Gas permeability of poly [1-(trimethylsilyl)-1-propyne] membranes modified by hexafluorobutyl methacrylate

Gas permeability of poly [1-(trimethylsilyl)-1-propyne] membranes modified by hexafluorobutyl methacrylate

Journal of Membrane hence, 82 (1993) 99-115 Elsevler Science Publishers B V , Amsterdam 99 Gas permeability of poly [l-(trimethylsilyl) -1-propyne]...

1MB Sizes 0 Downloads 2 Views

Journal of Membrane hence,

82 (1993) 99-115 Elsevler Science Publishers B V , Amsterdam

99

Gas permeability of poly [l-(trimethylsilyl) -1-propyne] membranes modified by hexafluorobutyl methacrylate Guanwen Chen*, Hans J. Griesser and Albert W.H. Mau** Dwlslon of Chemtcals and Polymers, CSIRO, Prwate Bag 10, Clayton 3168, VLC (Australaa) (Received August 19,1992, accepted m revised form February 26,1993)

Abstract

Poly [ l- (trrmethylsrlyl)-1-propyne] (PTMSP) membranes have a very high oxygen permeability but a low separation factor for oxygen and nitrogen Hexafluorobutyl methacrylate (HFBM) was mcorporated mto PTMSP membranes m order to increase the seleckvrty of the membrane for oxygen HFBM was rmblbed into the cast PTMSP membrane and polymerized by UV rrradration The resulting membranes were not fully symmetrical; the chemical modification took place to a larger extent m the membrane volume closer to the UV light The activation energy for the permeation of oxygen and nitrogen through the modtied PTMSP membranes was determined to be positive For HFBM modified PTMSP membranes, the higher the flux was, the lower the separation efficiency Values of the permeability coefficient P for oxygen at 2O”C, and of the separation factor for the modified membranes were P(0,) =l 39x10m7 cm3 (STP)-cm/cm’-set-cmHg, and P(O,)/P(N,) =5 4 for the sample with the highest separation efficiency, and P(0,) =5 56x lo-’ cm3 (STP)-cm/cm’-set-cmHg, and P(O,)/ P(N,) =2 0 for the sample with the highest permeability Both the unmodified and HFBM modified PTMSP membranes showed no dechne m P(Os) and P(O,)/P(N,) values over a penod of six months Key words

separation of oxygen and nitrogen, gas separation

Introduction Membranes have attracted much attention for use in gas separation processes such as the enrichment of oxygen from air. Poly [ l-(tnmethylsilyl) -l-propyne] (PTMSP) has been a material of particular interest because solvent cast membranes fabricated from this material were found to have an exceptionally high permeability for oxygen and other gases [l-7]. *Permanent address Institute of Chemrstry, Academia Smlca, Beqmg 100080, P R China To whom correspondence should be addressed

The permeability coefficient for oxygen, P (0,)) which depended somewhat on fabrication conditions, was determined to be of the order of 7 x lo-’ cm3 ( STP)-cm/cm2-set-cmHg at 25°C [ 11, exceeding by an order of magnitude the value of P( 0,) of poly (dimethyl siloxane) (PDMS ), which is 6.1 x lo-’ cm3 (STP) -cm/ cm2-set-cmHg [8]. As PDMS has the highest oxygen permeability among the commonly available commercial polymer materials, PTMSP is a promising material for use as a high flux gas separation membrane. The rapid oxygen permeation through PTMSP could provide the foundation for the

0376-7388/93/$06 00 0 1993 Elsevler Science Pubhshers B V All nghts reserved

100

design of commercially viable processes. However, as nitrogen also permeates readily, the separation factor for oxygen and nitrogen, P( 0,) /P ( N2), is low for PTMSP membranes. This separation factor is dependent on the membrane fabrication parameters and is approximately 1.7 [ 11. A high throughput is also observed for many other gases, resulting in a low separation factor for most gas combinations and thus compromising the potential for application of PTMSP membranes in most gas separation processes of commercial interest. Clearly, improvements are needed to the selectivity of the membrane in order to exploit its high throughput in practical applications. A further potential disadvantage of PTMSP is that the membrane performance has been reported not to be stable with time [ 3,9-111, with the permeability coefficients decreasing substantially within days following preparation of the membrane, even when stored in a vacuum container [9]. After one hundred days, the permeability of isobutane for instance had decreased to one percent of its original value, and decreases were also observed for the permeabilities of O2 and N2 [ 10 1. The reasons for this decrease in permeability are not clear, and different models have been proposed [9,11,12]. In marked contrast to these reports, however, one study found no aging effects in solvent cast PTMSP membranes [ 13 1. A number of researchers have studied ways of improving the permselectivity of PTMSP membranes [ 11,14-181. For instance, application of an ultrathin organosilicon copolymer layer on to the surface of a PTMSP membrane enabled the enrichment of oxygen from air by increasing the oxygen concentration from 21% to35%inonestep [Xi]. The relatively high solubility of oxygen in fluorinated polymers is well known, and hence it appears promising to modify PTMSP membranes by using fluorination to increase the separation efficiency while avoiding excessive

G Chen et al/J Membrane Sea 82 (1993) 99-115

reduction in permeability. Chemical modification of PTMSP membranes has been performed using Fz gas diluted in a nitrogen flow [16-U]. The modification proceeded to a depth of about 50 to 100 nm and fluorination occurred on the trimethylsilyl side group. The P(0,) value of the modified membranes was a little lower than that of PTMSP, but the transport selectivity for oxygen over nitrogen increased greatly, to values of P (0,) /P (N, ) in the range of 4 to 5. In this report we present an alternative, convenient approach based on the incorporation of hexafluorobutyl methacrylate (HFBM ) into PTMSP membranes, and report data on permeability and separation efficiency of the modified membranes. We have also reinvestigated the question of the stability of PTMSP membranes on storage and found that the performance of our membranes was stable over a period of six months. Experimental

