《1. Introduction》

1. Introduction

Gas separation is an essential industrial process. Compared with conventional separation processes such as distillation and absorption, membrane-based gas separation is a powerful approach that can be used in various applications to alleviate global environmental and energy crises [1–4]. The first commercialized membranebased gas separation system was installed in 1980 for hydrogen (H2) separation [5]. At present, membrane-based gas separation is diversified to include carbon dioxide (CO2) capture, natural gas sweetening, H2 production and purification, olefin/paraffin separation, and other petrochemical-related applications [6–10]. Diverse membrane materials have been developed to achieve outstanding separation performance, including inorganic membranes, twodimensional (2D) lamellar membranes, and polymeric membranes [11–20]. Polymeric membranes dominate the current gas separation market due to their low cost, easy processability, and high reproducibility [11,20]. However, the separation performance of polymeric membranes is usually constrained by the tradeoff relationship between permeability and selectivity; that is, more permeable polymers are usually less selective and vice versa (Fig. 1) [21–24]. This tradeoff is known as the ‘‘Robeson upper bound.”

《Fig. 1》

Fig. 1. Schematic representation of the tradeoff relationship between permeability and selectivity. Reproduced from Ref. [24] with permission.

The incorporation of porous fillers into a polymeric matrix has inspired the development of polymer/filler hybrid membranes— the so-called mixed-matrix membranes (MMMs) [25]. MMMs integrate the advantages of the flexibility and processability of polymers with the precision of porous fillers, resulting in high permeability and selectivity [25–28]. Although some MMMs exhibit superior separation abilities, their current performance is still far from the predicted values, since the rational matching of porous fillers and polymeric matrix is challenging. As a result, unfavorable morphologies can be generated in MMMs, including filler particle aggregation, nonselective interfacial voids, polymer chain rigidification, and filler pore blockage (Fig. 2) [13,26,27]. Aggregation of filler particles will produce nonselective voids or even macroscopic defects within the membrane, leading to decreased separation selectivity. Interfacial voids will form a bypass for gas molecules, resulting in negligible or even no separation selectivity. Polymer chain rigidification around fillers will create a barrier to gas transport due to lower polymer chain mobility, leading to decreased permeability, which is commonly encountered in rubbery polymer-based MMMs [29,30]. The pore-blocking of fillers by polymer chains or solvent molecules can turn porous fillers into impermeable particles, resulting in increased path tortuosity for gas transport and thus decreased permeability [13,31–33].

《Fig. 2》

Fig. 2. Schematic diagram of fabricated MMMs and the formation of unfavorable morphologies in MMMs. Reproduced from Ref. [27] with permission.

To mitigate these problems and fulfill the potential of MMMs, researchers have proposed numerous strategies for the rational matching of porous materials with polymeric matrixes to achieve improved interfacial morphology and compatibility [13,27,29]. Such approaches include adjusting the geometry and functionality of porous fillers and building connections between porous fillers and the polymeric matrix by forming electrostatic interactions, hydrogen bonds, coordination bonds, or even covalent bonds. The selected porous fillers include metal–organic frameworks (MOFs), covalent organic frameworks (COFs), porous aromatic frameworks, metal–organic cages (MOCs), and so forth [26,34–36]. MOFs, which possess tunable pore architectures, high stability, and diverse designability, are among the most studied porous fillers in MMMs and show excellent separation performance [26]. MOCs, which are discrete molecular compounds, are emerging porous fillers in this field [36]. Although they have not been studied extensively, their good solubility and processability endow MOCs with the potential capability to achieve homogeneous dispersion and good compatibility with polymeric matrixes at the molecular level, which could solve the long-standing interfacial issues of MMMs [36,37]. Considering the similarities between MOFs and MOCs, we provide a comprehensive overview of recent advances in MMMs containing MOFs and MOCs as porous fillers for gas separation. We discuss the strategies that have been developed for the fabrication of MOF-/MOC-based MMMs with enhanced interfacial compatibility and separation performance. Furthermore, we highlight the challenges and perspectives that may deserve consideration for the future development of MMMs. We hope that this review will inspire the design and selection of membrane materials and the construction of high-performance MMMs.

《2. Types of porous materials》

2. Types of porous materials

《2.1. Metal–organic frameworks》

2.1. Metal–organic frameworks

MOFs are hybrid porous materials composed of metal ions or clusters that are coordinated with organic linkers via coordination bonds to form extended crystalline structures [38–40]. Typical MOFs used in MMMs include zeolitic imidazolate frameworks (ZIFs), University of Oslo-66 (UiO-66) series, and Materials Institute Lavoisier (MIL) series (Fig. 3) [26,33]. ZIFs are composed of transition metal cations (e.g., Zn2+, Co2+) and anionic imidazolate linkers, with a structure resembling that of zeolites [41,42]. By changing the anionic imidazolate linkers, a series of ZIFs with different aperture sizes, such as ZIF-8 (0.34 nm) and ZIF-90 (0.35 nm), have been prepared and incorporated into polymeric matrixes for MMM fabrication [41,43–46]. Among them, ZIF-8, which is composed of Zn2+ and 2-methylimidazole linker, has been investigated in depth for membrane-based gas separation, especially for propylene/propane (C3H6/C3H8) separation [41,47,48]. UiO-66 is another popular porous filler that exhibits excellent stability in aqueous solutions and various organic solvents [38]. The original UiO-66, which is composed of terephthalic acid (BDC) linker, possesses octahedral and tetrahedral cavities [26,38]. By replacing the linker with other BDC derivatives, a series of UiO-66-type MOFs with different functional groups (e.g., –NH2, –OH, –SO3H) were obtained [49–54]. UiO-66-NH2, a type of amine-functionalized MOF, is particularly attractive owing to its inherent basic properties that afford good affinity toward acid gases and allow further modifications [29,38]. MIL-series MOFs exhibit interesting properties such as flexibility, high surface area, and high stability, introducing new possibilities when used as fillers in MMM fabrication [55–58]. For example, MIL-53 is known to have a ‘‘breathing” effect, which manifests as a structural transformation from narrow pores to large pores when adsorbing guest molecules [26,55]. MIL101(Cr), which has large mesoporous cavities (2.9 and 3.4 nm), possesses a high surface area, excellent hydrothermal stability, and a high CO2 uptake of 40 mol∙kg–1 (at 304 K and 5 MPa) [53,57,59]. 

