Membranes processes are very important in our every day life but also in industry, e.g., for water and waste water treatment, in medical applications or separation of petrochemicals. Classical separation methods used for purification of chemical products, notably distillation, extraction and crystallization are energy and cost intensive. Over 50% of the energy costs in the chemical industry are used for the separation of gaseous or liquid mixtures [1]. With membrane technology, the costs for difficult separations, e.g., of azeotropic mixtures, can be reduced significantly [2].

Membranes with pore sizes of more than 2 nm made out of ceramics, zeolithes, glass, metal or polymers are frequently used in practice [3,4,5]. Here, separation is based on size exclusion and thus these membranes are suited for separation of components with sufficient size difference, e.g., in dialysis, waste water treatment and functional clothing. For other applications, especially for separation of components with comparable sizes or the separation of ions from water, solution-diffusion membranes are used [6].

The solution diffusion mechanism is based on the principle that the mixture component with higher solubility and higher diffusion rate permeates preferentially through the membrane, independent on the component sizes [7,8]. Solution-diffusion membranes feature free volume sites which cannot be occupied by polymer chains due to restricted motion and packing density of the polymer chains. The components are transported through the membrane by successive movement between the transient free volume gaps close to the feed side to those close to the permeate side due to thermal motion of segments of the polymer chains [9].

The pervaporation process, which is used for the separation of liquid mixtures, is schematically shown in Figure 1. The liquid mixture (feed) whose components shall be separated is led over the membrane. Before permeation a phase transition from the liquid phase to the vapour phase is induced by thermal heating. The component which permeates preferentially is concentrated in the permeate; whereas the detained component is enriched in the retentate. Which component is preferentially permeating is dependent on the solubility of the components in the polymer matrix and the diffusion rate of the components through the membrane. The driving force for permeation is given by the difference of chemical potentials of the components on feed and permeate side respectively, depending on pressure, temperature and concentration difference [8,10,11]. In gas separation, the feed as well as retentate and permeate are gaseous and no phase transition occurs.

The separation characteristics of a membrane are classified in terms of permeability P and selectivity α. The permeability is a measure of the permeate's quantity while the selectivity describes its quality (mole fraction of preferentially permeating component in the permeate) [9].

There are different types of materials which are suitable for membrane application. Inorganic membranes like ceramic membranes of metal oxides (γ-Al₂O₃, ZrO₂, TiO₂ or ZrTiO₄), perovskites or zeolites, porous carbon membranes, metal membranes or porous glass membranes [13] have exceptionally high permeability and selectivity together with thermal and chemical stabilities. However, manufacturing procedures are more complicated (e.g., require support treatment, zeolite crystallization, thermal programming for pyrolysis and a controlled inert gas atmosphere in terms of flow, pressure and composition) and there is a lower reproducibility for membrane properties associated with high cost and low mechanical stability. This makes the production of inorganic membranes more difficult than for polymer membranes [13,14,15].

Therefore, most of the commercial solution-diffusion membrane units contain polymeric materials [16,17,18,19], which can be divided into rubbery and glassy polymers. While glassy polymers show very attractive separation characteristics, rubbery polymers show comparably low selectivity with high permeability for common gas pairs such as O₂/N₂, H₂/CH₄, CO₂/CH₄, etc. [20,21]. Of the glassy polymers, polyimides have been found to be very promising as membrane materials because they not only have better separation characteristics than the commonly used ones (e.g., polycarbonates) but also offer good thermal and chemical resistance and are easy to process [9]. Additionally they can be easily functionalized and therefore customized for any given separation problem.

Solution-diffusion membranes already find application in removal of CO₂ from natural gas, dehydration of organic solvents, desalination, removal of polar low molecular weight components in equilibrium reactions, oxygen or nitrogen enrichment from air as well as in vapour recovery systems and several more [22,23,24,25].

