Recent Advances in Polymer-Inorganic Mixed Matrix Membranes for CO2 Separation

Since the second industrial revolution, the use of fossil fuels has been powering the advance of human society. However, the surge in carbon dioxide (CO2) emissions has raised unsettling concerns about global warming and its consequences. Membrane separation technologies have emerged as one of the major carbon reduction approaches because they are less energy-intensive and more environmentally friendly compared to other separation techniques. Compared to pure polymeric membranes, mixed matrix membranes (MMMs) that encompass both a polymeric matrix and molecular sieving fillers have received tremendous attention, as they have the potential to combine the advantages of both polymers and molecular sieves, while cancelling out each other’s drawbacks. In this review, we will discuss recent advances in the development of MMMs for CO2 separation. We will discuss general mechanisms of CO2 separation in an MMM, and then compare the performances of MMMs that are based on zeolite, MOF, metal oxide nanoparticles and nanocarbons, with an emphasis on the materials’ preparation methods and their chemistries. As the field is advancing fast, we will particularly focus on examples from the last 5 years, in order to provide the most up-to-date overview in this area.


Introduction
Environmental and energy issues are two of the major challenges mankind are facing in the 21st century. In the past few decades, as the concentration of CO 2 in the atmosphere has continued to rise, the consequential effects of "global warming" and ecological destruction are threatening all living creatures on the Earth [1]. In 2021, the CO 2 concentration in the atmosphere reached a record high, approaching 420 ppm, and it is expected to continue to rise in the near future. As a result, it has been a global endeavor to limit CO 2 emission. In 2016, over 200 countries joined the United Nations to sign the Paris Agreement, indicating a consensus across the world about the critical need to reduce CO 2 emissions [2].
In general, there are three major CO 2 separation systems: (1) CO 2 /N 2 separation in flue gas stream, (2) CO 2 /H 2 separation in syngas, and (3) CO 2 /CH 4 separation in natural gas production [3,4]. In addition to those, the direct air capture of CO 2 from the atmosphere has also received a great amount of attention from both academia and industry [5]. Common CO 2 separation techniques include cryogenic separation, absorption, adsorption, and membrane separation [6]. Cryogenic separation is based on the condensation of gases with different boiling points, and involves phase change, which also entails a high consumption of energy and thus may not be suitable for large-scale deployment. Chemical absorption is another widely commercialized separation technique, which relies on using a basic solvent to neutralize the acidity of CO 2 by forming a weakly bonded intermediate. The absorption solvent can be further regenerated though heating (>100 • C). However, the absorption approach has several drawbacks, including equipment corrosion and degradation, high energy consumption, and potential environmental pollution [7]. Adsorption involves the However, the absorption approach has several drawbacks, including equipment corrosion and degradation, high energy consumption, and potential environmental pollution [7]. Adsorption involves the physical capture of a gas or liquid on the surface of solid porous adsorbents. The adsorbents can be reversibly regenerated though two processestemperature swing adsorption (TSA) and pressure swing adsorption (PSA). For TSA, the gas is attached to the adsorbent at high pressure and regenerated at low pressure. For PSA, the gas is attached to the adsorbent at low temperature and regenerated at an elevated temperature. However, both TSA and PSA also require significant amounts of energy and time to regenerate the adsorbents [8].
Membrane separation involves separating multiple types of gas molecules by taking advantage of the differences in permeability within the membrane under a given pressure. Although the commercialization of membrane separation only started in the 1970s, it has been almost two centuries since the discovery of the phenomenon of membrane separation. In 1829, Graham discovered the swelling of pig bladders in CO2 [9]. In 1831, Mitchell confirmed the difference between the diffusivity rates of H2 and CO2 in natural rubber [10]. In 1866, Graham proposed the solution-diffusion model, which explains the permeation of gas in polymers [11]. In the 1920s, Daynes invented a time lag method to measure the permeability coefficient [12]. In the 1930s, Barrer proposed dual mode sorption for gas permeation in glassy polymers [13]. Finally, in 1979, Monsanto commercialized a PDMS/PSf-based hollow fiber membrane under the name Prism ® , with high permeability and selectivity, for several applications, including N2/O2 separation, gas dehydration, bio gas upgrading, etc. In comparison to the other CO2 separation techniques, the membrane separation technique has the advantages of easy processing, low cost and low energy consumption. Recently, the area of membrane separation has attracted much attention.
According to the structures of the membranes, they can be classified into two categories-nonporous (dense) membranes and porous membranes [14]. The solution-diffusion model is used to describe the transport behavior of vapor and gas through a nonporous membrane, by which it is assumed that no pores exist, and different gas molecules are separated based on their solubility and diffusivity inside the membranes [15]. In a solution-diffusion process, first, the upstream gas from the inlet (Figure 1a) come into contact with and dissolves in the high-pressure face membrane ( Figure 1b). Then, the gas molecules diffuse through the membrane down a concentration gradient and desorb from the low-pressure face of the membrane (Figure 1c), where the diffusion is in a nonequilibrium state. Finally, as time goes on, eventually, the diffusion of gas reaches an equilibrium state (Figure 1d), and the one-dimensional gas flux (NA) across the membrane displays a linearity that can be described by Fick's law [16]: where DA is the diffusion coefficient of gas A inside the nonporous membrane, CA is the concentration of molecule A inside the membrane, and x is the thickness of the membrane, ranging from 0 to approximately l.  Under the equilibrium state, the permeability of gas A across the membrane can be described as: (2) where P A is the permeability of gas A, normally expressed in the unit of Barrer (1 Barrer = 1 × 10 −10 cm 3 (STP)cm/(cm 2 s cmHg)), p 2 is the gas pressure on the upstream side (feed), and p 1 is the gas pressure on the downstream side (permeate). According to Henry's law, the gas concentrations at the high-pressure face (upstream) and the lowpressure face (downstream) are C A1 and C A2 , and can be described as: where S A is the solubility of gas A in the membrane in an equilibrium state. Therefore, the permeability of gas A can be expressed as: When there are two types of gases-A and B-the selectivity of gas A over gas B can be expressed as: For glassy polymers, the major contribution to selectivity is made by the ratio of the diffusion coefficient, which is dependent on the molecular sizes of the penetrant. For rubbery polymers, the major contributor to the selectivity is the ratio of solubility (sorption coefficient), which is dependent on the condensability of the gases in the membranes, as well as the gas-polymer intermolecular interactions.
For highly condensable gases, the concentration of the gas inside the membrane can be described by the Flory-Huggins model: where p sat is the saturated vapor pressure under test temperature, χ is the Flurry-Huggins coefficient, and φ 2 is the volume fraction of gas dissolved inside the polymer. As the temperature changes, the solubility of light gases remains relatively stable, while the solubility of highly condensable gases undergoes greater fluctuation. The relationship between gas A's solubility inside the polymer and the temperature can be generalized as: and the diffusivity of the gas can be generalized as: where both S A0 and D A0 are constants, ∆H S is the adsorption enthalpy, and E D is the activation of gas diffusion [16].
On the other hand, porous inorganic membranes have continued to receive attention from both academia and industry due to their well-known thermal and chemical stability, and their much higher gas flux and/or selectivity. When penetrant gases pass through a porous membrane, the gas flux is mainly determined by the pore sizes, and the selectivity hinges on the difference between the kinetic diameters of the gases and the pore size. In general, the separation mechanism can be described by the four following mechanisms: (1) Knudsen diffusion-when the upstream pressure is much higher than the downstream pressure, the selectivity is the square root of the ratio of the molecular weights of the penetrants; (2) capillary condensation-when the mean free path of the gases is between the smallest and largest pore sizes, the gases can be treated as a viscous fluid in terms of their permeability, and this is also determined by Knudsen diffusion; (3) surface diffusionthe driving force of gas diffusion is the concentration gradient of the gas near the wall of the pores; (4) molecular sieving-the porous membranes will block gases with a kinetic diameter greater than the pore size, and will only allow the passing of a gas with a smaller kinetic diameter [14,17].
