Thin film nanocomposite embedded with polymethyl methacrylate modified multi-walled carbon nanotubes for CO₂ removal

Abstract

Over the past decades, carbon nanotubes (CNTs) have gained tremendous attention as nanofillers in nanocomposite membranes owing to their potential to improve the physical properties and gas separation performance. In this work, polyamide–ethylene oxide (PA–EO) thin film nanocomposite (TFN) membranes embedded with polymethyl methacrylate (PMMA) grafted multi-walled carbon nanotubes (MWNTs) were successfully fabricated. The TFNs were fabricated via an interfacial polymerization (IP) technique to allow the formation of a very thin selective skin. The effects of the incorporation of nanofiller within the coating and selective thin film layers on the membrane morphologies and gas separation performance have been highlighted. The TFN incorporating milled PMMA-MWNTs within its coating layer showed a 29% increment in CO₂ permeance (70.5 gas permeation units (GPU)) with 47% and 9% enhancement in CO₂/N₂ and CO₂/CH₄ selectivity respectively compared to its thin film composite counterpart. While the improvement in gas separation performance can be primarily attributed to the presence of highly diffusive channels rendered by the CNTs, PMMA grafting is also believed to play an important role to ensure good nanofiller dispersion and good filler–polymer compatibility. Uncovering the construction of membrane fabrication could pave facile yet versatile ways for the development of effective membranes for greenhouse gas removal.

---

1. Introduction

The synergy between polymeric and inorganic materials has given rise to a range of composite membranes which inherit the high processability¹ of the polymer phase and the superior separation performance of the inorganic phase. Recently, many studies have demonstrated that the fabrication of thin film nanocomposites (TFNs)² through a two-step interfacial polymerization (IP) method holds very promising potential to significantly enhance the membrane properties such as permeability, selectivity and stability in various membrane separation processes. The IP approach allows the skin and the support layers to be tailored and optimized discretely,³ hence enabling a more versatile membrane to be produced. Apart from that, the self-sealing and self-termination properties of the polymerization process allow an intact thin skin layer with thickness ranging from 0.05 to 0.25 μm to be formed.⁴ Moreover, the penetration of a selective layer into the porous substrate pores during IP also prevents the delamination of the skin layer from the support layer.⁵ Recently, TFNs have been more commonly applied for liquid separation, however, the relevant literature reporting on their application to CO₂ separation is rather limited and the results exhibited by these TFNs were not up to expectation.^6–8 As such, it is of great interest to investigate the potential of TFN for CO₂ gas removal application.

In this study, CNTs were used as the TFN nanofiller owing to their good thermal and mechanical properties which could greatly enhance the strength⁹ and thermal stability of the polymer matrix. Best of all, the inner pores of the CNTs which are formed from a rolled-up sheet of graphene, are extremely smooth and can serve as diffusion channels that allow rapid mass transport (few orders of magnitude greater than that of zeolite with similar pore size) when incorporated into the polymer matrix.¹⁰ Nevertheless, despite the advantages offered by this nanomaterial, past studies have witnessed poor compatibility between CNT fillers and the polymer matrix which leads to the formation of defects such as interface voids, rigidification of the polymer surrounding the nanomaterial and blockage of filler pores.¹¹ This limitation has been recognized as the major hiccup to the development of CNT–polymer composites because the overall transport properties of the nanocomposite are critically dictated by the interface morphology at the nanoscale.¹² In addition, CNTs tend to entangle with one another into bundles of rope-like crystalline structures that are held strongly together by van der Waals forces due to their nano-sized tube-like structure and high aspect ratio. This presents a great challenge to achieve good dispersion and good interaction between the CNTs and the polymer matrix.¹³ Due to an increase in nanotube surface curvature,¹⁴ the aggregation tendency increases with decreasing number of graphene layers which follows the order of multi-walled carbon nanotube (MWNT) < double-walled carbon nanotube (DWNT) < single-walled carbon nanotube (SWNT). Thus, MWNTs were used in this study.

Functionalization of CNTs has been widely adopted to overcome the abovementioned complications. Shen et al.¹⁵ demonstrated that polymethyl methacrylate grafted MWNTs (PMMA-MWNTs) showed good compatibility with polyamide hence facilitated the formation of a dense TFN selective thin film layer. Furthermore, PMMA-MWNTs have good solubility in organic solvents, thus ensuring good dispersion of the nanofillers in the organic solvent used in the IP process.¹⁶ PMMA-MWNTs with high grafting density can be easily synthesized via environmentally friendly¹⁷ in situ emulsion polymerization to promote strong polymer–CNT interactions.¹⁸

Since the pristine MWNTs used in this study have dimensions of outer diameter 10 nm ± 1 nm × inner diameter 4.5 nm ± 0.5 nm × length 3–6 μm, in which the length of the MWNTs is greater than the average thickness of the thin film formed (0.15 μm),⁶ some of the randomly oriented nanotubes might protrude from the membrane surface if they were directly incorporated into the skin layer. In order to minimize the unfavourable CNT protrusion, the functionalized MWNTs were incorporated into the PDMS coating layer (sub-layer beneath the thin film) during the TFN fabrication. The nanotubes were also subjected to 8 h mechanical ball milling in order to assess the effectiveness of ball milling to shorten the nanotubes¹⁹ and minimize the protrusion.

