Polymer of Intrinsic Microporosity (PIM-1) Membranes Treated with Supercritical CO2

Polymers of intrinsic microporosity (PIMs) are a promising membrane material for gas separation, because of their high free volume and micro-cavity size distribution. This is countered by PIMs-based membranes being highly susceptible to physical aging, which dramatically reduces their permselectivity over extended periods of time. Supercritical carbon dioxide is known to plasticize and partially solubilise polymers, altering the underlying membrane morphology, and hence impacting the gas separation properties. This investigation reports on the change in PIM-1 membranes after being exposed to supercritical CO2 for two- and eight-hour intervals, followed by two depressurization protocols, a rapid depressurization and a slow depressurization. The exposure times enables the impact contact time with supercritical CO2 has on the membrane morphology to be investigated, as well as the subsequent depressurization event. The density of the post supercritical CO2 exposed membranes, irrespective of exposure time and depressurization, were greater than the untreated membrane. This indicated that supercritical CO2 had solubilised the polymer chain, enabling PIM-1 to rearrange and contract the free volume micro-cavities present. As a consequence, the permeabilities of He, CH4, O2 and CO2 were all reduced for the supercritical CO2-treated membranes compared to the original membrane, while N2 permeability remained unchanged. Importantly, the physical aging properties of the supercritical CO2-treated membranes altered, with only minor reductions in N2, CH4 and O2 permeabilities observed over extended periods of time. In contrast, He and CO2 permeabilities experienced similar physical aging in the supercritical treated membranes to that of the original membrane. This was interpreted as the supercritical CO2 treatment enabling micro-cavity contraction to favour the smaller CO2 molecule, due to size exclusion of the larger N2, CH4 and O2 molecules. Therefore, physical aging of the treated membranes only had minor impact on N2, CH4 and O2 permeability; while the smaller He and CO2 gases experience greater permeability loss. This result implies that supercritical CO2 exposure has potential to limit physical aging performance loss in PIM-1 based membranes for O2/N2 separation.


Introduction
Polymers of intrinsic microporosity (PIMs) are attractive for polymeric membranes, because of their very high fractional free volume and favourable interconnectivity between micro-cavities [1,2]. For many gas pairs, PIMs-based membranes are on or above the Robeson's upper bound, the criterion denoting current state-of-art performance in gas separation membranes [3]. This high performance is the result of the spirobisindane moiety, creating rigid ladder-type polymeric chain structures, with significant steric hindrance preventing chain rotation and limiting chain packing. However, PIMs-based polymeric membranes suffer from a reduction in separation performance over time, known

Materials and Methods
PIM-1, the polycondensation product of ultrahigh purity monomers of 5,5 ,6,6 -tetrahydroxyl-3,3,3 ,3 -tetramethyl-1,1 -spirobisindane (TTSBI) and 2,3,5,6-tetra fluoroterephthalonitrile (TFTPN), was synthesised following established procedures by Budd et al. [16]. Membranes of PIM-1 were cast from solutions of dichloromethane through controlled evaporation. The final film thickness was between 63 and 78 µm. All films were annealed at 150 • C for 1 day under vacuum to remove dichloromethane, and then cooled to room temperature overnight. The original membrane was washed with methanol and then allowed to dry to reverse any aging effects, while those membranes exposed to scCO 2 were not [9,17]. Washing with methanol treatment of the scCO 2 treated membranes reverts their morphology, similar to reversing aging effects.
Membrane densities were determined through standard procedures [18]. Gas sorption measurements of CO 2 were undertaken on a gravimetric sorption analyser (GHP-FS, VTI Instruments) operating at 35 • C. The pressure was incrementally adjusted from 0 to 20 atm, with helium used for buoyancy correction [19]. Single gas permeabilities were undertaken on a variable pressure constant volume apparatus as previously described [20], with feed pressures of 8 atm and 35 • C. The permeability values were the average of three single gas measurements per gas with a fresh membrane each time; error margins corresponding to two standard deviations in the permeability data set. ScCO 2 is achieved above 31.1 • C and 72.9 atm [21]. The scCO 2 treatment was undertaken in an autoclave equipped with the inlet of CO 2 and a backpressure regulator; membranes were exposed at 246.7 atm and at 50 • C for 2 or 8 h. Depressurization back to ambient pressure was achieved through two mechanisms, rapid depressurization, which occurred over a few minutes (depressurization rate: 118 atm/min), and gradual depressurization, over 150 min (depressurization rate: 1.7 atm/min).
Hence, the scCO 2 treatment process was designed to investigate both the impact of exposure time and removal rate of CO 2 on the underlying PIM-1 membrane's performance.

