Journal of Membrane Science 155 (1999) 231±240
Polyetherimide/polyvinylpyrrolidone vapor permeation membranes. Physical and chemical characterization Richard J. Cranford, Hans Darmstadt, Jin Yang, Christian Roy* Institut Pyrovac inc., Parc technologique du QueÂbec meÂtropolitain, 333, rue Franquet, Sainte-Foy, QueÂ., Canada G1P 4C7 Received 29 May 1998; accepted 1 October 1998
Abstract Integrally skinned capillary tube membranes were prepared by the wet-phase inversion method. A series of polyetherimide (PEI, Ultem 1000) membranes were prepared with varying amounts of polyvinylpyrrolidone (PVP) in the casting solution. The surfaces of the membranes were analyzed by electron spectroscopy for chemical analysis (ESCA). It was found that the molecular structure of PEI, both with and without PVP, changes considerably during membrane preparation. The ESCA results indicated that the amount of PEI nitrogen remaining fully imidized at the surface varied in the range 63±86%. The PVP/PEI mass ratio at the membrane surface was found to increase linearly from 0 to 0.10 as the ratio was increased from 0 to 0.43 in the casting solution. The PVP/PEI mass ratio in the membrane bulk was determined by thermogravimetric analysis (TGA) to reach a maximum of 0.067. Vapor permeation experiments were done with a water/n-propanol mixture. The addition of PVP increased the membrane selectivity ( PA/PB, Awater, B1-propanol) from 76 to 810, while the permeance for water remained relatively constant at 1.310ÿ6 mol/m2 s Pa. # 1999 Elsevier Science B.V. All rights reserved. Keywords: Skin; Vapor permeation; Blend membrane; ESCA; Water treatment
1. Introduction Vapor permeation membranes are used to separate gas-phase mixtures by exploiting differences in sorption and diffusion properties of the compounds present. In general, water vapor permeates these membranes at a rate orders of magnitude higher than permanent gases and organic compounds. This permits applications such as water puri®cation, dehydration of organic vapors and permanent gases, latent heat recovery from dryer exhausts, and breaking of azeo*Corresponding author. Tel.: +1-418-656-7406; fax: +1-418656-2091; e-mail:
[email protected]
tropes. Polyimides are attractive polymers for the preparation of these and other types of membranes due to their high glass-transition temperatures. In general, polyimides are insoluble in organic solvents and are cast into membranes in their polyamic acid form. Polyetherimide is an exception and dissolves in some solvents permitting direct casting. PVP has been primarily used to increase casting solution viscosity and to assist other spinning parameters for reverse-osmosis and ultra®ltration membranes [1]. It has also been used to increase selectivity of vapor permeation membranes [2]. Since PVP is water soluble, it is leached to some extent when water is used as the gelation medium. It is important to know
0376-7388/99/$ ± see front matter # 1999 Elsevier Science B.V. All rights reserved. PII: S0376-7388(98)00316-0
232
R.J. Cranford et al. / Journal of Membrane Science 155 (1999) 231±240
the extent of PVP retention in order to improve knowledge of how it increases selectivity performance. Integrally skinned membranes, with a thin skin supported on a porous interior, form in a single casting step as opposed to composite membranes where an additional thin coating is applied. The addition of PVP may increase selectivity in a manner similar to how thin coatings increase selectivity in composite membranes, i.e. by reducing the relative transport rate through large pores or defects at the membrane surface. Since PVP is hydrophilic, its presence may also increase the sorption coef®cient of water. Miyano et al. [3] suggested the possibility of a non-uniform distribution of PVP from the active surface to the underlying support. They showed that PVP is retained at the surface of polyethersulfone (PES, Vitrex) membranes. They used internal re¯ection Fourier transform infrared spectroscopy (FTIR-IRS) which measures the outer 500 nm layer of the membrane surface. This work will use ESCA which probes the outer 5 nm layer and bulk measurements to determine the concentration in the underlying support to attempt to determine if there is indeed a non-uniform distribution.