Generally, transport through membranes is governed by diffusivity (depends on physical character) and solubility (depends on chemical properties) of the specific gas species. The transport rate of a targeted gas species (CO₂ in this case) could be elevated by the presence of reactive groups (amide and free-amine groups within the polymer matrix in this case) within the membrane chains whereas non-reactive gases rely on the diffusion mechanism to pass through the membrane. The benefit of facilitated transport for improving the selectivity is even more noticeable in dense membranes whereby the tightly packed polymer chains impose great resistance toward gas diffusion. Since CO₂ has strong affinity toward the EO groups,⁶ it is pulled towards the polymer chains and boosts the transport process of this gas. Thus, the permeance and selectivity of CO₂ could be enhanced. Therefore, in this study, monomers containing ethylene oxide (EO) groups i.e. diethylene glycol bis(3-aminopropyl) ether (DGBAmE) and trimesoyl chloride (TMC) were selected as active monomers for the formation of the selective layers to render high CO₂ permeance and selectivity. The effects of CNT modification (oxidation, PMMA grafting and ball-milling) and the incorporation of fillers at different layers (PDMS sub-layer and skin layer) towards the morphology and separation performance of TFN were evaluated. The CO₂ separation performance of the resultant TFNs was benchmarked with the neat thin film composite (TFC).

2. Experimental

2.1 Materials

MWNTs (98% carbon basis, OD × ID × L: 10 ± 1 nm × 4.5 ± 0.5 nm × 3–6 μm), polyvinylpyrrolidone (PVP, K90) and DGBAmE (97%) were purchased from Sigma Aldrich whereas sulfuric acid (H₂SO₄, 95–97%), nitric acid (HNO₃, 65%), methyl methacrylate (MMA, stabilized for synthesis), N-methyl-2-pyrrolidone (NMP, 99%), n-hexane (99%) and analytical grade potassium persulfate (KPS), sodium hydrogen carbonate (NaHCO₃) and sodium carbonate (Na₂CO₃) were purchased from Merck Milipore. Other chemicals include cetyltrimethylammonium bromide (CTAB, 99%, Acros Organics), trimesoyl chloride (TMC, 98%, Acros Organics), methanol (99.8%, Fisher Scientific), ethanol (96%, Fisher Scientific), polydimethylsiloxane coating solution (PDMS, Sylgard 184, Dow Corning) and polysulfone (PSf, Udel P-3500, Solvay).

2.2 Chemical oxidation of MWNTs

1.0 g of MWNTs was added into 200 mL of H₂SO₄ : HNO₃ (each 3 M) mixed acid solution with a volume ratio of 3 : 1 at 80 °C for 6 h to introduce carboxylic acid functional groups onto the MWNT surface. The dispersion solution was then filtered and the MWNTs residue was rinsed with deionized (DI) water until pH 6 was obtained. Lastly, the oxidized MWNTs were rinsed with 200 mL ethanol and left to dry overnight at 60 °C.

2.3 Synthesis of PMMA-MWNTs

10 g of CTAB, 10 g of MMA, 1.5 g f-MWNTs and 300 mL deionized (DI) water were mixed and sonicated for 20 min. The solution was then poured into a 500 mL three-neck round bottom flask that was equipped with a condenser, a mechanical stirrer and a nitrogen inlet and placed in a 70 °C oil bath. The air in the flask was replaced by a stream of nitrogen and the mixture was stirred at 500 rpm. 3 g of KPS and 2 g of NaHCO₃ were added into the mixture and the reaction was allowed to take place for 21 h with continuous stirring until a greyish odourless (all MMA reacted) solution was produced. The latex dispersion was drawn out from the flask and added into methanol solution and allowed to stand overnight. Finally, the precipitate was filtered via vacuum filtration and washed with excess methanol and DI water before dried in an oven at 60 °C.

2.4 Ball milling of functionalized MWNTs

1 g of functionalized MWNTs were placed in a 100 mL Scott bottle containing around 200 g of 1/4 inch steel balls. The bottle was then capped tightly and placed inside a bench-top laboratory roller mill and allowed to roll for 8 h.