《Fig. 3》

Fig. 3. Crystal structures of the representative MOFs used in MMMs.

《2.2. Metal–organic cages》

2.2. Metal–organic cages

MOCs, also referred to as metal–organic polyhedra (MOPs), are a class of discrete molecular assemblies formed by coordination bonds between metal nodes and organic ligands [60–64]. Unlike MOFs, MOCs do not form extended frameworks. When MOCs are dissolved, discrete MOC molecules can be present in solutions, since no bonds need to be broken [65]. Zirconium-based MOCs (Zr-MOCs), which are composed of trinuclear zirconium clusters and carboxylate ligands, have gained increasing scholarly attention due to their outstanding aqueous and chemical stability [64,66,67]. Bis(cyclopentadienyl)zirconium dichloride (Cp2ZrCl2) is the most commonly used zirconium (Zr)-based reagent in the synthesis of Zr-MOCs, wherein the cyclopentadiene (Cp) group at the vertices prevents the extended assembly of cages (Fig. 4(a)) [68]. By tuning the ligand type, researchers have successfully synthesized ZrMOCs with different functional groups (e.g., –NH2, –SO2, –SO3Na), which are expected to aid in the building of crosslinked networks and enhance the networks’ affinity to the gases of interest during separation [63,68–71]. In particular, amino-functionalized Zr-MOC (i.e., ZrT-1-NH2), which is synthesized from Cp2ZrCl2 and 2-aminoterephthalic acid (NH2-BDC), has been studied widely for the fabrication of membranes and extended architectures (Fig. 4(a)) [66,68,71,72]. The ionic character of Zr-MOC renders it soluble in organic solvent/water mixed systems, such as N,N-dimethylformamide (DMF)/water, methanol/water, acetone/water, and acetonitrile/water [66]

Copper-based MOCs (Cu-MOCs), which are composed of 12 dinuclear Cu(II) paddlewheel units and 24 bridging linkers, have a cub-octahedral structure (Fig. 4(b)) [63,73]. The first Cu-MOC was synthesized in 2001 using isophthalic acid (IPA) as the ligand [74]. Subsequently, a series of Cu-MOCs were successfully synthesized by functionalizing the 5-position of the IPA ligand [61,62,75]. The introduction of long alkyl chains or polyethylene glycol (PEG) chains on the ligand has proven to be an efficient way to increase the solubility of Cu-MOCs in commonly used organic solvents, such as tetrahydrofuran (THF), dichloromethane (DCM), and DMF [65,76–79]. Cu-MOCs usually have poor hydrolytic stability and thus suffer from decomposition in the presence of water, which restricts their application in water-related systems [73,77,80]. In contrast to Cu-MOCs, isostructural rhodium-based MOCs (RhMOCs) possess high hydrolytic stability due to the robust Rh–Rh bonds in the paddlewheel units (Rh–Rh bond energy: 16.5 kcal∙mol–1 , 1 kcal ≈ 4184 J) [81]. The activity of the axial Rh–Rh paddlewheels and the structural stability of Rh-MOCs allow researchers to modulate the solubility and functionality of RhMOCs by means of either post-assembly modification (PAM) or coordination with N-donor molecules, which provides opportunities for the construction of MOC-based functional materials with well-defined porosity [82,83].

《Fig. 4》

Fig. 4. (a) Crystal structure of the amino-functionalized Zr-MOC (ZrT-1-NH2) formed by 2-aminoterephthalic acid (NH2-BDC) and Cp2ZrCl2; (b) crystal structure of the hydroxyl-functionalized Cu-MOC formed by 5-hydroxyisophthalic acid and dinuclear Cu(II) paddlewheel units. (a) Reproduced from Ref. [68] with permission.

The design and selection of MOCs with excellent stability are necessary for MMM industrial applications. CO2 capture—one of the most important separation applications for MMMs—involves issues of water stability. MOCs with poor hydrolytic stability will undergo structural collapse and therefore lose their separation abilities when challenged with real flue gas. To address this issue, PAM is a useful strategy for overcoming the poor stability of MOCs constructed by labile metal ions, such as Cu2+ [63,84]. Decorating a hydrophobic shielding layer onto the outer surface of MOCs or crosslinking neighboring MOCs to construct stronger coordination bonding may help stabilize the porous structure, although this may sacrifice the solution processability of MOCs to some extent [85,86].

《3. Fabrication of MMMs》

3. Fabrication of MMMs

As the library of MOFs and MOCs grows, more advanced MMMs using MOFs or MOCs as porous fillers have been developed [26,34,36,87,88]. Appropriate combination methods for porous fillers and polymers are essential in realizing favorable interfacial compatibility and excellent separation performance [25,27]. Herein, we classify the strategies for fabricating MOF-/MOCbased MMMs into two categories: fabrication strategies using physical mixing; and fabrication strategies using covalent linkages. The former involves a simple mixture with only spontaneous or weak interactions (e.g., electrostatic interactions, coordination bonds, and hydrogen bonds) at the interface, while the latter involves covalent bonds between porous fillers and polymers.

