- Email: [email protected]
Development and characterization of disk supported carbon membrane prepared by one-step coating-carbonization cycle
Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Contents lists available at ScienceDirect
Journal of Industrial and Engineering Chemistry journal homepage: www.elsevier.com/locate/jiec
Development and characterization of disk supported carbon membrane prepared by one-step coating-carbonization cycle N.H. Ismaila,b , W.N.W. Salleha,b,* , N. Sazalia,b , A.F. Ismaila,b,* a b
Advanced Membrane Technology Research Centre (AMTEC), Universiti Teknologi Malaysia, 81310 Skudai, Johor Darul Takzim, Malaysia Faculty of Chemical and Energy Engineering (FCEE), Universiti Teknologi Malaysia, 81310 Skudai, Johor Darul Takzim, Malaysia
A R T I C L E I N F O
Article history: Received 2 June 2017 Received in revised form 15 August 2017 Accepted 19 August 2017 Available online 25 August 2017 Keywords: Supported carbon membrane Co-polyimide P84 Carbon dioxide Gas separation Spray coating
A B S T R A C T
Different dope solutions for the alumina disk supported carbon membrane preparation for CO2/N2 and CO2/CH4 separation was formulated in this study. The prepared polymeric membrane made of commercial co-polyimide BTDA-TDI/MDI (P84) was carbonized at 700 C under N2 gas flow. A defect-free membrane was obtained when high polymer composition was used. The disk supported carbon membrane with CO2/N2 and CO2/CH4 selectivity of 15 and 45, respectively, and CO2 permeance of 400 Barrer were obtained by one-step spray coating technique. The polymer composition of 12 wt% was concluded to be the optimum composition for the alumina disk supported carbon membranes. © 2017 The Korean Society of Industrial and Engineering Chemistry. Published by Elsevier B.V. All rights reserved.
Introduction The factor of impurity composition of acidic gas in natural gas as well as low quality gas up to 80% makes the natural gas purification crucial [1]. This situation had prompted researchers to study and enhance the performance of the available technologies towards gas separation application [2]. Gas separation membranes have been used in various applications, such as oxygen or nitrogen enrichment, hydrogen recovery, acid gas treatment, and natural gas dehydration [3–5]. Polymeric membranes have emerged as a potentially superior membrane for gas separation. However, it needs to be improved as it has limitations in its performance, such as poor thermal and chemical resistance [6,7]. Polymeric membrane that employs solution diffusion as its separation mechanism is not an effective mechanism to apply for gas separation. Carbon membrane is one of the promising membrane materials that proved to give high gas separation performance and can overcome the disadvantages of polymeric membranes [8]. There are four different types of separation mechanism such as Knudsen diffusion (>10 Å), surface diffusion (<50 Å), capillary condensation (>30 Å), and molecular sieving (<6 Å) [6]. Carbon membrane is fabricated by carbonizing polymeric membrane at high temperature (600–1000 C) where most of the heteroatoms
* Corresponding authors at: Faculty of Chemical and Energy Engineering (FCEE), Universiti Teknologi Malaysia, 81310 Skudai, Johor Darul Takzim, Malaysia. E-mail addresses: [email protected] (W.N.W. Salleh), [email protected] (A.F. Ismail).
that present in the polymeric membrane will be released and replaced by carbon in the backbone of its molecular structure [9]. During the carbonization process, there is a possibility where the polymer precursor will melt due to the high temperature applied. To prevent this problem, thermosetting polymer is the best candidate to overcome this matter. The chosen polymer precursor must meet a number of criteria, such as high aromatic carbon content, high glass transition temperature (Tg), excellent thermal resistance, mechanical stability, and chemical stability, and provides high separation properties [10]. Polyimide has high possibility to fulfil those requirements and it has various classes which is determined by their different dianhydrides, which are pyromelliticdianhydride group (PMDA) such as Kapton [9], benzophenonetetracarboxylicdianhydride (BTDA) such as Matrimid and P84 [9–11], 3,30,4,40-biphenyltetracarboxylic dianhydride (BPDA) such as UIPR and UIP-S [11], and hexafluoroisopropylidene (6FDA) such as pyralin [12]. Among these classes, P84 that has superior gas separation performance [13–16] was chosen as the polymer precursor for this study. P84 is generally made by the polycondensation of aromatic acid dianhydrides and diamines. It was reported that its performance was influenced by the chemical structure of the constituent monomers as shown in Fig. 1. P84 co-polyimide (BTDA-TDI/MDI, co-polyimide of 3,30 ,4,40 ,-benzophenone tetracarboxylic dianhydride and 80% methyl phenylene- diamine + 20% methylene diamine) exhibits an excellent chemical and thermal resistance. This polymer is not only suitable for gas separation but also excellent in ultrafiltration and nanofiltration [10,17].
http://dx.doi.org/10.1016/j.jiec.2017.08.038 1226-086X/© 2017 The Korean Society of Industrial and Engineering Chemistry. Published by Elsevier B.V. All rights reserved.
314
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Fig. 1. Chemical structure of P84 (BTDA-TDI/MDI).
In this study, supported carbon membrane was prepared by introducing porous disk alumina as a support material to overcome the brittleness of the carbon membrane. Obtaining a defect-free thin layer film of carbon membrane is crucial as several parameters need to be considered, such as composition of the polymer precursor, coating method, and parameters during the heat treatment process, which includes temperature, heating rate, gas flow rate, environment gas, and soaking time [16]. Spin, spray, slip, and dip coating are the common methods applied to fabricate the supported carbon membrane. Past studies would suggested to implement several time of coating-carbonization cycle in order to obtain a defect-free carbon membrane layer that reflect in the gas separation performance [18,19]. However, these repeating methods are time and cost consuming. The one-step spray coating technique was proposed in this study by investigating the effect of polymer composition on gas permeation performance. The polymer precursor was distributed uniformly on the porous alumina support material. Until now, there are no studies reported on supported carbon membrane derived from P84 via spray coating technique. In 1999, Archarya and Foley applied spray coating method to form a fine mist of PFA on the porous support using airbrush that took less than 15 s to complete for one sample [20]. Most of the previous literatures were reported on dip and spin coating techniques, and recently, slip coating technique with several time of coating-carbonization cycles was also reported to reduce defect on the supported carbon membrane [9]. Air spray method can produce ultra-thin selective layer which can provide high selectivity without deteriorating the gas permeability [21]. The distribution of the polymer solution on the support is significantly affected by the polymer solution composition. Generally, dilute solution (2–5 wt% of polymer) is used when applying spray coating method to form uniform and thin layer. Furthermore, it is convinced that the spray coating method has a high potential for reproducibility of the thin and uniform carbon membrane layer [22]. Thus, the objective of this research was to investigate the effect of polymer compositions on the gas separation performance of the supported carbon membrane. The membrane was characterized by means of TGA, FTIR, BET, XRD, SEM images, and single gas permeation measurement. Experimental Materials Commercial co-polyimide BTDA-TDI/MDI (P84) powder purchased from Sigma–Aldrich (CAS#: 58698-66-1) was used as polymer precursor. N-Methyl-2-pyrrolidone (NMP) procured from Merck (Germany) was chosen as the solvent. Commercial symmetric porous alumina disk was utilized as supporting material with diameter of 47.0 mm, thickness of 1.0 0.05 mm, and mean pore size of 0.14 mm was bought from Shanghai Gongtao Ceramics Co., Ltd.
Carbon membrane preparation P84 polyimide powder was dried in an oven at 60 C for one day to remove water vapour. The polymeric solution was prepared by dissolving four different compositions of P84 at 6, 9, 12, and 15 wt% in NMP. The polymeric solution was stirred and heated at 70 C until homogenous solution was formed. It was sonicated for 3–4 h to eliminate bubble formed during stirring process. The supporting material was polished using fine silicon carbide (SiC) paper of grit P2000 before dry it in an oven for 3 h prior coating process. Porous alumina was coated by homogeneous solution via spray coating method at 1 bar at room temperature. The dope solution was sprayed directly towards the supporting material using air spray with distance of the spray nozzle and the alumina support at 20 cm. The coated alumina support was dried in an oven at 60 C overnight. The coated porous alumina disk with polymer precursor membrane was placed in the centre of the Carbolite horizontal tubular furnace for heat treatment process. It was carbonized at 700 C with heating rate of 3 C/min under nitrogen, then it was left to cool down to room temperature. This process took almost 4 h to achieve the final carbonization temperature and another 4 h to reach room temperature after the carbonization process. The nomenclature of the resultant disk supported carbon membranes was given in the form of P-polymer composition for polymeric membrane (P-6, P-9, P-12, P-15) and CM-polymer composition for carbon membrane (CM-6, CM-9, CM-12, CM-15). Membrane characterization Thermal behaviour of the membrane was obtained from thermogravimetric analysis (TGA 2050), carried out at temperatures between 50 C and 1000 C at a rate of 10 C/min under nitrogen gas at 50 ml/min. Elemental analysis was carried out using elemental analyzer model Vario Micro Cube indicate the percentage of the present elements in the membrane sample. While the functional groups that influence the gas performance were evaluated via Fourier transform infrared spectroscopy (FTIR), using single reflection diamond for Spectrum Two spectrometer (PerkinElmer, L1600107). The X’Pert PRO X-ray different diffractometer (XRD) from PANalytical with the 2u diffraction angle of 10–90 was performed using Cu Ka radiation of 1.54 Å wavelength. Bragg’s Law (nl = 2d sin u ) was adapted to determine inter planar distance (d-spacing) between the individual layers of the carbon. Brunauer–Emmett–Teller (BET) were obtained by utilizing equipment model Micromeritics 3 Flex Surface Characterization Analyzer. This nitrogen (N2) adsorption method was applied to obtain surface area and pore volume of porous materials from polymeric and carbon membrane. Surface and cross sectional morphologies of the carbon membrane was observed by scanning electron microscopy (SEM) model EOL JSM-5610LV. The prepared samples were coated with gold by sputter coating under vacuum to create neutral charge during SEM characterization. The viscosity of the polymeric solution was measured using a viscometer (Brookfield).