2.5 Preparation of PSf supports

PSf (15 wt%) and PVP (3 wt%) were dissolved in NMP (82 wt%) with stirring until a homogeneous dope solution was obtained. The two opposing ends of a clean dried glass plate were each taped with three layers of cellulose tapes to act as guide during casting of the PSf support. The dope solution was poured onto the prepared glass plate slowly and casted using a clean glass roller. Then the glass plate was immediately immersed into a coagulation bath (water) at room temperature and allowed to sit for 10 min to ensure a complete phase inversion process. After that, the fabricated PSf membranes were stored in DI water for 24 h to remove solvent residue. The PSf supports were post-treated by immersing the membranes in ethanol solutions for 5 min followed by in n-hexane for 1 min and left to dry under ambient conditions for 12 h.

2.6 Preparation of composite membrane TFNs

The PSf membrane supports were sandwiched in between a glass plate and a Viton rubber frame followed by coating with PDMS solution for 10 min (step 1) then dried under ambient conditions for 12 h. After that, the coated PSf supports were sandwiched in the same setup again and impregnated with organic phase for 10 min (step 2). Excess solution was carefully removed from the surface before polymerizing the impregnated supports with aqueous phase for 3 min (step 3) to form the polyamide thin film. The resultant composite membranes were dried under ambient conditions for 12 h. Using the same procedure, TFC and TFNs embedded with O-MWNTs, PMMA-MWNTs, milled oxidized MWNTs (m-O-MWNTs) or milled PMMA-MWNTs (m-PMMA-MWNTs) were produced by varying the solution composition as presented in Table 1. The modified MWNTs are incorporated in the coating layer of TFN-O, TFN-mO, TFN-P and TFN-mP-c whereas for TFN-mP-f, the nanofillers are incorporated in the polyamide thin-film. All solutions containing nanofillers were sonicated for 1 h before use.

Table 1 Membrane sample descriptions

Sample ID	Step 1: coating (10 min)	Step 2: organic phase (10 min)	Step 3: aqueous phase (3 min)
TFC	2 wt% PDMS in hexane	0.28% w/v TMC in hexane	0.35% w/v DGBAmE + 0.4% w/v Na2CO3 in water
TFN-O	0.5 g L^−1 O-MWNTs + 2 wt% PDMS in hexane	0.28% w/v TMC in hexane	0.35% w/v DGBAmE + 0.4% w/v Na2CO3 in water
TFN-mO	0.5 g L^−1 m^−1-O-MWNTs + 2 wt% PDMS in hexane	0.28% w/v TMC in hexane	0.35% w/v DGBAmE + 0.4% w/v Na2CO3 in water
TFN-P	0.5 g L^−1 PMMA-MWNTs + 2 wt% PDMS in hexane	0.28% w/v TMC in hexane	0.35% w/v DGBAmE + 0.4% w/v Na2CO3 in water
TFN-mP-c	0.5 g L^−1 m^−1-PMMA-MWNTs + 2 wt% PDMS in hexane	0.28% w/v TMC in hexane	0.35% w/v DGBAmE + 0.4% w/v Na2CO3 in water
TFN-mP-f	2 wt% PDMS in hexane	0.5 g L^−1 m^−1-PMMA-MWNTs + 0.28% w/v TMC in hexane	0.35% w/v DGBAmE + 0.4% w/v Na2CO3 in water

---

2.7 Characterization of functionalized MWNTs and TFNs

The prepared samples were examined by attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR, Nicolet 5700, Thermo Electron Corporation) to identify the functional groups present in the samples upon the CNT functionalization and thin film formation. Thermogravimetric analysis (TGA, DSC822^e, Mettler Toledo) conducted in the temperature range of 25–600 °C and under nitrogen gas flow at heating rate of 10 °C min⁻¹ was used to determine the degree of MWNT functionalization. Transmission electron microscopy (TEM, HT7700, Hitachi) was used to inspect the structure of the pristine and modified MWNTs. The surface and cross section morphologies of the fabricated membranes were investigated using field emission scanning electron microscopy (FESEM, SU8020, Hitachi). In order to observe the cross section of the membranes, the samples were fractured in liquid nitrogen to obtain a clean break across the membranes.

