PIM-1 membranes containing POSS - graphene oxide for CO2 separation

Separation and Purification Technology 298 (2022) 121447 Available online 15 June 2022 1383-5866/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). PIM-1 membranes containing POSS - graphene oxide for CO 2 separation Sajjad Mohsenpour a , Ahmed W. Ameen a , Sebastian Leaper a , Clara Skuse a , Faiz Almansour a , Peter M. Budd b , Patricia Gorgojo a , c , d , * a Department of Chemical Engineering, Faculty of Science and Engineering, The University of Manchester, Manchester M13 9PL, UK b Department of Chemistry, Faculty of Science and Engineering, The University of Manchester, Manchester M13 9PL, UK c Instituto de Nanociencia y Materiales de Arag´ on (INMA) CSIC-Universidad de Zaragoza, C/ Mariano Esquillor s/n, 50018 Zaragoza, Spain d Departamento de Ingeniería Química y Tecnologías del Medio Ambiente, Universidad de Zaragoza, C/ Pedro Cerbuna 12, 50009 Zaragoza, Spain ARTICLE INFO Keywords: Graphene Oxide POSS, PIM-1 membrane Mixed matrix membrane CO 2 separation ABSTRACT PIM-1 mixed matrix membranes (MMMs) were fabricated with polyhedral oligomeric silsesquioxane (POSS) and graphene oxide (GO) functionalized with POSS (GO-POSS), and tested for CO 2 /N 2 (single gas) and CO 2 /CH 4 (1:1, v:v gas mixture). The CO 2 permeability of the best performing fresh MMM (containing 0.05 wt% GO-POSS) was ~ 12000 Barrer, which is 69% higher than that of the neat PIM-1 membrane, with about the same selectivity (CO 2 /CH 4 selectivity ~ 12 and CO 2 /N 2 selectivity ~ 20). In both cases, the gas separation data surpass the 2008 Robeson upper bound. In addition to the initial CO 2 permeability enhancement, the use of GO-POSS is an efficient strategy to slow down physical aging. The MMM at a filler loading of 0.75 wt% showed less than half of the reduction in CO 2 permeability than the neat PIM-1 membrane 160 days after preparation (26% for the MMM vs 58% for the purely polymeric one). 1. Introduction It has been widely accepted that the release of greenhouse gases such as carbon dioxide (CO 2 ) into the atmosphere from human activities is a major contributor to climate change. This in turn can increase the frequency and severity of extreme weather events such as heatwaves, floods and droughts [1]. According to the 2015 Paris agreement, global warming must be kept below 1.5 ◦C to prevent catastrophic effects [2]. One strategy to achieve this is to capture CO 2 from the atmosphere or directly from industrial processes using carbon capture and storage (CCS) technologies. However, these are currently expensive and energy intensive [3]. CO 2 presents additional problems for the production of electricity from lower-emission energy sources such as natural gas and biogas (which are mostly CH 4 ). Both natural gas and biogas contain CO 2 (2–50%) which must be removed because it is corrosive to gas pipelines and reduces the calorific value of the fuel [4]. Therefore, there is an immediate need for technologies that can separate CO 2 from methane (CH 4 ) and air (nitrogen (N 2 )). The adoption of polymeric gas separation membranes has grown within industrial applications due to their low cost, good processability and ease of fabrication [5]. However, the tradeoff between selectivity and permeability is the major drawback of the polymeric membranes, holding back their wider use for these applications. Polymers of intrinsic microporosity (PIMs) are superglassy polymers with high free volume that have attracted considerable attention as membrane materials for gas separations over the past two decades [610]. PIM-1, the most studied PIM, has high CO 2 permeability and is soluble in a few common organic solvents, including chloroform and tetrahydrofuran (THF), which allows for the fabrication of dense membranes via casting and solvent-evaporation. Physical aging and plasticization in high free-volume membranes are the main limiting factors when it comes to real-life applications. Aging is caused by the relaxation of the non-equilibrium polymer chains and results in freevolume collapse and substantial gas permeability decrease over time. The so-called plasticization, on the other hand, is caused by the sorption of condensable gases at high pressures that induces swelling, and leads to decreased selectivity [11]. To minimize aging and plasticization, different approaches have been proposed including polymer structure modification, crosslinking and addition of non-porous and microporous fillers (mixed matrix membranes (MMMs)) [12]. Rational selection of the filler with an appropriate structure (particle size, pore orientation, etc.) and resistance to aggregation (i.e. minimization of the interfacial * Corresponding author at: Departamento de Ingeniería Química y Tecnologías del Medio Ambiente, Universidad de Zaragoza, C/ Pedro Cerbuna 12, 50009 Zaragoza, Spain. E-mail address: [email protected] (P. Gorgojo). Contents lists available at ScienceDirect Separation and Purification Technology journal homepage: www.elsevier.com/locate/seppur https://doi.org/10.1016/j.seppur.2022.121447 Received 29 April 2022; Received in revised form 4 June 2022; Accepted 5 June 2022 Separation and Purification Technology 298 (2022) 121447 2 incompatibility between the filler and the polymer) is crucial to obtain MMMs with superior properties [13]. Porous aromatic frameworks (PAFs) [14], carbon nanotubes (CNTs) [15], polyhedral oligomeric silsesquioxane (POSS) [11,16], graphene [17-19], and metal organic frameworks (MOFs) [20] are some examples of filler that have been used for the preparation of PIM-1-based MMMs. Graphene oxide (GO) derivatives have been used as fillers in PIM-1 and other polymer matrices due to their high aspect ratio (>1000), high stability and abundant functional groups in GO for surface functionalization [21]. When GO flakes are dispersed in common organic solvents (such as those used to dissolve PIM-1) they do not form a stable colloidal suspension, and thus GO functionalization is needed. Our research group incorporated in a previous work octylamine (OA) and octadecylamine (ODA)- functionalized GO into PIM-1 membranes and slower rates of aging were reported [18]. This was due to the large lateral flake size of the functionalized GO, which induced rigidification of PIM-1 polymer chains. However, CO 2 permeability decreased with the addition of the fillers for freshly-prepared membranes, as a consequence of the non-porous nature of GO (i.e. increased tortuosity = increased path length of diffusing gas molecules). Work on the incorporation of other