Facile modification of ZIF-8 mixed matrix membrane for CO₂/CH₄ separation: synthesis and preparation

Abstract

Metal–organic frameworks (MOFs) possess tunable characteristics that allow various modifications to be practiced and suitability for specific applications. In this study, zeolitic imidazole framework 8 (ZIF-8) was synthesized and subjected to ammonia modification under different temperatures and ammonia solution loading. The presence of the N–H group observed after the modification indicates the successful modification of ZIF-8. The modified ZIF-8 showed notable changes with increased phase crystallinity, micropore volume and BET surface area. The unmodified and modified ZIF-8s were then dispersed into a polysulfone (PSf) matrix, and the MMMs were prepared via dry/wet phase inversion. No apparent differences in the membrane’s morphology and thermal stability were noticed between the neat PSf membrane and the MMMs. The MMMs were further subjected to pure CO₂ and CH₄ gas permeation experiments. CO₂ permeance decreased while CO₂/CH₄ selectivity increased as a result of ZIF-8 modification, due to the decrease in mesopore contribution and the increase in micropore contribution to the gas permeation path. The affinity of the N–H group in the modified ZIF-8 to CO₂ also contributed to the increase of CO₂ permeance. For example, when the ZIF-8 modified in 25 mL ammonia solution at 60 °C (Z25c) was dispersed in the PSf matrix, the CO₂/CH₄ selectivity increased to 72% and CO₂ permeability to 43% compared to the neat PSf membrane.

---

1. Introduction

The depletion of natural gas resources as well as their worsening quality are great concern nowadays. It was reported that natural gas usually consists of up to 80% impurities, depending on geographical factors.¹ Among the impurities in natural gas, CO₂ is considered as the most crucial species that needs to be intensively removed. Aside from environmental issues such as being the main culprit of climate change, the presence of CO₂ in the process streams would decrease the heating value, pipeline corrosion, and consequently lead to an increase in operational and maintenance costs. Challenges arise to ensure that the presence of CO₂ in pipelines is below 0.2% based on U.S. pipeline specifications before the product is delivered to customers. Hence, an efficient CO₂ removal system is necessary. The well-developed processes for CO₂ removal such as amine absorption, pressure swing adsorption and cryogenic distillation have been implemented over the years and have demonstrated to be reliable in large scale operation and adaptable for many industries.^2,3 However, limitations of the conventional processes such as high solvent losses and degradation, unit corrosion, high maintenance, high energy requirement, and refrigerant toxicity^4,5 have driven the search for more economical and efficient gas separation systems.

Over the years, attention has been focused on polymeric membranes for natural gas separation. Polymeric membranes are advantageous in terms of versatility in a wide-range of CO₂ concentrations, low operating cost, modest energy requirement, and space efficiency compared to the conventional processes.⁶ In addition, membrane separation as a pressure-driven process would be practical for high pressure systems in natural gas processing and requires no additional compression units to transport the retentate stream.⁵ However, polymeric membranes are bound by the Robeson trade-off limits between permeability and selectivity, which has hindered its potential application in gas separation.⁷ Although new classes of polymer such as Thermally Rearrange (TR) polymers and polymers of intrinsic microporosity (PIM) have shown to be promising candidates for gas separation,⁸ their development is still under investigation and unlikely to be commercially available for large scale membrane production in the next few years.

Limitations suffered by polymeric membranes have led to research for inorganic membranes as alternatives. Inorganic membranes such as zeolites and carbon membranes offer several advantages over conventional polymeric membranes in terms of mechanical strength, thermal stability and resistance towards a wide range of chemicals. Particularly, inorganic membranes provide better gas permeability and gas pair selectivity, exceeding Robeson’s upperbound suffered by polymeric membranes. Despite their prevalence, fabricating continuous defect free inorganic membranes is a challenge due to their brittle structure. As well, requirement of extensive fabrication energy is one of the obstacles that hinder their diverse applications. Hence, recent membrane development has been focused on mixed matrix membranes (MMMs), a combination of a polymeric membrane as the continuous phase with an inorganic particle as the disperse phase. Incorporating filler into the polymer matrix can either improve permeability while maintaining gas pair selectivity⁹ or improve gas pair selectivity without the expense of permeability¹⁰ or improve both permeability and selectivity.¹¹ Besides, the two phases compensate for each phase’s limitations and improved processibility, mechanical strength, thermal and chemical stability also motivate further research on MMMs.¹² Previous studies have shown promising results by utilizing zeolites,^10,13 carbon nanotubes (CNTs),¹⁴ carbon molecular sieves (CMSs),¹⁵ activated carbon¹⁶ and metal- organic frameworks (MOFs)¹⁷ as the disperse phase.

Development of defect-free MMMs remains a challenging task as polymer–filler compatibility is often insufficient to form homogeneous interfaces.¹⁸ The polymer–filler incompatibility would provoke filler to repulse the continuous phase and form a “sieve-in-cage” morphology, which is responsible for unselective voids. The unselective voids are more preferable for penetrant to permeate across the membrane since it provides negligible mass transport resistance without discriminating penetrants. Hence, selection of the dispersed phase is a crucial factor to develop defect-free MMMs. Compared to other classes of fillers, MOFs have shown to have good interactions with the polymer matrix through its organic ligands. MOFs are crystalline compounds that consist of metal ions and secondary building unit (SBU) or organic ligands. Besides, large surface area, high micropore volume, various pore sizes, crystallinity and a high metal content has let MOFs emerge as spectacular porous materials for diverse applications.¹⁹ Zeolite Imidazole Framework-8 (ZIF-8), a product between Zn²⁺ with 2-methylimidazole (2-MeIM), is one of the most investigated MOFs. ZIF-8 properties are widely studied to show very good chemical stability against polar and nonpolar solvents,²⁰ reorientation of its structure at high pressure²¹ and high mechanical strength.²²