Trimethylsilyl-1-propyne (TMSP), obtained from Shinetsu Chemical Industry Ltd., Tokyo, Japan, was refluxed over CaH, under a protective nitrogen atmosphere and distilled three times at normal pressure. The boiling point was 99°C to 100” C. The purified monomer was polymerized following the method described by Masuda [ 191, using a TaC& catalyst in toluene solution at 30°C. A 100 ml flask equipped with a three-way stopcock was flushed with dry nitrogen and charged with TMSP (55 mmol, 6.1 g, 8.3 ml) and toluene (19 ml) using a hypodermic syringe. This monomer solution was heated to 80” C. Tantalum pentachloride (1.0 mmol, 360 mg) was placed in another 106 ml flask equipped with a three-way stopcock, the flask was flushed with dry nitrogen and 25 ml of toluene added using a syringe. The metal

101

G Chen et al/J Membrane Set 82 (1993) 99-115

chloride was completely dissolved by keeping the flask at 80°C for 15 min, to give a yellow solution. Using a syringe, 25 ml of the monomer solution was then added to the catalyst solution. Polymerization was performed at 80” C for 3 hr. As the polymerization proceeded, the solution turned brown and solidified. The polymerization reaction was terminated by adding 25 ml of a 1: 9 v/v mixture of methanol and toluene while stirring the mixture. The formed polymer was separated off and redissolved in 2 litres of toluene with stirring. The solution was poured little by little into 20 litres of methanol with stirring. The precipitated polymer was allowed to stand overnight, filtered off, washed with methanol and dried to a constant weight. The product was a white tib&like polymer. The yield was 95%. The weight average molecular weight (M,) of the polymer was determined to be 7.45 x 10’ by gel permeation chromatography. A sample of a related copolymer material, designated CP in the following, was kindly supplied by Prof. T. Higashimura (Department of Polymer Chemistry, Kyoto University, Japan). It consisted of 95% TMSP and 5% of the unit: -7 CH3

=

y~(CH32 742 wqJ3

This material was used in the same way as PTMSP for the fabrication of membranes. Membmne fabrication and modification Membranes for gas permeation measurements were prepared by casting a dilute solution (l-2 wt.% in toluene) in a glass dish at room temperature. The solvent was allowed to

evaporate very slowly over about one week. The membranes were then dried in vacuum. The thickness of the dry membranes was in the range of 30 to 80 p, depending on the initial concentration. The membranes thus fabricated appeared uniform and dense. Hexafluorobutyl methacrylate (HFBM ) was obtained from Polysciences Inc., Warrington, PA. PTMSP membranes were modified using either the pure liquid, or a solution of HFBM in absolute ethyl alcohol. Two methods were used to accomplish the incorporation of HFBM. Method 1 comprised application of HFBM onto the membrane surface. Pure HFBM, or a solution of HFBM in absolute ethyl alcohol, was applied in measured quantities onto one surface of PTMSP membranes from a pipette. The wetting and spreading was good, ensuring that the membrane surface was covered evenly with HFBM, which was completely absorbed by the membrane rather than remaining on the surface as a visible layer. Polymerization of HFBM was then induced by UV irradiation (254 nm) using a low pressure mercury lamp at an intensity of 0.01 mW/cm2 for 3 hr. Membranes modified by this treatment are designated as “PTMSP/HFBM”. Method 2 comprised membrane immersion in HFBM solution. PTMSP membranes were soaked in solutions of various concentrations of HPBM in ethanol, and then W irradiated for 6 hr under a protective nitrogen blanket, using the same irradiation conditions as above. Membranes thus modified are labelled “PTMSP/HFBM (S )“. The modified membranes were dried for 12 hr in vacuum at room temperature, then in vacuum at 60°C for another 12 hr. PTMSP membranes without HFBM were also subjected to UV irradiation in order to study possible effects arising from irradiation alone. Permeabilitymeasurements The permeability coefficients were determined using a permeability test device K-315N-

102

G Chen et al /J Membrane Scr 82 (1993) 99-115

01 from Rikaseki Kogyo Ltd., Tokyo, Japan. The oxygen to nitrogen permselectivity of the membrane was characterised by the ratio of P(G,) to P(N,).

Infrared spectra were measured on a BOMEM MB-100 FTIR spectrometer in the transmission mode.

Spectroscope: characterizatwn of membranes

Results

XPS analyses were performed on a VG Escalab V spectrometer with a non-monochromatic Al-Ka source at a power of 200 W and a hemispherical analyser. The total pressure in the main vacuum chamber during analysis was typically lo-’ mbar. The binding energy scale was calibrated with data from sputter cleaned nickel (Fermi edge at 0 eV), gold (Au 4fTj2 at 83.98 eV) and copper (Cu 2p,,, at 932.67 eV) foils. Elements present were identified from survey spectra. High resolution spectra were recorded from individual peaks at 30 eV pass energy in the fured analyser transmission mode. The elemental composition of the surface was determined based on a first principles approach [ 201; atomic ratios were calculated from integral peak intensities using a nonlinear Shirley type background and published values for photoionization cross-sections [ 211. The inelastic mean free path of the photoelectrons was assumed to be proportional to E” ‘, where E is the kinetic energy [ 221. The transmission function of the analyser was determined to be proportional to E - O5. The random error in the quantitative analysis of elemental compositions is between 5% and 10% (usually 7-8% ) on this XPS unit. Specimens were analyzed at two different photoelectron emission angles, 0 ’ and 60” as measured from the surface normal, to obtain information about the variation with depth of the chemical composition. The effects of sample decomposition under non-monochromatized X-ray radiation have been considered: specimens were exposed to radiation for less than one minute before the start of the data acquisition, and the total exposure time was kept below 30 min.