Of those applications, natural gas treatment is of the highest industrial interest [26]. Currently, less than 5% of the plants processing natural gas employ membrane technology, which is mainly used for removal of carbon dioxide [27]. As mentioned above, processing costs can be reduced significantly by substitution or extension of the commonly used amine absorption plants [28] by membrane separation processes. Natural gas is composed of mainly methane (75%–90%) and some higher hydrocarbons. Unfortunately, a significant number of natural gas resources cannot be exploited due to the high amount of CO₂ (up to 30%) and other undesirable impurities (water, nitrogen, hydrogen sulfide, etc.) contained in the mixture, since they lead to corrosion of the pipelines [23]. The specifications for natural gas delivery to the U.S. national pipeline grid [29] demand carbon dioxide contents of less than 2%. First polymeric membranes for removal of CO₂ from natural gas consisted of cellulose acetate [24]. Cellulose acetate membranes have a CO₂/CH₄ selectivity of about 12–15 under normal operating conditions, which is not sufficient for industrial application [23]. Today, more and more polyimides are used as membrane material instead of cellulose derivatives, since the separation properties are better (selectivities of 20–25) [9,23]. However, the low durability of unmodified polyimide membranes poses a problem, which can be overcome by crosslinking [9,30].

Besides natural gas treatment, separation of aromatics and aliphatics is a promising field for application of solution-diffusion membrane technology. In Europe, the benzene percentage in motor fuel is limited to 1 vol.% by law [31]. Reduction of the benzene content of reformates is usually done by extraction with usage of a tetraethylene glycol/water mixture as the extracting solvent. Here, the separation factor for the aromatics is between 2 and 3 [32]. By implementing pervaporation membranes, the cost-intensive extracting unit can be replaced. The pervaporation properties of many different polymers were studied in respect to an application as membrane material for aromatics/aliphatics separation, i.e., polyimides, polyphosphonates, acetyl cellulose, benzoylated chitosan or polyurethane-silica hybrid membranes. The benzene/cyclohexane selectivity obtained with those polymers as membrane material typically lies between 5 and 20, with a normalized flux of 5–10 kg µm m⁻² h⁻¹ [33,34,35]. It is important that the membrane material is stable and that the separation characteristics will not change drastically in dependency on the concentration of aromatics in the reformate. Although all of those polymers show promising pervaporation characteristics, the stability often recedes at high aromatic concentrations and elevated feed temperatures. Again, crosslinking of the polymer can lead to an increased durability of the membrane [32]. As was shown for methyl methacrylate/methacrylic acid copolymer membranes ionically crosslinked with metal ions, the separation factor and normalized flux are not smaller than those of the non-crosslinked polymers [36].

The separation of olefin/paraffin mixtures is rather difficult because of the small differences in physical properties. Currently, such separations are carried out by low temperature distillation, which is rather cost intensive. Although membrane based hybrid processes would be less cost intensive they are not applied, since the separation factors of the membrane materials currently commercially available (e.g., silicone rubber, polysulfone, cellulose acetate, 1,2-polybutadiene or polyethylene) are too low [37,38]. Very attractive separation characteristics can be achieved with polyimides as membrane materials [39,40,41,42,43], but these materials are very sensitive to plasticization. To obtain membrane materials suitable for olefin/paraffin separation on industrial scales, the problem of plasticization has to be resolved, first. Some strategies for increase of membrane durability are presented in this review.

Gas separation membranes used in industrial applications are usually integrated as hollow fibers or flat sheets packaged as spiral-wound modules [27]. They are composed of a thin nonporous gas-selective layer (typically less than 0.5 µm [23]) deposited onto a highly porous carrier layer [8,11,44]. The carrier material increases the mechanical stability of the membrane, without influencing the separation characteristics [17]. A promising advance is the usage of composite membranes, in which different materials are used for the two different layers. This way, materials for both layers can be optimized separately and the high-cost polymer materials necessary for good separation properties can be applied more economically, since they are only needed for the very thin selective layer [27].

A membrane material is well suited for separation, when permeability and selectivity both are high. By modification of the polymer, these parameters can be increased. Unfortunately, improvement of the permeability often is gained in conjunction with a loss of selectivity, or the other way around [17,20,45]. For several polycarbonates, poly(ether sulfone)s and polyimides the solubility selectivities for O₂/N₂ and CO₂/CH₄ were found to be very similar, independent on the polymer structure, while the selectivities varied significantly [11,46]. That implies, that diffusion selectivity D_x/D_y is of higher importance in gas separation than solubility selectivity S_x/S_y, according to Equation (1) [17].

(1)

Therefore, to optimize both permeability and selectivity, a narrower free volume distribution in combination with a higher chain stiffness should be attempted [47].