《3.1. Conventional MMMs》

3.1. Conventional MMMs

3.1.1. MOF-based conventional MMMs

The physical mixing of MOFs with polymers is a versatile and efficient way to fabricate MMMs without the need for complicated reactions and fabrication processes. To achieve better dispersion of MOFs within the polymeric matrix and, accordingly, good separation performance, researchers usually focus on regulating MOF geometry and MOF functionality. Downsizing the MOF particle size to several tens of nanometers is beneficial to improve MOF particle dispersity [89,90]. For example, nano-sized UiO-66 (20–30 nm) exhibited improved dispersion in a polymer of intrinsic microporosity (PIM-1) matrix, resulting in a 40% increase in CO2/N2 selectivity [91]. Large agglomerates of nano-sized UiO-66 were not observed within the PIM-1 matrix, in contrast to micro-sized UiO-66-based MMMs. Nano-sized ZIF-67 (25–35 nm) was also prepared and incorporated into a PIM-1 matrix to fabricate MMMs with a 69% increase in CO2/CH4 selectivity, while micro-sized ZIF67-based MMMs only exhibited a 35% increase of selectivity [92]. Recently, He et al. [93] synthesized nano-sized ZIF-8 (~100 nm) in a PIM-1 matrix using in situ methods. Due to excellent interfacial compatibility, massive MOF loadings of up to 67.2 wt% were achieved. These ultrahigh MOF loadings effectively enhanced the CO2 solubility, as confirmed by complementary density functional theory (DFT) simulation, which was quite different from other published results in which MOFs generally promoted gas diffusion.

In addition to particle size, MOF morphology affects the separation performance of MMMs. For example, compared with a pristine Matrimid 5218 membrane, MIL-53 nanoparticle-based MMMs exhibited an improved CO2 permeability from 8.01 to 9.03 Barrer (1 Barrer = 3.35 × 10–16 mol∙m∙(m2 ∙s∙Pa)–1 ), while MIL-53 nanorod- and MIL-53 microneedle-based MMMs exhibited a decreased CO2 permeability of 7.52 Barrer [56]. The researchers indicated that these results might be due to better disruption of the polymer chains by the MIL-53 nanoparticles, producing more free volume in the polymeric matrix [56]. ZIF-8 with five different shapes was synthesized and incorporated into a polyethylene oxide (PEO) matrix to fabricate MMMs [94]. Gas permeability measurements showed that nanorod ZIF-8-based MMMs exhibited the best performance for C3H6/C3H8 separation (permeability PC3H6 = 16.6 Barrer, separation selectivity αC3H6/C3H8 = 9.2), probably due to the intensification of the molecular-sieving effect of the nanorod ZIF-8 framework [94]. MOFs with hollow interior structures are preferred in order to minimize the transport resistance of gas through the core of the fillers, while their molecularsieving shells afford high selectivity. For example, ZIF-67/PIM-1 MMMs with hollow ZIF-67 particles exhibited an improved CO2 permeability of 37%, compared with ZIF-67/PIM-1 MMMs with solid ZIF-67 particles [95].

MOF functionalization is a popular way to engineer the MOF– polymer interface. The construction of MOFs with rich functional groups (e.g., –NH2, –OH, –CN) via direct assembly or PAM is useful for fabricating high-performance MMMs and has been investigated extensively in the past decade [91,96–103]. Recently, Jiang et al. [104] prepared an imidazole-2-carboxyaldehyde-functionalized UiO-66-NH2 (i.e., UiO-66-NH2@ICA) via PAM. Compared with pristine UiO-66-NH2, the as-synthesized UiO-66-NH2@ICA showed a higher CO2 adsorption capacity due to the increased density of the CO2-philic nitrogen (N) atoms decorating the pore environment. Accordingly, the as-fabricated UiO-66-NH2@ICA/Matrimid 5218 MMM with 10 wt% of filler loading exhibited 40% increased CO2/CH4 selectivity compared with the UiO-66-NH2/Matrimid 5218 MMM. A novel ‘‘delayed linker addition” protocol was proposed by Hillman et al. [105] to prepare a hybrid ZIF-8 framework containing unsubstituted imidazolate linkers. The as-fabricated MMMs exhibited excellent C3H6 permeability (PC3H6 = 111.9 Barrer) and C3H6/C3H8 separation selectivity (αC3H6/C3H8 = 14.3) [105]. Molecular simulations further revealed that the improved permeability with increasing unsubstituted imidazolate linkers was due to the increasing number of larger and more flexible apertures within the hybrid ZIF-8 framework. 