Nonetheless, MOF membranes suffer from low intrinsic CO₂/CH₄ selectivity and therefore they have been mainly studied as adsorption media.^23–25 In the case of ZIF-8, it offers intrinsic CO₂/CH₄ selectivity <5, which is significantly lower than other inorganic membranes such as zeolite T²⁶ and carbon membranes.²⁷ Hence, modifications are necessary to further improve its affinity towards CO₂ before implementing ZIF-8 as a membrane filler for CO₂/CH₄ separation. MOFs are known for their tunable properties and ease of modification. Modifications of MOFs are practiced to alter their pore structure, pore aperture, surface functional groups, and chemical stability, depending on specific applications.^28–31 Notably, post-synthetic modifications have been demonstrated as a straightforward modification approach with significant improvement of MOF properties. For example, metal ion doping into the MOF to provide open metal sites has significantly improved the gas adsorption capacity of the MOF. Metal ions such as potassium, sodium and lithium are used to provide additional interactions with the quadrupole moments of gases while enhancing molecular sieving through the restricted pore size of a modified MOF.^24,32 On the other hand, modifying the MOF through ligands also proved to be a convincing approach. Introducing functional groups such as pyridine and amine to the MOF structure would increase its affinity towards quadrupole moment gases such as CO₂ and H₂.^23,33 For example, Zhang et al.³⁴ demonstrated that ammonia impregnated ZIF-8 provides additional basic sites within the pores which as a result increased CO₂ uptake up to 50% as compared to virgin ZIF-8 without changes in the crystal integrity. In short, enhancing the tunability of MOFs would significantly improve their properties and their potential as fillers for MMMs and would give distinctive factors for gas separations.

Although ammonia-based modifications of ZIF-8 have been demonstrated to be an attractive approach to increase its affinity towards CO₂, the implementation of the modified particle as the dispersed phase in a MMM is yet to be investigated. This issue is crucial to be investigated since modifications may result in closed pores, which would significantly deteriorate the MMM performance. Herein, we are to report the preparation and characterization of an improved CO₂ selective MMM using modified ZIF-8 for CO₂/CH₄ separation. The ammonia modified ZIF-8 particles were prepared under different modification protocols and their properties were stringently studied before being dispersed into the polymer matrix. Polysulfone (PSf) was selected as the polymeric phase since it is an abundant and cheap material while providing well balanced between CO₂ permeability and CO₂/CH₄ selectivity. Asymmetric MMMs were fabricated using a dry/wet phase inversion method by incorporation of modified ZIF-8s under different modification protocols. The effect of the ZIF-8 embodiment on the overall membrane properties and the CO₂/CH₄ separation is evaluated.

2. Experimental

2.1 Materials

Zinc nitrate hexahydrate (Zn(NO₃)₂·6H₂O) was purchased from Alfa Aesar. Chemicals such as 2-methylimidazole (2-MeIM), n-hexane, polydimethylsiloxane (PDMS) and triethylamine (TEA) were obtained from Sigma Aldrich. Polysulfone (PSf Udel® P-1700) with a density of 1.24 g cm⁻³ was procured from Solvay Plastic. Ammonium hydroxide solution (25%), N,N-methylpyrrolidone (NMP), tetrahydrofuran (THF) and methanol were purchased from Merck. All chemicals were used without further purification.

2.2 ZIF-8 synthesis

ZIF-8 was prepared by following the same procedure as described in our previous work³⁵ with the ratio of Zn(NO₃)₂ : 2-MeIM : H₂O 1 : 6 : 500. Briefly, metal salt solution was prepared by dissolving 2 g of Zn(NO₃)₂·6H₂O (6.72 mmol) in 12.11 g of deionized water (20% of total deionized water). For the ligand solution, 2-MeIM (3.312 g, 40.34 mmol) was dissolved into 48.45 g of deionized water and subsequently 3.0 mL of TEA was added. The metal salt solution was gently added to the ligand solution, resulting in a cloudy solution mixture. The solution was stirred vigorously for 30 minutes, followed by centrifugation of the reaction product. The obtained product was washed several times with deionized water to remove excess reactants and subsequently dried in an oven at 60 °C for 12 hours. The collected powder was ground into fine particles before further drying in an oven at 100 °C for a minimum of 12 hours to further evacuate the guest molecules in the ZIF-8 pores.

2.3 ZIF-8 modification parameters

The synthesized ZIF-8s were indigenously modified as follows. Firstly, 1.0 g of the prepared ZIF-8 was uniformly dispersed into a predetermined volume of ammonium hydroxide solution (Table 1) with an additional 10 mL of deionized water. The suspension was then sonicated for 60 minutes to break the particle bulks before being vigorously stirred for 24 h at the specific temperatures. The sample was collected via centrifugation and washed with deionized water (3 times) before being dried in an oven (60 °C) overnight. The ZIF-8 modification parameters are listed in Table 1.