The most common approaches for membrane modification are shown in Figure 2. Usually at least two polymers are blended together, to take advantage of the different properties of those components. For example, it was found that monomers having bulky substituents, e.g., the 6FDA dianhydride which contains -CF₃ groups (see Figure 8 for structure), restrict chain mobility and simultaneously chain packing, and lead to significantly improved selectivity as well as permeability [46,48]. Today, most copolymers consist of alternating bulky and flexible monomer units are used and by now a variety of monomers is available [17].

Although permeability and selectivity are the main criteria for classification of the separation characteristics of a membrane, other properties of the material have to be considered as well, as already mentioned above. The membranes are required to be thin but nonetheless stable, and they are desired to be low in cost. Additionally to mechanical stability, the chemical resistance against the feed mixture is of great importance.

Membrane fouling, that is adsorption of impurities on the membrane surface or into the membrane, can lead to significant decreases in permeability of the membrane. By surface modification, e.g., grafting adequate monomers on the surface or plasma treatment, the surface roughness, hydrophilicity and charge can be adjusted to the particular feed mixture and thus prevent or confine fouling [49].

Native, i.e., unmodified, polymeric membranes often show strong plasticization effects when exposed to high partial pressures of CO₂ [50,51], hydrocarbons [52,53] or ethylene oxide [54]. This leads to an increase in intermolecular distance and molecular motion of the polymer chains due to a decrease of inter- and/or intramolecular attractive forces. Consequently, the permeability for all feed components increases whereas the selectivity decreases. Strong plasticization can even lead to a partial dissolution of the membrane [32]. The hydrocarbons present in natural gas can reduce the selectivity by 30%–50% [50,55]. The effect is larger for rubbery polymers than for glassy polymers, which implies that plasticization is a problem especially for polyimides. This can be confined by implementing either covalent or ionic crosslinks of the polymer backbone. In both cases the plasticization resistance [30,53,54,56] as well as the separation efficiency can be improved [57,58,59,60] compared to the non-crosslinked polymer for pervaporation [36,61] and gas separation [62,63,64,65].

Another approach to increase the separation characteristics of a membrane is the implementation of additives. Mixed-matrix membranes (MMMs) are constructed from an inorganic material in the form of discrete micro- or nanoparticles incorporated into a polymeric matrix as the continuous phase (see Figure 3). The utilization of two materials with different flux and selectivity opens the possibility to improve a gas separation membrane by allowing the synergistic combination of polymers with their easy processability and the superior gas separation performance of inorganic materials.

A strong improvement of permeability could be expected, if porous inorganic additives, e.g., zeolites, metal organic frameworks (MOFs) [66,67,68,69,70,71,72], carbon nanotubes (CNTs) [73,74,75,76,77,78,79,80,81] or porous layered oxides (PLOs) [82,83,84,85,86,87,88] with a pore diameter larger than the kinetic diameter of one gas component but smaller than the other one could be applied to the organic polymer.

This would be economically very attractive for large scale applications if the selectivity does not change. First MMMs were studied in the 1980s and are more and more investigated in recent years [89,90,91,92,93]. MMMs receive attention as a possibility to enhance the properties of pure polymer membranes [94]. Separation properties with MMMs can be well above the Robeson upper bond (see Figure 4) [20,21], which is a plot of permeability versus selectivity for most industrial relevant gas mixtures. Porous inorganic fillers can overcome the tradeoff between selectivity and permeability which is typical for continuous (pure) polymer membranes.

Carbon nanotubes (CNTs) have been potentially suggested for their application in molecular separation by considering their hollow channels. In fact, simulations and detailed predictions demonstrated that the diffusivities of light gases inside the pores of CNTs are highly rapid compared to other porous materials [73]. According to this potential application, the use of CNT/polymer nanocomposites in membrane based separation processes is growing rapidly and considerable efforts were undertaken for the development in last decade [74,75,76,77,78,79,80] CNT particles are added as fillers to create preferential permeation pathways and thereby improve the separation performance in terms of the gas permeability [76,81]. The processability of polymeric membranes is combined with the higher permeability of nanotubes in order to achieve a synergistic separation performance that surpasses that of conventional polymeric and inorganic membranes.

To increase the gas permeability and also enhance the selectivity behavior of polymer membranes, addition of nanoporous fillers have been intensively investigated [90,93,95,96,97,98]. However, zeolites particle face the problem of formation of aggregates creating defects, especially if the polymer matrix is consisting of a glassy polymer [99]. Several attempts of modifying the surface of the zeolites to chemically bond them to the polymer chains and avoid macrovoid formation have been made, which often led to a considerable decrease of the gas permeability due to the rigidification of the polymer chains near the zeolite surface [100,101]. Also modification of the zeolite surface by using amino functionalized siloxan to induce the formation of hydrogen bonds has been used [102].

