Current Trends in Metal–Organic and Covalent Organic Framework Membrane Materials

Abstract Metal–organic frameworks (MOFs) and covalent organic frameworks (COFs) have been thoroughly investigated with regards to applications in gas separation membranes in the past years. More recently, new preparation methods for MOFs and COFs as particles and thin‐film membranes, as well as for mixed‐matrix membranes (MMMs) have been developed. We will highlight novel processes and highly functional materials: Zeolitic imidazolate frameworks (ZIFs) can be transformed into glasses and we will give an insight into their use for membranes. In addition, liquids with permanent porosity offer solution processability for the manufacture of extremely potent MMMs. Also, MOF materials influenced by external stimuli give new directions for the enhancement of performance by in situ techniques. Presently, COFs with their large pores are useful in quantum sieving applications, and by exploiting the stacking behavior also molecular sieving COF membranes are possible. Similarly, porous polymers can be constructed using MOF templates, which then find use in gas separation membranes.


Introduction
Membranes as adisruptive technology are able to reduce the global energy consumption in the chemical separation of raw materials,a sw ell as actively reduce greenhouse gases actively,a nd thus form the basis for as ustainable future. [1,2] Membrane technology in the petrochemical sector alone could replace distillation processes and save up to 80 % energy in separation processes,w hich could lead to 8% savings in the global energy consumption. More than half of the separations are to gas separations ( Figure 1). [1][2][3] Porous membranes have come along way from the first description of metal-organic framework (MOF) mixed-matrix membranes (MMMs) using MOF-5, [4] the first neat MOF membranes starting with Mn-(HCO 2 ) 2 in 2007, [5] and the development of the first ZIF-8 membranes in 2009, [6] to todayss tate-of-the-art membranes.C ovalent organic frameworks (COF) were used much later for gas separation membranes,since water stability was one of their early issues. [7,8] Nevertheless,t he first neat and 3D COF membranes comprising COF-320 date back to 2015, [9] whereas the first experimental CO 2 -separating MMMs using exfoliated NUS-2 and NUS-3 sheets were reported in 2016. [10] When MOF membranes were first developed, the aim was to make these membranes as thick as possible and membranes of 20-300 mmt hickness were synthesized. This originated more or less from experience with zeolite membranes,w hich was drastically changed for thin films. [6,11,12] Especially for gas separation and purification, MOFs and COFs show great potential that needs to be unlocked. Today we know that thinner layers are better due to two main factors:higher flux and better selectivity.However,d efects are still an issue with thinner films,and single crystals might be considered for recording permeation data. Nevertheless,t he goal is downsizing with the preparation of thin films on the nanometer scale and the use of nanoparticles with the best possible polymer-filler interactions in MMMs.A dditionally,t he development of methods for high reproducibility and control over processes is needed. Theaim of this Minireview is to give an overview and highlight trends for the next steps in MOF and COF membrane research, paying particular attention to novel and very hot topics.
Metal-organic frameworks (MOFs) and covalent organic frameworks (COFs) have been thoroughly investigated with regards to applications in gas separation membranes in the past years.More recently,new preparation methods for MOFs and COFs as particles and thin-film membranes,a swell as for mixed-matrix membranes (MMMs) have been developed. We will highlight novel processes and highly functional materials:Zeolitic imidazolate frameworks (ZIFs) can be transformed into glasses and we will give an insight into their use for membranes.Inaddition, liquids with permanent porosity offer solution processability for the manufacture of extremely potent MMMs.Also,MOF materials influenced by external stimuli give new directions for the enhancement of performance by in situ techniques. Presently,C OFs with their large pores are useful in quantum sieving applications,a nd by exploiting the stackingb ehavior also molecular sieving COF membranes are possible.Similarly,porous polymers can be constructed using MOF templates,w hichthen find use in gas separation membranes. Figure 1. An simplified depiction of the world energy consumption and the amount used only for separation tasks in the production of primary chemicals. [1][2][3] [*] B. HosseiniMonjezi, Dr.M .T sotsalas, Dr.A . Knebel Institute of Functional Interfaces (IFG) Thep rocessability of MOF and COF materials is of increasing importance and has caught the attention of the scientific community,l eading to many derivative materials with extreme potential, such as porous liquids, [13][14][15] amorphous,p orous MOF-based glasses, [16][17][18] and porous organic polymers, [19][20][21] bringing MOFs/COFs up to the next level. Also,f rom am ore fundamental point of view it is important to step away from the random testing of materials.W ew ill show published data that leads to ad eeper understanding of the materials properties,from experiment and theory. [22,23] To actually find experimental model systems for membrane separation, single-crystal permeation testing is necessary. [24,25] Also we will address stimuli-responsive MOF materials, where gas transport has been followed and in situ and framework effects such as gate-opening, vibrational modes, [26] and electrostatic interactions between guests,l inkers,a nd metal centers could be investigated. [27] Forr eal-life applications,g ood processability and performance of MOF and COF membranes is crucial, which is more amatter of post-processing rather than the original material. Park et al. recently published ap aper where they show that material development is the most important step towards good performing membranes. [28] Making new materials out of existing ones by novel processing methods leads to advanced materials. [29] Advanced separation techniques and devices will be highlighted here as well, such as quantum sieving with MOF and COF membranes for isotope separation. [30,31] Separation processes are among the greatest challenges worldwide and using membranes could help save the planet [2] by reducing greenhouse gas emission, either actively in CO 2 separations,o rpassively by saving energy. fiber. Making polymer-filler MMMs from MOF particles is in general simple and cheap. [32] Since most COFs grow to form sheet-like structures anyway,w ew ill dive deeper into MOFs here,w here obtaining sheet-like particles is not so trivial. Owing to the high structural variety that is offered by the chemistry of MOF materials,special techniques are needed to prepare sheet-like particles.

