Preparation and Characterisation of Blend Cellulose Acetate Membrane for CO2/N2 Separation

Among the carbon dioxide (CO2) separation methods, the membrane separation technology has been mostly applied to reduce the CO2 emission. Since most CO2 are emitted from power generation plants, the CO2/nitrogen (N2) separation was selected. In this project, the blend cellulose acetate (CA) membranes were prepared by mixing CA with 39% acetyl concentration (CA-39) and 56% acetyl content (CA-56) through the wet-phase inversion method. The CO2/N2 separation performance was determined by evaluating the permeation of gases and the CO2/N2 selectivity by modifying the polymer concentration. The characterisation of the membrane was carried out by using the scanning electron spectrometry (SEM) and attenuated total reflectance Fourier transform infrared spectrometry analysis (ATR-FTIR). The optimal membrane obtained was M3 with polymer concentration of 4:6 (CA-39:CA-56) that had defect-free membrane surface with smooth dense skin layer. The functional groups present enhanced the gas permeance rate. With regards to the gas permeation test, M3 presented the best CO2/N2 separation performance with CO2 permeance of 99.26 ± 3.08 GPU, N2 permeance of 87.12 ± 0.81 GPU and CO2/N2 selectivity of 1.139 ± 0.037. It is expected that the results obtained from this work can be applied to industries, such as the gas separation field or power generation plant, in order to reduce CO2 emission.


INTROduCTION
Membrane technology is one of the gas separation methods that can effectively reduce the carbon dioxide (CO 2 ) emission. 1,2 It has advantages such as being high energy efficiency, cheap, environmentally-friendly and simplicity. 3,4 Since CO 2 are mostly emitted from power generation plant, the CO 2 /nitrogen (N 2 ) separation should be more effective to reduce the CO 2 emission rate. 5,6 Based on Feron et al., the cellulose acetate (CA) holds the greatest promise in membrane material selection because it has the highest CO 2 /N 2 selectivity compared with other polymeric materials. 7 Moreover, CA has many advantages such as being low cost and environmentally friendly, which causes CA to be more preferable compared to the other polymeric materials. 8,9 There are two types of CA polymers used in this study, which are CA with acetyl content of 56% (CA-56) and CA with acetyl content of 39% (CA-39).
To improve the gas separation performance, the most important parameter is the polymer concentration. Aroon et al. reported that the polymer concentration increased with the gas selectivity but formed a thicker skin layer and thus, reduced the gas permeability. 10 According to Hacarlioglu et al., a higher polymer concentration was preferable; however, it cannot exceed certain limits due to the high permeability and low selectivity of polymeric membranes. 11 Many works summarised that the increment in polymer concentration decreased the membrane permeability and increased the gas selectivity. [10][11][12] Therefore, the main aim of this research is to develop a thin, defect-free blend CA-39/CA-56 membrane, which can enhance the permeance of CO 2 and increase the selectivity of CO 2 /N 2 . Up to date, the study on polymer concentration for blend CA membranes with various acetyl concentrations (CA-39 and CA-56) has not been investigated.

CA Membrane Fabrication
A mixture of 10 wt% of CA, 63 wt% of acetic acid and 27 wt% of deionised water was stirred and heated to 55°C by a heating plate for 3 h until the CA polymer completely dissolved in the solution. Thereafter, the solution was cooled until it reached ambient temperature of 27°C. After that, the cooled solution at ambient temperature was stirred for another 12 h and then sent for ultrasonic degassing (ELMASONIC S60H, Germany) to prevent the formation of bubbles. The casting thickness of the sample membrane was adjusted by using the Automatic Film Applicator (Malaysia) at ambient room temperature. Then, the sample membrane was submerged into a deionised water bath for 5 min. Next, it was moved to another water bath and immersed for one day in order to eliminate any excess solvent remaining in the sample membrane. To dry the membrane, the sample membrane was submerged into fresh ethanol for 4 h and then submerged into n-hexane solution for another 1 h. The final CA membrane was located between two glass plates and dried for one day. Finally, the final CA was stored and ready for further testing. 13 The percentage of polymer-solvent mixture was determined from previous works. 14,15 The casting thickness of the membrane was fixed at 250 μm and the composition for the membranes were illustrated in Table 1.

