Investigation of cellulose acetate and ZIF-8 mixed matrix membrane for CO2 separation from model biogas

ABSTRACT The separation of CO2 from the biogas mixtures (CH4/CO2) is essential for biogas upgradation. However, polymer membranes used for CO2 separation exhibit low permeability. Mixed Matrix Membranes (MMMs) incorporating inorganic filler in the polymer enhance CO2 separation. In this work, bio-degradable cellulose acetate (CA) based MMMs with varying filler weight percentages (2–20 wt.%) of ZIF-8 were studied for the separation of CO2 from a model biogas (CH4/CO2) mixture. The MMMs were characterized by analysis of TGA and DSC for thermal stability and FTIR for alteration or formation of any new functional group. FESEM was done to evaluate the dispersion and interaction of ZIF-8 in the CA polymer matrix. Considering the economic aspect, the fabricated MMMs were tested for gas separation performance at reasonably lower feed pressure (1.5, 2 bar). MMM with 5 and 10 wt.% of ZIF-8/ CA MMMs showed the best performance with CO2 permeability of 9.65 Barrer and 9.5 Barrer, approximately two folds as compared to pure CA, and CO2/CH4 selectivity was 10.37 and 15.3. The experimental results were compared with the predicted gas permeation results determined using MMM transport predictive models, and found that the permeabilities were higher than the model predictions. GRAPHICAL ABSTRACT