Sheet-Like MOF Particles
MOF nanosheets in the production of ultrathin MMMs in general lead to high performance in separation applications. An alignment in thin polymer composite films is guaranteed due to the shearing forces from the casting approach, making sheet-like particles extremely interesting for polymer composite films.
Avery interesting example for the preparation of sheets was published by Peng et al. in 2014, [33] where the lamellar structure of Zn 2 (bIm) 4 allows the soft physical exfoliation of sheets by wet ball-milling and mild chemical delamination (Figure 2A-C). In addition to the use of particles for MMMs, they used afiltration technique to deposit the MOF particles as an ultrathin film on av ery rough, porous Al 2 O 3 support ( Figure 2C). Achieving alayer of 5nmthickness on asupport with that amount of roughness is not possible by solvothermal growth methods or layer-by-layer deposition. Asolvothermal growth technique will always result in ag reater thickness to form adense and gas-separating layer. [33] Many lamellar growing crystals can be exfoliated chemically,which was also shown by Pustovarenko et al. in 2018. [34] They used as urfactant-assisted approach in the synthesis of nanosheets.T he first solution contains Al(NO) 3 ·9 H 2 Oa nd the surfactant hexadecyltrimethylammonium bromide (CTAB);t he other solution contains the deprotonated 1,4benzodicarboxylic acid (BDC) linker together with 2-aminoterephtalic acid (2-ATA)asapromoter. After both solutions are heated, nucleation is induced by blending them. The CTAB forms al amellar phase and the MOF grows as approximately 100 nm 100 nm sheets between the surfactant lamellas. [34] Another approach is diffusion-mediated synthesis at at wo-phase interface,a sr eported by Rodenas et al. [35] ( Figure 2D-F). MOF-2(Cu), which already grows as alamellar MOF,issynthesized at atwo-liquid interface.Bydiffusion control, the sheets grow along the polar/nonpolar interface. AFM (atom force microscopy) analysis proves to be al ot more accurate than SEM (scanning electron microscopy) imaging for determining sheet thickness. [35] We highlight the production of MOF sheets here,since the morphology has as trong impact on the gas-separation performance of the membrane.K ang et al. [36] report big differences for [Cu 2 (ndc) 2 (dabco)] n in bulk, cubic,and sheetlike morphologies.They evaluated the different morphologies by the MMM performance in precombustion hydrogen separation and find that 1) downsizing to nanocrystals increases the performance drastically,w hereas 2) the use of nanosheets increases the performance further and leads to Figure 2. A) SEM image and B) crystal structure of Zn 2 (bIm) 4 ;C)SEM image of Zn 2 (bIm) 4 sheets deposited by filtration as athin membranefilm. From Y. Peng et al. [33] reprinted with permission from AAAS 2014. D) SEM image showing the layered morphology of CuBDC. E, F) AFM analysis of delaminated CuBDC sheets. Reprinted from T. Rodenas et al. [35] with permission from SpringerN ature 2014. benchmark performances. [37,36] Also the Tsapatsis group reported astrong increase in the selectivity and permeability by almost 70 %w hen sheet-like particles are used. Their approach is to directly synthesize Cu(BDC) nanosheets 2.5 mmi nl ength and/or width and only 25 nm thickness.F or CO 2 /CH 4 separation, they find higher values with nanosheets than with spherical particles;t hey also predicted performances theoretically and came to the same conclusion. [38] A rather important aspect of the incorporation of MOFs into MMMs is the compatibility of filler and polymer. [38,39] The implementation of simulations could be ag ood hint towards defect density in MOF membranes [38,40] as it is already widely used to predict the best polymer-filler pairs. [41]