functionalized GO materials: (PIM-1)-functionalized (3-aminopropyl)triethoxysilane GO (APTS-GO) [19] and (PIM-1)-functionalized holey GO [22] followed, but decrease CO 2 permeability for non-aged membranes as compared to neat PIM-1 was also observed. To overcome the reduction in gas permeability, we present in this work the decoration of GO nanosheets with porous nanoparticles (NPs) that can create additional gas transport channels and still aid in reducing the physical aging in PIM-1 matrices. Porous aminopropyl isobutyl POSS NPs (AM0265, referred to here as POSS) anchored to GO and as received (without GO) have been used as fillers in PIM-1 membranes. Unlike GO, the POSS and the GO-POSS NPs (with a cage-like three-dimensional structure of 1–3 nm in size [23,24] and chemical structure shown in Figure S1) can both be dispersed in chloroform and THF, which are the organic solvents used to dissolve PIM-1. The high aspect ratio of the GOPOSS is expected to increase the interaction with the polymer chains, and thus aid in suppressing the physical aging phenomena while maintaining a high gas separation performance. 2. Experimental 2.1. Materials PIM-1 polymer was synthesized with monomers of 3,3,3′,3′-tetramethyl-1,1′-spirobisindane-5,5′,6′,6′-tetrol, 97% (TTSBI, Alfa Aesar (UK)) and 2,3,5,6-tetrafluoroterephthalonitrile, 99% (TFTPN, Fluorochem Ltd.). TTSBI was used after a purification step. For this, 15 g TTSBI was dissolved in 400 ml methanol (MeOH) (VWR International Ltd) and the temperature was increased to 120 ◦C while stirring the solution in an open beaker until 300 ml of MeOH had evaporated. Then 200 ml dichloromethane (DCM) (Sigma Aldrich, UK) was added to the beaker and the precipitate was filtered and dried under vacuum (10 mbar) at 50 ◦C for 5 h (~60% yield). Potassium carbonate (K 2 CO 3 ) was purchased from Fisher Scientific UK Ltd. Both K 2 CO 3 and TFTPN were dried in the vacuum oven at 25 ◦C overnight prior to use. Chloroform was provided by Fisher Chemical. N,N-dimethylacetamide (DMAc), toluene, and 1,4-dioxane were purchased from Sigma Aldrich (UK). Acetone was purchased from VWR International Ltd. Deionized water (DI) used in this study was produced by a Milli-Q integral system (Merck Millipore, Ireland). For GO functionalization with POSS, GO (1 wt% aqueous suspension) was purchased from William Blythe (Lancashire, UK), POSS powder was purchased from Hybrid Plastics (US), N,N’-dicyclohexylcarbodiimide (DCC), and tetrahydrofuran (THF) were purchased from Sigma Aldrich (Germany). 2.2. PIM synthesis PIM-1 synthesis was based on the procedure described in the work by Ameen et al. [7]. Briefly, TTSBI (10.79 g, 0.031 mol), TFTPN (6.26 g, 0.031 mol), anhydrous K 2 CO 3 (12.85 g, 0.093 mol), DMAc (60 ml) and toluene (30 ml) were added to a three neck round bottom flux and reacted under a continuous flow of dry N 2 using a mechanical stirrer at 160 ◦C. At the end of the reaction (after 36 min), the solution was added to MeOH to precipitate out the PIM-1, which was filtered, dried and redissolved in 300 ml chloroform overnight and subsequently reprecipitated using MeOH. The resulting polymer was washed with DI water overnight and subsequently with 1,4-dioxane, acetone, and MeOH. Finally, PIM-1 was dried in a vacuum oven at 120 ◦C for 2 days (yield ~ 95%). 2.3. Graphene oxide-POSS functionalization POSS functionalized GO was synthesized according to the method described by Xue et al. [25] 100 mg dried GO was dispersed in 50 ml of THF using a sonication bath (Elmasonic, 80 kHz frequency at 100% power) for 2 h. This was transferred to a 250 ml round bottom flask along with 2 g POSS and 100 mg DCC and sonicated for another 10 min, and then refluxed at 80 ◦C for 48 h. At the end of the reaction, the solution was poured into MeOH (500 ml). The precipitate was collected by vacuum filtration (homemade polyacrylonitrile filter (0.2 µm pore size)). The filtration procedure was repeated several times to remove unreacted POSS from the final filter powder (Figure S2) and the obtained powder was labelled GO-POSS48. In order to synthesise GO-POSS with higher POSS content, the functionalization reaction was repeated with higher initial POSS content (4 g) and longer reaction time (72 h), and the obtained powder was labelled GO-POSS72. The schematic of GO functionalization with POSS is shown in Figure S1. In contrast to GO, GO-POSS is soluble in chloroform (Figure S3). 2.4. Membrane fabrication To prepare the MMMs, first a GO-POSS/chloroform solution (0.8 mg ml −1 ) was prepared using a sonication bath (80 kHz frequency at 100% power) for 2 h. Aliquots of this solution were diluted with chloroform (total volume 4.5 ml) and stirred for 1 h. Following this, 0.14 g of PIM-1 powder was dissolved in each solution. The obtained solutions were stirred for 6 h, followed by 10 min sonication. Immediately after the sonication, the solutions were filtered through glass wool (Sigma Aldrich) and cast in 5 cm petri dishes. To reduce the solvent evaporation rate, the petri dishes were covered with a glass lid with a minimal gap and left in a fume cupboard to dry at room temperature for 18 h. Then, the solidified membranes were soaked in MeOH for 8 h and subsequently dried under vacuum (10 mbar) at 65 ◦C overnight (~12–14 h). MeOH treatment helps to remove the chloroform residue and increases the fractional free volume (FFV) and gas permeability [26]. The residue of solvent in the membrane affects its performance and it is much easier to remove MeOH from the polymer matrix than chloroform [27]. The PIM1 membranes prepared with GO-POSS48 and GO-POSS72 were labelled 48PGP(x) and 72PGP(x), respectively where x (=0.01, 0.05, 0.1, 0.25, 0.5, 0.75) referred to the percentage (wt%) of GO-POSS to PIM-1, and PGP are the initials in PIM-1/GO-POSS. In addition, to study the role of potentially available oxygen functional groups of GO on the gas separation performance, GO-POSS72 was reduced at 120 ◦C for 8 h following the procedure as described in a previous publication [28] and labelled rGO-POSS72. A digital photo of rGO-POSS72 dispersed in chloroform is shown in Figure S3. The MMMs prepared by this filler were labelled PrGP(x) (x =0.01, 0.05 and 0.1 wt%). Purely polymeric and PIM-1 membranes with only POSS (no GO) were also prepared and used as the control membranes. They were prepared following the same procedure as for the PGP membranes described in the previous paragraph, and were labelled PP(x) (x =0.02, S. Mohsenpour et al. Separation and Purification Technology 298 (2022) 121447 3 0.03, 0.05, 2, 5 wt%). It should be noted that GO or rGO could not be homogeneously dispersed in chloroform so they were not used as a filler. The thicknesses of the membranes were measured using a digital micrometer (Mitutoyo Corporation) with an accuracy of ±0.5 µm. For each membrane, 3–5 measurements were carried out and the average results are reported in Figure S4. In order to preserve the integrity of the membranes and be able to retest for gas separation after several weeks/ months, membrane discs were sandwiched between two circular pieces of aluminium with concentric circular holes and sealed with epoxy resin (Araldite Rapid, Industrial MTCE Suppliers). The effective membranes areas available for gas permeation after the sealing were calculated by ImageJ software and were ~ 0.1 cm 2 . 