2.2. Membrane preparation The asymmetric capillary tubes were prepared by the wet-phase inversion technique. Polyetherimide (PEI) and polyvinylpyrrolidone (PVP) were used as polymers. Their structures are shown in Scheme 1. The solvent was N-methylpyrrolidone (NMP), and the coagulation medium was water. The spinning velocity was 10 cm/s and the air-gap height was 10 cm. Therefore the external surface had a one-second residence time in ambient air. The capillary tubes were left in the coagulation medium overnight then rinsed and dried under ambient conditions. The wall thickness was adjusted primarily by changing the ori®ce gap of the spinneret. The capillary tubes had outside diameter of 1.7 mm. A detailed experimental procedure for the membrane preparation can be found elsewhere [4]. Three series of membranes were prepared as shown in Table 1. Membranes HF 52±56 had a constant PVP/ PEI ratio in the casting solution and different total polymer concentrations. Membranes HF 56±60 had constant composition and different wall thickness. Membranes HF 61±64 had constant total polymer concentration and different PVP/PEI ratios. 2.3. Membrane characterization
2. Experimental 2.1. Materials Polyetherimide (PEI, Ultem, 1000) in pellet form was supplied by General Electric Corporation. Nmethylpyrrolidone (NMP), 99% grade and Polyvinlypyrrolidone (PVP, MW ca. 10 000) in powder form were supplied by Aldrich.
Thermogravimetric analysis (TGA) of the membrane samples was performed with a thermobalance SSC/5200H TG/DTA(220) from Seiko. In this series of experiments, samples of approximately 5 mg of the membrane were heated from room temperature to a ®nal temperature of 9008C at a heating rate of 58C/min under a nitrogen atmosphere. By ®tting differential thermogravimetric (DTG) curves to the curves of pure
Scheme 1. PVP and PEI chemical structures.
R.J. Cranford et al. / Journal of Membrane Science 155 (1999) 231±240
233
Table 1 Composition of the casting solution and wall thickness of the membranes Membrane
HF HF HF HF HF HF HF HF HF HF HF HF HF
52 53 54 55 56 57 58 59 60 61 62 63 64
Casting solution composition NMP (mass%)
PEI (mass%)
PVP (mass%)
79.0 80.0 81.0 82.0 83.0 83.0 83.0 83.0 83.0 80.0 80.0 80.0 80.0
18.50 17.62 16.74 15.86 14.98 14.98 14.98 14.98 14.98 14.00 16.00 18.00 20.00
2.50 2.38 2.26 2.14 2.02 2.02 2.02 2.02 2.02 6.00 4.00 2.00 0.00
PVP and a pure PEI membrane (sample HF 64), the bulk concentrations of the two polymers were obtained. The method and the mathematical treatment have been described by Yang et al. [5]. ESCA measurements were done with an ESCALAB MK II spectrometer equipped with a Microlab system from Vacuum Generators. The capillary tube membranes were slit open and pressed ¯at in order to perform ESCA experiments on the internal and external surfaces. The ESCA experiments were performed at a temperature of ÿ608C and a pressure of 10ÿ7 Pa. The speci®c surface area and the pore-size distribution were obtained by nitrogen adsorption at 77 K. An OMNISORP 100 apparatus from OMICROM was used for the measurements. Micrographs were taken with a JSM-840A scanning electron microscope from JEOL. The separation performance of the membranes was determined with a mixture of water and n-propanol (1:1 by mass) in the gas phase. Capillary tube bundles with surface area of approximately 100 cm2 were used. Testing was done at 858C with feed at the bore side of the capillary tubes. The composition of the feed mixture, the retentate and the permeate were determined by refractometry and total carbon analysis with a Beckman model 915A TOC analyzer. Feed and permeate pressures were 33 and 7 kPa, respectively. The membranes were characterized with the solution± diffusion model.