Furthermore, interest has risen in sheet-shaped porous layer oxides (PLOs) such as TiO₂ [89], SiO₂ [103,104,105,106] and Na₄Ti₂Si₈O₂₂·4H₂O [87,88] as fillers which are also called flakes or nanosheets if one of the dimensions is in the nanometer range. The aforementioned materials in lamellar form have been used as inorganic fillers in mixed-matrix membranes [87,88]. Such sheet-shaped or lamellar fillers have a high aspect ratio, which allows casting of thin mixed-matrix membranes. The selectivity of the membrane can be improved by size exclusion due to their microporous structure. Such filler materials are subdivided into selective and non-selective fillers depending on the permeability of the feed components through these materials. Also, such sheet-shaped fillers can be incorporated into MMMs in random or distinct parallel orientation. It was previously shown that the orientation of non-permeable inorganic sheets in different polymers has a great influence on the permeability [82,83]. Similar improvement of the separation characteristics can be achieved with smaller amounts of oriented fillers as with higher loads of non-oriented filler. Predictions of the permeation properties of mixed-matrix membranes with sheet-shaped inorganic fillers like porous layered oxides have been made for oriented non-selective sheets [84] as well for selective flakes [85]. Another approach to oriented sheet membranes dealt with fully inorganic MCM-22/silica nanocomposite membranes [86].

Crosslinking of polymer membranes can be achieved by different approaches, depending on the functional groups of the polymer and the desired separation properties of the membrane. The aim is to increase the durability of the membrane by circumvention of plasticization without causing losses in selectivity and permeability. Ideally, the separation characteristics of the membrane, especially the selectivity, are enhanced by implementation of crosslinks.

Often there is a variety of approaches to induce crosslinking of a polymer. For instance, a detailed review of different crosslinking methods for poly(vinyl alcohol) (PVA) membranes is provided by Bolto et al. [107]. PVA membranes show high stability in highly acidic or alkaline environments [108]. They are often chosen for dehydration procedures, where swelling due to adsorption of water presents a problem.

By performing several freeze-thaw cycles, crystalline regions are formed that act as crosslinks [109]. With increasing crystallinity the swelling is reduced. Although the physical crosslinks are not as strong or as stable as chemical ones, membranes modified this way last several months without a change of separation properties [107,110]. Reduction of plasticization of PVA can also be achieved by heat treatment at temperatures up to 160°, but the permeability is decreased as well. This, too, is explained by a change in crystallinity rather than the introduction of covalent crosslinks [111,112].

Treatment of PVA with strongly oxidizing radical producers (e.g., K₂S₂O₈) leads to formation of polymer radicals, which form crosslinks by radical coupling. This has also been shown to reduce the solubility of the polymer [113]. By varying the amount of persulfate the crosslinking degree can be regulated [114].

Metal-organic frameworks (MOFs) are of increasing interest as porous filler in mixed-matrix membranes (see Figure 3). MOFs offer various advantages over zeolites or other porous inorganic additives [167,168,169,170,171]: The organic ligands as an inherent part of MOFs allow them to interact well with the polymer material and its functionalities. This way, the formation of gaps between the “inorganic” filler and the organic polymer phase, which would cause losses in selectivity, can be avoided. Also the MOF surface properties can be easily tuned by functionalisation with various organic molecules if necessary. Moreover, MOFs also have higher pore volumes and lower density than zeolites, and therefore their effect on the membrane properties can be more pronounced for a given mass loading. For MMMs a perfect interaction between the two components is highly important in order to achieve optimized separation properties of the hybrid material.

Incorporation of the MOF 4,4'-bipyridine-hexafluorosilicate-copper(II) (Cu-BPY-HFS) (see Figure 11a) into a dense Matrimid^® 5218 membrane increased the gas permeabilities and selectivities for pure gases (from single gas experiments) and gas mixtures of CH₄/N₂ but decreased the ideal CO₂/CH₄ and H₂/CH₄ selectivities [172].