2.5. Materials and membranes characterization 2.5.1. PIM-1 polymer characterization The molecular weight of the synthesized PIM-1 polymer was measured using gel permeation chromatography (GPC, Viscotek GCPmax VE 2001 chromatograph (Malvern, UK)) [29]. The results were analyzed by OmniSEC software. 1 H Nuclear Magnetic Resonance (HNMR) spectroscopy was performed using a Bruker DPX 400 MHz spectrometer. Matrix assisted laser desorption/ionization-time of flight (MALDI-TOF) mass spectroscopy was carried out on a Shimadzu Biotech Axima Confidence instrument. Nitrogen adsorption isotherms were acquired using a Micromeritics ASAP 2020 sorption analyzer according to the procedure reported by Ameen et al. [7]. 2.5.2. Characterization of fillers and membranes The functionalization of GO with POSS was evaluated through a range of characterization techniques: X-ray photoelectron spectroscopy (XPS), attenuated total reflectance Fourier transform infrared (ATRFTIR) spectroscopy, X-Ray diffraction (XRD), thermogravimetric analysis (TGA), atomic force microscopy (AFM) and scanning and transmission microscopies. Some of these techniques (XRD, TGA and scanning electron microscopy (SEM)) were also used for the characterization of the prepared membranes. XPS was conducted using an Axis Ultra spectrometer (Kratos Analytical Limited, Manchester, UK) with a monochromatic Al K α source (1486.7 eV). ATR-FTIR was carried out in an iDS Nicolet iS5 spectrometer (Thermo Scientific, UK) with a wavenumber range of 400–4000 cm −1 . XRD measurements were carried out to analyse the interlayer space (d-spacing) of GO and GO-POSS, according to Bragg’s law (equation (1)) [30], and also to investigate the crystallinity of the prepared membranes. Samples were measured in a PANalytical X’Pert Pro equipment with Cu K α and Kβ emission lines in a 2 θ range of 4◦-50◦. d=λ 2sinθ (1) where λ is the wavelength of the X-ray beam (0.154 nm in this equipment) and θ (degree) is the diffraction angle. A TGA550 thermal analyser (TA Instruments) was utilized to study the chemical functionalization of GO-POSS and the thermal stability of the fillers and the MMMs. The analyses were performed at a heating rate of 10 ◦C min −1 under N 2 flow from 40 to 800 ◦C. The average size of GO-POSS flakes was measured using SEM (FEI Quanta 250 FEG-SEM). Transmission electron microscopy (TEM, FEI Tecnai G2 20) was conducted to investigate structural features of GO and GO-POSS flakes. The samples were prepared by drop-casting on a lacey carbon film on a 200 mesh copper grid (Agar Scientific). In addition, the thickness and surface morphology of GO and GO-POSS nanosheets were assessed using an AFM. The samples were prepared by dropcasting 30 µl of a 100 ppm GO-POSS solution on clean mica substrates. To study the morphology of prepared membranes, cross-sectional and surface SEM images were taken using an FEI Quanta 650 SEM. 2.6. Gas separation performance Single gas separation tests were conducted for different gas molecules including H 2 , CO 2 , CH 4 , and N 2 using the constant-pressure variable volume method. The permeate gas composition was analyzed using a gas chromatograph column (Agilent 490 microGC) which has two channels, a molecular sieve (MS) and a PoraPlotU (PPU) column. H 2 and N 2 were quantified from the peaks obtained in the MS, and the CH 4 and CO 2 permeability from the peaks in the PPU. Mixed gas separations were also carried out in the same setup, but feeding a mixture (CO 2 /CH 4 at 50:50 vol%) into the membrane cell. For CO 2 /N 2 separation, singlecomponent gases were used. The transmembrane pressure was ~ 1 bar and the membranes were placed inside an oven at a constant temperature of 25 ◦C. Argon (Ar) or helium (He) were used to sweep the permeate gas and send it to the MS and the PPU channels, respectively. The feed and sweep gas flow rates were 80 ml min −1 and 10 ml min −1 , respectively, to maintain the stage cut (permeate flow feed flow ) always below 2 wt% and constant feed composition. More details on the gas separation unit can be found elsewhere [7]. The gas permeability was calculated according to the following equation: Pi=QiL A(yipf−xipp)×1010 (2) where, P i is the permeability in Barrer unit (1 Barrer =10 -10 cm 3 (STP) cm cm −2 s −1 cmHg −1 ), Q i (cm 3 s −1 ) is the gas flow rate of component i, L (cm) is the membrane thickness, A (cm 2 ) is the effective membrane area. pf and pp (cmHg) are the total pressures in the feed and permeate side, respectively. yi and xi are the mole fraction of component i in the feed and the permeate side, respectively. The selectivity ( α ) was calculated by dividing the permeability of gas A (which shows higher permeability) by the permeability of gas B according to equation (3). α A/B=PA PB (3) For each membrane variant, 3 to 5 samples were prepared and average results along with the standard deviations have been reported. The tested fresh membranes were stored in a sealed petri dish. 48GP MMMs were re-tested after 7, 21, 55, 125, and 160 days and 72PGP MMMs were retested after 110 and 160 days in order to evaluate the physical aging. The CO 2 permeability drop and relative CO 2 permeability were calculated according to equations (4) and (5). Permeability drop =Pfresh CO2−Paged CO2(4) Relative PCO2=Paged CO2 Pfresh CO2 (5) To obtain more insight into the role of GO-POSS in the PIM-1 membranes, the diffusion coefficient (D, cm 2 s −1 ) and solubility (S, cm 3 (STP) cm −3 .cmHg −1 ) for CO 2 and CH 4 in selected MMMs and pristine PIM-1 membranes were obtained using a time-lag apparatus (constant-volume variable pressure method) at ambient temperature (~20 ◦C) and a feed pressure of 1.2 bar. The apparatus is described in more detail elsewhere [7]. The permeability (Pi(Barrer)) can be calculated by equations (6) and (7) and D and S were