Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A smart and responsive crystalline porous organic cage membrane with switchable pore apertures for graded molecular sieving

A smart and responsive crystalline porous organic cage membrane with switchable pore apertures... Articles https://doi.org/10.1038/s41563-021-01168-z A smart and responsive crystalline porous organic cage membrane with switchable pore apertures for graded molecular sieving 1,6 2,3,6 1 4 4 1 Ai He , Zhiwei Jiang , Yue Wu    , Hadeel Hussain , Jonathan Rawle , Michael E. Briggs , 1 2,3 1,5  ✉  ✉ Marc A. Little    , Andrew G. Livingston and Andrew I. Cooper    Membranes with high selectivity offer an attractive route to molecular separations, where technologies such as distillation and chromatography are energy intensive. However, it remains challenging to fine tune the structure and porosity in membranes, particularly to separate molecules of similar size. Here, we report a process for producing composite membranes that com- prise crystalline porous organic cage films fabricated by interfacial synthesis on a polyacrylonitrile support. These membranes −1 exhibit ultrafast solvent permeance and high rejection of organic dyes with molecular weights over 600 g mol . The crystal- line cage film is dynamic, and its pore aperture can be switched in methanol to generate larger pores that provide increased −1 methanol permeance and higher molecular weight cut-offs (1,400 g mol ). By varying the water/methanol ratio, the film can be switched between two phases that have different selectivities, such that a single, ‘smart’ crystalline membrane can perform graded molecular sieving. We exemplify this by separating three organic dyes in a single-stage, single-membrane process. 1,2 15 orous organic cages (POCs) are discrete molecules with polymerization . This produces amorphous polymer net- intrinsic cavities that can create porosity in molecular crys- works with a modest degree of pore tunability. There is a strong 1 3 4 Ptals , amorphous solids and porous liquids . The adsorp- demand to develop membranes with more tunable and modu- tion properties of POCs can sometimes be predicted in silico from lar pore structures. Various porous solids, including zeolites , 5,6 1,2 17 18 knowledge of their molecular structures in isolation . However, the POCs , organic polymers , metal–organic frameworks , cova- adsorption properties of POC materials are also affected by their lent organic frameworks (COFs) and hydrogen-bonded organic 2,7 20 solid-state packing . For example, extrinsic pores in POC crystals frameworks have been explored. Banerjee et al. reported COF can selectively adsorb guests, including rare gases . Indeed, ineffi- films with 1.4 to 2.6 nm pores that showed good performance cient packing of POCs can generate solids with considerably more in dye rejection . Dichtel et al. reported COF films with 3.4 nm 2,7 porosity than would be expected from the cage cavities alone . This pores and tunable thicknesses over the range of 100 μm to 2.5 nm combination of intrinsic and extrinsic porosity determines the func- that rejected Rhodamine WT from water . The same team also tionality of POC-based materials in selective adsorption processes. reduced the effective pore size of their COF membrane to 3.3 23 24,25 Most separation studies involving POCs have used molecular and 3.2 nm using reticular chemistry . In addition to COFs , 2,7 crystals , which can exhibit slow adsorption kinetics. Also, many metal–organic frameworks and their composites have been used 24,26 POC crystals rely on selective adsorption governed by thermo- to produce membranes . However, it remains challenging to dynamics, rather than kinetics, which limits their practical use produce continuous nanofiltration membranes with extended in size- and shape-selective membrane filters. Given their solu- porous frameworks that perform exclusively as size-based molec- tion processability, however, there is scope to develop crystalline ular sieves rather than selective adsorbents . POCs are solu- POC-based membranes that operate by selectively removing guests tion processable and their solid-state structures are defined by that are either too large or that have the wrong shape to diffuse non-covalent intermolecular interactions, which can be switched 28,29 through the POC pore structure. using chemical stimuli to alter their bulk porosity . As such, There is growing interest in membrane technologies that per- POCs are intriguing but relatively unexplored candidates for new 30–37 form industrial and environmentally relevant separations where types of membrane materials . two or more solutes are separated one from each other, as in distil- Many practically important molecular separations involve ter- lation or chromatography, as opposed to separations where a whole nary systems or more complex mixtures—for example, separating set of solutes is concentrated, such as in evaporation or seawater multiple hydrocarbon fractions from light crude oil by distillation, 9–13 38,39 reverse osmosis . A major advantage of membranes is that they pervaporation or organic solvent reverse osmosis ; purification 40,41 can perform separations in the liquid phase, which is often more of fatty acids , such as the practical recovery of omega-3 poly- practically useful than vapour phase separations. unsaturated fatty acids from fish oil by nanofiltration ; or sieving Membranes for liquid separations are typically produced using out by-products from reactions, for example in the liquid-phase 14 43 phase inversion, which can be followed by coating or interfacial peptide synthesis of pharmaceuticals . To achieve equivalent 1 2 Department of Chemistry and Materials Innovation Factory, University of Liverpool, Liverpool, UK. Department of Chemical Engineering, Imperial College 3 4 London, South Kensington, London, UK. School of Engineering and Materials Science, Queen Mary University of London, London, UK. Diamond Light 5 6 Source, Didcot, UK. Leverhulme Research Centre for Functional Materials Design, University of Liverpool, Liverpool, UK. These authors contributed equally: Ai He, Zhiwei Jiang. e-mail: a.livingston@qmul.ac.uk; aicooper@liverpool.ac.uk NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials 463 Nature Materials Articles Water MeOH Water Crystalline thin film CC3α phase in water CC3γ′ phase in MeOH b c CH Cl 2 2 MeOH Water CC3 ≡ ≡ CHDA TFB CH Cl Water MeOH 2 2 Intrinsic cavity CC3α CC3γ′ Extrinsic channel Fig. 1 | s ynthesis of a crystalline CC3 film and its crystal structures. a, Scheme showing the interfacial synthesis method used to fabricate crystalline CC3 films, which were subsequently adhered to a pAN support. These crystalline cage films can cycle between two different forms, CC3α-pAN and CC3γ′-pAN, by cycling the solvent between water and MeOH. CH Cl , dichloromethane. b, CC3α structure with its 3D pore network shown in yellow. c, The CC3γ′ 2 2 structure, formed by soaking in MeOH, has additional extrinsic solvent-filled channels, shown here in orange, that open up additional porosity in the membrane in response to the MeOH solvent. separations for complex mixtures using membranes, cascades of Fabrication of crystalline CC3 films membranes with graded molecular weight cut-offs (MWCOs) Continuous films with highly crystalline domains of CC3 were pro- have been developed , using phase inversion (polymeric mem- duced using a combined interfacial condensation reaction and crys- 45 46 branes) or sol–gel processing (ceramic membranes) by manip- tallization process at a water–dichloromethane interface (Fig. 1a). ulating the recipe for dope solution or fabrication conditions to This interfacial process allows the two-component reaction of CC3, produce multiple membranes with a variety of pore sizes. This which is synthesized via a [4 + 6] cycloimination reaction using places membranes at a disadvantage for ternary and higher sepa- 1,3,5-triformylbenzene (TFB) and (1R,2R)-1,2-diaminecyclohexane rations—by contrast, a single distillation or chromatography col- (CHDA), while simultaneously directing the formation of CC3 umn can produce multiple fractions with differing compositions. films at the interface (Methods). Continuous and free-standing CC3 Separating more than binary solute systems using a membrane films were transferred from the liquid–liquid interface onto various cascade requires multiple pumped recycle streams and complex substrates (for example, glass, steel mesh, carbon tape and silicon fluid controls . While solvent gradients are used in chromatog- wafers; Supplementary Fig. 1) for further analysis of the crystallin- raphy to modulate solid–liquid interactions, to the best of our ity and surface morphology. Before performing permeance and dye knowledge, there are as yet no reports of membranes that respond rejection studies, the CC3 film was coated onto a PAN support by to solvent gradients by changing their solute selectivity. filtration to form the composite membrane (Fig. 2a). The resulting Here, we report the fabrication of close-packed and defect-free membrane, referred to hereafter as CC3-PAN, was free of macro- films of a shape-persistent imine POC, CC3, which grow at the liq- scopic defects up to at least 7.4 cm in diameter using this prepa- uid–liquid interface between water and dichloromethane (Fig. 1a). ration process, with no evidence of delamination after cutting the These films comprise highly crystalline domains of CC3 in its most membrane into smaller pieces (Supplementary Fig. 2). The CC3 thermodynamically stable polymorph, CC3α (Fig. 1b). By coating film was characterized by Fourier transform infrared spectroscopy the CC3α film on polyacrylonitrile (PAN), we produce a continu- (Supplementary Fig. 3), Raman spectroscopy, nuclear magnetic res- ous membrane (CC3α-PAN) that has excellent permeance for both onance (NMR) spectroscopy (Supplementary Fig. 4), scanning elec- −2 −1 −1 polar and non-polar solvents, including water (43.0 l m h bar ) tron microscopy (SEM), focused ion beam SEM (FIB-SEM), X-ray −2 −1 −1 and toluene (55.9 l m h bar ). Furthermore, we found that it is diffraction and atomic force microscopy (AFM). For spectroscopic possible to rapidly and reversibly switch the membrane pore aper- measurements, a crystalline CC3α sample was used as a reference . ture using common solvents. Exposure of the non-covalent crystal CC3α has a three-dimensional (3D) diamondoid pore structure packing of CC3 in methanol (MeOH) induces a rapid phase transi- and is the thermodynamically most stable polymorph CC3 (ref. ). tion from CC3α to a different crystalline phase, CC3γ′, which is A Raman map was performed on an 80 × 80 μm CC3 film depos- less densely packed. This systematically increases the effective pore ited on glass (Fig. 2e,f and Supplementary Fig. 5), which indicated aperture of the resulting membrane, CC3γ′-PAN (Fig. 1c). This that the CC3 film comprised crystalline domains with the same switching property of the CC3γ′-PAN membrane allows the perme- solid-state structure as the CC3α polymorph (Fig. 2i). SEM images ation of larger organic dyes that can be rejected in water while the showed a continuous, apparently defect-free film in the CC3-PAN large pore apertures are turned ‘off ’ in the CC3α-PAN membrane. composite (Fig. 2b and Supplementary