Preparation of Mixed-Matrix Membranes
In addition to material choice and particle preparation, ag ood procedure for MMM preparation is necessary to achieve the best possible performance.However,gaining the optimal interaction between inorganic fillers and polymers is challenging.
Thea dhesion between polymer and filler materials is strongly dependent on the ratio of inorganic to organic components in the MOF material. Fori nstance,t he MIL-96 material with av ery high amount of inorganic Al-m 3 -oxocentered trinuclear clusters shows avery poor polymer-filler interaction;i tforms agglomerates and even shows crystal ripening in operando,l eading to void formation. [39] in their paper on zeolite 4A, Moore and Koros [42] reported several cases of membrane defects that can occur as ar esult of nonideal effects (Figure 3). [42] Thedistribution of the MOF filler also plays acritical role, because the formation of percolation defects can occur, as shown by Castro-MuÇoz et al. [43] Small defects can have ah uge impact and MMM procedures should aim for the perfect embedding of fillers (Case 0) by improving the polymer-filler interaction. Even tiny problems in the compatibility of the components can lead to cracks and defects in the resulting composite membranes,e specially when ah igh content of filler is used. [44] We think that several factors play arole when apoorly performing MMM results:1)The solvent used for the polymer does not give stable MOF dispersion. This leads to low MOF loading capacity and bad performance due to agglomeration. [39] 2) Thep roportion of inorganic buildings units and organic linkers is suboptimal. [42] 3) A procedure for good polymer compatibility is not followed. Some recipes consist of complicated mixing procedures,such as the stepwise addition of specific small amounts of the polymer to the colloidal solution to form astabilizing polymer shell surrounding the nanoparticles. [45]

Porous Liquids for Liquid Processing of MMMs
Porous liquids (PLs) are anovel class of porous materials that have been known for only afew years.First proposed by the James group in 2007, [46] they reported the experimental breakthrough in 2015. [15] PLs are materials with as pecial feature:p orous cage structures with am aximum pore . Non-ideal effects in an MMM lead to ad rastic change in performance. Case 0: The ideal case-selectivity and permeability increase. Case 1: Ar igidified polymer layer around the filler.C ase II:V oids form around the filler,gas breaks through. Case III:"Halo" defects accelerate transport through polymer without transport through the filler.C ase IV:C logged pores exclude transport through the sieve. Case V: Formation of aregion with reduced permeability.R eprinted from Moore and Koros [42] with permission from Elsevier 2012.