The functionalization of MOF surfaces with a second porous filler in order to construct core–shell composite structures can tune the surface characteristics of MOFs with increased surface roughness, surface area, or affinity to the surrounding polymeric matrix [106,107]. Due to the wholly organic nature of COFs, Cheng et al. [108] prepared a MOF@COF core–shell filler (i.e., UiO-66- NH2@TpPa-1) by means of a two-step polymerization and crystallization process, in which the COF outer layers efficiently avoided the formation of nonselective voids (Fig. 5). The as-fabricated UiO-66-NH2@TpPa-1/polysulfone (PSf) MMM with 5 wt% of filler loading exhibited a CO2/CH4 selectivity of 46.7, which was significantly greater than that of the pristine PSf membrane and the UiO66-NH2/PSf MMM. MOF@MOF core–shell fillers have also been prepared and introduced into a polymeric matrix for MMM fabrication. Song et al. [107] utilized a layer-by-layer deposition method to prepare a UiO-66-NH2@ZIF-8 core–shell filler. Compared with the UiO-66-NH2/PSf MMM, the CO2/N2 selectivity of the asfabricated UiO-66-NH2@ZIF-8/PSf MMM increased to 39 due to the smaller pore size of the ZIF-8 shell, which enhanced the molecular sieving of the core–shell filler [107]. Other types of MOF@MOF core–shell fillers have also been successfully synthesized using ZIFs as a shell [109–111]. Recently, Wu et al. [112] reported the growth of a sub-20 nm uneven MOF-74 shell on a MOF surface (e.g., MOF808, UiO-66, UiO-66-NH2) (Fig. 6). Compared with the pristine 4,4' - (hexafluoroisopropylidene) diphthalic anhydride (6FDA)–durene membrane, the ethylene/ethane (C2H4/C2H6) selectivity of the asfabricated MOF-801@MOF-74/6FDA–durene MMM dramatically increased to 5.91, surpassing the 2013 Robeson upper bound. This outstanding separation performance stemmed from the highdensity open metal sites on the MOF-74 shell, which allowed the filler to coordinatively crosslink with the polymeric matrix, thereby ensuring good interfacial compatibility and enhanced C2H4/C2H6 selectivity [112]. A molecular dynamics simulation was carried out to characterize the interactions between MOF-74 and 6FDA–durene polymer. Subsequently, the researchers utilized similar chemistry to construct a MOF@nanocapsule core–shell filler, in which the open metal sites on the nanocapsule (i.e., PgC5Cu) coordinatively crosslinked with the polar functional groups on polymers, including PSf, PIM, and polyimide (PI), achieving improved dispersity [113]

《Fig. 5》

Fig. 5. Synthetic route of (a) UiO-66-NH2, (b) TpPa-1, and (c) UiO-66-NH2@TpPa-1 composite filler. Tp: triformylphloroglucinol; Pa-1: p-phenylenediamine. Reproduced from Ref. [108] with permission.

《Fig. 6》

Fig. 6. Schematic diagrams of (a) MOF@MOF-74 composite fillers and (b) gas transport through the pore channels of MOF@MOF-74 composite fillers. (c) CO2/CH4 separation performance of pristine ODPA–DAM, UiO-66-NH2(x)ODPA (x = 9 and 24), and UiO-66-NH2@Ni74(x)ODPA (x = 10 and 22) membranes. (d) C2H4/C2H6 separation performance of pristine 6FDA–durene, pristine 6FDA–DAM, MOF-801(x)durene (x = 9 and 22), MOF-801@Ni74(x)durene (x = 10, 16, and 26), MOF-801@Ni74(x)DAM (x = 8, 10, and 18), and NiMOF-74(x)durene (x = 2.4 and 18) membranes. The x in parentheses represents the weight percentages of the fillers in the MMMs. For each membrane, the light to dark symbol colors represent low to high filler loadings in the membrane, respectively. ODPA: 4,4' -oxidiphthalic anhydride; DAM: 2,4,6-trimethyl-1,3-phenylenediamine; 6FDA: 4,4' -(hexafluoroisopropylidene)diphthalic anhydride. Reproduced from Ref. [112] with permission.

MOF functionalization by means of macromolecules is another popular and efficient method for improving interfacial compatibility and separation performance. This is because the polymer nature of macromolecules presents the possibility of filling the gaps between the porous fillers and polymeric matrix. For example, ZIF-8 was coated with a polydopamine (PDA) layer to obtain ZIF8@PDA nanoparticles [114]. The formation of hydrogen bonds between ZIF-8@PDA and the PI matrix helped eliminate undesirable interfacial voids and enhance the gas separation selectivity of the MMMs while slightly sacrificing permeability. Directly grafting identical macromolecules that match the polymeric matrix is beneficial for engineering the interfacial interaction. For example, Wang et al. [115] reported the grafting of polyvinylamine (PVAm) on a MIL-101(Cr) surface. Improved interfacial compatibility was observed between the PVAm-modified MIL-101(Cr) particles and the PVAm matrix due to the formation of hydrogen bonds, which were also helpful for the fabrication of a defect-free ultrathin membrane [115]. Using a novel gravity-induced interface selfassembly technique, ultrathin MMMs (thickness: ~200 nm) with unobstructed gas transport channels through the selective layer were successfully fabricated, exhibiting a CO2 permeance of 823 gas permeation units (GPU) and a CO2/N2 separation selectivity of 242. By using PI as a matrix, 4,4' -oxidiphthalic anhydride (ODPA)– 2,4,6-trimethyl-1,3-phenylenediamine (DAM) PI oligomers were covalently grafted onto a UiO-66-NH2 surface to enhance MOF dispersity in the PI matrix [116]. The resulting MMMs showed excellent CO2/CH4 and CO2/N2 separation performance (PCO2 = 142 Barrer,  αCO2/CH4 = 43, αCO2/N2 = 27) and improved CO2 plasticization resistance (Fig. 7) [116]. Similar strategies have also been adopted by Qian et al. [117] and Dai et al. [118] to covalently graft PI oligomers onto the surface of UiO-66-NH2 and MIL-101(Cr), respectively. 