2.8 Gas permeation measurement

All membrane samples were saturated with water for 5 min prior to the testing. The pressure-normalized fluxes of the membranes were determined by a variable volume method using pure (N₂), methane (CH₄) and CO₂ at room temperature. The measurement was conducted with an input stream pressure of 2 bar while the outlet stream was exposed to atmospheric conditions. A circular membrane disc with effective permeation area of 13.5 cm² was used. Volumetric gas permeation rates were determined with a simple soap bubble flow meter. The upstream of the chamber was always purged with the test gas prior to the permeation measurement. The gas permeation rate was measured after the steady state was reached and each set of data was determined as an average of three replicates. The gas permeance was determined using the following expression:

(P/l)_i = Q_i/(ΔpA) (1)

where, Q_i (cm³ s⁻¹) is the volumetric flow rate of gas ‘i’ or ‘j’ at standard temperature and pressure, l is the membrane thickness (cm), A is the effective membrane area (cm²) and Δp (cmHg) is the trans-membrane pressure difference. Permeances are expressed in gas permeation units, GPU, where:

1 GPU = 1 × 10⁻⁶ cm³ (STP) (cm² s cmHg)⁻¹ (2)

The pure gas selectivity was obtained by taking the ratio of the pure gas permeabilities:

αⁱ_j = (P/l)_i/(P/l)_j (3)

3. Results and discussion

3.1 Characterization of functionalized MWNTs

As depicted in Fig. 1, all the spectra show broad absorption bands around 3200 cm⁻¹ and 3600 cm⁻¹ (area shaded in grey color) which are attributed to –OH stretching, indicating the presence of hydroxyl groups in all the MWNTs. For PMMA-MWNTs, the absorption peak of carbonyl groups (C=O) is detected at 1726 cm⁻¹ while the peaks at 1156 cm⁻¹ and 1046 cm⁻¹ corresponding to the C–O stretching in PMMA can be clearly observed. Characteristic peaks of –CH₂ and –CH₃ bend vibration are also detected at 1464 cm⁻¹ and 1384 cm⁻¹ respectively. The spectra obtained clearly indicated that PMMA has been successfully grafted on the nanotube surface.^20,21

Fig. 1 FTIR spectra of pristine MWNT, O-MWNT and PMMA-MWNT.

Based on Fig. 2, below 100 °C, there was a mass loss of less than 2% in all samples due to the removal of absorbed water. At the decomposition temperature below 600 °C, the pristine MWNTs experienced a minute mass loss of about 1.7% due to the breakdown of carboxylic and hydroxyl groups whereas the oxidized sample underwent more than 6% loss in mass.²² This is a good indication that the pristine MWNTs were successfully oxidized. The mass of PMMA-MWNTs sample decreased rapidly from 200 °C to 450 °C due to the decomposition of grafted PMMA. Based on the difference in mass loss between PMMA-MWNTs and O-MWNTs at 450 °C by taking their respective mass loss from evaporated moisture content into account, a PMMA grafting degree of 11.9% was estimated. This value is in good agreement with literature that employed a similar functionalization procedure.¹⁸

Fig. 2 TGA profiles of pristine MWNTs, O-MWNTs and PMMA-MWNTs.

As the introduction of functional groups onto the nanotubes via covalent functionalization will inevitably affect the MWNTs structural integrity,^23–25 TEM analysis was performed to evaluate the extent of the nanotube structural changes. Fig. 3 presents the TEM images of the pristine, oxidized and PMMA grafted MWNT. TEM images revealed that amorphous carbon was deposited (indicated by arrow) along the surface of the pristine MWNTs (Fig. 3b) and most of the nanotube tips remained closed (indicated by red box in Fig. 3a). Upon acid oxidation, the amorphous deposition was removed and the nanotubes tips were partially or completely opened (indicated by red circles in Fig. 3c). From Fig. 3h and j, the deposition of a PMMA layer on the grafted MWNTs could not be visibly observed probably due to the low concentration of the PMMA monomer used in the functionalization. Nevertheless, the m-PMMA-MWNTs (Fig. 3i) appeared to be less entangled compared to the pristine (Fig. 3a) and m-O-MWNTs (Fig. 3e) samples, which indicates the effectiveness of the modification to enhance the dispersion of the nanotubes. Overall, the modified nanotubes still retained their tubular structure integrity with no visible wall damages.

Fig. 3 TEM images of (A and B) pristine MWNT, (C and D) O-MWNTs, (E and F) m-O-MWNTs, (G and H) PMMA-MWNTs and (I and J) m-PMMA-MWNTs.

Based on the images in Fig. 4, it is obvious that oxidized MWNTs dispersed poorly in PDMS coating solution (Fig. 4(Ai and Bi)) due to nanotube aggregation (Fig. 4(Aii and Bii)) which leads to non-uniform embedment of the fillers within the TFNs (Fig. 4(Aiii and Biii)). It is also noticeable that the milled PMMA-MWNTs exhibited poor dispersion in the TMC–hexane organic solution (Fig. 4Ei). This might be due to the low degree of PMMA grafting on the nanotubes which could not completely shield the effect of hydrogen bonding resulting from the interactions between the remaining surface hydroxyl and carboxyl groups. However, when the PMMA grafted MWNTs were mixed into the coating solution, good dispersion was observed (Fig. 4(Ci and Di)), which in turn resulted in uniform dispersion of nanotubes within the TFN (Fig. 4(Ciii and Diii)).