Angewandte
Chemie diameter smaller than that of the solvent molecules surrounding them. Thus,t hey remain empty and accessible for gases while in the liquid state. [47] PLs can be categorized in three different types [46,48] (Figure 4): Ty pe 1PLs are cage materials that are liquid by themselves.T he only example known to us thus far are polyether-functionalized coordination cages that act as ionic liquids. [49] Carefully said, MOF-based melts,f or example,m ade of ZIF-62 might also be regarded as type 1p orous liquids (see below). [50][51][52] Ty pe 2p orous liquids are organic cages that can be dissolved in asterically demanding solvent. Fori nstance,o rganic cages could be obtained by cycloimination of (15S,16S)-1,4,7,10,13-pentaoxacycloheptadecane-15,16-diamine with the cross-linker benzene-1,3,5trialdehyde.H ere,1 5-crown-5 serves as the solvent for the cages. [15] Organic cages were recently used for propylene/ propane separation with great results. [53] In contrast, aT ype 3P L, is ac olloidal solution of solid framework particles.T here are reports of ZIF-8 and zeolite ZSM-5 dispersed in ionic liquids that are Ty pe 3P Ls. [13,54] A new finding also suggests MOFs and zeolites dissolved in long-chain organic oils and silicon oils yield Ty pe 3P Ls. [55] MOF-based PLs are able to load MMMs with ahigh wt. % due to high colloidal stability.R ecently,Z IF-67 and ZIF-8 nanoparticles could be functionalized on their outer surface by N-heterocyclic carbenes (NHCs) to make them solution processable.The NHC-functionalized ZIFs were able to form monodisperse,h ighly stable colloidal solutions in nonpolar solvents,inwhich many polymers are prepared. Using ZIF-67 and ZIF-8 with NHC functionalization also leads to av ery good interaction with the polymer matrix (6FDA-DHTM-Durene and 6FDA-DAM), enabling very high MOF loadings of up to 47.5 wt. %inside the polymer,while also being able to separate gases in the liquid state. [56]

MOF Glasses for Membranes
As already mentioned, some MOFs in the ZIF family melt and form stable liquids,when heated under inert atmosphere (typically Ar or N 2 ). [17,18] Thei nert atmosphere is crucial in order to prevent thermal oxidation and decomposition of the ZIF melt. Ap rototypical example is ZIF-4, which melts at % 590 8 8Cb efore thermal decomposition at % 600 8 8C ( Figure 5A). [17] Molecular dynamics simulations on the thermal behavior of ZIF-4 yielded further insights into the process of MOF melting. TheZ n À Nb onds dissociate on the ps timescale,g enerating undercoordinated Zn 2+ cations. [52] Subsequently,new ZnÀNbonds are formed by association of other imidazolate linkers.T hese simulations suggest that the liquid ZIF-4 still possesses microporosity similar to the crystalline phase,b ut there is currently no experimental proof that the pores in liquid ZIF-4 are accessible.N evertheless,t he liquid ZIF-4 can be regarded as av ariation of aT ype IP L. Quenching the liquid ZIF-4 to room temperature generates aglass denoted a g ZIF-4 (a g = amorphous glass). [17,18] Theglass features af rozen atomic configuration of the supercooled liquid state.X-ray total scattering experiments show that the glass is amorphous (i.e.d oes not possess long-range order), but it possesses alocal structure that is identical to that of the crystalline ZIF. [17,18,57] In the past few years,anumber of other (mixed-linker) ZIFs have also been shown to melt and form glasses. [18,50,51,58] Prominent examples include ZIF-62 and TIF-4, which are structurally closely related to the prototypical ZIF-4 and feature the same cag network topology,b ut as econdary imidazolate linker. Importantly,t hese mixed-linker ZIFs generally feature am uch lower melting point than conventional ZIF-4. As demonstrated for ZIF-62, the melting point of the crystals,a swell as the glass transition temperature of  [49] with permission from Springer Nature 2020. B) PL Type 2: Organic cages with permanent porosity in the crown ether.Reprinted from Giri et al. [15] with permission from Springer Nature 2015. C) PL Type 3: Colloidally disperse, NHC-functionalized ZIF-67 in the non-penetrating solvent mesitylene. Reprinted from Knebel et al. [55] with permission from Springer Nature 2020.

