Highly Permeable Matrimid®/PIM-EA(H2)-TB Blend Membrane for Gas Separation

The effect on the gas transport properties of Matrimid®5218 of blending with the polymer of intrinsic microporosity PIM-EA(H2)-TB was studied by pure and mixed gas permeation measurements. Membranes of the two neat polymers and their 50/50 wt % blend were prepared by solution casting from a dilute solution in dichloromethane. The pure gas permeability and diffusion coefficients of H2, He, O2, N2, CO2 and CH4 were determined by the time lag method in a traditional fixed volume gas permeation setup. Mixed gas permeability measurements with a 35/65 vol % CO2/CH4 mixture and a 15/85 vol % CO2/N2 mixture were performed on a novel variable volume setup with on-line mass spectrometric analysis of the permeate composition, with the unique feature that it is also able to determine the mixed gas diffusion coefficients. It was found that the permeability of Matrimid increased approximately 20-fold with the addition of 50 wt % PIM-EA(H2)-TB. Mixed gas permeation measurements showed a slightly stronger pressure dependence for selectivity of separation of the CO2/CH4 mixture as compared to the CO2/N2 mixture, particularly for both the blended membrane and the pure PIM. The mixed gas selectivity was slightly higher than for pure gases, and although N2 and CH4 diffusion coefficients strongly increase in the presence of CO2, their solubility is dramatically reduced as a result of competitive sorption. A full analysis is provided of the difference between the pure and mixed gas transport parameters of PIM-EA(H2)-TB, Matrimid®5218 and their 50:50 wt % blend, including unique mixed gas diffusion coefficients.


Introduction
Significant progress has been made in the development of new polymers for the fabrication of gas separation membranes. Polymers used for gas separation membranes can be either rubbers, generally characterized by a relatively low selectivity but high permeability and thus enabling a great productivity, or glassy polymers, characterized by low permeability but high gas-pair selectivity and offering a good separation efficiency. This trade-off between permeability and selectivity, first introduced [1], and updated by Robeson [2] and then by Pinnau et al. [3] is typical for

Membranes Characterization
Chemical and morphological analysis of membranes were performed by scanning electron microscopy (SEM) on a Phenom Pro X desktop SEM, equipped with backscattering detector (Phenom-World B.V., Eindhoven, The Netherlands) and infrared spectroscopy (FTIR) analyses on a Spectrum Spotlight Chemical Imaging Instrument (PerkinElmer). Single gas permeation tests were carried out at 25 °C and at a feed pressure of 1 bar, using a fixed-volume pressure increase instrument Ref. [37] and [42], respectively.
The aim of this work to enhance the permeability of Matrimid by the addition of PIM-EA(H 2 )-TB, and to find the desired combination of the high permeability of the PIM and the high selectivity of the polyimide. Detailed analysis of the gas transport parameters under single and mixed gas permeation conditions provides deep insight into the role of gas diffusivity and solubility in the overall transport properties of the novel Matrimid ® 5218/PIM-EA(H 2 )-TB blend. In particular, a novel mixed gas permeation setup with the unique possibility to determine the mixed gas diffusion coefficients will provide unprecedented information on the coupling effect between CO 2 and CH 4 or CO 2 and N 2 during permeation of the respective mixtures in the neat polymers and the blend.

Materials
Matrimid ® 5218 was kindly supplied by Huntsman (Basel, Switzerland) and PIM-EA(H 2 )-TB was prepared as described previously [37] and the purified polymer, isolated as a powder, was used without any further treatment.

Preparation of Matrimid ® 5218/PIM-EA(H 2 )-TB Blend Membranes
The casting solutions of both pure polymers were prepared at a concentration of 2 wt % of the polymer in dichloromethane (DCM). Homogenous solutions were obtained under magnetic stirring overnight. The blend solution was prepared by mixing equal amounts of the two individual solutions. The blend membrane containing 50 wt % of both Matrimid ® 5218 and PIM-EA(H 2 )-TB was prepared pouring the appropriate amount of the blend solution into a metallic casting ring of 3 cm diameter, placed on a Teflon ® support. The solvent was evaporated at room temperature for 24 h, yielding optically defect-free membranes Figure 1.