Fig. 6a) with a thickness This switchable porosity is reversible, and surprisingly, it does not of ~80 nm measured on a free-standing film (Fig. 2d), which con- compromise the continuity of the membrane. This allowed us to tained embedded, octahedral CC3 crystals (Supplementary Fig. 7). separate three organic dyes with different sizes via graded sieving Cross-sectional SEM images were obtained after step-by-step FIB using a single membrane. trenching and polishing of both CC3-PAN (Supplementary Fig. 8) NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials 464 Height (nm) Height (nm) Nature Materials Articles a b c CC3 film Si 100 CC3 film 500 nm 500 nm 20 CC3 film 500 nm PAN Height profile PAN support 80 nm 40 1 cm 500 nm 0 0.5 1.0 1.5 Distance (μm) e f g h 5 μm Stage 1: 4 h 20 μm 20 μm 5 μm CC3α film Stage 2: 16 h 48 h 24 h Crystalline CC3α 16 h reference 5 μm 8 h Stage 3: 24 h 4 h Amorphous CC3 reference 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 5 μm 2θ (°) 500 1,000 1,500 2,000 2,500 3,000 Stage 4: >48 h –1 Raman shift (cm ) Fig. 2 | Characterization of a CC3α film. a, photograph of composite membrane CC3α-pAN with a diameter of 7.4 cm. b, SEM image of CC3α-pAN showing the surface morphology of the CC3α film. Shown below is the cross-sectional FIB-SEM image of CC3α-pAN. c, AFM height image (top) and the height profile (bottom) of CC3α film transferred onto a silicon (Si) wafer. d, SEM image of a free-standing CC3α film, where the film was deliberately buckled to show its thickness. e,f, Raman microscope image (e) and Raman map (f) of a CC3α film on a glass support, where we purposely scratched the film before the measurement to expose the glass support (black stripe in f). The red regions on a CC3α film had comparable Raman spectra to the crystalline CC3α reference sample. g, SEM images of CC3α-pAN-X h-0.8% membranes formed at different reaction times, showing four stages of CC3α film formation. h, Out-of-plane GIXRD (wavelength, λ = 0.689 Å) patterns of CC3α-pAN-X h-0.8% membranes fabricated using reaction times between 4 and 48 hours 1 3 (2θ refers to the scattering angle). i, Raman spectra of CC3α film, a crystalline CC3α reference and an amorphous CC3 reference . and a CC3 film coated on a silicon wafer (Supplementary Fig. 9). The the CC3 film (Fig. 2g,h). This allowed us to create CC3 films from FIB-SEM images showed a clear boundary between the CC3 film the interfacial reaction that were four times thinner than the CC3 layered on top of the supports. After transferring the as-synthesized film created by spin coating . To further confirm the crystalline CC3 film onto a silicon wafer, we performed AFM measurements structure of the film, we performed a series of powder X-ray dif- to investigate the film thickness further. AFM again confirmed that fraction (PXRD) and grazing incidence X-ray diffraction (GIXRD) the CC3 film was continuous with a constant thickness of ~80 nm measurements on CC3-PAN (Methods). These diffraction mea- (Fig. 2c and Supplementary Figs. 6b, 8 and 9). surements revealed that the CC3 film was crystalline and had the A key advantage of interfacial synthesis is that it can create con- same structure as CC3α (Supplementary Figs. 10 and 11). 15,21 tinuous films of the product . Here, we also modified the reaction To further investigate the crystallization process of CC3 films at conditions to optimize the thickness, continuity and crystallinity of the solvent interface, we varied the reaction time from 4 to 96 hours NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials Intensity (a.u.) Intensity (counts) Nature Materials Articles a b c 1. Ethanol 7 2. Water 160 100 3. Toluene 6 80 4. Heptane 120 5. Hexane 6. Acetonitrile 4 60 7. Acetone Stage 1 Stage 3 4 h 24 h 2 48 h Stage 2 1 20 Stage 4 8 h 20 R = 0.9829 96 h 16 h 0 0 4 8 16 24 48 96 0 50 100 150 200 250 300 400 600 800 1,000 1,200 1,400 Reaction time (h) –2 –1 –1 δ η d Molecular weight (g mol ) d m d e f SO Na 100 3.0 Congo red 1,400 NH N N N N H2N 2.5 1,200 NaO3S 1,000 2.0 Retentate 1.5 Feed Permeate Applied pressure 1.0 Retentate 2 bar 20 bar Feed 5 bar 35 bar 0.5 Permeate 10 bar 0 0 0 5 10 15 20 25 30 35 400 600 800 1,000 1,200 1,400 400 500 600 700 800 –1 Wavelength (nm) Applied pressure (bar) Molecular weight (g mol ) Fig. 3 | Nanofiltration performance of CC3α membranes. a, plot showing pure solvent permeances versus their combined solvent properties (viscosity η, molar diameter d and solubility parameter δ ) for CC3α-pAN, where R is the coefficient of determination for the function. Hansen solubility parameter (δ) m d and the physical properties of each organic solvent are listed in Supplementary Table 2. b, Water permeance for CC3α-pAN-X h-0.8% membranes fabricated using reaction times that ranged between 4 and 96 hours. c, Dye rejection measurements for CC3α-pAN-X h-0.8% membranes in water. d,e, Water flux (d) and dye rejections (e) of a CC3α-pAN membrane under a range of applied pressures. f, Ultraviolet–visible absorption spectra of Congo red in water before (feed) and after (permeate and retentate) selectivity tests performed with CC3α-pAN. Insets show photographs of the feed, permeate and retentate solutions and the molecular structure of Congo red. Dye rejection was calculated using the intensity of the maximum absorption peak in the permeate and the feed, and equation (3) in the Methods. Mass balance calculations were performed using the maximum absorption peak values of the feed, permeate and retentate, with equation (4). All error bars depict the standard deviation (s.d.) of the data points obtained from at least three independent membranes. and manipulated the reagent concentrations from 0.2 to 2.5 wt%. the properties of CC3α-PAN-24 h-0.8%, referred to hereafter as We use the nomenclature CC3α-PAN-X h-Y% to refer to the mem- CC3α-PAN. branes made with X hours of reaction time and Y weight percent of the reagents. SEM, FIB-SEM and AFM revealed that thicker films Membrane performance of CC3α-PAN with larger crystals were produced as the reaction time and reagent To determine the permeance and dye rejection performance of concentrations were increased (Supplementary Figs. 12–21). By CC3α-PAN, we performed filtration experiments in dead-end cells contrast, using a reagent concentration of 0.2 wt% resulted in poorly using solvents and dyes with different sizes and chemical function- crystalline CC3 membranes (Supplementary Figs. 14 and 15). For alities (Supplementary Table 1 and Supplementary Figs. 22 and 23). the reactions with reagent concentrations of 0.8 wt%, the CC3 With a water contact angle of 94° (Supplementary Fig. 24), the film thickness increased with reaction time (30–600 nm from the CC3α-PAN membrane was stable in a range of polar and non-polar 4–60 h reactions; Supplementary Figs. 16 and 17), and the FIB-SEM solvents (Supplementary Fig. 25), proving that these solvents do images revealed triangle/octahedral-shaped crystals embedded not dissolve CC3. This led to ultrafast solvent permeances (Fig. 3a; in the CC3 films from the 32 and 48 h reactions (Supplementary Supplementary Fig. 26 for blank PAN data). We attribute this to the Figs. 18–20). By contrast, from the reactions with reagent concen- 3D interconnected porosity through the CC3α crystals in the film. trations of 2.0%, multiple CC3 films were found stacked on top of By comparison, an amorphous CC3 membrane prepared by spin one another (Supplementary Fig. 21). We suggest that the interfa- coating (Supplementary Section 1.2) exhibited a 30-fold lower sol- cial synthesis occurs in four stages (Fig. 2g): Stage 1 (0–4 hours), vent permeance under the same testing conditions (Supplementary interfacial polymerization of a continuous oligomeric film at the Fig. 27), although it should also be noted that the amorphous CC3 dichloromethane–water interface; Stage 2 (4–16 hours), self-sorting membrane was four times thicker . Apparently, the crystalline of the reactants and oligomers into the CC3α product and the for- CC3α-PAN provides sufficient robustness to support the intercon- mation of a partially reacted, semi-cage film; Stage 3 (24–48 hours), nected channels under high applied pressures. To further confirm crystallization of CC3α and the formation of octahedral crystals the importance of crystallinity, CC3α membranes with differ- in the film; and Stage 4 (48–96 hours), formation of defects in the ent crystallinity levels were fabricated at each of the four reaction film caused by larger octahedral crystals creating cracks and imper- stages simply by controlling the reaction time. A partially crystal- fections. GIXRD measurements demonstrated the crystallization line membrane (CC3α-PAN-8 h-0.8%) at Stage 2 exhibited a water −2 −1 −1 process across these stages, where the crystallinity increased with permeance of 3.0 l m h bar ; that is, an order of magnitude lower −2 −1 −1 a longer reaction time (Fig. 2h). We therefore focused attention on than the fully crystalline Stage 3 membrane (49.5 l m h bar for NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials –2 –1 Flux (l m h ) –2 –1 –1 Permeance (l m h bar ) Rejection (%) –2 –1 –1 Permeance (l m h bar ) Absorbance (a.u.) Rejection (%) –2 –1 –1 Permeance (l m h bar ) Nature Materials Articles a b CC3γ′ powdered reference Switchable pore aperture In water CC3 film in MeOH In MeOH CC3 film in water CC3 film in air CC3α powdered reference 0 200 400 600 800 1,000 1,200 1,400 1,600 2 3 4 5 6 –1 2θ (°) Molecular weight (g mol ) c d 40 60 1 2 3 4 5 Cycle 1 Cycle 2 Cycle 3 Cycle e f This work Cycle 2: in MeOH CC3γ′-PAN Upper bound CC3α-PAN MOF MC Cycle 2: in water CC3α-PAN NPs GO PTMSP Commercially available PA PEEK CC3γ′-PAN Cycle 1: in MeOH PEI PEO PANI PIM PI PIP Cycle 1: in water CC3α-PAN PE PAR 2 3 4 5 6 0 250 500 750 1,000 1,250 1,500 1,750 –1 2θ (°) Molecular weight cut-off (g mol ) Fig. 4 | X-ray diffraction characterization and switchable separation performance of CC3-PAN membranes. a, GIXRD of CC3α-pAN in air and water, and CC3γ′-pAN in MeOH. Experimental pXRD patterns of CC3α and CC3γ′ powders are included as references. b, MWCO curve for CC3α-pAN in water and CC3γ′-pAN in MeOH containing 20 ppm dye solutes. The MWCO was determined by interpolating from the plot of rejection against the molecular weight of the dyes and corresponds to the molecular weight for which rejection reaches 90%. All error bars depict the s.d. of the data points obtained from at least three independent membranes. c, Reversible dye rejection of BB and solvent permeance of the CC3-pAN membrane observed upon switching the feedstock solvent between water and MeOH. All error bars denote the s.d. for measurements from at least three independent membranes. d, photographs of CC3-pAN filtration dead-end cell captured from Supplementary Video 1 during the different cycles; BB is rejected in water by CC3α-pAN while CC3γ′-pAN does not reject BB in MeOH. e, In situ GIXRD patterns showing the reversible phase transition between CC3α-pAN and CC3γ′-pAN, by cycling between water and MeOH. f, Acetone permeance versus MWCO of general solutes in acetone for nanofiltration membranes reported in the literature and CC3α-pAN. MOF, metal–organic framework; MC, macrocycle; Nps, nanoparticles; GO, graphene oxide; pTMSp, poly(1-(trimethylsilyl)-1-propyne); pA, polyamide; pEI, polyethyleneimine; pANI, polyaniline; pI, polyimide; pE, polyethylene; pEEK, poly(ether