of some typical composite membranes, along with the fabrication conditions, is given in Table 1. The nomenclature describing the membrane reflects the following information, that is, 6FPAI/TET(50)/HCl(0.22) signifies hexafluoropoly(amide-imide)/mole percent of the titanium alkoxide (TET) with respect to the poly(amide-imide) repeat unit/catalyst concentration in mole HC1 per liter of solvent. It is quite obvious from Table 1 that the solvent has an extremely important effect on the membrane appearance. The use of dimethylacetamide, DMAC, generally leads to a phase-separated membrane, while THF favors the formation of optically transparent membranes. This observation can be explained in terms of hydrogen bonding interactions between the solvent and the polymer. DMAC can hydrogen bond with the polymer's amide group, hin-

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

69

Table 1 6FPAI/TiO2 membrane fabrication conditions No.

Membrane

Tit2 a composite % by wt

H20/Ti Mol. ratio

Solvent

Optical appearance

Other observations

1 2 3 4 5 6 7

6FPAI/TET(39)/HCI(0.22) 6FPAI/TET(55)/HCI(0.22) 6FPAI/TET(44)/HCI(0.15) 6FPAI/TET(44)/HCI(0.22) 6FPAIiTET(61 )/HC1(0.24) 6FPAI/TET(66)/HCl(0.20) 6FPAI/TET(75)/HCI(0.21)

6 11 7.3 7.3 13.4 16.3 22.5

9 7 7 10 10 8.7 9.3

DMAC DMAC THF THF THF THF THF

Translucent Translucent Opaque Transparent Transparent Transparent Transparent

Phase separated Phase separated

Brittle Brittle Brittle

a Assuming 100% conversion of TET to the TiO2.

dering the molecular interaction between the 6FPAI and the TiO2 domains, thus leading to a phase separated system. In comparison, weak polar interactions between the THF molecules and the polymer repeat units favors interaction between the 6FPAI and the TiO2 domains. Furthermore, an opaque solution can also occur when using THF solvent along with low HC1 concentration. This observation can be explained by the electric double layer theory. The PZC (Point of Zero Charge) of the TiO2 is pH = 6 [32]. The TiO2 surface is positively charged at pH values lower than 6. Accordingly, the low HCI concentration reduces the electric charge on the TiO2 surface, because it shifts the pH of the solution closer to the PZC. The decrease in the surface charge reduces, in turn, the (-potential that is a direct indicator of the repulsive force among the TiO2 particles. Therefore, an opaque solution is believed to arise from the weak repulsive forces among the TiO2 particles. This phenomenon is not expected to occur at relatively high HC1 concentrations because of higher repulsive forces between the TiO2 particles. Also, the solvent properties of THF will change with increasing HCI and H20 content. For this reason, we have maintained low pH conditions during the fabrication of all samples used in the characterization studies. Complete solvent removal was verified with TGA analysis. No weight losses due to the presence of a solvent were observed. In fact the TGA curves did not show any weight loss until above 400°C. The TiO2 concentration was also analyzed by TGA and is summarized in Table 2. This measurement was based on the fact that most polymers are essentially pyrolyzed before reaching 700°C in air, yet TiO2 remains thermally stable over 1000°C in air. Therefore, the

Table 2 TiO2 content determined by TGA at 750°C in an air atmosphere Membrane

6FPAI/TiO 2 (7.3%)

Theory (% by wt.) 7.3 Measured (% by wt.) 6.2 Loss of Tit2 1.1 % Incorporated 84.9

6FPAI/TiO2 (16.3%)

6FPAI/TiO2 (22.5%)

16.3 14.3 2.0 87.7

22.5 20.8 1.7 92.4

remaining residue reflects the actual TiO2 content. The high percentage of Tit2 incorporation shows the feasibility of the chosen fabrication methods. Some loss of Tit2 may have also occurred during the methanol extraction process. The WAXD profiles of the 6FPAI/TiO2 composites and the unfilled 6FPAI polymer are shown in Fig. 2. Although the scattering intensity decreases with the T i t 2 content due to the absorption of X-rays by the inorganic component, the location of the peaks remains unchanged. We believe that these peaks are associated with some type of short range molecular order in the polymer. Furthermore, Tit2 exists in an amorphous state within the 6FPAI, since no peaks associated with crystalline forms of the Tit2 [33] are observed. Thus, it is apparent that the structure of the 6FPAI crystals, most likely resulting from solvent induced crystallization, is unaffected by the presence of the Tit2 domains, although there may be an overall decrease in polymer crystallinity with the introduction of the inorganic component. DSC measurements were carried out to study the changes of the glass transition temperature with T i t 2 composition. However, because the results were complicated by crystallization features of the polymer, the

70

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

~.~

1

A. erPAu

c. 6FPA;/TIOe(16,0~}

%

£

...¢~-

~m'

z

~,

• "~

. ~ ~.'~_

,

..'~,.