Cu-BTC MOF (see Figure 11b) as additive in an asymmetric membrane of Matrimid^® 9725 or of 3:1 Matrimid^®/polysulfone Utrason S 6010 N blend showed a higher CO₂ permeance compared to the unfilled membrane in mixed gas permeation experiments [173,174]. The binary gas mixtures CO₂/CH₄ and CO₂/N₂ were studied for CO₂ concentrations from 10 vol.% to 75 vol.%. The permeance of the preferential permeating gas CO₂ and, hence, the CO₂ selectivity increased with the filler loading. Noteworthy, the CO₂/CH₄ selectivity α_CO2/CH4 dropped linearly when the CO₂ content increased from 10 vol.% to 75 vol.%. The slope of this selectivity drop did not depend on the [Cu₃(BTC)₂] additive content [173,174]. This could be due to plasticization phenomena with higher CO₂.

Cu-BTC with polyimide (PI) prepared from 4,4-oxydianiline (ODA) and pyromellitic dianhydride (PMDA) was successfully spun into MMM hollow fibers by a dry/wet-spinning method. H₂ permeance and selectivity of H₂ with respect to N₂, CO₂, O₂ and CH₄ increased with increased Cu-BTC loading. At 6 wt.% Cu-BTC, the permeance of H₂ was higher by 45%, and its ideal selectivity from other gases was up by a factor of 2–3 compared to pure PI [175].

Cu-BTC (see Figure 11b) or ZIF-8 MOF (see Figure 12a) and the zeolite silicalite-1 (S1C) were combined in a polysulfone Udel^® P-3500 MMM [176] for the separation of CO₂/N₂, CO₂/CH₄, O₂/N₂ and H₂/CH₄ mixtures. For some of these gas mixtures, the combined-filler MMM showed a synergetic enhancement in selective gas transport when compared either to the pure polymer or to the MMM with only one filler type. All fillers, at the same loading of 16 wt.%, increase the CO₂ permeability when compared to that of the bare polymer in the two CO₂-containing mixtures. However, the maximum separation selectivities for CO₂/CH₄ and CO₂/N₂ mixtures (α_CO2/CH4= 22.4 with P = 8.9 barrer for CO₂, and α_CO2/N2= 38.0 with P = 8.4 barrer for CO₂) were only achieved when Cu-BTC was combined with S1C in a PSF MMM [176].

When the same Cu-BTC/ZIF-8/S1C single- or double-additive MMMs were applied to O₂/N₂ and H₂/CH₄ separations, for which mechanism is based mainly on diffusion and not on adsorption differences, the combination of Cu-BTC and S1C significantly enhanced the selectivities of O₂ or H₂, respectively. This may be due to an improved diffusion through zeolite crystals from the synergy with Cu-BTC [176].

MOF-5 as nanocrystals showed a strong affinity to the polymer matrix of Matrimid^® 5218. In MOF-5, six carboxylate groups coordinate to the six edges of the tetrahedral {Zn₄O} unit in octahedral geometry and the terephthalate linker then form the edges of a cube in the primitive cubic network (see Figure 11c). Single gas permeabilities of H₂, CO₂, O₂, N₂ and CH₄ doubled approximately with MOF-5 loadings of 30 wt.% from those of the pure polymer. This increase was ascribed to the MOF-5 porosity. Tests with binary CH₄-containing gas mixtures for the MMM with 30% MOF-5 in the polyimide revealed an increase in selectivity for CH₄. The larger solubility of CO₂ and N₂ in the polymer matrix was seen as the reason to favor enhanced CH₄ transport [177].

Zeolitic imidazolate framework-8 (ZIF-8,) with Matrimid^® 5218 MMMs was prepared with loadings up to 80% (w/w), which are much higher than the typical loadings achieved with selected zeolite materials [178]. In ZIF-8, the 2-methylimidazolate ligands connect individual zinc atoms and form the edges of a cuboctahedral -cage in a sodalite network (see Figure 12a). The ZIF-8/Matrimid^® MMMs permeabilities were tested for H₂, CO₂, O₂, N₂, CH₄, C₃H₈, and gas mixtures of H₂/CO₂ and CO₂/CH₄ and increased as the ZIF-8 loading increased to 40% (w/w). This increase in permeability suggested that the addition of ZIF-8 nanoparticles to Matrimid^® increases the distance between polymer chains creating higher free volume in the polymer matrix. For the majority of gas pairs (O₂/N₂, CH₄/N₂, H₂/O₂, H₂/CO₂, H₂/N₂), there was no significant change in the ideal selectivity until 40 wt.% ZIF-8 loading. However, at higher loadings of 50 wt.% and 60 wt.%, the permeability decreased for all gases, and the selectivities increased. Ideal selectivities of gas pairs containing gases of different size, such as H₂/O₂, H₂/CO₂, H₂/CH₄, CO₂/CH₄, CO₂/C₃H₈ and H₂/C₃H₈, showed improvement with the 50 wt.% ZIF-8 loading through a shift from polymer-driven to ZIF-8-controlled gas transport process. ZIF-8 can selectively transport smaller gas molecules, such as H₂ and CO₂ [178].