Gas Permeation Test
To examine the permeation of gas, a gas permeation test was conducted based on previous published work. 13 The volume displacement method was used to determine the flowrate of both output streams individually by using soap bubble flow meters. The permeability of membrane (P/l) was determined by Equation 1 and indicated in GPU. 15 where: l = membrane thickness (cm) A = effective membrane area (cm 2 ) Q = Measured volumetric flowrate in standard temperature and pressure (cm 3 s -1 ) ∆p = Pressure difference through the membrane (cmHg) where α CO 2 /N 2 is the ideal selectivity of CO 2 /N 2 .

SEM
The surface morphology of the synthesised membranes was investigated through a scanning electron microscopy (SEM) (Hitachi TM3000). The membrane samples were broken into small pieces and frozen at -80°C for one day to obtain an ordinarily steady and neat cut. The samples were then sputter-coated with thin gold film to avoid sample charging, which caused the contrasting structure. Next, the samples were placed inside the SEM to collect the cross-sectional structure of the membranes. A minimum of five membrane samples were used in this characterisation to examine the consistency of the samples. 17,18

ATR-FTIR Spectroscopy Analysis
The attenuated total reflectance-Fourier transform infrared (ATR-FTIR) was used to record the spectra, amend their baselines, systematise the spectra, and detect the peaks value through theNicolet IS10 (USA) spectrometer with wave numbers ranging from 450 cm -1 to 4000 cm -1 at 4 cm -1 resolution. 19 Spectra specimens were collected with 32 scans setting. The spectra wavenumbers of membrane specimens were recorded at ambient room temperature and repeated few times for every specimen. 14,20

Effects of Polymer Concentration of CA with different Acetyl Contents
In order to understand the interaction effects of the gas separation membranes, the membrane properties were determined and membrane characterisation was carried out for developing a new enhanced membrane-based gas separation. In this study, the CA membranes were blended at different polymer ratios of CA-39 and CA-56 Moreover, Figure 1 illustrated that the intensity of absorbance decreased from M1, M2 to M3 but dramatically increased in M4. When the content of CA-39 increased in the blend membranes, it caused reduction in the absorbance intensity for each main band and demonstrated that fewer functional groups were included in the hydrogen bonding of the blend membranes. 22 In this regard, the presence of polar functional groups influenced the gas permeance as well. Lee et al. stated that the dipole moments of the polar functional groups such as the hydroxyl (O-H) group and the carbonyl (C=O) group in the CA membranes had strong interaction between the CO 2 molecules and the polar functional groups, which affected the absorbance and enhanced the CO 2 permeance. 23 Gassensmith et al. also found that the polar functional groups in the CA membrane bonded with the CO 2 molecules by physisorption manner through dipole interactions. 24 Furthermore, a lower intensity of absorbance indicated that less hydroxyl groups were chemically bonded with acetyl groups and formed polymeric membrane structures with less compactness. 25 However, the increment in the intensity of the absorbance of M4 might be due to the increment in the content of CA-39, which formed strong hydrogen bonding between the water and polymer during membrane formation. This caused disorderly packed polymer chain, which led to a higher intensity of absorbance. 26 In addition, the SEM was used to investigate the membrane morphology. Thus, the high-resolution cross-section and surface SEM micrographs of CA-39/ CA-56 blend membrane for M1 (1 wt% of CA-39), M2 (2 wt% of CA-39), M3 (4 wt% of CA-39) and M4 (5 wt% of CA-39) were presented in Figure 2.