.

1o

.

zo

,....~~ 40

30

~o

20(degree) Fig. 2. WAXD patterns for the unfilled 6FPAI and the 6FPAI/TiO/ composites. The number in the brackets is the TiO2 content by weight based on 100% conversion from TET to TiO2.

10000.

B. 6FFAk'fiO1 (16 ~%) c. 6F~'AgnOz al.sw,)

points to a heterogeneous crystallization process, possibly resulting from polymer crystallization originating in the amorphous phase and crystallization near the TiO2 domains which may serve as nucleating centers. TEM results showing the morphology of the membranes with varying TiO2 content are illustrated in Fig. 4. The dark portions represent the TiO2 rich phase. Although the size of the TiO2 domains increases with the TiO2 content, the maximum size of the TiO2 domains achieved, that is, in 6FPAI/TiO2 (22.0%), is still less than 5 nm. This suggests that the binary system is highly dispersed and compatible, possibly as a result of specific molecular interactions between the two components. We shall examine this more closely in the next section.

~1:10,

3.2. Spectroscopic studies 6000.

41~)-

~0-

0

/

&

&

Tcmperat ur© ~eC)

Fig. 3. DSC results of the second heat for the unfilled 6FPAI and the 6FPAI/TiO2 composites.

changes m Tg were inconclusive. As suggested by the WAXD data, the DSC results also indicate that there may be an overall decrease in polymer crystallinity with the introduction of the TiO2 component. As shown in Fig. 3, a significant crystallization in these materials occurs from 275 to 420°C. Since the TGA results show virtually no weight loss in this temperature range, these exotherms cannot be associated with removal of organics associated with continuing polymerization of the alkoxide. While the unfilled 6FPAI has only one broad crystallization exotherm, the composite materials have two crystallization exotherms. The exotherms increase with increasing TiO2 content suggesting that polymer crystallinity, depressed by the presence of TiO2 domains during fabrication conditions, can only be realized at crystallization temperatures above the glass transition temperature. The presence of the two exotherms further

Transmission spectra of the unfilled and filled 6FPAI samples are shown in Fig. 5. The regions that do not show IR absorption are omitted. The bands in the region from 800 to 5 0 0 c m -~ increase with increasing TiO2 content, due to the presence of a broad band associated with the vibration of the Ti-O bond [34]. Other spectral regions of interest include the N - H stretching region near 3300 cm -1 and the amide C=O region around 1700 cm l, which are replotted in an expanded version in Fig. 6. The bands at 1780 cm -1 and 1730 cm - l are associated with the imide carbonyl band [35] and are quite insensitive to the presence of the TiO2 component. Other characteristic bands include the absorption at 1600 cm - t due to the benzene ring stretch and a contribution from O - H bonding in monomeric water, which also has a band at 1630 cm - t [35]. A careful examination reveals that the intensity of this latter band, although very weak, increases with an increase in the TiO2 content. Additional bands include the free amide C=O absorption located at 1690 cm -1 and the hydrogen bonded C=O found at 1666 cm -~. Although the intensity of this shoulder is weak, an increase in the TiO2 content causes this shoulder to disappear, suggesting that the incorporated TiO2 domains disrupt any self-hydrogen bonding between chains, that is, between the C=O and the N - H group in the poly(amide-imide) itself. Thus, there appears to be no hydrogen bonding present between the amide C=O group in the polymer and the

Q. Hu et aL /Journal of Membrane Science 135 (1997) 65-79

71

(A)

Fig. 4. TEM of the pure 6FPAIand the 6FPAI/TiO2composites. (A) Unfilled 6FPAI, (B) 6FPAI/TiO2 (7.3%), (C) 6FPAI/TiO2 (16.3%), (D) 6FPAI/TiO2 (22.5%).