Fig. 4 Dispersion of (Ai) O-MWNTs, (Bi) m-O-MWNTs, (Ci) PMMA-MWNTs and (Di) m-PMMA-MWNTs in PDMS coating solutions and (Ei) m-PMMA-MWNT in organic phase after 1 h sonication and being left to stand for 24 h in sample bottles. (Aii, Bii, Cii, Dii and Eii) Dispersion of 10 mL of each of the solutions that were spread on petri dishes, (Aiii, Biii, Ciii, Diii and 3iii) the resultant TFNs and images of dispersed (Biv and Bv) m-O-MWNTs and (Div and Dv) m-PMMA-MWNTs in PDMS coating layer.

The FESEM images in Fig. 4(Biv, Bv, Div and Dv) evidently depict the difference in aggregation tendency of the functionalized MWNTs in the coating layer. The m-O-MWNTs tend to agglomerate severely into dense balls of entangled nanotubes (Fig. 4Biv) whereas the m-PMMA-MWNTs tend to form networks of tubes (Fig. 4Div) that were adhered by polymeric linkages (Fig. 4Dv) similar to the findings reported by Liu et al.¹⁸ This indicates that the nanotubes were made compatible with the PDMS coating through the introduction of PMMA chains on the MWNTs.

Although the loading of all functionalized MWNTs has been fixed at 0.5 g L⁻¹, it is observed that the height of the settled m-O-MWNTs was lower than that of O-MWNTs (compare Fig. 4Ai and 4Bi). This indirectly indicates that the ball milling has successfully reduced the nanotube lengths because shorter tubes are able to form smaller and more compact aggregates (compare Fig. 4Aii and 4Bii) that occupied less volume compared to their counterparts of the same mass.

3.2 Characterization of TFN

ATR-FTIR spectra of all the composite membranes in Fig. 5 show that interfacial polymerization has successfully taken place since strong characteristic peaks of C=O (amide I) at 1655 cm⁻¹, N–H (amide II) at 1543 cm⁻¹ and C–N at 1243 cm⁻¹ which correspond to the amide groups in polyamide thin film were detected.

Fig. 5 ATR-FTIR spectra of fabricated composite membranes.

Fig. 6 shows the FESEM images of TFC and TFNs. The surface of TFC was covered by dense nodular structures which have been commonly reported for typical polyamide thin films⁴ whereas the surface morphologies of the TFNs were similar to those reported by Ma et al.²⁶ Generally, the incorporation of nanofillers into the thin film increased the size of the nodules with broadened ridge and valley features (Fig. 6(b–f)). While the nodules of TFN-O and TFN-mO tend to grow into leaf-like structures that overlap with one another (Fig. 6(b and e)), the nodules of TFN-P and TFN-mP-c formed linkages among themselves into a net-like structure (Fig. 6(c and f)).¹⁵ Although both TFN-mO (Fig. 6e) and TFN-mP-c (Fig. 6f) seemed to have uniform surface morphology, the polyamide structures of TFN-mP-c were more intricate and have denser coverage.

Fig. 6 Surface morphologies of (a) TFC, (b) TFN-O, (c) TFN-P, (d) TFN-mP-f, (e) TFN-mO and (f) TFN-mP-c obtained at × 5k magnification and cross sections of (g) TFC, (h) TFN-mP-f, (i) TFN-O, (j) TFN-P, (k) TFN-mO, and (l) TFN-mP-c obtained at × 20k magnification using FESEM.

Other than that, it can be obviously seen that the surface structures on TFN-O (Fig. 6b) and TFN-P (Fig. 6c) are coarser compared to the TFNs embedded with milled nanotubes (Fig. 6(e and f)). Since the coarse structures are attributed to the agglomeration of nanotubes, this observation suggests that mechanical milling can suppress the formation of aggregates and allow the formation of a uniform thin film layer with much more refined structures. In the case of TFN-mP-f, some areas were not completely covered by the polyamide (Fig. 6d) which could lead to the deterioration of the membranes gas separation performance, particularly the gas selectivity. Based on the cross section images, the thickness of the skin layer of all the resultant composite membranes is estimated to be around 110 ± 10 nm.

3.3 Gas separation performance of TFNs

It should be noted that the gas permeation test was also performed on the dry polyamide composite membranes and the separation behaviour of these dry samples was similar to that reported by Andrew Lee et al.²⁷ in which their performance was not as good as the wetted ones. Once the composite membranes were wetted, the CO₂ permeance and CO₂/N₂ selectivity increased while the N₂ permeance decreased.