calculated with equations (8) and (9), respectively. Qi=VpϑSTPθi RT (6) Pi=QiL AΔp×1010 (7) Di=L2 6 τ i (8) S. Mohsenpour et al. Separation and Purification Technology 298 (2022) 121447 4 Si=Pi Di (9) where, Vp (53.68 cm 3 ) is the permeate side volume, ϑSTP (22400 cm 3 mol −1 ) is the molar volume of the gas at standard temperature and pressure. R (6236.36 cmHg.cm 3 K −1 .mol −1 ) is gas constant and T (K) is the absolute temperature. Δp (cmHg) is the average transmembrane pressure. L (cm) is the membrane thickness, θi (cmHg s −1 ) is the build-up of permeate side pressure and τ i (s) is time-lag which has been obtained from the permeate side pressure vs time graph. In this study, the permeability of the prepared MMMs was modelled using Maxwell-Wagner-Sillars equation (equation (10)). PMMMs =Pc[nPd+ (1−n)Pc− (1−n)(Pc−Pd)∅f nPd+ (1−n)Pc+n(Pc−Pd)∅f](10) where,PMMMs is the predicted permeability in Barrer. Pc is the permeability of the continuous phase (P PIM-1 , Barrer), Pd is the permeability of the dispersed phase (PGO−POSS, Barrer) and ∅f denotes the fractional volume of the dispersed phase. n is the filler shape factor ranging from 0 to 1. For prolate ellipsoids 0 ≤n ≤1/3 and for oblate ellipsoids 1/3 ≤ n ≤1. When the value of nis assumed 1/3 it means the fillers are spherical and equation (10) is known as the Maxwell equation. The limits of n =0 and n =1 represent the permeation of gas through parallel plates (horizontal) and laminate series plates (vertical), respectively [31]. This simple model has been widely applied for the prediction of permeation performance for membranes at low particle loadings (volume fraction ≤0.2) [31], and it has been successfully used for membranes containing 2D fillers [7]. To convert the mass fraction of the dispersed phase to the volume fraction, densities of PIM-1, GO and POSS were considered 1.1, 1.8 and 1.2 g cm −3 , respectively. The reported permeance of a 50 nm thick GO-POSS laminate membrane is 35 GPU [32]. That gives a permeability value of 1.75 Barrer, which is much lower than that of the PIM-1 membrane and can therefore be neglected. 3. Results and discussion 3.1. PIM-1 characterization The synthesized PIM-1 exhibited a weight-average molar mass (M w ) and number-average molar mass (M n ) of 153000 g mol −1 and 67500 g mol −1 , respectively, with dispersity (M w /M n ) of 2.26. The BET surface area was 756 m 2 g −1 according to the N 2 adsorption isotherms (Figure S5 of the supporting information), which is in agreement with the range reported in the literature (400–800 m 2 g −1 ) [33]. The 1 H NMR spectrum of the PIM-1 powder is shown in Figure S6, and the structure of the PIM-1 polymer is discussed in the supporting information. The 1 H NMR shows that the synthesized polymer has a proportion of branched structure. Figure S7 displays the MALDI TOF mass analysis of the synthesized PIM-1, which confirms its cyclic structure at low molar mass. The fragmentation product peaks are also observed which may originate from the presence of the branched structure that is consistent [7] with the 1 H NMR results. The weight percentages of carbon (C) and nitrogen (N) for the synthesized PIM-1 are 74.19 and 6.11, respectively with C/N ratio of 12.14, which is slightly lower than the theoretical value of C/N for a typical PIM-1 (12.43). This may be the result of the presence of the polymer structures terminated with a fluoro monomer (which has nitrogen in its structure) [34]. 3.2. GO-POSS characterization Unlike GO, the prepared GO-POSS samples can be dispersed in chloroform (pictures in Figure S3). XPS confirms the presence of silicon and nitrogen elements for the GO-POSS nanohybrids; Si 2p, Si 2 s and nitrogen peaks are identified in the XPS survey of GO-POSS72 (Figure S8b) at 99.5, 150 and 399.5 eV, respectively. Furthermore, the high-resolution spectra of N 1 s for GO-POSS72 (Fig. 1a) and GO-POSS48 (Figure S8d) show a peak at 399.9 eV that corresponds to amide bonds (O=C−NH) [32,35,36], and confirms covalent bond formation between GO and POSS (see proposed molecular model in Figure S1). The high-resolution spectrum of C 1 s for GO-POSS72 (Fig. 1b) shows a peak at 285.3 eV that corresponds to C-N [37-39], which is not found for GO (C1s spectrum in Figure S8e) and is further evidence of POSS presence in the prepared GO-based nanohybrids. The oxygen-containing functional groups in GO (carboxylic and epoxy groups) provide sites for the nucleophilic substitution reaction with amine groups in POSS. Therefore, the decline in oxygen percentage of GO-POSS in comparison with GO (atomic percentage values are shown in Table S1 of the supporting information) can be attributed to the chemical functionalization between amine and oxygen functional groups, as shown by the presence of amide bonds in the N 1 s spectra and as reported in other studies [40-42]. Some of the oxygen loss may be also ascribed to partial reduction of the GO at 80 ◦C (temperature of GOPOSS functionalization reaction). The higher Si and N content in the GO-POSS72 sample as compared to GO-POSS48 suggests higher POSS concentration at longer reaction times. FTIR spectra (Figure S9) are consistent with XPS results. The average interlayer distance of GO and GO-POSS nanosheets was calculated from the XRD data (Fig. 1c) using Bragg’s law (equation (1)); the characteristic broad GO peak at 11.5 ◦ gives a d-spacing of 0.77 nm, in agreement with reported values in the literature for GO [43]. For the GO-POSS nanohybrids the characteristic GO peak broadens and decreases its intensity, shifting to lower 2θ values of 10.1 ◦ (d-spacing = 0.87 nm) and 9.1 ◦ (d-spacing =0.97 nm) upon functionalization for 48 h (GO-POSS48) and 72 h (GO-POSS72), respectively. The increase in the interlayer space upon reaction suggests that covalently-bonded POSS moieties have been intercalated between the GO sheets, as reported elsewhere [25]. However, the higher increase in the interlayer space for GO-POSS72 is accompanied by a lower crystallinity, suggesting that longer reaction times lead to a more disordered structure. It is also worth noting that the characteristic (002) peak at 2θ ~ 25 ◦ that is normally reported upon reduction of GO is not observed for the GO-POSS nanohybrids (nor for the reduced sample rGO-POSS72), which further confirms successful intercalation of POSS, rather than reduction, at the reaction conditions. The TGA curves of GO and GO-POSS samples are shown in Fig. 1d. The GO weight reduced by ~10% up to 100 ◦C due to desorption of adsorbed water molecules, while no significant loss was detected for GOPOSS48 and