Wall thickness (mm)
Specific surface (m2/g)
170 200 170 139 136 228 72 77 93 155 120 130 129
33.5 35.3 40.9 45.5 41.4 37.9 50.5 ± 44.0 39.0 50.4 42.5 34.6
3. Results and discussion 3.1. Bulk composition The bulk composition of the membranes was determined by thermogravimetry. PVP and PEI produce different decomposition peaks in the DTG curves. By analysis of the area of the various peaks, the composition of the membrane samples can be determined. PEI is more thermally stable and decomposes at higher temperature than PVP as shown in Fig. 1. The differentiated weight loss curve of the aliphatic PVP showed a maximum at 4288C. The differentiated weight loss curve of the as-received PEI pellets showed a maximum at 5248C. The PEI membrane decomposed at a slightly lower temperature than the pellet and showed two maxima at 5088C and 5748C. There are two possible explanations for the differences between the DTG curves of the PEI samples. First, the morphology of the PEI membrane is porous and the asreceived PEI pellets (10 mg/pellet) are dense. Therefore, the PEI membrane has less mass-transfer resistance. Secondly, the ESCA spectra indicated that the PEI chemistry changed during membrane preparation (see Section 3.2). In the thermogravimetric rate versus temperature curves for membranes HF 61±63, a maximum in the peak is present corresponding to PVP and two maxima corresponding to PEI in the membrane. Since the pure PVP sample was in powder form, its morphology was
234
R.J. Cranford et al. / Journal of Membrane Science 155 (1999) 231±240
Fig. 1. Thermogravimetric rate versus temperature for PVP and PEI.
more similar to the porous membrane than the PEI pellets. The samples had maxima in their DTG curves corresponding to PVP of 4228C which was only slightly lower than the maximum for PVP (4288C). The composition of the membranes determined by thermogravimetry for the bulk is given in Table 2. Extrapolation of the data to the origin by ®tting it with a smooth curve shows that nearly all of the PVP is retained for low PVP/PEI ratios. However, for high PVP/PEI ratios, extrapolation of the data shows that a maximum of 0.070 is attained. 3.2. Surface composition Based on the ESCA data the determination of the PVP/PEI ratio at the interior and exterior membrane surfaces was attempted by four methods. The ®rst method is based on the different elemental composition of the two polymers. The nitrogen conTable 2 Determined PVP/PEI ratio in the membrane bulk and at the surfaces Membrane
HF 61 HF 62 HF 63
PVP/PEI mass ratio Casting solution
Bulk
Internal surface
External surface
0.429 0.250 0.111
0.067 0.063 0.055
0.104 0.055 0.028
0.096 0.063 0.035
centration in PVP (12.5%) is considerably higher than in PEI (4.4%). Thus it is possible to calculate the concentration of the two polymers on the membrane surface from the nitrogen concentration on the surface. Methods 2, 3 and 4 are based on the different chemical nature of the nitrogen, oxygen and carbon atoms in the two polymers. Values determined by Beamson and Briggs [6] for pure PVP and pure PEI were compared with results obtained with the membranes. Method 2 was based on the resolution of the N1s components. PEI and PVP nitrogen have different binding energy (400.4 and 399.9 eV, respectively). Method 3 was based on resolution of the O1s components. PEI has two components (531.9 and 533.5 eV), and PVP has one (531.3 eV). Method 4 was based on resolution of the C1s component into two curves: one curve for the combined carbon contribution from PEI and one curve for the combined carbon contribution from PVP. 3.2.1. ESCA elemental analysis Method 1, based on the elemental analysis of the surface (Table 3), gave higher than theoretical nitrogen concentrations. This is probably largely due to inherent inaccuracy of the measurements, but may also be partly due to a non-homogeneous composition. Since the ESCA sampling depth is only approximately 5 nm, it is sensitive to orientation effects. Therefore, this method may yield incorrect results if the elemental composition changes with increasing distance from
R.J. Cranford et al. / Journal of Membrane Science 155 (1999) 231±240 Table 3 Elemental composition of the membrane surfaces C (%)
O (%)
N (%)
PEI (theoretical) PVP (theoretical)
82.2 75.0
13.3 12.5
4.5 12.5
HF HF HF HF
61 62 63 64
int. int. int. int.
78.4 79.0 79.4 77.0
14.8 15.0 14.5 16.8
6.8 6.0 6.1 6.2
HF HF HF HF
61 62 63 64
ext. ext. ext. ext.