TiO2 component. On the other hand, the N - H stretching band, visible as a shoulder at 3275 cm -~ and representative of hydrogen bonded N - H groups, increases slightly in intensity with increasing TiO2 content, suggesting hydrogen bonding between the N - H groups in the polymer and the Ti--O-Ti or T i - O - H groups of the inorganic component. Results of ATR spectral analysis of samples which have been treated with solvent extraction, shown in Fig. 7, also exhibit a

small shoulder at 3275 cm -1 which, however, is not as obvious as in the transmission spectra. Since solvent extraction seems to also remove most of the monomeric water observed at 1630 cm -1, it is possible that some of the hydrogen bonding interactions may have actually occurred between the N - H groups in the 6FPAI and the monomeric water bound to the TiO2 domains. This interaction is particularly important during the membrane fabrication process itself.

72

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

(B)

Fig. 4. (Continued)

3.3. Permeation studies

Permeation studies were only carried out with the unfilled 6FPAI and the 6FPAI/TiO2 (7.3%) composite membranes. The experimental results obtained for various gases are summarized in Table 3(a-d). The transport parameters are averaged over the three feed pressures, namely, 2, 4 and 6 atmospheres, as the variation in the calculated values were within the

accuracy of the experimental measurements (5%). The value in the brackets represents the maximum variation observed when comparing the average calculated value with the measured values at the three feed pressures. The deviations in parameter values for CO2 are larger than for other gases because a finite decrease in CO2 permeability was observed with increasing pressure. Furthermore, we do not report any experimental error for hydrogen because the

73

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

(c)

Fig. 4. (Continued)

permeability was evaluated only at one pressure. Also, the time lag observed during hydrogen permeation was too small to permit accurate calculation of its diffusion coefficient. In general, the data indicate that permeabilities and diffusion coefficients decrease and solubilities and certain selectivities increase with the presence of the TiO2 component. The temperature effects on gas permeability for the 6FPAI and 6FPAI/TiO2 (7.3%) membranes were

analyzed by using the Arrhenius equation [37,38]. P = P0exp ( ~ T P )

(1)

From least squares fit of the experimental data, the apparent activation energy, Ep, and the pre-exponential factor, P0, values were obtained and are tabulated in Table 4. The correlation coefficient, r, which repre-

74

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

(D)

Fig. 4. (Continued)

sents of the quality of the regression fit, is also given in Table 4 as the permeability values used for least squares fit were averaged over three different feed pressures for each temperature (see above). The relatively good fit confirms the accuracy of the experimental values and the reasoning behind the mean values quoted. The number in the brackets represents the standard deviation observed for the curve fit.

The activation energy results obtained are consistent with apparent activation energies observed in similar polyimide systems [38]. With the exception of CO2 and H2, all other gases have a higher value of Ep in the 6FPAI/TiO2 membrane than in the pure 6FPAI membrane. The increase in the apparent activation energy for permeation suggests that the composite membrane has a more rigid or denser structure than the unfilled 6FPAI membrane. In part this

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

75

////

i .0

0.8

0.6

0.4

o.0 4000

. P'~/%~///,.. 3500

!

,

3000

I

1000

1500

WAVENUMBERS. cm

500

-i

Fig. 5. F'I'IR spectra of the unfilled 6FPAI and the 6FPAI/TiO2 composites at 35°C. (A) Unfilled 6FPAI, (B) 6FPAI/TiO2 (7.3%), (C) 6FPAI/TiO2 (22.5%).

0.15

- ~* /

1.2

375

0.12 ~

1¢~

i

16oo

1600

0.9

o.oo

~ ~

o.6 o.o~

~

~ B

0.03 0.00 3600

~

3300

3000

1800 WAVENUMBERS,

0.3 1700

| 600

1500

0.0 1400

c r n ~l

Fig. 6. F'FIR spectra of the unfilled 6FPAI and the 6FPAI/TiO2 composite at 35°C. (A) Unfilled 6FPAI, (B) 6FPAI/TiO2 (7.3%), (C) 6FPAI/TiO2 (22.5%).