ether ketone); pEO, poly(ethylene oxide); pIM, polymers of intrinsic microporosity; pIp, piperazine; pAR, polyacrylate (Supplementary Table 9 for full details). CC3α-PAN-48 h-0.8%) prepared with prolonged reaction times through the membranes (Fig. 3c). By comparison, amorphous (Fig. 3b). CC3α-PAN-8 h-0.8% and CC3α-PAN-48 h-0.8% exhib- oligomeric membranes produced in Stage 1 (CC3α-PAN-4 h-0.8%) ited the same MWCO, as determined by filtering a range of dyes and cracked, highly crystalline membranes produced in Stage 4 NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials Rejection (%) Intensity (a.u.) Intensity (a.u.) Water MeOH –2 –1 –1 Rejection (%) Acetone permeance (l m h bar ) MeOH Water Nature Materials Articles a b Mixture = NP + BB + DR NP (P) Permeate = NP Add MeOH BB (P) 100 80 60 40 20 BB Permeate = Water content in MeOH (vol%) DR (R) Retentate = DR c d In water In water NP NP BB Mixture NP (P) DR In MeOH MeOH added (90 vol%) Mixture BB BB BB (P) DR DR (R) 300 400 500 600 700 800 Wavelength (nm) Fig. 5 | Mixture fitting and graded sieving using a single switchable membrane. a, BB rejection in mixtures of water and MeOH (v/v) for CC3-pAN (top), and photographs of the permeates (bottom). All error bars depict the s.d. of the data points obtained from at least three independent membranes. The red dashed line was fitted as the logistic function (y = 1/(1 + exp(−16.1(x – 0.617))); Supplementary Section 1.4). b, photographs showing the ternary molecular separation in a filtration dead-end cell, the nascent mixture feedstock, the permeate (p) collected in the first and second step, and the retentate (R) collected in the second step. c, Scheme showing ternary molecular separation of three dyes (DR, BB and Np) using one single membrane (CC3-pAN) in a continuous process: Step 1, CC3α-pAN in water (blue background) allows permeation of only Np, leaving BB and DR in the retentate. Step 2, 90 vol% MeOH (green background) was added into the retentate to transform the membrane structure to CC3γ′-pAN, which allows permeation of only BB, leaving DR in the retentate. d, Ultraviolet–visible absorption spectra of the mixture containing three molecules in water, permeate from water, mixture and permeate from 90 vol% of MeOH in water and the remaining retentate. Note, the maximum absorbance wavelength for BB is 551 nm in water and 587 nm in MeOH; BB also shows absorbance at 305 nm in MeOH, while Np shows its maximum absorbance at 312 nm in the same solvent. (CC3α-PAN-96 h-0.8%) exhibited unexpectedly higher water per- CC3α-PAN membranes might be more competitive for separations meances but failed to achieve comparable separation performances, that require higher pressures. indicating that they contained physical defects. To confirm that dye adsorption did not contribute to the selectiv- Two limitations of membranes produced from other crystalline ity performance of CC3α-PAN, mass balance calculations were used porous materials, such as COFs, are poor stability at high pressures to measure the dye concentration in the retentate. After permeating and the interference of adsorption processes . Here, the CC3α-PAN 48 ml of Congo red from the 100 ml feed, the absorption intensity membrane was tested under a range of applied pressures, up to a of Congo red in the retentate increased from 1.24 to 2.53, while its maximum of 35 bar. The water flux increased linearly with increas- absorption intensity in the permeate was only 0.02. In combination, ing applied pressure (Fig. 3d) without affecting the MWCO these values are consistent with ~100% dye rejection (Fig. 3f and (Fig. 3e). Longer duration studies demonstrated the mechani- Supplementary Figs. 29 and 30). These measurements agree with cal robustness of CC3α-PAN and showed consistent dye rejection the colourless membrane surface observed after the dye filtration −2 −1 −1 (99.7% for rose bengal) and water permeance (~43 l m h bar ) (Supplementary Fig. 2). In addition, soaking powdered crystals of over 20 hours (Supplementary Fig. 28). The applied pressure of CC3α (100 mg) in the dye solution (100 ml) did not lead to adsorp- 35 bar is an order of magnitude higher than that used for liquid tion in the crystals after seven days (Supplementary Fig. 31). These filtration through COF membranes , which suggests that these results all indicate that the dyes were rejected by the membrane. NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials Rejection (%) Absorbance (a.u.) Nature Materials Articles s witchable pore aperture for graded sieving in situ GIXRD measurements on solvated CC3 films, while per- Previous studies showed that certain POCs can be switched between forming two consecutive cycles, we found that the composite mem- 28,29 multiple polymorphs to modify their porosity . The solid-state brane transformed cleanly between CC3α-PAN and CC3γ′-PAN structure of CC3 has been directed into different polymorphs by when the solvent was switched between water and MeOH and back crystallization from specific solvents , but until now, the solid-state again (Fig. 4e). We attributed this switching phenomenon solely to transformation of CC3 crystals was not explored, to the best of our the phase transition of CC3 films, rather than swelling of the mem- knowledge. branes. To validate this, the CC3-PAN membrane was soaked in We found that both air-dried and water-solvated films exhibited acetone and acetonitrile, and the nanofiltration tests were repeated. the same diffraction patterns as the reference peaks of CC3α pow- CC3-PAN exhibits comparable MWCOs in acetone and acetoni- ders measured by PXRD (Fig. 4a). A series of GIXRD patterns were trile to those observed in water (Supplementary Figs. 41 and 46) then recorded after submerging the membrane in various organic because the same phase, CC3α-PAN, is present in these solvents solvents (Supplementary Figs. 32 and 33). The crystalline CC3α (Supplementary Figs. 32 and 33). Remarkably, the acetone perme- −2 −1 −1 film transformed into a new structure when submerged in MeOH ance of CC3α-PAN reached 177 l m h bar with a MWCO of −1 (Fig. 4a). By indexing the GIXRD pattern, we confirmed this was ~600 g mol , which is well above the upper bound performance a MeOH-solvated CC3 phase (CC3γ′; Supplementary Fig. 34) that for nanofiltration membranes reported in the literature (Fig. 4f, was isolated previously by crystallizing CC3 from dichloromethane Supplementary Fig. 47 and Supplementary Table 9). and MeOH (ref. ). The CC3γ′ structure is very different from its A series of water and MeOH feedstocks containing the dye BB thermodynamically most stable polymorph, CC3α (ref. ), where were used to determine the dynamic transformation between the cage packs in a window-to-window arrangement to generate a CC3α-PAN and CC3γ′-PAN (Fig. 5a). Understanding this dynamic diamondoid pore network (yellow channels in Fig. 1b). By contrast, transformation allowed us to manipulate the pore aperture in a sin- the CC3 molecules in the CC3γ′ phase are packed less densely, thus gle CC3-PAN membrane by simply adjusting the water concentra- providing large extrinsic pores between hexagonally arranged CC3 tion in a water–MeOH mixture, without any activation processes molecules (orange channels in Fig. 1c). or the use of multiple membranes . To demonstrate this, we per- To investigate the structural transformation between CC3α-PAN formed a graded sieving experiment to separate molecules from a and CC3γ′-PAN, we performed a series of in situ GIXRD measure- ternary mixture using a single membrane. Initially, a water feedstock −1 ments while dosing the membrane surface with solvent vapour containing three dyes, 4-nitrophenol (NP; yellow, 139 g mol ), BB −1 −1 and after coating the membrane surface in a thin solvent layer (blue, 826 g mol ) and direct red 80 (DR; red, 1,373 g mol ) was fil- (Methods). CC3γ′-PAN formed by immersion in MeOH trans- tered through the CC3-PAN membrane (Fig. 5c). Since CC3-PAN formed back into CC3α-PAN after being immersed in water (Fig. adopts its CC3α-PAN structure in water, the narrower pore aperture 4e), with evidence of both phases found when the membrane was allowed only the smallest molecule, NP, to diffuse through the mem- immersed in a mixture of water and MeOH (Supplementary Figs. 35 brane, while the larger molecules, BB and DR, were rejected. Excess and 36). High-resolution PXRD also confirmed that CC3γ′ cleanly water was used for flushing residual NP from the retentate, and this transforms into CC3α after thermally desolvating a powdered sam- process was repeated until the NP concentration in the permeate was ple of CC3γ′ suspended in MeOH in a capillary (Supplementary below 1%. Subsequently, 90 vol% of MeOH was added into the water Fig. 37). retentate to generate a feedstock that transformed the membrane We next used MeOH rather than water to dissolve the dyes and structure to CC3γ′-PAN with the larger pore aperture. BB could filtered these solutions through the CC3-PAN membrane under the then diffuse through the membrane alone, while DR was retained −1 same conditions. Interestingly, the MWCO shifted from 600 g mol in the cell (Fig. 5b,d). Finally, excess MeOH was used to flush any −1 in water to 1,400 g mol in MeOH for the same membrane (Fig. residual BB from the cell to leave only DR in the retentate, where it 4b and Supplementary Figs. 38 and 39). By contrast, a commercial could be collected in pure form (Supplementary Section 1.4). Synder NDX nanofiltration membrane with a comparable MWCO −1 (500–700 g mol ) exhibited similar rejection behaviour in both water Conclusion and outlook and MeOH (Supplementary Figs. 40 and 41). We attribute this dra- Continuous, defect-free POC membranes can achieve high per- matic change in MWCO to the phase transformation to CC3γ′-PAN meances for a range of organic solvents—in some cases exceeding in MeOH. We further investigated how crystallinity influences the upper performance bounds—while also showing excellent separa- switchable pore aperture by measuring dye rejection of CC3-PAN tion performances. These highly ordered crystalline POC mem- membranes with lower crystallinity (fabricated using lower con- branes exhibit a switchable phase transition between two crystalline centrations or shorter reaction times; Supplementary Figs. 42–45). forms, CC3α-PAN and CC3γ′-PAN. This allows graded sieving to CC3-PAN-4 h-0.8% rejected 78.2% of brilliant blue (BB) dye from separate a mixture of three organic dyes using a single, smart mem- water compared to 52.7% from MeOH, while the less crystalline brane and creates a membrane-based parallel to the widespread and CC3-PAN-4 h-0.2% had a less distinct BB rejection performance highly effective use of solvent gradients in chromatography . POC (68.6% from water versus 52.8% from MeOH). Hence, the high crys- membranes with switchable pore apertures could also lead to new 53 54 tallinity in the CC3 membrane is essential for regulating its separation applications in triggered drug delivery , biosensors or fermenta- performance after switching its pore aperture using a solvent stimulus. tion/fractionation processes . We next performed molecular separations while cycling between While the current synthesis process makes it challenging to scale CC3α-PAN and CC3γ′-PAN using a single membrane and water and implement these POC membranes in commercial processes, and MeOH feedstocks containing the BB dye. We found that both it is conceivable that a more scalable production method might be water and MeOH permeances remained high after cycling between developed by exploiting the solution processability of these molecu- CC3α-PAN and CC3γ′-PAN (Fig. 4c, Supplementary Tables 5–7 lar cages. Future efforts will focus on using computational methods, and Supplementary Video 1). More importantly, the rejection of BB such as crystal structure prediction, to design POC crystals with switches between ~100% in water to ~0% in MeOH in each cycle specific properties that can be designed from first principles. (Fig. 4c,d); that is, the membrane can be switched ‘on’ and ‘off ’ using a solvent. The reversible transition between CC3α-PAN and Online content CC3γ′-PAN appears to be complete within the one minute it takes Any methods, additional references, Nature Research report- to switch the feedstock (Supplementary Video 1) and creates alter- ing summaries, source data, extended data, supplementary infor- native diffusion pathways through the membrane structure. From mation, acknowledgements, peer review information; details of NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials 469 Nature Materials Articles 33. Zhu, G., O’Nolan, D. & Lively, R. P. Molecularly mixed composite membranes: author contributions and competing interests; and statements of challenges and opportunities. Chem. Eur. J. 26, 3464–3473 (2020). data and code availability are available at https://doi.org/10.1038/ 34. Alexandre, P.