77.0 79.3 79.1 78.9
16.3 15.3 15.2 15.7
6.7 5.4 5.7 5.4
the surface in the volume probed by ESCA. The fact that the oxygen concentrations were determined to be higher than the theoretical values is thought to be due to the presence of excess oxygen from hydrolysis or water adsorption. Another source of error with this method is due to the fact that the sampling depth differs slightly from element to element. 3.2.2. ESCA nitrogen spectra The N1s spectrum of PEI shows one peak for imide nitrogen at 400.4 eV [6]. In the N1s spectra of the membrane which was prepared from a PEI solution without PVP (sample HF 64), this peak and two additional peaks were found (see Table 4). The presence of the two additional peaks at 399.1 (N1) and
235
399.9 eV (N2) indicates that a part of PEI reacted during membrane preparation. Polyimides are known to be susceptible to hydrolysis which can lead to isoimide formation and chain scission. In addition, there is a mobile imide±isoimide equilibrium, and the equilibrium fraction of isoimide decreases as temperature increases [7]. A possible rearrangement mechanism for transformation of the imide nitrogen in PEI to isoimide nitrogen is shown in Scheme 2. A peak with a BE similar to the N1 peak is often found in the spectra of polyimides and was assigned to nitrogen atoms in isoimide groups [8]. Chain termination amine groups may also contribute to the N1 peak. The N2 peak has a BE which is typical for nitrogen atoms next to C=O groups. In the case of PEI, the N2 peak can be assigned to amic acid groups shown in Scheme 2. Residual NMP solvent would contribute to the N2 peak. However, the thermogravimetric curves showed no weight loss due to solvent for any of the membranes, so the contribution of residual NMP solvent to the N2 peak was considered negligible. The N1s spectrum of PVP only shows an N2 peak [6]. In the N1s spectra of the membranes prepared from casting solutions with different PVP/PEI ratios, the area of the N2 peak increased with increasing PVP/ PEI ratio in the casting solution. In order to determine the PVP/PEI ratios, it was assumed that in all these membranes the contribution of PEI to the N2 peak was the same. The PEI contribution to the N2 peak was calculated to be 9.1% from an iterative procedure
Table 4 Relative area of ESCA N1s peaks for membranes prepared from casting solutions with different PVP/PEI ratios Sample
PVP/PEI mass ratio Casting solution
Relative peak area (%) Membrane surface
PEI (theoretical) PVP (theoretical)
N1 (399.1 eV) 0 0
N2 (399.9 eV) 0 100
N3 (400.4 eV) 100 0
HF HF HF HF
61 62 63 64
int. int. int. int.
0.429 0.250 0.111 0
0.08 0.04 ± 0
14.9 13.8 12.2 24.4
28.2 18.7 1.6 13.1
56.9 67.5 86.2 62.5
HF HF HF HF
61 62 63 64
ext. ext. ext. ext.
0.429 0.250 0.111 0
0.04 0.03 0.01 0
19.8 15.2 19.0 18.7
20.1 18.0 11.8 9.1
60.1 66.8 69.2 72.2
18.2
9.1
72.7
PEIa a
Average altered PEI contribution at the membrane surface.
236
R.J. Cranford et al. / Journal of Membrane Science 155 (1999) 231±240
Scheme 2. PEI rearrangement mechanism.