Permeabdity of oxygen and nrtrogen Table 1 lists the permeability coefficients of nitrogen, P(N,), and oxygen, P(O,), for PTMSP, PTMSP/HFBM and PTMSP/ HFBM( S) membranes, and the separation factors for oxygen and nitrogen. The incorporation of HFBM caused considerable increases in the 02/Nz separation efficiency of the PTMSP membranes. The highest value for measured was 5.4. P(G,)IP(N,) The permeability coefficient decreased on HFBM incorporation, regardless of whether this was done by applying pure HFBM directly on to the surface of the membrane (Method I ) or by immersing the membrane m a solution of HFBM in absolute ethanol (Method II). However, the permeability still remained very high, being in the range of 10m7cm3 (STP) -cm/cm”set-cmHg. The fastest permeation was obtained on a sample modified by Method I using a 1.0 vol.% HFBM/EtOH solution; this led to a P(O,)/P(N,) value of the modified membrane which was very close to the value of PDMS while the P(0,) value was 10 times larger than that of PDMS. Better separation was obtained at oxygen fluxes which were still larger than those of PDMS, making PTMSP/ HFBM a better membrane material than commercial PDMS under all conditions. The permeability coefficients did not vary in a simple manner with fabrication conditions; the relationships between fabrication and performance are in need of further, detailed study. Generally, however, it appears that for a given concentration of HFBM in ethanol, immersion in solution results in a smaller permeability

G Chen et al/J

Membrane Set 82 (1993)99-115

103

TABLE 1 Permeabhty coefficients of nitrogen, P( N, ) , and oxygen, P( O,), and oxygen/nitrogen separation factor, P( 0,) /P( N,) , at 20” C, of PTMSP and CP membranes before and after modlficatlon w&b HFBM Treatment solution (% HFBM m EtOH)

Sample ID

PTMSP PTMSP/HFBM

PTMSP/HFBM

CP CP/HFBM

(S )

0 05 01 05 1 2 10 20 30 100 01 05 1 30 100

P(Qi?

P(N,)”

P(O,)IP(N,)

7 87 2 78 199 107 5 56 143 152 132 163 0 77 139 197 391 7 52 0 91 177

506 105 0 73 0 34 2 74 049 0 49 0 40 0 37 0 16 0 26 0 57 158 4 82 0 27 0 50

155 26 27 31 20 29 31 33 44 48 54 35 25 156 34 35

“m umta of lo-’ cm3 (‘STP)-cm/cm*-aec-cmHg

and a higher separation efficiency compared with fabrication by application on to one side of the membrane. Increasing the W exposure time was thought likely to increase the crosslink density of the incorporated HFBM and thereby decrease the permeation rate, with perhaps an increase in the selectivity. Two samples were fabricated by duplicating the conditions used for the samples that gave the highest permeation rates in Table 1, except that the exposure to UV was prolonged. The first sample was prepared by immersion in a solution of 0.1% HFBM in ethanol, as for the sample which gave a P (0, )/P ( N2 ) value of 5.4, but then exposed for 12 hr. P( OZ) / P(N,) for this membrane was 4.0, with permeation rates P( 0,) of 0.96 x 10m7cm3 (STP)cm/cm2-set-cmHg and P(N,) of 0.24~10~~ cm3 (STP) -cm/cm2-set-cmHg. The second sample was exposed to UV for 13 hr after treatment of PTMSP with pure HFBM; this led to a substantial drop in performance: the oxygen

c&ficient

permeability was reduced by a factor of four, and the separation factor was 2.6, compared with the value of 4.8 for the membrane exposed for 3 hr. Hence, in both cases the effect of increasing the W exposure time was detrimental. In another experiment, pure HFBM was polymerized by W irradiation for 3 hr. The polyHFBM layer thus formed was cast on to the upper surface of a PTMSP membrane, so that a membrane consisting of separate PTMSP and poly-HFBM layers was obtained. The oxygen/ nitrogen separation factor of the bilayer membrane was 3.1, with values of 7.27 x lOma cm3 (STP) -cm/cm2-set-cmHg and 2.35 x lOmacm3 (STP ) -cm/cm2-set-cmHg at 20 ’ C for P (0,) and P ( N2 ) respectively. These permeability coefficients are comparable with those of a PTMSP/HFBM membrane fabricated by applying pure HFBM monomer (Table 1); the latter, however, possessed a separation factor of 4.8. The PTMSP/poly-HFBM bilayer structure thus does not seem to offer performance

104

benefits over HFBM incorporation into PTMSP. Membranes fabricated from the copolymer CP changed in the same way as PTMSP membranes on HFBM incorporation. Table 1 also lists the permeability properties of CP copolymer membranes for oxygen and nitrogen before and after treatment with HFBM. As for PTMSP, the permeability coefficients decreased and the separation efficiency increased. The extent of oxygen enrichment was not as large for modified CP copolymer membranes as it was for modified PTMSP homopolymer membranes when using the same concentration of HFBM/EtOH solution.

Alternative modif~ations of PTMSP membranes Preliminary experiments were done to investigate the potential of other modification approaches. One sample was fabricated using, instead of HFBM, pure hexafluorobutyl acrylate (HFBA), applied from a pipette such that it was spread evenly on the surface of a PTMSP membrane and then polymerized by W irradiation for 3 hr (Method I). Another sample was fabricated using methyl methacrylate (MMA) in the same way. The permeation properties of these membranes are listed in Table 2. For PTMSP/HFBA, the difference in permeation rates compared with PTMSP/ HFBM is small and probably related to fabrication variability rather than intrinsic differences. For PTMSP/MMA, the separation factor was inferior, indicating that the fluorinated structure of the HFBM and HFBA monomers is essential for best performance. This result is in agreement with the generally high affinity for oxygen of fluorinated polymers. We note, therefore, that the oxygen permeability coefficient of the membrane modified by MMA incorporation also was still relatively high, at 1.22 x 10v7 cm3 (SIP)-cm/cm’-set-cmHg, and

G Chen et al /J Membrane Scz 82 (1993) 99-115 TABLE 2 Permeab&y coefficients of nitrogen, P(N,), and oxygen, P(O,), and oxygen/mtrogen separation factor, P(O,)/ P(Nz),at20”C,ofPTMSPmembraneem~ed~thvarIOUB other compounds and treatments Sample ID

WO,)”

WN,)”

P(O,)IP(N,)