CO 2 Sorption Isotherm of PIM-1
The sorption isotherm of CO 2 in the original PIM-1 membrane is provided in Figure 1, as a function of pressure at 35 • C. This isotherm was comparable to literature and followed standard dual-sorption model behaviour [22], with a significant sorption of CO 2 at low pressures attributed to the micro-cavities within the membrane morphology being filled, while at higher pressures the micro-cavities became saturated and additional sorption was limited to the polymeric matrix. The CO 2 concentration (C) within the membrane can be modelled by dual-sorption theory [22]: where p is the pressure, k D the Henry's law constant, C' H the maximum Langmuir adsorption capacity and b the Langmuir affinity. The evaluated parameters are provided in Table 1.

CO2 Sorption Isotherm of PIM-1
The sorption isotherm of CO2 in the original PIM-1 membrane is provided in Figure 1, as a function of pressure at 35 °C. This isotherm was comparable to literature and followed standard dualsorption model behaviour [22], with a significant sorption of CO2 at low pressures attributed to the micro-cavities within the membrane morphology being filled, while at higher pressures the microcavities became saturated and additional sorption was limited to the polymeric matrix. The CO2 concentration (C) within the membrane can be modelled by dual-sorption theory [22]: where p is the pressure, kD the Henry's law constant, C'H the maximum Langmuir adsorption capacity and b the Langmuir affinity. The evaluated parameters are provided in Table 1. These dual-sorption parameters were comparable to literature and reveal that the PIM-1 membrane studied here had similar morphology to those previous studies [10,23]. Interestingly, the kD and b values for PIM-1 was significantly lower than other polymers investigated for CO2 separation, such as cellulose triacetate and polyimides [15]. This indicates that the PIM-1 polymer chain did not have strong affinity for CO2 relative to other polymers, and that the strong sorption was mainly attributed to the micro-cavities, which had a capacity significantly higher than other polymers. The sorption analyser's maximum pressure was 20 atm, and; therefore, determining the sorbed amount of CO2 up to the critical pressure could not be measured. However, the isotherm clearly followed standard dual-sorption mode behaviour, and hence it was possible to extrapolate the sorption behaviour to higher pressures to indicate how much CO2 could be sorbed into the membrane. This extrapolation is provided in Figure 1 only as a guide, since non-linear deviation in CO2 sorption is anticipated at significantly high pressures because of polymer plasticization and condensation of scCO2 in the membrane. Interestingly, the extrapolation suggests that PIM-1 membrane's sorption of CO2 at high pressures was not substantial, only a doubling of the amount of CO2 sorbed at 20 atm. This is due to the relatively poor affinity the polymeric matrix has for CO2,  These dual-sorption parameters were comparable to literature and reveal that the PIM-1 membrane studied here had similar morphology to those previous studies [10,23]. Interestingly, the k D and b values for PIM-1 was significantly lower than other polymers investigated for CO 2 separation, such as cellulose triacetate and polyimides [15]. This indicates that the PIM-1 polymer chain did not have strong affinity for CO 2 relative to other polymers, and that the strong sorption was mainly attributed to the micro-cavities, which had a capacity significantly higher than other polymers.
The sorption analyser's maximum pressure was 20 atm, and; therefore, determining the sorbed amount of CO 2 up to the critical pressure could not be measured. However, the isotherm clearly followed standard dual-sorption mode behaviour, and hence it was possible to extrapolate the sorption behaviour to higher pressures to indicate how much CO 2 could be sorbed into the membrane. This extrapolation is provided in Figure 1 only as a guide, since non-linear deviation in CO 2 sorption is anticipated at significantly high pressures because of polymer plasticization and condensation of scCO 2 in the membrane. Interestingly, the extrapolation suggests that PIM-1 membrane's sorption of CO 2 at high pressures was not substantial, only a doubling of the amount of CO 2 sorbed at 20 atm. This is due to the relatively poor affinity the polymeric matrix has for CO 2 , compared to other polymers used for gas separation membranes [23]. Hence, the amount of CO 2 sorbed at critical pressure would be comparable to cellulose triacetate [15], a midrange polymeric membrane with a permselectivity that is lower than the Robeson's upper bound, rather than other high performing polymers.