increase can be attributed to the increase in the activation energy of diffusion, Ed, as our data indeed show an appreciable loss in diffusivity with the incorporation of the ceramic component. The decrease in the apparent activation energy of CO2 can be explained by the large negative contribution of the enthalpy of sorption, AH~, in the well-known equation [37]: Ep = Ed + /kHs

(2)

The interaction between residual OH groups on the TiO2 component and the polar COz molecules can give rise to a higher enthalpy of sorption for CO2 in the

6FPAI/TiO2 composite membrane as compared to the unfilled 6FPAI membrane. This is also demonstrated by the increase in the apparent solubility of CO2 in the composite membrane as shown in Table 3(c). On the other hand, because the kinetic diameter of H2 is small compared to other gases, one would not expect to see very large changes in the activation energy of diffusion with the addition of the inorganic component to the membrane. Smaller molecules are generally affected to a lesser extent by changes in the structural rigidity [38,39]. Thus, the observed decrease in the apparent activation energy of H2 permeation must also be associated with a higher enthalpy of sorption of H2

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

76 0.10

,

,

/,/

;

I

I

l,O

I

3356 0.08

I

3275

0.8

ll 686

b3 0.08

0.6

o {n == 0.04

0.4

0.02

0.2

z m

~d z o

i

0.00 3600

,

3300

3000

1800

I

,

1700

I

,

1600

I

1500

,

0.0 1400

WAVENUMBERS, em -I Fig. 7. ATR spectra of the unfilled 6FPAI and the 6FPAI/TiO2 composite at 35°C. (A) Unfilled 6FPAI, (B) 6FPAI/TiO2 (22.5%).

in the 6FPAI/TiO2 composite membrane. In fact, recent studies of hydrogen adsorption/absorption reaction and transport through TiO2 film indicate that hydrogen interacts strongly with the anions in the oxide [40]. In Table 3(d), with the exception of O2/N 2 gas pair, where the gas molecules have similar size, we can see an improvement in the permselectivities of selected gas pairs with the incorporation of the TiO2 component. In fact, the improvement is quite large when we compare the permselectivities of the HE/CH 4 gas pair. The uncertainty in the selectivity values was calculated using the propagation of error technique [41 ] and is not reported for gas pairs containing H E (for reasons mentioned above). Despite this, the overall performance of the composite membrane is still moderately under the upper bound [36]. For example, the upper bound of the permselectivity of CO2/CH 4 is ".~50 at COz permeability of 50 barrers [36], while the permselectivity of CO2/CH4 for the 6FPAI/TiO2 (7.3%) composite membrane is 33.3 at a CO2 permeability of 44.7 barrers. Despite this observation, it is worth pointing out that the TiO2 weight percentage of 7.3% corresponds to a volume percentage of only 2.4%, on the basis of 6FPAI density of 1.4 g/cc [28] and a TiO2 density of 4.26 [42]. In this regard, the improvement in permselectivity brought about by the incorporation of a TiO2 component in the

poly(amide-imide) membrane can be considered significant. Furthermore, it is possible to fabricate composite membranes with higher Ti02 content; however, these have to be supported on highly porous membrane supports, a task, which presents its own problems. Finally, as observed in the pure poly(amide-imide), the permselectivities of the composite membranes still decrease with increasing temperature. Hence, the presence of the Ti02 component does not stabilize the change in permselectivity with temperature as was originally hoped. This is due to the fact that restrictions on molecular relaxations in the polymer matrix are limited to the interfacial regions between the two components. Furthermore, any hydrogen-bonding interactions present in these regions will decrease with increasing temperature. Hence, a hybrid system with direct covalent linkages between the polymeric and ceramic components, instead of physical interactions, would form a much better high performance membrane. We are in the process of investigating such materials.