PTMSP/HFBA PTMSP/MMA PTMSP W irradiated PTMSP y-nradlated PTMSP/H202 PTMSP/HCOOH PTMSP/HBF, PTMSP/BPFB

0.75 122 9 88 6 84 860 a71 625 2 34

019 043 650 412 555 558 377 133

40 28 15 17 15 16 17 18

4n un1t.eof 10m7cm3 (STP ) -cm/cm’-set-cmHg

the separation efficiency was markedly improved compared with that of the untreated PTMSP membrane. PTMSP membrane samples were exposed to either W light or y-irradiation (Co5’). In the latter case, a new IR absorption band at 1750 cm -’ indicated that a substantial extent of oxidation of PTMSP had taken place, probably as a consequence of reaction of free radicals created by y-irradiation with atmospheric oxygen. These treatments had very little effect on the permeability properties; P( 0,) /P( N,) was determined tobe 1.52 and 1.66, respectively, for the PTMSP samples subjected to the two treatments (Table 2). The oxygen affinity of the membrane might conceivably also improve on partial oxidation of the polymeric structure. Treatment of PTMSP membrane samples with hydrogen peroxide, formic acid, and HBF4, however, affected the permeability properties of the membranes very little. Following these treatments, the values for P( 0,) /P( N2) were between 1.55 and 1.66. Finally, PTMSP was treated with bromopentafluorobenzene (BPFB ) and UV irradiated for 3 hr. This treatment also did not yield promising results, as the oxygen permeability coefficient dropped by a factor of slightly

105

G Chen et al /J Membrane Scr 82 (1993) 99-115

morethan3,andP(0,)/P(Na)wasratherpoor at 1.75 (Table 2).

Temperature dependence of permeation The permeability coefficients of nitrogen and oxygen were determined at several temperatures. The values measured for a PTMSP/ HFBM(S) membrane fabricated by immersion in a 0.1% (v/v) solution in HFBM in ethanol, and the oxygen/nitrogen separation factor P( 0,) /P( N,), at various temperatures are listed in Table 3. Figure 1 shows an Arrhenius plot of the permeability coefficients, from which activation energies, E,, for the permeation of these gases were calculated. Table 3 shows that the P( 0,) value increased and the separation factor decreased with increasing temperature. The activation energy for O2 and N, permeation is positive, with E,=11.6 kJ/ mol for oxygen and E,= 16.4 kJ/mol for nitrogen. With a positive activation energy both for oxygen and nitrogen, the permeability-temperature relationship of PTMSP/HFBM is akin to that typically observed for dense membranes, and differs markedly from results of an earlier study which found the activation energy of permeation through PTMSP of gases such as oxygen and nitrogen to be negative, with E, TABLE 3 Permeabrhty coefticrents of nitrogen, P( N, ) , and oxygen, P(C),), and oxygen/nitrogen separation factor, P(O,)/ P(N,), of a PTMSP/HFBM( S) membrane at varrous temperatures Membrane fabncaton by Immersion m a 0 1 % (v/v) solutron of HFBM m ethanol Temperature (“C)

P(O,)a

P(N,)”

P(O,)/P(N,)

20 30 50 70 90

096 105 149 181 2 36

0 24 0 31 046 0 62 090

40 34 32 29 26

Pn umts of 10m7 cm3 (STP) -cm/cm2-see-cmHg

I

28

29

30

3.1

3.2

33

38

lo3 (KIT)

Frg 1 Arrbenms dragram of the permeabrhty coefficients of nitrogen and oxygen for a PTMSP/HFBM (S) membrane fabncated by rmmerslon m a 0 1% (v/v) solutron of HFBM m ethanol

values of - 1.26 kJ/mol for 0, and - 0.64 kJ/ mol for Nz [ 31.

Stab&y of the modifuzd membranes Several studies [3,9-111 have reported that the permeability of PTMSP membranes decreased rapidly with time after fabrication even when the membranes were kept in vacuum. Such behaviour was not observed for our membranes; they appeared stable over a period of six months stored in air at 20 ’ C (Table 4). The relatively small changes listed in Table 4 probably relate more to instrument drifts over this time and measurement uncertainty than chemical changes in the samples. Two PTMSP/HFBM membranes were redissolved in toluene and membranes were cast again from these solution. The resultant membranes were found to have a permeability to oxygen and nitrogen higher than before re-casting, but the separation efficiency was much reduced, as listed in Table 5. Evidently, the structure of the membrane obtained by HFBM incorporation is not of a homogeneous nature and subject to disturbance by solvents. Although the PTMSP/HFBM membranes can be

G Chen et d/J

106

Membrane Scz 82 (1993) 99-115

TABLE 4 Permeablhty properties determined at 20°C of untreated and HFBM modified membranes, freshly prepared and after SIX months storage m ambient air Sample ID

Storage time (months)

P(0,)”

P(N,)”

P(O,)IP(N,)

PTMSP

0 6 0 6 0 6

7 81 7 25 163 191 177 1 16

5 06 4 64 0 37 046 0 50 034

155 156 44 42 35 34

PTMSP/HFBM CP/HFBM

“m units of 10e7 cm3 (STP)-cm/cm’-set-cmHg TABLE 5 Permeahhty properties determmed at 20 oC of two PTMSP/ HFBM membranes before and after dlssolutlon m toluene and re-castmg Sample ID

P(0,)”

P(N,)”

P(O,)IP(N,)

PTMSP/HFBM 1 before after PTMSP/HFBM 2 before after

0 77 242 0 75 50

016 128 0.186 284

48 19 40 18

4n umta of 10e7 cm3 (STP)-cm/cm’-set-cmHg

stored under ambient atmospheric conditions, they cannot be re-cast without a considerable decrease in separation performance. Spectroscopic analysis of PTMSP membranes modcfied by HFBM Transmission infrared spectra of membranes cast from PTMSP and the copolymer CP are shown in Fig. 2. IR bands assignable to the PTMSP structure are located at 2950,2850, 1540, 1240, and 820 cm-‘. Figure 2(b) also shows a band at 1050 cm-’ which is characteristic of Si-C-Si bonding and assignable to the minor component of the copolymer; although the copolymer contains only 5 mole percent of this material, it is clearly evident in the spectra.