Membranes Characterization
Chemical and morphological analysis of membranes were performed by scanning electron microscopy (SEM) on a Phenom Pro X desktop SEM, equipped with backscattering detector (Phenom-World B.V., Eindhoven, The Netherlands) and infrared spectroscopy (FTIR) analyses on a Spectrum Spotlight Chemical Imaging Instrument (PerkinElmer). Single gas permeation tests were carried out at 25 • C and at a feed pressure of 1 bar, using a fixed-volume pressure increase instrument (ESSR, Geestchacht, Germany), described elsewhere [43]. Permeability coefficients, P, and diffusion coefficients, D, were determined by the time-lag method [44]. The simplest model of permeation through dense polymeric films describes permeability as the product of diffusion coefficient and solubility coefficient. Thus, the apparent solubility, S, was indirectly calculated as S = P/D. The ideal selectivity is the ratio of permeability of two species, α (A/B) = P A /P B . Mixed gas permeation tests were carried out using a custom made constant pressure/variable volume instrument, described elsewhere [45,46], equipped with a quadrupole mass filter (HPR-20 QIC Benchtop residual gas analysis system, Hiden Analytical). Measurements were carried out as a sequence of increasing and subsequently decreasing pressure steps in the range from 1-6 bar(a).

Chemical and Morphological Characterization
The chemical interaction between Matrimid ® 5218 and PIM-EA(H 2 )-TB was studied by means of FTIR-ATR. The IR-spectrum for neat Matrimid ® 5218, neat PIM-EA(H 2 )-TB and for the blend were shown in Figure 2. The distinctive imide peaks of Matrimid ® 5218 appear at 1712 for C=O stretching and at 1361 cm −1 for C-N stretching, these peaks are also found in the spectrum of the blend. The characteristic peaks of PIM-EA(H 2 )-TB are the CH 2 asymmetric stretch vibrations of the ethanoanthracene (EA) unit at 2960 cm −1 and the scissoring vibrations at 1420 cm −1 . The water peak (3370 cm −1 ) in the neat PIM-EA(H 2 )-TB and in the blend demonstrate the relatively hydroscopic nature of TB-PIMs [23]. The good compatibility of Matrimid ® 5218 and PIM-EA(H 2 )-TB was also deduced from the high optical transparency of the film and high mechanical resistance. The cross-sectional SEM image shows very few separate domains (less than 1% of the area) confirming the excellent miscibility between PIM-EA(H 2 )-TB and Matrimid ( Figure 2). (ESSR, Geestchacht, Germany), described elsewhere [43]. Permeability coefficients, P, and diffusion coefficients, D, were determined by the time-lag method [44]. The simplest model of permeation through dense polymeric films describes permeability as the product of diffusion coefficient and solubility coefficient. Thus, the apparent solubility, S, was indirectly calculated as S = P/D. The ideal selectivity is the ratio of permeability of two species, α(A/B) = PA/PB. Mixed gas permeation tests were carried out using a custom made constant pressure/variable volume instrument, described elsewhere [45,46], equipped with a quadrupole mass filter (HPR-20 QIC Benchtop residual gas analysis system, Hiden Analytical). Measurements were carried out as a sequence of increasing and subsequently decreasing pressure steps in the range from 1-6 bar(a).

Chemical and Morphological Characterization
The chemical interaction between Matrimid ® 5218 and PIM-EA(H2)-TB was studied by means of FTIR-ATR. The IR-spectrum for neat Matrimid ® 5218, neat PIM-EA(H2)-TB and for the blend were shown in Figure 2. The distinctive imide peaks of Matrimid ® 5218 appear at 1712 for C=O stretching and at 1361 cm −1 for C-N stretching, these peaks are also found in the spectrum of the blend. The characteristic peaks of PIM-EA(H2)-TB are the CH2 asymmetric stretch vibrations of the ethanoanthracene (EA) unit at 2960 cm −1 and the scissoring vibrations at 1420 cm −1 . The water peak (3370 cm −1 ) in the neat PIM-EA(H2)-TB and in the blend demonstrate the relatively hydroscopic nature of TB-PIMs [23]. The good compatibility of Matrimid ® 5218 and PIM-EA(H2)-TB was also deduced from the high optical transparency of the film and high mechanical resistance. The crosssectional SEM image shows very few separate domains (less than 1% of the area) confirming the excellent miscibility between PIM-EA(H2)-TB and Matrimid ( Figure 2).