A typical composite membrane for gas separation is multilayered, whereby, on top of a microporous support layer (for mechanical stability), there are three layers of coating-a protective layer, a selective polymer layer and a gutter layer [18]. Commonly used porous support layers include polysulfone (PSf), polyacrylonitrile (PAN), cellulose acetate (CA), poletherimide (PEI), polyethersulfone (PES), polyvinylidene fluoride (PVDF), polyphenylene oxide (PPO), polypropylene (PP), polytetrafluoroethylene (PTFE), Teflon, etc. [19]. Both the gutter layer and the protective layer are highly permeable, and the desired gas selectivity is related to the selective polymer layer, which is usually 0.1 to 1 micron in thickness. In this paper, we will focus solely on the discussion of the selective polymer layer, especially in the context of the last 5 years. An ideal selective polymer layer should possess both high selectivity and permeability. However, membrane materials generally make a trade-off, with either high selectively/low permeability or vice versa. This phenomenon is exemplified by the Robeson-type trade-off plot, as shown in Figure 2, wherein the selectivity of gas pairs is logarithmically plotted against the permeability of the faster permeating gas. Tremendous effort has been made to create membrane materials that can exceed the Robeson limitation, as indicated in Figure 2. smallest and largest pore sizes, the gases can be treated as a viscous fluid in terms of their permeability, and this is also determined by Knudsen diffusion; (3) surface diffusion-the driving force of gas diffusion is the concentration gradient of the gas near the wall of the pores; (4) molecular sieving-the porous membranes will block gases with a kinetic diameter greater than the pore size, and will only allow the passing of a gas with a smaller kinetic diameter [14,17]. A typical composite membrane for gas separation is multilayered, whereby, on top of a microporous support layer (for mechanical stability), there are three layers of coating-a protective layer, a selective polymer layer and a gutter layer [18]. Commonly used porous support layers include polysulfone (PSf), polyacrylonitrile (PAN), cellulose acetate (CA), poletherimide (PEI), polyethersulfone (PES), polyvinylidene fluoride (PVDF), polyphenylene oxide (PPO), polypropylene (PP), polytetrafluoroethylene (PTFE), Teflon, etc. [19]. Both the gutter layer and the protective layer are highly permeable, and the desired gas selectivity is related to the selective polymer layer, which is usually 0.1 to 1 micron in thickness. In this paper, we will focus solely on the discussion of the selective polymer layer, especially in the context of the last 5 years. An ideal selective polymer layer should possess both high selectivity and permeability. However, membrane materials generally make a trade-off, with either high selectively/low permeability or vice versa. This phenomenon is exemplified by the Robeson-type trade-off plot, as shown in Figure 2, wherein the selectivity of gas pairs is logarithmically plotted against the permeability of the faster permeating gas. Tremendous effort has been made to create membrane materials that can exceed the Robeson limitation, as indicated in Figure 2. Pure polymer membranes are particularly susceptible to the permeability-selectivity trade-off relationship, as depicted by the Robeson upper bounds [21]. A higher permeability with high selectivity allows for the use of a less polymeric material, ultimately reducing the cost, while a high selectivity results in the higher purity of the treated gas. Depending on the specific application case, a suitable combination of permeability/selectivity needs to be applied. In general, a commercially useful membrane should be in the upper right area of Figure 2. Many inorganic membranes, such as zeolite, have already reached this area. However, inorganic membranes are often compromised by brittleness and their difficult processing. Considering the nature of both polymer membranes and inorganic membranes, mixed matrix membranes (MMMs) [22], which consist of a polymer matrix as the continuous phase and a porous inorganic material dispersed in the polymer matrix as a filler, are becoming particularly appealing, as they combine the advantages of flexibility and high permeability, while canceling out the disadvantages of the two types Pure polymer membranes are particularly susceptible to the permeability-selectivity trade-off relationship, as depicted by the Robeson upper bounds [21]. A higher permeability with high selectivity allows for the use of a less polymeric material, ultimately reducing the cost, while a high selectivity results in the higher purity of the treated gas. Depending on the specific application case, a suitable combination of permeability/selectivity needs to be applied. In general, a commercially useful membrane should be in the upper right area of Figure 2. Many inorganic membranes, such as zeolite, have already reached this area. However, inorganic membranes are often compromised by brittleness and their difficult processing. Considering the nature of both polymer membranes and inorganic membranes, mixed matrix membranes (MMMs) [22], which consist of a polymer matrix as the continuous phase and a porous inorganic material dispersed in the polymer matrix as a filler, are becoming particularly appealing, as they combine the advantages of flexibility and high permeability, while canceling out the disadvantages of the two types of membranes, such as the trade-off of selectivity and permeability in polymer membranes and the brittleness in inorganic membranes.
Depending on the nature (porosity) of the inorganic fillers, there are two types of MMMs. When the inorganic filler is porous, the resulting MMMs can have drastically improved selectivity against smaller gases through the size exclusion effect [23], or against easily condensable gases by increasing the affinity of their inorganic nanopores towards the gas molecules [24]. If the inorganic filler is nonporous, such as nonporous oxide nanoparticles, the membrane's free volume can be increased, especially for semi-crystalline polymers, such that the permeability of the membrane can be increased [25]. For MMMs with low inorganic loading, the effective permeability (P eff ) can be described by the Maxwell model: where P c is the permeability of the polymer (continuous) phase, P d is the permeability of the inorganic (dispersed) phase, and φ d is the volume fraction of the inorganic phase, which can be calculated using the following equation: where ρ c is the density of the polymer phase and ρ d is the density of the inorganic phase. w c and w d are the weight fraction of the polymer and the inorganic filler. For porous fillers whose permeability is significantly higher than that of the polymer phase (P d P c ), the effective permeability can be simplified as: For nonporous fillers whose permeability is significantly lower than that of the polymer phase (P d P c ), the permeability can be simplified as However, when the inorganic loading exceeds a certain threshold (normally~30 vol%), interconnection between the inorganic filler particles starts to occur. Under such circumstances, the MMMs act similarly to microporous membranes, with their inorganics connecting to form nanochannels [26]. Not only can the Maxwell model be used to estimate/predict the permeability of the MMMs with known P d , P c and φ d (to help in understanding the structure-property relationship) [17], it can also be used to study the permeability and diffusivity of the inorganic fillers, which may be highly valuable when developing effective inorganic fillers. For example, Liu et al. developed an MMM based on 6FDA-DAM polyimide and a submicron-sized rare-earth (RE) MOF with a face-centered cubic (fcu) topology (RE-fcu-MOF). Such an MMM demonstrated exceptionally enhanced separation performance for the removal of CO 2 and H 2 S from natural gas and the separation of butane isomers. In order to measure the permeability of different gases in the inorganic filler, the P eff of the MMM and the P c of the pure polymer phase were first experimentally measured. With known φ d , the P d values of RE-fcu-MOF were calculated. Furthermore, with the solubility of gas in the MOF filler obtained via the measured sorption isotherm, gas diffusivities in RE-fcu-MOF, including for He, H 2 , CO 2 , O 2 , N 2 , CH 4 , C 3 H 6 , n-C 4 , i-C 4 and SF 6 , can be obtained [27][28][29].

Choice of Polymers and Inorganics
The appropriate combination of polymer phase and inorganic phase, as well as preparation methods, are key to the performance of the resulting MMM. In general, there are four potential scenarios when mixing a polymer with inorganic fillers. Ideally, when the fillers are embedded in the polymer matrix without generating extra free volume and interaction with the polymer chains, the performance of the MMMs depends solely on the respective properties of the two phases, and can be described by the Maxwell model (Case 1 in Figure 3). However, especially for glassy polymers, the addition of fillers can generate extra free volume, as shown in case 2 of Figure 3, due to poor interfacial adhesion. As gases always prefer to move through the space of least resistance, permeability can increase at the cost of selectivity, as a result of "leaky interface". Moreover, sometimes the addition of inorganic fillers can cause the rigidification of the surrounding polymers, thus decreasing the permeability with a non-significant increase in selectivity, as indicated in case 3. On the other hand, the polymer chains can either partially or completely block the pores in the inorganic fillers. Partially blocked pores may have a size exclusion effect, while completely blocked pores can turn the porous filler into nonporous filler [17,30].
Polymers 2021, 13, 2539 6 of 33 fillers are embedded in the polymer matrix without generating extra free volume and interaction with the polymer chains, the performance of the MMMs depends solely on the respective properties of the two phases, and can be described by the Maxwell model (Case 1 in Figure 3). However, especially for glassy polymers, the addition of fillers can generate extra free volume, as shown in case 2 of Figure 3, due to poor interfacial adhesion. As gases always prefer to move through the space of least resistance, permeability can increase at the cost of selectivity, as a result of "leaky interface". Moreover, sometimes the addition of inorganic fillers can cause the rigidification of the surrounding polymers, thus decreasing the permeability with a non-significant increase in selectivity, as indicated in case 3. On the other hand, the polymer chains can either partially or completely block the pores in the inorganic fillers. Partially blocked pores may have a size exclusion effect, while completely blocked pores can turn the porous filler into nonporous filler [17,30]. CO2/N2 separation is often required for the treatment of flue gas from fossil fuel-fired power plants. Depending on the fuel source, flue gas often contains hot impurities, in addition to CO2 (10-15%) and N2 (78-80%), such as O2, CO, H2O, SO2, NOx, HCl, and Hg complexes, as well as considerable amounts of particles [31]. CO2/CH4 separation is employed mostly in the production of natural gas, the consumption of which has been increasing every year. Depending on the reservoir sources, raw natural gas contains various amount of CO2, hydrocarbons, H2S, N2, He, etc. [32]. Interest in CO2/H2 separation is relatively sparse compared to CO2/N2 and CO2/CH4 separation. The production of H2 is achieved through steam methane reforming (SMR), followed by water-gas shift (WGS), which leads to the combination of H2 and significant amounts of CO2 [33]. There are two types of CO2/H2 separation membranes-CO2-selective membrane, which rely on the difference of solubility in the membrane, and H2-selective membranes, which rely on the diffusivity of the gases in the membrane [19]. In order to meet the requirement of CO2 separation in different scenarios, the choice of polymer phase needs to satisfy several criteria, including (1) high gas permeance or permeability to reduce the membrane area; (2) high selectivity of CO2 over other gases so that the amount of gas to be circulated can be reduced; (3) high thermal and chemical stability due to the often harsh environment in CO 2 /N 2 separation is often required for the treatment of flue gas from fossil fuel-fired power plants. Depending on the fuel source, flue gas often contains hot impurities, in addition to CO 2 (10-15%) and N 2 (78-80%), such as O 2 , CO, H 2 O, SO 2 , NO x , HCl, and Hg complexes, as well as considerable amounts of particles [31]. CO 2 /CH 4 separation is employed mostly in the production of natural gas, the consumption of which has been increasing every year. Depending on the reservoir sources, raw natural gas contains various amount of CO 2 , hydrocarbons, H 2 S, N 2 , He, etc. [32]. Interest in CO 2 /H 2 separation is relatively sparse compared to CO 2 /N 2 and CO 2 /CH 4 separation. The production of H 2 is achieved through steam methane reforming (SMR), followed by water-gas shift (WGS), which leads to the combination of H 2 and significant amounts of CO 2 [33]. There are two types of CO 2 /H 2 separation membranes-CO 2 -selective membrane, which rely on the difference of solubility in the membrane, and H 2 -selective membranes, which rely on the diffusivity of the gases in the membrane [19]. In order to meet the requirement of CO 2 separation in different scenarios, the choice of polymer phase needs to satisfy several criteria, including (1) high gas permeance or permeability to reduce the membrane area; (2) high selectivity of CO 2 over other gases so that the amount of gas to be circulated can be reduced; (3) high thermal and chemical stability due to the often harsh environment in which it will be used, such as high temperatures and corrosive acid impurities; (4) resistance to CO 2 plasticization and aging to prevent the worsening of the separation efficiency over the long term; (5) low-cost and easy manufacturing for large-scale production [19].