Angewandte
Chemie the corresponding glasses,c an be adjusted precisely by the amount of secondary linker,r esulting in am elting point of only % 372 8 8C, more than 200 8 8Cl ower than that of ZIF-4. [59] Most importantly,m ixed-linker ZIF glasses feature permanent porosity for av ariety of gases,s uch as CO 2 ,H 2 ,a nd several hydrocarbons ( Figure 5B). [50,51,59,60] Even though the sorption capacity of the ZIF glasses is typically approximately 50 %l ower than the capacity of their crystalline parent frameworks,t his finding sets the stage for the application of glassy ZIFs in gas separation. Kinetic sorption measurements of propane and propylene in a g ZIF-62 materials showed that propylene is adsorbed much faster than propane,demonstrating the potential of ZIF glasses for gas separation applications ( Figure 5C). [59] As an important consequence of their liquid-state processability, [61] ZIF glasses can easily form composites with other materials.B ennett and co-workers prepared MOFcrystal-glass composites of crystalline MIL-53 in am atrix of a g ZIF-62. [62] When it comes to membrane applications,M OF glasses have two conceptual advantages:1)The glasses can be easily processed and deposited in their liquid state and 2) there are no grains or grain boundaries in the isotropic glass.G rain boundaries are unavoidable in polycrystalline MOF membranes and represent af undamental problem, since these boundaries represent defects or microscopic cracks that can significantly compromise the selectivity of the membrane. Wang and Jin et al. reported the first ZIF glass membrane made of an a g ZIF-62 film on ap orous a-Al 2 O 3 support ( Figure 6). [63] In analogy to the ZIF bulk glasses,the ZIF glass membrane was prepared by melt-quenching asolvothermally synthesized polycrystalline ZIF-62 film (thickness % 70 mm) on a a-Al 2 O 3 support under inert atmosphere.T he original ZIF-62 film featured intergrown microcrystals associated with gaps,pinholes,and grain boundaries.The melt-quenched a g ZIF-62 membrane is smooth and defect-free without any grain structure ( Figure 6). TheZ IF-62 glass membrane showed enhanced gas separation properties,w ith separation factors of 50.7 (H 2 /CH 4 ), 34.5 (CO 2 /N 2 ), and 36.6 (CO 2 / CH 4 ). [63] Another recent proof-of-concept study reported an MMM consisting of a g ZIF-62 imbedded in ap olyimide matrix. [64] AZ IF-62/polyimide-MMM containing 20 wt. % ZIF-62 showed an improvement of its CO 2 /N 2 selectivity by % 27 %upon thermal transformation of the crystalline ZIF-62 to a g ZIF-62 at 440 8 8C.

Neat MOF Membranes
Neat MOF membranes are usually grown on ceramic supports by solvothermal methods.Since ceramic membranes are 1000 times more expensive (per m 2 )than polymeric films, neat MOF membranes are hard to apply in industrial settings. [11] MOF membranes on ceramic supports cannot be used for antifouling treatment such as decoking,since MOFs would burn as well-a huge disadvantage.T he one-time use of these membranes would be ah uge cost factor.N evertheless,M OF membranes have long been synthesized on ceramic supports,a nd we do not want to exclude potential applications.F rom af undamental perspective,e specially for understanding the transport properties of the MOF itself,itis  [52] with permission from Springer Nature 2017. B) C 3 H 6 and C 3 H 8 sorption isotherms of a g ZIF-62 glasses containing various amounts (x)o fthe bim À linker.C)Kinetic sorption profiles for C 3 H 6 and C 3 H 8 .R eprinted from Frentzel-Beyme et al. [59] with permissionf rom the American Chemical Society 2019. Energy-dispersive Xray mapping shows Zn distribution (green). Reprinted from Wang et al. [63] with permissionf rom Wiley-VCH 2020.

Angewandte Chemie
Minireviews of great importance to produce and measure crystalline intergrown layers.H owever,o fp articularly interest is the gathering of "true permeation data" from single-crystalline membranes (Figure 7). [25,65] The" true" separation properties can be measured using single crystals,and diffusion constants and real permeation data can be obtained. [66] Although singlecrystalline membranes would be the ultimate goal of membrane science,t hey cannot be obtained on al arge scale. Therefore,t he layer-by-layer growth of surface-anchored metal-organic frameworks (SURMOFs) could be ak ey approach, because it produces an almost perfect layer. This technique offers large-scale processability of neat MOF layers with highly defined thickness. [67] Thec rystallinity can be so high that HKUST-1 films become transparent, since the characteristic blue color centers are missing. [68] It has been demonstrated that this technique also offers applications for neat MOF membranes,with the first example being ZIF-8. [69] Thedefined heteroepitaxial growth of ZIF-67 and ZIF-8 with exactly the same layer height has been shown. [70] The techniqueso nly current drawback is the limited number of accessible MOF structures,d ue to the solvent and temperature limitations of the method. Recently,U iO-66-NH 2 has been made available by liquid-phase epitaxy [71] and many more complicated frameworks will be available as SURMOFs in the near future.T he SURMOF technology will set standards as atool since it is possible to follow the exact growth of neat MOFs step by step. [72]