Pure Gas Transport Properties
Single gas permeation measurements were carried out in the order He, H2, O2, N2, CH4 and CO2 at 25 °C on time-stabilized membranes [47]. Figure 3 shows a plot of the gas transport parameters of 30 days aged PIM and blend membranes, and a >1 year aged Matrimid membrane as a function of the membrane composition. The quantitative values are reported in Table 1

Pure Gas Transport Properties
Single gas permeation measurements were carried out in the order He, H 2 , O 2 , N 2 , CH 4 and CO 2 at 25 • C on time-stabilized membranes [47]. Figure 3 shows a plot of the gas transport parameters of 30 days aged PIM and blend membranes, and a >1 year aged Matrimid membrane as a function Matrimid ® 5218/PIM-EA(H2)-TB blend than in neat Matrimid membrane, thanks to the higher free volume of the PIM.
The gas permeability and diffusivity are greater for the neat PIM, while there is little difference in solubility between the blend and the neat PIM. The effect of composition is largest for diffusivity, which increases about two orders of magnitude from Matrimid to PIM, whereas the solubility increases less than an order of magnitude. The time lag of H 2 and He is too short to be measured accurately in the relatively thin neat Matrimid ® 5218 membrane, and thus the related diffusion coefficient and solubility of this membrane could not be determined. The combined effect of S and D is reflected in an exceptionally high permeability of the neat PIM membrane, with an approximately three orders of magnitude higher N 2 and CH 4 permeability compared to the Matrimid membrane. On the other hand, the selectivity is generally higher in Matrimid ® 5218 ( Figure 3b and Table 1), especially for gas pairs with very different kinetic diameters, like H 2 /N 2 , He/N 2 , mainly as a result of the much higher diffusion selectivity (Figure 3d). The O 2 /N 2 selectivity is higher in Matrimid ® 5218 due to a slightly higher diffusion selectivity and solubility selectivity, whereas the higher CO 2 /N 2 selectivity in Matrimid ® 5218 must be ascribed mainly to the higher solubility selectivity (Figure 3e).
The blend membrane with 50 wt% of each polymer exhibits intermediate properties with respect to the two individual polymers, with a roughly linear trend in permeability on a logarithmic scale. This trend suggests that the two polymers have good compatibility and form a homogeneous blend, because Robeson [48] anticipated that that the permeability, P b , of a homogeneous polymer blend can be expressed as: where φ 1 and φ 2 are the volume fractions of the two polymers in the blend, and P 1 and P 2 are their respective permeabilities. The minor deviations of the experimental data from linearity are likely due to the use of the weight fraction ( Figure 3) instead of volume fraction (Equation (1)). The volume fraction of the PIM, which has a lower density than Matrimid ® 5218, is higher than the weight fraction.
To some extent, the nonlinearity may also be due to slight differences in the degree of physical ageing in the three samples, typically observed for PIMs but less in common glassy polymers with lower free volume.

Mixed Gas Permeability
The membrane performance for two relevant industrial separations was investigated via mixed gas permeability measurements on the pristine Matrimid ® 5218, PIM-EA(H 2 )-TB and on the Matrimid ® 5218/PIM-EA(H 2 )-TB blend (See Appendix A Table A1). Measurements were performed from 1 to 6 bar(a) with two binary gas mixtures of CO 2 /CH 4 (35:65 vol %) and CO 2 /N 2 (15:85 vol %), in order to simulate the biomethane purification process from biogas and CO 2 capture from flue gas, respectively ( Figure 4, Table 2). selectivity decreases in these two membranes. The stronger effect of CO2 on methane than on nitrogen is likely due to its higher concentration (35 vol %) in the CO2/CH4 mixture, compared to 15 vol % in the CO2/N2 mixture. All membranes were aged for at least a month before the measurements, so that further ageing during the experiments should not be expected. Only neat PIM-EA(H2)-TB showed a slightly lower permeability in the pressure decrease steps, but the difference was hardly more than the experimental error, so may not be necessarily attributed to physical ageing. On the contrary, the Matrimid ® 5218/PIM-EA(H2)-TB blend showed weak hysteresis for CO2 and CH4, with slightly higher permeability of both gases in the pressure decrease steps. This suggests a slight dilatation of the polymer matrix and/or removal of trace amounts of residual solvent at elevated CO2 partial pressures.

