Introduction

The reduction of carbon dioxide (CO2) emissions resulting from combustion of fossil fuels is recognized as a crucial issue because of its significant impact on global climate change and other associated problems.1, 2 Therefore, CO2 capture and separation is becoming a serious concern and of great importance for mitigating such environmental issues. Hydrogen (H2) purification,3 natural gas treatment4 and the removal of CO2 from gas streams are related technologies for efficient fossil fuel utilization.5 Up to now, four methods have been proposed and used in CO2 capture and separation: absorption in a liquid phase,6, 7, 8 adsorption on highly porous materials,9, 10 membrane separation6 and cryogenic condensation.11 Among them, membrane separation technology requires smaller operational energy and lower running costs as compared with the other more energy consumptive methods, especially in small- and medium-scale separations.12, 13

For membrane separation of CO2, supported ionic liquid membranes (SILMs),14, 15, 16, 17 polymeric membranes18, 19, 20, 21, 22 and ion-gel membranes23, 24, 25, 26, 27 have been developed. SILMs are polymeric or inorganic porous materials filled with ionic liquids (ILs). They exhibit higher permeability and selectivity of CO2 than other polymeric membranes, which generally exhibit limited CO2 diffusion in their rigid polymer matrix. These SILMs are favorable for gas separation because of their high CO2 selectivity, which is owing to the higher solubility of CO2 in IL than other gases, such as O2, H2, N2 and CH4.14, 15, 16, 17, 28 The unique properties of ILs, such as their negligible vapor pressure, high thermal and chemical stability, and low flammability,29, 30, 31 also contribute to the stability and durability of SILMs. However, SILMs cannot hold ILs under pressurized conditions because ILs in the pores are physically held by weak capillary forces, which is a serious problem for gas separation membranes.17 Ion gels, which are polymeric networks swollen by an IL or by IL mixtures,32 have recently been developed as promising materials for CO2 separation membranes to solve the problem of SILMs.

Generally, polymer gels are classified into two categories, that is, chemical gels and physical gels. The same classification can also be made for ion gels. Chemical ion gels are formed by covalent crosslinking of IL-compatible polymers. For example, we developed a series of chemical ion gels by in situ polymerization of vinyl monomers in ILs.33 Physical ion gels are typically formed by microphase-separated structures of amphiphilic block copolymers,34, 35, 36, 37, 38, 39, 40 in which IL-incompatible segments41, 42 aggregate and serve as the physical crosslinking points. Both classes of ion gels have been applied to CO2 separation membranes and both exhibit favorable separation properties without any leakage issues. However, because of their lower mechanical properties, both classes of ion-gel membranes need to be relatively thick (>1 mm) to maintain membrane integrity. This thickness leads to lower permeability because the permeability of separation membranes depends on the ratio of the permeability coefficient to the membrane thickness.43 We, therefore, were interested in developing mechanically reliable CO2 separation membranes.

Up to now, we reported a series of thin and tough composite membranes (<100 μm) consisting of an IL and a sulfonated polyimide (SPI) having a cation in common with the IL.44, 45 These membranes consist of excellent mechanical properties, which have never been achieved for conventional ion-gel membranes, in addition to good thermal stability and high ionic conductivity.46 These composite membranes have been applied to non-humidified intermediate temperature fuel cells47 and actuators.48 Thus, we considered that these composite membranes can also be possible candidate as CO2 separation membranes.

In this work, we prepared IL/SPI composite membranes comprising a hydrophobic IL, either [C4mim][NTf2] or [C4mim]PF6, and an SPI containing [C4mim] as a counter cation. Membranes were fabricated using the solution casting method. These membranes were characterized via their thermal and mechanical property measurements. Furthermore, CO2 separation performance of the membranes was explored.