Polymer/Zeolite Mixed Matrix Membranes
There are several major types of inorganic fillers that have been widely used to prepare MMMs: (1) Zeolite-a class of microporous, aluminosilicate minerals commonly used as commercial adsorbents for separation and as catalysts due to their high surface area and chemical stability [52,53]; (2) metal organic framework (MOF)-a class of polymers composed of metal ions or clusters coordinating as organic ligands to form highly porous structures with a high surface area, low density and structural tunability [54,55]; (3) zeolite imidazolate framework (ZIF)-a subclass of MOF materials, composed of metal ions connected by imidazole with tetrahedral structures, that have flexible structures and adjustable pore sizes and better thermal, chemical and moisture stability than other types of MOFs [56]; (4) oxide nanoparticles, such as non-porous silica or metal oxide nanoparticles, mesoporous silica, polyoligosilsesquioxanes (POSS), etc. [57][58][59][60]; (5) Barrer), 1030% (56.5) and 3633% (56) compared to the neat membrane [65]. Jusoh et al. reported an MMM with 1 wt.% zeolite T embedded in 6FDA-durene polyimide. The composite membrane showed a CO 2 permeability of 843. 6 Barrer and an ideal CO 2 /CH 4 selectivity of 19.1, which values are 80% and 172% higher than those of the neat membrane, along with enhanced resistance to plasticization up to 20 Bar [66]. Ahmad et al. studied the effect of adding 25 wt.% zeolite 4A to polyvinyl acetate (PVAc), and found that the addition of zeolite 4A increased the CO 2 /N 2 selectivity by 75% (100.54), although with a reduction in permeance [67]. Adams et al. studied the high loading of zeolite 4A (up to 50 vol.%) onto PVAc, and found that even with such high inorganic loading and at high pressure (440 psi), an improved selectivity (63%) for CO 2 /CH 4 could be observed [68]. Zhao et al. reported a type of MMM based on Matrimid ® 5218 polyimide and a partially lithiated Na-ZSM-25 zeolite (LNZ-25). Due to the complete blocking of CH 4 and the high admission of CO 2 achieved by LNZ-25, the MMM with 5 wt.% zeolite loading showed a high CO 2 /CH 4 selectivity of 169 at 5 bar, as well as plasticization resistance behavior [69]. Zhang et al. reported a tertiary-component MMM based on doping Pebax polymer with PEG-600 and NaY zeolite. The MMMs with 20 wt.% PEG and 30 wt.% NaY zeolite showed CO 2 /N 2 selectivity of 107.9 and CO 2 permeability of 172.6 Barrer. Furthermore, with the CO 2 /N 2 mixed gas, the designed membrane could directly enrich the concentration of CO 2 from 15% to 96.7% in one step [70]. Li et al. recently reported a carbon molecular sieve-type membrane based on a combination of hierarchically porous zeolite 5A and a thermally carbonized Matrimid ® polyimide polymer. The hierarchical zeolite 5A with both micropores and mesopores can facilitate CO 2 diffusion and reduce transport resistance at the same time; the thermal carbonization further improved free volume and pore aperture. This carbon molecular sieve-type MMM showed 450 Barrer of CO 2 permeability and 19.3 Barrer of CO 2 /CH 4 selectivity [71].

Polymer/Zeolite Mixed Matrix Membranes
In order to enhance the compatibility between the zeolite and the polymer matrix, zeolite modification or the adding of a third phase agent were often employed. Mashhadikahn et al. functionalized the zeolite 13X with aminosilane, which reacted with the hydroxyl groups on the zeolite particles. The functionalized zeolite particles, with an average pore size of 2.57 nm, were mixed with 6FDA-durene polyimide and fabricated into a membrane though solution-casting. With 15 wt.% loading of the aminosilanized zeolite, the polyimidebased MMM showed a high CO 2 permeability of 887 Barrer and a CO 2 /N 2 selectivity of 25.3 Barrer at a feed pressure of 0.2 MPa at room temperature [72]. Similarly, Chen et al. modified EMC-2 zeolite with 3-aminopropyltriethoxysilane (APTES) as the amino agent and APTMDS (bis(3-aminopropyl)-tetramethyldisiloxane) as the crosslinker [64]. The crosslinked MMM with 6FDA-ODA at 25 wt.% zeolite showed excellent CO 2 /CH 4 separation at 150 psi and 35 • C, with a CO 2 permeability of 40.9 Barrer and a CO 2 /CH 4 selectivity of 80.2 Barrer [73]. Amooghin et al. created defect-free Matrimid ® 5218/NaY zeolite by using 3-aminopropyl (diethoxy)methylsilane (APDEMS) as the crosslinking agent ( Figure 4a) [65]. The obtained results revealed that at 15 wt.% of filler loading, the CO 2 permeability increased from 8. 34 Barrer for the neat membrane to 9.70 Barrer, and the CO 2 /CH 4 selectivity was considerably increased from 36.3 to 57.1 Barrer (about 57%) due to the improved dispersion of the inorganic fillers [74]. Khan et al. developed a coupling approach by using APTMS to crosslink zeolite 3A with arylate-terminated PSf, as shown in Figure 4b [66]. The crosslinked PSf/zeolite 3A MMM with 40 wt.% zeolite loading exhibited a H 2 permeance of 15.1 GPU and H 2 /CO 2 selectivity of 7.12 Barrer, which is an approximately 5-fold increase compared to the neat membrane [75]. A similar approach was used to prepare the crosslinked PSf/MCM-41 MMMs, which showed high ideal selectivities for CO 2 /N 2 and CO 2 /CH 4 of 32.97 and 31. 48 Barrer, respectively [76]. ers 2021, 13, 2539 10 of 33 In another approach to enhancing the compatibility between zeolite and the polymer matrix, Ahmad [70]. It was evident that with the addition of IL, the interfacial voids were reduced, with a consequentially improved CO2 selectivity against both N2 and H2 [78]. Singh et al. reported an MMM containing poly(ionic) liquid (PIL) crosslinked with divinylbenzene, room temperature ionic liquid (RTIL) and SAPO-34 zeolite particles [71]. Such a system showed high solubility for CO2 and high mechanical stability, and thus a CO2/CH4 selectivity of 90 and CO2 permeability of 26 Barrer [79]. Dunn et al. evaluated the zeolite-PIL-RTIL MMM under a binary feed of up to 40 bar at 50 °C [68]. It was found that the SAPO-34-PIL-RTIL MMM with an inorganic loading of 30 wt.% showed a CO2/CH4 selectivity of 43, with a CO2 permeability of 202 Barrer, under an equimolar binary feed [80].

Polymer/MOF Mixed Matrix Membranes
MOFs represent a class of highly ordered crystalline porous materials [81,82]. MOFs are considered promising molecular sieves for various separation applications due to their exceptionally high surface area, tunable pore/channel sizes, and tailorable functionalities [82]. MOF-based mixed matrix membranes (MOF-MMMs) are developed as a compromise to overcome these practical challenges in membrane fabrication and utilization [83]. Promisingly, thanks to the better chemical compatibility between MOFs and polymers, the interfacial issue is less severe in MOF-MMMs than in zeolite-MMMs [20,83]. In this regard, numerous MOF-MMMs have been reported in recent years. The remarkable progresses in, and remaining challenges for, MOF-MMMs have been extensively summarized in several insightful reviews [84][85][86][87]. In this miniature review, we mainly focus on the most commonly studied MOFs involved in the development of MOF-MMMs for CO2-   [70]. It was evident that with the addition of IL, the interfacial voids were reduced, with a consequentially improved CO 2 selectivity against both N 2 and H 2 [78]. Singh et al. reported an MMM containing poly(ionic) liquid (PIL) crosslinked with divinylbenzene, room temperature ionic liquid (RTIL) and SAPO-34 zeolite particles [71]. Such a system showed high solubility for CO 2 and high mechanical stability, and thus a CO 2 /CH 4 selectivity of 90 and CO 2 permeability of 26 Barrer [79]. Dunn et al. evaluated the zeolite-PIL-RTIL MMM under a binary feed of up to 40 bar at 50 • C [68]. It was found that the SAPO-34-PIL-RTIL MMM with an inorganic loading of 30 wt.% showed a CO 2 /CH 4 selectivity of 43, with a CO 2 permeability of 202 Barrer, under an equimolar binary feed [80].