Introduction
The growing world population's high energy consumption has led to unprecedented greenhouse gas (GHG) emissions, resulting in global warming and a host of negative impacts, including melting glaciers, rising sea levels, climate change, and loss of biodiversity.Among the contributing GHGs, carbon dioxide (CO 2 ) is particularly significant.Therefore, it is essential to take measures to reduce GHG emissions, such as replacing conventional fuels with renewable fuels (e.g.biogas), carbon dioxide capture, storage, and utilization.Several conventional techniques, such as adsorption, absorption, and cryogenic separations, have been developed for carbon dioxide separation and capture [1].However, these methods require high energy consumption and significant investment.As a result, alternative methods, such as membrane separation, an energyefficient and portable method of carbon dioxide capture, are being explored [2].
Membranes can be fabricated from polymers and inorganic materials and are the semi-permeable barriers that allow different gas species to transport at differing rates.Polymers, owing to their ease of fabrication, mechanical stability, and separation properties, are the first choice of researchers and commercial membrane separation.Cellulose acetate (CA), polyethersulfone (PES), and polyimide (PI) are employed commercially for gas separation membranes [3,4].The performance of membranes is characterized using metricspermeability and selectivity.Permeability describes the productivity of the membrane, and selectivity of the membrane describes the efficiency of separation for a gas species.However, there is a well-known trade-off between permeability and selectivity for polymer membranes, as demonstrated in two literature surveys in 1991 and 2008, with a relation called Robeson's upper bound [5,6].
Novel membranes incorporating inorganic nanoparticles in the polymer matrix, known as Mixed Matrix Membranes, are being investigated extensively to overcome this trade-off and enhance the separation.Several combinations of inorganic nanoparticles and polymers are investigated for their compatibility and separation performance [7].Among the inorganic nanofillers, zeolites, silica, Carbon Molecular sieves, and Carbon nanotubes (CNTs) were explored [1].However, the compatibility and defect-free MMMs showing high permeability and high selectivity remain a challenge [8].Recently, there has been a great interest in Metal-Organic Frameworks (MOF) as nanofillers in MMMs due to their high porosity and surface area, and tunable pore size.
MOFs are metal nodes linked with organic linkers.Several MOFs are being extensively researched, such as ZIF-based MOFs (Zeolite Imidazolate Frameworks), UiO-66 (University of Oslo), and MIL (Material Institute of Lavoisier)-based MOFs, among others [9].ZIFs fall under a subclass of the MOF family for their zeolite-like topology with narrow apertures linking large cavities is known for their excellent transport properties.A bond angle of approximately 145°was found, similar to the Si-O-Si bond angle in the alumina-silicate zeolite.ZIFs are extensively studied for CO 2 separation owing to their high porosity, tunable large pore volume, and high surface area, and high thermal and chemical endurance.The pore size of the ZIF-8 is 3.4 Å, whereas CO 2 is 3.3 Å and CH 4 is 3.8 Å; thus, molecular sieve separation of CO 2 /CH 4 is possible.Synthesis methods for ZIF-8 particles are simple and do not require heat treatments like those for zeolites [10].It was reported that reducing the size from micro (11 µm) to nano-sized Cu-BTC MOF (35 nm) demonstrated enhancement in the CO 2 permeability and CO 2 /CH 4 selectivity from 138.6 to 178.9 Barrer and 33.9-34.6,respectively [11].Thus, synthesized ZIF-8 filler that is further subjected to size reduction can give better results.
Gas transport through polymeric membrane takes place by solution-diffusion mechanism, where gas species are dissolved or absorbed at the higher pressure side (feed side) and diffuses through the membrane owing to the difference in chemical potential, and then desorbs at the lower pressure side (permeate side).In MMMs, the second transport mechanism is due to the dispersed phase, i.e. the inorganic fillers.The filler is highly porous, showing a higher CO 2 permeability [12].Molecular sieving appears in the transport mechanism when filler having a kinetic diameter in a range of the two species is dispersed in the polymer matrix.If the filler adsorbs and desorbs one gas species faster than the second gas species, then a dual sorption mechanism of gas transport is observed [13].
In this study, MMMs were fabricated from ZIF-8 nanoparticles and bio-degradable cellulose acetatea combination of MMM not reported in the literature.Cellulose acetate polymer has been used extensively for membrane-based research.CA is the first polymer used in water-based separation studies such as ultrafiltration and reverse osmosis [14,15].CA possesses other suitable membrane properties such as ease of casting, low cost, and good thermal and mechanical stability.In this study, MMMs with filler loadings of 2-20 wt% were fabricated.MMMs were characterized to investigate the filler dispersion in the polymer matrix, thermal degradation, and change in crystallinity.Finally, the separation performance of MMMs was tested at two feed pressures of 1.5 and 2 bar for a model biogas.The experimental results of CO 2 permeability and CO 2 /CH 4 selectivity were compared with predicted data employing well-known transport models such as Maxwell, Lewis-Nielsen, and Bruggeman.Finally, the experimental results were compared with literature data.

Synthesis of ZIF-8
ZIF-8 was synthesized using the procedure reported in the literature [16], briefly described here.First, 1.5 g of Zinc nitrate hexahydrate (Zn(NO 3 ) 2 .6H 2 O) and 1.67 g of 2-Methylimidazole were dissolved in 50 ml of methanol separately.Once the solids were completely dissolved, the mixtures were combined and stirred for 1 h.The resulting milky white solution mixture was centrifuged at 7000 rpm for 5 min, and the solids obtained were washed with methanol and centrifuged again; this process was repeated three times.The final solid particles were dried in an air oven at 85°C for 6 h and then in a vacuum oven at 100°C overnight.The obtained particles were fed to a planetary ball mill and ground for 1 h, and the final ZIF-8 in powder form was stored for further analysis.