Table 1 ZIF-8 modification parameters

ZIF-8 samples	Ammonia solution (mL)	Temperature	MMM
Z-0	—	—	M0
Z25a	25	Ice bath (4 °C)	M25a
Z25b	25	Room temperature	M25b
Z25c	25	60 °C	M25c
Z50a	50	Ice bath (4 °C)	M50a
Z50b	50	Room temperature	M50b
Z50c	50	60 °C	M50c

---

2.4 Asymmetric flat sheet membrane preparation

An asymmetric flat sheet MMM was prepared from the solution that consisted of PSf (25 wt%), NMP (60 wt%), THF (15 wt%) and ZIF-8 (0.5 wt% of the total solids unless otherwise stated). ZIF-8 was dispersed into NMP/THF solvent and approximately 10% of the polymer was added under stirring for the purpose of priming. The remaining polymer was gradually added and the mixture was kept stirring until the solution became homogeneous. The casting dope was then hand-cast using a casting bar to a thickness of 120–190 μm. After standing on the glass plate for approximately 3 min, the cast film together with the glass plate were immersed in a coagulation bath (water at 27 °C) where the membrane was solidified. Then, the membrane was transferred to fresh water and kept there for 1 day to remove the residual solvent completely. Finally, the membrane was solvent-exchanged by immersing it progressively in methanol and n-hexane, each for 2 h, before being air-dried for 24 hours in the ambient atmosphere. The pristine PSf membrane was prepared by the same process without adding ZIF-8. The surface of the membrane was then brought into contact with 3 wt% PDMS/n-hexane solution for 10 min to seal possible pinholes before it was “cured” at 60 °C for 12 hours. Sample abbreviations are listed in Table 1.

2.5 Characterization

Attenuated total reflectance infrared spectroscopy (ATR-IR) analysis was conducted using a Perkin Elmer UATR (Single Reflection Diamond) for Spectrum Two to observe the functional groups of the modified ZIF-8s.

A differential scanning calorimeter (DSC) was used to determine the glass transition temperature (T_g) of the prepared membranes after incorporating the filler into polymer matrix using a Mettler Toledo DSC 822e. The membrane sample was cut into small pieces, weighed and placed into a pre-weighed aluminium crucible. Then, the sample was heated from 50 to 400 °C at a heating rate of 10 °C min⁻¹ in the first cycle to remove the thermal history. The sample was cooled from 400 to 30 °C at a rate of 10 °C min⁻¹. The same heating protocol was repeated in the next heating cycle. T_g of the sample was determined as the midpoint temperature of the transition region in the second heating cycle.

Thermogravimetric analysis (TGA) was used to characterize the thermal stability of the prepared samples. TGA records the weight changes of the sample when heated continuously. The sample was heated from 50 to 900 °C at a heating rate of 10 °C min⁻¹ under a nitrogen atmosphere with a nitrogen flow rate of 20 mL min⁻¹.

XRD analysis was used to confirm that the phase of the ZIF-8 was similar to that reported in the literature and to monitor the changes in ZIF-8 crystallinity after modification. X-Ray Diffraction (XRD) analysis using a Siemens D5000 Diffractometer is non-destructive analysis to identify the structure of the sample by measuring 2θ angles. The XRD will emit X-rays towards the sample and the X-rays are diffracted at different angles and intensities by CuKα radiation with a wavelength (λ) = 1.54 Å at room temperature.

The specific surface area and pore volume of the virgin ZIF-8 and modified ZIF-8 crystals were measured using a Micromeritics gas adsorption analyzer ASAP2010 instrument equipped with commercial software for calculation and analysis. The BET surface area was calculated from the adsorption isotherms using the standard Brunauer–Emmett–Teller (BET) equation. The total pore volume was evaluated by converting the adsorption amount at p/p₀ = 0.95 to a volume of liquid adsorbate. The mesopore volume was obtained using a BJH plot while the micropore volume was obtained using the t-plot method of Lippens and de Boer with the adsorption data.

Scanning electron microscopy (SEM) was used to observe the membrane structure and morphology. Membrane samples were fractured cryogenically in liquid nitrogen. The samples were coated with gold before they were imaged and photographed by employing a scanning electron microscope (TM3000, Hitachi) with a potential of 10 kV under magnifications ranging from 1000 to 20 000.

A transmission electron microscope (TEM) (JEOL, JSM-6701FJEOL 1230) was applied to observe the macrostructures of the ZIF-8. Samples were prepared by dispersing ZIF-8 powder in methanol. A drop of methanol was used for the dispersion of ZIF-8 on carbon-coated copper grids operating at 300 kV.

Gas permeation tests were performed using a constant pressure–variable volume system described elsewhere.³⁶ The membranes were placed into the permeation cell with an effective permeation area of 12.5 cm² and exposed to pure CH₄ or CO₂ gas. The feed pressure was 4 bar gauge and the temperature was 27 °C. The pressure-normalized flux (permeance) of gas i was calculated as follows:

where i represents the gas penetrant, V_i is the volume of gas permeated through the membrane (cm³, STP), A is the effective membrane area (cm²), t is the permeation time (s) and Δp is the transmembrane pressure drop (cmHg). The permeance is expressed in gas permeation units, GPU, as

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

The selectivity was obtained using eqn (2):

where α_i/j is the selectivity of gas penetrant i over gas penetrant j, P_i/l and P_j/l are the permeance of gas penetrants i and j, respectively.

3. Results and discussion

3.1 Characterization of virgin and modified ZIF-8

The crystallinity of the prepared ZIF-8 was identified via XRD (Fig. 1a). The presence of strong peaks implies a high crystallinity of the prepared ZIF-8 showing good agreement with previous experimental work³⁷ and the simulated pattern;²⁰ where the peaks at 2θ = 7.30, 10.35, 12.70, 14.80, 16.40 and 18.00° correspond to planes (110), (200), (211), (220), (310), and (222), respectively. Fig. 1b represents the TEM image of the ZIF-8. The prepared ZIF-8 possesses rhombic dodecahedron morphology; showing good agreement with the literature.³⁷ The particle size of the prepared samples was estimated from a series of TEM images to be around 133 nm.

Fig. 1 (a) XRD pattern of the virgin ZIF-8 as compared with the literature³⁷ and the simulated pattern,²⁰ and (b) morphology of the prepared ZIF-8.