4. Conclusions

Nano-composite membranes, based on a fluorinated poly(amide-imide) polymer matrix and a TiO2 component, have been successfully fabricated by a

77

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

Table 3 Permeabilities, diffusivities and solubilities of the 6FPAI and 6FPAI/TiO2 (7.3%) averaged over three pressures 6FPAI 35°C

6FPAI/TiO2 (7.3%) 55°C

75°C

35°C

55°C

75°C

2.60 (0.04) 3.2 (0.2) 13.7 (0.9) 55.8 (1.9) 87.6

3.60 (0.02) 4.3 (0.3) 16.8 (1.4) 63.9 (2.0) 108.3

1.30 (0.01) 1.80 (0.03) 9.2 (0.2) 44. 7 (3.7) 66.2

2.00 (0.01) 2.50 (0.06) 11.3 (0.3) 48.7 (2.0) 81.7

2.90 (0.02) 3.4 (0.1) 13.8 (0.4) 50 (2.0) 100.0

(b) Diffusion (cm2/s) (1 × 10-8) 02 9.2 (0.3) N2 2.3 (0.2) CO2 2.9 (0.1) CH4 0.73 (0.2)

21.3 (0.9) 5.7 (0.1) 7.1 (1.3) 1.3 (0.3)

46.7 (2.5) 11.3 (1.0) 10.9 (0.5) 2.8 (0.2)

7.7 (1.1) 1.7 (0.2) 2.8 (0.5) 0.34 (0.02)

12.7 (0.3) 3.4 (0.2) 4.5 (0.4) 0.76 (0.04)

20.3 (0.6) 4.7 (0.4) 6.6 (0.6) 1.60 (o.04)

(c) Apparent solubility (cc(STP)/cc-atm) 02 0.92 (0.05) N2 0.76 (0.1) CO2 13.7 (0.8) CH4 2.0 (0.6)

0.49 (0.02) 0.43 (0.02) 6.2 (1.4) 1.5 (0.3)

0.27 (0.03) 0.29 (0.05) 4.5 (0.3)

0.68 (0.01) 0.58 (0.03) 8.2 (1.2) 2.0 (o.1)

0.52 (0.02) 0.55 (0.04) 5.8 (0.7)

(0.1)

0.92 (0.1) 0.81 (0.06) 12.3 (3.0) 3.0 (0.1)

4.3 (0.4) 21.7 (0.8) 33.7

3.7 (0.5) 17.6 (0.6) 30

5.0 (0.1) 33.3 (3.0) 50.8

4.5 (0.2) 24.5 (1.0) 41.0

4.0 (0.2) 17.3 (0.7) 34.5

(a) Permeability (in Barrer):

CH4 N2 02 CO2 H2

1.80 (0.03) 2.30 (0.06) 11.2 (0.9) 52.7 (0.9) 67.0

1.0

1.3

(0.04)

(d) Selectivity

PA/PB O2/N2 CO2/CH4 Hz/CI-h

4.9 (0.4) 29.3 (0.7) 37.2

s o l - g e l method. A p p a r e n t activation energies for p e r m e a t i o n of the various gases e x a m i n e d suggest that the n a n o - c o m p o s i t e m e m b r a n e has a denser and a more rigid structure w h e n compared to the corresponding pure p o l y ( a m i d e - i m i d e ) m e m b r a n e . Furthermore, there appear to be specific interm o l e c u l a r interactions b e t w e e n gases such as CO2, H2 and the TiO2 domains. Finally, the composite

m e m b r a n e has shown higher permselectivities for selected gas pairs w h e n c o m p a r e d to the pure p o l y ( a m i d e imide) m e m b r a n e , even at very low v o l u m e concentration of the TiO2 c o m p o n e n t . Such results are encouraging, because they suggest that possibly higher selectivities could be achieved at increasingly higher concentrations of the ceramic component.

78

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79

Table 4 Arrhenius equation parameters for various gases 6FPAI Po Oz

269.4

N2

587.6

6FPAI/TiO2 (7.3%)

Ep 1.95 (.05) 3.40

r

Po

Ep

r

-0.9997

329.6

-0.9997

-0.9999

453.8

2.20 (0.05) 3.38

(.04) CO2

273.9

CH4

801.9

1.02 (0.3) 3.74

0.9640

122.1

-0.9996

1106

(0.3) Hz

4433

2.57

-0.9999

(.02) 0.61 (0.2) 4.12

-0.9654 -0.9997

(0.1) -0.9995

2394

(0.1)

2.20 (0.05)

-0.9997

Ev: kcal/mol; Po: Barter; r: correlation coefficient.