GO-POSS72 because of their hydrophobic nature. For GO, significant weight losses in the range of 180–240 ◦C and above 550 ◦C were observed and ascribed to decomposition of oxygen functional groups and the graphite skeleton, respectively [19]. GO-POSS showed better thermal stability than POSS, which decomposes completely at ~300 ◦C. The thermal stability for GO is also improved after the reaction with POSS (final weight for GO at 800 ◦C ~ 2 wt% and >40 wt% for the GO-POSS samples) as amide bonds in GO-POSS prevent thermal degradation of the GO carbon skeleton [25,37]. The higher final weight percentage of GO-POSS72 than GO-POSS48 at 800 ◦C (46 and 40 wt%, respectively) suggests a slightly higher degree of functionalization (i.e. more amide bonds) for the sample obtained after reaction with 72 h. The average flake size of GO-POSS (0.79 ±0.32 µm) was calculated using Gaussian curve fitting from SEM images (Figure S10). The structural properties and height profile of GO and GO-POSS have been studied by AFM and TEM and some images are shown in Figures S 11&12. The higher height profiles for GO-POSS72 (Figure S11b) and GO-POSS48 (Figure S11c) as compared to the one for GO (Figure S11a) also confirm the presence of POSS on GO for GO-POSS samples. 3.3. Membrane characterization FTIR was used to study the possible interaction of the fillers (GOS. Mohsenpour et al. Separation and Purification Technology 298 (2022) 121447 5 POSS and POSS) and the PIM-1 polymer (Figure S13). The amide groups in GO-POSS and amine groups in POSS have –NH bonds in their structures which can form hydrogen bonds with cyano (-C≡N) groups in the PIM-1. It has been reported that the formation of hydrogen bonds can make peaks shift to lower wavenumbers (cm −1 ) with a decrease in peak intensity [44]. However, herein, no changes were observed between the FTIR spectra of PIM-1/GO-POSS, PIM/POSS and neat PIM-1 membranes, possibly due to the low concentration of fillers in the polymer matrix. It is worth noting that the N-H peak at 1463 cm −1 observed for POSS (Figure S9) is also observed in the GO-POSS samples. Nevertheless, for the MMMs the CH 3 /CH 2 stretching at 1440 cm −1 overlaps with the N-H peak and no information can be inferred from it. XRD spectra of the neat PIM-1 and the 72PGP MMMs are shown in Fig. 2a. The two main peaks of PIM-1 at 13.7 ◦ and 18.41 ◦ correspond to the micropores within the polymer structure created by the ladder-type backbone structure of PIM-1 and the chain-to-chain distance of spacepacked polymer matrix, respectively [45,46]. For the MMMs, the two peaks at 2θ =21.5 ◦ and 23.8 ◦ correspond to POSS. By increasing the GOPOSS content the intensity of the first PIM-1 peak decreased. Moreover, the second peak of PIM-1 and the two characteristic peaks of POSS moved gradually towards lower 2θ values (the position of the second peak of PIM-1 is shown in Fig. 2a. This can be attributed to the interrupted chain packing of PIM-1 [47] and greater chain d-spacing of the membrane due to the presence of GO-POSS [11,48]. Moreover, the XRD spectra of selected 48PGP MMMs (Figure S14a) confirm the same behaviour as 72PGP MMMs, however, changes in the peak locations are smaller. XRD spectra of all PIM/POSS (PP) MMMs are displayed in Figure S14b, and no shifts are observed. Neat PIM-1 and MMMs exhibit similar thermal stability according to the TGA results (Fig. 2b). The weight loss up to 460 ◦C corresponds to the removal of adsorbed moisture and volatile organic molecules [49]. The significant weight loss from 460 to 800 ◦C is due to polymer chain degradation and mostly ether linkage removal [19]. The cross-sectional SEM images of the neat PIM-1 and one of the MMMs (72PGP0.25) as an example are shown in Fig. 2 c&d. In certain areas of both membranes, wrinkles and polymer veins are visible, which are related to the cleanness of the fracturing method as described in section 2.5.2. The presence of the filler can increase the chance of these wrinkles appearing, with more appearing as the additive concentration goes up due to the interaction of polymer and additive [50,51]. A higher magnification SEM image of 72PGP0.25 is shown in Fig. 2d. The SEM images show no visible agglomeration of GO-POSS. The surface and cross-sectional SEM images of the other MMMs are shown in Figure S15. In some cases, the rapid evaporation of the low boiling point solvent (chloroform) resulted in small pinholes on the surface. However, the depth of the pinholes is usually less than 100 nm and they do not affect the performance of the membranes [52]. Fig. 1. High-resolution spectra of N 1s (a) and C 1s (b) for GO-POSS72, normalized XRD spectra (c) and TGA curves (d) of GO, POSS, GO-POSS48, and GO-POSS72. The XRD spectrum of the reduced sample rGO-POSS72 is also included in (c). S. Mohsenpour et al. Separation and Purification Technology 298 (2022) 121447 6 3.4. Gas separation performance 3.4.1. Fresh membranes Single-gas permeability tests were conducted at a transmembrane pressure of 1 bar at 25 ◦C using gases with different kinetic diameters including H 2 (2.89 Å), CO 2 (3.3 Å), N 2 (3.64 Å), and CH 4 (3.8 Å). The gas permeability vs kinetic diameter for PIM-1 and selected PIM-1-based membranes containing GO-POSS fillers (loadings of 0.05 and 0.25 wt % of GO-POSS48 and GO-POSS72) is plotted in a graph in Figure S16. Permeability values for all tested membranes do not follow the same trend as with size: PCO2>PH2 >PCH4 >PN2, which can be related to higher solubilities for CO 2 and CH 4 as compared to H 2 and N 2 , respectively, as reported elsewhere [53,54]. PIM-1 has polar groups (cyano and ether groups) in its structure that can encourage the sorption of gases such as CO 2 [53]. The PIM-1 membrane containing 0.05 wt% of GO-POSS72 deserves special attention, as it shows a more pronounced increase in CO 2 and CH 4 permeabilities as compared to all the other membranes. Solution-diffusion is the widely-accepted gas transport mechanism in PIM-1 membranes, and thus single gas measurements were performed on a constant volume/pressure increase instrument to calculate the solubility (S) and diffusivity (D) terms of the membranes (values shown in Table 1). It is observed that membrane 72PGP0.05 has ~ 4 times larger D than plain PIM-1, and reduced S (~half of the one for PIM-1). Therefore, the increase in permeability is due to a much faster diffusion of CO 2 through the MMM. It is worth noting that the GO-POSS flakes consist of non-porous GO flakes and porous POSS and the distance between two adjacent silicon atoms in POSS (4.4 Å) is bigger than the