Recovery of free volume in PIM-1 membranes through alcohol vapor treatment

Physical aging is currently a major obstacle for the commercialization of PIM-1 membranes for gas separation applications. A well-known approach to reversing physical aging effects of PIM-1 membranes at laboratory scale is soaking them in lower alcohols, such as methanol and ethanol. However, this procedure does not seem applicable at industrial level, and other strategies must be investigated. In this work, a regeneration method with alcohol vapors (ethanol or methanol) was developed to recover permeability of aged PIM-1 membranes, in comparison with the conventional soaking-in-liquid approach. The gas permeability and separation performance, before and post the regeneration methods, were assessed using a binary mixture of CO2 and CH4 (1:1, v:v). Our results show that an 8-hour methanol vapor treatment was sufficient to recover the original gas permeability, reaching a CO2 permeability > 7000 barrer.


Introduction
Since the commercialization of the first membrane-based system for gas separation applications three decades ago, a wide range of polymers have been investigated and developed, but less than ten have made it to commercial use [1]. Non-network polymer of intrinsic microporosity (PIM-1), synthesized for the first time at the University of Manchester two decades ago [2], has shown great potential as a membrane material at laboratory scale, due to its high chemical, thermal and mechanical stability, as well as high gas permeability (arising from its high free volume) and fair selectivity values. All these properties also make it attractive for commercial gas separation applications [3,4]. However, as generally happens to high free volume polymers, PIM-1 is prone to lose the void space rapidly due to physical aging, which results in the loss of membrane performance over time. The polymer chains are initially packed in a non-equilibrium state, leaving excess free volume. Over time, loss of free volume occurs by the gradual rearrangement of the polymer chains towards an equilibrium state, leading to an increase of polymer density. As a consequence the permeability of the membrane is reduced, compromising its long-term stability. This effect is more pronounced at the beginning of its lifetime, with a significant decline in permeability, the rate of which decreases with time [5,6]. Similarly to other polymers [7], the rate of physical aging in PIM-1 membranes is also thickness-dependent, being faster in thin films compared to thicker films [8,9]. Other factors, such as temperature, pressure, atmosphere and storage conditions, also play an important role in the physical aging rate [10][11][12].
It is worth noting that several other solution-processable PIMs synthesized over the past few years have overcome the well-known Robeson 2008 upper bound for certain gas pairs [13,14]. However, as with PIM-1, they also suffer from physical aging [15][16][17], and PIM-1 is still the most investigated non-network PIM. Different approaches have been explored to prevent or minimize physical aging in PIM-1 membranes, including: 1) modification and synthesis of new polymer structures [6,18], 2) post-modification treatments (e.g., thermal oxidative crosslinking [19], the use of supercritical CO 2 , and ultraviolet treatment [20]), and 3) addition of fillers to the polymer matrix, giving socalled mixed matrix membranes [21][22][23]. Physical aging is a reversible process, unlike others such as degradation, chemical aging and contamination that affect membrane properties irreversibly [1]. In some glassy polymers, such as polysulfone and polyimide, the membrane performance is recovered by annealing the sample above its glass transition temperature (T g ), followed by rapid quenching to room temperature [24]. However, in polymers with very high or undetectable T g , the group to which PIM-1 belongs, this is not possible [1]. In this case, the 'rejuvenation' of the membrane performance is typically undertaken by soaking the films in lower alcohols, such as methanol or ethanol [8,12,13,25]. Alcohols swell PIM-1, causing an enhancement of the molecular motion of the polymer chains that leads to an increase in the free volume of the polymer [26]. In addition, the alcohol treatment also flushes out any residual solvents and contaminants trapped within the polymer and erases the past processing history [12]. Similarly, other studies have reported the posttreatment of membranes with a range of solvents to enhance their performance in applications such as organic solvent nanofiltration [27,28], reverse osmosis [29], nanofiltration [30], ultrafiltration [31,32] and pervaporation [33].
Albeit soaking in alcohol is a simple procedure, which is typically done after membrane casting, it requires the removal of the membrane from the gas separation apparatus when it is used to recover permeability of aged membranes. This might be acceptable for small scale and fundamental studies, however, at the industrial level, this procedure would be completely unattainable; membrane modules used in gas separation are not designed to cope with liquid, and a treatment that does not require membrane dismantling would be desirable. Moreover, we envisage that the treatment with alcohol vapor would not cause any further disturbance in the gas separation system, and the use of an inert gas or air following the vapor treatment could eliminate any traces of alcohol vapor and leave the membrane module ready to use. Therefore, in this work we present an alternative that consists in exposing aged PIM-1 films to ethanol or methanol vapor four months after casting, measuring their gas permeability before and after the alcohol treatment as a means of tracking recovery of free volume. One of the most promising applications of PIM-1 membranes at commercial scale is the sweetening of natural gas. Therefore, the separation of the relevant gas pair CO 2 /CH 4 was investigated.
2.2 Synthesis of PIM-1 PIM-1 was synthetized following the procedure proposed by Du et al. [34]. Firstly, 10.00 g of TFTPN, 17.29 g of TTSBI and 20.73 g of K 2 CO 3 were reacted in a mixed solvent of 100 mL of dimethylacetamide and 50 mL of toluene under N 2 at 160°C. Reflux and constant mechanical stirring of the solution was set for 40 min, after which a highly viscous solution was produced. This solution was soaked in methanol and left overnight. The yellow crude product was collected by vacuum filtration and washed three times with methanol and a final wash with acetone. Then, it was dissolved in chloroform and reprecipitated in methanol. The recovered product was refluxed for 12 h in water at 100°C and then dried under vacuum at 110°C overnight.