Polymer/MOF Mixed Matrix Membranes
MOFs represent a class of highly ordered crystalline porous materials [81,82]. MOFs are considered promising molecular sieves for various separation applications due to their exceptionally high surface area, tunable pore/channel sizes, and tailorable functionalities [82]. MOF-based mixed matrix membranes (MOF-MMMs) are developed as a compromise to overcome these practical challenges in membrane fabrication and utilization [83]. Promisingly, thanks to the better chemical compatibility between MOFs and polymers, the interfacial issue is less severe in MOF-MMMs than in zeolite-MMMs [20,83]. In this regard, numerous MOF-MMMs have been reported in recent years. The remarkable progresses in, and remaining challenges for, MOF-MMMs have been extensively summarized in several insightful reviews [84][85][86][87]. In this miniature review, we mainly focus on the most commonly studied MOFs involved in the development of MOF-MMMs for CO 2 -related separation, and discuss several practical factors affecting membrane separation performance. Some examples of MOF structures are summarized in Figure 5.  Indeed, among the thousands of MOFs reported, only a small portion have been utilized as efficient fillers for CO2/H2, CO2/N2 and CO2/CH4 separations. This is probably due to the similar physical and chemical properties among the gas molecules, requiring the precise control of MOF structures to achieve optimal molecular sieving affinity. Nevertheless, in the past 5 years, x-UiO-66 (x = functional group), x-MIL-53, and ZIF-8 have received extensive interest as the three most studied MOF fillers for CO2-related separations, based on our analysis of accessible publication records, accounting for over half of the total. On the one hand, the former two kinds of MOFs, i.e., NH2-UiO-66 and NH2-MIL-53, capture CO2 from mixtures as a result of favorable CO2-host interaction affinities, which can be further rationally tuned by the highly amendable functionality (x) of the linkers. On the other hand, ZIF-8 separates gas molecules through its ideal pore aperture (~3.4 Å), allowing target molecules to pass through with relative ease, but hindering other Indeed, among the thousands of MOFs reported, only a small portion have been utilized as efficient fillers for CO 2 /H 2 , CO 2 /N 2 and CO 2 /CH 4 separations. This is probably due to the similar physical and chemical properties among the gas molecules, requiring the precise control of MOF structures to achieve optimal molecular sieving affinity. Nevertheless, in the past 5 years, x-UiO-66 (x = functional group), x-MIL-53, and ZIF-8 have received extensive interest as the three most studied MOF fillers for CO 2 -related separations, based on our analysis of accessible publication records, accounting for over half of the total. On the one hand, the former two kinds of MOFs, i.e., NH 2 -UiO-66 and NH 2 -MIL-53, capture CO 2 from mixtures as a result of favorable CO 2 -host interaction affinities, which can be further rationally tuned by the highly amendable functionality (x) of the linkers. On the other hand, ZIF-8 separates gas molecules through its ideal pore aperture (~3.4 Å), allowing target molecules to pass through with relative ease, but hindering other molecules. Table 2 listed some of the recent examples of MOF/polymer MMMs. Urban and co-workers examined high NH 2 -UiO-66 loadings (~50 wt.%) in polysulfone (PSf), and observed the evolution from single-mode to dual transport regimes [88]. A new percolative pathway though the MOF is formed, acting as a molecular highway for gases when the MOF loading increases above 30 wt.%. As a result, CO 2 permeability dramatically increased from 18 to 46 Barrer, while the CO 2 /CH 4 selectivity was well maintained (~24-26). Zhu and co-workers functionalized the NH 2 -UiO-66 structure by converting the -NH 2 into -CN groups, which covalently linked with the PIM-1 via thermal reaction. The UiO-66-CN@PIM-1 membranes showed exceptionally high CO 2 permeability (>15,433 barrer) and high CO 2 /N 2 selectivity (>24) [89].
Friess and coworkers incorporated NH 2 -UiO-66 into a polyethylenimine (PEI) matrix with an MOF loading of 18 wt.%, and observed a CO 2 permeability of 394 Barrer with a CO 2 /H 2 selectivity of~12 [90]. Interestingly, Xiang and coworkers combined polyvinylpyrrolidone-modified PA-UiO-66 particles (PVP + MOF) with poly(acrylic acid) (PAA) through H-bonding-assisted layer-by-layer assembly, and achieved a H 2 /CO 2 selectivity of~20, indicating the highly amendable nature of mixed matrix membranes for separating the mixture of H 2 and CO 2 [91].
Chen and coworkers fabricated Pebax-based composite hollow fiber membranes by incorporating NH 2 -UiO-66 into the thin Pebax selective layer, and observed excellent CO 2 permeance (~325 GPU) and CO 2 /N 2 selectivity (~56), as well as CO 2 /CH 4 selectivity (~20), with a loading of 50 wt.%. In addition, the authors found that the presence of UiO-66 in Pebax significantly improved the membrane's operational stability under high pressure [92].
Rodrigue and coworkers fabricated MMMs using NH 2 -MIL-53(Al) particles as fillers and rubbery Pebax as the matrix, via a solution casting method [93]. The authors achieved enhanced CO 2 permeability, as well as CO 2 /N 2 and CO 2 /CH 4 selectivity, by introducing the MOF. The highest ideal selectivity values for CO 2 /CH 4 and CO2 2 /N 2 separations are 23.3 and 59.4 Barrer, respectively, with 10 wt.% MOF loading, while the CO 2 permeability was as high as 149 Barrer. Fong and coworkers optimized the spinning parameters of NH 2 -MIL-53(Al)/CA hollow fiber MMMs for gas separations [94]. The authors concluded that the optimization of spinning parameters can be considered a feasible and efficient method to fabricate hollow-fiber MOF-MMMs with improved separation performance.
Li and coworkers tailored the interfacial interaction between ZIF-8 and the polymer matrix, and observed an exceptional CO 2 /CH 4 selectivity of 58 Barrer with a CO 2 permeability of 20 Barrer [95]. The authors revealed that the Zn 2+ post-modification of the polymer matrix enables enhanced polymer/ZIF-8 interaction, while controlling the Zn 2+ treatment conditions can further tailor the CO 2 permeability and CO 2 /CH 4 selectivity.
Wey and coworkers used thermally annealed ZIF-8 nanoparticles as fillers in poly(styreneco-butadiene) (SBC) to improve the permselectivity of MMMs [96]. The thermal annealing treatment created polar atoms by breaking the Zn-N bonds in the ZIF-8 framework, and these are beneficial in promoting both CO 2 permeability and CO 2 /N 2 selectivity. The fabricated MMMs with a ZIF-8 loading of 20 wt.% exhibit a CO 2 permeability of~40 Barrer, with a CO 2 /N 2 selectivity of~22 Barrer. More examples of ZIF-MMMs will be discussed in Section 5.
Several other MOFs have also been widely studied in MMM development. For instance, Guiver and coworkers constructed ultrathin unobstructed gas transport channels in MMMs by using the gravity-induced interface self-assembly of poly(vinylamine) and polymer-modified MIL-101(Cr) [97]. Promisingly, the MMMs achieved a high CO 2 permeance of 283 GPU (gas permeation units), and a CO 2 /N 2 selectivity of 242 Barrer at 0.5 MPa.
Koros and coworkers investigated the CO 2 /CH 4 separation performance of a Y-fumfcu-MOF/6FDA-DAM MMM over a wide range of temperatures from 233 K to 338 K [98]. The authors revealed that temperature can significantly affect the membrane performance due to the different roles that sorption and diffusion factors played in the separation process. Remarkably, decreasing the temperature from 308 K to 233 K affords an extremely high CO 2 /CH 2 selectivity (~130), with an associated CO 2 permeability above 1000 Barrer. Besides the selection of MOF fillers and polymer matrices, the interfacial compatibility between the MOF and the polymer is another key consideration to realizing the full potential of MOF-MMMs for gas separation. Li and coworkers reported a new approach using polyimide brushes to covalently graft the MOF surface to the engineer interface, which resulted in a stand-alone membrane with 88 wt.% MOF loading [99]. The same group further reported a dual-interfacial engineering approach using a sub-20 nm polycrystalline MOF-74 shell as a transition phase to engineer the MOF-polymer interface [100]. Following this novel approach, the NH 2 -UiO-66-based MMMs manifested a simultaneous increase in CO 2 /CH 4 selectivity and CO 2 permeability with increases in MOF loading.
In summary, while there is as yet no commercialized MOF-based MMM, they still hold huge potential for the gas separation industry. Of the polymers adopted as the matrix, the commercial polymer Pebax has received major attention. In addition, 6FDA-based polyimide is another hot candidate for use as a polymer matrix. On the other hand, ZIF and UiO-66 are the most popular MOF materials.

Polymer/ZIF Mixed Matrix Membranes
ZIFs, a subfamily of MOFs, possess a zeolite topology with tunable pore structures. ZIFs consist of point metal nodes in a center connected to imidazolate linkers, and can adopt numerous three-dimensional structures ( Figure 6) [101]. Before now, 15 ZIF types have been found among 239 known types, and the relationship between imidazolates and topologies has been described in detail in the literature [102].