Mixed Gas Diffusion and Solubility
As described recently, a unique feature of our mixed gas permeation setup with on-line massspectrometric analysis of the permeate composition is that it allows the determination of, not only, the permeability but also the diffusion coefficients of the individual gases in the mixture [45,46]. The results for the three membranes are given in Table 3. The diffusion coefficient is determined by a time lag method for gas mixtures [45], and the solubility is approximated indirectly as S = P/D. The most obvious result for the blend and PIM membrane is that, while the CH4 and N2 permeability change relatively little in the mixture (Table 2), their diffusion coefficients increase substantially and since the change in diffusion coefficient of CO2 is much smaller, the diffusion selectivity decreases  The Matrimid ® 5218 membrane showed very weak pressure-dependence of permeability and selectivity for both gas pairs, CO 2 /CH 4 and CO 2 /N 2 in the given pressure range. It does not show significant hysteresis between the pressure increase steps and the pressure decrease steps, which means that neither substantial physical ageing, nor CO 2 induced dilation of the polymer takes place. On the other hand, for the Matrimid ® 5218/PIM-EA(H 2 )-TB blend and the neat PIM-EA(H 2 )-TB membranes, the CO 2 permeability decreases as function of the feed pressure for both gas mixtures. This is typical for polymers with distinct dual mode behaviour, which is indeed very common for PIMs [49]. The N 2 permeability decreases in a similar fashion as a function of pressure, and thus the CO 2 /N 2 selectivity is virtually constant in all three samples. In the blend and the pure PIM, the methane permeability is slightly less pressure-dependent than that of CO 2 , and therefore the CO 2 /CH 4 selectivity decreases in these two membranes. The stronger effect of CO 2 on methane than on nitrogen is likely due to its higher concentration (35 vol %) in the CO 2 /CH 4 mixture, compared to 15 vol % in the CO 2 /N 2 mixture. All membranes were aged for at least a month before the measurements, so that further ageing during the experiments should not be expected. Only neat PIM-EA(H 2 )-TB showed a slightly lower permeability in the pressure decrease steps, but the difference was hardly more than the experimental error, so may not be necessarily attributed to physical ageing. On the contrary, the Matrimid ® 5218/PIM-EA(H 2 )-TB blend showed weak hysteresis for CO 2 and CH 4 , with slightly higher permeability of both gases in the pressure decrease steps. This suggests a slight dilatation of the polymer matrix and/or removal of trace amounts of residual solvent at elevated CO 2 partial pressures.

Mixed Gas Diffusion and Solubility
As described recently, a unique feature of our mixed gas permeation setup with on-line mass-spectrometric analysis of the permeate composition is that it allows the determination of, not only, the permeability but also the diffusion coefficients of the individual gases in the mixture [45,46]. The results for the three membranes are given in Table 3. The diffusion coefficient is determined by a time lag method for gas mixtures [45], and the solubility is approximated indirectly as S = P/D. The most obvious result for the blend and PIM membrane is that, while the CH 4 and N 2 permeability change relatively little in the mixture (Table 2), their diffusion coefficients increase substantially and since the change in diffusion coefficient of CO 2 is much smaller, the diffusion selectivity decreases significantly. The gas solubility shows exactly the opposite trend: the CH 4 and N 2 solubility decrease in the presence of CO 2 . This provides clear evidence of competitive sorption by CO 2 at the expense of the less condensable gases.

Robeson Plots and Comparison with Literature Blend-Data
The gas permeability data of neat Matrimid ® 5218, PIM-EA(H 2 )-TB, and the blended membrane are plotted in the Robeson diagrams for CO 2 /N 2 , CO 2 /CH 4 , O 2 /N 2 and H 2 /CH 4 ( Figure 5). From Matrimid ® 5218 to the blend and to pure PIM-EA(H 2 )-TB, the diagrams show a strong increase in the pure gas permeability, accompanied by a modest decrease in ideal selectivity, as a common trend for all gas pairs. Literature data of pure Matrimid and other PIM/Matrimid blend membranes for CO 2 /N 2 and CO 2 /CH 4 separation are plotted for reference. Most of the data of our samples lie inside the data cloud near its top range. The cloud of Matrimid data derives from the different measurement conditions, membrane preparation and conditioning.
Only the CO 2 /N 2 selectivity of the neat PIM and the blend lie significantly higher than the cloud and these samples are positioned much closer to the most recent upper bound than Matrimid. Especially for CO 2 /N 2 , with PCO 2 = 198 Barrer and αCO 2 /N 2 = 29, the Matrimid ® 5218/PIM-EA(H 2 )-TB blend presents a better trade-off between permeability and selectivity compared to other Matrimid ® 5218/PIM (50:50) blends reported in the literature, such as PIM-1/Matrimid (PCO 2 = 155 Barrer; αCO 2 /N 2 = 27) and cPIM-1/Matrimid (PCO 2 = 145 Barrer; αCO 2 /N 2 = 24) [32,34]. Of note is that the Matrimid permeability was increased approximately 20-fold by the addition of 50 wt % of PIM-EA(H 2 )-TB. This offers the potential for the preparation of asymmetric or thin film composite membranes with much higher permeability, without compromising selectivity too much.