kinetic diameters of all the tested gases [48,55]. Therefore, the GO-POSS at such optimum loading must be creating additional free volume in the PIM-1 polymer matrix and/or creating additional faster permeation Fig. 2. Characterization of the neat PIM-1 and selected MMMs: XRD spectra (a) and TGA curves (b) of neat PIM-1 and 72PGP membranes (the dashed lines are a guideline to show the movement of the first peak of PIM-1 and POSS peaks in XRD), cross-sectional SEM image of a PIM-1 membrane (c) and 72PGP0.25 (higher magnification SEM image is shown inset) (d). Table 1 Gas permeability (P), diffusion (D) and solubility (S) coefficients and selectivity obtained with the time-lag setup for fresh membranes. Membrane P (Barrer) D ×10 7 (cm 2 s −1 ) S ×10 3 (cm 3 (STP) cm −3 .cmHg −1 ) CO 2 /CH 4 ideal selectivity CO 2 CH 4 CO 2 CH 4 CO 2 CH 4 PIM-1 6688 394 10.90 6.210 613.5 63.52 17.0 48PGP0.01 7321 444 41.14 7.576 178.0 58.62 16.5 48PGP0.05 8753 514 42.76 31.78 204.7 16.18 17.0 48PGP0.25 5168 637 13.55 2.908 381.5 219.1 8.1 72PGP0.01 8378 525 48.48 24.77 172.8 21.2 15.9 72PGP0.05 13,944 987 39.67 6.375 351.5 154.9 14.1 72PGP0.25 5347 325 25.13 6.992 212.8 46.44 16.4 PP0.05 11,029 648 39.36 16.82 280.3 38.52 17 PP5 5895 398 12.99 6.630 453.8 60.12 14.8 S. Mohsenpour et al. Separation and Purification Technology 298 (2022) 121447 7 paths through the POSS NPs, while reducing available sorption sites for CO 2 in the PIM-1. The membranes were tested for CO 2 /CH 4 (50:50 v:v) separation one day after preparation and the results are plotted in a bar-like graph in Fig. 3a. The neat PIM-1 membrane showed average CO 2 and CH 4 permeabilities of 7195 ±907 Barrer and 590 ±93 Barrer, respectively with a CO 2 /CH 4 selectivity of 12.3 ±1.2. A wide range of permeabilities (4000–7000 [16,56]) and CO 2 /CH 4 selectivity (10–20 [18]) have been reported in literature for PIM-1 due to varying polymer structures and experimental conditions. The PIM-1 synthesis can give rise to different chain structures and molecular weights, [34,57] and gas separation tests can incur a certain degree of error due to inaccurate thickness measurements, and pressure, temperature and feed composition-dependence behaviour of the membranes, which are rarely the same in literature. To determine the GO-POSS arrangement in the polymer matrix, the filler shape factor (n parameter in equation (10)) was optimized to obtain the best fit between predicted CO 2 permeability and experimental data (Figure S17). n value was found to be ~ 0.99 which means most GOPOSS nanosheets were arranged horizontally [7]. The gas separation performance of the membranes was also studied using single gases of CO 2 and N 2 at a transmembrane pressure of 1 bar, and a temperature of 25 ◦C. The results are presented in Fig. 3b. The neat PIM-1 membrane showed average CO 2 and N 2 permeabilities of 7323 ± 784 Barrer and 387 ±22 Barrer, respectively with CO 2 / N 2 selectivity of 18.9 ±1.1 which is in the range of CO 2 /N 2 selectivity that has been reported in the literature (16–25) [16,53]. CO 2 permeability of the MMMs indicated the same trend as CO 2 /CH 4 separation. However, for CO 2 /N 2 separation, all the PGP membranes showed slightly higher selectivity in comparison with the neat PIM-1 membrane, but the selectivities are in the range of error. The permeability and selectivity of all prepared membranes are presented in Table S2 of the supporting information. It should be noted that 72PGP0.75 was prepared but it was fragile and could not be tested. According to Fig. 3a, by increasing the GO-POSS48 and GO-POSS72 loadings up to 0.05 wt%, the CO 2 permeability increased to 8025 ±857 Barrer (CO 2 /CH 4 selectivity: 13.3 ±1.3) and 12185 ±1547 Barrer (CO 2 /CH 4 selectivity: 12.0 ±0.2), respectively. The gas transport through the membranes can happen by three paths: 1) PIM-1 free volume, 2) interlayer space of GO-POSS, 3) interface between filler and PIM-1 polymer chains. Further increases in filler content decreased the CO 2 permeability, possibly due to the rigidification of the polymer chains, as well as the increased tortuosity resulting from the high aspect ratio of GO laminates (deposited horizontally inside the polymer matrix as confirmed with the Maxwell equation). In addition, higher concentrations of GO-POSS may cause pore blockage, decreases the effective porosity as well as decreasing the membrane’s gas solubility due to the relative reduction in polymer content (Table 1) [11,58,59]. It has been reported that nonporous fillers can restrict the diffusion of larger molecules but they may provide near frictionless channels for smaller molecules and increase the selectivity of the membrane [60]. However, different behaviours have been reported by introducing nonporous fillers in the PIM-1 polymer structure. Luque-Alled et al. [19] prepared (PIM-1)-functionalized GO derivatives/PIM-1 MMMs, and all the membranes (regardless of the filler concentrations) showed lower permeability and higher CO 2 /CH 4 selectivity in comparison with neat PIM-1 membrane, however, on adding GO-OA and GO-ODA both permeability and selectivity decreased [18]. In this study, all the PGP MMMs have almost the same or higher selectivity than neat PIM-1. This can be attributed to the good compatibility of filler and polymer chains that can minimize interfacial voids. Moreover, the good solubility of GOPOSS in chloroform can facilitate adequate distribution of filler in the MMMs. The diffusion (D) and solubility (S) values of the fresh neat PIM-1 and selected MMMs for CO 2 and CH 4 are shown in Table 1. It should be noted that in the time-lag instrument, the membranes were first tested for CO 2 for 5 days and then for CH 4 . This may be the reason for the higher singlegas selectivity (due to aging) compared to the gas mixture (Fig. 3a). The highest permeability and diffusion coefficient occurred at the filler loading of 0.05 wt% for both 48PGP and 72PGP MMMs (same as the CO 2 /CH 4 gas mixture separation). Also, the D values of all MMMs are greater than neat PIM-1 with a maximum at 0.05 wt% of the filler. This confirms the higher FFV of the MMMs at low filler concentrations (caused by the disruption in the packing of the polymer chains) as well as PIM-1 pores blockage at higher loadings (0.25 wt% for PGP and 5 wt % for PP), which counteracts the first effect and thereby decreasing the gas diffusion coefficient. The permeability enhancement at low filler loadings can be due to the higher D values of the MMMs (due to chain packing disruption) [11,61] and also the larger d-spacing of GO-POSS compared to GO (meaning gas molecules can diffuse between adjacent GO-POSS sheets). It should be noted that the d-spacing of GO-POSS72 is larger than GO-POSS48 as confirmed by XRD results (section 3.2). According to Table 1, there is no specific relationship between the filler loadings and solubility coefficients; however, S values for all MMMs are smaller than that for neat PIM-1. This indicates that some of the sorption sites of PIM-1 are not accessible due to the presence of the filler [7]. PIM/POSS (PP) MMMs (without GO) were fabricated as described in the experimental section. The POSS loadings in the PP membranes were 0.02, 0.03, 0.05, 2 and 5 wt%. The range was selected based on the estimated content of just POSS in the best performing PGP membranes. The membrane with the highest permeability among the prepared PGP Fig. 3. Effect of GO-POSS48, GO-POSS72 and POSS on the gas separation performance of the MMMs for CO 2 /CH 4 (50:50 v:v) (a) and CO 2 /N 2 (single gases) (b). 