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Gas permeation study The gas permeation properties of the prepared membrane was evaluated using pure gas permeation system, tested at room temperature and feed pressure of 4 bar using bubble soap flow meter. The values of permeance and selectivity are determined using Eqs. (1) and (2). Permeance, P Pi Qi Q ¼ i2 DP ¼ l A DP pr
1GPU ¼ 1 106
ð1Þ
cm3 ðSTPÞ cm2 cmHg
where (Pi/l) is the gas permeance of a membrane (GPU), i is the gas species penetrating the membrane, Qi is the volumetric flow rate of gas i at standard temperature and pressure (sm3/s, STP), A is the membrane surface area (cm2), DP is the pressure difference between the feed side and permeation side of the membrane (cmHg), and r is the radius of the alumina disk (cm). Selectivity, a P =l ai=j ¼ i Pj =l
ð2Þ
where i/j i = is the selectivity of species gas i to species gas j, Pi/l and Pj/l are the permeance of gas i and j, respectively. Results and discussion Thermal behaviour analysis TGA with N2 atmosphere at flow rate of 50 ml/min and heating rate of 10 C/min was applied to assess the weight loss for the precursors of 6, 9, 12, and 15 wt% without support material. Fig. 2 illustrates the weight variation of polymer during heating process in the temperature range of 50–800 C. There are four stages of phenomena that occurred during the degradation of P84. Stage 1 that refers to the temperature of below 200 C is attributed to the existence of residue solvent and the release of adsorbed water from the precursor [10,16,17,23]. The second thermal degradation stage is due to the thermal rearrangement of original structure of P84, while the third stage represents the initial stages of polymer decomposition towards the formation of carbon membranes [24]. At 420 C, P-6, P-9, and P-15 experiences a sudden drop
315
approximately 5% from their original weight, while a sudden drop occurs for P-12 at 7% from its original weight. Higher Td with low weight loss is a good candidate to employ in carbon membrane fabrication as it can reduce melting possibility when exposed at high temperature (heat treatment process), which in this study refers to 12 wt% of P84. It is stated that the glass temperature (Tg) of the P84 polyimide is 315 C and as pointed out from the TGA result, Td for P84 is 420 C. Stabilization temperature can be determined from the Tg and Td results. For this study, 350 C is the stabilization temperature as it lies between both temperatures. When implementing high temperatures towards the long chain backbone components (polymer), it lead the chain to separate and react with one another to change the properties of the polymer and contributes in obtaining high gas separation performance [25]. Moreover, chemical and mechanical reactions involve during thermal degradation lead to physical and optical changes relative to the initially specified properties, including the reduction of ductility, colour changes (from light yellow to black), and physical properties. At final stage, the weight loss from 670 to 800 C is small at 5%. In this study, at the carbonization temperature 700 C, the 15 wt% polymer composition resulted in the highest weight loss of 46.3%, followed by 9 wt% (46.0%), 6 wt% (45.0%), and 12 wt% composition (43.0%). Elemental analysis The elemental analysis (percentage of atoms) of P84 (powder form), polymeric and carbon membrane at 12 wt% polymer precursors is presented in Table 1. The result of P84 and its polymeric membrane shows a different value, where P-12 has higher value in C element but lower value of H, N and O element as compared to P84. This phenomenon was due to the present of solvent element in the polymeric membrane, where the solvent was not fully evaporated during drying process and remain in the polymeric membrane. As the polymeric membrane contains the same functional group, the amount of C element for all the polymeric membrane are assumed to be similar. Theoretically, the C content in polymeric membrane should be lower in comparison to its carbon membrane as it has been occupied by greater value of N and O elements. Most of the heteroatoms (N and O elements) are released during the heat treatment process, resulting in small amount of them in carbon membrane matrix [9]. These phenomena affect the end product structure of the fabricated membrane that enhance the gas separation performance [26]. It is predicted that, when applying higher temperature, the greater amount of C element will be obtained as it has enough time to clearly eliminate heteroatoms. Based on Table 1, after the polymeric membrane was carbonized, the element of C content increased by 16% from its original level. Meanwhile, H, N, and O decreased by 34%, 18%, and 26%, respectively. Fourier transform infrared spectroscopy (FTIR) The FTIR spectrum of the polymeric and carbon membrane prepared at different polymer compositions are shown in Fig. 3. Based on the results, similar peaks of functional groups are observed for all the polymeric membranes. The key absorption Table 1 Elemental analysis of P84, polymeric and carbon membrane at 12 wt. % of P84. Membrane
P84 P-12 CM-12 Fig. 2. TGA analysis of P84-based polymeric membrane.
Element content (%) C
H
N
O
58.2 63.7 73.6
6.0 3.8 2.5
8.1 7.6 5.6
27.2 24.5 18.2
316
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Fig. 3. FTIR spectra of polymeric membrane and carbon membrane prepared at different polymer compositions of P84.
Fig. 4. XRD pattern of polymeric and carbon membrane prepared at difference polymer compositions of P84.
peaks for polymeric membrane is observed at 725 cm1 is assigned to C H (alkenes), 1100 cm1 and 1360 cm1 exhibited to OC N CO (imide II), 1510 cm1 show O¼C N (amide II) and 1720 cm1 and 1780 cm1 represent asymmetric OC NCO (imide I) [27,28]. A weak vibration at 3615 cm1 was corresponding to OH (carboxylic acid groups) is observed where the carbonyl stretching of the acid group is overlapped by the absorption of the imidic carbonyl [29,30]. However, after the carbonization process at 700 C under N2 gas flow, most of the absorption peaks disappear for all the carbon membranes. The peak for carbon membrane for P84 appears at 1140 cm1 (C N, amines). Due to the uncertain peaks of alkynes and carboxylic acid, both of the groups are ignored for present work because most of the original heteroatoms (O and N) are released during the carbonization process and results in the formation of cross-linked and stiff carbon matrix [6,9]. The peaks appeared in the range of 2000– 2500 cm1 mainly covers weak bond stretching modes, due to the existence of CO and CO2 in the beam. The CO and CO2 obviously seen only in carbon membrane due to the breaking of the original bond in its polymer precursor. Carbonization temperature induced the breakage of the presence bond in the polymeric membrane, resulted to new bonding as the previous bond was rearranged [31]. It is observed that FTIR was unable to characterize a few volatile compounds such as O2, N2, and H2, however the presence of volatile compounds (CO2, CO, O2, N2, benzene, aliphatic hydrocarbons, and other aromatic compounds) was clearly observed during thermal degradation [10].
(100) plane in graphite. Based on the XRD spectra, by converting polymer membrane to carbon membrane, it is revealed that the peak is shifted to the right that results in a decrease in the dspacing value [28]. Although various loading of polymer precursor in composition solvent were applied, the angles (2u) of the membranes were almost similar, which concluded that different amount of polymer precursor did not have significant affect towards the XRD characterization. Increasing P84 composition in the carbon membrane will obtain clear peak in the XRD due to the existence of the P84 that can be easily detected. However, a further study is required to ensure estimation more accurate result. Gas separation requires low d-spacing that result in low distance of the atomic planes and subsequently obtain a superior gas separation selectivity by separation mechanism of molecular sieving (<5 Å) [22].
Wide-angle X-ray diffraction patterns The XRD characterization was used to analyze the microstructure and determine the interlayer distance of the carbon matrix of the polymeric and carbon membranes. Fig. 4 illustrates the XRD pattern for polymeric and carbon membranes fabricated at different polymer compositions. The polymeric membrane shows peak at 18 with d-spacing value of 5 Å. After polymer precursor has been carbonized, the spectra of the carbon membranes exhibit a peak at 2u = 23 with (002) plane in graphite [32]. The theta values are shifted when the polymeric membrane is carbonized due to the presence of the graphitic and turbostratic structure for the efficient diffusion path for gas molecules to pass through [24]. In addition, the carbon membrane is the result of an amorphous carbon matrix with a mixture of sp2 and sp carbon components, which contribute a narrow pore size distribution [22,26,27]. Another weak and broad peak of the carbon membrane is found at 2u = 44 with d-spacing of 2.1 Å and characteristic peak of the
Pore structure analysis The pore structure analysis was determined by characterizing the polymeric and carbon membranes without their support material. Nitrogen adsorption isotherms is adopted to gain BET surface area, the micropore volume, and total pore volume. BET characterization is conducted to a relative P/P0 of approximately 0.3 at 196 C (77 K), where P0 is the saturation pressure. Table 2 shows the carbon membrane obtains a higher BET surface area in comparison to its original form (polymeric membrane). The BET surface is referred to the value of the surface area involves during the gas adsorption. The differences in the membrane morphological structure of the polymeric and carbon membrane lead to a huge gap of the BET surface area, thus the total pore volumes are also different [26]. The carbonization process that went through by the carbon membrane managed to break the original bond of the polymeric precursor and led to a new arrangement of atoms as proved in the FTIR characterization. Fig. 5 interprets the N2 adsorption of polymeric and carbon membrane prepared from 12 wt% of the polymer precursor P84. The initial N2 adsorption for the polymeric membrane is flat and low in quantity however, at approximately 0.9 P/Po, there is a very narrow range of a sudden increase in N2 adsorption. This represents the adsorption isotherm type I which describes the P-12 of either a chemisorption isotherm (final upswing at high pressures may not be present) or physical sorption on a material that has extremely fine pores (micropores). While carbon membrane marked by CM-12 has an acute increase in its initial N2 adsorption and thereafter the isotherms become flat. Then a slight increase of the N2 isotherms was observed. The carbon membrane exhibited the adsorption isotherms type II that
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
317
Table 2 Effect of membrane pore structure for polymeric and carbon membrane.