Conclusions
The pure and mixed gas permeation measurements of neat polymer of intrinsic microporosity PIM-EA(H2)-TB, Matrimid ® 5218 and their 50/50 wt % blend demonstrated that the permeability of Matrimid ® 5218 can be increased dramatically by the addition of the PIM whilst maintaining a reasonably high selectivity. Mixed gas permeation experiments reveal a comparable CO2 permeability and a slightly higher selectivity than the ideal one for the CO2/CH4 and CO2/N2 gas pairs, and addition of the PIM moves the performance of Matrimid ® 5218 closer to the Robeson upper bound for these gas pairs. Interestingly, in spite of the slightly higher mixed gas permselectivity, the mixed gas diffusion selectivity is significantly lower than the ideal diffusion selectivity. This indicates that the presence of CO2 favours the diffusion of the slower gases N2 and CH4, but that the effect of competitive sorption, which reduces their solubility, dominates the overall performance of the membranes. All membranes show a moderate to weak decrease in either the permeability or the selectivity for the CO2/CH4 and CO2/N2 gas pairs with increasing pressure, typical for dual mode sorption behaviour. The very strong effect of PIM-EA(H2)-TB offers the possibility to tailor the permeability of Matrimid ® 5218 over a wide range and opens perspectives for making high permeability thin film composite membranes with the mechanical resistance of Matrimid and PIM-like permeabilities.  ixed gas permeation measurements of neat polymer of intrinsic microporosity rimid ® 5218 and their 50/50 wt % blend demonstrated that the permeability of be increased dramatically by the addition of the PIM whilst maintaining a lectivity. Mixed gas permeation experiments reveal a comparable CO2 lightly higher selectivity than the ideal one for the CO2/CH4 and CO2/N2 gas f the PIM moves the performance of Matrimid ® 5218 closer to the Robeson upper airs. Interestingly, in spite of the slightly higher mixed gas permselectivity, the electivity is significantly lower than the ideal diffusion selectivity. This indicates O2 favours the diffusion of the slower gases N2 and CH4, but that the effect of , which reduces their solubility, dominates the overall performance of the branes show a moderate to weak decrease in either the permeability or the 2/CH4 and CO2/N2 gas pairs with increasing pressure, typical for dual mode The very strong effect of PIM-EA(H2)-TB offers the possibility to tailor the trimid ® 5218 over a wide range and opens perspectives for making high lm composite membranes with the mechanical resistance of Matrimid and ies.

Conclusions
The pure and mixed gas permeation measurements of neat polymer of intrinsic microporosity PIM-EA(H 2 )-TB, Matrimid ® 5218 and their 50/50 wt % blend demonstrated that the permeability of Matrimid ® 5218 can be increased dramatically by the addition of the PIM whilst maintaining a reasonably high selectivity. Mixed gas permeation experiments reveal a comparable CO 2 permeability and a slightly higher selectivity than the ideal one for the CO 2 /CH 4 and CO 2 /N 2 gas pairs, and addition of the PIM moves the performance of Matrimid ® 5218 closer to the Robeson upper bound for these gas pairs. Interestingly, in spite of the slightly higher mixed gas permselectivity, the mixed gas diffusion selectivity is significantly lower than the ideal diffusion selectivity. This indicates that the presence of CO 2 favours the diffusion of the slower gases N 2 and CH 4 , but that the effect of competitive sorption, which reduces their solubility, dominates the overall performance of the membranes. All membranes show a moderate to weak decrease in either the permeability or the selectivity for the CO 2 /CH 4 and CO 2 /N 2 gas pairs with increasing pressure, typical for dual mode sorption behaviour. The very strong effect of PIM-EA(H 2 )-TB offers the possibility to tailor the permeability of Matrimid ® 5218 over a wide range and opens perspectives for making high permeability thin film composite membranes with the mechanical resistance of Matrimid and PIM-like permeabilities. Acknowledgments: Phenom-World B.V., Eindhoven (NL), is gratefully acknowledged for providing a Phenom Pro X desktop SEM for evaluation. Huntsman (Basel, Switzerland) is gratefully acknowledged for providing a sample of Matrimid ® 5218.

Conflicts of Interest:
The authors declare no conflict of interest.