using all the samples. This was done by ®rst estimating the PEI contribution to the N1 and N2 peaks to determine the PVP concentrations which in turn was used to determine the average PEI contribution to the N1 and N2 peaks, etc. 3.2.3. ESCA oxygen spectra In addition to additional nitrogen peaks and broadening of existing peaks, PEI reactions lead to additional types of oxygen which consequently makes resolution of the O1s component of PVP dif®cult. The O1s component was resolved into three peaks. The ®rst peak (531.8 eV) included both the PEI and PVP carbonyl oxygen and was used to estimate the PVP concentrations. The O1s spectra of PEI shows two peaks at 531.87 and 533.33 eV for the oxygen atoms in >N±CO± and C±O±C groups, respectively [6]. Theoretically, the ratio of their areas should be 2:1. However, a lower ratio (1.5) was reported for the spectrum of PEI [6]. For PVP, the O1s spectrum showed a peak at 531.3 eV for its oxygen (>N±CO± group), a second peak at approximately 532.7 eV was assigned to adsorbed water. In this work, the oxygen spectra of the membranes were ®tted to three peaks: a peak for oxygen in >N± CO± groups (O1, BE531.8 eV), a peak for adsorbed water (O2, BE532.6 eV) and a peak for oxygen in C±
O±C groups (O3, BE533.5 eV). As outlined above, the ESCA N1s spectra suggest that imide groups in PEI react to form isoimide. This rearrangement also in¯uences the O1s spectra. The oxygen in the isoimide group between the C=O and C=N± group probably contributes to the O3 peak and the C=O oxygen in the isoimide group is probably shifted to higher BE and contributes to the O2 peak. For the spectra of the internal and external surfaces of the pure PEI membrane, HF 64, the ratio of the O1 / O3 peak areas was close to the theoretical value of 2 (2.1 and 1.7, respectively, from Table 5). In the spectra of the membranes prepared from a casting solution containing PEI and PVP, the area of the O1 peak increased with increasing concentration of PVP in the casting solution. Since the PVP portion contributes only to the O1 peak, it is possible to determine the surface concentrations of the two polymers. A summary of the results from the four methods for the determination of the surface composition can be found in Table 2. The results indicate that the PVP concentrations at the interior and exterior surfaces are nearly equal, and that the concentration increases fairly linearly with an increase in the concentration of PVP in the casting solution. The results indicate that for high PVP/PEI ratios in the casting solution, there is greater PVP retention at the surface than in the bulk, but for low concentrations the opposite is true.
R.J. Cranford et al. / Journal of Membrane Science 155 (1999) 231±240
237
Table 5 Relative area of ESCA O1s peaks for membranes prepared from casting solutions with different PVP/PEI ratios Sample
PVP/PEI mass ratio Casting solution
Relative peak area (%) Membrane surface
PEI (theoretical) PVP (theoretical)
O1 (531.8 eV) 67 100
O2 (532.6 eV) 0 0
O3 (533.5 eV) 33 0
HF HF HF HF
61 62 63 64
int. int. int. int.
0.429 0.250 0.111 0
0.10 0.06 0.01 0
62.3 59.6 54.7 57.4
12.2 9.2 15.1 15.4
25.5 31.2 30.2 27.2
HF HF HF HF
61 62 63 64
ext. ext. ext. ext.
0.429 0.250 0.111 0
0.12 0.09 0.02 0
63.9 61.9 56.3 50.7
11.9 12.6 13.0 19.4
24.2 25.5 30.7 29.9
60 54.0
0 15.7
40 30.3
PEIa PEIb a
Values determined for fully imidized PEI by Beamson and Briggs [6]. Average altered PEI contribution at the membrane surface.