《Fig. 7》

Fig. 7. (a) Synthetic route of UiO-66-NH2@PI. (b) CO2/N2 and (c) CO2/CH4 separation performance of the pristine PI (opened square), UiO-66-NH2@PI/ODPA–DAM MMM (closed triangle), and UiO-66-NH2/ODPA–DAM MMM (closed circle) with different loadings of 5% (yellow symbols), 9% (orange symbols), 17% (red symbols), or 27% (black symbols). Reproduced from Ref. [116] with permission.

Polyethyleneimine (PEI), which has rich –NH2 groups on its backbone, is considered to be a good modifying agent for tuning MOF surface properties and interfacial compatibility. Xin et al. [57] utilized a vacuum-assisted method to prepare PEI-functionalized MIL-101(Cr), in which PEI brushes were attached to the internal and external surfaces of MIL-101(Cr) particles. Electrostatic interactions and hydrogen bonds formed between the –SO3H group on the sulfonated poly(ether ether ketone) (SPEEK) matrix and the –NH2 group on the PEI brushes [57]. Accordingly, the as-fabricated MIL101(Cr)@PEI/SPEEK MMM exhibited a 128.1% enhanced CO2/CH4 selectivity in comparison with the MIL-101(Cr)/SPEEK MMM. Using glutaraldehyde as a covalent cross-linker, PEI brushes have also been applied to graft on a Christian-Albrecht University (CAU)-1 surface [119]. Recently, Wang et al. [120] revealed the potential of grafting polymer brushes with various configurations in order to affect the membrane separation performance (Fig. 8). The grafting of branched PEI is helpful in enhancing themembrane selectivity,while the grafting of block copolymer polyether block amide (Pebax) is conducive to increasing the membrane permeance [120]. Other types of macromolecules, such as PEG, poly(ionic liquid), and poly(dimethylsiloxane) (PDMS), have also been adopted to functionalize MOF surfaces, resulting in improved interfacial compatibility and separation performance [121–123]. The gas separation performance of MOFbased conventional MMMs in recent studies is summarized in Table 1 [45,56,57,91–95,104,105,107,108,112,114–120,124].

《Fig. 8》

Fig. 8. (a) Synthetic route of Pebax-, PVAm-, and PEI-functionalized UiO-66; (b) schematic diagram of gas transport through the pore channels of Pebax-, PVAm-, and PEIfunctionalized UiO-66; (c) CO2/N2 separation performance of the fabricated MMMs. UKX, UKM, and UKI were species that synthesized via post-modification of UiO-66 by Pebax, PVAm, and PEI, respectively. KH560: 3-glycidyloxypropyltrimethoxysilane; mPSf: modified polysulfone; PPO: poly(phenylene oxide); MMP: metal-induced ordered microporous polymer; [Emim][BF4]: 1-ethyl-3 methylimidazolium tetrafluoroborate; GO: graphene oxide; BUPP: bridging UiO-66-NH2-poly(ethylene glycol) diglycidyl ether (PEGDE)-PVAm. Reproduced from Ref. [120] with permission.

《Table 1》

Table 1 Summary of the gas separation performance of MOF-based conventional MMMs.

XLPEO: crosslinked poly(ethylene oxide); TBDA2: 3,9-diamino-4,10-dimethyl-6,12-dihydro-5,11-methanodibenzo[b,f][1,5]diazocine; DDS: 4,4' -diaminodiphenyl sulfone; ICA: imidazolate-2-carboxyaldehyde; PIEM: poly(isocyanatoethyl methacrylate).

3.1.2. MOC-based conventional MMMs

Although the use of MOCs in MMMs is relatively new, MOCs are promising in demonstrating similar properties to MOFs in MMM fabrication. In particular, the solubility and processibility of MOCs enable them to be integrated into industrial membrane fabrication processes. In contrast to MOF particles, which form suspensions in solvent media, MOCs act as discrete molecules and disperse uniformly within a polymeric matrix at the molecular level. A key step in the fabrication of MOC-based MMMs is to match the solubility of the MOCs with that of the selected polymer. In this respect, CuMOCs have been greatly investigated for the fabrication of MOCbased MMMs due to their suitable pore size, easy functionality, and good solubility in common solvents.

By using PI as a matrix, a –NH2 functionalized Cu-MOC (i.e., MOP-15) was dissolved in dimethyl sulfoxide (DMSO) and then incorporated into a 6FDA–DAM PI matrix during MMM fabrication [125]. In contrast to MOF-based MMMs, which usually attain optimal performance at filler loadings of 10–20 wt%, the addition of just 1.6 wt% of MOC resulted in a 40% improvement in CO2 permeability, with good aging resistance. However, further increasing the MOC loading to 7.4 wt% led to a dramatically decreased selectivity due to the aggregation of MOCs at this loading. Focused ion beam scanning electron microscopy (FIB-SEM) further provided visualization of the MOC dispersion within the polymeric matrix. At a loading of 1.6 wt%, a homogeneous membrane with discrete MOC molecules (2–3 nm) embedded in the matrix could be clearly observed. To enhance the solubility of MOCs in commonly used low-boiling-point solvents, an alkyl chain-decorated Cu-MOC (i.e., MOP-18), which is highly soluble in chloroform (CHCl3), was incorporated into a Matrimid 5218 matrix for MMM fabrication [65]. Scanning electron microscopy (SEM) revealed that the MOP18 did not aggregate within the matrix even at loadings as high as 80 wt%, which is quite distinct from MOF-based MMMs. 