Membrane fabrication
2.2.1.Pure CA membrane fabrication Cellulose acetate powder was dried in a hot air oven at 60°C for 4 h to remove the adsorbed moisture.The dense membrane was prepared using 15 wt% of the polymer solution in DMF solvent.During pre-experiments, it was noted that the solution preparation procedure and drying of the casted membrane had greatly influenced the permeability and selectivity of the pure membrane.Here, the optimized membrane fabrication is reported.Approximately 10% of the total polymer powder was initially added to the solvent (DMF) and stirred for 2 h.This is called the priming method, which allows the uniform mixing of polymer in a solvent [17]; the remaining polymer is added stepwise.The solution mixture was stirred overnight at ambient temperature; then, the solution was sonicated for 2 h to break any polymer lumps.The polymer solution was then kept in a vacuum oven at room temperature for 1 h to remove any air bubbles formed during stirring.Finally, the polymer solution was cast on a flat glass petri-dish cleaned with methanol.Then, the petri dish was covered with a lid and kept in an air oven at 60°C for 4 h.Then the petri-dish was transferred to a vacuum oven with a lid removed and kept for drying at 60°C for 2-3 days until the film was peeled off.After peeling off the petri-dish, the membrane was dried in a vacuum oven at 60°C for 12 h.Finally, obtained pure CA membrane was stored for further characterization and gas permeation testing.

ZIF-8/CA membrane fabrication
ZIF-8/CA-based mixed matrix membranes were prepared with varying wt% (2,5,10,15,20) fillers.Firstly, a predefined amount of as-synthesized ZIF-8 was added to DMF and sonicated for 1 h to ensure proper mixing.Then, CA of 3 g was added to the mixture following the same priming technique used for pure CA membrane and stirred overnight at room temperature.The MOF-polymer mixture was sonicated for 2 h to break the polymer lumps and ensure more uniform dispersion.The dope solution was cast on a petri dish, and a similar drying procedure was followed as pure CA membrane.The filler wt.% was calculated as the weight of the filler over the total weight of the polymer and filler.

Membrane characterization
X-Ray Diffraction (XRD) analysis confirmed the crystal structure of ZIF-8 using a Rigaku Miniflex X-Ray diffractometer with Cu Kα radiation (k = 1.54 Å) with 30 kV and 15 m.The data was collected at 5-40°at a scanning rate of 2°C/min.Thermogravimetric analysis (TGA) was conducted using Perkin-Elmer TGA-4000.The pure MOF, pure CA, and MMMs samples were heated from 30-600°C at the rate of 10 °C/min and cooled at the same rate under N 2 atmosphere.Differential Scanning Calorimetry (DSC) (Perkin Elmer, DSC 4000) analysis was carried out to study the thermal properties of pure CA and MMMs.The membranes were heated from 20 to 250°C with a heating and cooling rate of 10°C/min in the presence of nitrogen.The second heating scan of the membrane was used for analysis.Fourier Transform Infrared (FTIR) characterization was done using a Perkin Elmer Frontier Spectrometer.The sample preparation was done using a KBr press Model M-15.The FTIR was conducted by applying a spectrometer in the IR range of 4000-400 cm −1 .The evaluation of functional groups of ZIF-8, pure CA, and MMMs was done.Further characterization of ZIF-8 and MMMs was done by Field-Emission Scanning Electron Microscope (FESEM) (FEI, ApreoLoVac model).The crystallite size and surface morphology, as well as filler distribution and interaction with polymer, were studied.The samples were coated with a gold layer and placed under the microscope for analysis.

Gas permeation experiments
All membranes were tested at room temperature (∼30°C ) and feed side pressure of 1.5 and 2 bar.Gas permeation testing was conducted in a module made of SS316 in which the membrane is placed between two Viton® O-rings; leak testing was performed to ensure no leaks or bypass.The setup schematic is previously reported [18] and given in supplementary information, figure S1.A mixture of model biogas, i.e.CH 4 :CO 2 (60:40) was used as the feed gas, while pure N 2 was used as the sweep gas.The retentate side was connected to a back pressure regulator (BPR) to maintain pressure inside the permeation cell.The retentate and permeate outlet streams were connected to the gas chromatography (Shimadzu, GC-2014) equipped with Thermal Conductivity Detector (TCD).The outlet concentrations obtained were used to calculate the permeability of CO 2 (P CO2 ) and selectivity of CO 2 /CH 4 (a CO 2 /CH 4 ) for each membrane.To ensure repeatability, experiments were repeated twice, and the average data is reported here.
Two performance metrics of membrane separation permeability and selectivity are calculated.The permeability (Barrer units, 1 Barrer = 10 −10 cm 3 (STP) of gas species is given by Equation (1).
Where, Q = Flow rate of the species on the permeate side, l = Thickness of the membrane, A = effective area of the membrane, Δp = partial pressure difference of the gas species on permeate and retentate side.The selectivity a CO 2 /CH 4 is given by Equation (2).