Some ZIF materials, such as ZIF-8, ZIF-11, ZIF-67, ZIF-71, ZIF-90, ZIF-95 and ZIF-301, have large cavities with narrow pore apertures close to the kinetic diameter of CO2 (0.33 nm), and this allows the inclusion of a suitable filler material in sole ZIF-based MMMs for CO2 separation. Table 3 compiles the research studies carried out in the past 5 years on the fabrication of sole ZIF-based MMMs.
Among the ZIF series, ZIF-8, with a diameter of ca. 1.1 nm and 0.34 nm windows, is frequently studied for MMM separation performance [125,126,[128][129][130][131][132][133]136,139]. Two major challenges ((1) high ZIF-8 loading and (2) the fabrication of ultrathin (~100 nm) ZIF-8based MMMs) need to be overcome in order to commercialize ZIF-8-based membranes Figure 6. The crystals of ZIFs. The largest cage in each ZIF is shown with the ZnN 4 tetrahedra in blue and the CoN 4 in pink. The yellow ball is included in the structure for clarity and to indicate space in the cage. Reprinted with permission from [101], copyright 2008 American Association for the Advancement of Science.
Some ZIF materials, such as ZIF-8, ZIF-11, ZIF-67, ZIF-71, ZIF-90, ZIF-95 and ZIF-301, have large cavities with narrow pore apertures close to the kinetic diameter of CO 2 (0.33 nm), and this allows the inclusion of a suitable filler material in sole ZIF-based MMMs for CO 2 separation. Table 3 compiles the research studies carried out in the past 5 years on the fabrication of sole ZIF-based MMMs. Among the ZIF series, ZIF-8, with a diameter of ca. 1.1 nm and 0.34 nm windows, is frequently studied for MMM separation performance [125,126,[128][129][130][131][132][133]136,139]. Two major challenges ((1) high ZIF-8 loading and (2) the fabrication of ultrathin (~100 nm) ZIF-8-based MMMs) need to be overcome in order to commercialize ZIF-8-based membranes [117]. ZIF-11 is a promising ZIF, with structural features (connected via apertures with a diameter of 0.3 nm) that make it appropriate for H 2 or CO 2 separation. Ehsani et al. utilized ZIF-11 as the filler and Pebax 2533 as the membrane matrix to fabricate ZIF-11/Pebax 2533 rubbery MMMs [137]. The resulting membranes exhibited the highest CO 2 permeability (402.9 Barrer) and selectivity for CO 2 /CH 4 (12.5), H 2 /N 2 (12.9) and H 2 /CH 4 (4.8) at 50-70 wt.% loading. ZIF-67 has also attracted considerable attention due to its natural family of MOFs. An et al. [139] first produced ZIF-67/6FDA-DAM MMMs for C 3 H 6 /C 3 H 8 separation, and Meshkat et al. successfully incorporated ZIF-67 in a Pebax 1657 rubbery membrane to obtain an MMM [130]. Such a membrane not only improved gas permeability (130% increase in CO 2 permeability), but also enhanced membrane separation performance (58% increase in CO 2 /CH 2 selectivity), compared to a pristine Pebax membrane at a feed pressure of 11 bar and at 35 • C. ZIF-95 with large voids (24.0 Å) is also an extremely important material in gas adsorption and purification applications. In the work of Ilicak et al., ZIF-95 was used as filler in the preparation of polyimide-based MMMs [135]. ZIF-95/Matrimid 5218 (70/30) MMMs achieved good CO 2 separation performance (CO 2 /CH 2 selectivity of 58 and H 2 /CH 2 selectivity of 192.0). These results indicate filler-polymer compatibility, and the possibility of forming additional channels and enhancing the free Matrimid volume.
In order to improve the interfacial compatibility between ZIFs and polymers in MMMs, and thus membrane separation performance, many methods of modifying and functionalizing ZIFs so as to build covalent bonds between ZIFs and polymers in the membrane matrix have been explored. Table 4 compiles the research studies carried out in the past 5 years on the fabrication of modified and/or hybrid ZIF-based MMMs for gas separation. The strategy of ZIF modification not only greatly broadens the scope of choice for polymer matrices and fillers; it also further enhances the membrane separation performance due to the extra functional groups introduced to overcome the problems of weak polymer-filler interfacial interactions [118]. ZIF-7 is another promising member of the ZIF family for CO 2 separation. Gao et al. incorporated functionalized OH-ZIF-7 filler into a Pebax 2533 matrix [156]. The resulting OH-ZIF-7/Pebax MMMs showed high CO 2 permeability of 273 Barrer and a CO 2 /N 2 separation factor of 38. In the work of Wang et al., a thin and uniform polydopamin (PD) layer of well-controlled thickness was coated on the surface of ZIF-8 nanocrystals via a selfpolymerization process of dopamine, in order to improve the interfacial interaction of ZIF-8 nanocrystals and polyimide with a Tröger's base functional group (TBDA2-6FDA-PI) [124]. Such MMMs achieved excellent gas permeation performances and comprehensive gas separation performances, surpassing the Robeson upper bound for gas pairs of H 2 /N 2 and H 2 /CH 4 , and approaching the upper bound for O 2 /N 2 gas pairs. Several researchers fabricated robust MMMs using amino-functionalized ZIFs (e.g., NH 2 -ZIF-7 and NH 2 -ZIF-8) with rubbery polymers (e.g., XLPEG, Pebax) [142,158] and glassy polymers (e.g., 6FDA-PI, PSF, polyamide) [128,143,156]. These amino groups enhanced filler affinity for CO 2 , resulting in significantly improved separation performances (CO 2 /N 2 and CO 2 /CH 4 selectivities) in the resultant MMMs, compared to the neat polymer under the same testing conditions. In order to improve gas molecule transportation in the membrane matrix between ZIFs and polymers, ZIF hybrids-a combination of ZIFs with other inorganic fillers (e.g., SWCNTs, MWCNTs, GO, SiO 2 , etc.)-were developed and used as fillers for gas separation [125,126,[145][146][147][148][149][150]152,153,157,159]. Li et al. developed novel MMMs by incorporating ZIF-8, in situ-inserted by MWCNTs, into a Pebax 1657 membrane matrix [150]. The functionalized MWCNTs inserted into the ZIF-8 particles not only prevented the aggregation of ZIF-8 particles, but also enhanced the membrane permeation properties (CO 2 permeability of 186 Barrer and CO 2 /N 2 selectivity of 61.3 at feed pressure of 0.5 MPa and 35 • C) and membrane mechanical strength. Huang et al. prepared a ZIF-8@GO hybrid using a facile in situ growth method, and accordingly fabricated ZIF-8@GO/Matrimid MMMs [153]. Such MMMs containing 20 wt.% ZIF-8@Go hybrid exhibit the highest selectivity of up to 65 for CO 2 /N 2 , with a CO 2 permeability of 238 Barrer, which surpasses the Robeson upper bound for 2008.
Most ZIFs are particle-like and have a 3D framework [101,102]. Recently, new kinds of ZIFs with other shapes have been prepared and used as fillers for the fabrication of MMMs. Wang et al. fabricated MMMs by dispersing ZIF-8 hollow nanotubes (H-ZIF-8) (Figure 7) in a Pebax 1657 membrane matrix [125]. Compared with neat Pebax, the MMMs containing 5% H-ZIF-8 achieved promising CO 2 separation performances (CO 2 permeability of 147 Barrer and CO 2 /N 2 selectivity of 68), approaching the 2008 Robeson upper limit and surpassing most reported CNTs/Pebax-based membranes.