The incorporation of Cu-MOC has also been found to improve the anti-aging property of poly(1-trimethylsilyl-1)propyne (PTMSP) membranes, which have been known to be susceptible to such phenomena (Fig. 9) [126]. By comparing the anti-aging property of MMMs constructed using four types of Cu-MOCs (i.e., tert-butyl (tBu) MOP, diethylene glycol (DEG) MOP, triethylene glycol (TEG) MOP, and MOP-18) with different lengths of non-polar hydrocarbons or polar PEG chains on the ligand, the researchers concluded that the anti-aging property of the MMMs was mainly caused by the length—rather than the chemistry—of the MOC side chain [126].

《Fig. 9》

Fig. 9. (a) Structure of the PTMSP matrix; (b) schematic diagram of the pristine PTMSP membrane and the Cu-MOC-based MMM before and after aging. (c–f) Crystal structures of the selected Cu-MOCs: (c) tert-butyl (tBu) MOP; (d) diethylene glycol (DEG) MOP; (e) triethylene glycol (TEG) MOP; and (f) MOP-18. AIM: anti-aging intercalated membrane. Reproduced from Ref. [126] with permission.

PEO, a highly CO2-philic polymer, has also been applied as a matrix to fabricate MOC-based MMMs. Cu-MOC with polar –SO3Na groups on its side chain (i.e., MOP-3) was incorporated into a PEO matrix for CO2 separation [127]. Increasing the Cu-MOC loading was found to increase the CO2 permeability of the MMMs, although a slightly decreased selectivity was observed. Unlike the fabrication of dense MMMs, Sohail et al. [128] prepared ultrathin MOCbased MMMs via an atom transfer radical polymerization (ATRP)- based continuous assembly of polymers (CAP) technique. The presence of unsaturated metal sites and PEG chains on Cu-MOC facilitated CO2 transport within the PEO matrix, resulting in enhanced CO2 solubility and CO2 solubility selectivity, as confirmed by sorption analysis (Fig. 10) [128]. Accordingly, the ultrathin MMMs exhibited a high CO2 permeance of 448 GPU and a high CO2/N2 selectivity of 30. The thickness of the membrane was around 291 nm, and the thickness of the selective layer could be controlled to be less than 50 nm, highlighting the significance of the ATRPbased CAP technique for ultrathin MMM fabrication. MOCs with other topologies or metals have also been incorporated into polymeric matrixes for efficient gas separation [129]. The gas separation performance of MOC-based conventional MMMs in recent studies is summarized in Table 2 [65,125,127–130].

《Fig. 10》

Fig. 10. (a) Synthetic route, (b) schematic diagram, and (c) cross-sectional SEM image of the fabricated EG3-MOP/PEG9DMA/PDMS thin film composite MMM. PAN: polyacrylonitrile; EG3-MOP: triethylene oxide-modified metal–organic polyhedral; PEG9DMA: poly(ethylene glycol)dimethacrylate. Reproduced from Ref. [128] with permission.

《Table 2》

Table 2 Summary of the gas separation performance of MOC-based conventional MMMs.

PVDF: polyvinylidine fluoride; PolyPDXLA: poly(poly(1,3-dioxolane) acrylate).

《3.2. Covalently bonded MMMs》

3.2. Covalently bonded MMMs

3.2.1. MOF-based covalently bonded MMMs

Covalently bonding MOFs with polymers via facile reactions is considered to be a powerful method to reinforce the interfacial interactions in order to achieve defect-free membranes. PI and PIM are commonly studied glassy polymers used to form covalent linkages with MOFs. Yu et al. [101] reported the chemical crosslinking of PIM-1 and UiO-66-CN to fabricate UiO-66- CN@sPIM-1 MMMs via the thermal rearrangement of PIM-1 (sPIM-1) (Fig. 11). The as-fabricated MMMs showed a high CO2 permeability of 12 063 Barrer and a CO2/N2 selectivity of 53.5. The fluoride on PIM-1 can also react with the –OH group on MgMOF-74, forming an inter-connected micropore network [131]. Accordingly, Mg-MOF-74@PIM-1 MMM exhibited a simultaneous improvement of CO2 permeability to 21 269 Barrer and CO2/CH4 selectivity to 19.1 [131]. [Cd2L(H2O)]2·5H2O (Cd-6F) MOF, synthesized with a 6FDA ligand, was incorporated into a 6FDA–oxydianiline (ODA) PI matrix for MMM fabrication via in situ polymerization [132]. Improved interfacial compatibility and better separation performance were observed due to the interaction between the uncoordinated –COO group on Cd-6F and the –NH2 group on the 6FDA–ODA matrix. 

《Fig. 11》

Fig. 11. (a) Synthetic route to fabricate UiO-66-CN@sPIM-1 MMM; (b) schematic diagram of gas transport within UiO-66-CN@sPIM-1 MMM; (c) CO2/N2 separation performance of the fabricated UiO-66-CN@sPIM-1 MMM. Reproduced from Ref. [101] with permission.