Models for predicting gas separation performance
The experimental data are compared with existing mathematical models such as Maxwell, Lewis-Nielsen, and Bruggeman models for predicting gas permeability.These models generally existed for electric or thermal conductivity models due to the similarities in analogies between electric or thermal conduction where the permeation of species through the composites, such conductivity models are used to predict gas permeability through the MMMs.Maxwell model, in particular, was one of the first models used to predict the permeability of gases in MMMs.It can reasonably predict the permeability of gas at lower filler concentrations, preferably less than 20% volume.Another limitation of the Maxwell model is that it does not consider the factors like particle size, distribution, and agglomeration of filler, which usually happens at higher filler loading.The Maxwell model is given by Equation (3).
Where, P MMM , P f, P p are permeabilities of MMMs, filler (dispersed phase), and polymer (continuous phase), respectively.Ø f is the volume fraction of filler in MMMs.The pure ZIF-8 permeability for CO 2 and CH 4 are taken as 2819.6 Barrer and 698.9 Barrer, respectively [19].Lewis-Nielsen model was originally developed to predict the elastic modulus in composite materials.Unlike Maxwell and Bruggeman's model, this model considers factors regarding the dispersion of fillers in the continuous phase by accounting for the factor for maximum filler loadings.Lewis-Nielsen's model considers filler morphology.The Lewis-Nielsen equation is given by Equation ( 4) and maximum filler volume fraction Ψ m is calculated by Equation (5), For non-ideal MMMs with moderate filler loadings, the Bruggeman model was applied.This model was derived from the dielectric model in heterogeneous material and was developed using a differential effective medium approach, which is a further modification of the Maxwell model.However, the model is an implicit model, which makes solving equations slightly complex.The Bruggeman model is given by Equation (8).

Optical images of membranes
Figure S2 represents the images of MMMs with varying filler weight% (0-20 wt.%).The pure cellulose acetate (CA) membrane was clear and transparent, while the MMMs changed their colour from transparent to milky white with increased wt.% of ZIF-8.The MMMs with 20 wt.% ZIF-8 in CA was brittle and completely opaque.