-E. et al. A robust porous quinoline cage: transformation of a s41563-021-01168-z. [4+6] salicylimine cage by Povarov cyclization. Angew. Chem. Int. Ed. 59, 19675–19679 (2020). Received: 12 April 2021; Accepted: 11 November 2021; 35. Mastalerz, M., Schneider, M. W., Oppel, I. M. & Presly, O. A salicylbisimine cage compound with high surface area and selective CO /CH adsorption. Published online: 10 January 2022 2 4 Angew. Chem. Int. Ed. 50, 1046–1051 (2011). 36. Avellaneda, A. et al. Kinetically controlled porosity in a robust organic cage References material. Angew. Chem. Int. Ed. 52, 3746–3749 (2013). 1. Tozawa, T. et al. Porous organic cages. Nat. Mater. 8, 973–978 (2009). 37. Elbert, S. M. et al. Shape-persistent tetrahedral [4+6] boronic ester 2. Zhang, G. & Mastalerz, M. Organic cage compounds – from cages with different degrees of fluoride substitution. Chem. Eur. J. 24, shape-persistency to function. Chem. Soc. Rev. 43, 1934–1947 (2014). 11438–11443 (2018). 3. Jiang, S. et al. Molecular dynamics simulations of gas selectivity in amorphous 38. Suganuma, S. & Katada, N. Innovation of catalytic technology for upgrading porous molecular solids. J. Am. Chem. Soc. 135, 17818–17830 (2013). of crude oil in petroleum refinery. Fuel Process. Technol. 208, 106518 (2020). 4. Giri, N. et al. Liquids with permanent porosity. Nature 527, 216–220 (2015). 39. Jang, H.-Y. et al. Torlon® hollow fiber membranes for organic solvent reverse 5. Sturluson, A., Huynh, M. T., York, A. H. P. & Simon, C. M. Eigencages: learning osmosis separation of complex aromatic hydrocarbon mixtures. AIChE J. 65, a latent space of porous cage molecules. ACS Cent. Sci. 4, 1663–1676 (2018). e16757 (2019). 6. Jelfs, K. E. & Cooper, A. I. Molecular simulations to understand and to design 40. Lv, E., Ding, S., Lu, J., Yi, W. & Ding, J. Separation and purification of fatty porous organic molecules. Curr. Opin. Solid State Mater. Sci. 17, 19–30 (2013). acids by membrane technology: a critical review. Int. J. Chem. React. Eng. 18, 7. Little, M. A. & Cooper, A. I. The chemistry of porous organic molecular 20190224 (2020). materials. Adv. Funct. Mater. 30, 1909842 (2020). 41. Gilmer, C. M., Zvokel, C., Vick, A. & Bowden, N. B. Separation of saturated 8. Chen, L. et al. Separation of rare gases and chiral molecules by selective fatty acids and fatty acid methyl esters with epoxy nanofiltration membranes. binding in porous organic cages. Nat. Mater. 13, 954–960 (2014). RSC Adv. 7, 55626–55632 (2017). 9. Nunes, S. P. & Peinemann, K.-V. Membrane Technology: in the Chemical 42. Ghasemian, S., Sahari, M. A., Barzegar, M. & Gavlighi, H. A. Concentration Industry (John Wiley & Sons, 2006). of omega-3 polyunsaturated fatty acids by polymeric membrane. Int. J. Food 10. Kim, D., Salazar, O. R. & Nunes, S. P. Membrane manufacture for peptide Sci. Technol. 50, 2411–2418 (2015). separation. Green Chem. 18, 5151–5159 (2016). 43. Takahashi, D., Inomata, T. & Fukui, T. AJIPHASE®: a highly efficient synthetic 11. Marchetti, P., Jimenez Solomon, M. F., Szekely, G. & Livingston, A. G. method for one-pot peptide elongation in the solution phase by an Fmoc Molecular separation with organic solvent nanofiltration: a critical review. strategy. Angew. Chem. Int. Ed. 129, 7911–7915 (2017). Chem. Rev. 114, 10735–10806 (2014). 44. o Th mpson, K. A. et al. N-aryl–linked spirocyclic polymers for membrane 12. Sholl, D. S. & Lively, R. P. Seven chemical separations to change the world. separations of complex hydrocarbon mixtures. Science 369, 310–315 (2020). Nature 532, 435–437 (2016). 45. Sadrzadeh, M. & Bhattacharjee, S. Rational design of phase inversion 13. Elimelech, M. & Phillip, W. A. The future of seawater desalination: energy, membranes by tailoring thermodynamics and kinetics of casting solution technology, and the environment. Science 333, 712–717 (2011). using polymer additives. J. Memb. Sci. 441, 31–44 (2013). 14. Kabay, N., Arda, M., Trochimczuk, A. & Streat, M. Removal of chromate by 46. Lu, Y., Chen, T., Chen, X., Qiu, M. & Fan, Y. Fabrication of TiO -doped solvent impregnated resins (SIRs) stabilized by coating and chemical ZrO nanofiltration membranes by using a modified colloidal sol-gel process crosslinking. I. Batch-mode sorption studies. React. Funct. Polym. 59, and its application in simulative radioactive effluent. J. Memb. Sci. 514, 9–14 (2004). 476–486 (2016). 15. Jiang, Z., Karan, S. & Livingston, A. G. Water transport through ultrathin 47. Siew, W. E., Livingston, A. G., Ates, C. & Merschaert, A. Molecular polyamide nanolm fi s used for reverse osmosis. Adv. Mater. 30, 1705973 (2018). separation with an organic solvent nanofiltration cascade–augmenting 16. Cundy, C. S. & Cox, P. A. The hydrothermal synthesis of zeolites: precursors, membrane selectivity with process engineering. Chem. Eng. Sci. 90, intermediates and reaction mechanism. Micropor. Mesopor. Mat. 82, 299–310 (2013). 1–78 (2005). 48. Little, M. A., Chong, S. Y., Schmidtmann, M., Hasell, T. & Cooper, A. I. 17. o Th mas, A. Functional materials: from hard to soft porous frameworks. Guest control of structure in porous organic cages. Chem. Commun. 50, Angew. Chem. Int. Ed. 49, 8328–8344 (2010). 9465–9468 (2014). 18. Furukawa, H., Cordova, K. E., O’Keeffe, M. & Yaghi, O. M. The chemistry 49. Little, M. A. et al. Trapping virtual pores by crystal retro-engineering. and applications of metal-organic frameworks. Science 341, 1230444 (2013). Nat. Chem. 7, 153–159 (2015). 19. Côté, A. P. et al. Porous, crystalline, covalent organic frameworks. Science 50. Jones, J. T. A. et al. Modular and predictable assembly of porous organic 310, 1166–1170 (2005). molecular crystals. Nature 474, 367–371 (2011). 20. Lin, R.-B. et al. Multifunctional porous hydrogen-bonded organic framework 51. Karan, S., Jiang, Z. & Livingston, A. G. Sub–10 nm polyamide nanofilms materials. Chem. Soc. Rev. 48, 1362–1389 (2019). with ultrafast solvent transport for molecular separation. Science 348, 21. Dey, K. et al. Selective molecular separation by interfacially crystallized covalent 1347–1351 (2015). organic framework thin films. J. Am. Chem. Soc. 139, 13083–13091 (2017). 52. Aumann, L. & Morbidelli, M. A continuous multicolumn countercurrent 22. Matsumoto, M. et al. Lewis-acid-catalyzed interfacial polymerization of solvent gradient purification (MCSGP) process. Biotechnol. Bioeng. 98, covalent organic framework films. Chem 4, 308–317 (2018). 1043–1055 (2007). 23. Corcos, A. R. et al. Reducing the pore size of covalent organic frameworks in 53. Loftsson, T., Vogensen, S. B., Brewster, M. E. & Konráosdóttir, F. Effects of thin-film composite membranes enhances solute rejection. ACS Mater. Lett. 1, cyclodextrins on drug delivery through biological membranes. J. Pharm. Sci. 440–446 (2019). 96, 2532–2546 (2007). 24. Zhang, C., Wu, B.-H., Ma, M.-Q., Wang, Z. & Xu, Z.-K. Ultrathin metal/ 54. Kohli, P., Wirtz, M. & Martin, C. R. Nanotube membrane based biosensors. covalent–organic framework membranes towards ultimate separation. Chem. Electroanalysis 16, 9–18 (2004). Soc. Rev. 48, 3811–3841 (2019). 55. O’Brien, D. J. & Craig, J. C. Ethanol production in a continuous 25. Wang, Z., Zhang, S., Chen, Y., Zhang, Z. & Ma, S. Covalent organic fermentation/membrane pervaporation system. Appl. Microbiol. Biotechnol. frameworks for separation applications. Chem. Soc. Rev. 49, 708–735 (2020). 44, 699–704 (1996). 26. Venna, S. R. & Carreon, M. A. Metal organic framework membranes for carbon dioxide separation. Chem. Eng. Sci. 124, 3–19 (2015). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 27. Fenton, J. L., Burke, D. W., Qian, D., Olvera de la Cruz, M. & Dichtel, W. R. published maps and institutional affiliations. Polycrystalline covalent organic framework films act as adsorbents, not membranes. J. Am. Chem. Soc. 143, 1466–1473 (2021). Open Access This article is licensed under a Creative Commons 28. Jones, J. T. A. et al. On–off porosity switching in a molecular organic solid. Attribution 4.0 International License, which permits use, sharing, adap- Angew. Chem. Int. Ed. 50, 749–753 (2011). tation, distribution and reproduction in any medium or format, as long 29. Bera, S. et al. Porosity switching in polymorphic porous organic cages with as you give appropriate credit to the original author(s) and the source, provide a link to exceptional chemical stability. Angew. Chem. Int. Ed. 58, 4243–4247 (2019). the Creative Commons license, and indicate if changes were made. The images or other 30. Song, Q. et al. Porous organic cage thin films and molecular-sieving third party material in this article are included in the article’s Creative Commons license, membranes. Adv. Mater. 28, 2629–2637 (2016). unless indicated otherwise in a credit line to the material. If material is not included in 31. Brutschy, M., Schneider, M. W., Mastalerz, M. & Waldvogel, S. R. Porous the article’s Creative Commons license and your intended use is not permitted by statu- organic cage compounds as highly potent affinity materials for sensing by tory regulation or exceeds the permitted use, you will need to obtain permission directly quartz crystal microbalances. Adv. Mater. 24, 6049–6052 (2012). from the copyright holder. To view a copy of this license, visit http://creativecommons. 