Acknowledgements The authors wish to a c k n o w l e d g e the financial support f r o m the National S c i e n c e F o u n d a t i o n through Grant C T S - 9 6 2 2 4 7 3 . Special thanks are e x t e n d e d to Professor H e r v e M a r a n d and Dr. Srinivas Srivatsan in the department o f chemistry at VPI and S U for very helpful c o m m e n t s and to Dr. Stephen M c C a r t n e y at the N S F and ST C e n t e r at VPI and S U for his kind help in obtaining the T E M pictures. M a n y thanks are also e x t e n d e d to Drs. Pushpinder Puff and Keith M u r p h y f r o m A i r Products and C h e m i c a l s Co. for very helpful discussions concerning the permeability data.

[6]

[7]

[8]

[9]

[10]

References [11] [1] H.H. Huang, B. Orler and G.L. Wilkes, Structure-property behavior of new hybrid materials incorporating oligomeric species into sol-gel glasses, Macromolecules, 20 (1987) 1322-1330. [2] B. Wang and G.L. Wilkes, New Ti-PTMO and Zr-PTMO ceramer hybrid materials prepared by the sol-gel method: Synthesis and characterization, J. Polym. Sci.; Part A: Polym. Chem., 29 (1991) 905-909. [3] J.M. Mark, C.Y. Jiang and M.Y. Tang, Simultaneous curing and filling of elastomers, Macromolecules, 17 (1984) 26132616. [4] C.J.T. Landry, B.K. Coltrain and B.K. Brady, In situ polymerization of tetraethoxysilane in poly(methyl methacrylate); morphology and dynamic mechanical properties, Polymer, 33(7) (1992) 1486. [5] J.J. Fitzgerald, C.J. Landry and J.M. Pochan, Dynamic studies of the molecular relaxations and interactions in microcompo-

[12]

[13]

[14]

[15]

sites prepared by in-situ polymerization of silicon alkoxides, Macromolecules, 25 (1992) 3715-3722. J.E. Mark and S.J. Pan, Reinforcement of poly(dimethylsiloxane) networks by in-situ precipitation of silica. A new method for preparation of filled elastomers, Makromol. Chem. Rapid Commun., 3 (1982) 681-685. M. Nandi, J.A. Conklin, L. Salvati Jr. and A. Sen, Molecular level ceramic/polymer composites. 2. Synthesis of polymertrapped silica and titania nanoclusters, Chem. Mater., 3 (1991) 201-206. T. Saegusa, Organic polymer-silica gel hybrid: precursor of a highly porous silica gel, J. MacromoL Sci.-Chem., A28(9) (1991) 817-829. J.C. Shrotter, M. Smaihi and C. Guizard, Polyimide-siloxane hybrid materials: influence of coupling agents addition on microstructure and properties, J. Appl. Polym. Sci., 61(12) (1996) 2137-2149. J. Livage, M. Henry and C. Sanchez, Sol-gel chemistry of transition metal oxides, Progr. Solid State Chem., 18 (1988) 259-342. C.J. Brinker and G.W. Scherer, Sol~3el Science; the Physics and Chemistry of Sol-Gel Processing, Academic Press, San Diego, 1990. M. Smaihi, T. Jermoumi, J. Marigan and R.D. Noble, Organic-inorganic gas separation membranes: preparation and characterization, J. Membrane Sci., 116 (1996) 211220. J.C. Shrotter, S. Goizet, M. Smaihi and C. Guizard, Preparation and characterization of new hybrid organicinorganic materials for gas separation, Proceedings Euromembrane Bath 1995, Vol. 1, pp. I313-I318. C.J. Wung, Y. Yang, EN. Prasad and EE. Karasz, Poly(pphenylene vinylene)-silica composite: A novel sol-gel prcessed non-linear optical material for optical waveguides, Polymer, 32 (1991) 604-608. H. Krug and H. Schmidt, Organic-inorganic nanocomposites for micro-optical applications, New J. Chem., 18 (1994)

1125-1134.