Changes in the IR spectra were observed following the incorporation of HFBM and its polymerization by UV irradiation. The IR spectrum of PTMSP/HFBM membrane is shown in Fig. 3. Comparison with IR spectrum of PTMSP (Fig. 2a) shows several new bands for the modified membrane. The additional band located at 1750 cm-l is assigned to the carbony1 group of HFBM. Bands characteristic of -CF2- are located at 1280 cm-’ and 1320 cm-‘, and a band at 1120 cm-’ is characteristic of -0-CH- structures. Figure 3 indicates successful incorporation of substantial amounts of HFBM into the membrane; IR analysis in the transmission mode would not be expected to possess sufficient sensitivity to detect the bands if the treatment bad only grafted a thm HFBM layer onto the surface of the PTMSP membrane. The methyl methacrylate modified PTMSP membrane had a transmission IR spectrum similar to that of PTMSP/HFBM membranes except for the absence of the absorption bands assigned to the -CF,- structure at 1280 cm-l and 1320 cm-‘. The UV spectrum of HFBM in absolute ethyl alcohol showed an absorption band at 220 to 230 nm. The copolymer CP in cyclohexane absorbed at 250 nm. The W/Vis spectrum of PTMSP dissolved m cyclohexane had only a

G Chen et al/J Membrane

SCL

3000

I

I

I

I

4000

107

82 (1993) 99-115

1600

1600

1200

I

800

1

Wavenumber(cm-‘1

Fig 2 Transmlsslon IR spectra of membranes cast from (a) PTMSP and (b) CP

1

4000

3000

I

I

I

I800

1600

1200

I

600

1

Wavenumber(cm-I)

Fig 3 Transmxwon IR spectra of (a) PTMSP, and (b) PTMSP/HFBM

very small absorption in the UV region, with a maximum at 273 nm and an 6, of 120 I-mol- ‘cm-l, and no absorption above 340 nm. XPS analysis was used to assess the elemen-

membranes

tal composition of the surface of the membranes. At vertical takeoff of the photoelectrons (emission angle of 0 ’ ) , the method probes surface and near-surface layers to a depth of

G Chn et al /J Membrane Set 82 (1993) 99-115

108

BIndIng Energy (eV)

7

-

- 10 nm of the materials assuming an inelastic mean free path of the photoelectrons of 3 nm, which appearg to Ix a rea00nabie vahx GOT polymers. With an emission angle of 60 O,higher surface sensitivity is obtained, with a probe depth of N 2-3 nm, and it is possible to assess

whether the polymer possesses the same composition at the “surface” as in the “bulk”. Emissicm angle6 higher f&n &Q” Cxre& wa?ranted unless the material is very smooth, surface roughness, the finite photoelectron collection aperture, and a reduced signal-to-noise

G Ch-enet al/J

109

Membrane Scl 82 (1993) 99-115

TABLE 6 Composhonal analysis by XPS of PTMSP before and after UV irradiation Irradlatlon (hr)

Surface”

upper

upper upper lower lower lower

Emission angle (degrees)

0 60 0 60 0 60

Composition ( % )

Atomic ratios

C

Sl

0

F

Sl/C

WC

82 2 814 60 6 815 a2 1 82 0

14 2 13 1 88 16 5 15 9 15 5

36 55 306 20 20 25

-

0 0 0 0 0 0

0 044 0 068 0 51 0 025 0 024 0 031

17 16 15 20 19 19

‘See text for explanakon

ratio combine to decrease the accuracy of data collected at high emission angles. XPS spectra of a PTMSP membrane are shown in Fig. 4. Compositional data obtained from the peak intensities (Table 6) were in good agreement with expectations; the Si/C ratio obtained by XPS is identical within experimental error to the theoretical value of 0.167 derived from the presumed PTMSP structure. The small difference in the values obtained at the two emission angles is not significant in the light of an accuracy of between 5 and 10% for XPS compositional analysis in these experiments. In addition to the Si and C signals (hydrogen cannot be detected in XPS) a peak located at 532.8 eV and assigned to oxygen was detected. Evidently, a small extent of oxidation of the PTMSP material occurred. Angle dependent XPS indicated that the concentration of the oxygen containing groups was considerably higher at the surface. PTMSP modification by UV irradiation promoted HFBM incorporation raises the question as to the effects of the HFBM on the membrane as compared with the effects, if any, that the W irradiation may have on PTMSP alone. Accordingly, PTMSP was irradiated in the absence of HFBM and analyzed by XPS; the data are also listed in Table 6. The intensity of the contribution assigned to oxygen increased

markedly on irradiation, attesting to photo-oxidation occurring to a considerable extent. Selective oxidative cleavage and removal of the trimethylsilyl group did not occur as shown by the constancy of the Si/C ratio. It appears that if the formation of oxygen containing species by photo-oxidation results in removal of lower weight fragments, such removal is indiscriminate. At an emission angle of 60’) the O/C ratio was measured to be 0.52 which agrees within experimental error with the value of 0.51 determined at 0” emission. With the above mean escape depths for the photoelectron signals, the absence of an angle dependence indicates that the W induced photo-oxidation penetrated at least 10 nm into the polymer. However, the UV irradiation may not penetrate the full thickness of the PTMSP membrane without noticeable attenuation. We define the surface of the membrane which was exposed directly to the UV light as the “upper surface” and the opposite side of the membrane as the “lower surface”. Table 6 also hsts data taken on the two sides of membranes. Comparison of the PTMSP analysis data shows that on the lower surface, the amount of oxygen is markedly less than on the upper surface and comparable to the amount detected on samples which were not subjected to irradiation. Thus, exposure of PTMSP to UV irradiation &d not