Stimuli-Responsive Neat MOF Membranes
Whereas the response of MOFs to applied pressure and temperature is aw ell-known concept, [73] MOFs can also respond to other types of external stimuli, such as light and electric fields. [74] Thefirst conceptional proof of electric-fieldstimulated MOFs was shown theoretically by the group of Maurin et al. for the breathing behavior of MIL-53. [75] In general, electric fields should be able to align dipolar moments inside the MOF structures (e.g.t he linker molecules). [76] There are theoretical concepts that use strongly dipolar linkers to align dipolar moments in ah igh-voltage electric field. [77] Nevertheless,f or MOFs that are feasible for membranes or adsorptive separation techniques (ZIF-8, MIL-53 etc.), theory and practical work conclude that electric fields must exceed the breakthrough voltage before linker orientation occurs. [75,76,78] Nevertheless,s ome MOFs are known for ferroelectric effects,e ven when it is nearly impossible to measure the hysteresis at accessible temperatures. [79] This is the case for ZIF-8 with the space group of I4 3m,which is able to display as tructural transformation inside an electric field (500 Vmm À1 ). Thes ymmetry is reduced to the monoclinic space group Cm and the symmetry switches further to R3m at higher fields.A sc onsequence of the E-field-driven transformation, achange in rotational energy barriers and ahigher framework stiffness arises,w hich increases molecular sieving. [78] Zhou et al. demonstrated the direct synthesis of aZIF-8m embrane in an electric field. There,Z IF-8 crystallizes directly in the Cm space group during the growth process. This leads to av ery good molecular-sieving ZIF-8(Cm) membrane that outperforms the usual ZIF-8(I4 3m)m embrane. [80] Utilizing light-responsive molecules inside the pores of MOFs,either in the backbone or as aguest molecule,leads to controllable gas transport and adsorption. [27] When, for example,a zobenzene (AZB) is introduced in the pores as guest molecules,a no ld concept already used in zeolites, [81] molecular transport through the pores can be influenced by gating effects,a sc oncluded from in situ gas permeation results. [82] Another case shows AZB molecules as side chains on the backbone of MOFs,l eading to adsorptive separation differences due to cis-trans isomerization, which effects adsorption of CO 2 by hindering its diffusion via interaction with its quadrupolar moment. [27,83] Light response has also been shown for MMMs,w here JUC-6 and PCN-250 were used inside aM atrimid 5218 membrane.N evertheless, thermal effects in the MMM could also have an effect on the gas separation. [84] Thef ield of stimuli-responsive membranes and switches [85] is strongly growing and the visionary aim towards au niversal membrane system for all different kind of separations looks achievable.  [25] with permission from Elsevier 2019. B) Measuring anisotropic gas permeationt hrough [Cu 2 (bza) 4 (pyz)] n with 1D channels. The channels in the (100) direction are free for gas transport, no gas permeation occurs in the (001) direction. Reprinted from Takamizawa et al. [65] with permission from the American Chemical Society 2010. Chemie

Membranes Based on Covalent Organic Frameworks
Thef irst COFs reported by Yaghi et al. in 2005 [8] were constructed utilizing the reversible reaction of boronic acid trimerization to form boroxine COF (e.g.C OF-1), or their condensation with catechols to form boronate ester COFs (e.g. COF-5). Both reactions proceed with the evolution of H 2 Oa nd therefore the equilibrium in these reversible reactions is dependent on the water content and humidity stability is limited. Since then, many different reversible reactions have been used for the formation of COFs,a lso providing high chemical, thermal, and mechanical stabilities. Thev ery first example of aC OF studied in terms of gas separation was reported by Zhu et al. in 2013. [86] Them icroporous boronate ester 3D COF (MCOF-1) derived from tetra(4-dihydroxy-borylphenyl)methane and 1,2,4,5-tetrahydroxybenzene was synthesized and investigated as an adsorbent for various gases,such as methane,ethylene,ethane,and propane.Although no experimental data was provided in this study,itclearly demonstrated the high potential of COFs for gas separation, and since then research on COF membranes for gas separation has increased drastically. [86]