Q. Hu et al./Journal of Membrane Science 135 (1997) 65-79 [16] D. Khastgir, H.S. Maiti and P.C. Bandyopadhyay, Polystyrene-titinia composite as a dielectric material, Mater. Sci. Eng., 100 (1988) 245-253. [17] S. Wang and J.E. Mark, Generation of finely-divided nickel particles in a protective medium and their use as high-activity catalysts, Polym. Bull., 29 (1992) 343-348. [18] H. Schmidt and H. Wolter, Organically modified ceramics and their applications, J. Non-Cryst. Solids, 121 (1990) 428. [19] B. Wang and G.L. Wilkes, Novel hybrid inorganic-organic abrasion-resistant coatings prepared by a sol-gel process, J. Macromol. Sci. Pure Appl. Chem., A31 (1994) 249. [20] H. Kita, S. Hideki, K. Tanaka and K. Okamoto, Transport properties of polyimide/silica composite membranes, J. Photopolym. Sci. Technol., 8(2) (1995) 315-320. [21] M. Moaddeb and W.J. Koros, Gas transport properties of thin polymeric membranes in the presence of silicon dioxide particles, J. Membrane Sci., 125(2) (1997) 143-161. [22] T.H. Kim, W.J. Koros, G.R. Husk and K.C. O'Brian, Relationship between gas separation properties and chemical structure in a series of aromatic polyimides, J. Membrane Sci., 37 (1988) 45-62. [23] S.A. Stem, Y. Mi and H. Yamamoto, Structure/permeability relationship of polyimide membranes. Applications to the separation of gas mixtures, J. Polym. Sci.: Part B: Polym. Phys., 27 (1989) 1887-1909. [24] K. Matsumoto and P. Xu, Gas permeation of aromatic polyimides, J. Membrane Sci., 81 (1993) 23-30. [25] E Orgaz-Orgaz, Gel to glass conversion: Densification kinetics and controlling mechanism, J. Non-Cryst. Solids, 100 (1988) 115-141. [26] J. Brandrup and E.H. Immergut, Polymer Handbook, 3rd edn., John Wiley and Sons, New York, 1989. [27] Q. Hu, E. Marand and G.L. Wilkes, In situ formation of nanosized TiO2 domains within a poly(amide-imide) by a sol-gel process, J. Appl. Polym. Sci., (1997) submitted. [28] D. Fritsch and K.V. Peinemann, Novel highly permselective 6F-poly(amide-imide)s as membrane host for nano-sized catalysts, J. Membrane Sci., 99 (1995) 29-38.

79

[29] S. Dhingra and E. Marand, Mixed gas transport study through polymeric membranes. 1. A novel experimental technique, J. Membrane Sci., (1997) submitted. [30] J. Crank, The Mathematics of Diffusion, 2nd edn., Clarendon Press, Oxford, 1975. [31] L.C. Sawyer and D.T. Grubb, Polymer Microscopy, Chapman Hall, London, 1987. [32] R.J. Hunter, Zeta Potential in Colloid Science, Academic Press, New York, 1981. [33] T. Lopez, E. Sanchez, P. Bosch, Y. Meas and R. Gomez, Mater. Chem. Phys., 32 (1992) 141. [34] R.A. Nyquist and R.O. Kagel, Infrared Spectra of Inorganic Compounds, Academic Press, New York, 1971, p. 214. [35] N.L. Alpert, W.E. Keiser and H.A. Szymanski, IR - Theory and Practice of Infrared Spectroscopy, A Plenum/Rosetta Edition, New York, 1973. [36] L.M. Robeson, Correlation of separation factor versus permeability for polymeric membranes, J~ Membrane Sci., 62 (1991) 165-185. [37] V.T. Stannet, Diffusion in Polymers, John Wiley and Sons, New York, 1971. [38] L. Costello and W. Koros, Effect of structure on the temperature dependence of gas transport and sorption in a series of polycarbonates, J. Polym. Sci.: Part B: Polym. Phys., 32 (1994) 701-713. [39] L. Costello and W. Koros, Thermally stable polyimide isomers for membrane-based gas separations at elevated temperatures, J. Polym. Sci.: Par B: Polym. Phys., 33 (1995) 135-146. [40] S. Pyun and Y.G. Yoon, Hydrogen transport through TiO2 film prepared by plasma enhanced chemical vapour deposition (PECVD) method, in: Hydrogen Effects in Materials, Proc. 5th Inter. Conf. on the Effect of Hydrogen on the Behavior of Materials, Sept. 11-14, 1994, A.W. Thompson and N.R. Moody (Eds.), TMS publishers, pp. 261-269. [41 ] J.P. Holeman, Experimental Methods for Engineers, 6th edn., McGraw-Hill, New York, 1994. [42] D.R. Lide, Handbook of Chemistry and Physics, 71st edn., 1990-1991.