Membrane characterization
XRD of as-synthesized ZIF-8 is shown in Figure 1.The ZIF-8 showed several diffraction peaks at 7.5°(110), 11.2°( 200), 13.5°(211), 16°(220), and 17.5°(222), which indicates the crystal structure typical ZIF-8 which is reported in the literature [20].The peaks are sharp and welldefined, which indicates the sodalite structure of ZIF-8 without any impurities.Figure 2 represents the temperature vs. weight-loss curve for pure ZIF-8, pure CA, and MMMs.TGA of pure ZIF-8 showed weight loss of around ∼8% °C, corresponding to the removal of H 2 O molecule and solvent molecules, in the temperature range of 30-255°C.The gradual weight loss in the temperature range 200°C to 300°C is attributed to the removal of unreacted species i.e. 2-methylimidazole, from the nano-particle's surface [21].The TGA curve showed a long plateau in the temperature range from 300°C to 600°C, indicating the high thermal stability of ZIF-8 [22].For pure CA membrane, a minute weight loss of ∼3% at 100°C was observed, which can be attributed to moisture removal that might have been absorbed at room temperature.Then weight loss between 310°C -400°C was observed due to chain degradation of CA polymer.After 400°C, the weight loss observed was assigned to the carbonization of the sample [23].
The MMMs followed the same trend of weight loss.The MMMs start degrading at ∼200°C.Therefore, adding ZIF-8 in MMMs did not significantly alter the thermal properties.DSC analysis for pure ZIF-8, pure CA, and MMMs are given in Table 1.The glass transition temperature (T g ) for pure CA membrane was 183°C [24].As filler weight loading increased from 2 to 20 wt.%, the T g slightly shifted from 183 to 188°C.Adding filler in the polymer matrix may have increased the polymer chain stiffness.The addition of filler may cause less movement of the polymer chain.The increment trend in T g of MMMs with filler weight % was observed [24].
FTIR spectrum for pure ZIF-8, pure CA, and MMMs is given in Figure 3.The ZIF-8 spectra showed peaks of 2900-3100 cm −1 , indicating aromatic, alkane, and alkene stretch inside the imidazole unit.The peak that appeared at 1585 cm −1 was associated with C-N stretching and while the peaks around 1350-1500 cm −1 showed the entire ring stretching of the aromatic ring inside 2-methylimidazole.The peaks found in the range of 900-1350 cm −1 were because of the alkane in-plane blending, and peaks below 800 cm −1 were due to the out-plane blending of alkane.The Zn-N peak was found around 420 cm −1 [25,26].For pure CA membrane, FTIR peaks showed stretching in 3750-3500 cm −1 , representing the OH (hydroxyl) group.A -CH (aliphatic) bond was observed at a vibration peak around 3000 cm −1 and a 1600 cm −1 peak indicating the stretching of the C=O bond [27].The C-H peak in pure CA samples was intense compared to all MMMs.Also, it gradually weakens due to the increased concentration [28].Because of lower filler loading, no significant    alteration or addition of peaks was observed in the MMMs spectra [29].Figure 4a,b shows the FESEM images of ZIF-8.The ZIF-8 particles are spherical, and ball milling has reduced the particle size from ∼250 nm to ∼100 nm.The size reduction of filler has shown improved results. Figure 4c-n show the top and cross-sectional views of pure and MMMs to evaluate for morphology and filler distribution in the polymer matrix.The pure CA membrane was smooth and dense, with no cracks or pinholes.As ZIF-8 fillers are added, the morphology of MMMs changes from smooth to fibrous ones.The MMMs were also dense, with no cracks or pinholes.Also, the crosssectional view of MMMs showed uniform distribution of ZIF-8 filler.The uniform dispersion of ZIF-8 particles was due to the organic linker, which increased the compatibility between ZIF-8 and CA [30].The filler agglomeration started appearing at 15 wt.% and at 20 wt.% severe agglomeration was observed in both the top view and cross-sectional view.Additional cross-sectional FESEM images at various filler loadings are shown in Figure S3, showing the nanofiller dispersion in the polymer.