3–5 samples of each kind were tested and average values and standard deviations are reported. S. Mohsenpour et al. Separation and Purification Technology 298 (2022) 121447 8 membranes has a loading of GO-POSS72 of 0.05 wt% (membrane 72PGP0.05), and the POSS mass fraction of filler GO-POSS72 is 0.46 based on TGA results (section 3.2). Thus, the concentration of only POSS in this membrane is 0.02 wt% (0.05 ×0.46 =0.02). Much higher concentrations up to 5 wt% were prepared to study the possible interaction of amine groups in POSS and CO 2 gas molecules, as it has been reported that amine groups can react with CO 2 and improve the separation performance [62]. The highest permeability was obtained at a loading of 0.05 wt% POSS (10575 ±609 Barrer), which represents a 47% increase in comparison with that of the neat PIM-1 membrane (Fig. 3a). The higher permeability of 72PGP0.05 compared to PP0.02 (same POSS content for both), may be explained by the higher FFV of the PGP membranes as confirmed by XRD results. Even though the kinetic diameters of the tested gases are smaller than the pore size of POSS (4.4 Å), their permeability decreases as POSS concentration increases (following a similar trend to that for GO-POSS), with a final permeability that was slightly lower than that of the pure PIM-1 membrane. There are two possible explanations for this. First, rigidification of PIM-1 polymer chain at high POSS loadings [63] and second, partial blockage of POSS pores by polymer chain [64]. In addition, different types of POSS have been reported in PIM-1 [11] and poly[1-(trimethylsilyl)-1-propyne] (PTMSP) [65] membranes. These studies have reported that despite their porosity, POSS nanoparticles act as nonporous fillers, increasing tortuosity and extent of polymer pore blockage at high loadings and thus reducing membrane permeability. This is evidenced by the much lower value of CO 2 diffusivity for PP5 (Table 1) as compared to the values of PP0.05 and the rest of the tested MMMs. Figure S16j shows an inset with a high magnification cross-sectional SEM image of a PP5 membrane that suggests good dispersion and no agglomeration of POSS nanoparticles into PIM-1. Moreover, the CO 2 /CH 4 selectivity of the PP membranes slightly decreased by increasing POSS concentration from 0.05 to 5 wt% (11.8 vs 10.4). The addition of high loadings of up to 5 wt% of POSS (membrane PP5) does not lead to higher solubility diffusivities and thus, low interaction (or no interaction) of available amine groups in POSS with CO 2 can be expected. Yong et al. [11] fabricated PIM-1/DiSilanolIsobutyl POSS MMMs. The highest permeability was achieved at the lowest POSS loading (0.5 wt%), which showed a 35% enhancement in CO 2 permeability in comparison with a neat PIM-1 membrane. This was attributed to the higher FFV of MMMs as well as the reaction of Si-OH group on POSS with CO 2 . However, at higher POSS loadings (>2 wt%) the pores were blocked and the d-spacing of the polymer chains decreased, which induced an increase in CO 2 /CH 4 selectivity and a decline in CO 2 permeability. Konnertz et al. [66] indicated that the highest permeability of PIM-1 MMMs was at the lowest loading of 1 wt% PhenethylPOSS. However, the permeability of the MMMs at lower filler loadings (less than 1 wt%) was not investigated. The permeability enhancement at low filler loading was assigned to the disrupted polymer chains. The effect of a more intense thermal treatment of the GO-POSS nanohybrids (reduced sample rGO-POSS72 obtained at 120 ◦C for 8 h) on the gas separation performance of the MMMs was studied and results are plotted in a graph in Figure S18. Membranes containing rGO-POSS at a loading of 0.05 wt% also showed very similar CO 2 /CH 4 selectivity values and higher CO 2 permeability than purely PIM-1 membranes. However, the permeability increase was not as pronounced as that of MMMs with non-reduced GO-POSS72 nanohybrids. The XRD spectra of both fillers (Fig. 1c) indicate a slightly more disordered material for the reduced sample, but there are no significant differences in their structures which can account for the differences in permeability. Therefore, it is hypothesized that the thermal treatment at 120 ◦C can lead to the disappearance of remaining un-reacted oxygen-containing functional groups still present after the functionalization reaction (as reported elsewhere [28]), that can in fact affect the gas separation performance of the PIM-1 membranes. Yet, the mechanism for this is unclear and further research should be conducted. The effect of feed (CO 2 /CH 4 ) composition at the constant transmembrane pressure of 1 bar on the performance of selected membranes was studied and is shown in Figure S19. By increasing the volume fraction of CO 2 in the feed, the more condensable penetrant species (CO 2 ) tend to occupy the Langmuir sites (microvoids in the polymer matrix) and reduces the permeability of CH 4 . Therefore, CO 2 /CH 4 selectivity for a gas mixture is slightly higher than the ideal selectivity (single gases) [67]. In addition to competitive sorption on Langmuir sites, plasticization (which occurs at high CO 2 partial pressures) can also contribute to a change in selectivity. To study the effect of pressure, neat PIM-1, 72PGP0.25, and 48PGP0.25 membranes were tested at a total feed pressure in the range of 20–130 psi (1.4–9 bar), which is the limit in our setup, using a CO 2 /CH 4 (50:50 v:v) mixture as the feed (Figure S20). On increasing the pressure, the CO 2 permeability decreased and no sign of plasticization was observed up to 9 bar. This is in agreement with literature that shows no plasticization of PIM-1-based membranes up to 440 psi (30 bar) [11]. Lower permeability at higher pressures may be due to membrane compression (pore size reduction). It is worth noting that the permeability reduction is more significant for the neat PIM-1 membrane as compared to that of the MMMs, which can be due restriction in polymer chain mobility when the filler is added. The CO 2 / CH 4 selectivity, meanwhile, stays almost constant with increasing pressure. 