P-12 CM-12
BET surface area (m2/g)
Total pore volume of pores (cm3/g)
Average pore diameter (nm)
20 364
0.094 0.130
19.0 1.4
has high energy in adsorption. Moreover, these characteristics expressed that carbon membrane is mainly occupied by microporous material that gives impact towards the gas separation performance [23]. The difference of N2 adsorption isotherms between the carbon membrane and polymeric membrane is the contradicting with the pore size. Polymeric membrane that went through the carbonization process tends to obtain smaller pores as the pores would be shrinkage [33]. Morphological structure analysis The morphological structure view of the P84 polymeric membrane and P84 carbon membrane were determined by scanning electron microscopy (SEM). Fig. 6 shows the images of alumina disk support and polymeric membrane of P84 at different polymer precursor compositions. The bare alumina support structure is like a bricks arrangement that resulted in uneven surfaces. These uneven surfaces improve the adhesion mechanism between the membrane layer and disk supporting material. The pressure applied during the spray coating process contributes to a better adhesion strength between membrane and supporting material surface. Adhesion mechanism is crucial especially when the membrane undergoes carbonization process as the polymer tend to shrink and cause peeling problem. Non-uniform film layer of P-6 is observed, where most of the porous part of the alumina support is not fully covered because most of the solvent penetrate through the porous support. Increasing the polymer amount from 9 to 12 wt% makes the dope solution to be more uniformly distributed and most of the porous surface is covered [35]. High percentage of solvent has low viscosity (low fraction of solids) and low amount of solvent in dope solution has high fraction of solids that affect the distribution of solution on the porous alumina support during the spray coating method. This non-uniform thin film result in high permeance but low in selectivity. In addition, the thickness of the membrane increased when the weight percentage of the polymer increased. The distribution and thickness of the membrane layer is also depending on the viscosity of the polymeric solution. Table 3
presents the viscosity values for the polymeric solution at various P84 compositions. It clearly shows that the significant differences in the viscosity values affect the polymer distribution, membrane thickness, and gas performance. High viscosity is obtained when a high amount of polymer composition is used in the dope solution. The mass flow of the polymeric solution was influenced by the solution composition where low viscosity solution have a wide spread dispersion area. Theoretically, the low viscosity solution is faster to move and obtain a wide dispersion area, while increasing the viscosity value results in a slower mass flow and small dispersion area that directs the polymer solution to agglomerate [32]. In this study, the polymer composition of 12 wt% gives the best polymer distribution on the supporting material and has moderate thickness as compared to other samples. Fig. 7 shows the SEM images surface and cross section of carbon membranes which were originally from polymeric membrane P84 carbonized at 700 C in N2 gas flow of 200 ml/min with a heating rate of 3 C/min. As mentioned previously, heteroatoms that exist in polymeric membrane are replaced by carbon in the backbone of molecular structure during the carbonization process. The replacement of heteroatoms to carbon, forms the amorphous structure that possess large BET surface areas, controllable, and narrow pore size distribution [35– 37]. Based on carbon membrane surface, the distribution of thin film carbon membrane is better (uniform) compared to its polymeric membrane. After polymeric membrane has been carbonized, the carbon membrane become a denser microstructure compared to its original form, resulting in better morphological structure [25]. However, the defect formation of the supporting material was clearly observed from the carbon membrane surface. CM-6 has uneven surface image where the distribution of the membrane layer was not uniform and effected the thickness of the carbon membrane layer. As the P84 composition increased up to 9 wt%, CM-9 demonstrated smoother surface layer compared to CM-6. CM-12 and CM-15 have showed uniform carbon membrane layer because of their high viscosity which can prevent the polymeric solution from deep penetration inside the support material. It was reported that different composition of solid and solvent membrane created different morphologies in terms of their skin and sub-porous [38]. The thickness of the carbon membrane was lower compare to their polymeric membrane as it shrink after went through the carbonization process. Gas permeation measurements
Fig. 5. N2 adsorption isotherms for polymeric and carbon membrane.
In this study, both polymeric and carbon membrane were tested using pure gas permeation system. Among the polymeric membrane, P-12 obtained highest selectivity for CO2/N2 at 2.6 and CO2/CH4 at 7.7. All the prepared carbon membranes exhibited higher result compared to their original polymeric precursor membranes. After thermal decomposition of the polymeric precursor, carbon membrane obtains a highly aromatic structure comprising disordered sp2 hybridized hexagonal sheet, producing mostly char with turbostatic structure where pores are formed from the packing imperfections [17]. Carbon membrane mechanical properties was improved by the existence of these sp2 bonds which provide better hardness besides its superior gas separation performance. Moreover, the
318
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Fig. 6. SEM images of bare alumina supporting material surface and P84 polymeric membrane surface and cross section.
high gas separation performance of the carbon membrane is due to the micropores formation within the membrane structure that is influenced by thermal decomposition. Carbon membrane CM12 obtained highest selectivity values for CO2/N2 and CO2/CH4 at 15 and 45, respectively. The membrane separation mechanism depends on the pore size of the final product of the membrane. The separation mechanism of the polymeric and carbon
membrane are different due to the presence of this amorphous structure [21,39]. The permeance and selectivity of both polymeric and carbon membranes decreased and increased, respectively, when the polymer composition was increased from 6 to 12 wt%. It is because of the different concentrations of polymer precursor P84 in the polymer solution. Low composition polymer (conquered by
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321 Table 3 Viscosity of polymeric P84 compositions.
solution
at
different
Membrane
Viscosity (cP)
P-6 P-9 P-12 P-15
17 20 138 214
solvent) will penetrate through the porous alumina support and results in high mass transfer resistance that affect the gas separation performance [40]. Based on the SEM images obtained, low composition of P84 covers less porous alumina supporting material surface where the gap within the membrane was clearly seen. This drives to high permeability but low selectivity. On the other hand, when increasing the composition of P84 to 15 wt%, the selectivity drops, which is influenced by the distribution of the polymeric membrane on the supporting material surface. This high
319
polymer composition is affected by its high viscosity as shown by the viscosity test. Low viscosity solution lead to a very wide distribution of the fluid when it came out from the orifice. While for high viscosity solution, the spray coating pressure will not able to overwhelm the viscosity forces that affects the mass flow of the fluid [41]. Previous researchers preferred to use low composition of polymer solution when utilizing air spray coating technique [42,43]. CM-12 and CM-15 has achieved uniform surface layer but differ in membrane thickness due to the different of viscosity of the polymer composition. The lowest gas separation performance was obtained for CM-15 due to the higher mass transfer resistance during the gas transport compare to CM-12. This phenomenon was influenced by the membrane thickness where CM-15 has thicker membrane layer than CM-12. In this study, 12 wt% was preferred as it was easier to handle and exhibited a good gas separation selectivity performance. Fig. 8 shows the gas separation performances of CO2/N2 and CO2/CH4 for polymeric and carbon membrane with Robeson’s
Fig. 7. SEM images of surface and cross section of carbon membrane at different P84 composition.
320
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
The results showed that the CO2/N2 for both polymeric and carbon membrane was not able to exceed the Robeson upper bound line. For CO2/CH4 separation, CM-12 lied above the Robeson upper bound line that is the commercial region of interest [29]. The selectivity of selected supported carbon membranes was summarized and compared with the present work in Table 4. The selectivity performance of the present membranes are comparable as in the reported literatures. Most of the previous researchers preferred to fabricate their supported carbon membrane using either spin, dip, or slip coating technique, instead of using spray coating method [18,19,43,45–49]. Coating technique normally requires several number of coating-carbonization cycles to overcome the defect formation on the carbon membrane surface with extra time and care [16]. However, there are a few researchers claimed that a single coating-carbonization as the best method as it can reduce time and cost along with the fabrication of disk supported carbon membrane [50–52]. This present work applied one-step spray coating technique by employing spray coating method in one cycle of coatingcarbonization which took less time and reduced cost. The defect formation on the carbon membrane layer can be reduce by applying intermediate layer. In future, it was recommended to utilize intermediate layer on the porous support material as it could reduce the defect formation and increase the gas separation performance of the supported carbon membrane [21]. Conclusions
Fig. 8. Gas separation performances of (a) CO2/N2 and (b) CO2/CH4 for polymeric and carbon membrane with Robeson’s upper bound 2008.
upper bound 2008 [44]. The gas selectivity of CO2/N2 and CO2/CH4 increase for both polymeric and carbon membrane when the P84 composition increased from 6 wt% to 12 wt%. However, at 15 wt%, the selectivity was dropped due to high viscosity of the polymer solution which leads to thicker membrane layer. The membrane thickness played an important role during the gas transport process as it effects the gas separation performance. Based on Fig. 8, 12 wt% of polymeric and carbon membrane demonstrated the lowest gas permeance and the highest gas selectivity. Carbon membrane attained six times higher result compared to the polymeric membrane for CO2/N2 and CO2/CH4.