Gas Permeability in Original PIM-1 Membrane
The gas permeability through the original PIM-1 membrane is provided in Table 2 after seven days of aging, along with reported literature values. There was discrepancy between the reported gas permeabilities in the literature, as well as with those determined here, which is attributed to the casting history of the PIM-1 membrane. The earlier study of Budd et al. [24] had lower gas permeabilities compared to this work and that of Thomas et al. [25], because the earlier studies had different degrees of physical aging and had not undergone methanol restoration. The CO 2 permeability reported here had a reasonable correlation with that of Thomas et al. [25], though all gases were higher, especially CH 4 which was three times the magnitude. However, the CH 4 result was similar to the membrane reported by Starannikova et al. [26], and there was also comparison in the He and O 2 permeabilities of the two membranes. Hence, the wide variation in reported gas permeabilities for PIM-1 based membranes could be attributed to the differences in the morphology of the membrane, as a result of both the synthesising procedure and casting history. The selectivity of the original PIM-1 membrane is provided in Table 3, and displayed similar behaviour to literature, in that the membrane is clearly selective for CO 2 against CH 4 and N 2 ; as well as being selective for O 2 and He. However, the high CH 4 permeability of this membrane resulted in the CO 2 /CH 4 and He/CH 4 selectivity being lower than literature, and hence the membrane investigated here was not on the Robeson's upper bound for these gas pairs.

Density
The density of the original PIM-1 membrane and after treatment with scCO 2 for two and eight hours, as well as rapid and slow depressurizations, is provided in Table 4, after seven days of aging. The density of the original membrane, after methanol regeneration, was 1.114 g/cm 3 . After scCO 2 treatment, the density increased irrespective of the treatment protocol, and hence a denser morphology was obtained. This was associated with the ability of scCO 2 to solubilise the polymer, enabling enhanced chain mobilization and rearrangement. As a consequence, the fractional free volume of scCO 2 -treated PIM-1 membrane would have reduced. Importantly, the longer eight-hour exposure resulted in a denser membrane morphology, than the shorter two-hour exposure, supporting the conclusion that polymer solubilisation was the dominate factor. The depressurization rate clearly impacted the morphology, with the rapid depressurization resulting in a lower density structure, irrespective of exposure time, which has been observed for other polymeric membranes exposed to scCO 2 [15]. This dense morphology was clearly observed in SEM images for the rapid depressurization PIM-1 membranes, provided in Figure 2. The densities were greater than the original membrane, implying that the morphology changes due to rapid depressurization of the scCO 2 were not great enough to reverse the polymer solubilisation effect. This density increase within the PIM-1 membranes differed from that observed for cellulose triacetate membranes exposed to scCO 2 , irrespective of depressurization, as well as polyimide-based membranes that underwent rapid depressurization, but was similar to polyimide membranes that experienced slow depressurization [15]. Hence, scCO 2 treatment was polymer dependent.