The XRD patterns of ZIF-8s after modification are presented in Fig. 2. The pattern remains unchanged after modification, indicating a strong resistance towards alkaline solution regardless of modification protocols.^20,34 In the previous report on amine modification it was demonstrated that the XRD peak intensity decreased after modification presumably due to impregnation of the amine within the ZIF-8 pores causing destructive interference with the XRD peaks.³⁸ However, TGA analysis (see Fig. S1, ESI†) reveals that no weight loss occurred below 150 °C for the virgin and modified ZIF-8, which indicates ammonium hydroxide solution was not impregnated within the pores. Moreover, the XRD peaks were intensified after modification in this study, which further implies that ammonium hydroxide not only was absent within the ZIF-8 pores, but also caused the removal of guest molecules. It was also observed that there was no apparent difference between the peak width of the virgin and modified ZIF-8. This can postulate that the modification does not provide any significant changes in particle size of ZIF-8. A new reflection was also observed at 2θ = 10.96° in all the modified samples presumably due to cage reordering of the ZIF-8 structure.³⁹

Fig. 2 X-ray diffraction of the modified ZIF-8s.

N₂ adsorption analysis of the prepared ZIF-8s is presented in Table 2. The BET surface area of the prepared pure ZIF-8 is comparable with the literature values.^40,41 The increase in BET surface area of the modified ZIF-8 samples, except for Z50b and Z50c, is understandable since the removal of guest molecules increases pore availability as observed by the intensification of XRD peaks after modification. It was also observed that the total pore volume of ZIF-8 increased significantly, again except for Z50b and Z50c, after modification, indicating pore reopening⁴² and/or formation of new pores due to cage reordering.³⁹

Table 2 The pore textural properties of the virgin and modified ZIF-8

Sample	BET surface area (m^2 g^−1)	Mesopore volume (cm^3 g^−1)	Micropore volume (cm^3 g^−1)	Total pore volume (cm^3 g^−1)
Z0	1031.7	0.2427	0.2995	0.5422
Z25a	1210.7	0.9877	0.2576	1.2453
Z25b	1223.4	0.3141	0.3842	0.6983
Z25c	1250.5	0.1951	0.3852	0.5803
Z50a	1132.2	0.5559	0.2964	0.8523
Z50b	1019.5	0.2368	0.3113	0.5481
Z50c	966.2	0.4983	0.4805	0.9788

---

The types of pore formed/reopened during the modification were highly dependent on the modification temperature. At ice bath temperature, the availability of mesopores increased dramatically while the micropore accessibility was slightly decreased. As modification temperature increased, the decrease in mesopores was significant probably due to pore constriction or the changing of mesopores to micropores (see Fig. S2–S4, ESI†). It was also observed that the amount of ammonia solution used for modification plays an important role in the surface properties of ZIF-8. Introducing 25 mL ammonia solution into ZIF-8 prompted a higher micropore volume whilst constriction of the mesopores was less than with 50 mL ammonia solution. This behavior was presumably caused by the dilution of the ammonium hydroxide solution with 10 mL of deionized water; where a more diluted solution eases the dissolution of guest molecules to evacuate the pores and eases the occurrence of modifications. Hence, the total pore volume increased when a smaller amount of ammonia solution was added, with the exception of Z25c and Z50c due to pore constriction induced at a higher modification temperature.

The intensification of the XRD peaks and improved surface properties observed in this work after various modifications contradict with the previous report, where amine would reside within the pores to cause destructive interference on the XRD beam by the blocked pores.³⁴ Hence, further characterizations are necessary to confirm the modification by ammonium hydroxide. Fig. 3a shows the infrared spectrum of the virgin and modified ZIF-8s. Most of the spectra are related to the vibrations of the methylimidazole units and thus can be described based on the origin of the bonds. It was observed that the spectra of the samples are in good agreement with other studies.^20,34 The absorption bands between 3135 and 2929 cm⁻¹ can be attributed to the aromatic and the aliphatic C–H stretching of methylimidazole, respectively. The characteristic peak at 1584 cm⁻¹ was due to the C=N stretching mode, whereas bands between 1350–1500 cm⁻¹ can be assigned to the entire ring stretching.⁴³ The peak at 450 cm⁻¹ shows the distinct stretching vibration of Zn–N.

Fig. 3 FTIR analysis of the virgin and modified ZIF-8 with (a) overall spectrum, and spectra that represent (b) aliphatic C–N stretch (1250–1020 cm⁻¹), (c) N–H bend (1650–1580 cm⁻¹), and (d) N–H stretch (3400–3250 cm⁻¹).

There were no apparent differences after ZIF-8 was subjected to different modification protocols due to the presence of similar functional groups also in the virgin ZIF-8. Hence, the IR absorbance values of some specific functional groups were more thoroughly studied. The modification of the ZIF-8 has led to the intensification of the IR absorbance peaks coming from the C–N stretch (Fig. 3b), N–H bend (Fig. 3c), and N–H stretch (Fig. 3d), thus suggesting that ammonia modifications under various procedures were successful. It was also observed that the peak intensification was more prominent in the Z50-series compared to the Z25-series which was presumably related to the higher ammonium hydroxide content that provided more interactions with the ZIF-8 particles.

The characterization data presented that the ZIF-8 was successfully modified without ammonia solution being impregnated inside its pores. Liu et al.⁴⁴ simulated that in the ideal crystal structure of amine-modified ZIF-8, the amino group took its place near the C=C bond on the methylimidazole linker (Fig. 4) with a stable structure. The changes of the methylimdazole ligands, which were responsible for a pore opening of 3.4 Å in ZIF-8,⁴⁵ would significantly influence the textural properties as was observed during the N₂ adsorption analysis (see Table 2) and in the XRD patterns (see Fig. 2). It can be postulated that, during modification, the N–H functional group deprotonates the C=C in the imidazole linkers and leads to cage reordering while maintaining its overall structure. However, the simulation demonstrated that the surface area and pore volume should decrease as modification occurs which contradicts with this work. This contradiction is presumably caused by (1) further removal of the guest molecule and pore reopening after modification leading to the increasing surface area and pore volume, overwhelming the pore constriction by the new N–H group, and (2) randomized N–H group attachment (combination of H₂N–C=C–NH₂, H–C=C–NH₂ and/or none in one crystal unit) experienced in this work, whereas the simulation was based on the idealized structure of amine-modified ZIF-8. Nevertheless, this work shows that the amine-group can be introduced through a straightforward approach.