32. Bushell, A. F. et al. Nanoporous organic polymer/cage composite membranes. org/licenses/by/4.0/. Angew. Chem. Int. Ed. 52, 1253–1256 (2013). © The Author(s) 2021 NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials 470 Nature Materials Articles details). The retentate was collected after each measurement. Error bars (s.d.) Methods were calculated by the STDEV.P function using data obtained from at least three Interfacial synthesis of crystalline CC3 films. An aqueous solution of CHDA independent membranes. (0.26 g, 2.24 mmol, 0.8 wt%) in water (32 ml) was carefully layered on top of For the dye rejection measurements, a series of dye feedstock solutions a dichloromethane solution (30 ml) that contained TFB (0.24 g, 1.48 mmol, in different solvents (water, MeOH, acetone and acetonitrile) were prepared 0.8 wt%) and was stored in a glass dish with an inner diameter of 7.4 cm (Fig. 1a). with a dye concentration of 20 ppm using the following dyes: reactive red 120 e in Th terfacial reaction was covered and kept at room temperature (~19–21 °C) −1 −1 −1 −1 (1,470 g mol ), DR (1,373 g mol ), rose bengal (1,018 g mol ), BB (826 g mol ), for between 4 and 96 hours (typically, 24 hours). The continuous crystalline −1 −1 Congo red (697 g mol ), protoporphyrin IX disodium (607 g mol ), acid fuchsin CC3 film that grew at the dichloromethane–water interface was then isolated −1 −1 −1 (585 g mol ), sunset yellow (452 g mol ), methyl orange (327 g mol ), neutral as a free-standing film that could be layered directly onto different substrates, −1 −1 red (289 g mol ) and NP (139 g mol ; Supplementary Table 1 for full details). including glass, steel mesh, carbon tape and silicon wafers. To perform liquid Ultraviolet–visible spectroscopy was used to measure the dye concentration in permeation studies, the CC3 film was transferred onto a PAN support to form the permeate to calculate dye rejection performance. Dye rejection, R (%), of the composite CC3-PAN membranes, which were then soaked in pure water for 1 day membranes was calculated as follows: (Supplementary Figs. 48 and 49; Supplementary Section 1.2 for full experimental ( ) details and Supplementary Fig. 50 for the reaction set-up). Fabrication of PAN R%= 1 −C /C ×100 (3) p f supports via phase inversion is presented in Supplementary Section 1.2. where C and C represent the dye concentrations in the permeate (C ) and feed p f p X-ray diffraction. GIXRD measurements were performed using the I07 beamline (C ). Dye concentrations in the permeate and feed were determined using a Cary at Diamond Light Source in the United Kingdom (wavelength, λ = 0.689 Å), using 5000 ultraviolet/visible/near-infrared spectrometer with the wavelengths specified a vertical (2 + 2)-type diffractometer equipped with a Pilatus 100 K area detector . in Supplementary Table 1. The MWCO was determined by interpolating from Membrane samples were cut into 1 × 2 cm pieces and stuck onto glass supports, the plot of rejection against the molecular weight of the dyes and corresponds to which were then mounted on a hexapod (PI-Micos) to allow independent the molecular weight for which rejection is 90%. During these measurements, the alignment with six degrees of freedom during the data collection (Supplementary volume and the concentration of the permeate and the retentate were measured, Fig. 51a). The measurements were conducted by moving the detector while and the mass balance of the feed solution could be calculated as follows: maintaining a fixed sample position. The grazing incidence angle is set at 2°. Data collection was performed at room temperature using in-plane (over the 2θ range C ×V = C ×V +C ×V (4) p p r r f f 3–40°, 0.50° step size) and out-of-plane (over the 2θ range 2–40°, 0.25° step size) measurement geometries, and GIXRD scans were processed in DAWN 2 (ref. ). where C , C and C are the dye concentrations in parts per million (grams per f p r GIXRD patterns were refined by Pawley refinement through TOPAS Academic . litre) of the feed, permeate and retentate, respectively; V , V and V represent the f p r High-resolution synchrotron PXRD data were collected using the I11 beamline at volume of the feed, permeate and retentate in litres, respectively. Typically, 0.2 l Diamond Light Source (λ = 0.827 Å). The full PXRD details are presented in the of the feed solution was added into the cell, then 0.1 l permeate was collected and Supplementary Information. 0.1 l retentate was left in the cell. Reversible filtration tests, membrane absorption For the in situ GIXRD measurements performed on solvated samples, tests, long-term operation information, membrane stability tests, water and pieces of Mylar film were used to cover the membrane surface with a thin layer MeOH feedstock mixture separation experiments and graded sieving experiments of solvent (water, MeOH, acetone and acetonitrile) during the GIXRD scans for the ternary system are presented in Supplementary Section 1.4. The set-up of (Supplementary Fig. 51b). To investigate the reversible transformation between a commercial bench-scale dead-end stirred filtration unit with transparent cells CC3α-PAN and CC3γ′-PAN, a membrane sample was removed from water (a 50 ml transparent Merck Millipore Amicon dead-end stirred cell connected to without drying and covered with 1.0 ml of a MeOH solvent layer before recording an 800 ml Merck Millipore Amicon RC800 reservoir) is shown in Supplementary the GIXRD data (Supplementary Fig. 35). To more closely mimic the reversible Fig. 23. Reversible filtration measurement data in water and MeOH are shown in membrane separation experiment where the feedstock was cycled between Supplementary Tables 5–7, and dye rejection measurement data in a water and water and MeOH, a CC3-PAN sample was removed from water without drying, MeOH mixture are shown in Supplementary Table 8. soaked in 100 ml MeOH for 1 minute and covered with a thin layer of MeOH (1.0 ml) before the GIXRD measurement. The same process was repeated with Data availability the identical CC3-PAN sample using water or MeOH (Fig. 4e). For the in situ Source data are provided with this paper. In addition, source data are deposited in measurements performed using solvent vapours, nitrogen gas was bubbled −1 the University of Liverpool Research Data Catalogue (https://doi.org/10.17638/ through a 2 l bottle that contained the organic solvent at a flow rate of 10 l min . datacat.liverpool.ac.uk/1512). Further details can be obtained from the authors The ‘wet gas’ generated during this process was then continually flowed over the upon request. membrane sample during the full measurement and Mylar film was used to seal the sample environment. References Separation measurements. Solvent permeance and dye rejection measurements 56. Nicklin, C., Arnold, T., Rawle, J. & Warne, A. Diamond beamline I07: a were performed using a Sterlitech HP4750 dead-end membrane filtration system beamline for surface and interface diffraction. J. Synchrotron Radiat. 23, (Supplementary Fig. 22). We also used a commercial bench-scale 50 ml transparent 1245–1253 (2016). Merck Millipore Amicon dead-end stirred cell, which was connected to an 800 ml 57. Basham, M. et al. Data analysis workbench (DAWN). J. Synchrotron Radiat. Merck Millipore Amicon RC800 reservoir, to visualize the filtration process 22, 853–858 (2015). (Supplementary Fig. 23). During these measurements, the feedstocks were kept 58. Coelho, A. TOPAS Academic v.5 (Coelho Software, 2012). under a 10 bar nitrogen pressure (3 bar for Merck Millipore Amicon cells) at room temperature, and the feedstock was continually stirred using a stirring bar rotating Acknowledgements at 400 r.p.m. The Hansen solubility parameter (δ) and the physical properties of the A.I.C., M.A.L., M.E.B., Y.W. and A.H. acknowledge the Engineering and Physical organic solvents (Supplementary Table 2) were used to investigate the relationships Sciences Research Council (EP/N004884/1). A.I.C. and A.H. acknowledge the between pure solvent permeances and the combined solvent properties. Leverhulme Trust via the Leverhulme Research Centre for Functional Materials Design −2 −1 Flux J (l m h ) was calculated according to the following equation: for funding. A.H. acknowledges the China Scholarship Council for a studentship and the Royal Society of Chemistry for a Researcher Mobility Grant (M19–2442). Z.J. and J = ΔV/ (A ×Δt) (1) A.G.L. acknowledge the Engineering and Physical Sciences Research Council for funding where ∆V is the volume of permeate collected in litres in a given amount of time, A (EP/R018847/1). A.I.C., M.A.L., A.H., Y.W., H.H. and J.R acknowledge Diamond Light is the membrane surface area in square metres and ∆t is the time in hours between Source for access to beamlines I07 (SI24359) and I11 (CY23666) that contributed to the the start and end of the measurement. results presented here. We thank M. Liu for advice about the cage synthesis conditions, −2 −1 −1 Solvent permeance P (l m h bar ) was calculated according to the following H. Chen for assistance during the I07 measurements, P. Cui and H. Gao for assistance equation: during the I11 measurements, Y. Li for fitting the logical curve (Fig. 5a), R. Clowes for instrument support and setting up the filtration cell, H. Yang and T. Mitra for assistance P = ΔV/ (A ×Δt ×p) (2) collecting SEM images and K. Arnold for collecting the FIB-SEM images. where ∆V is the volume of permeate collected in litres in a given amount of Author contributions time, A is the membrane surface area in square metres, ∆t is the time in hours between the start and end of the measurement and p is the transmembrane A.I.C. and A.G.L. conceived the project. A.H. and M.A.L. conceived the graded molecular sieving part of the study and performed the associated measurements. pressure. To calculate solvent permeance, typically, 0.2 l of pure solvent or dye A.H. and Z.J. prepared the membranes. A.H. led the characterization and membrane feedstock (20 ppm dye concentration) was added to the feedstock tank. The cell was then pressurized to 10 bar under nitrogen. The solvent permeate was performance tests. Z.J. synthesized the PAN membrane support and conducted the AFM then calculated based on the amount of time it took ~0.1 l of pure solvent or dye measurements. M.E.B. helped with the cage synthesis. H.H. and J.R. designed and set up the in situ GIXRD system and helped with these measurements. A.H., M.A.L. and feedstock to flow through the membrane (Supplementary Tables 3 and 4 for full NAtuRE MA tERIALs | www.nature.com/naturematerials Nature Materials Articles Y.W. contributed to the GIXRD measurements and processed the X-ray data. A.G.L. Additional information provided facility support and insight into advanced membrane science. Z.J. designed the Supplementary information The online version contains supplementary material interfacial synthesis method for the membrane. M.A.L. cosupervised the project with available at https://doi.org/10.1038/s41563-021-01168-z. A.I.C. and A.G.L. and wrote the manuscript with A.H. and Z.J., with contributions from Correspondence and requests for materials should be addressed to all coauthors. All of the authors participated in the discussion of data and commented Andrew G. Livingston or Andrew I. Cooper. on the manuscript. Peer review information Nature Materials thanks Michael Mastalerz and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Competing interests Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing interests. NAtuRE MA tERIALs | www.nature.com/naturematerials http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nature Materials Springer Journals

A smart and responsive crystalline porous organic cage membrane with switchable pore apertures for graded molecular sieving

Loading next page...