Based on the results tabulated in Table 2, the TFCs performed best after 5 min of soaking in water. The enhancement was due to the facilitated transport mechanism of CO₂ which was activated in the presence of moisture content.²⁸ In this study, all the membrane samples were wetted for at least 5 min prior testing. To maintain the moisture content of the membranes, the permeance test was carried out for no longer than 30 min for each sample and the membranes are visually confirmed to remain in moist condition after the experiment.

Table 2 Gas separation performance of dry and wetted TFCs

Sample description	Soaking duration (s)	CO2 permeance (GPU)	CO2/N2 selectivity
0.2 wt% TMC in organic phase reacted with 0.25 wt% DGBAmE in aqueous phase on PDMS coated PSf	0	23.51	29.52
	60	61.72	32.50
	300	70.54	41.43
	600	70.54	40.00
0.28 wt% TMC in organic phase reacted with 0.35 wt% DGBAmE in aqueous phase on PDMS coated PSf	0	61.72	2.25
	60	82.30	41.67
	300	70.54	34.29
	600	70.54	45.71

---

Overall, the experimental results tabulated in Table 3 show that the TFN embedded with functionalized MWNTs experienced significant permeance improvement for all the tested gases. The increase in permeance was primarily due to the addition of nanotubes which permit faster gas flow through the hollow cavity. However, this observation could also imply the presence of defects that can be related to the deterioration in selectivity. TFN-O and TFN-P showed reduction in CO₂/N₂ and CO₂/CH₄ selectivity due to the formation of coarse-sized aggregates that hindered the formation of defect-free polyamide skin layer.

Table 3 Gas permeance, CO₂/N₂ and CO₂/CH₄ selectivity of resultant composite membranes^a

Sample ID	Permeance, P (GPU)			Percentage change of permeance (%)			Selectivity, α		Percentage change of selectivity (%)
	N2	CH4	CO2	ΔPN2	ΔPCH4	ΔPCO2	CO2/N2	CO2/CH4
a Percentage change in permeance of specific gas through TFN compared to TFC, ΔPi = (PiTFN − PiTFC)/PiTFC × 100%. Percentage change in selectivity of TFN compared to TFC, where i refers to N2, CH4 or CO2.
TFC	1.20	2.06	54.87	—	—	—	45.73	26.64	—	—
TFN-O	1.83	4.70	61.72	52.50	128.16	12.48	33.73	13.13	−26.24	−50.70
TFN-mO	1.14	2.18	67.33	−6.67	5.83	22.71	59.06	30.89	29.17	15.95
TFN-P	1.67	2.82	58.09	39.17	36.89	5.87	34.78	20.60	−23.93	−22.66
TFN-mP-c	1.05	2.43	70.54	−12.50	17.96	28.56	67.18	29.03	46.92	8.98
TFN-mP-f	2.10	4.11	89.78	75.00	99.51	63.62	42.75	21.84	−6.50	−17.99

---

On the hand, TFNs incorporating milled functionalized MWNTs at the coating layer demonstrated enhancement in CO₂ and CH₄ permeance without sacrificing the selectivity. Both TFN-mO and TFN-mP-c exhibited higher CO₂/N₂ and CO₂/CH₄ selectivity than TFC. In this case, the enhancement in CO₂ and CH₄ permeance of both TFNs can be directly related to the incorporation of functionalized MWNTs which have provided rapid diffusion channels for the passage of gas molecules.²⁹ It is postulated that the presence of ethylene oxide, amide and free-amine groups in the defect-free polyamide thin film favours the transport of CO₂. Besides, the kinetic diameter of CO₂ (3.3 Å) which is smaller than that of CH₄ (3.8 Å)³⁰ also allowed the former to diffuse faster through the polymer matrix as it required less activation energy.³¹ Thus, when the enhancement in CO₂ permeance was greater compared to CH₄, significant improvement in CO₂/CH₄ selectivity can be achieved. Surprisingly, permeance of N₂ has decreased upon addition of the milled functionalized MWNTs. One plausible explanation is the rigidification of EO containing polyamide chains around the nanotubes. Rigidification reduced the polymer chain mobility and increased the resistance toward mass transport.³² Consequently, the permeance of the inert N₂ cannot benefit from the nanotubes while the increase in local EO and amide concentration (more EO containing polyamide chains compacted together) around the MWNTs might induce greater affinity towards CO₂, thus enhancing its permeance. Furthermore, the addition of nanofillers within the coating layer has increased the tortuosity of the gas diffusion path and hence slowed down the transport of N₂ as N₂ molecules cannot utilize the nanotube as a diffusion path and can only diffuse around the tube within the polymer matrix of the membrane.³³ It is also important to note that the magnitude of N₂ permeance reduction and CO₂ permeance increment was greater for TFN-mP-c compared to that of TFN-mO. This could be an indication of good compatibility between PMMA grafted MWNTs and the polyamide chains which allowed the formation of polyamide with higher crosslinking compared to the TFN embedded with oxidized MWNTs.¹⁵