Neat COF Membranes and the Bilayer Approach
Due to their large pore sizes,COFs have found use in gas separation membranes as components of stacked bilayers on porous supports (e.g. porous a-Al 2 O 3 ,c ellulose acetate, Nylon). Theb ilayer approach uses two different materials stacked upon each other. This was realized in both bottom-up and top-down ways.T he first examples relied on a-Al 2 O 3 supports for the synthesis of a mm-thick film of azine-linked COF (ACOF-1) via the condensation of 1,3,5-triformylbenzene and hydrazine hydrate under solvothermal conditions. [87] Thep erformance of the resulting membranes was measured for CO 2 /CH 4 mixed-gas separation using aWicke-Kallenbach permeation apparatus and reached a(CO 2 /CH 4 ) = 97.1 under optimized conditions.InACOF-1 CO 2 strongly adsorbs to the polar framework and the permeance of CH 4 is significantly lowered in the mixture compared to the single-gas permeance,w hich was explained by competitive adsorption mechanisms.T his is af ine demonstration for the critical need of mixed-gas permeances as atrustable measure of the material performance.I nadouble-layer system, using imine-linked COF (LZU1) on top of ACOF-1, performance could be increased. [88] Due to the large pores of COF materials,m olecular sieving can be achieved using an interlaced layer ("gateclosing" approach) between two COFs;outstanding H 2 /CO 2 , H 2 /N 2 ,and H 2 /CH 4 selectivities were achieved for the LZU1/ ACOF-1 bilayer membrane.A ne legant way to realize as imilar approach was recently reported by the Zhao group for two different COFs-an anionic imine-based COF, containing sulfonate groups,a nd ac ationic imine-based COF,c ontaining N-alkylated phenanthridine bromide (Figure 8). Both were deposited by the Langmuir-Schaefer method as thin films on ap orous a-Al 2 O 3 support. [89] Strong electrostatic interactions led to the formation of ac ompact staggered stacked film with narrow pores which could achieve very high H 2 /CO 2 selectivity. [89] One unique application for COF membranes aims towards the quantum sieving effect for the separation of hydrogen isotopes at cryogenic temperatures. [30] COF-1, prepared at room temperature in the presence of pyridine,c ontains one pyridine molecule per boroxine ring, limiting the pore size and provide kinetic hindrance at the aperture at cryogenic temperatures.Separation of an H 2 /D 2 isotope mixture could be achieved at 26 mbar loading pressure for the temperature range between 20 and 100 Kw ith as electivity a(D 2 /H 2 )o f9 .7 AE 0.9 at T exp < 30 K and 3.1 AE 0.5 at 70 K, exceeding the selectivity of commercial cryogenic distillation processes.

COF-Based MMMs
In 2016 Zhao [10] and Gascon [90] showed the first COFbased MMMs in mixed-gas separation. Exploring different 2D COF materials,the Zhao group used NUS-2 and NUS-3, which have the ultimate advantage of high stability against water. [10] These COFs are derived from the condensation of triformylphloroglucinol with hydrazine hydrate (NUS-2) or 2,5-diethoxyterephthalohydrazide (NUS-3). Thep resence of -OH groups allows keto-enol tautomerization to act as al ocking mechanism for the labile imine bonds by transferring them into nondynamic and chemically inert b-ketoenamine bonds.This stable 2D COF could be exfoliated to COF nanosheets and incorporated up to 30 wt. %p olyetherimide (Ultem )a nd polybenzimidazole (PBI). Them ixed-gas separation performance of these MMMs using an equimolar H 2 /CO 2 gas mixture was a = 5.80 for NUS-2@Ultem and a = Figure 8. The COF bilayer approach using cationic and anionic COF sheets to prepare staggered bilayer membranes for H 2 /CO 2 separation. Reprinted from Ying et al. [89] with permission from the American Chemical Society 2020. Angewandte Chemie 31.40 for NUS-2@PBI. TheM MM prepared using 2D imine ACOF-1 in Matrimid displays ahigh selectivity for CO 2 /CH 4 gas mixtures and at wofold increased CO 2 permeability. [90] Significantly increased CO 2 permeability owing to electrostatic interactions in COFs seems to be common, since other imine-based COFs (e.g.those formed by the condensation of melamine and terephthaldehyde) in PIM-1 [91] also showed this effect.
With regards to the polymer-filler interaction described in Section 2.2, purely organic, covalently bonded COFs typically perform very well in MMMs,i nc ontrast to MOFs or zeolites. [92] Nevertheless,t he compatibility of COFs and polymers can be further improved. Fore xample,amatrix able to make van der Waals interactions (e.g. hydrogen bonds between COF and polymer chains) allows for better component mixing. [93][94][95] When an NH-rich imine-COF was combined with an NH-rich PBI matrix, aCOF loading of 50 wt. % was possible, [93] whereas OH-rich COF-5 showed good compatibility with PEG-containing polyether block amide (PEBAX or VESTAMID E). [95] TheW ang group [96] premodified the surface of 2D imine COF-LZU1 particles with polyvinylamine chains,w hich resulted in the good compatibility of the modified COF with polyvinylamine matrix.