Disk supported carbon membrane was successfully fabricated by adapting commercialized alumina disk as a supporting material with only one-step spray coating technique. Various polymer solution compositions were sprayed on alumina disk before the carbonization process at 700 C under N2 gas environment with the heating rate of 3 C/min. The final product of the carbon membrane is determined by the composition of the polymer precursor as it plays an important role on the membrane properties. The TGA and FTIR spectra show almost similar results due to the similar compounds exist in the prepared membranes. The gas separation performance data increased when the polymer compositions was increased from 6 wt% to 12 wt% due to the uniform distribution of the polymer solution on the porous support surface. Based on the result, it is revealed that the promising gas permeation properties was obtained for the carbon membrane prepared with 12 wt% polymer composition. The CO2/N2 and CO2/CH4 selectivity of 15 and 45 was achieved with 82% and 84% increment from their polymeric membrane, respectively.
Table 4 Comparison of gas permeability and selectivity of supported carbon membranes with other works in the literature. Polymer
Configuration
Coating technique
Coatingcarbonization cycle
Carbonization temperature
CO2 permeance
PEI – Matrimid PFA Matrimid PPO Phenolic resin PEI P84
Alumina disk Tubular alumina Tubular alumina Alumina disk Tubular alumina Alumina disk Stainless steel tubular Alumina disk Alumina disk
Spin – Dip Spin Dip Spin Slip Spin Spray
– – 3 2 3 – 3 1 1
600 700 850 700 850 600 700 600 700
1046.00a – 131.75b 772.10b 287.36b 147.50a 32.00b 426.10a 400.00a
a b
Barrer = 1010 cm3 (STP) cm/(cm2 s cm-Hg). GPU = 106 cm3(STP)/(cm2 s cm-Hg).
Selectivity
Reference
CO2/N2
CO2/CH4
– 1.80 – 14.30 – 20.04 88.00 – 15.00
27.6 – 75.82 – 87.34 19.30 347.40 56.40 45.00
[46] [45] [18] [19] [43] [47] [35] [48] Present work
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Acknowledgments The authors gratefully acknowledge the financial support from the Ministry of Higher Education and Universiti Teknologi Malaysia under Higher Institution Centre of Excellence Scheme (Project Number: R.J090301.7846.4J188), Fundamental Research Grant Scheme (Project Number: R.J130000.7846.4F846) and Research University Grant Scheme (Project Number: Q. J130000.2546.12H76). The authors would also like to acknowledge technical and management support from Research Management Centre (RMC), Universiti Teknologi Malaysia. References [1] S. Saedi, S.S. Madaeni, K. Hassanzadeh, A.A. Shamsabadi, S. Laki, J. Ind. Eng. Chem. 20 (2014) 1916. [2] Y.-C. Lee, S.M. Lee, W.G. Hong, Y.S. Huh, S.Y. Park, S.C. Lee, J. Lee, J.B. Lee, H.U. Lee, H.J. Kim, J. Ind. Eng. Chem. 23 (2015) 16. [3] E. Kim, E. Shin, G. Ko, S.H. Kim, O.H. Han, S.K. Kwak, Y. Seo, Chem. Eng. J. 306 (2016) 298. [4] R. Swaidan, X. Ma, E. Litwiller, I. Pinnau, J. Membr. Sci. 447 (2013) 387. [5] Q. Ma, H. Liu, H. Wang, L. Sun, S. Ma, Polym. Test. 54 (2016) 139. [6] W.N.W. Salleh, A.F. Ismail, T. Matsuura, M.S. Abdullah, Sep. Purif. Rev. 40 (2011) 261. [7] N. Sazali, W.N.W. Salleh, N.A.H.M. Nordin, A.F. Ismail, J. Ind. Eng. Chem. 32 (2015) 167. [8] L. Xu, M. Rungta, J. Hessler, W. Qiu, M. Brayden, M. Martinez, G. Barbay, W.J. Koros, Carbon 80 (2014) 155. [9] S.M. Saufi, A.F. Ismail, Carbon 42 (2004) 241. [10] P.S. Tin, T.-S. Chung, Y. Liu, R. Wang, Carbon 42 (2004) 3123. [11] S.S. Hosseini, M.M. Teoh, T.S. Chung, Polymers 49 (2008) 1594. [12] S.S. Hosseini, T.S. Chung, J. Membr. Sci. 328 (2009) 174. [13] S.S. Hosseini, M.R. Omidkhah, A. Zarringhalam Moghaddam, V. Pirouzfar, W.B. Krantz, N.R. Tan, Sep. Purif. Technol. 122 (2014) 278. [14] J. Ren, Z. Li, R. Wang, J. Membr. Sci. 309 (2008) 196. [15] T.A. Centeno, A.B. Fuertes, Sep. Purif. Technol. 25 (2001) 379. [16] S.C. Rodrigues, R. Whitley, A. Mendes, J. Membr. Sci. 459 (2014) 207. [17] M.J.C. Ordoñez, K.J. Balkus Jr., J.P. Ferraris, I.H. Musselman, J. Membr. Sci. 361 (2010) 28. [18] N. Sazali, W.N.W. Salleh, M. Nordin, N.A. Hadi, Z. Harun, A.F. Ismail, J. Appl. Polym. Sci. (2015) 132. [19] L.-H. Cheng, Y.-J. Fu, K.-S. Liao, J.-T. Chen, C.-C. Hu, W.-S. Hung, K.-R. Lee, J.-Y. Lai, J. Membr. Sci. 460 (2014) 1.
321
[20] M. Acharya, H.C. Foley, J. Membr. Sci. 161 (1999) 1. [21] N. Ismail, W. Salleh, N. Sazali, A. Ismail, Sep. Sci. Technol. (2017). http://dx.doi. org/10.1080/01496395.2017.1321671. [22] Y. Shen, A.C. Lua, J. Appl. Polym. Sci. 116 (2010) 2906. [23] J. Su, A.C. Lua, J. Membr. Sci. 305 (2007) 263. [24] X. Ma, R. Swaidan, B. Teng, H. Tan, O. Salinas, E. Litwiller, Y. Han, I. Pinnau, Carbon 62 (2013) 88. [25] B. Zhang, Y. Wu, Y. Lu, T. Wang, X. Jian, J. Qiu, J. Membr. Sci. 474 (2015) 114. [26] Y.-J. Fu, K.-S. Liao, C.-C. Hu, K.-R. Lee, J.-Y. Lai, Microporous Mesoporous Mater. 143 (2011) 78. [27] W. Qiu, C.-C. Chen, L. Xu, L. Cui, D.R. Paul, W.J. Koros, Macromolecules 44 (2011) 6046. [28] X. Qiao, T.S. Chung, AIChE J. 52 (2006) 3462. [29] E.M. Maya, A. Tena, J. de Abajo, J.G. de la Campa, A.E. Lozano, J. Membr. Sci. 349 (2010) 385. [30] E. Maya, A. Lozano, J. de Abajo, J. De la Campa, Polym. Degrad. Stab. 92 (2007) 2294. [31] J.B. Hamm, A.R. Muniz, L.D. Pollo, N.R. Marcilio, I.C. Tessaro, Carbon 119 (2017) 21. [32] P. Stähle, V. Gaukel, H.P. Schuchmann, Food Res. Int. 77 (Part 1) (2015) 55. [33] W.N.H.W. Zainal, S.H. Tan, M.A. Ahmad, Chem. Eng. Technol. 40 (2017) 94. [35] P.-S. Lee, D. Kim, S.-E. Nam, R.R. Bhave, Microporous Mesoporous Mater. 224 (2016) 332. [36] N. Bhuwania, Y. Labreche, C.S.K. Achoundong, J. Baltazar, S.K. Burgess, S. Karwa, L. Xu, C.L. Henderson, P.J. Williams, W.J. Koros, Carbon 76 (2014) 417. [37] P.J. Williams, W.J. Koros, Adv. Membr. Technol. Appl. (2008) 599. [38] S. Rafiq, Z. Man, A. Maulud, N. Muhammad, S. Maitra, J. Membr. Sci. 378 (2011) 444. [39] P.K. Chu, L. Li, Mater. Chem. Phys. 96 (2006) 253. [40] H.-H. Tseng, C.-T. Wang, G.-L. Zhuang, P. Uchytil, J. Reznickova, K. Setnickova, J. Membr. Sci. 510 (2016) 391. [41] G. Susanna, L. Salamandra, T.M. Brown, A. Di Carlo, F. Brunetti, A. Reale, Solar Energy Mater. Solar Cells 95 (2011) 1775. [42] T. Becherer, M. Vieira Nascimento, J. Sindram, P.-L.M. Noeske, Q. Wei, R. Haag, I. Grunwald, Prog. Org. Coat. 87 (2015) 146. [43] N.H. Ismail, W.N.W. Salleh, N. Sazali, A.F. Ismail, Chem. Eng. Trans. (2015) 1465. [44] L.M. Robeson, J. Membr. Sci. 320 (2008) 390. [45] N. Kruse, Y. Schießer, S. Kämnitz, H. Richter, I. Voigt, G. Braun, J.U. Repke, Sep. Purif. Technol. 164 (2016) 132. [46] H.-H. Tseng, K. Shih, P.-T. Shiu, M.-Y. Wey, J. Membr. Sci. 405–406 (2012) 250. [47] H.-H. Tseng, S.-H. Chang, M.-Y. Wey, J. Membr. Sci. 454 (2014) 51. [48] M.-Y. Wey, H.-H. Tseng, C.-k. Chiang, J. Membr. Sci. 453 (2014) 603. [49] W. Wei, G. Qin, H. Hu, L. You, G. Chen, J. Membr. Sci. 303 (2007) 80. [50] M. Teixeira, M. Campo, D.A. Tanaka, M.A. Tanco, C. Magen, A. Mendes, Chem. Eng. Res. Des. 90 (2012) 2338.