Density
The density of the original PIM-1 membrane and after treatment with scCO2 for two and eight hours, as well as rapid and slow depressurizations, is provided in Table 4, after seven days of aging. The density of the original membrane, after methanol regeneration, was 1.114 g/cm 3 . After scCO2 treatment, the density increased irrespective of the treatment protocol, and hence a denser morphology was obtained. This was associated with the ability of scCO2 to solubilise the polymer, enabling enhanced chain mobilization and rearrangement. As a consequence, the fractional free volume of scCO2-treated PIM-1 membrane would have reduced. Importantly, the longer eight-hour exposure resulted in a denser membrane morphology, than the shorter two-hour exposure, supporting the conclusion that polymer solubilisation was the dominate factor. The depressurization rate clearly impacted the morphology, with the rapid depressurization resulting in a lower density structure, irrespective of exposure time, which has been observed for other polymeric membranes exposed to scCO2 [15]. This dense morphology was clearly observed in SEM images for the rapid depressurization PIM-1 membranes, provided in Figure 2. The densities were greater than the original membrane, implying that the morphology changes due to rapid depressurization of the scCO2 were not great enough to reverse the polymer solubilisation effect. This density increase within the PIM-1 membranes differed from that observed for cellulose triacetate membranes exposed to scCO2, irrespective of depressurization, as well as polyimide-based membranes that underwent rapid depressurization, but was similar to polyimide membranes that experienced slow depressurization [15]. Hence, scCO2 treatment was polymer dependent.   Figure 2. Scanning electron microscope (SEM) images of the PIM-1 membranes after two-and eight-h exposure to scCO 2 , with rapid depressurization.

Gas Permeability
The gas permeability through the PIM-1 membrane after treatment with scCO 2 is provided in Table 4, after seven days of aging, for two and eight-hour exposure, as well as rapid and slow depressurization. All four scCO 2 treatments resulted in a reduction in the gas permeability through the membrane, compared to the original membrane (Table 2). This behaviour corresponded well with the increased density of the scCO 2 treated membranes, implying that the PIM-1 morphology change restricts gas permeance. The exception was N 2 , which was essentially unchanged, within error, of the original membrane result. The low permeability of N 2 in the original and scCO 2 treated membranes implies that the PIM-1 membrane morphology remained unfavourable for N 2 , and the interaction with scCO 2 did not alter this.
There was a clear trend in the scCO 2 treatment protocol on the membrane, with rapid depressurization having reduced gas permeabilities compared to slow depressurization, for the same exposure time. This reveals that rapidly removing the scCO 2 from the PIM-1 membrane alters the morphology to be less favourable for gas permeance compared to slow depressurization. This was counter to the observation for other polymeric membranes exposed to scCO 2 , notably cellulose triacetate and polyimide [15], where rapid depressurization resulted in swelling. This behaviour was also counter to the density measurements (Table 4), where a lower density usually corresponded to increased gas permeability. This difference was attributed to the higher fractional free volume of PIM-1 and the already high permeance of CO 2 enabling the scCO 2 to rapidly desorb through the established connecting pathways between micro-cavities. This limited the ability of scCO 2 to generate new pathways for desorption during rapid depressurization and, hence, the PIM-1 structure did not swell, as the resulting densities of the rapid depressurization membranes remained greater than the original membrane. Why the slower depressurization protocol resulted in higher gas permeabilities is unknown, given the denser morphology (Table 4); but the behaviour does suggest that the existing and established pathways through PIM-1 membranes' micro-cavities remained open during scCO 2 treatment, most likely because of the strong accumulation of scCO 2 in these free volume regions. The variability in gas permeability after scCO 2 treatment of various polymeric membranes [15] further establishes that the change in morphology and gas separation properties outcomes are polymer dependent.
The corresponding selectivity of the PIM-1 membranes after scCO 2 treatment are provided in Table 5. Compared to the original membrane there was clear deviation for the scCO 2 membranes. For separation from CH 4 (i.e., CO 2 /CH 4 and He/CH 4 ) the scCO 2 -treated membranes showed an increase in selectivity over the original membrane. This was a direct result of the CH 4 permeability reducing by 32%-39% after exposure to scCO 2 , while the CO 2 permeability was reduced by only 7%-32% and He permeability reduced by 4%-31%. In contrast, CO 2 /N 2 and He/N 2 selectivity decreased after exposure to scCO 2 , which was a direct result of the N 2 permeability, through the scCO 2 -treated PIM-1 membranes, remaining essentially constant with the original membrane. Furthermore, there was evidence that longer scCO 2 exposure time and slower depressurization result in higher selectivity than rapid depressurization.