 
/lp/springer-journals/a-smart-and-responsive-crystalline-porous-organic-cage-membrane-with-HFGaf7oquc

References (61)

Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2021
ISSN
1476-1122
eISSN
1476-4660
DOI
10.1038/s41563-021-01168-z
Publisher site
See Article on Publisher Site

Abstract

Articles https://doi.org/10.1038/s41563-021-01168-z A smart and responsive crystalline porous organic cage membrane with switchable pore apertures for graded molecular sieving 1,6 2,3,6 1 4 4 1 Ai He , Zhiwei Jiang , Yue Wu    , Hadeel Hussain , Jonathan Rawle , Michael E. Briggs , 1 2,3 1,5  ✉  ✉ Marc A. Little    , Andrew G. Livingston and Andrew I. Cooper    Membranes with high selectivity offer an attractive route to molecular separations, where technologies such as distillation and chromatography are energy intensive. However, it remains challenging to fine tune the structure and porosity in membranes, particularly to separate molecules of similar size. Here, we report a process for producing composite membranes that com- prise crystalline porous organic cage films fabricated by interfacial synthesis on a polyacrylonitrile support. These membranes −1 exhibit ultrafast solvent permeance and high rejection of organic dyes with molecular weights over 600 g mol . The crystal- line cage film is dynamic, and its pore aperture can be switched in methanol to generate larger pores that provide increased −1 methanol permeance and higher molecular weight cut-offs (1,400 g mol ). By varying the water/methanol ratio, the film can be switched between two phases that have different selectivities, such that a single, ‘smart’ crystalline membrane can perform graded molecular sieving. We exemplify this by separating three organic dyes in a single-stage, single-membrane process. 1,2 15 orous organic cages (POCs) are discrete molecules with polymerization . This produces amorphous polymer net- intrinsic cavities that can create porosity in molecular crys- works with a modest degree of pore tunability. There is a strong 1 3 4 Ptals , amorphous solids and porous liquids . The adsorp- demand to develop membranes with more tunable and modu- tion properties of POCs can sometimes be predicted in silico from lar pore structures. Various porous solids, including zeolites , 5,6 1,2 17 18 knowledge of their molecular structures in isolation . However, the POCs , organic polymers , metal–organic frameworks , cova- adsorption properties of POC materials are also affected by their lent organic frameworks (COFs) and hydrogen-bonded organic 2,7 20 solid-state packing . For example, extrinsic pores in POC crystals frameworks have been explored. Banerjee et al. reported COF can selectively adsorb guests, including rare gases . Indeed, ineffi- films with 1.4 to 2.6 nm pores that showed good performance cient packing of POCs can generate solids with considerably more in dye rejection . Dichtel et al. reported COF films with 3.4 nm 2,7 porosity than would be expected from the cage cavities alone . This pores and tunable thicknesses over the range of 100 μm to 2.5 nm combination of intrinsic and extrinsic porosity determines the func- that rejected Rhodamine WT from water . The same team also tionality of POC-based materials in selective adsorption processes. reduced the effective pore size of their COF membrane to 3.3 23 24,25 Most separation studies involving POCs have used molecular and 3.2 nm using reticular chemistry . In addition to COFs , 2,7 crystals , which can exhibit slow adsorption kinetics. Also, many metal–organic frameworks and their composites have been used 24,26 POC crystals rely on selective adsorption governed by thermo- to produce membranes . However, it remains challenging to dynamics, rather than kinetics, which limits their practical use produce continuous nanofiltration membranes with extended in size- and shape-selective membrane filters. Given their solu- porous frameworks that perform exclusively as size-based molec- tion processability, however, there is scope to develop crystalline ular sieves rather than selective adsorbents . POCs are solu- POC-based membranes that operate by selectively removing guests tion processable and their solid-state structures are defined by that are either too large or that have the wrong shape to diffuse non-covalent intermolecular interactions, which can be switched 28,29 through the POC pore structure. using chemical stimuli to alter their bulk porosity . As such, There is growing interest in membrane technologies that per- POCs are intriguing but relatively unexplored candidates for new 30–37 form industrial and environmentally relevant separations where types of membrane materials . two or more solutes are separated one from each other, as in distil- Many practically important molecular separations involve ter- lation or chromatography, as opposed to separations where a whole nary systems or more complex mixtures—for example, separating set of solutes is concentrated, such as in evaporation or seawater multiple hydrocarbon fractions from light crude oil by distillation, 9–13 38,39 reverse osmosis . A major advantage of membranes is that they pervaporation or organic solvent reverse osmosis ; purification 40,41 can perform separations in the liquid phase, which is often more of fatty acids , such as the practical recovery of omega-3 poly- practically useful than vapour phase separations. unsaturated fatty acids from fish oil by nanofiltration ; or sieving Membranes for liquid separations are typically produced using out by-products from reactions, for example in the liquid-phase 14 43 phase inversion, which can be followed by coating or interfacial peptide synthesis of pharmaceuticals . To achieve equivalent 1 2 Department of Chemistry and Materials Innovation Factory, University of Liverpool, Liverpool, UK. Department of Chemical Engineering, Imperial College 3 4 London, South Kensington, London, UK. School of Engineering and Materials Science, Queen Mary University of London, London, UK. Diamond Light 5 6 Source, Didcot, UK. Leverhulme Research Centre for Functional Materials Design, University of Liverpool, Liverpool, UK. These authors contributed equally: Ai He, Zhiwei Jiang. e-mail: a.livingston@qmul.ac.uk; aicooper@liverpool.ac.uk NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials 463 Nature Materials Articles Water MeOH Water Crystalline thin film CC3α phase in water CC3γ′ phase in MeOH b c CH Cl 2 2 MeOH Water CC3 ≡ ≡ CHDA TFB CH Cl Water MeOH 2 2 Intrinsic cavity CC3α CC3γ′ Extrinsic channel Fig. 1 | s ynthesis of a crystalline CC3 film and its crystal structures. a, Scheme showing the interfacial synthesis method used to fabricate crystalline CC3 films, which were subsequently adhered to a pAN support. These crystalline cage films can cycle between two different forms, CC3α-pAN and CC3γ′-pAN, by cycling the solvent between water and MeOH. CH Cl , dichloromethane. b, CC3α structure with its 3D pore network shown in yellow. c, The CC3γ′ 2 2 structure, formed by soaking in MeOH, has additional extrinsic solvent-filled channels, shown here in orange, that open up additional porosity in the membrane in response to the MeOH solvent. separations for complex mixtures using membranes, cascades of Fabrication of crystalline CC3 films membranes with graded molecular weight cut-offs (MWCOs) Continuous films with highly crystalline domains of CC3 were pro- have been developed , using phase inversion (polymeric mem- duced using a combined interfacial condensation reaction and crys- 45 46 branes) or sol–gel processing (ceramic membranes) by manip- tallization process at a water–dichloromethane interface (Fig. 1a). ulating the recipe for dope solution or fabrication conditions to This interfacial process allows the two-component reaction of CC3, produce multiple membranes with a variety of pore sizes. This which is synthesized via a [4 + 6] cycloimination reaction using places membranes at a disadvantage for ternary and higher sepa- 1,3,5-triformylbenzene (TFB) and (1R,2R)-1,2-diaminecyclohexane rations—by contrast, a single distillation or chromatography col- (CHDA), while simultaneously directing the formation of CC3 umn can produce multiple fractions with differing compositions. films at the interface (Methods). Continuous and free-standing CC3 Separating more than binary solute systems using a membrane films were transferred from the liquid–liquid interface onto various cascade requires multiple pumped recycle streams and complex substrates (for example, glass, steel mesh, carbon tape and silicon fluid controls . While solvent gradients are used in chromatog- wafers; Supplementary Fig. 1) for further analysis of the crystallin- raphy to modulate solid–liquid interactions, to the best of our ity and surface morphology. Before performing permeance and dye knowledge, there are as yet no reports of membranes that respond rejection studies, the CC3 film was coated onto a PAN support by to solvent gradients by changing their solute selectivity. filtration to form the composite membrane (Fig. 2a). The resulting Here, we report the fabrication of close-packed and defect-free membrane, referred to hereafter as CC3-PAN, was free of macro- films of a shape-persistent imine POC, CC3, which grow at the liq- scopic defects up to at least 7.4 cm in diameter using this prepa- uid–liquid interface between water and dichloromethane (Fig. 1a). ration process, with no evidence of delamination after cutting the These films comprise highly crystalline domains of CC3 in its most membrane into smaller pieces (Supplementary Fig. 2). The CC3 thermodynamically stable polymorph, CC3α (Fig. 1b). By coating film was characterized by Fourier transform infrared spectroscopy the CC3α film on polyacrylonitrile (PAN), we produce a continu- (Supplementary Fig. 3), Raman spectroscopy, nuclear magnetic res- ous membrane (CC3α-PAN) that has excellent permeance for both onance (NMR) spectroscopy (Supplementary Fig. 4), scanning elec- −2 −1 −1 polar and non-polar solvents, including water (43.0 l m h bar ) tron microscopy (SEM), focused ion beam SEM (FIB-SEM), X-ray −2 −1 −1 and toluene (55.9 l m h bar ). Furthermore, we found that it is diffraction and atomic force microscopy (AFM). For spectroscopic possible to rapidly and reversibly switch the membrane pore aper- measurements, a crystalline CC3α sample was used as a reference . ture using common solvents. Exposure of the non-covalent crystal CC3α has a three-dimensional (3D) diamondoid pore structure packing of CC3 in methanol (MeOH) induces a rapid phase transi- and is the thermodynamically most stable polymorph CC3 (ref. ). tion from CC3α to a different crystalline phase, CC3γ′, which is A Raman map was performed on an 80 × 80 μm CC3 film depos- less densely packed. This systematically increases the effective pore ited on glass (Fig. 2e,f and Supplementary Fig. 5), which indicated aperture of the resulting membrane, CC3γ′-PAN (Fig. 1c). This that the CC3 film comprised crystalline domains with the same switching property of the CC3γ′-PAN membrane allows the perme- solid-state structure as the CC3α polymorph (Fig. 2i). SEM images ation of larger organic dyes that can be rejected in water while the showed a continuous, apparently defect-free film in the CC3-PAN large pore apertures are turned ‘off ’ in the CC3α-PAN membrane. composite (Fig. 2b and Supplementary Fig. 6a) with a thickness This switchable porosity is reversible, and surprisingly, it does not of ~80 nm measured on a free-standing film (Fig. 2d), which con- compromise the continuity of the membrane. This allowed us to tained embedded, octahedral CC3 crystals (Supplementary Fig. 7). separate three organic dyes with different sizes via graded sieving Cross-sectional SEM images were obtained after step-by-step FIB using a single membrane. trenching and polishing of both CC3-PAN (Supplementary Fig. 8) NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials 464 Height (nm) Height (nm) Nature Materials Articles a b c CC3 film Si 100 CC3 film 500 nm 500 nm 20 CC3 film 500 nm PAN Height profile PAN support 80 nm 40 1 cm 500 nm 0 0.5 1.0 1.5 Distance (μm) e f g h 5 μm Stage 1: 4 h 20 μm 20 μm 5 μm CC3α film Stage 2: 16 h 48 h 24 h Crystalline CC3α 16 h reference 5 μm 8 h Stage 3: 24 h 4 h Amorphous CC3 reference 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 5 μm 2θ (°) 500 1,000 1,500 2,000 2,500 3,000 Stage 4: >48 h –1 Raman shift (cm ) Fig. 2 | Characterization of a CC3α film. a, photograph of composite membrane CC3α-pAN with a diameter of 7.4 cm. b, SEM image of CC3α-pAN showing the surface morphology of the CC3α film. Shown below is the cross-sectional FIB-SEM image of CC3α-pAN. c, AFM height image (top) and the height profile (bottom) of CC3α film transferred onto a silicon (Si) wafer. d, SEM image of a free-standing CC3α film, where the film was deliberately buckled to show its thickness. e,f, Raman microscope image (e) and Raman map (f) of a CC3α film on a glass support, where we purposely scratched the film before the measurement to expose the glass support (black stripe in f). The red regions on a CC3α film had comparable Raman spectra to the crystalline CC3α reference sample. g, SEM images of CC3α-pAN-X h-0.8% membranes formed at different reaction times, showing four stages of CC3α film formation. h, Out-of-plane GIXRD (wavelength, λ = 0.689 Å) patterns of CC3α-pAN-X h-0.8% membranes fabricated using reaction times between 4 and 48 hours 1 3 (2θ refers to the scattering angle). i, Raman spectra of CC3α film, a crystalline CC3α reference and an amorphous CC3 reference . and a CC3 film coated on a silicon wafer (Supplementary Fig. 9). The the CC3 film (Fig. 2g,h). This allowed us to create CC3 films from FIB-SEM images showed a clear boundary between the CC3 film the interfacial reaction that were four times thinner than the CC3 layered on top of the supports. After transferring the as-synthesized film created by spin coating . To further confirm the crystalline CC3 film onto a silicon wafer, we performed AFM measurements structure of the film, we performed a series of powder X-ray dif- to investigate the film thickness further. AFM again confirmed that fraction (PXRD) and grazing incidence X-ray diffraction (GIXRD) the CC3 film was continuous with a constant thickness of ~80 nm measurements on CC3-PAN (Methods). These diffraction mea- (Fig. 2c and Supplementary Figs. 6b, 8 and 9). surements revealed that the CC3 film was crystalline and had the A key advantage of interfacial synthesis is that it can create con- same structure as CC3α (Supplementary Figs. 10 and 11). 