While we have successfully demonstrated that the incorporation of milled PMMA-MWNTs within the coating layer could improve the separation performance of TFC, the attempt to produce a defect-free skin layer embedded with these nanofillers remains a challenge due to the unfavorable length of nanotubes used in this study. The TFN-mP-f results have a similar trend to that of TFN-P. It is believed that the incorporation of nanotubes within the skin layer has resulted in more severe breach of PMMA-MWNTs out from the selective skin layer. This prediction has been evidenced by the higher gas permeance of TFN-mP-f as compared to TFN-P and TFN-mP-c. Yet, it is worth mentioning that the selectivity loss of TFN-mP-f was much lower than that of TFN-P. This can be attributed to the substantial increase in CO₂ permeance (55% higher than TFN-P) compared to N₂ (26% higher than TFN-P) and CH₄ (46% higher than TFN-P).

Based on the results obtained and observations made throughout this study, possible mechanisms for the formation of TFNs are proposed in Fig. 7. As illustrated, TFN-P and TFN-mP-c were formed through coating of the PSf supports with PDMS solution containing PMMA grafted nanotubes (Fig. 7(a and d)). The coating layer shrunk upon drying (Fig. 7(b and e)). Since the nanotubes in TFN-mP-c were shorter than those in TFN-P, they tend to fill the narrow spaces hence allowed the formation of a much more compact layer with suppressed CNT protrusion. As such, when the interfacial polymerization took place atop of these coated supports, the resultant polyamide layer of TFN-mP-c (Fig. 7f) exhibited much more refined surface morphology compared to that of TFN-P (Fig. 7c). Furthermore, the tightly assembled m-PMMA-MWNTs have also impose greater resistance and more tortuous paths toward gas diffusion than the longer counterparts which eventually led to enhancement in their selectivity.

Fig. 7 Illustration of proposed formation of (a, b and c) TFN-P, (d, e and f) TFN-mP-c and (h, i, j and k) TFN-mP-f (not to scale).

However, during the impregnation process with TMC–hexane solution containing milled PMMA-MWNTs to form TFN-mP-f, the nanotubes distributed closer to the surface were found to adhere on the coating layer as depicted in Fig. 7h. Due to the relatively short mechanical milling duration, a mixture of long and short nanotubes might present in the sample in which the longer nanotubes tend to settle at a faster rate compared to the shorter ones. Therefore, it was expected that the coating surface was largely distributed with longer nanotubes. As the organic solution is poured away, nanotubes that were not properly adhered were removed whereas those remaining on the surface attached even more firmly during the drying process (Fig. 7i). As shown in Fig. 7(j and k), since the nanotubes were positioned at the outermost layer, the polyamide layer was formed according to the randomly arranged nanotubes and hence resulted in a rough surface that was filled with irregular coarse structures.

4. Conclusions

In this study, TFNs have been fabricated using the well-established IP technique. This study has highlighted the significance of monomer selection and functionalization of nanotubes to ensure the successful fabrication of TFNs with enhanced CO₂ gas removal performance. It is also imperative for one to take note of the details of the fabrication technique as we have demonstrated that the incorporation of the same milled PMMA-MWNTs at different layers of the TFN could result in completely different morphologies and gas separation performance. In conclusion, PMMA-MWNTs exhibited good dispersibility in PDMS coating solution and enabled the fabrication of TFN with enhanced gas separation performance. As a result, TFN embedded with milled PMMA-MWNTs in the coating layer has shown improvement of about 29%, 47% and 9% in CO₂ permeance, CO₂/N₂ selectivity and CO₂/CH₄ selectivity respectively compared to its TFC counterpart. Based on the findings of this study, it is anticipated that TFNs incorporating PMMA-MWNTs can serve as potential candidates for promising CO₂ gas removal application.

Acknowledgements

The authors would like to acknowledge the financial support obtained from Research University Grant (vot. no.: 04H86) and Fundamental Research Grant Scheme (vot. no.: 4F306).