An asymmetric MMM of Matrimid^® 9725 and ZIF-8 (or MIL-53(Al), see below) exhibited a higher CO₂ permeance than the unfilled membrane in mixed gas permeation experiments [174]. MOF fillers with 10 wt.%, 20 wt.% and 30 wt.% were loaded in Matrimid^®. Binary gas mixtures CO₂/CH₄ and CO₂/N₂ with CO₂ concentrations from 10 vol.% to 75 vol.% were investigated and the results compared to Cu-BTC/Matrimid^® [173] or MIL-53(Al) [174]. The permeance of the preferentially permeating gas CO₂ increased with the filler loading in the binary mixtures in all three cases but the CO₂ selectivity only increased slightly in the case of Cu-BTC or MIL-53(Al) and remained almost constant for ZIF-8. The CO₂/CH₄ selectivity α_CO2/CH4 showed a constant drop when the CO₂ content increased from 10 vol.% to 75 vol.% for all three additives. The selectivity drop was independent of the filler content. Increase in CO₂ permeance was assigned to the extra pore network of the MOF fillers. Scanning electron microscopy cross-section images showed the MOF filler well distributed and embedded in the polymer matrix even for the 30 wt.% loaded membranes [173,174].

MMMs made of poly-(1,4-phenylene ether-ether-sulfone) (PPEES) and ZIF-8 with filler loadings of 10 wt.%, 20 wt.% and 30 wt.% were employed in CO₂ diffusion studies using pulsed field gradient (PFG) NMR techniques [179]. The self-diffusion coefficient increased from 2.1 × 10⁻⁸ cm² s⁻¹ for the pristine PPEES membrane to 9.3 × 10⁻⁸ cm² s⁻¹ for the 30 wt.% ZIF-8/PPEES-MMM. ZIF-8 provides Langmuir adsorption sites for CO₂ molecules, thus, the gas adsorption in the MMM increases with the filler content. ZIF-8 contributes greatly to gas permeation by increasing the gas solubility in the composite membranes [179].

Combining both ZIF-8 and silicalite-1 in polysulfone Udel^® P-3500 (ZIF-8/S1C-PSF MMM) did not improve the separation results from either S1C-PSF or ZIF-8-PSF MMMs in the case of CO₂/CH₄ and CO₂/N₂ gas mixtures. Probably the relatively large silicalite-1 crystals could not be intercalated between small ZIF-8 particles. ZIF-8 alone produced the highest increase of CO₂ permeability, which can be attributed to its textural properties and its small particle size, albeit giving poorly dispersed aggregates [176]. For O₂/N₂ and H₂/CH₄ gas separation a ZIF-8-PSF MMM produced the best selectivity-permeability results compared to a Cu-BTC- or S1C-PSF MMM. This may be due to an increase in free volume (as suggested for ZIF-8-polyimide MMMs) [178] together with an efficient molecular separation effect (based on diffusion differences) because of the small pore aperture window of ZIF-8 (3.4 Å, 0.34 nm) compared to Cu-BTC (6 Å, 0.6 nm) and S1C (5.5 Å, 0.55 nm) [176].