Experimental procedures

Materials

1-Bromobutane (98%) and triethylamine (99%; Wako Chemicals, Takasaki, Japan) was used as received. 1-Methylimidazole (99%; Tokyo Chemical Industry, Tokyo, Japan) was distilled under vacuum before use. Lithium bis(trifluoromethanesulfonyl)amide (Morita Chemicals, Osaka, Japan) and sodium hexafluorophosphate (97%; Wako Chemicals) were stored in Ar atmosphere grove box. Bis[4-(3-aminophenoxy)-phenyl]sulfone (98%; Tokyo Chemical Industry) was recrystallized from ethanol/water solution. 1,4,5,8-Naphthalene-tetracarboxylicdianhydride (Sigma-Aldrich, St Louis, MO, USA) was poured into N,N-dimethylformamide and resulting suspension was stirred at 60 °C for a day to remove soluble impurity. Pure 1,4,5,8-naphthalene-tetracarboxylicdianhydride was corrected by filtration, washed with acetone several times and dried under vacuum. 2,2′-Benzidinedisulfonic acid (70%; Tokyo Chemical Industry) was purified as follows: purchased 2,2′-benzidinedisulfonic acid was dissolved in deionized water including excess amount of triethylamine and the solutions was filtered. Addition of 1 M HSO4 aqueous solution to the solution afforded precipitation of pure 2,2′-benzidinedisulfonic acid as white powder. The white power was washed with water several times and dried under vacuum.

Membrane preparation

[C4mim][NTf2] and [C4mim]PF6 were synthesized and purified according to a previously reported method,49 and their structure was identified by 1H NMR. SPI was also synthesized by a previously reported method,46 and its counter cation was exchanged with [C4mim] by immersing it in a methanol solution of [C4mim][NTf2].48 Composite membranes were prepared by the solution casting method according to a previously reported method.46 The chemical structures of [C4mim][NTf2], [C4mim]PF6 and the SPI in [C4mim] form and the appearance of the composite membranes are shown in Figure 1.

Figure 1
figure 1

Chemical structures of (a) sulfonated polyimide (SPI), (b) ionic liquid (ILs) ([C4mim][NTf2] and [C4mim][PF6]) and (c) a photograph of [C4mim][NTf2] composite membrane. A full color version of this figure is available at the Polymer Journal journal online.

Thermal analyses

Differential scanning calorimetry (DSC) measurements were carried out using a Hitachi High-Tech DSC 6220 (Hitachi High-Technologies, Tokyo, Japan) under a N2 atmosphere. The samples were tightly sealed in aluminum pans in a glove box ([H2O]<1 p.p.m.). The measuring pans in the DSC apparatus were heated to 150 °C, cooled to −150 °C and then heated again to 150 °C. The cooling and heating rates were 10 °C min−1. The DSC curves were recorded during the reheating scans. Thermogravimetric analysis of the composite membranes was carried out using a Seiko Instruments TG-DTA 6200C (Seiko Instruments, Chiba, Japan) under a N2 atmosphere. The samples were weighed and placed in aluminum pans. Then, they were heated to 100 °C and maintained at that temperature for 30 min under N2 atmosphere to remove included water. After cooling to room temperature, they were heated again to 550 °C at a heating rate of 10 °C min−1 and the thermogravimetric curves were recorded.

Tensile tests

Tensile tests were performed with a Shimadzu EZ-LX 10N (Shimadzu, Kyoto, Japan) under an ambient atmosphere. Samples were cut into dumbbell shapes (JIS K6251, 35 × 6 mm2 (total) and 12 × 2 mm2 (test area)) using a Kobunshi-Keiki dumbbell cutter (Koubunshi Keiki, Kyoto, Japan). Stress–strain curves were measured by stretching the samples at a head speed of 1 mm s−1.

Gas permeability measurement

CO2 and N2 permeability was measured with a gas permeation measurement apparatus, GTR-10XFKS (GTR-Tech, Kyoto, Japan), and the concentration of the permeated gases was quantified using a gas chromatograph, Yanaco G2700T (Yanaco Technical Science, Tokyo, Japan), with a thermal conductivity detector. Helium was used as the carrier gas for the measurement of CO2 and N2 concentrations. The membranes were placed at the center of the permeation cell (permeation area was 9.62 cm2), and the test gas (either CO2 or N2/mixed CO2 and N2 (2:1 by volume)) was introduced to one side of the membrane at a flow rate of 30 ml min−1. The carrier gas was introduced to the other side of the membrane at a flow rate of 50 ml min−1, and the concentration of the permeated gas was analyzed by the gas chromatograph. A pressure of carrier gas and a total pressure of test gas were set at 76 cm Hg (atmospheric pressure), respectively. The measurements were conducted five times for a single membrane to estimate gas permeability, and the average was adopted as the result. The gas permeability coefficients, and (Barrer=1 × 10−10 cm3 (STP) cm cm−2 s−1 cm Hg−1), were calculated using the following equation:

where Q (cm3) is the volume of the test gas permeated through the membrane, X (cm) is the thickness of the sample membrane, A (cm2) is the permeation area, t (s) is the sampling time, C (cm Hg) is the partial pressure of the test gas and T (K) is the absolute temperature.