Fig. 4 Idealized crystal structure of amine-modified ZIF-8 (adapted from Liu et al.⁴⁴). The NH₂–ZIF-8 represents the –NH₂ attached on a single side of C=C, while (NH₂)₂–ZIF-8 represents the –NH₂ group attached on both sides of C=C in the imidazole linkers.

3.2 MMM characterization

The morphologies of the prepared membranes are presented in Fig. 5. The cross-sectional image of the neat PSf membrane shows an asymmetric structure, with an apparent active layer accompanied by a sponge-like substructure (Fig. 5a). The dry phase inversion was introduced to form a thin-selective layer through evaporation of a volatile solvent (THF). The sponge-like substructure was also observed as a result of the solvent exchange between NMP and water during the wet inversion.

Fig. 5 Cross-sectional membrane morphology of (a) neat PSf, and (b) M0.

A similar cross-sectional morphology was observed for all the prepared MMMs (see Fig. 5b, S5 and S6, ESI†) since small amounts of fillers were incorporated and a similar preparation protocol was used for fabrication. It should be noted that nanoparticles tend to form a bigger cluster due to interactions between their surfaces, especially when a large number of the particles are incorporated. However, no apparent ZIF-8 particles were observed within the membrane cross sectional morphology regardless of the modification procedure. In addition, there were no changes in the membrane functional groups as observed via IR analysis (see Fig. S7, ESI†) after filler (virgin and modified ZIF-8) incorporation. The results suggest that (1) sonication introduced to the dope preparation was able to break the particle cluster; (2) a small amount of ZIF-8 was enough to make significant changes in the membrane morphology and functional group; and (3) uniform distribution of the ZIF-8 particles throughout polymer matrix was possible.

The thermal stability of the prepared membranes was investigated (see Fig. S8, ESI†) and no significant difference between the thermal stability of the pristine PSf and MMMs was observed, which suggested that low ZIF-8 loading did not affect overall thermal stability of the prepared membranes. The T_g values of the prepared membranes are summarized in Table 3. The T_g of the neat PSf membrane is comparable to the value previously reported,⁴⁶ indicating a high rigidity of the materials. Incorporation of virgin and modified ZIF-8s into the PSf matrix provided notable changes in the membrane’s T_g. The membrane’s T_g decreased after the virgin ZIF-8 was embedded, indicating increased segmental mobility as the particles reside among the polymer chains. Incorporating modified ZIF-8s into the PSf matrix showed a further decrease in the membrane’s T_g. Hence, it can be postulated that better dispersion of the modified ZIF-8s was enabled compared to the virgin ZIF-8.

Table 3 Glass transition temperature of the prepared membranes

Sample	Glass transition temperature (Tg)
Neat PSf	185
M0	175
M25a	165
M25b	156
M25c	160
M50a	151
M50b	146
M50c	167

---

3.3 Gas permeation

3.3.1 Influence of virgin and modified ZIF-8s on membrane performance. Table 4 shows the CO₂ permeance and CO₂/CH₄ selectivity of the neat PSf membrane and MMMs. The CO₂ permeance and CO₂/CH₄ selectivity of the neat PSf are 21.3 ± 6.4 GPU and 19.8, respectively, both lower than the previously reported values due to the difference in materials and fabrication history.^36,47,48 Upon incorporation of 0.5 wt% (total solids) virgin ZIF-8 (M0), CO₂ permeance and CO₂/CH₄ selectivity increased to 29.2 GPU and 23.16, respectively. These phenomena are expected since the open metal sites in the ZIF-8 pores have good interactions with the quadrupole moment of CO₂, while none with CH₄.⁴⁹ Besides, ZIF-8 possesses pore openings of 3.4 Å, which introduces molecular sieving as a separation mechanism allowing CO₂ to migrate easily while hindering CH₄ permeation (kinetic diameter of CO₂ is 3.3 Å while CH₄ is 3.8 Å).⁵⁰ Consequently, a significant increase in the CO₂/CH₄ selectivity was observed even when the ZIF-8 loading was low.

Table 4 The results of membrane permeation studies with the neat PSf membrane, MMM with unmodified ZIF-8 and MMMs with modified ZIF-8

Sample	Permeance^a (GPU)				Ideal CO2/CH4 selectivity^b
	CO2	±^c	CH4	±^c
a GPU = 1 × 10^−6 cm^3 cm^−2 s^−1 cmHg^−1.b Ideal selectivity is based on average of CO2 and CH4 permeance.c ± represents the standard deviation.
Neat	21.27	6.35	1.08	1.33	19.83
M0	29.19	3.68	1.26	0.05	23.16
M25a	27.05	5.99	1.18	0.18	22.66
M25b	21.21	1.26	0.73	0.01	29.04
M25c	21.16	4.84	0.62	0.228	34.09
M50a	28.19	7.87	1.26	0.35	22.61
M50b	21.74	4.94	1.02	0.39	22.72
M50c	22.41	6.77	0.77	0.17	28.66

---

The data are reproduced in Fig. 6 for better interpretation of the effect of the temperature on ZIF-8 modification. The figure shows that, regardless of the modification temperature, all the modified ZIF-8 containing MMMs show CO₂ permeance lower than the virgin ZIF-8 containing MMM (M0), despite the enhancement of BET surface area and total pore volume after modification, as observed by the adsorption experiments. This decrease in the membrane permeance is not due to polymer chain rigidification, since the T_g of the MMM decreased when modified ZIF-8 was incorporated. Plausible explanations are (1) CO₂ entrapment within the filler due to stronger quadrupole-π electron interactions on the N–H group of the modified ZIF-8 (ref. 34 and 44) which hindered its diffusion; and (2) a decrease in mesopore contribution to the permeation path.