Gas permeation
MMMs gas separation performance was tested for mixed gas feed, CO 2 :CH 4 (40:60).Figure 5 shows the separation performance of pure and MMMs for two different pressures (1.5 bar and 2 bar).Figure 6a shows the CO 2 permeability (P CO2 ), Figure 6b shows the CH 4 permeability (P CH4 ), and Figure 6c shows the selectivity a CO 2 /CH 4 .It can be noted in cases of MMMs at various filler percentages that the CO 2 /CH 4 selectivity is higher than the With the incorporation of ZIF-8 fillers in CA, the permeability of CO 2 increased until 15 wt.% and then decreased at 20 wt.%.The permeability of CO 2 was increased by 41% at 1.5 bar.This increment is attributed to the molecular sieving effect of ZIF-8 with a pore aperture size of 3.4 Å, between the molecular size of CO 2 , and CH 4 of 3.3 and 3.8 Å, respectively.A similar increment in CO 2 permeability was observed at 2 bar.Due to low filler loading, the CO 2 permeability of 2% MMMs was almost comparable with pure CA.The CO 2 permeability was increased from 4.829 to 9.65 Barrer (more than two folds) at 1.5 bar and 6.398 to 9.5 Barrer.It can be observed an increased permeability but decreased CO 2 /CH 4 selectivity compared to pure CA.This can be due to the complete frigid saturation and limited flexibility, where ZIF-8 absorbs more gases of kinetic diameter ∼3.4 Å [31].
Pure CA polymeric membrane is a dense membrane and exhibits a solution-diffusion mechanism for gas transport [31].Initially, CO 2 molecule adsorbs at the feed side of the membrane.Further, due to the concentration gradient, it diffuses through the membrane and finally desorbs at another side of the membrane.Adding ZIF-8 provides a faster pathway for gas transport by providing a molecular sieving effect.Thus, we can see improved CO 2 permeability with adding ZIF-8 in CA except at 2 wt.% loadings and 20 wt.% loadings.At 2 wt.%, decreased CO 2 permeability can be associated with restricted polymer chain mobility, which is not observed at moderate filler loadings (5 wt.%, 10 wt.%, and 15 wt.%).At 20 wt.%, the decreased CO 2 permeability is associated with the agglomerates formed, which can adversely affect the performance.
A lower selectivity of MMMs was observed compared with the pure CA membrane.This may be attributed to nano-defects at the polymer-filler interface [32].As discussed above, gas transport through pure CA membrane is due to solution-diffusion mechanism.Thus, pure CA membranes possess lower permeability.Whereas CA inherently shows higher CO 2 affinity, resulting in a faster CO 2 transport rate (as compared to CH 4 ) through it, resulting CO 2 /CH 4 selectivity.The selectivity was retained at ∼10 for 1.5 bar and ∼12 for 2 bar for filler loading of 0 to 20 wt %.The maximum selectivity obtained at 5 wt% MMMs was 10.37 for 1.5 bar and 15.3 for 2 bar.Among the fabricated ZIF-8/CA MMMs, 5 wt.% and 10 wt.% membranes showed the best gas separation performance with CO 2 permeability of 9.65 Barrer, 9.5 Barrer, and CO 2 /CH 4 selectivity of 10.73, 15.3 at 1.5 bar and 2 bar, respectively.MMMs demonstrated a slight decrement in the CO 2 permeability around 8.76 Barrer and 8.63 Barrer at higher filler loadings compared to maximum CO 2 permeability, which can be due to the ZIF-8 particle agglomeration observed in SEM images.With increased feed pressure (from 1.5 bar to 2 bar), an improvement in the selectivity of ZIF-8/CA MMMs was observed.This is attributed to more permeation of penetrant CO 2 than CH 4 [33].