Regarding MMM fabrication with rubbery polymers, Xu et al. [133] fabricated ultrathin MMMs by chemically crosslinking functionalized UiO-66-NH2 with a PVAm matrix. UiO-66-NH2 was first modified with poly(ethylene glycol) diglycidyl ether (PEGDE) to obtain epoxy-group-terminated MOF particles (i.e., PEG-UiO-66- NH2), which were further covalently bonded with a PVAm matrix via a facile epoxide–amine reaction (Fig. 12). The as-fabricated MMMs, which had a membrane thickness of 410 nm, exhibited a high CO2 permeance of 1295 GPU and a CO2/N2 selectivity of 91, which were better than those of UiO-66-NH2/PVAm MMM. By using the acrylate group on PEG precursors, crosslinked MMMs were fabricated via the copolymerization of PEG precursors with isopropenyl-functionalized UiO-66 or vinyl-functionalized Beijing University of Chemical Technology (BUCT) MOFs [134,135]. Crosslinked PDMS-based MMMs have also been successfully fabricated. Katayama et al. [136] prepared a modified UiO-66 via the PAM of hydride-terminated PDMS, which was further reacted with a PDMS matrix via a hydrosilylation reaction [136]. A defect-free MMM with 50 wt% of filler loading was obtained and exhibited good mechanical flexibility. Gao et al. [137] modified UiO-66-NH2 with cis-5-norbornene-exo-2,3-dicarboxylic anhydride (ND), which further participated in the ring-opening metathesis polymerization (ROMP) of norbornene for MMM fabrication. The as-fabricated MMM with 20 wt% of filler loading exhibited a significant improvement in mechanical toughness. A similar ROMP strategy was adopted by Hossain et al. [29] to form a covalent linkage between norbornene-modified UiO-66 (i.e., UiO-66-NB) and a PEG/polypropylene glycol (PPG)–PDMS copolymeric matrix (Fig. 13). The as-fabricated MMMs with 3 wt% of filler loading showed a CO2 permeability of 585 Barrer and a CO2/N2 selectivity of 53, approaching the 2019 Robeson upper bound [29]. An outstanding anti-plasticization property (up to 2.53 MPa) and a stable antiaging property (up to 11 months) were also observed. The gas separation performance of MOF-based covalently bonded MMMs in recent studies is summarized in Table 3 [29,101,131–133,135,137].

《Fig. 12》

Fig. 12. (a) Synthetic routes of PEG-UiO-66-NH2 nanoparticles; (b) the fabricated PEG-UiO-66-NH2/PVAm MMM. Reproduced from Ref. [133] with permission.

《Fig. 13》

Fig. 13. (a) Synthetic route of the fabricated UiO-66-NB-n@x(PEG/PPG–PDMS) MMM. (b, c) CO2/N2 separation performance of the fabricated MMMs at different (b) temperatures and (c) pressures (1 atm = 101 325 Pa). xMMM: cross-linked MMM. Reproduced from Ref. [29] with permission.

《Table 3》

Table 3 Summary of the gas separation performance of MOF-based covalently bonded MMMs.

MA: methacrylic anhydride.

3.2.2. MOC-based covalently bonded MMMs

As with MOF-based MMMs, using simple mixing strategies may not be sufficient to avoid MOC aggregation or phase separation during the fabrication of MOC-based MMMs. The judicious design of the intermolecular interactions between MOCs and polymers is conducive to realizing the full potential of MOCs. Although the covalent hybridization of MOCs with polymers has been investigated extensively as a means of constructing MOC/polymer hybrid materials, the fabrication of MOC-based MMMs using covalent bonding has rarely been reported, especially in the field of gas separation [24,60,63,138]. To achieve covalent bonding, MOCs must have functional groups that can react with monomers or polymeric precursors for membrane fabrication, along with adequate solubility for subsequent reaction and membrane fabrication processes. Recently, our group reported the fabrication of homoporous hybrid membranes (HHMs) by the chemical crosslinking of a polymerizable Zr-MOC (i.e., ZrT-1-AA) with PEO precursors via an ultraviolet (UV)-induced radical polymerization (Fig. 14) [139]. With only 1 wt% of MOC added, the obtained hybrid membrane exhibited an improved CO2 permeability without compromising CO2/CH4 selectivity. The addition of MOCs helped to increase CO2 solubility and diffusivity, as confirmed by sorption analysis. Since the membranes could be cured quickly under UV treatment, MOC aggregation could be avoided during the membrane fabrication process. The fluorescence technique provided evidence for the structural integrity and homogeneous distribution of the MOCs within the membrane architecture at the molecular level. This crosslinking method is expected to be applicable to other types of MOCs with different topologies and pore sizes, yielding defect-free membranes with high loadings. Considering the analogous crystal engineering of MOFs and MOCs, strategies adopted in constructing MOF-based covalently bonded MMMs may also be extended to fabricate MOC-based covalently bonded MMMs. For example, MOCs with –NH2 groups are expected to crosslink with isocyanate-terminated oligomers or epoxy groupterminated oligomers to fabricate crosslinked MMMs under mild conditions [133,140]. 

《Fig. 14》

Fig. 14. (a) Synthetic route of ZrT-1-AA. (b) Schematic diagram and (c) synthetic route of HHMs constructed by polymerizable ZrT-1-AA and PEO polymers. PEGMEA: poly(ethylene glycol) methyl ether acrylate; PEGDA: poly(ethylene glycol) diacrylate; Reproduced from Ref. [139] with permission.

《4. Challenges and perspectives》

4. Challenges and perspectives

《4.1. Extension of gas separation applications》

4.1. Extension of gas separation applications

Existing studies on MMMs mainly focus on CO2-related separations, including CO2/N2, CO2/CH4, and H2/CO2 separations. Future work is encouraged to extend applications to more challenging separations such as C2H4/C2H6 and C3H6/C3H8 separations [43,141,142]. In addition to focusing on commonly studied porous materials such as ZIF-8 and ZIF-67 for C3H6/C3H8 separation, it is recommended to propose novel strategies and exploit more advanced porous materials and membrane materials. The construction of favorable sorption sites for C2H4 or C3H6 in MOFs (e.g., defective MOF, MOF/ionic liquid composite) would be conducive to improving C2H4/C2H6 or C3H6/C3H8 separation selectivity [105,124]. The fabrication of ultrahigh MOF-based MMMs will help to enhance gas permeability and fully utilize the characteristics of porous materials.