Aging Study
The change in He permeability through PIM-1 membranes over an extended period of time is provided in Figure 3, for both the original membrane and the four scCO 2 -treated protocols. For all five membranes the He permeability decreased with time, indicative of physical aging. There were differences between the permeabilities presented in Figures 3-7, and those presented in Table 4, as they represent different PIM-1 membranes that were measured at different times after scCO 2 treatment. Over the 63 days studied, all five membranes experienced similar aging behaviour, losing~200 barrer in He permeability at the 63-day mark. Interestingly, there was no difference between the original and scCO 2 -treated membranes in terms of He aging. This suggests scCO 2 exposure had not changed the mechanism of physical aging; that of lattice contraction within the PIM-1 structure [8], as polymer chains rearrange to a denser state. A comparable increase in density of the PIM-1 membranes was observed in the aging period, with the two hours slow depressurization PIM-1 membrane experiencing a density increase of 5%. However, He was not a good indicator of micro-cavity change, because being the smallest molecule enabled He to permeate more readily through the micro-cavities and polymeric matrix compared to other gases.   The change in CO 2 permeability through PIM-1 membranes over an extended period of time is provided in Figure 4, for both the original membrane and the four scCO 2 -treated protocols. Again, there was clear evidence of physical aging, with an initial pronounced loss in permeability over the first two weeks, which progressed to a gradual reduction over the longer time period. The effect of physical aging on He and CO 2 permeabilities was clearly different, which is attributed to the relative sizes of the gases. CO 2 is significantly large that its permeance through the PIM-1 morphology is dominated by transport through the micro-cavities, which becomes restricted as the membrane ages. The original membrane CO 2 permeability was reduced by~2500 barrer over the 49 days studied, while the scCO 2 -treated membranes experienced a larger reduction of 2500 to 3500 barrer over the aging period. Hence, the scCO 2 treated process had enhanced the physical aging impact on CO 2 permeability. This is attributed to the denser membrane structure (Table 4), resulting from scCO 2 treatment having reduced the micro-cavities in which CO 2 transverses through the PIM-1 membrane, and hence physical aging in the remaining micro-cavities was more pronounced on CO 2 permeability, which appeared as an enhancement of physical aging.
The change in CH 4 permeability through PIM-1 membranes over time is provided in Figure 5, for both the original membrane and the four scCO 2 -treated protocols. For CH 4 , there was a clear difference in the physical aging behaviour of the original membrane and those treated with scCO 2 . The original membrane CH 4 permeability was reduced by~500 barrer over 50 days, while membranes exposed to scCO 2 for two hours experienced a physical aging loss of~300 barrer, and membranes treated for eight hours experienced less than 200 barrer loss in CH 4 permeability over 63 days. Similar behaviour was also clearly observed for N 2 and O 2 permeability in PIM-1 membranes over time, as provided in Figures 6 and 7, respectively. Hence, exposure to eight hours of scCO 2 resulted in a morphology that underwent minor physical aging in terms of CH 4 , N 2 and O 2 permeability, while two hours scCO 2 exposure gave rise to physical aging that was significantly reduced compared to the original membrane for the same gases. This was attributed to the scCO 2 treatment reducing the larger micro-cavities, by creating a denser morphology, in which CH 4 , N 2 and O 2 would previously permeate through PIM-1 (as evident by the decrease in permeability of CH 4 and O 2 between the original and scCO 2 treated membranes). Hence, further micro-cavity contraction due to physical aging had only a minor impact on CH 4 , N 2 and O 2 , as they were already size excluded from the micro-cavities in which CO 2 and He permeate. This became notable in the change in selectivity of the membrane over time, with the CO 2 /N 2 selectivity of the eight-hour exposed membrane (irrespective of depressurization protocol) decreasing to 10.3 over the 63 days. Similar, changes were also observed in the CO 2 /CH 4 selectivity, which decreased to 4.6 over the 63 days; highlighting the magnitude of the decrease in CO 2 permeability relative to the larger gases.        The physical aging of gas permeability in PIM-1 membranes can be described by a power law, as established by Bernardo et al. [27]: where P 0 is the initial permeability of the membrane at t = 1 and β P the permeability aging rate constant. The determined β P values for the original and scCO 2 -treated PIM-1 membranes are provided in Figure 8, as a function of the squared effective diameter of the gases studied, CO 2 is at 0.091 nm 2 and N 2 is at 0.092 nm 2 . The original PIM-1 membrane had very similar β P values to that reported by Bernardo et al. [27], and hence comparable physical aging with that study, which included an ethanol treatment post-fabrication. For the scCO 2 -treated PIM-1 there was a clear reduction in the aging constant for those membranes exposed for eight hours, along with a loss in correlation associated with gas diameter. The explanation to this behaviour is attributed to CO 2 solubilising the polymer chain and creating a denser morphology the longer PIM-1 is exposed; however, scCO 2 when desorbing leaves behind micro-cavities of sufficient size to enable depressurization. These micro-cavities then undergo physical aging, which was observed in the permeability loss of He and CO 2 ; while CH 4 , N 2 and O 2 are larger molecules and thus size restricted from these post scCO 2 micro-cavities, and hence do not experience physical aging to the same degree as their permeabilities are already reduced. The physical aging of gas permeability in PIM-1 membranes can be described by a power law, as established by Bernardo et al. [27]: where P0 is the initial permeability of the membrane at t = 1 and βP the permeability aging rate constant. The determined βP values for the original and scCO2-treated PIM-1 membranes are provided in Figure 8, as a function of the squared effective diameter of the gases studied, CO2 is at 0.091 nm 2 and N2 is at 0.092 nm 2 . The original PIM-1 membrane had very similar βP values to that reported by Bernardo et al. [27], and hence comparable physical aging with that study, which included an ethanol treatment post-fabrication. For the scCO2-treated PIM-1 there was a clear reduction in the aging constant for those membranes exposed for eight hours, along with a loss in correlation associated with gas diameter. The explanation to this behaviour is attributed to CO2 solubilising the polymer chain and creating a denser morphology the longer PIM-1 is exposed; however, scCO2 when desorbing leaves behind micro-cavities of sufficient size to enable depressurization. These micro-cavities then undergo physical aging, which was observed in the permeability loss of He and CO2; while CH4, N2 and O2 are larger molecules and thus size restricted from these post scCO2 micro-cavities, and hence do not experience physical aging to the same degree as their permeabilities are already reduced.