The SEM surface images of each membrane, shown in Figures 2(a, c, e and g) show smooth and defect-free surfaces for CA-39/CA-56 blend membranes. The nonporous and dense structure of the CA membrane allowed the build-up of pressure, which is applicable for CO 2 /N 2 separation. The smooth and nonporous dense skin layers formed on the upper surface of the membranes were due to the perfect demixing of the CA blend membranes. 27 Besides, the hydrophilic properties of the CA membranes allowed the formation on the membrane surface with good regularity. 28 It was because of the hydrophilic chains, the hydrophilic CA polymer enriched the membrane surface during the phase separation process through the formation of hydrogen bonding with water molecules. 28,29 Based on Figure 1, the ATR-FTIR analysis showed that the CA membrane had high hydrophilicity due to numerous polar functional groups included in the membranes. Hence, the high hydrophilicity of the CA membranes caused better dispersion in dope solution and formed smooth membrane structures due to the strong hydrogen bonding between the polymer and solvent. 29 Based on Figures 2(b, d, f and h), M1 (6.16 ± 0.07 μm), M2 (5.71 ± 0.07 μm) and M3 (5.47 ± 0.06 μm) had smoother dense skin layer compared to M4 (3.97 ± 0.11 μm). The smooth dense skin layer was formed due to the perfect demixing and precipitation mechanism that tightened the polymer molecular chain. 30 Based on Figure 1, the ATR-FTIR analysis indicated that the strong hydrogen bonding formed between the acetyl group from the CA polymer and carbonyl group from the acetic acid caused stronger interaction between polymer-solvent mixture which led to smoother dense skin layer formation. 13 As observed in Figure 2(h), the M4 demonstrated nonuniformity dense skin layer. It may be due to the increment in CA-39, which was highly hydrophilic than CA-56. The membranes that had high hydrophilic properties with more hydroxyl groups caused irregularities on the dense skin layer due to imperfect demixing mechanism. 26,28,30 The higher hydrophilicity caused water to flow towards the membrane during membrane formation and led to high surface energy and high chances for nonuniform membrane. 31 This was due to the high hydrophilicity of CA-39 leading to strong hydrogen bonding between the water and polymer, thereby, causing water to spread over the membrane during membrane fabrication. The hydrophilic polymer moved towards the membrane surface during the membrane formation process due to its high affinity towards water molecules. 29 Thus, it caused disorderly packed molecules and formed uneven dense skin layer. 26 In polymeric membranes, the formation of dense skin layer is necessary for CO 2 /N 2 separation. The dense skin layer formed by the molecular orientation concentrated on the upper surface of the membrane. 32 As displayed in the SEM images, shown in Figures 2(b, d, f and h), the dense skin layer was observed with reducing thickness from 6.16 ± 0.07 μm (M1) to 3.97 ± 0.11 μm (M4) when the polymer concentration of CA-39 increased from 1 wt% (M1) to 5 wt% (M4).
Pinnau and Freeman reported that the higher polymer concentration of polymer with low acetyl content (CA-39) led to thinner dense skin layer formation. 33 When comparing CA-39 with CA-56, the latter contained higher acetyl group content while CA-39 consisted more hydroxyl group which means that the membranes with more CA-39 had higher hydrophilicity. 34 The membrane with higher hydrophilicity, as indicated by the ATR-FTIR analysis, enhanced the polymer and solvent exchange rate when immersing the membrane into distilled water bath. This led to extreme rapid demixing, thus, producing a less dense membrane. 17 Moreover, the increment in polymer concentration of CA-39 increased the hydrophilicity of the polymeric membrane that led to less advanced gelation in polymer/nonsolvent/solvent phase and formed thinner dense skin layer. 17,19 Therefore, the M4 with more CA-39 and less CA-56 resulted in thinner dense skin layer formation.