Recently, 2-D nanoporous nanosheets have attracted significant attention due to their beneficial features associated with lower diffusion resistance and large external surface area, the latter of which allows the exposure of more active sites. The 2-D ZIFs are a new kind of ZIF, with a cushion-shaped cavity between the layers that has dimensions of 0.94 nm × 0.7 nm × 0.53 nm and a leaf-shaped structure (named ZIF-L in Figure 8; reported and prepared by Chen et al.) [161]. Feng et al. used 2-D ZIF and 2-D ZIF-derived nanoporous nanosheets in various applications due to their special morphology and pore features [162], including metal removal [163], water purification [164], solvent separation [111], gas separation [165][166][167][168], and so on. Recently, 2-D nanoporous nanosheets have attracted significant attention due to their beneficial features associated with lower diffusion resistance and large external surface area, the latter of which allows the exposure of more active sites. The 2-D ZIFs are a new kind of ZIF, with a cushion-shaped cavity between the layers that has dimensions of 0.94 nm × 0.7 nm × 0.53 nm and a leaf-shaped structure (named ZIF-L in Figure 8; reported and prepared by Chen et al.) [161]. Feng et al. used 2-D ZIF and 2-D ZIF-derived nanoporous nanosheets in various applications due to their special morphology and pore features [162], including metal removal [163], water purification [164], solvent separation [111], gas separation [165][166][167][168], and so on.   Recently, 2-D nanoporous nanosheets have attracted significant attention due to their beneficial features associated with lower diffusion resistance and large external surface area, the latter of which allows the exposure of more active sites. The 2-D ZIFs are a new kind of ZIF, with a cushion-shaped cavity between the layers that has dimensions of 0.94 nm × 0.7 nm × 0.53 nm and a leaf-shaped structure (named ZIF-L in Figure 8; reported and prepared by Chen et al.) [161]. Feng et al. used 2-D ZIF and 2-D ZIF-derived nanoporous nanosheets in various applications due to their special morphology and pore features [162], including metal removal [163], water purification [164], solvent separation [111], gas separation [165][166][167][168], and so on. In a recent work by Deng et al., the effect of the morphology of ZIF on the CO2 separation ability of ZIFs/Pebax 2533 MMMs was investigated [168]. Three ZIFs with different shapes-particles (0D), microneedles (1D) and leaves (2D)-were studied. The results show that the increment in CO2 permeability was related to the ZIF morphology, in the order of 0-D < 1-D < 2-D. The effects of morphology on selectivity seem to be the opposite, with the 0-D ZIF showing the highest selectivity. Zhang et al. synthesized 2-D ZIF-IL nanosheets using zinc salt and 2-methylimidazole aqueous solution, and incorporated these into a carboxymethyl cellulose (CMC) solution to form a steadily mixed aqueous suspension through one-step solution blending (Figure 9) [166]. The resulting MMMs with a 30 wt.% loading of ZIF-IL exhibited the highest separation ability (the selectivities of H2/CO2, H2/N2, CO2/CH2 and N2/CH2 were 10.62, 21.54, 17.87 and 8.93, respectively). In a recent work by Deng et al., the effect of the morphology of ZIF on the CO 2 separation ability of ZIFs/Pebax 2533 MMMs was investigated [168]. Three ZIFs with different shapes-particles (0D), microneedles (1D) and leaves (2D)-were studied. The results show that the increment in CO 2 permeability was related to the ZIF morphology, in the order of 0-D < 1-D < 2-D. The effects of morphology on selectivity seem to be the opposite, with the 0-D ZIF showing the highest selectivity. Zhang et al. synthesized 2-D ZIF-IL nanosheets using zinc salt and 2-methylimidazole aqueous solution, and incorporated these into a carboxymethyl cellulose (CMC) solution to form a steadily mixed aqueous suspension through one-step solution blending (Figure 9) [166]. The resulting MMMs with a 30 wt.% loading of ZIF-IL exhibited the highest separation ability (the selectivities of H 2 /CO 2 , H 2 /N 2 , CO 2 /CH 2 and N 2 /CH 2 were 10.62, 21.54, 17.87 and 8.93, respectively).
Zhu et al. proposed the functionalized auxiliary etching of ZIF-L fakes (hZIF-IL) by tannic acid (TA) to enlarge the surface or the pores, and enhance the hydrophilicity [165]. The as-prepared Pebax 1657/hZIF-IL MMM with 5 wt.% loading of hZIF-IL showed significant improvement in membrane transport properties (40% increase in CO 2 permeability and 70% increase in CO 2 /CH 4 selectivity), compared to the neat Pebax 1657 membrane.  In summary, Pebax, 6FDA-based polyimide and PSf are the best options for use as the polymer matrix for ZIF-based MMMs. ZIF-7 and ZIF-8, together with their functionalized counterparts, received the most attention due to their excellent separation performances.

Polymer/Oxide Nanoparticles Mixed Matrix Membranes
Oxide nanoparticles, such as silica, alumina, TiO2, etc., are often used as filler in MMMs because the particles could modify the properties of neighbor polymers and the interface between the fillers and the polymers, and increase or decrease the path to be traveled by one of the gases, leading to increased permeability [169]. Table 5 summarizes the recent separation performance of oxide nanoparticles-based MMMs. Hosseini et al. prepared an MMM based on a combination of MgO nanoparticles surfaced-modified with Ag + and Matrimid ® 5218 polyimide matrix. The surfaces of MgO nanoparticles (NPs) have high affinity towards CO2, and surface-anchored Ag + further enhances the solubility of CO2 through the formation of reversible π-complexation with double bonds. The obtained Matrimid-MgO-Ag + MMM with 20 wt.% inorganic loading showed 50% and 35% increases in selectivity for CO2/CH4 and CO2/N2, respectively, compared to the neat In summary, Pebax, 6FDA-based polyimide and PSf are the best options for use as the polymer matrix for ZIF-based MMMs. ZIF-7 and ZIF-8, together with their functionalized counterparts, received the most attention due to their excellent separation performances.

Polymer/Oxide Nanoparticles Mixed Matrix Membranes
Oxide nanoparticles, such as silica, alumina, TiO 2 , etc., are often used as filler in MMMs because the particles could modify the properties of neighbor polymers and the interface between the fillers and the polymers, and increase or decrease the path to be traveled by one of the gases, leading to increased permeability [169]. Table 5 summarizes the recent separation performance of oxide nanoparticles-based MMMs. Hosseini et al. prepared an MMM based on a combination of MgO nanoparticles surfaced-modified with Ag + and Matrimid ® 5218 polyimide matrix. The surfaces of MgO nanoparticles (NPs) have high affinity towards CO 2 , and surface-anchored Ag + further enhances the solubility of CO 2 through the formation of reversible π-complexation with double bonds. The obtained Matrimid-MgO-Ag + MMM with 20 wt.% inorganic loading showed 50% and 35% increases in selectivity for CO 2 /CH 4 and CO 2 /N 2 , respectively, compared to the neat membranes [170]. Ahn et al. reported a preparation of MMM comprising silica NPs and polymer with intrinsic microporosity (PIM). The permeability of the MMM increased with the addition of silica NPs due to the void cavities created by the impermeable particles. However, the selectivity of CO 2 /N 2 in the membrane decreased due to the addition of the NPs [171]. Ahmad et al. reported an MMM comprising TiO 2 NPs with an average particle size of 21 nm and a PVAc matrix. With up to 10 wt.% of the TiO 2 loaded, the membrane showed improvements in both thermal stability and gas permeation. At 2 bar and 30 • C, the permeability of O 2 , CO 2 and H 2 was increased by 95%, 79% and 62%, respectively. The selectivities of the gas pairs O 2 /N 2 , H2/N 2 , and CO 2 /N 2 were also increased by 38%, 26.5% and 14%, respectively [172]. Farashi et al. studied the effect of adding ZnO to the PEBAX matrix. With a 10 wt.% loading of 18 nm ZnO NPs on Pebax-1657, the CO 2 permeability and CO 2 /CH 4 selectivity increased about 13 and 21%, respectively [173]. Ameri et al. investigated the effects of adding alumina NPs to polyurethane membranes on gas separation performance. It was found that with the increase in Al 2 O 3 , the permeability of the gases decreased, along with an increase in selectivity. With 30 wt.% of alumina, the selectivity of CO 2 /CH 4 and CO 2 /N 2 increased from 7.77 and 25.84 to 23.48 and 67.88, respectively, at 760 mmHg [174]. Molki et al. found similar effects of nickel oxide (NiO) NPs on the permselectivity of polyurethane (PU). When the NiO content was low (1 wt.%), the CO 2 /N 2 selectivity increased by 79.21%, with a slight increase (~1%) in CO 2 permeability. However, when the inorganic load increased to 5 wt.%, the CO 2 /N 2 selectivity increased by 161.1%, while the CO 2 permeability was reduced by 3.31% [175]. Mesoporous particles with relatively large pores could be an ideal filler for an MMM due to the extra channel provided for gas diffusion. Wang et al. explored the use of mesoporous KIT-6 silica, which has a larger pore size (~6 nm) and a three-dimensional interconnected cubic pore structure that favors gas diffusion. With an optimum loading of 2 wt.%, the permeability and selectivity of CO 2 /N 2 simultaneously increased because the condensation of CO 2 enabled its passing through the otherwise unpassable, partially blocked mesopores ( Figure 10). This effect is even stronger with more condensable gases, such as C 4 H 10 , with which the permeability and selectivity against N 2 increased by 464% and 248% [183]. Surface modification of silica nanoparticles by APTES can enhance the permeability without reducing the selectivity. Su et al. reported that in an MMM comprising crosslinked PEG and SiO2 NPs, the permeability was increased by 20% to 133 Barrer with 2.7 wt.% inorganic loading [176]. To further increase the permeability of SiO2-containing MMMs, Hasebe et al. developed a type of polyimide-based MMM with a permeable nanospace by anchoring dendritic amino groups onto the surfaces of silica nanoparticles (Figure 11). The surface modification prevented particle aggregation, while the dendritic amino groups provided extra nanospace, which enhances the solubility and permeability of CO2. An MMM comprising 6FDA-DABA polyimide and 25 wt.% inorganic content Surface modification of silica nanoparticles by APTES can enhance the permeability without reducing the selectivity. Su et al. reported that in an MMM comprising crosslinked PEG and SiO 2 NPs, the permeability was increased by 20% to 133 Barrer with 2.7 wt.% inorganic loading [176]. To further increase the permeability of SiO 2 -containing MMMs, Hasebe et al. developed a type of polyimide-based MMM with a permeable nanospace by anchoring dendritic amino groups onto the surfaces of silica nanoparticles ( Figure 11). The surface modification prevented particle aggregation, while the dendritic amino groups provided extra nanospace, which enhances the solubility and permeability of CO 2 . An MMM comprising 6FDA-DABA polyimide and 25 wt.% inorganic content showed an improved CO 2 permeability of 1920 Barrer without significant reductions in CO 2 /N 2 selectivity (23) [58]. Building on the idea of using NPs with peripheral nanospace, Kudo et al. increased the loading of the silica NPs with nanospace up to 50 wt.%, so that the particles would start to aggregate anisotropically to form aggregations of pearl necklacelike structures. This gave rise to much nanospace, together with newly formed mesopores that can increase permeability without sacrificing selectivity. The MMMs with 50 wt.% dendritic amino silica in PIM showed permeabilities of 152,000, 2850, 916 and 1340 Barrers, respectively for CO 2 , O 2 , N 2 and CH 4 , and a CO 2 /N 2 selectivity of 16.6, a CO 2 /CH 4 selectivity of 11.4, and an O 2 /N 2 selectivity of 3.1, surpassing the 2008 Robeson upper bound [177]. Reprinted with permission from [58], copyright 2017 Elsevier. (b) Change from spherical particles to pearl necklace particles for the formation of efficient gas transport channels via the assembly of surface-modified and structure-specific nanoparticles at high concentrations. Reprinted with permission from [177], copyright 2020 Elsevier.