15,21 tinuous films of the product . Here, we also modified the reaction To further investigate the crystallization process of CC3 films at conditions to optimize the thickness, continuity and crystallinity of the solvent interface, we varied the reaction time from 4 to 96 hours NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials Intensity (a.u.) Intensity (counts) Nature Materials Articles a b c 1. Ethanol 7 2. Water 160 100 3. Toluene 6 80 4. Heptane 120 5. Hexane 6. Acetonitrile 4 60 7. Acetone Stage 1 Stage 3 4 h 24 h 2 48 h Stage 2 1 20 Stage 4 8 h 20 R = 0.9829 96 h 16 h 0 0 4 8 16 24 48 96 0 50 100 150 200 250 300 400 600 800 1,000 1,200 1,400 Reaction time (h) –2 –1 –1 δ η d Molecular weight (g mol ) d m d e f SO Na 100 3.0 Congo red 1,400 NH N N N N H2N 2.5 1,200 NaO3S 1,000 2.0 Retentate 1.5 Feed Permeate Applied pressure 1.0 Retentate 2 bar 20 bar Feed 5 bar 35 bar 0.5 Permeate 10 bar 0 0 0 5 10 15 20 25 30 35 400 600 800 1,000 1,200 1,400 400 500 600 700 800 –1 Wavelength (nm) Applied pressure (bar) Molecular weight (g mol ) Fig. 3 | Nanofiltration performance of CC3α membranes. a, plot showing pure solvent permeances versus their combined solvent properties (viscosity η, molar diameter d and solubility parameter δ ) for CC3α-pAN, where R is the coefficient of determination for the function. Hansen solubility parameter (δ) m d and the physical properties of each organic solvent are listed in Supplementary Table 2. b, Water permeance for CC3α-pAN-X h-0.8% membranes fabricated using reaction times that ranged between 4 and 96 hours. c, Dye rejection measurements for CC3α-pAN-X h-0.8% membranes in water. d,e, Water flux (d) and dye rejections (e) of a CC3α-pAN membrane under a range of applied pressures. f, Ultraviolet–visible absorption spectra of Congo red in water before (feed) and after (permeate and retentate) selectivity tests performed with CC3α-pAN. Insets show photographs of the feed, permeate and retentate solutions and the molecular structure of Congo red. Dye rejection was calculated using the intensity of the maximum absorption peak in the permeate and the feed, and equation (3) in the Methods. Mass balance calculations were performed using the maximum absorption peak values of the feed, permeate and retentate, with equation (4). All error bars depict the standard deviation (s.d.) of the data points obtained from at least three independent membranes. and manipulated the reagent concentrations from 0.2 to 2.5 wt%. the properties of CC3α-PAN-24 h-0.8%, referred to hereafter as We use the nomenclature CC3α-PAN-X h-Y% to refer to the mem- CC3α-PAN. branes made with X hours of reaction time and Y weight percent of the reagents. SEM, FIB-SEM and AFM revealed that thicker films Membrane performance of CC3α-PAN with larger crystals were produced as the reaction time and reagent To determine the permeance and dye rejection performance of concentrations were increased (Supplementary Figs. 12–21). By CC3α-PAN, we performed filtration experiments in dead-end cells contrast, using a reagent concentration of 0.2 wt% resulted in poorly using solvents and dyes with different sizes and chemical function- crystalline CC3 membranes (Supplementary Figs. 14 and 15). For alities (Supplementary Table 1 and Supplementary Figs. 22 and 23). the reactions with reagent concentrations of 0.8 wt%, the CC3 With a water contact angle of 94° (Supplementary Fig. 24), the film thickness increased with reaction time (30–600 nm from the CC3α-PAN membrane was stable in a range of polar and non-polar 4–60 h reactions; Supplementary Figs. 16 and 17), and the FIB-SEM solvents (Supplementary Fig. 25), proving that these solvents do images revealed triangle/octahedral-shaped crystals embedded not dissolve CC3. This led to ultrafast solvent permeances (Fig. 3a; in the CC3 films from the 32 and 48 h reactions (Supplementary Supplementary Fig. 26 for blank PAN data). We attribute this to the Figs. 18–20). By contrast, from the reactions with reagent concen- 3D interconnected porosity through the CC3α crystals in the film. trations of 2.0%, multiple CC3 films were found stacked on top of By comparison, an amorphous CC3 membrane prepared by spin one another (Supplementary Fig. 21). We suggest that the interfa- coating (Supplementary Section 1.2) exhibited a 30-fold lower sol- cial synthesis occurs in four stages (Fig. 2g): Stage 1 (0–4 hours), vent permeance under the same testing conditions (Supplementary interfacial polymerization of a continuous oligomeric film at the Fig. 27), although it should also be noted that the amorphous CC3 dichloromethane–water interface; Stage 2 (4–16 hours), self-sorting membrane was four times thicker . Apparently, the crystalline of the reactants and oligomers into the CC3α product and the for- CC3α-PAN provides sufficient robustness to support the intercon- mation of a partially reacted, semi-cage film; Stage 3 (24–48 hours), nected channels under high applied pressures. To further confirm crystallization of CC3α and the formation of octahedral crystals the importance of crystallinity, CC3α membranes with differ- in the film; and Stage 4 (48–96 hours), formation of defects in the ent crystallinity levels were fabricated at each of the four reaction film caused by larger octahedral crystals creating cracks and imper- stages simply by controlling the reaction time. A partially crystal- fections. GIXRD measurements demonstrated the crystallization line membrane (CC3α-PAN-8 h-0.8%) at Stage 2 exhibited a water −2 −1 −1 process across these stages, where the crystallinity increased with permeance of 3.0 l m h bar ; that is, an order of magnitude lower −2 −1 −1 a longer reaction time (Fig. 2h). We therefore focused attention on than the fully crystalline Stage 3 membrane (49.5 l m h bar for NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials –2 –1 Flux (l m h ) –2 –1 –1 Permeance (l m h bar ) Rejection (%) –2 –1 –1 Permeance (l m h bar ) Absorbance (a.u.) Rejection (%) –2 –1 –1 Permeance (l m h bar ) Nature Materials Articles a b CC3γ′ powdered reference Switchable pore aperture In water CC3 film in MeOH In MeOH CC3 film in water CC3 film in air CC3α powdered reference 0 200 400 600 800 1,000 1,200 1,400 1,600 2 3 4 5 6 –1 2θ (°) Molecular weight (g mol ) c d 40 60 1 2 3 4 5 Cycle 1 Cycle 2 Cycle 3 Cycle e f This work Cycle 2: in MeOH CC3γ′-PAN Upper bound CC3α-PAN MOF MC Cycle 2: in water CC3α-PAN NPs GO PTMSP Commercially available PA PEEK CC3γ′-PAN Cycle 1: in MeOH PEI PEO PANI PIM PI PIP Cycle 1: in water CC3α-PAN PE PAR 2 3 4 5 6 0 250 500 750 1,000 1,250 1,500 1,750 –1 2θ (°) Molecular weight cut-off (g mol ) Fig. 4 | X-ray diffraction characterization and switchable separation performance of CC3-PAN membranes. a, GIXRD of CC3α-pAN in air and water, and CC3γ′-pAN in MeOH. Experimental pXRD patterns of CC3α and CC3γ′ powders are included as references. b, MWCO curve for CC3α-pAN in water and CC3γ′-pAN in MeOH containing 20 ppm dye solutes. The MWCO was determined by interpolating from the plot of rejection against the molecular weight of the dyes and corresponds to the molecular weight for which rejection reaches 90%. All error bars depict the s.d. of the data points obtained from at least three independent membranes. c, Reversible dye rejection of BB and solvent permeance of the CC3-pAN membrane observed upon switching the feedstock solvent between water and MeOH. All error bars denote the s.d. for measurements from at least three independent membranes. d, photographs of CC3-pAN filtration dead-end cell captured from Supplementary Video 1 during the different cycles; BB is rejected in water by CC3α-pAN while CC3γ′-pAN does not reject BB in MeOH. e, In situ GIXRD patterns showing the reversible phase transition between CC3α-pAN and CC3γ′-pAN, by cycling between water and MeOH. f, Acetone permeance versus MWCO of general solutes in acetone for nanofiltration membranes reported in the literature and CC3α-pAN. MOF, metal–organic framework; MC, macrocycle; Nps, nanoparticles; GO, graphene oxide; pTMSp, poly(1-(trimethylsilyl)-1-propyne); pA, polyamide; pEI, polyethyleneimine; pANI, polyaniline; pI, polyimide; pE, polyethylene; pEEK, poly(ether ether ketone); pEO, poly(ethylene oxide); pIM, polymers of intrinsic microporosity; pIp, piperazine; pAR, polyacrylate (Supplementary Table 9 for full details). CC3α-PAN-48 h-0.8%) prepared with prolonged reaction times through the membranes (Fig. 3c). By comparison, amorphous (Fig. 3b). CC3α-PAN-8 h-0.8% and CC3α-PAN-48 h-0.8% exhib- oligomeric membranes produced in Stage 1 (CC3α-PAN-4 h-0.8%) ited the same MWCO, as determined by filtering a range of dyes and cracked, highly crystalline membranes produced in Stage 4 NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials Rejection (%) Intensity (a.u.) Intensity (a.u.) Water MeOH –2 –1 –1 Rejection (%) Acetone permeance (l m h bar ) MeOH Water Nature Materials Articles a b Mixture = NP + BB + DR NP (P) Permeate = NP Add MeOH BB (P) 100 80 60 40 20 BB Permeate = Water content in MeOH (vol%) DR (R) Retentate = DR c d In water In water NP NP BB Mixture NP (P) DR In MeOH MeOH added (90 vol%) Mixture BB BB BB (P) DR DR (R) 300 400 500 600 700 800 Wavelength (nm) Fig. 5 | Mixture fitting and graded sieving using a single switchable membrane. a, BB rejection in mixtures of water and MeOH (v/v) for CC3-pAN (top), and photographs of the permeates (bottom). All error bars depict the s.d. of the data points obtained from at least three independent membranes. The red dashed line was fitted as the logistic function (y = 1/(1 + exp(−16.1(x – 0.617))); Supplementary Section 1.4). b, photographs showing the ternary molecular separation in a filtration dead-end cell, the nascent mixture feedstock, the permeate (p) collected in the first and second step, and the retentate (R) collected in the second step. c, Scheme showing ternary molecular separation of three dyes (DR, BB and Np) using one single membrane (CC3-pAN) in a continuous process: Step 1, CC3α-pAN in water (blue background) allows permeation of only Np, leaving BB and DR in the retentate. Step 2, 90 vol% MeOH (green background) was added into the retentate to transform the membrane structure to CC3γ′-pAN, which allows permeation of only BB, leaving DR in the retentate. d, Ultraviolet–visible absorption spectra of the mixture containing three molecules in water, permeate from water, mixture and permeate from 90 vol% of MeOH in water and the remaining retentate. Note, the maximum absorbance wavelength for BB is 551 nm in water and 587 nm in MeOH; BB also shows absorbance at 305 nm in MeOH, while Np shows its maximum absorbance at 312 nm in the same solvent. (CC3α-PAN-96 h-0.8%) exhibited unexpectedly higher water per- CC3α-PAN membranes might be more competitive for separations meances but failed to achieve comparable separation performances, that require higher pressures. indicating that they contained physical defects. To confirm that dye adsorption did not contribute to the selectiv- Two limitations of membranes produced from other crystalline ity performance of CC3α-PAN, mass balance calculations were used porous materials, such as COFs, are poor stability at high pressures to measure the dye concentration in the retentate. After permeating and the interference of adsorption processes . Here, the CC3α-PAN 48 ml of Congo red from the 100 ml feed, the absorption intensity membrane was tested under a range of applied pressures, up to a of Congo red in the retentate increased from 1.24 to 2.53, while its maximum of 35 bar. The water flux increased linearly with increas- absorption intensity in the permeate was only 0.02. In combination, ing applied pressure (Fig. 3d) without affecting the MWCO these values are consistent with ~100% dye rejection (Fig. 3f and (Fig. 3e). Longer duration studies demonstrated