Based on Figure 2, the final thickness of M1, M2, M3 and M4 were 280.8 ± 1.4 μm, 308.1 ± 1.6 μm, 263.1 ± 0.8 μm and 262.7 ± 1.1 μm, respectively. The decreased membrane thickness led to increase in gas permeation. However, the dense skin layer was the controlling factor for this asymmetric membrane. The gas separation performance was determined through the gas permeation test. Figure 3 demonstrates the CO 2 permeance rate of CA-39/CA-56 blend membranes for M1 (1:9), M2 (2:8), M3 (4:6) and M4 (5:5) with polymer concentration of CA-39 at 1 wt%, 2 wt%, 4wt% and 5 wt%, respectively. It showed that the CO 2 permeance improved from 85.64 ± 2.46 GPU (M1) to 328.23 ± 3.09 GPU (M4), when the polymer concentration of CA-39 increased from 1 wt% (M1) to 5 wt% (M4). According to the FTIR analysis and SEM results, indicated in Figure 1 and Figure 2, thinner dense skin layer formation from 6.16 µm to 3.97 µm and lower intensity of absorbance has been observed. This led to higher CO 2 permeance when the polymer content of CA-39 increased. As CA-39 contained less acetyl (CH 3 CO) group and more hydroxyl (O-H) group compared to CA-56, more CO 2 was solvated when the polymer content of CA-39 increased, thereby, improving the CO 2 permeance as well. 35 The increment in the hydrophilic CA-39 improved the intramolecular hydrogen bonding between the hydroxyl groups but reduced the intermolecular hydrogen bonding and enhanced the polymer chain rigidity, thus, leading to higher CO 2 permeance. 36 In addition, the increment in the hydrophilic CA-39 formed a stronger interaction between the polar functional groups of the CA polymers and quadrupolar moment of CO 2 which caused improvement in the CO 2 permeance rate. 18 Additionally, the CO 2 permeance of M4 (328.23 ± 3.09 GPU) increased dramatically due to the irregular dense skin layer, as demonstrated in Figure 2(h). The nonuniform dense skin layer formed because of the loss of volatile solvent during the membrane fabrication process resulting in reduction in the volatile solvent. 37 The irregular skin layer allowed the CO 2 molecules to pass through the thinner parts of the membrane easily and resulted in high CO 2 permeance. 38 Figure 4 shows that the N 2 permeation rate increased from about 74.94 ± 1.36 GPU (M1) to 307.05 ± 0.62 GPU (M4) when the polymer content of CA-39 increased from 1 wt% (M1) to 5 wt% (M4). This phenomenon occurred due to the decrease in the dense skin layer from 6.16 ± 0.07 μm, shown in Figure 2(b) to 3.97 ± 0.11 μm, shown in Figure 2(f). Ismail and Yean proved that the membrane with high permeation rate was achieved with thinner dense skin layer. 39 This has been explained by Wijmans and Baker who stated that when the polymer concentration of CA-39 increased, the polymer molecule concentration decreased, thus, causing improvement in gas permeance due to the lesser dense chain packing. 40 Furthermore, the FTIR results showed that when the CA-39 increased, less hydroxyl groups were chemically bonded with the acetyl groups, hence, forming polymeric membrane structure with less compactness that resulted in enhanced N 2 permeance. 25 In addition, the sudden increment in N 2 permeation of M3 (87.12 ± 0.81 GPU) to M4 (307.05 ± 0.62 GPU) may be due to the irregular dense skin layer thickness of M4. The formation of irregular dense skin layer caused faulty polymer chain distribution and allowed the N 2 particles to permeate through the membrane easily. This may due to the insufficient dispersion of the polymer in dope solution and imperfect polymer chain packing. 29,39 Based on Figure 5, the CO 2 /N 2 selectivity for CA-39/CA-56 blend membrane decreased when the polymer concentration of CA-39 increased from 1 wt% (M1) to 5 wt% (M4). The CO 2 /N 2 selectivity of M1, M2 and M3 were 1.1428 ± 0.02, 1.1406 ± 0.034 and 1.139 ± 0.037, respectively. Madaeni et al. explained that the increment in gas permeance rate lead to lower selectivity for the polymeric membrane. 12 Furthermore, the dense skin layer thickness reduced from 6.16 ± 0.07 μm, shown in Figure 2(b), to 3.97 ± 0.11 μm, shown in Figure 2(f). When the dense skin layer thickness decreased, the layers formed in membrane formation were reduced in each membrane and led to lower resistance against the passing gas that caused lower membrane selectivity. 13 The results also illustrated that M4 had the lowest selectivity, which was 1.069 ± 0.008. This was due to the irregular dense skin layer formation that enabled both the CO 2 and N 2 gas particles to permeate through the membrane easily. Ahmad et al. reported that the selectivity of membrane strongly depended on the integrity of the dense skin layer. 18 Therefore, to obtain a high selectivity membrane, the dense skin layer formed must be uniform and dense.