Contents lists available at ScienceDirect
Journal of Industrial and Engineering Chemistry journal homepage: www.elsevier.com/locate/jiec
Development and characterization of disk supported carbon membrane prepared by one-step coating-carbonization cycle N.H. Ismaila,b , W.N.W. Salleha,b,* , N. Sazalia,b , A.F. Ismaila,b,* a b
Advanced Membrane Technology Research Centre (AMTEC), Universiti Teknologi Malaysia, 81310 Skudai, Johor Darul Takzim, Malaysia Faculty of Chemical and Energy Engineering (FCEE), Universiti Teknologi Malaysia, 81310 Skudai, Johor Darul Takzim, Malaysia
A R T I C L E I N F O
Article history: Received 2 June 2017 Received in revised form 15 August 2017 Accepted 19 August 2017 Available online 25 August 2017 Keywords: Supported carbon membrane Co-polyimide P84 Carbon dioxide Gas separation Spray coating
A B S T R A C T
Different dope solutions for the alumina disk supported carbon membrane preparation for CO2/N2 and CO2/CH4 separation was formulated in this study. The prepared polymeric membrane made of commercial co-polyimide BTDA-TDI/MDI (P84) was carbonized at 700 C under N2 gas flow. A defect-free membrane was obtained when high polymer composition was used. The disk supported carbon membrane with CO2/N2 and CO2/CH4 selectivity of 15 and 45, respectively, and CO2 permeance of 400 Barrer were obtained by one-step spray coating technique. The polymer composition of 12 wt% was concluded to be the optimum composition for the alumina disk supported carbon membranes. © 2017 The Korean Society of Industrial and Engineering Chemistry. Published by Elsevier B.V. All rights reserved.
Introduction The factor of impurity composition of acidic gas in natural gas as well as low quality gas up to 80% makes the natural gas purification crucial [1]. This situation had prompted researchers to study and enhance the performance of the available technologies towards gas separation application [2]. Gas separation membranes have been used in various applications, such as oxygen or nitrogen enrichment, hydrogen recovery, acid gas treatment, and natural gas dehydration [3–5]. Polymeric membranes have emerged as a potentially superior membrane for gas separation. However, it needs to be improved as it has limitations in its performance, such as poor thermal and chemical resistance [6,7]. Polymeric membrane that employs solution diffusion as its separation mechanism is not an effective mechanism to apply for gas separation. Carbon membrane is one of the promising membrane materials that proved to give high gas separation performance and can overcome the disadvantages of polymeric membranes [8]. There are four different types of separation mechanism such as Knudsen diffusion (>10 Å), surface diffusion (<50 Å), capillary condensation (>30 Å), and molecular sieving (<6 Å) [6]. Carbon membrane is fabricated by carbonizing polymeric membrane at high temperature (600–1000 C) where most of the heteroatoms
* Corresponding authors at: Faculty of Chemical and Energy Engineering (FCEE), Universiti Teknologi Malaysia, 81310 Skudai, Johor Darul Takzim, Malaysia. E-mail addresses: [email protected] (W.N.W. Salleh), [email protected] (A.F. Ismail).
that present in the polymeric membrane will be released and replaced by carbon in the backbone of its molecular structure [9]. During the carbonization process, there is a possibility where the polymer precursor will melt due to the high temperature applied. To prevent this problem, thermosetting polymer is the best candidate to overcome this matter. The chosen polymer precursor must meet a number of criteria, such as high aromatic carbon content, high glass transition temperature (Tg), excellent thermal resistance, mechanical stability, and chemical stability, and provides high separation properties [10]. Polyimide has high possibility to fulfil those requirements and it has various classes which is determined by their different dianhydrides, which are pyromelliticdianhydride group (PMDA) such as Kapton [9], benzophenonetetracarboxylicdianhydride (BTDA) such as Matrimid and P84 [9–11], 3,30,4,40-biphenyltetracarboxylic dianhydride (BPDA) such as UIPR and UIP-S [11], and hexafluoroisopropylidene (6FDA) such as pyralin [12]. Among these classes, P84 that has superior gas separation performance [13–16] was chosen as the polymer precursor for this study. P84 is generally made by the polycondensation of aromatic acid dianhydrides and diamines. It was reported that its performance was influenced by the chemical structure of the constituent monomers as shown in Fig. 1. P84 co-polyimide (BTDA-TDI/MDI, co-polyimide of 3,30 ,4,40 ,-benzophenone tetracarboxylic dianhydride and 80% methyl phenylene- diamine + 20% methylene diamine) exhibits an excellent chemical and thermal resistance. This polymer is not only suitable for gas separation but also excellent in ultrafiltration and nanofiltration [10,17].
http://dx.doi.org/10.1016/j.jiec.2017.08.038 1226-086X/© 2017 The Korean Society of Industrial and Engineering Chemistry. Published by Elsevier B.V. All rights reserved.
314
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Fig. 1. Chemical structure of P84 (BTDA-TDI/MDI).
In this study, supported carbon membrane was prepared by introducing porous disk alumina as a support material to overcome the brittleness of the carbon membrane. Obtaining a defect-free thin layer film of carbon membrane is crucial as several parameters need to be considered, such as composition of the polymer precursor, coating method, and parameters during the heat treatment process, which includes temperature, heating rate, gas flow rate, environment gas, and soaking time [16]. Spin, spray, slip, and dip coating are the common methods applied to fabricate the supported carbon membrane. Past studies would suggested to implement several time of coating-carbonization cycle in order to obtain a defect-free carbon membrane layer that reflect in the gas separation performance [18,19]. However, these repeating methods are time and cost consuming. The one-step spray coating technique was proposed in this study by investigating the effect of polymer composition on gas permeation performance. The polymer precursor was distributed uniformly on the porous alumina support material. Until now, there are no studies reported on supported carbon membrane derived from P84 via spray coating technique. In 1999, Archarya and Foley applied spray coating method to form a fine mist of PFA on the porous support using airbrush that took less than 15 s to complete for one sample [20]. Most of the previous literatures were reported on dip and spin coating techniques, and recently, slip coating technique with several time of coating-carbonization cycles was also reported to reduce defect on the supported carbon membrane [9]. Air spray method can produce ultra-thin selective layer which can provide high selectivity without deteriorating the gas permeability [21]. The distribution of the polymer solution on the support is significantly affected by the polymer solution composition. Generally, dilute solution (2–5 wt% of polymer) is used when applying spray coating method to form uniform and thin layer. Furthermore, it is convinced that the spray coating method has a high potential for reproducibility of the thin and uniform carbon membrane layer [22]. Thus, the objective of this research was to investigate the effect of polymer compositions on the gas separation performance of the supported carbon membrane. The membrane was characterized by means of TGA, FTIR, BET, XRD, SEM images, and single gas permeation measurement. Experimental Materials Commercial co-polyimide BTDA-TDI/MDI (P84) powder purchased from Sigma–Aldrich (CAS#: 58698-66-1) was used as polymer precursor. N-Methyl-2-pyrrolidone (NMP) procured from Merck (Germany) was chosen as the solvent. Commercial symmetric porous alumina disk was utilized as supporting material with diameter of 47.0 mm, thickness of 1.0 0.05 mm, and mean pore size of 0.14 mm was bought from Shanghai Gongtao Ceramics Co., Ltd.
Carbon membrane preparation P84 polyimide powder was dried in an oven at 60 C for one day to remove water vapour. The polymeric solution was prepared by dissolving four different compositions of P84 at 6, 9, 12, and 15 wt% in NMP. The polymeric solution was stirred and heated at 70 C until homogenous solution was formed. It was sonicated for 3–4 h to eliminate bubble formed during stirring process. The supporting material was polished using fine silicon carbide (SiC) paper of grit P2000 before dry it in an oven for 3 h prior coating process. Porous alumina was coated by homogeneous solution via spray coating method at 1 bar at room temperature. The dope solution was sprayed directly towards the supporting material using air spray with distance of the spray nozzle and the alumina support at 20 cm. The coated alumina support was dried in an oven at 60 C overnight. The coated porous alumina disk with polymer precursor membrane was placed in the centre of the Carbolite horizontal tubular furnace for heat treatment process. It was carbonized at 700 C with heating rate of 3 C/min under nitrogen, then it was left to cool down to room temperature. This process took almost 4 h to achieve the final carbonization temperature and another 4 h to reach room temperature after the carbonization process. The nomenclature of the resultant disk supported carbon membranes was given in the form of P-polymer composition for polymeric membrane (P-6, P-9, P-12, P-15) and CM-polymer composition for carbon membrane (CM-6, CM-9, CM-12, CM-15). Membrane characterization Thermal behaviour of the membrane was obtained from thermogravimetric analysis (TGA 2050), carried out at temperatures between 50 C and 1000 C at a rate of 10 C/min under nitrogen gas at 50 ml/min. Elemental analysis was carried out using elemental analyzer model Vario Micro Cube indicate the percentage of the present elements in the membrane sample. While the functional groups that influence the gas performance were evaluated via Fourier transform infrared spectroscopy (FTIR), using single reflection diamond for Spectrum Two spectrometer (PerkinElmer, L1600107). The X’Pert PRO X-ray different diffractometer (XRD) from PANalytical with the 2u diffraction angle of 10–90 was performed using Cu Ka radiation of 1.54 Å wavelength. Bragg’s Law (nl = 2d sin u ) was adapted to determine inter planar distance (d-spacing) between the individual layers of the carbon. Brunauer–Emmett–Teller (BET) were obtained by utilizing equipment model Micromeritics 3 Flex Surface Characterization Analyzer. This nitrogen (N2) adsorption method was applied to obtain surface area and pore volume of porous materials from polymeric and carbon membrane. Surface and cross sectional morphologies of the carbon membrane was observed by scanning electron microscopy (SEM) model EOL JSM-5610LV. The prepared samples were coated with gold by sputter coating under vacuum to create neutral charge during SEM characterization. The viscosity of the polymeric solution was measured using a viscometer (Brookfield).