Different sonication powers were applied on ZIF-8 nanoparticles for the preparation of ZIF-8/Matrimid^® nanocomposite membranes with the aim of investigating the effect of typical membrane processing conditions on the structure, the interfacial morphology and the gas separation performance. It was shown that ultrasonication generates significant changes in the shape, size distribution, and structure of ZIF-8 particles suspended in an organic solvent during membrane processing. Although there are significant changes in the particle morphology, there are only minor losses in crystallinity and microporosity as proven from powder X-ray diffraction, synchrotron X-ray pair distribution function analysis and nitrogen physisorption. Dynamic light scattering and electron microscopy show that ZIF-8 nanoparticles undergo substantial Ostwald ripening when subjected to high intensity ultrasonication. Composite films prepared with both direct (high-intensity) and indirect (low-intensity) sonication show good adhesion between the polymer and ZIF-8 phases. However, films prepared using indirect sonication exhibit drastic agglomeration of nanoparticles while direct sonication produced ripened nanoparticles with variable dispersion. The ripened particles give lower pore volumes and lower surface areas compared to the as-synthesized material. ZIF-8/Matrimid^® composite membranes prepared from the two different sonication methods show significant differences in microstructure. Permeation measurements in membranes fabricated with high-intensity sonication show strong enhancement in permeability of CO₂ and increased CO₂/CH₄ selectivity, in agreement with the Maxwell model. In contrast, composite membranes prepared with low-intensity sonication are found to be defective [180].

Small and less agglomerated ZIF-20 with 8 wt.% in a polysulfone Udel^® P-3500 mixed matrix membrane gave a better separation of an equimolar O₂/N₂ mixture than the pure polymer. As in ZIF-8 the imidazolate moiety in the purinate ligands connects individual zinc atoms and forms the edges of a cuboctahedral -cage but in a zeolite-A network (see Figure 12b). The increase in O₂/N₂ selectivity from 4.7 (±0.4) to 6.7 (±0.5) in the MMM could be justified by the small difference in kinetic size between O₂ (kinetic diameter d_k = 0.343 nm) and N₂ (d_k = 0.368 nm) [181].

Submicrometer-sized particles of ZIF-90 (see Figure 12b, sodalite network again as in ZIF-8) were used to fabricate nanocomposite membranes with three different polyimides [Ultem^® 1000, Matrimid^® 5218 and 6FDA-DAM (6FDA: 2,2-bis(3,4-carboxyphenyl)hexafluoro-propane dianhydride, DAM: diaminomesitylene)] [171]. Scanning electron microscopy revealed an excellent adhesion of ZIF-90 crystals with the polyimides with no interfacial voids and well dispersed MOF crystals. Ultem^® and Matrimid^® MMMs showed significantly enhanced CO₂ permeability without any loss of CO₂/CH₄ selectivity. ZIF-90 with the highly permeable polymer 6FDA-DAM showed significant enhancements in both CO₂ permeability and CO₂/CH₄ selectivity. Membranes containing smaller particles showed slightly better results. The performance of ZIF-90/6FDA-DAM MMM exceeded the polymer upper bound for polymeric membrane performance from 1991 [20], and reaches the technologically attractive region (see Figure 4) [171].

The CO₂/CH₄ binary mixture gas-permeation properties of pure 6FDA-DAM and 15 wt.% ZIF-90/6FDA-DAM membranes revealed an enhanced gas-separation performance of the MMM. The CO₂/CH₄ mixed-gas selectivity of the ZIF-90 MMM was higher than the ideal selectivity measured by single-component gas permeation, presumably because of selective sorption and diffusion of CO₂ in the ZIF-90 crystals [171]. Also, a ZIF-90/6FDA-DAM membrane showed an ideal CO₂/N₂ selectivity of 22 compared to 14 for pure 6FDA-DAM indicating the possibility for separation of CO₂ from flue gases [171].

The MOF Mn(HCOO)₂, where each formiate ligand bridges between three manganese atoms (see Figure 13a), has very small channels of ~2 Å diameter only, and hence, as PSF-MMM showed high adsorption affinity for H₂ only. Higher loadings reduced the gas solubility, but increased the permeability, indicating defective membranes with interfacial voids [182].

An asymmetric membrane from MIL-53(Al) (or Cu-BTC or ZIF-8, see above) and Matrimid® 9725 showed a higher CO₂ permeance than the unfilled membrane in mixed gas permeation experiments [174]. The framework of MIL-53(Al) is a flexible, 'breathing'-type network, that is, it can assume different shapes and porosities depending on guest presence or absence. The BDC ligand bridges between four Al atoms. The hydroxo-bridging takes place along the metal chains in the b direction (see Figure 13b). The permeance of the preferentially permeating CO₂ in the binary gas mixtures CO₂/CH₄ and CO₂/N₂ with CO₂ concentrations from 10 vol.% to 75 vol.% increased with the filler loading for all three MOFs but the CO₂ selectivity slightly increased only in the case of Cu-BTC or MIL-53(Al) and remained almost constant for ZIF-8. Under the same conditions the CO₂ permeability remained invariant with the type of MOF for the CO₂/N₂ gas mixture. The CO₂/CH₄ selectivity α_CO2/CH4dropped linearly when the CO₂ content increased from 10 to 75 vol.% for all three MOFs. The decrease in selectivity was essentially independent of the MOF filler content [173,174].