Conclusions
The permeability and selectivity of PIM-1 membranes for gas separation are impacted by exposure to scCO2, due to the process creating a denser membrane morphology. He, CH4, O2 and CO2 all experience a reduction in permeability through the membrane, while N2 permeability remains relatively constant, when PIM-1 is exposed to scCO2 for two or eight hours, at the end of which the process undergoes rapid or slow depressurization. This is attributed to scCO2 solubilising the PIM-1 polymer chain, enabling polymer chain rearrangement and altering the micro-cavity environment.

Conclusions
The permeability and selectivity of PIM-1 membranes for gas separation are impacted by exposure to scCO 2 , due to the process creating a denser membrane morphology. He, CH 4 , O 2 and CO 2 all experience a reduction in permeability through the membrane, while N 2 permeability remains relatively constant, when PIM-1 is exposed to scCO 2 for two or eight hours, at the end of which the process undergoes rapid or slow depressurization. This is attributed to scCO 2 solubilising the PIM-1 polymer chain, enabling polymer chain rearrangement and altering the micro-cavity environment. Interestingly, the physical aging of PIM-1 membranes is altered by exposure to scCO 2 . He and CO 2 permeabilities are observed to decrease significantly over extend time periods due to contraction of the micro-cavities within PIM-1; however, N 2 , CH 4 and O 2 permeabilities experience only a small reduction over the same time period, especially if the PIM-1 membrane had been exposed to scCO 2 for eight hours. This is attributed to the initial scCO 2 exposure contracting the larger micro-cavities, reducing N 2 , CH 4 and O 2 permeabilities because of size exclusion, as such subsequent aging in those micro-cavities does not impact those gases permeabilities because they are already excluded. Hence, scCO 2 exposure is a procedure that has potential implications in reducing the impact of physical aging in PIM-1-based membranes for N 2 , CH 4 and O 2 separation.