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Gas permeation study The gas permeation properties of the prepared membrane was evaluated using pure gas permeation system, tested at room temperature and feed pressure of 4 bar using bubble soap flow meter. The values of permeance and selectivity are determined using Eqs. (1) and (2). Permeance, P Pi Qi Q ¼ i2 DP ¼ l A DP pr
1GPU ¼ 1 106
ð1Þ
cm3 ðSTPÞ cm2 cmHg
where (Pi/l) is the gas permeance of a membrane (GPU), i is the gas species penetrating the membrane, Qi is the volumetric flow rate of gas i at standard temperature and pressure (sm3/s, STP), A is the membrane surface area (cm2), DP is the pressure difference between the feed side and permeation side of the membrane (cmHg), and r is the radius of the alumina disk (cm). Selectivity, a P =l ai=j ¼ i Pj =l
ð2Þ
where i/j i = is the selectivity of species gas i to species gas j, Pi/l and Pj/l are the permeance of gas i and j, respectively. Results and discussion Thermal behaviour analysis TGA with N2 atmosphere at flow rate of 50 ml/min and heating rate of 10 C/min was applied to assess the weight loss for the precursors of 6, 9, 12, and 15 wt% without support material. Fig. 2 illustrates the weight variation of polymer during heating process in the temperature range of 50–800 C. There are four stages of phenomena that occurred during the degradation of P84. Stage 1 that refers to the temperature of below 200 C is attributed to the existence of residue solvent and the release of adsorbed water from the precursor [10,16,17,23]. The second thermal degradation stage is due to the thermal rearrangement of original structure of P84, while the third stage represents the initial stages of polymer decomposition towards the formation of carbon membranes [24]. At 420 C, P-6, P-9, and P-15 experiences a sudden drop
315
approximately 5% from their original weight, while a sudden drop occurs for P-12 at 7% from its original weight. Higher Td with low weight loss is a good candidate to employ in carbon membrane fabrication as it can reduce melting possibility when exposed at high temperature (heat treatment process), which in this study refers to 12 wt% of P84. It is stated that the glass temperature (Tg) of the P84 polyimide is 315 C and as pointed out from the TGA result, Td for P84 is 420 C. Stabilization temperature can be determined from the Tg and Td results. For this study, 350 C is the stabilization temperature as it lies between both temperatures. When implementing high temperatures towards the long chain backbone components (polymer), it lead the chain to separate and react with one another to change the properties of the polymer and contributes in obtaining high gas separation performance [25]. Moreover, chemical and mechanical reactions involve during thermal degradation lead to physical and optical changes relative to the initially specified properties, including the reduction of ductility, colour changes (from light yellow to black), and physical properties. At final stage, the weight loss from 670 to 800 C is small at 5%. In this study, at the carbonization temperature 700 C, the 15 wt% polymer composition resulted in the highest weight loss of 46.3%, followed by 9 wt% (46.0%), 6 wt% (45.0%), and 12 wt% composition (43.0%). Elemental analysis The elemental analysis (percentage of atoms) of P84 (powder form), polymeric and carbon membrane at 12 wt% polymer precursors is presented in Table 1. The result of P84 and its polymeric membrane shows a different value, where P-12 has higher value in C element but lower value of H, N and O element as compared to P84. This phenomenon was due to the present of solvent element in the polymeric membrane, where the solvent was not fully evaporated during drying process and remain in the polymeric membrane. As the polymeric membrane contains the same functional group, the amount of C element for all the polymeric membrane are assumed to be similar. Theoretically, the C content in polymeric membrane should be lower in comparison to its carbon membrane as it has been occupied by greater value of N and O elements. Most of the heteroatoms (N and O elements) are released during the heat treatment process, resulting in small amount of them in carbon membrane matrix [9]. These phenomena affect the end product structure of the fabricated membrane that enhance the gas separation performance [26]. It is predicted that, when applying higher temperature, the greater amount of C element will be obtained as it has enough time to clearly eliminate heteroatoms. Based on Table 1, after the polymeric membrane was carbonized, the element of C content increased by 16% from its original level. Meanwhile, H, N, and O decreased by 34%, 18%, and 26%, respectively. Fourier transform infrared spectroscopy (FTIR) The FTIR spectrum of the polymeric and carbon membrane prepared at different polymer compositions are shown in Fig. 3. Based on the results, similar peaks of functional groups are observed for all the polymeric membranes. The key absorption Table 1 Elemental analysis of P84, polymeric and carbon membrane at 12 wt. % of P84. Membrane
P84 P-12 CM-12 Fig. 2. TGA analysis of P84-based polymeric membrane.
Element content (%) C
H
N
O
58.2 63.7 73.6
6.0 3.8 2.5
8.1 7.6 5.6
27.2 24.5 18.2
316
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Fig. 3. FTIR spectra of polymeric membrane and carbon membrane prepared at different polymer compositions of P84.
Fig. 4. XRD pattern of polymeric and carbon membrane prepared at difference polymer compositions of P84.
peaks for polymeric membrane is observed at 725 cm1 is assigned to C H (alkenes), 1100 cm1 and 1360 cm1 exhibited to OC N CO (imide II), 1510 cm1 show O¼C N (amide II) and 1720 cm1 and 1780 cm1 represent asymmetric OC NCO (imide I) [27,28]. A weak vibration at 3615 cm1 was corresponding to OH (carboxylic acid groups) is observed where the carbonyl stretching of the acid group is overlapped by the absorption of the imidic carbonyl [29,30]. However, after the carbonization process at 700 C under N2 gas flow, most of the absorption peaks disappear for all the carbon membranes. The peak for carbon membrane for P84 appears at 1140 cm1 (C N, amines). Due to the uncertain peaks of alkynes and carboxylic acid, both of the groups are ignored for present work because most of the original heteroatoms (O and N) are released during the carbonization process and results in the formation of cross-linked and stiff carbon matrix [6,9]. The peaks appeared in the range of 2000– 2500 cm1 mainly covers weak bond stretching modes, due to the existence of CO and CO2 in the beam. The CO and CO2 obviously seen only in carbon membrane due to the breaking of the original bond in its polymer precursor. Carbonization temperature induced the breakage of the presence bond in the polymeric membrane, resulted to new bonding as the previous bond was rearranged [31]. It is observed that FTIR was unable to characterize a few volatile compounds such as O2, N2, and H2, however the presence of volatile compounds (CO2, CO, O2, N2, benzene, aliphatic hydrocarbons, and other aromatic compounds) was clearly observed during thermal degradation [10].
(100) plane in graphite. Based on the XRD spectra, by converting polymer membrane to carbon membrane, it is revealed that the peak is shifted to the right that results in a decrease in the dspacing value [28]. Although various loading of polymer precursor in composition solvent were applied, the angles (2u) of the membranes were almost similar, which concluded that different amount of polymer precursor did not have significant affect towards the XRD characterization. Increasing P84 composition in the carbon membrane will obtain clear peak in the XRD due to the existence of the P84 that can be easily detected. However, a further study is required to ensure estimation more accurate result. Gas separation requires low d-spacing that result in low distance of the atomic planes and subsequently obtain a superior gas separation selectivity by separation mechanism of molecular sieving (<5 Å) [22].
Wide-angle X-ray diffraction patterns The XRD characterization was used to analyze the microstructure and determine the interlayer distance of the carbon matrix of the polymeric and carbon membranes. Fig. 4 illustrates the XRD pattern for polymeric and carbon membranes fabricated at different polymer compositions. The polymeric membrane shows peak at 18 with d-spacing value of 5 Å. After polymer precursor has been carbonized, the spectra of the carbon membranes exhibit a peak at 2u = 23 with (002) plane in graphite [32]. The theta values are shifted when the polymeric membrane is carbonized due to the presence of the graphitic and turbostratic structure for the efficient diffusion path for gas molecules to pass through [24]. In addition, the carbon membrane is the result of an amorphous carbon matrix with a mixture of sp2 and sp carbon components, which contribute a narrow pore size distribution [22,26,27]. Another weak and broad peak of the carbon membrane is found at 2u = 44 with d-spacing of 2.1 Å and characteristic peak of the
Pore structure analysis The pore structure analysis was determined by characterizing the polymeric and carbon membranes without their support material. Nitrogen adsorption isotherms is adopted to gain BET surface area, the micropore volume, and total pore volume. BET characterization is conducted to a relative P/P0 of approximately 0.3 at 196 C (77 K), where P0 is the saturation pressure. Table 2 shows the carbon membrane obtains a higher BET surface area in comparison to its original form (polymeric membrane). The BET surface is referred to the value of the surface area involves during the gas adsorption. The differences in the membrane morphological structure of the polymeric and carbon membrane lead to a huge gap of the BET surface area, thus the total pore volumes are also different [26]. The carbonization process that went through by the carbon membrane managed to break the original bond of the polymeric precursor and led to a new arrangement of atoms as proved in the FTIR characterization. Fig. 5 interprets the N2 adsorption of polymeric and carbon membrane prepared from 12 wt% of the polymer precursor P84. The initial N2 adsorption for the polymeric membrane is flat and low in quantity however, at approximately 0.9 P/Po, there is a very narrow range of a sudden increase in N2 adsorption. This represents the adsorption isotherm type I which describes the P-12 of either a chemisorption isotherm (final upswing at high pressures may not be present) or physical sorption on a material that has extremely fine pores (micropores). While carbon membrane marked by CM-12 has an acute increase in its initial N2 adsorption and thereafter the isotherms become flat. Then a slight increase of the N2 isotherms was observed. The carbon membrane exhibited the adsorption isotherms type II that
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
317
Table 2 Effect of membrane pore structure for polymeric and carbon membrane.