Fig. 6 Gas separation performance of the prepared membranes with (i) CO₂ permeance and (ii) ideal CO₂/CH₄ selectivity. The MMM containing virgin ZIF-8 is compared with modified ZIF-8 (a) under ice bath, (b) at room temperature, and (c) at 60 °C.

Regarding the M25 series, CO₂ permeance decreases progressively from virgin (M0) to ammonium modification at 4 °C (M25a) to 27 °C (M25b) and to 60 °C (M25c). It is recalled that the mesopore volume also decreased progressively from 0.9877 cm³ g⁻¹ of M25a to 0.1951 cm³ g⁻¹ of M25c (see Table 2), while micropore volume increased from 0.2576 cm³ g⁻¹ of M25a to 0.3852 cm³ g⁻¹ of M25c. Therefore, the decrease of CO₂ permeance from M25a to M25c is likely due to the decreased contribution of mesopores and increased contribution of micropores to the gas transport channel.

The decrease in CH₄ permeance from M25a to M25c seems more severe than the decrease in CO₂ permeance, since CO₂/CH₄ selectivity keeps increasing from M25a to M25c. This is also due to the increase in the micropore contribution from M25a to M25c. Considering the kinetic diameters of CO₂ (3.3 Å) and CH₄ (3.8 Å), and the size of the micropores that are likely to be originating from the 6-ring window aperture of 3.4 Å,²⁰ it is obvious that CH₄ transport through the micropores is prohibited while that for CO₂ is allowed. Hence, an increase in the micropore contribution leads to an increase in selectivity. The same trend is observed for the M50 series. However, unlike the M25 series, the selectivity does not change from M50a to M50b but increases significantly from M50b to M50c. This trend parallels the trend in the micropore volume M-50a (0.2964 cm³ g⁻¹) ≈ M50b (0.3113 cm³ g⁻¹) < M50c (0.4805 cm³ g⁻¹). Therefore, suppression of CH₄ transport by micropores becomes most effective for M50c.

Hence, the latter two effects, the enhanced interaction between CO₂ and CH₄ and the hindering effect on modified ZIF-8, have contributed to the significant increase up to 72% in ideal CO₂/CH₄ selectivity compared to the neat PSf membrane.

3.3.2 Influence of filler loading. An attempt to further improve the gas separation and permeation performance was implemented by increasing the filler loading. Z25c was chosen for the filler, since M25c demonstrated promising performance with higher selectivity and practically unchanged CO₂ permeance as compared to the neat PSf membrane (see Table 3 and Fig. 6).

Quite disappointingly, both CO₂ permeance and selectivity decreased as Z25c loading was increased from 0.5 wt% to 2.0 wt% as shown in Fig. 7. This is due to the increase in tortuosity of the permeation path in the presence of a large number of filler particles, which increases the mass transport resistance. In addition, severe Z25c agglomeration that might take place at higher loading limits the penetrant’s access to the ZIF-8 pores⁵⁰ and dwindles its selective features. Consequently, a simultaneous decrease in both CO₂ permeance and ideal CO₂/CH₄ selectivity was observed at higher Z25c loading. Hence, only 0.5 wt% (total solids) of Z25c is necessary to improve the membrane selectivity.

Fig. 7 Separation performance of PSf/Z25c at different filler loading.

3.4 Comparison with literature

MMMs in which ZIF-8 was incorporated in different polymers have been reported during the past years. Most of the polymer/ZIF-8 MMMs were prepared as dense membranes in earlier studies focusing on the changes in gas solubility and diffusivity with ZIF-8 incorporation and CO₂ permeability reported in Barrer. Whereas, asymmetric membranes were prepared in this study since it is more relevant to practical applications and CO₂ permeance is reported in GPU. To enable a performance comparison with other ZIF-8 incorporated MMMs, the selective layer thickness of the neat PSf and M25c was evaluated from the FESEM images, as shown in Fig. 8. The selective layer thicknesses for PSf and M25c were estimated to be around 0.238 and 0.341 μm, respectively, and the CO₂ permeabilities were calculated to be 5.06 and 7.26 Barrer, respectively.

Fig. 8 Skin layer thickness of (a) the neat PSf membrane, and (b) M25c.

The intrinsic CO₂ permeability and CO₂/CH₄ selectivity of PSf are reported to be 4.4 Barrer and 30.0, respectively.⁵¹ Compared to the reported values, the selectivity obtained in this study (19.8) is significantly lower and compensated by higher CO₂ permeability (5.06 Barrer). The performances of various polymer/ZIF-8 MMMs reported in the literature are tabulated in Table 5 and compared with M25c of this study. Most notable is the work by Ordoñez et al.⁵⁰ where 50 wt% (total solids) of ZIF-8 was incorporated in Matrimid®. They achieved CO₂/CH₄ selectivity of 124.89, which is 187% better than the neat Matrimid® membrane. This level of high ZIF-8 loading is typical of dense MMMs without suffering from defective polymer–filler interface and the dispersed ZIF-8 can be fully utilized to facilitate the separation performance. For this reason, the dense MMMs could surpass the Robeson upper bound limit.