Fabrication of membranes
Thirteen freestanding PIM-1 membranes were prepared through a solvent evaporation method. First, the synthesized PIM-1 powder was dissolved in chloroform at a concentration of 4 wt-% under magnetic stirring for 24 h. Then, the dope solution was cast on Steriplan ® petri dishes (5 cm diameter) that were immediately covered and left in the fume cupboard for 3 d at room temperature. The thickness of the membranes was measured using a digital micrometer screw gauge with an accuracy of AE0.5 mm (Mitutoyo IP65 Coolant Proof, UK). At least five measurements were performed on each membrane to obtain an average thickness.

Membrane treatment
From the prepared PIM-1 membranes, three were tested without any alcohol treatment and were labelled as PIM-1 as-cast ( Table 1). The rest were subjected to treatment with methanol or ethanol the day after collection from the petri dish (day 1). For the treatment, membranes were immersed in alcohol for 24 h at room temperature, and subsequently placed in a vacuum oven at 120°C for 10 h to remove any trapped residual solvents. Further treatment with alcohol vapor was carried out for two of the ethanol-treated membranes and two of the methanol-treated ones, using the same vaporized alcohol as that in the initial solvent treatment. PIM-1 membranes were placed in a membrane cell and exposed to the alcohol vapor generated by the heating of either ethanol or methanol at 50°C. A vacuum of 10 mbar was applied on the downstream side of the membrane in order to force the passage of the alcohol vapor across the membrane. This procedure is illustrated in Fig. 1 with details on the membrane cell used. Table 1 displays the codes given to the membranes.
These membranes were treated either with liquid ethanol or liquid methanol the day after collection from casting petri dishes (day 1). Some of these membranes were further aged and some were treated with methanol or ethanol vapor after at least 114 d have passed. Aging time is the number of days after removal from petri dishes. As-cast PIM-1 membranes were also tested for comparison. The membranes were labelled using the following code: PIM-1-(alcohol used in the treatment: MeOH or EtOH)-L(day of soaking in liquid alcohol)-V(day of vapor alcohol treatment)-a or b to designate different membranes. For example membrane PIM-1-MeOH-L(1)-V(120)-a is a   membrane first soaked in methanol at day 1 and treated with methanol vapor 119 d later.