Notes and references

1. J. K. Adewole, A. L. Ahmad, S. Ismail and C. P. Leo, Int. J. Greenhouse Gas Control, 2013, 17, 46–65.
2. W. J. Lau, A. F. Ismail, N. Misdan and M. A. Kassim, Desalination, 2012, 287, 190–199.
3. S. Sorribas, P. Gorgojo, C. Téllez, J. Coronas and A. G. Livingston, J. Am. Chem. Soc., 2013, 135, 15201–15208.
4. X. Yu, Z. Wang, Z. Wei, S. Yuan, J. Zhao, J. Wang and S. Wang, J. Membr. Sci., 2010, 362, 265–278.
5. J. Zhao, Z. Wang, J. Wang and S. Wang, J. Membr. Sci., 2006, 283, 346–356.
6. S. Li, Z. Wang, C. Zhang, M. Wang, F. Yuan, J. Wang and S. Wang, J. Membr. Sci., 2013, 436, 121–131.
7. S. Sridhar, B. Smitha, S. Mayor, B. Prathab and T. M. Aminabhavi, J. Mater. Sci., 2007, 42, 9392–9401.
8. J. Albo, J. Wang and T. Tsuru, J. Membr. Sci., 2014, 449, 109–118.
9. H. Cong, J. Zhang, M. Radosz and Y. Shen, J. Membr. Sci., 2007, 294, 178–185.
10. L. Ge, L. Wang, A. Du, M. Hou, V. Rudolph and Z. Zhu, RSC Adv., 2012, 2, 5329–5336.
11. D. Bastani, N. Esmaeili and M. Asadollahi, J. Ind. Eng. Chem., 2013, 19, 375–393.
12. T.-S. Chung, L. Y. Jiang, Y. Li and S. Kulprathipanja, Prog. Polym. Sci., 2007, 32, 483–507.
13. S. M. Sanip, A. F. Ismail, P. S. Goh, T. Soga, M. Tanemura and H. Yasuhiko, Sep. Purif. Technol., 2011, 78, 208–213.
14. X. Liu, M. Wang, S. Zhang and B. Pan, J. Environ. Sci., 2013, 25, 1263–1280.
15. J. n. Shen, C. c. Yu, H. m. Ruan, C. j. Gao and B. Van der Bruggen, J. Membr. Sci., 2013, 442, 18–26.
16. S. Shang, L. Li, X. Yang and Y. Wei, Compos. Sci. Technol., 2009, 69, 1156–1159.
17. F.-A. Zhang, D.-K. Lee and T. J. Pinnavaia, Polymer, 2009, 50, 4768–4774.
18. X.-Q. Liu, S.-Q. Shen, R. Wen, W. Yang, B.-H. Xie and M.-B. Yang, Composites, Part B, 2013, 53, 9–16.
19. A. Fonseca, S. Reijerkerk, J. Potreck, K. Nijmeijer, Z. Mekhalif and J. Delhalle, Desalination, 2010, 250, 1150–1154.
20. H.-X. Wu, X.-Q. Qiu, W.-M. Cao, Y.-H. Lin, R.-F. Cai and S.-X. Qian, Carbon, 2007, 45, 2866–2872.
21. H. Xia, Q. Wang and G. Qiu, Chem. Mater., 2003, 15, 3879–3886.
22. F. Avilés, J. V. Cauich-Rodríguez, L. Moo-Tah, A. May-Pat and R. Vargas-Coronado, Carbon, 2009, 47, 2970–2975.
23. N. G. Sahoo, S. Rana, J. W. Cho, L. Li and S. H. Chan, Prog. Polym. Sci., 2010, 35, 837–867.
24. S. W. Kim, T. Kim, Y. S. Kim, H. S. Choi, H. J. Lim, S. J. Yang and C. R. Park, Carbon, 2012, 50, 3–33.
25. M. A. Aroon, A. F. Ismail, M. M. Montazer-Rahmati and T. Matsuura, J. Membr. Sci., 2010, 364, 309–317.
26. N. Ma, J. Wei, R. Liao and C. Y. Tang, J. Membr. Sci., 2012, 405–406, 149–157.
27. S. Andrew Lee, G. W. Stevens and S. E. Kentish, J. Membr. Sci., 2013, 429, 349–354.
28. S. Saedi, S. S. Madaeni, F. Seidi, A. A. Shamsabadi and S. Laki, Chem. Eng. J., 2014, 236, 263–273.
29. L. Ge, Z. Zhu and V. Rudolph, Sep. Purif. Technol., 2011, 78, 76–82.
30. Y. Zhang, J. Sunarso, S. Liu and R. Wang, Int. J. Greenhouse Gas Control, 2013, 12, 84–107.
31. A. Singh and W. J. Koros, Ind. Eng. Chem. Res., 1996, 35, 1231–1234.
32. A. F. Ismail, N. H. Rahim, A. Mustafa, T. Matsuura, B. C. Ng, S. Abdullah and S. A. Hashemifard, Sep. Purif. Technol., 2011, 80, 20–31.
33. M. A. Aroon, A. F. Ismail, T. Matsuura and M. M. Montazer-Rahmati, Sep. Purif. Technol., 2010, 75, 229–242.

---