A variant of MIL-53(Al), namely compound NH₂-MIL-53(Al) with the 2-aminobenzene-1,4-dicarboxylate linker (NH₂-BDC) but the same structure as MIL-53(Al) was used to fabricate nanocomposite membranes with PSF Udel^® P-3500. The homogeneously distributed NH₂-MIL-53(Al) particles showed excellent adhesion with the polysulfone without any additional compatibilization.

In contrast to most reported membranes, CO₂/CH₄ separation selectivity of NH₂-MIL-53(Al)/PSF-MMM increased with pressure due to the flexibility of the NH₂-MIL-53(Al) filler [183].

Mixed-matrix membranes with the water-stable MOF MIL-101 in polysulfone Ultrason S 6100 N exhibited a remarkable four-fold increase (compared to pure PSF) in the permeability of O₂ to technically needed values above 6 barrer and a simultaneous high selectivity for O₂ over N₂ of 5–6. The largest cage in this network with MTN zeolite topology is shown here with an inner diameter of ~34 Å and pore aperture windows up to ~16 Å. The benzene-1,4-dicarboxylate ligands bridge between trinuclear {Cr₃O} building units (see Figure 13c). The MIL-101 particles showed very good adhesion with polysulfone and long term stability [184].

Single gas experiments of MIL-101/PSF membranes with CO₂, CH₄ and N₂ at different MOF loadings showed increases in gas permeabilities with increasing MIL-101 weight percentage in PSF (see Figure 14). CO₂ is the preferentially permeating gas with permeability increases from about 5 to over 35 barrer from pure PSF to 20 wt.% MIL-101/PSF. The increase for CO₂ also raises the ideal selectivities for CO₂/CH₄ and CO₂/N₂ from about 20 to 25.

The CO₂ permeability is rather invariant to the thickness of the membrane while the CH₄ and N₂ permeabilities decrease when the membrane becomes thicker for the same MIL-101 loading (see Figure 15). Hence, the ideal CO₂/CH₄ and CO₂/N₂ selectivities strongly increase with the membrane thickness, in particular for CO₂/CH₄ (see Figure 15c).

Covalent crosslinking with maleimide derivatives as crosslinking unit can be investigated in situ in the liquid phase as well as in the solid phase using nanosecond transient UV/Vis absorption and FTIR measurements. Results point to maleimide anion formation via the triplet state and ionic dimerization to the cycloadduct as main reaction path upon UV irradiation. Additionally, the 2-2’-adduct is formed either from the educt or in a consecutive reaction from the cycloadduct. For 6FDA-ODA/6FDA-DABA 4:1 with MI the photocrosslinking yield was determined to be ~6%. The yield is restricted by the stiffness of the polymer backbone, which inhibits the photoreaction due to permanent spatial separation of the crosslinker side chains.

It has been shown for the separation of high-pressure mixtures of CO₂/CH₄ as well as for toluene/cyclohexane mixtures that the photocrosslinked copolyimide membranes are much more plasticization resistant than noncrosslinked copolyimide membranes. Generally covalent crosslinking results in much higher plasticization resistance and better selectivity than ionic crosslinking.

Mixed matrix membranes with metal-organic frameworks as additives (fillers) exhibit enhanced permeabilities and possibly also selectivities when compared to the underlying pure polymer. Matrimid^® and polysulfone are popular polymer matrices for MOF fillers. MOF particle adhesion to these organic polymers does not represent a problem. Permeability increases can be traced to MOF porosity. Crucial permeability-selectivity factors worth further investigations are, inter alia, filler content, filler (nano)size and dispersion, gas-binding functional groups in filler (e.g., amines for CO₂), water-stability of MOF-filler [185,186,187,188] as well as mixed-gases with different volume fractions instead of single gases. Addition of MOFs to polymers in MMMs easily yields performances similar to the best polymeric membranes and selectivities 10 times higher than those reported to date for any pure MOF membrane for the respective separation. In the end MOF-polymer MMMs allow for an easier synthesis and handability compared to pure MOF membranes [189,190,191,192,193].