PIM-1 powder and membrane characterization
Weight-average molecular weight (M w ), number-average molecular weight (M n ) and polydispersity index (PDI) of the PIM-1 polymer was determined by gel permeation chromatography (GPC). This analysis was carried out on a Viscotek GPC mac VE 2001 instrument using tetrahydrofuran as the eluent at a flow rate of 1 mL•min -1 and an injection volume of 100 mL using Viscotek GPC VE2001 columns and Viscotek 3580 refractive index detector. Polystyrene standards were used for the calibration of the system. The apparent surface area of PIM-1 was determined by Brunauer-Emmett-Teller (BET) analysis of the N 2 adsorption isotherm at 77 K. For that, PIM-1 powder was exposed to nitrogen at 77 K using a Micromeritics ASAP 2020 volumetric adsorption analyser (Micromeritics, USA). PIM-1 sample was first degassed under high vacuum at 60°C for 12 h.
Thermogravimetric Analysis (TGA) was conducted to study the thermal stability of the PIM-1 polymer powder and the prepared films. A TGA 550 thermal analyser (TA instruments, USA) was used, heating from room temperature to 800°C at a rate of 10°C·min -1 .
The morphology of the freestanding PIM-1 membranes was studied by scanning electron microscopy (SEM) using a FEI Quanta 200 microscope (FEI, USA). The cross sectional samples were prepared via cryo-fracturing in liquid nitrogen. The samples were sputtered with Platinum nanoparticles using an Emitech sputter coater (Quorom Technologies, UK) before imaging.
The chemical structure of PIM-1 was qualitatively studied using attenuated total reflection Fourier transform infrared spectroscopy (FTIR). The spectra were acquired with an iDS Nicolet iS5 instrument (Thermo Scientific, UK), equipped with a Ge crystal over the wavenumber range of 600-4000 cm -1 and step size of 0.5 cm -1 .

Gas permeation measurements
Gas permeation measurements were performed for CO 2 /CH 4 binary mixture (1:1, v:v) with a flow rate of 25 mL•min -1 of each gas. The feed solution was pressurized at 3 bar, while the permeate side was kept at atmospheric pressure. Helium (flow rate of 20 mL•min -1 ) was used as a sweep gas to dilute the permeating gases and direct them to a gas chromatography system (490 micro GC, Agilent, USA) equipped with a PoraPLOT (PPU) column for the analysis of the permeate composition. The experiments were conducted at 25°C. The gas permeability for each gas (P i ) was calculated using Eq. (1): where i is either CO 2 or CH 4 , Q i is the permeate flow of gas i (cm 3 (STP)•s -1 ), l is the membrane thickness (cm), A is the effective membrane area (cm 2 ) and Dp is a the partial pressure difference across the membrane (cmHg) for gas component i. The results are given in barrer units (1 barrer = 10 -10 •cm 3 [STP]•cm•cm -2 •s -1 •cmHg -1 ). The selectivity for the gas pair is calculated as the ratio of the permeabilities of the most permeable gas (P CO 2 ) over the less permeable gas (P CH 4 ), as shown in Eq. (2): 3 Results and discussion

PIM-1 and membrane characterization
The M w , M n and PDI of the PIM-1 synthesized in this study are 118800 g•mol -1 , 32900 g•mol -1 and 3.6, respectively. Figure 2(a) shows the N 2 adsorption/desorption isotherms of PIM-1 powder at 77 K. It can be observed that PIM-1 shows high N 2 uptake at very low relative pressures. The BET surface area of PIM-1 is 657 AE 8 m 2 •g -1 , which is in the typical range of values reported for PIM-1 in the literature. The thermal stability of PIM-1 powder and freestanding PIM-1 membranes was investigated through TGA. Thermogravimetric curves are shown in Fig. 2(b). It can be observed that both samples present similar thermal behavior. Below 120°C any weight loss was due to the removal of residual solvents that might be weakly adsorbed by the material [35]. Both PIM-1 powder and PIM-1-freestanding membranes are thermally stable up to 450°C, as the mass loss is not significant up to that temperature. The surface and cross-section of a liquid-ethanol treated PIM-1 membrane before aging (PIM-1-EtOH-L(1)-V (120)-b), are shown in micrographs a and c in Fig. 3, respectively, and confirm the preparation of defect-free dense membranes. In addition, the cross section of an ascast PIM-1 membrane is shown in Fig. 3(b). Some elongated deformations are present in both cross sections, with more intense creases for the as-cast membrane; these features can typically arise from the "freeze-fracture" process during sample preparation and do not indicate the presence of defects or voids within the dense membrane structure, as suggested by the gas separation performance discussed in section 3.2. The amount of residual solvent within the polymer matrix, after solvent evaporation at ambient conditions or even after vacuum treatment, influences the quality of the prepared samples; the presence of trapped solvent can lead to a more effective freezing of the polymer membrane and therefore, samples with less artificial topology, as observed in Fig. 3(c). The thickness measured for the membranes via SEM is in accordance with that measured with the micrometer.