Polyhedral oligomeric silsequioxane (POSS) is a silica nanoparticle containing an inorganic silica and oxygen-caged core with organic functional groups attached. The organic functional end groups allow for POSS to be compatible with a wide range of polymers. Therefore, POSS can be used as an organic filler nanocomposite in membrane design due to its excellent mechanical properties, ability for dimensional building blocks, and flexibility for enhanced separation. Various polymer-POSS interactions have been studied for gas separation in CO2 capture from methane CH4, nitrogen (N2) and hydrogen (H2) gas [185].
Chemically manipulated POSS nanoparticles with different polymers can help enhance CO2 separation in membranes. A recent study examined the CO2/CH4 separation ability of a pure graphene oxide (GO) membrane. Typically, GO membranes are restricted based on the size and interlayer interactions of the 2D structure. Yang et al. researched the incorporation of amino-functional POSS into GO membrane structures [179]. POSS-NH2 has a rigid structure that can regulate the interlayer structure of the GO and overall increase the selectivity while decreasing swelling. The GO-POSS-NH2 membrane yielded a Reprinted with permission from [58], copyright 2017 Elsevier. (b) Change from spherical particles to pearl necklace particles for the formation of efficient gas transport channels via the assembly of surface-modified and structure-specific nanoparticles at high concentrations. Reprinted with permission from [177], copyright 2020 Elsevier.
Polymer-grafted nanoparticles have recently been explored in the context of CO 2 separation due to their increased mean free volume and higher permeability compared to neat polymers [184]. However, the increase in permeability is often associated with a drastic drop in selectivity. Bilchak et al. recently found that the distribution of free volume is spatially heterogeneous, with larger pores residing on the far-end of the grafted polymer chain. They discovered that by adding long free polymer chains in a polymergrafted NP matrix, such larger pores can be blocked to prevent the diffusion of larger gas molecules without significantly affecting the permeability of smaller gas molecules. Based on this design, an MMM was prepared comprising 3% long free poly(methyl acrylate) (PMA) chains (molecular weight (MW) = 96,000) and PMA-grafted silica NPs, which are synthesized via surface-initiated reversible addition−fragmentation chain transfer polymerization (SI-RAFT), with a chain length of MW of 100,000 and grafting density of 0.5 chains/nm 2 . Such an MMM showed a CO 2 permeability of 42.6 Barrer and a CO 2 /CH 4 selectivity of 23.2 at an equimolar binary feed ratio [178].
Polyhedral oligomeric silsequioxane (POSS) is a silica nanoparticle containing an inorganic silica and oxygen-caged core with organic functional groups attached. The organic functional end groups allow for POSS to be compatible with a wide range of polymers. Therefore, POSS can be used as an organic filler nanocomposite in membrane design due to its excellent mechanical properties, ability for dimensional building blocks, and flexibility for enhanced separation. Various polymer-POSS interactions have been studied for gas separation in CO 2 capture from methane CH 4 , nitrogen (N 2 ) and hydrogen (H 2 ) gas [185].
Chemically manipulated POSS nanoparticles with different polymers can help enhance CO 2 separation in membranes. A recent study examined the CO 2 /CH 4 separation ability of a pure graphene oxide (GO) membrane. Typically, GO membranes are restricted based on the size and interlayer interactions of the 2D structure. Yang et al. researched the incorporation of amino-functional POSS into GO membrane structures [179]. POSS-NH 2 has a rigid structure that can regulate the interlayer structure of the GO and overall increase the selectivity while decreasing swelling. The GO-POSS-NH 2 membrane yielded a CO 2 /CH 4 selectivity of 74.5, compared to the pure GO membrane, which had a CO 2 /CH 4 selectivity of 10.
Kim et al. synthesized co-polymer POSS-PEG with a side chain of polymethlhydrosiloxane (PMHS), which was used as the starting polymer and was casted on a polyacrylonitrile (PAN) and polydimethylsiloxane (PDMS) membrane support composite [180]. Increased amounts of allyl-POSS caused the membranes to become brittle due to their high crystallinity, whereas high amounts of allyl-PEG enhanced the flexibility. However, large amounts of PEG resulted in membranes unable to be cast as free-standing. The copolymer PEG-POSS with a ratio of 80:20 resulted in the best CO 2 /N 2 permeability of 679 Barrer and a selectivity of 38.1. Ahmad et al. reported amino-functionalized POSS with polysulfone (PSf)-soaked ionic liquid (IL) [181]. IL is an excellent CO 2 adsorbent and has high thermal stability. IL-PSf/POSS membranes displayed an increase in selectivity for CO 2 /N 2 of about 24.12, and a CO 2 permeance of 34.98 GPU.
Optimizing the content of POSS can also improve CO 2 /H 2 separation. Ansaloni et al. fabricated ammonium chloride-functionalized POSS with 6FDA reactive precursors for tubular and disk polyPOSS-imide membranes [182]. Pure and mixed gas tests in N 2 , CH 4 , H 2 and CO 2 showed that polyPOSS-imide had a CO 2 /N 2 selectivity of 18. The permeation of the H 2 decreased compared with pure gas conditions, potentially due to the concentration polarization in the feed gas. However, Chung et al. showed that Pebax/POSS had the best CO 2 permeability of 135 Barrer at 8 atm pressure, with na CO 2 /H 2 selectivity of 52.3 for mixed gas [186]. It was observed that integrating Pebax/POSS into an MMM enabled a better separation performance in mixed gas than pure gas due to CO 2 plasticization. The plasticization helped increase the free volume and chain mobility.
Polymer-POSS allows for greater tunability and flexibility. Future studies will continue to better our understanding of different polymers with POSS, enabling us to achieve better CO 2 separation from either N 2 , CH 4 or H 2. These polymer/nanocomposites can be incorporated into MMMs and other membrane supports to enhance certain desired properties for CO 2 gas separation.
Of these reported MMMs, PU-based MMMs and commercial polymer Matrimid-based MMMs gave some of the best selectivities, while PIM-based MMMs achieved some of the best permeability values. POSS-based inorganic fillers tend to require relatively low amounts of loading compared to other oxide-based nanoparticles.

Polymer/Nano Carbon Mixed Matrix Membranes
In the past three decades, nanocarbon materials have emerged as promising filler materials for making MMMs due to their unique surface functionality, variety of 3D structures, and other superior physical properties. These carbon materials include 0-D carbon quantum dots, 1-D carbon nanotubes (CNT) and carbon fibers, 2-D graphene and graphene oxide (GO), 3-D carbon molecular sieves and activated carbons, etc. Table 6 lists some of the recent examples of polymer/carbon MMMs.  [198] Quantum dots are 0-D nanomaterials of a few nanometers with different properties from larger particles due to quantum mechanics. Shi et al. investigated the effect of adding polymer-like quantum dots (PQD) and graphitic-like QDs (GQD) to a PEBAX matrix on gas separation performance ( Figure 12). They found that, while PQD destroyed the crystalline structures of the polymer matrix, GQD with fewer functional groups caused an enhancement in microphase separation of PEBAX with only 0.05 wt.% loading, which led to more segregated domains for selective CO 2 permeation. As a result, only 0.05 wt.% of GQD loading is needed to achieve an optimal gas separation property, while 1 wt.% is needed for PQD [187]. The 1-D carbon nanotubes achieve transport diffusivity of gases that is orders of magnitudes higher than that of zeolites with comparable pore sizes, possibly due to their atomically smooth inner walls, which lead to less friction for mass transport [199]. Therefore, MMMs based on the use of either single-walled carbon nanotubes (SWCNTs) or multiwalled carbon nanotubes (MWCNT) have been reported. Kim et al. reported MMMs based on dispersing SWCNT-modified long alkyl chains in a PSf polymer matrix. It was found that adding SWCNT significantly increased the permeability, but with a reduced selectivity for CO2/CH4 [188]. Ahmad   The 1-D carbon nanotubes achieve transport diffusivity of gases that is orders of magnitudes higher than that of zeolites with comparable pore sizes, possibly due to their atomically smooth inner walls, which lead to less friction for mass transport [199]. Therefore, MMMs based on the use of either single-walled carbon nanotubes (SWCNTs) or multi-walled carbon nanotubes (MWCNT) have been reported. Kim et al. reported MMMs based on dispersing SWCNT-modified long alkyl chains in a PSf polymer matrix. It was found that adding SWCNT significantly increased the permeability, but with a reduced selectivity for CO 2 /CH 4 [188]. Ahmad et al. prepared β-CD-functionalized MWCNT/cellulose acetate (CA) MMMs via the wet-phase inversion technique. Such MMMs with only 0.1 wt.% filler loading showed improved gas permeability and selectivity for CO 2 /N 2 [189]. Zhao et al. prepared an MMM comprising MWCNT-NH 2 in a Pebax 1657 matrix. The incorporation of the MWCNT-NH 2 increased the amorphous region of Pebax, which was beneficial for gas permeability. With the addition of 33 wt.% of MWCTN, the permeability of CO 2 increased 3-folds, characterized by a significant decrease in the activation energy of permeation for CO 2 and no loss of selectivity for CO 2 /N 2 , CO 2 /H 2 or CO 2 /CH 4 [190]. Zhang et al. created a hydrogel-based MMM by wrapping MWCNT with a polyNIPAM hydrogel layer within a Pebax 1657 matrix. The MWCNT served as a highway for gas transport, while the hydrogel layer favors both CO 2 permeability and selectivity. Such MMMs in a humidified state showed a high CO 2 permeability of 567 Barrer, and selectivities of 35 and 70 for CO 2 /CH 4 and CO 2 /N 2 [191].