b
There are a number of factors which must be considered to explain this result. Firstly, the coagulation rate is higher at the surface than in the interior, and therefore contracting PEI macromolecules would entrap a greater amount of PVP macromolecules before they diffuse out of the polymer rich phase. Secondly, at the surface, there is a higher driving force for diffusion since the forming membrane is in direct contact with the coagulation medium, water. Therefore, PVP retention would be lower than in the interior. Thirdly, for these experiments the membrane pore volume increased from 0.72 to 0.83 as the PVP/PEI ratio increased from 0 to 0.43. Therefore, this factor would cause relatively lower resistance to leaching PVP from the interior for higher PVP/PEI ratios. In addition, during membrane spinning, the bore water is under pressure which may lead to convective type diffusion of PVP out through the external surface of the forming capillary tube membranes. 3.3. Membrane bulk morphology The scanning electron micrographs showed that the membranes are composed of spherical nodule aggregates with a diameter of approximately 100 nm. SEMs of the membrane cross sections showed the presence of radial channels formed extending inward from both surfaces. The inner skin was composed of a very thin
(10 nm) dense polymer matrix supported on a monolayer of tightly packed nodule aggregates in turn supported by more loosely packed nodule aggregates. The interior of the membrane had a porous spongelike structure, which consisted of fused nodule aggregates. The outer skins were generally denser and thicker than the inner skins. SEM shows that nodule aggregates are the basic building units of the membranes. A model based on this observation was used for a nitrogen adsorption study. If one assumes that the whole membrane consists of spherical nodule aggregates and the whole external surface of these aggregates is accessible, the speci®c surface area, A, of the membrane is given by A
3 ; r
(1)
where is the density of the polymer and r the radius of the nodule aggregate. The SEM experiments showed that the radius of a nodule aggregate was approximately 50 nm. The density of PEI is 1.27 g/ cm3. With these values, a speci®c surface area for the membrane of 47.2 m2/g is obtained. This value was close to the experimental value (Table 1). This suggests that most of the external surface of the nodule aggregates was accessible to the nitrogen, otherwise a smaller speci®c surface would have been obtained. Furthermore, a signi®cant amount of pores between
238
R.J. Cranford et al. / Journal of Membrane Science 155 (1999) 231±240
the nodules of the nodules aggregates would have caused a higher speci®c surface area and can be ruled out. Thus, a reasonable model for the membrane structure is one of non-porous nodule aggregates which are fused together. The speci®c surface area of the membranes increased from 33.5 to 45.5 m2/g when the total polymer concentration in the casting solution, at a ®xed PVP/PEI ratio of 0.135, decreased from 21 to 18 wt%. However, when the polymer concentration was further reduced to 17 wt%, the surface area decreased to 41.4 m2/g. A correlation was found between the diameter of the nodule aggregates, seen with SEM micrographs, and the speci®c surface. The correlation is in approximate agreement with Eq. (1). In contrast to our results, Bodzek [9] found that the speci®c surface area increased as the polymer concentration increased for both polysulfone and polyacrylonitrile membranes. The speci®c surface area also was found to depend on the PVP/PEI ratio in the casting solution. The speci®c surface area increased from 34.6 to 50.4 m2/g when the PVP/PEI ratio increased from 0 to 0.25, and the total polymer concentration was maintained constant. However, a further increase of the PVP/PEI ratio to 0.428 resulted in a smaller speci®c surface area of 41.1 m2/g. Since only about 25% of the PVP was retained and the total polymer concentration was maintained constant for this series, an increase in the PVP/PEI ratio also implies a reduction in the ®nal polymer concentration. Therefore this result is similar to the result above for the series HF 52±56. Fig. 2 shows the variation in the bulk pore size distribution for these membranes. It is likely that pores determined by nitrogen adsorption are the interstices of nodule aggregates [10]. The broader distribution for HF 61 and its deviation from the trend in speci®c surface suggest that it has a slightly different structure than the other membranes in this series. As the membrane wall thickness decreased, the speci®c surface increased as seen in the data in Table 1. Since the average distance from the surface decreases as the wall thickness decreases, this result indicates that the membrane speci®c surface is higher near the membrane surfaces. This result is consistent with the fact that slow-phase separation which occurs in the interior leads to tighter nodule aggregate fusion, as shown by SEM, and hence lower speci®c surface.
Fig. 2. Pore size distribution for membranes with different PVP concentration in the casting solution.
Extrapolation of the data to low wall thickness indicates much higher speci®c surface near the membrane surfaces. 3.4. Membrane permeability The permeation rates for both water and n-propanol decreased with increasing total polymer concentration in the casting solution as shown in Table 6. The membrane selectivity, the ratio of the permeances, increased with increasing polymer concentration. These results are as expected due to a reduction in the number of large pores or defects at the membrane skin and increased skin thickness with increasing polymer concentration in the casting solution. As can be seen in Table 7, the permeance for n-propanol decreased by an order of magnitude while the permeance for water remained relatively constant as the PVP/PEI ratio increased. 4. Conclusion PVP was found to be retained in both the outer 5 nm region near the surfaces and in the membrane bulk.