Advanced MOF/Polymer and COF/Polymer Hybrids
MOFs,COFs,and classical polymers feature different and often complementary properties in terms of their stability, surface area, and regular structure,a sw ell as their processability. [97] We want to provide ab rief summary of recent approaches going one step further in the combination of MOF/COF and polymers,i nw hich polymer species are inserted inside the MOF or COF pores and serve as precursors for MOF or COF growth, or in which MOFs/ COFs template the synthesis of porous polymer networks. [98] These new approaches can lead to improved performance and stability in avariety of membrane or separation applications, including water treatment and gas separation, for example, for CO 2 sequestering. [21,99] Theformation of advanced MOF-polymer hybrid devices and membranes can be divided into three main approaches categorized as:a )polymer synthesis within the MOF pores, b) PolyMOFs,and c) crosslinked MOFs (Figure 9). [19] Thed escribed approaches intend to either enhance the properties of the MOF or COF membranes by the combination of polymers with enhanced processability and stability,or enhance the performance of polymeric materials by the advantages of MOFs such as their high degree of order across multiple length scales,making it possible to implement highthroughput computational screening approaches. [23,100] We envision that these new concepts in the tight integration of MOF and COF materials on the one hand and polymer materials on the other hand will be further exploited in the future to tackle real-world separation challenges.

Perspectives
There is acritical need for disruptive technologies such as membranes to lower the energy use of the chemical industry and reduce greenhouse gas emission worldwide.M OFs and COFs are materials with extraordinary properties to help separations in the petrochemical sector, such as propylene/ propane,aswell as in direct CO 2 capture and the sustainable production of CH 4 .T om ake use of the potentially best materials for these processes,atargeted material development, rather than synthesis of more and more novel materials, is crucial. We have described the non-ideal polymer-filler effects in MMMs,w hich are already known but too often neglected. [42] We have recounted many pioneering studies of material development that are highly suitable for membrane science and we encourage people to work on these:p orous liquids and the liquid processability of MOF and COF particles in the production of polymer composite membranes; the formation of glasses composed of molecular-sieving ZIFs, opening up totally new perspectives,such as grain-boundaryfree films and the production of hollow fiber membranes made of neat MOF-glass.O nt he other hand, we think it is acrucial step and the main task of science to do fundamental research, determine material parameters (e.g.u sing single crystals for diffusion studies) and go to next-level separations such as quantum sieving.E ven though, for some of these processes finding an application is rather challenging,there is al ot to learn fundamentally:A sa ne xample,s timuliresponsive materials taught us al ot about MOFs and gas transport stimulation, whereas applications as "universal" membranes switching to the desired application are yet futuristic. Thep rerequisite here is that the phenomena be fully understood and that process integration is available for the spectrum of MOF,C OF,a nd polymer materials.A combined theoretical and experimental approach is necessary to develop these materials towards ak ey technology and transfer them to industry. Figure 9. Strategies for the tight integration of MOFsand COFsw ith polymer materials by a) integrating the polymer chain inside the pores, b) using polymericl inkers as precursors, or c) crosslinking the linker molecules post-synthetically.Reprinted from Begum et al. [98] with permission from the American ChemicalS ociety 2020.