110

G Chen et al /J Membrane SCL82 (1993) 99-115

SI 2p 2s

BIndIng

Energy (eVl

Blndlng Energy [eV) Fig 5 XPS spectra of a PTMSP/HFBM

membrane

(a) Survey spectrum, and (b) C 1s spectral reson

noticeably affect the composition of the lower surface. In agreement with the IR spectra, XPS also confirmed incorporation of HFBM into the PTMSP membranes. Figure 5 reproduces the XPS spectra of one PTMSP/HFBM sample. By comparison with Fig. 4, it is evident that the

exposure of PTMSP to HFBM has given rise to the presence in the modified membrane of an Fls peak. The XPS Cls spectral region of PTMSP/HFBM, shown in Fig. 5 (b), differed from that of PTMSP (Fig. 4b ) by the presence of carbon-oxygen bonds in the modified membrane. On the basis of their spectral positions,

111

G Chen et al /J Membrane Scl 82 (1993) 99-115 TABLE 7 Composltlonal analysrs by XPS of PTMSP/HFBM Method*

Surfaceb

upper lower upper upper lower

Emlsslon (degrees)

0 0 0 60 60

samples nradrated for 3 hr

Composition ( % )

Atomic ratios

C

Sl

0

F

Sl/C

o/c

F/C

62 0 79 9 649 646 818

a4 160 58 66 160

28 5 39 27 9 26 7 20

11 02 14 21 02

0 14 0 20 0 09 0 10 0 20

046 0 05 0 43 0 41 002

0 017 0003 0 022 0 033 0 002

“A annhcatlon of nure HFBM (on to the unner __ __ surface of the PTMSP membrane), B apphcatlon of a solution of HFBM/ EtOH (20/80 v/v) bas for Table 6

the new components to the Cls signal were assigned to C-O and O-C=0 contributions, which is consistent with incorporation of ester groups. The fluorine signal may lead one to assign the two new contributions to the ester group of incorporated HFBM. However, as we have shown above that UV irradiation alone causes incorporation of oxygen containing groups, the new spectral features may also contain a contribution from groups produced by oxidative processes taking place in the course of W irradiation. The two aspects cannot be disentangled at present; the above data relating to UV irradiation of pure PTMSP do not enable assessment of the relative importance of the photooxidative pathway in the irradiation of the mixed PTMSP/HFBM material. Addition of HFBM is likely to decrease the photo-oxidation of the PTMSP component because HFBM is a stronger absorber and will thus partly “screen” the TMSP moiety from UV damage. XPS does not allow determination of the relative amounts of oxygen associated with incorporated HFBM and photo-oxidatlvely produced groups, respectively, because of excessive spectral overlap. The compositional data for PTMSP/HFBM membranes are listed in Table 7. XPS was also used to study whether the ad-

dition of HFBM led to the fabrication of an asymmetric membrane. One may expect that HFBM incorporation by application to one side of the PTMSP membrane is likely to produce a membrane composition that is characterized by a higher relative concentration of HFBM at the upper surface. Table 7 also lists compositional data measured on the lower surface of the same PTMSP/HFBM membrane. It is evident that some penetration of HFBM occurred and the PTMSP membrane became modified throughout its thickness, but it also is clear that the amount of HFBM is considerably smaller at the lower surface. The membrane therefore is asymmetric but not to the extent of being a laminate of PTMSP/HFBM on a layer of remaining, pure PTMSP. Angle dependent data (not listed for brevity) showed that the fluorine content varied slowly; for both surfaces, the percentage of fluorine was similar at 0 ’ and 60’ emission angles. Penetration of HFBM into PTMSP might be facilitated by a solvent. Application of HFBM in ethanol gave, however, similar XPS data to those obtained on a sample prepared by application of pure HFBM (Table 7). On the upper surface, a somewhat higher amount of fluorine was observed but on the lower surface the fluorine concentration remained very small, m-

112

G Chen et al /J Membrane Scr 82 (1993) 99-115

dicating that ethanol did not markedly facilitate the transport of HFBM through PTMSP prior to UV polymerization. Discussion Incorporation of HFBM and HFBA caused considerable increases in the 0,/N, separation efficiency of the PTMSP membranes. The permeability coefficient decreased somewhat on modification, regardless of whether this was done by applying HFBM on to the surface of the membrane or by immersing the membrane in a solution of HFBM. However, the permeability still remained very high, being in the range of 10m7 cm3 (STP) -cm/cm2-set-cmHg, thus achieving separation efficiency of an order reported for other membranes at lower fluxes. As usual, a tradeoff between permeation rate and separation efficiency also exists for PTMSP/ HFBM membranes (Fig. 6 ) , but it appears that they offer outstanding potential in achieving reasonable separation at high flux, and higher separation at flux values which are not much reduced compared with pure PTMSP. In addition, our data document that the

.I

0

I

I

I

I

2

4

6

a

Oxygen

Permeabhty

lb

Coeffwent

6 The oxygen/nitrogen separation efficiency as a function of the oxygen permeahllity (m units of 10e7 cm3 ( STP )-cm/cm2-sec-cmHg) , for ( q ) PTMSP, ( n ) PDMS, (Cl) PTMSP/HFBM, (0) PTMSP/HFBM(S), ( l ) CP, (0 ) CP/HFBM, and ( + ) PTMSP membranes modified m various other ways Fig

modified membranes retained their performance over extended periods of time. The central question is how the incorporation of HFBM alters the PTMSP membrane structure towards stability and higher selectivity. A number of studies have focused on unravelling the reasons for the exceptionally high fluxes of various gases through PTMSP membranes. The high oxygen permeability has been attributed to a very high oxygen solubility [ 231, which is typical for silicone based polymers. There are, however, fundamental differences between PTMSP and PDMS membranes. PTMSP is a glassy polymer under the environmental conditions of most envisaged applications, with a glass transition temperature of about 200°C. Its density of 0.75 g/cm3 is substantially lower than that of other commonly studied membrane polymers. The cast PTMSP membrane contains a substantial number of cavities of about 1 pm diameter [ 9,241; the low density of the membrane is likely to be related to this unusual cavity structure. The permeability coefficient does not vary inversely as the square root of the molecular weight of the permeate [ 31; this indicates that the membrane is non-porous, that is, there are not continuous channels between the cavities, and between the surfaces and cavities. One may thus envisage PTMSP in terms of micrometer-sized cavities which do not contact the membrane surface and are separated from each other by polymeric separating walls; a structure which resembles a foam or a “Swiss cheese”. This structure differs markedly from that of the perfluorinated polymer Nafion, in which the cavities are interconnected by very narrow pores [ 251. Volkov has reported that the cavities occupy N 20% of the volume of PTMSP membranes [ 261; thus, isolated, 1 pm sized cavities are separated on average by solid PTMSP walls of thicknesses of similar magnitude. In Nafion, permeants can permeate through the membrane by migrating along the pores and