Besides, the reduction of CO 2 /N 2 selectivity occurred due to lower molecular concentration when more CA-39 was used. It caused less dense polymer chain packing and low complexity on side group, with more hydroxyl groups and less acetyl groups, leading to poor polymer chain distribution and low selectivity. 42 Due to the increment in gap size between polymer chain, the selectivity for the membrane decreased because of the low resistance to gas permeance. 41 CO 2 /N 2 selectivity (-)  In summary, Figure 6 was exhibited to summarise the average CO 2 permeance, N 2 permeance and CO 2 /N 2 selectivity of every membrane. It proved that the CO 2 had higher permeation rate than N 2 for each membrane. This result was presented because the high condensable CO 2 had better interaction with the polymeric chain and functional groups of the CA membrane compared with N 2 , due to its quadrupolar moment. 18 Hence, this resulted in a higher CO 2 permeance than N 2 permeance. Based on the results, the percentage of decrement in CO 2 /N 2 selectivity among M1 (1 wt% of CA-39), M2 (2 wt% of CA-39) and M3 (4 wt% of CA-39) were negligible. However, the CO 2 /N 2 selectivity exhibited reduction between M3 (4 wt% of CA-39) and M4 (5 wt% of CA-39) from 1.139 ± 0.037 to 1.069 ± 0.008, respectively. Therefore, although M4 had extremely high CO 2 permeance and N 2 permeance, it was still eliminated. Thus, M3 was selected as the best membrane due to its relatively high CO 2 permeance, N 2 permeance and CO 2 /N 2 selectivity (99.26 ± 3.08 GPU, 87.12 ± 0.81 GPU and 1.139 ± 0.037, respectively).  Table 2 demonstrated that higher CO 2 permeance and N 2 permeance were found for the optimal blend CA membrane (M3) in the present work, while comparing with Farrukh et al. 17 However, the CO 2 /N 2 selectivity was lower than the available membrane. This was due to the extremely strong interaction formed between the hydroxyl groups and the acetyl groups which caused high compactness membrane that led to low gas permeance and high membrane selectivity. 22,25 Furthermore, the presence of CA-56 with more acetyl groups in the CA polymer matrix caused movement in the polymeric chains, which improved the gas permeability. 1,20,35 This study was at preliminary level. Therefore, more researches are required to understand the blending chemistry of CA polymers in order to improve the gas separation performance effectively.

CONCLuSION
In this work, new membranes were successfully fabricated by blending the CA-39 and CA-56. The gas separation performance for the CA-39/CA-56 blend membranes were proven to be affected by the polymer concentration. The optimal CA-39/CA-56 blend membrane, M3, with the polymer ratio (CA-39:CA-56) at 4:6, was determined. The ATR-FTIR spectrum successfully indicated that the CA-39/CA-56 blend membrane consisted of polar functional groups that favoured the CO 2 permeance. The SEM results indicated that M3 formed a membrane that had relatively thin dense skin layer with smooth surface and uniform structure which demonstrated integrity of membrane structure and allowed the application of solution-diffusion mechanism. Besides, the M3 had the best CO 2 permeance, N 2 permeance and CO 2 /N 2 selectivity which were 99.26 ± 3.08, 87.12 ± 0.81 and 1.139 ± 0.037, respectively. Therefore, M3 demonstrated higher CO 2 and N 2 permeation rates but relatively lower selectivity for CO 2 /N 2 separation. In the future, it is expected that the mixed matrix membrane will improve the gas separation performance by combining the advantages of both polymeric and inorganic components while hollow fibre membrane performs high compactness and self-supported membrane structure.