Table 5 Comparison between the virgin ZIF-8-based MMM and this work

Polymer	Membrane configuration	Loading (wt% total solids)	% changes in CO2 permeability	% changes in CO2/CH4 selectivity	Reference
a Volumetric percent.
Matrimid	Dense	50	−54%	187%	50
PPEES	Dense	10	39%	29%	52
Matrimid	Dense	30	219%	15%	17
Matrimid	Dense	5	25%	11%	55
PIM-1	Dense	28^a	−3%	31%	53
PMPS	Dense	8.3	171%	−12%	54
PSf	Asymmetric	0.5	43%	72%	This work

---

In contrast, the fabrication of an asymmetric membrane does not allow high ZIF-8 loading. In this study, the loading of the modified ZIF-8 was as low as 0.5 wt%, which significantly limited the potential capacity of the ZIF-8 filler.^50,52–54 Thus, the MMM fabricated in this study does not stand out among the available data of ZIF-8 MMMs. Nevertheless, the performance of the M25c membrane is considered a remarkable improvement from the neat PSf membrane. The ammonia modification has improved surface area and micropore volume of ZIF-8 with the presence of the N–H functional group. The resulted MMM is more CO₂ permeable (7.26 Barrer) and CO₂ selective (34.09), than the neat PSf membrane despite its limited filler loading (0.5 wt%). These improvements are in good agreement with simulation⁴⁴ and experimental^34,42 works on CO₂ adsorption in amine-modified ZIF-8. Hence, the straightforward modification through ammonium solution on ZIF-8 proposed in this work will motivate research towards the future development of MOF-based MMMs for CO₂ removal.

4. Conclusion

ZIF-8s which underwent ammonia modification under various conditions exhibited significant changes in their properties, i.e. phase crystallinity, pore properties and BET surface area increased, while preserving the overall structure. The MMM with modified ZIF-8 has shown a substantial increase in ideal CO₂/CH₄ selectivity, especially for ZIF-8 modified under 25 mL ammonium hydroxide solution at 60 °C. The formation of new micropores while the disappearance of mesopores have hindered CH₄ permeation across the membrane. This, combined with the introduction of the N–H functional group within the ZIF-8 after modification resulted in significant improvements of 43% in CO₂ permeability and 72% in ideal CO₂/CH₄ selectivity with only minimal loading (0.5 wt%). Increasing the filler loading did not necessarily improve the MMM performances in CO₂/CH₄ separation due to particle agglomeration that restricts penetrant access to the pores. Hence, the ammonia modification proposed in this study has proven to be a straightforward approach to simultaneously increase CO₂ permeability and ideal CO₂/CH₄ selectivity, and would be beneficial for future works focused on gas separation MMM, especially for CO₂ separation.

Acknowledgements

The authors gratefully acknowledge the Ministry of Education (MOE) for the scholarship and Long-Term Research Grant Scheme (LRGS) Program under Universiti Teknologi Malaysia with the grant number Q.J130000.2452.04H71.