Single-gas permeation tests were performed by introducing a pure tested gas, whereas mixed-gas permeation tests were performed by introducing CO2/N2 (2/1 by volume) mixed gases.

Results and discussion

Membrane fabrication

In this paper, the IL/SPI composite membranes are denoted as SPI-A (X), where A is the anionic structure of the ILs and X is weight percent (wt%) of the ILs in the composite membranes. For example, SPI-NTf2 (50) means [C4mim][NTf2]/SPI composite membrane including 50 wt% of [C4mim][NTf2]. Figure 1c shows the appearance of a composite membrane. For all the compositions except for SPI-PF6 (75), transparent, flexible and uniform membranes were obtained. Previously, based on the results of atomic force microscopy tapping images, we reported nanophase separation of the ionic soft domain from the non-ionic hard domain in the IL/SPI composite membranes; we also reported that the ILs appeared to be preferentially incorporated into the ionic domain.45 [C4mim]PF6 had lower compatibility with the SPI in comparison with [C4mim][NTf2], and thus macroscopic phase separation occurred at higher IL contents. However, when the SPI and the ILs were compatible, homogeneous membranes having thicknesses of 20–150 μm with no leakage of the ILs were obtained. In this study, these membranes were examined using thermal analysis, tensile tests and gas permeation tests.

Differential scanning calorimetry

Figures 2a and b show the DSC curves of the pure ILs, the pure SPI and the composite membranes. [C4mim][NTf2] exhibited phase transitions at −87, −22 and −3.1 °C, which can be assigned to the glass transition, crystallization and melting temperatures, respectively, whereas [C4mim]PF6 exhibited those transitions at −73, −38, −10 and 15 °C, respectively. These values are comparable to those reported in previous studies.49, 50, 51 On the other hand, no phase transition except for a slight baseline shift, assignable to the glass transition, was observed for each composite membrane at temperatures below −50 °C. The ILs appear to be confined into the nanosize ionic domains, and thus crystallization of the ILs is inhibited (confinement effect).47

Figure 2
figure 2

Differential scanning calorimetry (DSC) curves of the pure sulfonated polyimide (SPI) membrane and composite membranes: (a) SPI-NTf2 and (b) SPI-PF6.

TG curves of the ILs, SPI and the composite membranes are shown in Figure 3, and the decomposition temperatures are listed in Table 1. The SPI weight loss was observed at 333 °C, whereas those of [C4mim][NTf2] and [C4mim]PF6 were observed at 413 and 358 °C, respectively. For the composite membranes, the residual weight at the upper limit of the measurement (at 550 °C) changed as a function of the weight ratio of the SPI to the IL. However, the composite membranes did not clearly show two-step weight loss because of the close decomposition temperatures between the SPI and the ILs. It is clear that the composite membranes have sufficient thermal stability for CO2 gas separation.

Figure 3
figure 3

Thermogravimetric (TG) curves of the pure sulfonated polyimide (SPI) membrane and the composite membranes: (a) SPI-NTf2 and (b) SPI-PF6.

Table 1 The results of TGA and tensile tests

Tensile tests

Strain–stress curves measured for the SPI and the composite membranes are shown in Figure 4, and the parameters of the mechanical properties are listed in Table 1. The stress at break, elongation at break and Young’s modulus measured for the SPI membrane were 58.7 MPa, 24.1% and 1.58 GPa, respectively. With the addition of the ILs, the yield behavior of the SPI gradually disappeared. These phenomena clearly demonstrate that the ILs functioned as plasticizers in the composite membranes. However, even for SPI-NTf2 (75), the stress at break and Young’s modulus were 3.16 and 9.6 MPa, respectively, values that are much greater than those reported for conventional ion gels: It was reported that poly(methyl methacrylate)/TEOS-based composite membrane including 76 wt% of [C4mim][NTf2] exhibits ca. 0.3 MPa of Young’s modulus,52 while polyether-based composite membrane exhibits 0.58 MPa.53 These superior mechanical properties of the IL/SPI composite membranes are preferable for practical applications. Differences in the mechanical behavior of the SPI-NTf2 and SPI-PF6 composite membranes were observed but were not significant.