P-12 CM-12
BET surface area (m2/g)
Total pore volume of pores (cm3/g)
Average pore diameter (nm)
20 364
0.094 0.130
19.0 1.4
has high energy in adsorption. Moreover, these characteristics expressed that carbon membrane is mainly occupied by microporous material that gives impact towards the gas separation performance [23]. The difference of N2 adsorption isotherms between the carbon membrane and polymeric membrane is the contradicting with the pore size. Polymeric membrane that went through the carbonization process tends to obtain smaller pores as the pores would be shrinkage [33]. Morphological structure analysis The morphological structure view of the P84 polymeric membrane and P84 carbon membrane were determined by scanning electron microscopy (SEM). Fig. 6 shows the images of alumina disk support and polymeric membrane of P84 at different polymer precursor compositions. The bare alumina support structure is like a bricks arrangement that resulted in uneven surfaces. These uneven surfaces improve the adhesion mechanism between the membrane layer and disk supporting material. The pressure applied during the spray coating process contributes to a better adhesion strength between membrane and supporting material surface. Adhesion mechanism is crucial especially when the membrane undergoes carbonization process as the polymer tend to shrink and cause peeling problem. Non-uniform film layer of P-6 is observed, where most of the porous part of the alumina support is not fully covered because most of the solvent penetrate through the porous support. Increasing the polymer amount from 9 to 12 wt% makes the dope solution to be more uniformly distributed and most of the porous surface is covered [35]. High percentage of solvent has low viscosity (low fraction of solids) and low amount of solvent in dope solution has high fraction of solids that affect the distribution of solution on the porous alumina support during the spray coating method. This non-uniform thin film result in high permeance but low in selectivity. In addition, the thickness of the membrane increased when the weight percentage of the polymer increased. The distribution and thickness of the membrane layer is also depending on the viscosity of the polymeric solution. Table 3
presents the viscosity values for the polymeric solution at various P84 compositions. It clearly shows that the significant differences in the viscosity values affect the polymer distribution, membrane thickness, and gas performance. High viscosity is obtained when a high amount of polymer composition is used in the dope solution. The mass flow of the polymeric solution was influenced by the solution composition where low viscosity solution have a wide spread dispersion area. Theoretically, the low viscosity solution is faster to move and obtain a wide dispersion area, while increasing the viscosity value results in a slower mass flow and small dispersion area that directs the polymer solution to agglomerate [32]. In this study, the polymer composition of 12 wt% gives the best polymer distribution on the supporting material and has moderate thickness as compared to other samples. Fig. 7 shows the SEM images surface and cross section of carbon membranes which were originally from polymeric membrane P84 carbonized at 700 C in N2 gas flow of 200 ml/min with a heating rate of 3 C/min. As mentioned previously, heteroatoms that exist in polymeric membrane are replaced by carbon in the backbone of molecular structure during the carbonization process. The replacement of heteroatoms to carbon, forms the amorphous structure that possess large BET surface areas, controllable, and narrow pore size distribution [35– 37]. Based on carbon membrane surface, the distribution of thin film carbon membrane is better (uniform) compared to its polymeric membrane. After polymeric membrane has been carbonized, the carbon membrane become a denser microstructure compared to its original form, resulting in better morphological structure [25]. However, the defect formation of the supporting material was clearly observed from the carbon membrane surface. CM-6 has uneven surface image where the distribution of the membrane layer was not uniform and effected the thickness of the carbon membrane layer. As the P84 composition increased up to 9 wt%, CM-9 demonstrated smoother surface layer compared to CM-6. CM-12 and CM-15 have showed uniform carbon membrane layer because of their high viscosity which can prevent the polymeric solution from deep penetration inside the support material. It was reported that different composition of solid and solvent membrane created different morphologies in terms of their skin and sub-porous [38]. The thickness of the carbon membrane was lower compare to their polymeric membrane as it shrink after went through the carbonization process. Gas permeation measurements
Fig. 5. N2 adsorption isotherms for polymeric and carbon membrane.
In this study, both polymeric and carbon membrane were tested using pure gas permeation system. Among the polymeric membrane, P-12 obtained highest selectivity for CO2/N2 at 2.6 and CO2/CH4 at 7.7. All the prepared carbon membranes exhibited higher result compared to their original polymeric precursor membranes. After thermal decomposition of the polymeric precursor, carbon membrane obtains a highly aromatic structure comprising disordered sp2 hybridized hexagonal sheet, producing mostly char with turbostatic structure where pores are formed from the packing imperfections [17]. Carbon membrane mechanical properties was improved by the existence of these sp2 bonds which provide better hardness besides its superior gas separation performance. Moreover, the
318
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Fig. 6. SEM images of bare alumina supporting material surface and P84 polymeric membrane surface and cross section.
high gas separation performance of the carbon membrane is due to the micropores formation within the membrane structure that is influenced by thermal decomposition. Carbon membrane CM12 obtained highest selectivity values for CO2/N2 and CO2/CH4 at 15 and 45, respectively. The membrane separation mechanism depends on the pore size of the final product of the membrane. The separation mechanism of the polymeric and carbon
membrane are different due to the presence of this amorphous structure [21,39]. The permeance and selectivity of both polymeric and carbon membranes decreased and increased, respectively, when the polymer composition was increased from 6 to 12 wt%. It is because of the different concentrations of polymer precursor P84 in the polymer solution. Low composition polymer (conquered by
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321 Table 3 Viscosity of polymeric P84 compositions.
solution
at
different
Membrane
Viscosity (cP)
P-6 P-9 P-12 P-15
17 20 138 214
solvent) will penetrate through the porous alumina support and results in high mass transfer resistance that affect the gas separation performance [40]. Based on the SEM images obtained, low composition of P84 covers less porous alumina supporting material surface where the gap within the membrane was clearly seen. This drives to high permeability but low selectivity. On the other hand, when increasing the composition of P84 to 15 wt%, the selectivity drops, which is influenced by the distribution of the polymeric membrane on the supporting material surface. This high
319
polymer composition is affected by its high viscosity as shown by the viscosity test. Low viscosity solution lead to a very wide distribution of the fluid when it came out from the orifice. While for high viscosity solution, the spray coating pressure will not able to overwhelm the viscosity forces that affects the mass flow of the fluid [41]. Previous researchers preferred to use low composition of polymer solution when utilizing air spray coating technique [42,43]. CM-12 and CM-15 has achieved uniform surface layer but differ in membrane thickness due to the different of viscosity of the polymer composition. The lowest gas separation performance was obtained for CM-15 due to the higher mass transfer resistance during the gas transport compare to CM-12. This phenomenon was influenced by the membrane thickness where CM-15 has thicker membrane layer than CM-12. In this study, 12 wt% was preferred as it was easier to handle and exhibited a good gas separation selectivity performance. Fig. 8 shows the gas separation performances of CO2/N2 and CO2/CH4 for polymeric and carbon membrane with Robeson’s
Fig. 7. SEM images of surface and cross section of carbon membrane at different P84 composition.
320
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
The results showed that the CO2/N2 for both polymeric and carbon membrane was not able to exceed the Robeson upper bound line. For CO2/CH4 separation, CM-12 lied above the Robeson upper bound line that is the commercial region of interest [29]. The selectivity of selected supported carbon membranes was summarized and compared with the present work in Table 4. The selectivity performance of the present membranes are comparable as in the reported literatures. Most of the previous researchers preferred to fabricate their supported carbon membrane using either spin, dip, or slip coating technique, instead of using spray coating method [18,19,43,45–49]. Coating technique normally requires several number of coating-carbonization cycles to overcome the defect formation on the carbon membrane surface with extra time and care [16]. However, there are a few researchers claimed that a single coating-carbonization as the best method as it can reduce time and cost along with the fabrication of disk supported carbon membrane [50–52]. This present work applied one-step spray coating technique by employing spray coating method in one cycle of coatingcarbonization which took less time and reduced cost. The defect formation on the carbon membrane layer can be reduce by applying intermediate layer. In future, it was recommended to utilize intermediate layer on the porous support material as it could reduce the defect formation and increase the gas separation performance of the supported carbon membrane [21]. Conclusions
Fig. 8. Gas separation performances of (a) CO2/N2 and (b) CO2/CH4 for polymeric and carbon membrane with Robeson’s upper bound 2008.
upper bound 2008 [44]. The gas selectivity of CO2/N2 and CO2/CH4 increase for both polymeric and carbon membrane when the P84 composition increased from 6 wt% to 12 wt%. However, at 15 wt%, the selectivity was dropped due to high viscosity of the polymer solution which leads to thicker membrane layer. The membrane thickness played an important role during the gas transport process as it effects the gas separation performance. Based on Fig. 8, 12 wt% of polymeric and carbon membrane demonstrated the lowest gas permeance and the highest gas selectivity. Carbon membrane attained six times higher result compared to the polymeric membrane for CO2/N2 and CO2/CH4.