3.4.2. Aged membranes The aging performance of PGP and PP MMMs for CO 2 /N 2 and CO 2 / CH 4 separation was studied and the results are shown in Table S2 of the supporting information. For each membrane variant, 3 to 5 samples were prepared. For some of the aged membranes only one sample survived after 160 days, probably due to loading/unloading into the gas separation cell several times, and therefore standard deviations are not provided for those samples. Graphs with CO 2 permeability and CO 2 /CH 4 selectivity versus aging time for the neat PIM-1 membrane and selected MMMs are displayed in Fig. 4 a&b. Values for the CO 2 permeability drop (calculated with equation (4)) and the relative CO 2 permeability (calculated with equation (5)) are shown in Fig. 4 c&d, respectively. For each type of filler (GO-POSS48, GO-POSS72 and POSS), the results of two membranes that have the highest initial permeability (48PGP0.05, 72PGP0.05 and PP0.05) and the best aging behaviour (48PGP0.75, 72PGP0.25 and PP2) are shown. In addition, the results of 48PGP0.25 are indicated for a fair comparison with 72PGP0.25. Polymer chain relaxation over time decreased the free volume of the membrane, and this affected the diffusion of molecules with a larger kinetic diameter more than the smaller molecules, [68] which translated into the general trend of increased CO 2 /CH 4 selectivity upon aging that was observed (Fig. 4b). It is worth noting that there were random decreases in selectivity for some of these samples in between the testing period, that could be due to defects from loading/unloading of the membrane coupons several times into the gas separation cell. For each membrane, 3 to 5 coupons were prepared and tested, for instance, 5 fresh samples were tested for 48PGP0.25, but only two samples survived after 125 days. It is therefore possible that small defects formed in some of the aged membranes during handling, specially in the epoxy resin/ membrane region (thus increasing permeability), that became more evident in the next aging test (due to further membrane handling) and thus that coupon was disregarded for that last and subsequent data points. According to Fig. 4c, the permeability drop was more pronounced during the first 21 days, followed by a gradual decrease, which is in agreement with the literature [68]. After 160 days of aging, the membranes containing 0.05 wt% of fillers (48PGP0.05, 72PGP0.05 and PP0.05) showed a higher CO 2 permeability drop than that of neat PIM-1. However, the permeability of all MMMs was higher than that of purely polymeric PIM-1 membranes. In addition, the CO 2 permeability reduction of the aged PP MMMs was higher than that of PGP membranes (at S. Mohsenpour et al. Separation and Purification Technology 298 (2022) 121447 9 the same POSS concentrations), which confirms the positive effect GO flakes has on physical aging inhibition. Both the initial CO 2 permeability and the permeability drop decreased by increasing the concentration of the filler above 0.0.5 wt%. 48PGP0.75 exhibited the lowest permeability drop (1153 Barrer) and the highest relative CO 2 permeability (0.74), which was attributed to the higher rigidification effects (interactions of the filler with the polymer chain) at higher filler loading [22]. It should be noted that from the operational point of view a membrane with higher stability is preferred over a membrane with higher initial permeability but higher permeability drop. The stability is vital for thin film composite (TFC) PIM-1 membranes because of the more pronounced effect of physical aging; for instance, a 200 nm thick PIM-1 membrane can lose 95% of its initial CO 2 permeability in less than three months after preparation [69]. The aging behaviour of PIM-1 membranes can be affected by different parameters including the PIM-1 polymer chain structure, [7,70] membrane thickness, [71] post-treatments (alcohol (MeOH and ethanol) treatment), [68] and membrane storage conditions. Foster et al. [34,70] studied the effect of the presence of branched PIM-1 chains on the performance and aging behaviour of neat PIM-1 membranes. They found that the PIM-1 membrane with the branched-chain experienced slower physical aging compared to the PIM-1 membrane with a linear polymeric structure. In our study, the presence of the branched PIM-1 structure is confirmed by NMR and MALDI-TOF characterizations. This can be the reason for the higher relative CO 2 permeability of the aged neat PIM-1 membrane in this investigation compared to our previous study (Table 2) [18]. Moreover, the alcohol-treated membranes demonstrate a slight increase in thickness accompanied by size shrinkage. This can be caused by stress-relaxation of the membranes and induces an enhancement in preventing physical aging [7]. In addition, thermal treatment (membrane drying after alcohol treatment) aids the polymer chain relaxation and brings about a lower permeability drop over time [68]. Different nanosheets and POSS have been used as a filler in the literature to suppress the physical aging of the PIM-1 membrane. The initial and relative CO 2 permeability of these membranes after a specific aging time are compared in Table 2. It should be noted that the highest loading of the filler does not always increase the inhibition of physical aging due to the filler agglomeration above the optimum concentration. Alberto et al. [18] prepared PIM-1/graphene-like materials MMMs at the concentration range of 0.01–0.25 wt% and the best physical aging impediment was achieved at a low loading of the filler (0.05 wt%). Ameen et al. [7] prepared PIM/boron nitride nanosheets (BNNs) at two different BNNs concentrations of 0.5 wt% and 0.8 wt%. The best membrane performance in terms of aging was achieved at 0.5 wt% BNNs. The CO 2 /CH 4 and CO 2 /N 2 separation performances of the selected prepared membranes have been compared with some of the MMMs Fig. 4. Evolution up to 160 days of CO 2 permeability (a), CO 2 /CH 4 selectivity (b), CO 2 permeability drop (c), relative CO 2 permeability (d) of PIM-1 and selected MMMs. S. Mohsenpour et al.