The 2-D graphene oxides (GO) have also been studied as fillers in MMMs for CO 2 separation [200]. However, individual GO nanosheets are impermeable to gas and liquids due to their high electron density. As a result, the conventional blending of GO with polymers results in MMMs with relatively low permeability. For example, Ha et al. fabricated PDMS/GO composite elastomer membranes. With 8 wt.% of GO added, the MMM showed a more than 99.9% reduction in gas permeation towards various gases, including H 2 , O 2 , N 2 , CH 4 and CO 2 , as well as an increase in CO 2 /N 2 and CO 2 /CH 4 selectivity over two order of magnitudes greater than neat PDMS [192]. On the other hand, due to the abundance of oxygen-based functional groups and the increased interfacial gaps, with a minor addition of GO nanosheets, the solubility of the CO 2 can be considerably increased, resulting in higher selectivity as well as higher permeability. Shin et al. added 0.3 wt.% GO nanosheets to a Pebax/PEG matrix, and the MMM showed both improved CO 2 /N 2 selectivity (55.8) and CO 2 permeability (~650 Barrer). Moreover, such GO-based MMMs showed outstanding anti-plasticization resistance (up to 9 bar) and long-term stability (~100 days) under mixed gas conditions [193]. Using GO as additives to achieve anti-aging was also demonstrated in MMMs using PIM-1 as a polymer phase, which is a type of membrane material notorious for fast aging. Luque-Alled et al. developed chemically crosslinked PIM-1/GO hybrids MMM by covalently attaching PIM-1 to GO functionalized with APTS (APTS-GO), and they found that with 10 wt.% addition of APTS-GO, the membrane maintains its initial CO 2 permeability after 150 days, which is 9 times longer than the pure PIM-1 membrane [201].
One way to achieve higher permeability and selectivity is by functionalizing GO nanosheets with amino groups. Zhang et al. modified GO nanosheets with aminosilane to obtain amino-GO/Pebax MMM. The surface modification improved the filler distribution and membrane mechanical strength. More importantly, the surface amino groups constructed a better CO 2 transport highway. In the dry state, with the addition of 0.5 wt.% of the amino-functionalized GO nanosheet, the CO 2 permeability increased from 95.4 Barrer to 172 Barrer. The performance is even more impressive under the humid state. With 0.9 wt.% of inorganic filler in the humid state, the MMM showed improved CO 2 permeability from 424 Barrer to 934.3 Barrer at 2 bar and 35 • C, while the CO 2 /CH 4 selectivity increased to 40.9 and the CO 2 /N 2 selectivity increased to 71.1 [194].
Another way to improve the permeability of GO-based MMMs is by using porous reduced GO (PRG). Dong et al. prepared PRG-based MMMs based on a Pebax 1657 matrix with a finely tuned degree of GO reduction to ensure sufficient residual amounts of functional groups on the reduced GO. The narrow gas flow galleries with an average gap of 0.34 between neighboring reduced GO sheets ensured a high molecular sieving effect, while the generated mesopores on the laminate produced rapid gas transport pathways. As a result, the MMM showed substantially improved CO 2 permeability (119) as well as CO 2 /N 2 selectivity (104) at 0.2 MPa and 30 • C [195]. In another work, He et al. reported a layer-by-layer CO 2 -philic Pebax MMM that was functionalized with both porous graphene oxide (PGO) and a "N 2 -phobic, CO 2 -philic" o-hydroxyazo-hierarchical porous organic polymer (o-POPs) ( Figure 13). Such MMMs exhibited a high CO 2 permeability of 232.7 Barrer, with an ideal selectivity of CO 2 /N 2 of 80.7 [196].
while the generated mesopores on the laminate produced rapid gas transport pathways. As a result, the MMM showed substantially improved CO2 permeability (119) as well as CO2/N2 selectivity (104) at 0.2 MPa and 30 °C [195]. In another work, He et al. reported a layer-by-layer CO2-philic Pebax MMM that was functionalized with both porous graphene oxide (PGO) and a "N2-phobic, CO2-philic" o-hydroxyazo-hierarchical porous organic polymer (o-POPs) ( Figure 13). Such MMMs exhibited a high CO2 permeability of 232.7 Barrer, with an ideal selectivity of CO2/N2 of 80.7 [196]. CNT and GO can be used simultaneously in an MMM to give both high permeability and selectivity by taking advantage of the smooth inner wall of CNT and the molecular sieving effect of GO. Li et al. prepared a Matrimid polyimide-based MMM with 5 wt.% MWCNT and 5 wt.% GO. Compared to the neat membrane, such MMMs showed an increase in CO2 permeability, CO2/CH4 selectivity and CO2/N2 selectivity of 331%, 149% and 147% respectively [197].
3-D carbon materials, such as micron-sized activated carbon (CA) particles, have also been explored as effective fillers in MMMs. Anson et al. created MMMs by adding micromesoporous activated carbon (AC) particles, with a particle size of 0.9 micron and surface area of 3271 m 2 /g, into an acrylonitrile-butadiene-styrene (ABC) polymer matrix. At 20 °C, with 10 wt.% of the AC, the MMM showed an increase in CO2 permeability from 2.87 to 10.81 Barrer, with a CO2/CH4 selectivity increase from ~24 to 35. Alternatively, by adding 40 wt.% of AC with a particle size of 4.47 micron and surface area of 818 m 2 /g, the MMM achieved a CO2 permeability of 20.5 with a CO2/CH4 selectivity of 51 at 20 °C [198].
In these nanocarbon-based MMMs, the commercial polymers Pebax and Matrimid are among the most widely adopted polymer matrix. GO-based membranes require the lowest CNT and GO can be used simultaneously in an MMM to give both high permeability and selectivity by taking advantage of the smooth inner wall of CNT and the molecular sieving effect of GO. Li et al. prepared a Matrimid polyimide-based MMM with 5 wt.% MWCNT and 5 wt.% GO. Compared to the neat membrane, such MMMs showed an increase in CO 2 permeability, CO 2 /CH 4 selectivity and CO 2 /N 2 selectivity of 331%, 149% and 147% respectively [197].
3-D carbon materials, such as micron-sized activated carbon (CA) particles, have also been explored as effective fillers in MMMs. Anson et al. created MMMs by adding micro-mesoporous activated carbon (AC) particles, with a particle size of 0.9 micron and surface area of 3271 m 2 /g, into an acrylonitrile-butadiene-styrene (ABC) polymer matrix. At 20 • C, with 10 wt.% of the AC, the MMM showed an increase in CO 2 permeability from 2.87 to 10.81 Barrer, with a CO 2 /CH 4 selectivity increase from~24 to 35. Alternatively, by adding 40 wt.% of AC with a particle size of 4.47 micron and surface area of 818 m 2 /g, the MMM achieved a CO 2 permeability of 20.5 with a CO 2 /CH 4 selectivity of 51 at 20 • C [198].
In these nanocarbon-based MMMs, the commercial polymers Pebax and Matrimid are among the most widely adopted polymer matrix. GO-based membranes require the lowest level of loading, and have a major impact on the improvement of selectivity. On the other hand, CNT-based MMMs tend to lead to huge improvement in terms of permeability.

Conclusions
In this mini review, we have examined recent advances in mixed matrix membranes based on five major types of fillers: zeolite, MOF, ZIF, oxide nanoparticles and nanocarbons. While attentions have been focused on single-gas permeabilities and ideal gas selectivities, it is worth noting that studies examining MMMs under practical conditions are extremely important. Moreover, mechanical strength, resistance to aging and plasticization, and cost/easiness of fabrication are equally important for developing practically viable membranes. While most MMMs are based on non-covalent blending, MMMs based on covalent grafting have shown unique advantages, especially given that thermodynamic phase separation could occur in non-covalent systems. Moreover, current MMMs are based on an inorganic-in-polymer design, but with the advances in membrane fabricating techniques and polymer chemistries, polymer-in-inorganic MMM could be possible, which would achieve both high permselectivities beyond the Robeson upper bound, and good mechanical integrity.
Because of the promise of MMMs, many companies, such as NuMat Technologies, have started the application of such materials in commercial business cases. However, several barriers need to be overcome before commercial success can be realized. For example, most research papers are performed under lab conditions, and more aggressive gas testing is needed to meet industry conditions, focusing on long-term durability, high pressure, and the presence of impurity gases. Additionally, as the fabrication process starts to be scaled up, aggregations tend to occur during membrane preparation, especially for membranes with high inorganic loadings. In addition, the fabrication cost and materials prices of potential MMMs need to be further reduced in order to compete with pure polymer alternatives.