Evaluation of aging in PIM-1 membranes via gas permeation
Soaking membranes in polar solvents such as ethanol and methanol is often used to remove any trapped solvent or contaminants within the PIM-1 microstructure, and also to reverse the effects of aging of the films. When alcohols penetrate PIM-1 they induce swelling, polymer chains relax and this allows residual solvent to be flushed out [13]. After the removal and evaporation of the alcohol, films present an enhancement of gas permeability due to the increase of free volume. In this regard, the gas separation performance was studied for CO 2 /CH 4 mixtures of untreated (as-cast) PIM-1 membranes, and methanol-and ethanol-treated membranes (fresh and aged ones). The CO 2 permeabilities and CO 2 /CH 4 selectivities obtained for all the membranes are shown in Table 1 and plotted in permeability and selectivity graphs in Fig. 5. Untreated fresh PIM-1 membranes showed the lowest average CO 2 permeability (2755 AE 137 barrer) due to the blockage of the pores by residual casting solvent still present in the microstructure of the membranes, as has previously been reported in the literature [13,37]. When soaked in lower alcohols, the average CO 2 permeability of fresh PIM-1 membranes increased, more when methanol was used (6272 AE 142 and 4513 AE 1228 barrer for methanol and ethanol, respectively). As expected, at day 27 (ca. 4 weeks after casting), the CO 2 permeability of methanol and ethanol-treated membranes dropped by 39.3% and 29.7%, respectively. This behavior has been previously reported in the literature [21,38] and confirms the tendency to lose free volume within the polymer structure (i.e., physical aging). It is worth noting that in a previous paper from our group a very similar CO 2 permeability loss was reported, 35.9% at day 35 for freestanding PIM-1 membranes initially soaked in liquid methanol [21].
As can be observed in Table 1, liquid methanol-treated PIM-1 membrane showed higher overall CO 2 permeability right after the alcohol treatment, and also ca. 4 weeks later, as compared to ethanol-treated PIM-1 membranes. As a consequence of physical aging, the average CO 2 permeability dropped from 6272 AE 142 barrer down to 3810 AE 868 barrer in just 26 d, and the CO 2 /CH 4 selectivity increased to 18.1 AE 4.0. These results show a superior overall performance of methanol-treated membranes compared to ethanol-treated ones, as more free volume was created upon methanol treatment. This was also observed by Mason et al. [6]. In that study, methanolsoaked aged PIM-1 films showed not only a full recovery of their gas permeabilities but also an increase compared to ethanol-treated PIM-1 films. McDermott et al. studied the effect of ethanol and methanol treatment of aged PIM-1 films on their small-angle X-ray scattering and wide-angle X-ray scattering patterns [25]. Since the porosity in PIMs, which is correlated to free volume, affects the scattering patterns, these patterns can be used to study the physical aging. Larger absolute intensity corresponds with larger porosity and, consequently, larger free volume. Their findings showed that methanol-treated PIM-1 films have larger scattering intensity compared to ethanol-treated ones, which in turn is associated with higher porosity and therefore correlated to free volume. This suggests that methanol use might be more effective as a rejuvenating procedure [25]. The effectiveness of methanol treatment in reversing the physical aging effects can also be related to the higher vapor pressure of methanol (16.96 kPa at 25°C) compared with ethanol (7.85 kPa at 25°C). The higher the vapor pressure, the stronger the driving force for the alcohol to evacuate the polymer structure once it is removed from the liquid, leaving additional free volume in the PIM-1. The rate of evaporation is related to the time needed for the chains to rearrange themselves in the matrix. Therefore, the higher the evaporation rate, the less time the chains have to reorient which causes an increase in free volume [11].