Figure 4
figure 4

Strain–stress curves of the sulfonated polyimide (SPI) and the composite membranes: (a) SPI-NTf2 and (b) SPI-PF6.

Gas permeability

CO2 and N2 permeation properties measured under single-gas conditions are listed in Table 2. At 30 °C, the CO2 permeability of the SPI membrane was 0.58 Barrer. On the other hand, the N2 permeability could not be measured correctly because the N2 gas permeation of the membrane was too low to be detected by our instrument. This result indicates that the SPI membranes have low gas permeability, comparable to those of typical polyimide membranes54, 55 (e.g., Kapton), which have high Tg and very low gas permeability. This suggests that the substitution of sulfonic acid groups into a polyimide structure does not greatly affect the gas barrier properties. The much higher CO2 permeability than N2 permeability through the SPI membrane can be explained by the higher solubility of CO2 into the SPI membrane due to the stronger interaction of CO2 with the SPI by a quadrupole moment than non-polar N2. It should be noted that the gas permeability greatly increased by the addition of the IL and that the and values of SPI-NTf2 (75) at 30 °C reached 412 Barrer and 15.4 Barrer, respectively, whereas those of SPI-PF6 (66) were 65.8 Barrer and 1.77 Barrer, respectively. Interestingly, the separation coefficient (α=) seems to depend on the IL structures. The α-values of SPI-NTf2 and SPI-PF6 were 27–29 and 33–37, respectively. No significant decrease in α was observed, even with the increase in gas permeability, accompanied by increasing IL content. These results suggest that the permeability of the gases is mainly dominated by the solubility of the gases in the membranes, where the ionic groups form domains, that the ILs are preferentially incorporated into the ionic domains and that both gases selectively permeate in the ionic domain.

Table 2 The results of gas permeability measurements performed under single- and mixed-gas conditions

The gas permeation properties under mixed-gas conditions were also investigated for the SPI-NTf2 composite membrane, and the results are listed in Table 2. As compared with the gas permeability under single-gas conditions, decreased, whereas increased. This phenomenon has been often reported for mixed-gas measurements. Two explanations are generally accepted. One is based on plasticization effect of CO2 molecule for matrix membrane. If membrane is plasticized by CO2 molecule, free volume in the membrane increases and thus gas permeation coefficient increases. Therefore, CO2 permeation coefficient may decrease with decreasing partial pressure of CO2 in test gas. Simultaneously, in the case of CO2/N2 mixture, N2 permeation coefficient will increase with decreasing partial pressure of N2, since partial pressure of CO2 increases. The other explanation is based on dragging effect. When two gases having different diffusion coefficients are used for mixed-gas measurements, faster diffusing gas molecule is dragged by slower diffusing one and becomes slower. In contrast, slower diffusing gas is dragged by faster diffusing one and becomes faster. As a result, the separation coefficient decreases. Actually, both mechanisms may cause the same result. To conclude which mechanism is plausible for IL/SPI composite membrane, further investigation is required. Finally, the separation performance of our composite membranes is demonstrated in Robson plots, shown in Figure 5, where the upper limits of the gas permeation polymer membranes are shown with solid lines.56 The results are still below the upper limit; however, they approach the limit with increasing IL content.

Figure 5
figure 5

Robeson plot of sulfonated polyimide/ionic liquid (SPI/IL) composite membrane. Solid line indicates upper limit for CO2 separation polymer membranes reported in Robeson.56

Conclusion

A series of composite membranes, SPI-NTf2 and SPI-PF6, were prepared and their thermal, mechanical and gas permeation properties were investigated. SPI showed good compatibility with ILs and thus ILs functioned as plasticizer of SPI. However, SPI composite membrane exhibited much high Young’s modulus (≈10 MPa) even when it includes 75 wt% of ILs. This characteristic is convenient for fabricate thin membrane. The SPI membrane behaved as a gas barrier membrane ( for SPI membrane is 0.58 Barrer). However, the addition of ILs greatly enhanced the gas permeability to be 412 Barrer for SPI-NTf2 (75). The SPI-PF6 composite membranes slightly lower the gas permeability than SPI-NTf2 composite membranes at same IL contents, but higher separation coefficient (α= ). In the permeation tests under mixed-gas conditions, and values were lower and higher, respectively, than those under single-gas conditions. However, α-value was maintained in the range 12–16.