《4.2. Fabrication of ultrathin MMMs》

4.2. Fabrication of ultrathin MMMs

Although dense MMM materials with inspiring separation performance have been greatly exploited, the fabrication of ultrathin MMMs (less than 1 lm thick) with high permeance and selectivity is highly desirable for large-scale industrial applications [7,143]. When fabricating ultrathin MMMs, the interfacial compatibility and other issues in dense MMMs may be magnified. For example, it has been demonstrated that ultrathin membranes may undergo an accelerated aging process in comparison with dense membranes [144]. Since the membrane becomes thinner, porous fillers with smaller particle sizes (i.e., less than 20 nm) will be conducive to eliminating undesirable interface issues and maintaining separation selectivity. In this case, discrete MOCs will be more promising than bulky MOF particles, since the particle size of MOCs is usually less than 10 nm. The development of advanced membrane fabrication techniques is also crucial for preparing defect-free ultrathin MMMs. Interfacial polymerization (IP), gravity-induced interface self-assembly, or ATRP-based CAP techniques may help in the fabrication of ultrathin MMMs [11].

《4.3. Exploitation of characterization methods》

4.3. Exploitation of characterization methods

With the increased complexity of fabricated MMMs, there is a need to improve characterization techniques in order to better understand and regulate the dispersity of porous fillers within polymeric matrixes. In general, it is relatively easy to characterize a crystalline MOF lattice within an amorphous polymer using traditional techniques, such as transmission electron microscopy (TEM) and powder X-ray diffraction (PXRD). However, it is not straightforward to characterize MOCs, due to their monodispersity and nanoscale particle size. In this case, FIB-SEM or focused ion beam transmission electron microscopy (FIB-TEM) affords a comprehensive visualization of the filler orientation and dispersion at the nanoscale level [125,126,145,146]. Fluorescence techniques can provide a three-dimensional (3D) resolution of membrane structures in order to demonstrate the structural integrity of MOCs within a matrix, although only a subset of MOCs exhibit intrinsic fluorescence [139,147]. In addition, the accessible porosity of fillers within a polymeric matrix cannot be directly proven by these methods. Positron annihilation lifetime spectroscopy (PALS) is a widely accepted technique for determining the pore size distribution, although it may not be available to all researchers. Exploring other auxiliary techniques (e.g., Raman spectroscopy) is necessary in order to gain more information on the pore size distribution. Understanding the interfacial interactions between porous fillers and a polymeric matrix is also necessary in order to guide material design and fully exploit the potential of each component. Solidstate nuclear magnetic resonance (ssNMR) is a powerful technique for unveiling molecular-level interactions between porous fillers and polymers but may be limited by a high concentration of polymer [31,135,148,149]. The synergistic application of characterizations and molecular simulations is highly desirable in order to elucidate the interfacial interactions within MMMs at an atomistic level [150].

《4.4. Study of the structure–performance relationship》

4.4. Study of the structure–performance relationship

Although extensive experiments have been carried out to evaluate the gas separation performance of MMMs, the structure– performance relationship between the intrinsic properties of MOFs or MOCs and the gas performance of their MMMs has not yet been fully understood, and it is time-consuming to investigate the tremendous expanse of combinations solely by experiments. To address this issue, a combination of computation (e.g., grand canonical Monte Carlo simulation) and experiments will be helpful in finding effective predictors of membrane performance [143,151,152]. In addition to computer simulation, the exploitation of machine learning, which is a class of statistical models that predicts properties based on a set of data, would provide a significant advantage to researchers in the assessment of the structure– performance relationship of membrane materials, since machine learning requires neither specialized equipment and experimental environments nor expensive computing clusters and supercomputers [153]. Machine learning has been utilized to discover novel polymers and predict the gas permeability of polymer membranes [153–155]. We believe that it is possible to extend machine learning to predict the separation performance of MMMs at an early stage of experimental activities and thereby help accelerate the design of membrane materials and membrane process engineering.

《5. Conclusions》

5. Conclusions

In conclusion, MMMs with MOFs and MOCs as porous materials have demonstrated prominent achievements in membrane-based gas separation. Although these two types of materials share many similarities, MOFs are more popular than MOCs in MMM fabrication to date, and the importance of MOCs in MMMs remains to be explored. Developing strong interactions between MOFs/MOCs and polymeric matrixes by forming electrostatic interactions, hydrogen bonds, coordination bonds, or covalent bonds is important to enhance the interfacial compatibility and improve the gas separation performance of the resulting MMMs. A deep understanding of the structure–performance relationship of MMMs, the exploitation of advanced characterization techniques and ultrathin MMMs, and the extension of applications to more challenging gas separations are also required in order to promote the widespread development of MMMs in the future.

《Acknowledgments》

Acknowledgments

This work was supported by the Ministry of Education, Singapore (MOE2019-T2-1-093 and MOE-T2EP10122-0002), the Energy Market Authority of Singapore (EMA-EP009-SEGC-020), the Agency for Science, Technology and Research (U2102d2004 and U2102d2012), and the National Research Foundation Singapore (NRF-CRP26-2021RS-0002).

《Compliance with ethics guidelines》

Compliance with ethics guidelines

Ziqi Yang, Zhongjie Wu, Shing Bo Peh, Yunpan Ying, Hao Yang, and Dan Zhao declare that they have no conflict of interest or financial conflicts to disclose.