Vapor solvent treatment has been previously reported to enhance the mechanical properties and separation performance of electrospun membranes [39,40]. However, to the best of our knowledge, this is the first study that addresses the use of vapor solvent to reverse the effect of physical aging. CO 2 and CH 4 gas permeabilities and CO 2 /CH 4 selectivity of methanol and ethanol vapor-treated membranes were monitored as a function of time. The values obtained are shown in Table 1 and also displayed in Fig. 5. It can be observed that all aged PIM-1 membranes show a decrease in gas permeability and increase in selectivity over time, similar to what is reported in other studies [9,21]. As previously explained, this is due to the loss of excess free volume and consequent rearrangement and packing of the polymer chains, and follows the trade-off between permeability and selectivity commonly seen in polymeric membranes [41]. According to Bernardo et al. [42], the effect of physical aging on the permeability, P, of aged PIM-1 membranes can be described by Eq. (3): where P 0 is the initial permeability at t = 1 and β p is the permeability aging rate constant. Figure 6 displays β p values for PIM-1 membranes aged for ca. 4 months (last four membranes in Table 1 [43], which are also plotted in Fig. 6 for CO 2 , CH 4 along with other gases. Only membrane PIM-1-EtOH-L(1)-V(120)-b shows higher β p values, possibly due to the lower extent of opening up free volume with the initial ethanol soaking; it is worth noting that this membrane shows the lowest initial gas permeability of all, which upon aging for 120 d, falls below 2000 barrer. When exposed to alcohol vapor ca. 4 months after casting, aged PIM-1 membrane PIM-1-MeOH-L(1)-V (120)-a and PIM-1-MeOH-L(1)-V(114)-b showed a full recovery of their CO 2 permeability, with values of 7020 and 7165 barrer, respectively, which are slightly above those reported initially right after the liquid-methanol treatment. On the other hand, membranes PIM-1-EtOH-L (1)-V(124)-a and PIM-1-EtOH-L(1)-V(120)-b, which were initially treated with liquid ethanol and aged for about the same period of time, also recovered their initial CO 2 permeability but the values were more scattered and ca. 50% lower. These results are again in accordance with the literature that reports the higher effectiveness of using methanol for making free volume of PIM-1 more accessible [6]. The fact that the permeability of methanol through freestanding PIM-1 membranes is approximately 3-fold the permeability of ethanol [44], also suggest a more effective treatment when the one-carbon solvent is used. Figure 7 shows the behavior of the methanol-treated membranes PIM-1-MeOH-L(1)-V(120)-a and PIM-1-   MeOH-L(1)-V(114)-b on a CO 2 /CH 4 selectivity vs CO 2 permeability double logarithmic plot. As anticipated, all PIM-1 membranes lie along the 2008 upper bound [45], since it was one of the polymers Robeson used to revisit the previous upper bound from 1991 [46]. As membranes age over time the data points move towards lower CO 2 permeability and higher CO 2 /CH 4 selectivity, following the 2008 trade-off upper bound. At days 114 and 120, membranes PIM-1-MeOH-L(1)-V(120)-a and PIM-1-MeOH-L(1)-V(114)-b were exposed to methanol vapor for 8 h, respectively, and gas performance evaluated the following day. The initial performance was recovered in both cases, with CO 2 permeability values in the 7000 barrer range and CO 2 /CH 4 selectivity around 11.

Conclusions
This study shows an effective procedure to reverse the effect of physical aging on PIM-1 membranes by exposure to vaporized alcohols (methanol or ethanol). It is anticipated that this could be done in situ at large scale gas separation plants. While the results obtained in this study are promising, more membranes need to be tested under different conditions, especially at higher feed pressures and at gas compositions closer to those of natural gas, in order to have a deeper and better understanding of the reversal of the physical aging effects under real operating conditions. Also, it would be useful to explore if reversibility of physical aging by means of alcohol vapor treatment would only be required upon installation of the membrane modules and after plants shutdowns; a contrasting effect of aging, i.e., plasticization, which typically occurs in glassy polymer membranes due to the presence of CO 2 at high pressures of a few tens of bar (typical operation conditions of sweetening of natural gas), most likely will impede physical aging during membrane operation.
These findings suggest the effectiveness of the vapor methanol treatment as a 'rejuvenation' process which may bring the use of PIM-1 membranes in industrial applications one step closer.  [47]. The arrows show the aging process over the tested period and they point towards the most aged membranes.