G Chen et al/J

Membrane SCL 8.2 (1993) 99-115

cavities, and they need not enter the “bulk” of the polymeric solid. In PTMSP, an analogous migration path does not exist as the voids are not interconnected. Accordingly, dissolution into and transport inside the solids PTMSP walls between cavities will have a major influence on the transport rates. The free volume thus plays a role: random, thermally driven motions of polymer chains open up temporary, sub-nanometre gaps between polymer chains, and it is by way of these free volume gaps that the permeate diffuses through the solid PTMSP phase between cavities. Rapid transport is facilitated by the presence of the easily movable trimethylsilyl side groups and the large temporary spaces between the flexible molecular chains. In addition, we believe that the cavity structure also contributes to the high permeability; between cavities, the permeant traverses relatively thin regions of PTMSP. Hence, a hypothetical straight diffusion path through the membrane may encompass solid material to only about half of the macroscopic membrane thickness. The structure of the unmodified PTMSP membrane is not compact, but with its cavities can be considered to represent an intermediate state between porous (with continuous pores from one membrane surface to the other) and dense membrane structures. The central question now is how HFBM is incorporated into this structure. We have not yet been able to gain sufficient insight into the microstructure of PTMSP/HFBM membranes; analysis of the distribution and microstructure of such small amounts of imbibed material is inherently very difficult. Nevertheless, we now wish to present a putative model for the mode of incorporation of HFBM into PTMSP, a model which we are attempting to assess the work in progress. One possibility is the segregation and polymerization of HFBM in the cavities inside the PTMSP polymer; hence, a heterogeneous structure would be formed, with microdomains of poly-

113

HFBM lodged within a PTMSP matrix. Another possibility is incorporation of I-IFBM into the large free volume created by the large trimethylsilyl side groups in the polymer structure. Such HFBM incorporation into the “bulk” structure, in-between PTMSP chains, is expected to reduce the chain segmental flexibility and the large free volume which facilitates permeation. A third, and in our opinion the most likely, possibility is that some of the HFBM is incorporated into the “bulk” of the PTMSP, thus altering the permeability of the solid walls between cavities, and additional HFBM, in excess of its solubility, separates into the voids and polymerizes there to form a poly-HFBM coating on the surface of the cavities, thus reducing their diameter and effectively creating a bilayer membrane structure in a cavity form. PTMSP membranes have been reported to suffer from deterioration of properties on aging [3,9-111; interestingly, our data (Table 4) are not in agreement with these earlier observations but support another study which found the performance of solvent cast PTMSP membranes to be stable [13]. Various models have been proposed for the deterioration of the performance of PTMSP membranes [9,11,12]. The sizes and amounts of the cavities in the membrane were, however, observed not to change on storage in a vacuum container; the time dependent decrease in the gas permeability was thus related to changes in the fine structure of the PTMSP polymer. It is not clear from the previous work what exactly these changes in the fine structure should be. We postulate, however, that the decrease in permeability with time demonstrated by PTMSP membranes studied by others may be a result of slow crystallization of PTMSP chains on storage of amorphous, freshly cast membranes; crystalline regions possess less flexibility and thus decreased permeability. The rate and extent of such crystallization would depend much on the casting and storage conditions; hence the ap-

114

parent contradictions in the stability of PTMSP membranes prepared by the various groups. The crystallization process is a manifestation of a thermally unstable condition. Our apparently “stable” PTMSP membranes are likely not to be thermodynamically stable either; they are only kinetically stable because the crystallization process has become so slow as not to be manifest on the time scale studied. Such crystallization occurring in the PTMSP “walls” would not markedly affect the void structure; this crystallization hypothesis is therefore also consistent with the experimentally observed retention of cavities on storage. Incorporation of HFBM into the PTMSP free volume would be expected to slow down the rate of crystallization; acting as an impurity, HFBM would interfere with the crystallization process and thus prevent, or at least slow down, the drop in permeability. Observation of such increased stability would provide indirect evidence for HFBM incorporation into the PTMSP bulk structure as opposed to complete HFBM segregation into the voids. As the unmodified membrane itself was stable over six months, however, this model could not be tested within the time frame of the present study but is the subject of ongoing work. Our spectroscopic analyses indicate that the resultant membranes are not homogeneous. XPS analysis provided values of the F/C ratio measured on the upper surface of PTMSP/ HFBM membranes 6 to 10 times larger than the values measured on the lower surface. Thus, HFBM was incorporated nonuniformly through the modified membrane following its application on to the upper surface. Presumably a continuous gradient was obtained, with a higher degree of HFBM incorporation near the upper surface, which was also facing the UV light source. Super-imposed on this variation in HFBM content across the thickness of the membrane is, we believe, a micro-heterogeneity arising from partial segregation of HFBM