Comparison with transport models for MMMs
The Maxwell, Lewis Nielsen, and Bruggeman models were applied for predicting the CO 2, and a comparison of experimental and predicted permeability is shown in Figure 6.The predicted permeability for CH 4 and comparison with the experimental data is given in supplementary information Figure S4.The permeability of CO 2 increases for the MMMs until 15 wt.% for 1.5 bar and 2 bar and decreases afterward at higher filler loadings due to the particle agglomeration, which can be supported by FESEM analysis.As per Maxwell, Lewis-Nielsen Higuchi, and Bruggeman models, permeability for both CO 2 and CH 4 are expected to increase.The experimental results reported in this study show improved permeability initially for moderate filler loadings (5 and 10 wt.%) following the increasing trend of the Maxwell, Lewis-Nielsen Higuchi model, and Bruggeman model.Further, at 20 wt.% filler in MMMs, permeability starts decreasing and follows the decreasing trend.Increased CO 2 permeability can be due to two reasons.First, ZIF-8 possess inherent size selection property for CO 2 gas molecule, resulting from 3.4 Å pore size.Second, the dispersion morphology of ZIF-8 in CA.The experimental selectivity of MMMs drops initially as compared to pure CA but increases as the filler content increases in MMMs.At moderate filler content, the increased selectivity is associated with size exclusion property of MOF particles, which ultimately increases the CO 2 transport through the MMMs.Initially, the gas permeation was due to the solution-diffusion mechanism at ∼2 wt.% filler content.The addition of filler disturbed the diffusion pathways through the membrane and provided a faster pathway for CO 2 transport through MMMs.
At filler loading 5 and 10 wt.% in MMMs, the experimental values follow the Maxwell and Lewis-Nielsen.The Maxwell model being simple model does not consider any defects arising due to interface and shape variation of filler while Lewis-Nielsen considers defects arising from shape of filler.This demonstrates that at filler loading 10wt.% and below, the filler particles are well dispersed in MMMs with no defects and voids.This was also confirmed from SEM images.At higher filler (15 wt.% and above) loading in MMMs, the experimental values follow Lewis Nielsen, and Bruggeman models.The Bruggeman model is for non-ideal MMMs formed due to rigidification of polymer and voids.Thus, higher filler loadings, the agglomeration of filler is the occurring which is also supported by SEM data, and the rigidification of polymer at nanoscale around the filler-polymer interface may have started.Experimental selectivity is lower than predicted by models, which is attributed to the presence of a very thin nanoscale nonselective interfacial region between filler and polymer.This non-selective interface being thin nano-scale layer acts as pathway for gas molecules, but also offers resistance to the gas flow as compared to the pathway provided by the channels through the ZIF-8 porous structure.Thus, MMMs shows better CO 2 permeability but exhibit lower selectivity performance when compared to predicted ones.
Most of the literature reports ideal selectivity, higher than the mixed gas selectivity, as a partial pressure gradient acts as the driving force for the transport [40].Furthermore, gas transport in the presence of competing gases in mixed gas feed differs from the pure gas feed [41,42].In the present study, no grafting or interfacial modification, or functionalization of the ZIF-8 filler is performed.Therefore, it may be one of the reasons for the decrease in selectivity, although an increase in permeability.MMMs synthesized in this work were tested for mixed gas feed and lie below Robeson's upper bound.However, the fabricated MMMs are environmentally friendly and can be potentially used at low operating pressure.

Conclusions
This work reports the fabrication of MMMs by incorporating the ZIF-8 nanoparticles (2-20 wt.%) in the bio- degradable CA polymer matrix.The XRD of as-synthesized ZIF-8 nano-particles is similar to that in the literature.An improvement in the glass transition temperature is observed, from 183°C to 188°C for filler loadings from 2 to 20 wt% indicating the polymer chain stiffness of MMMs improved by incorporating ZIF-8 nanoparticles.Also, MMMs followed the same trend of TGA when compared to pure CA membrane.All membranes start degrading at ∼200°C, demonstrating that the incorporation of filler did not affect the thermal stability of MMMs.FTIR for MMMs showed all peculiar bands for CA, but with increasing filler weight %, the peak intensity of MMMs dropped.SEM images showed the uniform dispersion of ZIF-8 nanoparticles in CA polymer was observed until 10 wt%.Agglomerations started appearing in the MMMs after 15 wt%.The 5 wt.% and10 wt% MMMs showed the best performance with CO 2 permeability of 9.65 Barrer and 9.5 Barrer (more than two folds as compared to pure CA) and CO 2 /CH 4 selectivity of 10.37 15.3 at 1.5 and 2 bar, respectively.The fabricated MMMs do not cross Robeson's upper bound.Due to the simple and easy fabrication method, use of a bio-degradable polymer matrix, and operation at lower pressure, this membrane can be used for CO 2 separation at the industrial level.

Figure 7 .
Figure 7. Graph for Robeson's plot for permeability for this work and literature.

Table 1 .
The glass transition temperature of pure CA and MMMs with 2-20 wt.% filler.