References

1. T. Cnop, D. Dortmundt and M. Schott, UOP Continued Development of Gas Separation Membranes Tech Paper, 2007.
2. M. Mofarahi, Y. Khojasteh, H. Khaledi and A. Farahnak, Energy, 2008, 33, 1311–1319.
3. M. Wang, A. Lawal, P. Stephenson, J. Sidders and C. Ramshaw, Chem. Eng. Res. Des., 2011, 89, 1609–1624.
4. T. E. Rufford, S. Smart, G. C. Y. Watson, B. F. Graham, J. Boxall, J. C. Diniz da Costa and E. F. May, J. Pet. Sci. Eng., 2012, 94–95, 123–154.
5. R. W. Baker, in Membrane Technology and Applications, John Wiley & Sons, Ltd, 2012, pp. 325–378.
6. R. W. Baker and B. T. Low, Macromolecules, 2014, 47, 6999–7013.
7. L. M. Robeson, J. Membr. Sci., 2008, 320, 390–400.
8. P. M. Budd and N. B. McKeown, Polym. Chem., 2010, 1, 63–68.
9. S. Kim, L. Chen, J. K. Johnson and E. Marand, J. Membr. Sci., 2007, 294, 147–158.
10. R. T. Adams, J. S. Lee, T.-H. Bae, J. K. Ward, J. R. Johnson, C. W. Jones, S. Nair and W. J. Koros, J. Membr. Sci., 2011, 367, 197–203.
11. K. Diíaz, L. Garrido, M. Loópez-Gonzaález, L. F. del Castillo and E. Riande, Macromolecules, 2010, 43, 316–325.
12. Y. Yampolskii, Macromolecules, 2012, 45, 3298–3311.
13. C. Casado-Coterillo, J. Soto, M. T. Jimaré, S. Valencia, A. Corma, C. Téllez and J. Coronas, Chem. Eng. Sci., 2012, 73, 116–122.
14. 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.
15. D. Q. Vu, W. J. Koros and S. J. Miller, J. Membr. Sci., 2003, 221, 233–239.
16. M. Anson, J. Marchese, E. Garis, N. Ochoa and C. Pagliero, J. Membr. Sci., 2004, 243, 19–28.
17. S. Basu, A. Cano-Odena and I. F. J. Vankelecom, Sep. Purif. Technol., 2011, 81, 31–40.
18. M. A. Aroon, A. F. Ismail, T. Matsuura and M. M. Montazer-Rahmati, Sep. Purif. Technol., 2010, 75, 229–242.
19. J. Yao and H. Wang, Chem. Soc. Rev., 2014, 43, 4470–4493.
20. K. S. Park, Z. Ni, A. P. Cote, J. Y. Choi, R. Huang, F. J. Uribe-Romo, H. K. Chae, M. O’Keeffe and O. M. Yaghi, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 10186–10191.
21. S. A. Moggach, T. D. Bennett and A. K. Cheetham, Angew. Chem., 2009, 48, 7087–7089.
22. J. C. Tan, T. D. Bennett and A. K. Cheetham, Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 9938–9943.
23. Y.-S. Bae, O. K. Farha, J. T. Hupp and R. Q. Snurr, J. Mater. Chem., 2009, 19, 2131.
24. Y.-S. Bae, B. G. Hauser, O. K. Farha, J. T. Hupp and R. Q. Snurr, Microporous Mesoporous Mater., 2011, 141, 231–235.
25. S. R. Caskey, A. G. Wong-Foy and A. J. Matzger, J. Am. Chem. Soc., 2008, 130, 10870–10871.
26. Z. Y. Yeo, S.-P. Chai, P. W. Zhu and A. R. Mohamed, Microporous Mesoporous Mater., 2014, 196, 79–88.
27. W. N. W. Salleh and A. F. Ismail, Sep. Purif. Technol., 2011, 80, 541–548.
28. J.-R. Li, Y. Ma, M. C. McCarthy, J. Sculley, J. Yu, H.-K. Jeong, P. B. Balbuena and H.-C. Zhou, Coord. Chem. Rev., 2011, 255, 1791–1823.
29. F.-J. Ma, S.-X. Liu, D.-D. Liang, G.-J. Ren, F. Wei, Y.-G. Chen and Z.-M. Su, J. Solid State Chem., 2011, 184, 3034–3039.
30. Q.-L. Zhu and Q. Xu, Chem. Soc. Rev., 2014, 43, 5468–5512.
31. Z. Wang and S. M. Cohen, Chem. Soc. Rev., 2009, 38, 1315–1329.
32. K. L. Mulfort and J. T. Hupp, Inorg. Chem., 2008, 47, 7936–7938.
33. S. Choi, T. Watanabe, T.-H. Bae, D. S. Sholl and C. W. Jones, J. Phys. Chem. Lett., 2012, 3, 1136–1141.
34. Z. Zhang, S. Xian, H. Xi, H. Wang and Z. Li, Chem. Eng. Sci., 2011, 66, 4878–4888.
35. N. A. H. M. Nordin, A. F. Ismail, A. Mustafa, P. S. Goh, D. Rana and T. Matsuura, RSC Adv., 2014, 4, 33292–33300.
36. A. F. Ismail and P. Y. Lai, Sep. Purif. Technol., 2003, 33, 127–143.
37. A. F. Gross, E. Sherman and J. J. Vajo, Dalton Trans., 2012, 41, 5458.
38. B. Arstad, H. Fjellvåg, K. Kongshaug, O. Swang and R. Blom, Adsorption, 2008, 14, 755–762.
39. F. Schröder, D. Esken, M. Cokoja, M. W. E. van den Berg, O. I. Lebedev, G. Van Tendeloo, B. Walaszek, G. Buntkowsky, H.-H. Limbach, B. Chaudret and R. A. Fischer, J. Am. Chem. Soc., 2008, 130, 6119–6130.
40. Y. Pan, Y. Liu, G. Zeng, L. Zhao and Z. Lai, Chem. Commun., 2011, 47, 2071–2073.
41. J. Cravillon, S. Münzer, S.-J. Lohmeier, A. Feldhoff, K. Huber and M. Wiebcke, Chem. Mater., 2009, 21, 1410–1412.
42. Z. Zhang, S. Xian, Q. Xia, H. Wang, Z. Li and J. Li, AIChE J., 2013, 59, 2195–2206.
43. Y. Hu, H. Kazemian, S. Rohani, Y. Huang and Y. Song, Chem. Commun., 2011, 47, 12694–12696.
44. D. Liu, Y. Wu, Q. Xia, Z. Li and H. Xi, Adsorption, 2013, 19, 25–37.
45. L. Hertäg, H. Bux, J. Caro, C. Chmelik, T. Remsungnen, M. Knauth and S. Fritzsche, J. Membr. Sci., 2011, 377, 36–41.
46. G. C. Kapantaidakis, S. P. Kaldis, X. S. Dabou and G. P. Sakellaropoulos, J. Membr. Sci., 1996, 110, 239–247.
47. B. Zornoza, A. Martinez-Joaristi, P. Serra-Crespo, C. Tellez, J. Coronas, J. Gascon and F. Kapteijn, Chem. Commun., 2011, 47, 9522–9524.
48. A. F. Ismail and W. Lorna, Sep. Purif. Technol., 2003, 30, 37–46.
49. R. Banerjee, H. Furukawa, D. Britt, C. Knobler, M. O’Keeffe and O. M. Yaghi, J. Am. Chem. Soc., 2009, 131, 3875–3877.
50. M. J. C. Ordoñez, K. J. Balkus, J. P. Ferraris and I. H. Musselman, J. Membr. Sci., 2010, 361, 28–37.
51. M. Mulder, Basic Principles of Membrane Technology, Kluwer Academic Publisher, Netherlands, 2nd edn, 1996.
52. K. Díaz, M. López-González, L. F. del Castillo and E. Riande, J. Membr. Sci., 2011, 383, 206–213.
53. A. F. Bushell, M. P. Attfield, C. R. Mason, P. M. Budd, Y. Yampolskii, L. Starannikova, A. Rebrov, F. Bazzarelli, P. Bernardo, J. Carolus Jansen, M. Lanč, K. Friess, V. Shantarovich, V. Gustov and V. Isaeva, J. Membr. Sci., 2013, 427, 48–62.
54. L. Diestel, X. L. Liu, Y. S. Li, W. S. Yang and J. Caro, Microporous Mesoporous Mater., 2014, 189, 210–215.
55. Q. Song, S. K. Nataraj, M. V. Roussenova, J. C. Tan, D. J. Hughes, W. Li, P. Bourgoin, M. A. Alam, A. K. Cheetham, S. A. Al-Muhtaseb and E. Sivaniah, Energy Environ. Sci., 2012, 5, 8359–8369.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra02230d

---