G Chen et al/J

Membrane Scr 82 (1993) 99-115

into the voids which exist in the native PTMSP structure. We thus postulate that, as HFBM accumulates and polymerizes both in-between PTMSP chains and in the voids, a semi-interpenetrating polymer network structure is formed between poly-HFBM and PTMSP. Thus, HFBM incorporation modifies the membrane performance in two ways: first, the bulk material itself is modified as the PTMSP walls between cavities accept from HFBM, and second, permeation through the cavities is altered as the segregated HFBM polymerizes inside the cavity. With the relatively small amounts of HFBM incorporated, cavities are not filled but, rather, only receive a coating lining the walls, thus leaving a smaller void. The permeation then becomes a multi-step process through sequences of three different phases: modified PTMSP, pure HFBM lining, and void; the PTMSP/HFBM membrane thus acts as a multi-step separator akin to a membrane consisting of multiple flat layers. The performance of a PTMSP/poly-HFBM bilayer membrane (with no HFBM inside the PTMSP) was inferior, suggesting that a multi-step permeation through repeat void/cavity wall structures is an efficient way to enhance separation compared with a continuous, single, thin enrichment layer. Acknowledgements We thank Prof. T. Higashimura for a sample of the CP copolymer, Prof. T. Masuda for a preprint of Ref. 19, and T. Gengenbach for the XPS analyses. G.C. gratefully acknowledges support under the CSIRO-Academic Sinica Exchange Scheme. References 1

T Masuda, E Isobe, T HlgashnnuraandK Takada, Poly [ l- (tnmethylsllyl)-1-propyne] A new high polymer synthesized with transltlon metal catalysta

115

G Chen et al /J Membrane Scr 82 (1993) 99-115 and characterized by extremely high gas permeability, J Am Chem Sot ,105 (1983) 7473 2 T Masuda, E Isobe and T Hlgashlmura, Polymenzatlon of l- (trunethylsllyl)-1-propyne by halides of mobum(V) and tantalum(V) and polymer properties, Macromol ,18 (1985) 841 3 K Takada, H Matauya, T Masuda and T Hlgashlmura, Gas permeablhty of polyacetylenes carrying substituents, J Appl Polym Scl ,30 (1985) 1605 4 T Masuda, B -2 Tang and T Hlgashlmura, Thermal degradation of polyacetylenes carrymg substltuents, Macromol ,18 (1985) 2369 5 T masuda and T Hlgashlmura, Synthesis of high polymers from substituted acetylenes exploitation of molybdenum- and tungsten-based catalysts, Act Chem Res ,17 (1984) 51 6 T Masuda, B -Z Tang and T Hlgashlmura, Mechanical properties of substituted polyacetylenes, Macromol ,19 (1986) 1459 7 T Masuda, T Tanahashl, A Nlkl and T Hlgashlmura, Polymerization of ahphatic acetylenes X Polymerization of internal octynes and hexynes by halides of mobmm (V) and tantalum (V), J Polym Scl Chem ,24 (1986) 809 8 H K Lonsdale, The growth of membrane technology, J Membrane Scl , 10 (1982) 81 9 H Shlmomura, K Nakamshr, H Odam, M Kurata, T Masuda and T Hlgashlmura, Permeation of gases m poly [ l- (tnmethylwlyl) -1-propyne), Kobunshl Ronbunshu, 43 (1986) 747 10 S Asakawa, Y Salt0 and M Sekl, Gas permeability of poly (1-tr~ethylsllyl-1-propyne) composite membrane as an oxygen separation membrane, Polym Prepr , Jpn ,36 (1987) 1952 11 T Nakagawa, T Salto, S Asakawa and Y Salto, Polyacetylene derwatlzes as membrane for gas separation, Gas Sep Punf ,2 (1988) 3 12 H Odam and T Uyeda, Theories of sorption and transport m polymer membrane, Polym J ,23 (1991) 467 13 M Langsam and L M Robeson, Substituted propyne polymers Part II Effects of agmg on the gas permeability properties of poly [ l- (tnmethylsllyl) -l-propyne] for gas separation membranes, Polym Eng Scl , 29 (1989)44

14

15

16

17

18

19

20 21

22

23

24 25

26

S Asakawa, Y Salto, K Waragai and T Nakagawa, Composite membrane of poly [ 1-trimethylsllyl) -lpropyne] as a potential oxygen separation membrane, GasSep Purlf,3 (1989) 117 S Asakawa, New type oxygen separation membrane with high permeability, New Mater (Japan), 6 ( 1988) 23 M Langsam, Fluorinated polymeric membranes for gas separation processes, US Patent 4 657 564 (April 1987) M Langsam, M Anand and E J Karwackl, Chemical surface modification of poly [ l- (tnmethylsdyl)-lpropyne] for gas separation membranes, Prepr Second Ann NAMS Meeting, Syracuse, NY, June l-3, 1988,~ 49 M Langsam, M Anand and E J Karwackl, Substituted polymers I Chemical surface modification of poly [ l- (trrmethylsllyl)-1-propyne] for gas separation Membranes, Gas Sep Punf ,2 (1988) 162 T Masuda, Poly [ l- (tnmethylalyl) -1-propyne], m D A Turell (Ed ), Macromolecular Syntheses, Vol 11, MRG Polymer Press, Hattlesbury, MS, 1992, p 45 J T Grant, Methods for quantitative analysis m XPS and AES, Surf Interf Anal, 14 (1989) 271 J H Scofield, Hartree-Slater subshell photolomzatlon cross-sections at 1254 and 1487 eV, J Electron Spectrosc Rel Phenom ,8 (1976) 129 M P Seah and WA Dench, Quantitative electron spectroscopy of surfaces A standard data base for electron melastlc mean free paths m sohds, Surf Interf Anal, 1 (1979) 2 Y Ichuaku, S A Stern and T Nakagawa, An mvestlgatlon of the high gas permeability of poly [ l- (trimethylalyl)-l-propyne], J Membrane Scl ,34 (1987) 5 G W Chen, unpublished results T D Glerke and W Y Hsu, The cluster-network model of the ion clustermg m perfluorosulfonated membranes, m A Elsenberg and H L Yeager (Eds ), Perfluormated Ionomer Membranes, ACS Syrnp Ser No 180, American Chemical Society, Washington, DC, 1982,~ 283 V V Volkov, Free volume structure and transport properties of glassy polymers - Materials for separatmg membranes, Polym J ,23 (1991) 457