R.J. Cranford et al. / Journal of Membrane Science 155 (1999) 231±240
239
Table 6 Permeation rates as a function of the total polymer concentration (PVP/PEI ratio constant) Membrane
Total polymer in casting solution (mass%)
Water permeance (10ÿ7 mol/m2 s Pa)
N-propanol permeance (10ÿ9 mol/m2 s Pa)
HF HF HF HF HF
17 18 19 20 21
19.6 14.3 9.7 8.8 7.5
107 43.5 5.7 1.7 1.6
56, HF 57 55 54 53 52
Table 7 Permeances as a function of PEI/PVP ratio Membrane
PVP in casting solution (mass%)
PEI in casting solution (mass%)
Water permeance (10ÿ7 mol/m2 s Pa)
N-propanol permeance (10ÿ9 mol/m2 s Pa)
HF 64 HF 63 HF 62
0 2 4
20 18 16
12.8 14.2 12.1
16.8 3.7 1.5
This ®nding together with the results of Miyano et al. [3], who found similar concentrations in the outer 500 nm region, implies that there is a fairly uniform distribution of PVP from the active surface layer to the underlying support. However, Miyano's results involved ultra®ltration membranes which have a more porous skin. Their results may not necessarily be valid for less porous integrally skinned membranes as those studied in this paper. There may be larger concentration of PVP immediately below the outer skin trapped by molecular sieving effects during membrane casting. In addition, there may be microdomains of high PVP concentration in the membrane and at the surfaces. Polyimide vapor permeation membranes can be made with high water permeation rates and excellent selectivities and thermal resistance. In addition, polyimides exist which are insoluble in organic solvents which is a practical necessity for applications involving treatment of organic vapors on an industrial scale. However, further work needs to be done to determine the hydrolytic stability of polyimide membranes. Acknowledgements The authors are grateful to Dr. Serge Kaliaguine of Universite Laval in QueÂbec for use of analytical
equipment; to Dr. Takeshi Matsuura at the University of Ottawa for his scienti®c collaboration; and to Dow Chemical and Ciba-Geigy for gratuitously supplying epoxy resin and hardener, respectively. References [1] I. Cabasso, E. Klein, J.K. Smith, Polysulfone hollow fibers. I. Spinning and properties, J. Appl. Polym. Sci. 20 (1976) 2377. [2] R.J. Cranford, C. Roy, T. Matsuura, Vapour permeation applied for the separation of water from organic compounds and gases using asymmetric polyetherimide/polyvinylpyrrolidone capillary tubes, Can. J. Chem. Eng. 75 (1997) 471. [3] T. Miyano, T. Matsuura, D.J. Carlson, S. Sourirajan, Retention of polyvinylpyrrolidone swelling agent in the poly(ether p-phenylenesulfone) ultrafiltration membrane, J. Appl. Polym. Sci. 41 (1990) 407. [4] R.J. Cranford, Vapour permeation applied to the vacuum pyrolysis process for the separation of water from organic compounds, Ph.D. Thesis, Universite Laval, DeÂpartement de geÂnie chimique, QueÂbec, Canada, 1996. [5] J.S. Yang, S. Kaliaguine, C. Roy, Improved quantitative determination of elastomers in tire rubber by kinetic simulation of DTG curves, Rubber Chem. Technol. 66 (1993) 213. [6] G. Beamson, D. Briggs, High Resolution XPS of Organic Polymers, Wiley, New York, 1992. [7] F.P. Gay, C.E. Berr, Polypyromellitimides: details of pyrolysis, J. Polym. Sci., Part A-1 6 (1968) 1935.
240
R.J. Cranford et al. / Journal of Membrane Science 155 (1999) 231±240
[8] P.L. Buchwalter, A.I. Baise, ESCA analysis of PMDA±ODA polyimide, in: K.L. Mittal (Ed.), Polyimides, Synthesis, Characterization, and Applications, Plenum Press, New York, 1984, pp. 537±545.
[9] M. Bodzek, Physico-chemical characterization of ultrafiltration membranes, Polish J. Chem. 57 (1983) 919. [10] L. Zeman, G. Tkacik, Pore volume distribution in ultrafiltration membranes, Polym Mater. Sci. Eng. 50 (1984) 169.