Disk supported carbon membrane was successfully fabricated by adapting commercialized alumina disk as a supporting material with only one-step spray coating technique. Various polymer solution compositions were sprayed on alumina disk before the carbonization process at 700 C under N2 gas environment with the heating rate of 3 C/min. The final product of the carbon membrane is determined by the composition of the polymer precursor as it plays an important role on the membrane properties. The TGA and FTIR spectra show almost similar results due to the similar compounds exist in the prepared membranes. The gas separation performance data increased when the polymer compositions was increased from 6 wt% to 12 wt% due to the uniform distribution of the polymer solution on the porous support surface. Based on the result, it is revealed that the promising gas permeation properties was obtained for the carbon membrane prepared with 12 wt% polymer composition. The CO2/N2 and CO2/CH4 selectivity of 15 and 45 was achieved with 82% and 84% increment from their polymeric membrane, respectively.
Table 4 Comparison of gas permeability and selectivity of supported carbon membranes with other works in the literature. Polymer
Configuration
Coating technique
Coatingcarbonization cycle
Carbonization temperature
CO2 permeance
PEI – Matrimid PFA Matrimid PPO Phenolic resin PEI P84
Alumina disk Tubular alumina Tubular alumina Alumina disk Tubular alumina Alumina disk Stainless steel tubular Alumina disk Alumina disk
Spin – Dip Spin Dip Spin Slip Spin Spray
– – 3 2 3 – 3 1 1
600 700 850 700 850 600 700 600 700
1046.00a – 131.75b 772.10b 287.36b 147.50a 32.00b 426.10a 400.00a
a b
Barrer = 1010 cm3 (STP) cm/(cm2 s cm-Hg). GPU = 106 cm3(STP)/(cm2 s cm-Hg).
Selectivity
Reference
CO2/N2
CO2/CH4
– 1.80 – 14.30 – 20.04 88.00 – 15.00
27.6 – 75.82 – 87.34 19.30 347.40 56.40 45.00
[46] [45] [18] [19] [43] [47] [35] [48] Present work
N.H. Ismail et al. / Journal of Industrial and Engineering Chemistry 57 (2018) 313–321
Acknowledgments The authors gratefully acknowledge the financial support from the Ministry of Higher Education and Universiti Teknologi Malaysia under Higher Institution Centre of Excellence Scheme (Project Number: R.J090301.7846.4J188), Fundamental Research Grant Scheme (Project Number: R.J130000.7846.4F846) and Research University Grant Scheme (Project Number: Q. J130000.2546.12H76). The authors would also like to acknowledge technical and management support from Research Management Centre (RMC), Universiti Teknologi Malaysia. References [1] S. Saedi, S.S. Madaeni, K. Hassanzadeh, A.A. Shamsabadi, S. Laki, J. Ind. Eng. Chem. 20 (2014) 1916. [2] Y.-C. Lee, S.M. Lee, W.G. Hong, Y.S. Huh, S.Y. Park, S.C. Lee, J. Lee, J.B. Lee, H.U. Lee, H.J. Kim, J. Ind. Eng. Chem. 23 (2015) 16. [3] E. Kim, E. Shin, G. Ko, S.H. Kim, O.H. Han, S.K. Kwak, Y. Seo, Chem. Eng. J. 306 (2016) 298. [4] R. Swaidan, X. Ma, E. Litwiller, I. Pinnau, J. Membr. Sci. 447 (2013) 387. [5] Q. Ma, H. Liu, H. Wang, L. Sun, S. Ma, Polym. Test. 54 (2016) 139. [6] W.N.W. Salleh, A.F. Ismail, T. Matsuura, M.S. Abdullah, Sep. Purif. Rev. 40 (2011) 261. [7] N. Sazali, W.N.W. Salleh, N.A.H.M. Nordin, A.F. Ismail, J. Ind. Eng. Chem. 32 (2015) 167. [8] L. Xu, M. Rungta, J. Hessler, W. Qiu, M. Brayden, M. Martinez, G. Barbay, W.J. Koros, Carbon 80 (2014) 155. [9] S.M. Saufi, A.F. Ismail, Carbon 42 (2004) 241. [10] P.S. Tin, T.-S. Chung, Y. Liu, R. Wang, Carbon 42 (2004) 3123. [11] S.S. Hosseini, M.M. Teoh, T.S. Chung, Polymers 49 (2008) 1594. [12] S.S. Hosseini, T.S. Chung, J. Membr. Sci. 328 (2009) 174. [13] S.S. Hosseini, M.R. Omidkhah, A. Zarringhalam Moghaddam, V. Pirouzfar, W.B. Krantz, N.R. Tan, Sep. Purif. Technol. 122 (2014) 278. [14] J. Ren, Z. Li, R. Wang, J. Membr. Sci. 309 (2008) 196. [15] T.A. Centeno, A.B. Fuertes, Sep. Purif. Technol. 25 (2001) 379. [16] S.C. Rodrigues, R. Whitley, A. Mendes, J. Membr. Sci. 459 (2014) 207. [17] M.J.C. Ordoñez, K.J. Balkus Jr., J.P. Ferraris, I.H. Musselman, J. Membr. Sci. 361 (2010) 28. [18] N. Sazali, W.N.W. Salleh, M. Nordin, N.A. Hadi, Z. Harun, A.F. Ismail, J. Appl. Polym. Sci. (2015) 132. [19] L.-H. Cheng, Y.-J. Fu, K.-S. Liao, J.-T. Chen, C.-C. Hu, W.-S. Hung, K.-R. Lee, J.-Y. Lai, J. Membr. Sci. 460 (2014) 1.
321
[20] M. Acharya, H.C. Foley, J. Membr. Sci. 161 (1999) 1. [21] N. Ismail, W. Salleh, N. Sazali, A. Ismail, Sep. Sci. Technol. (2017). http://dx.doi. org/10.1080/01496395.2017.1321671. [22] Y. Shen, A.C. Lua, J. Appl. Polym. Sci. 116 (2010) 2906. [23] J. Su, A.C. Lua, J. Membr. Sci. 305 (2007) 263. [24] X. Ma, R. Swaidan, B. Teng, H. Tan, O. Salinas, E. Litwiller, Y. Han, I. Pinnau, Carbon 62 (2013) 88. [25] B. Zhang, Y. Wu, Y. Lu, T. Wang, X. Jian, J. Qiu, J. Membr. Sci. 474 (2015) 114. [26] Y.-J. Fu, K.-S. Liao, C.-C. Hu, K.-R. Lee, J.-Y. Lai, Microporous Mesoporous Mater. 143 (2011) 78. [27] W. Qiu, C.-C. Chen, L. Xu, L. Cui, D.R. Paul, W.J. Koros, Macromolecules 44 (2011) 6046. [28] X. Qiao, T.S. Chung, AIChE J. 52 (2006) 3462. [29] E.M. Maya, A. Tena, J. de Abajo, J.G. de la Campa, A.E. Lozano, J. Membr. Sci. 349 (2010) 385. [30] E. Maya, A. Lozano, J. de Abajo, J. De la Campa, Polym. Degrad. Stab. 92 (2007) 2294. [31] J.B. Hamm, A.R. Muniz, L.D. Pollo, N.R. Marcilio, I.C. Tessaro, Carbon 119 (2017) 21. [32] P. Stähle, V. Gaukel, H.P. Schuchmann, Food Res. Int. 77 (Part 1) (2015) 55. [33] W.N.H.W. Zainal, S.H. Tan, M.A. Ahmad, Chem. Eng. Technol. 40 (2017) 94. [35] P.-S. Lee, D. Kim, S.-E. Nam, R.R. Bhave, Microporous Mesoporous Mater. 224 (2016) 332. [36] N. Bhuwania, Y. Labreche, C.S.K. Achoundong, J. Baltazar, S.K. Burgess, S. Karwa, L. Xu, C.L. Henderson, P.J. Williams, W.J. Koros, Carbon 76 (2014) 417. [37] P.J. Williams, W.J. Koros, Adv. Membr. Technol. Appl. (2008) 599. [38] S. Rafiq, Z. Man, A. Maulud, N. Muhammad, S. Maitra, J. Membr. Sci. 378 (2011) 444. [39] P.K. Chu, L. Li, Mater. Chem. Phys. 96 (2006) 253. [40] H.-H. Tseng, C.-T. Wang, G.-L. Zhuang, P. Uchytil, J. Reznickova, K. Setnickova, J. Membr. Sci. 510 (2016) 391. [41] G. Susanna, L. Salamandra, T.M. Brown, A. Di Carlo, F. Brunetti, A. Reale, Solar Energy Mater. Solar Cells 95 (2011) 1775. [42] T. Becherer, M. Vieira Nascimento, J. Sindram, P.-L.M. Noeske, Q. Wei, R. Haag, I. Grunwald, Prog. Org. Coat. 87 (2015) 146. [43] N.H. Ismail, W.N.W. Salleh, N. Sazali, A.F. Ismail, Chem. Eng. Trans. (2015) 1465. [44] L.M. Robeson, J. Membr. Sci. 320 (2008) 390. [45] N. Kruse, Y. Schießer, S. Kämnitz, H. Richter, I. Voigt, G. Braun, J.U. Repke, Sep. Purif. Technol. 164 (2016) 132. [46] H.-H. Tseng, K. Shih, P.-T. Shiu, M.-Y. Wey, J. Membr. Sci. 405–406 (2012) 250. [47] H.-H. Tseng, S.-H. Chang, M.-Y. Wey, J. Membr. Sci. 454 (2014) 51. [48] M.-Y. Wey, H.-H. Tseng, C.-k. Chiang, J. Membr. Sci. 453 (2014) 603. [49] W. Wei, G. Qin, H. Hu, L. You, G. Chen, J. Membr. Sci. 303 (2007) 80. [50] M. Teixeira, M. Campo, D.A. Tanaka, M.A. Tanco, C. Magen, A. Mendes, Chem. Eng. Res. Des. 90 (2012) 2338.