the mechani- Supplementary Figs. 29 and 30). These measurements agree with cal robustness of CC3α-PAN and showed consistent dye rejection the colourless membrane surface observed after the dye filtration −2 −1 −1 (99.7% for rose bengal) and water permeance (~43 l m h bar ) (Supplementary Fig. 2). In addition, soaking powdered crystals of over 20 hours (Supplementary Fig. 28). The applied pressure of CC3α (100 mg) in the dye solution (100 ml) did not lead to adsorp- 35 bar is an order of magnitude higher than that used for liquid tion in the crystals after seven days (Supplementary Fig. 31). These filtration through COF membranes , which suggests that these results all indicate that the dyes were rejected by the membrane. NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials Rejection (%) Absorbance (a.u.) Nature Materials Articles s witchable pore aperture for graded sieving in situ GIXRD measurements on solvated CC3 films, while per- Previous studies showed that certain POCs can be switched between forming two consecutive cycles, we found that the composite mem- 28,29 multiple polymorphs to modify their porosity . The solid-state brane transformed cleanly between CC3α-PAN and CC3γ′-PAN structure of CC3 has been directed into different polymorphs by when the solvent was switched between water and MeOH and back crystallization from specific solvents , but until now, the solid-state again (Fig. 4e). We attributed this switching phenomenon solely to transformation of CC3 crystals was not explored, to the best of our the phase transition of CC3 films, rather than swelling of the mem- knowledge. branes. To validate this, the CC3-PAN membrane was soaked in We found that both air-dried and water-solvated films exhibited acetone and acetonitrile, and the nanofiltration tests were repeated. the same diffraction patterns as the reference peaks of CC3α pow- CC3-PAN exhibits comparable MWCOs in acetone and acetoni- ders measured by PXRD (Fig. 4a). A series of GIXRD patterns were trile to those observed in water (Supplementary Figs. 41 and 46) then recorded after submerging the membrane in various organic because the same phase, CC3α-PAN, is present in these solvents solvents (Supplementary Figs. 32 and 33). The crystalline CC3α (Supplementary Figs. 32 and 33). Remarkably, the acetone perme- −2 −1 −1 film transformed into a new structure when submerged in MeOH ance of CC3α-PAN reached 177 l m h bar with a MWCO of −1 (Fig. 4a). By indexing the GIXRD pattern, we confirmed this was ~600 g mol , which is well above the upper bound performance a MeOH-solvated CC3 phase (CC3γ′; Supplementary Fig. 34) that for nanofiltration membranes reported in the literature (Fig. 4f, was isolated previously by crystallizing CC3 from dichloromethane Supplementary Fig. 47 and Supplementary Table 9). and MeOH (ref. ). The CC3γ′ structure is very different from its A series of water and MeOH feedstocks containing the dye BB thermodynamically most stable polymorph, CC3α (ref. ), where were used to determine the dynamic transformation between the cage packs in a window-to-window arrangement to generate a CC3α-PAN and CC3γ′-PAN (Fig. 5a). Understanding this dynamic diamondoid pore network (yellow channels in Fig. 1b). By contrast, transformation allowed us to manipulate the pore aperture in a sin- the CC3 molecules in the CC3γ′ phase are packed less densely, thus gle CC3-PAN membrane by simply adjusting the water concentra- providing large extrinsic pores between hexagonally arranged CC3 tion in a water–MeOH mixture, without any activation processes molecules (orange channels in Fig. 1c). or the use of multiple membranes . To demonstrate this, we per- To investigate the structural transformation between CC3α-PAN formed a graded sieving experiment to separate molecules from a and CC3γ′-PAN, we performed a series of in situ GIXRD measure- ternary mixture using a single membrane. Initially, a water feedstock −1 ments while dosing the membrane surface with solvent vapour containing three dyes, 4-nitrophenol (NP; yellow, 139 g mol ), BB −1 −1 and after coating the membrane surface in a thin solvent layer (blue, 826 g mol ) and direct red 80 (DR; red, 1,373 g mol ) was fil- (Methods). CC3γ′-PAN formed by immersion in MeOH trans- tered through the CC3-PAN membrane (Fig. 5c). Since CC3-PAN formed back into CC3α-PAN after being immersed in water (Fig. adopts its CC3α-PAN structure in water, the narrower pore aperture 4e), with evidence of both phases found when the membrane was allowed only the smallest molecule, NP, to diffuse through the mem- immersed in a mixture of water and MeOH (Supplementary Figs. 35 brane, while the larger molecules, BB and DR, were rejected. Excess and 36). High-resolution PXRD also confirmed that CC3γ′ cleanly water was used for flushing residual NP from the retentate, and this transforms into CC3α after thermally desolvating a powdered sam- process was repeated until the NP concentration in the permeate was ple of CC3γ′ suspended in MeOH in a capillary (Supplementary below 1%. Subsequently, 90 vol% of MeOH was added into the water Fig. 37). retentate to generate a feedstock that transformed the membrane We next used MeOH rather than water to dissolve the dyes and structure to CC3γ′-PAN with the larger pore aperture. BB could filtered these solutions through the CC3-PAN membrane under the then diffuse through the membrane alone, while DR was retained −1 same conditions. Interestingly, the MWCO shifted from 600 g mol in the cell (Fig. 5b,d). Finally, excess MeOH was used to flush any −1 in water to 1,400 g mol in MeOH for the same membrane (Fig. residual BB from the cell to leave only DR in the retentate, where it 4b and Supplementary Figs. 38 and 39). By contrast, a commercial could be collected in pure form (Supplementary Section 1.4). Synder NDX nanofiltration membrane with a comparable MWCO −1 (500–700 g mol ) exhibited similar rejection behaviour in both water Conclusion and outlook and MeOH (Supplementary Figs. 40 and 41). We attribute this dra- Continuous, defect-free POC membranes can achieve high per- matic change in MWCO to the phase transformation to CC3γ′-PAN meances for a range of organic solvents—in some cases exceeding in MeOH. We further investigated how crystallinity influences the upper performance bounds—while also showing excellent separa- switchable pore aperture by measuring dye rejection of CC3-PAN tion performances. These highly ordered crystalline POC mem- membranes with lower crystallinity (fabricated using lower con- branes exhibit a switchable phase transition between two crystalline centrations or shorter reaction times; Supplementary Figs. 42–45). forms, CC3α-PAN and CC3γ′-PAN. This allows graded sieving to CC3-PAN-4 h-0.8% rejected 78.2% of brilliant blue (BB) dye from separate a mixture of three organic dyes using a single, smart mem- water compared to 52.7% from MeOH, while the less crystalline brane and creates a membrane-based parallel to the widespread and CC3-PAN-4 h-0.2% had a less distinct BB rejection performance highly effective use of solvent gradients in chromatography . POC (68.6% from water versus 52.8% from MeOH). Hence, the high crys- membranes with switchable pore apertures could also lead to new 53 54 tallinity in the CC3 membrane is essential for regulating its separation applications in triggered drug delivery , biosensors or fermenta- performance after switching its pore aperture using a solvent stimulus. tion/fractionation processes . We next performed molecular separations while cycling between While the current synthesis process makes it challenging to scale CC3α-PAN and CC3γ′-PAN using a single membrane and water and implement these POC membranes in commercial processes, and MeOH feedstocks containing the BB dye. We found that both it is conceivable that a more scalable production method might be water and MeOH permeances remained high after cycling between developed by exploiting the solution processability of these molecu- CC3α-PAN and CC3γ′-PAN (Fig. 4c, Supplementary Tables 5–7 lar cages. Future efforts will focus on using computational methods, and Supplementary Video 1). More importantly, the rejection of BB such as crystal structure prediction, to design POC crystals with switches between ~100% in water to ~0% in MeOH in each cycle specific properties that can be designed from first principles. (Fig. 4c,d); that is, the membrane can be switched ‘on’ and ‘off ’ using a solvent. The reversible transition between CC3α-PAN and Online content CC3γ′-PAN appears to be complete within the one minute it takes Any methods, additional references, Nature Research report- to switch the feedstock (Supplementary Video 1) and creates alter- ing summaries, source data, extended data, supplementary infor- native diffusion pathways through the membrane structure. From mation, acknowledgements, peer review information; details of NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials 469 Nature Materials Articles 33. Zhu, G., O’Nolan, D. & Lively, R. P. Molecularly mixed composite membranes: author contributions and competing interests; and statements of challenges and opportunities. Chem. Eur. J. 26, 3464–3473 (2020). data and code availability are available at https://doi.org/10.1038/ 34. Alexandre, P.-E. et al. A robust porous quinoline cage: transformation of a s41563-021-01168-z. [4+6] salicylimine cage by Povarov cyclization. Angew. Chem. Int. Ed. 59, 19675–19679 (2020). Received: 12 April 2021; Accepted: 11 November 2021; 35. Mastalerz, M., Schneider, M. W., Oppel, I. M. & Presly, O. A salicylbisimine cage compound with high surface area and selective CO /CH adsorption. Published online: 10 January 2022 2 4 Angew. Chem. Int. Ed. 50, 1046–1051 (2011). 36. Avellaneda, A. et al. Kinetically controlled porosity in a robust organic cage References material. Angew. Chem. Int. Ed. 52, 3746–3749 (2013). 1. Tozawa, T. et al. Porous organic cages. Nat. Mater. 8, 973–978 (2009). 37. Elbert, S. M. et al. Shape-persistent tetrahedral [4+6] boronic ester 2. Zhang, G. & Mastalerz, M. Organic cage compounds – from cages with different degrees of fluoride substitution. Chem. Eur. J. 24, shape-persistency to function. Chem. Soc. Rev. 43, 1934–1947 (2014). 11438–11443 (2018). 3. Jiang, S. et al. Molecular dynamics simulations of gas selectivity in amorphous 38. Suganuma, S. & Katada, N. Innovation of catalytic technology for upgrading porous molecular solids. J. Am. Chem. Soc. 135, 17818–17830 (2013). of crude oil in petroleum refinery. Fuel Process. Technol. 208, 106518 (2020). 4. Giri, N. et al. Liquids with permanent porosity. Nature 527, 216–220 (2015). 39. Jang, H.-Y. et al. Torlon® hollow fiber membranes for organic solvent reverse 5. Sturluson, A., Huynh, M. T., York, A. H. P. & Simon, C. M. Eigencages: learning osmosis separation of complex aromatic hydrocarbon mixtures. AIChE J. 65, a latent space of porous cage molecules. ACS Cent. Sci. 4, 1663–1676 (2018). e16757 (2019). 6. Jelfs, K. E. & Cooper, A. I. Molecular simulations to understand and to design 40. Lv, E., Ding, S., Lu, J., Yi, W. & Ding, J. Separation and purification of fatty porous organic molecules. Curr. Opin. Solid State Mater. Sci. 17, 19–30 (2013). acids by membrane technology: a critical review. Int. J. Chem. React. Eng. 18, 7. Little, M. A. & Cooper, A. I. The chemistry of porous organic molecular 20190224 (2020). materials. Adv. Funct. Mater. 30, 1909842 (2020). 41. Gilmer, C. M., Zvokel, C., Vick, A. & Bowden, N. B. Separation of saturated 8. Chen, L. et al. Separation of rare gases and chiral molecules by selective fatty acids and fatty acid methyl esters with epoxy nanofiltration membranes. binding in porous organic cages. Nat. Mater. 13, 954–960 (2014). RSC Adv. 7, 55626–55632 (2017). 9. Nunes, S. P. & Peinemann, K.-V. Membrane Technology: in the Chemical 42. Ghasemian, S., Sahari, M. A., Barzegar, M. & Gavlighi, H. A. Concentration Industry (John Wiley & Sons, 2006). of omega-3 polyunsaturated fatty acids by polymeric membrane. Int. J. Food 10. Kim, D., Salazar, O. R. & Nunes, S. P. Membrane manufacture for peptide Sci. Technol. 50, 2411–2418 (2015). separation. Green Chem. 18, 5151–5159 (2016). 43. Takahashi, D., Inomata, T. & Fukui, T. AJIPHASE®: a highly efficient synthetic 11. Marchetti, P., Jimenez Solomon, M. F., Szekely, G. & Livingston, A. G. method for one-pot peptide elongation in the solution phase by an Fmoc Molecular separation with organic solvent nanofiltration: a critical review. strategy. Angew. Chem. Int. Ed. 129, 7911–7915 (2017). Chem. Rev. 114, 10735–10806 (2014). 44. o Th mpson, K. A. et al. N-aryl–linked spirocyclic polymers for membrane 12. Sholl, D. S. & Lively, R. P. Seven chemical separations to change the world. separations of complex hydrocarbon mixtures. Science 369, 310–315 (2020). Nature 532, 435–437 (2016). 45. Sadrzadeh, M. & Bhattacharjee, S. Rational design of phase inversion 13. Elimelech, M. & Phillip, W. A. The future of seawater desalination: energy, membranes by tailoring thermodynamics and kinetics of casting solution technology, and the environment. Science 333, 712–717 (2011). using polymer additives. J. Memb. Sci. 441, 31–44 (2013). 14. Kabay, N., Arda, M., Trochimczuk, A. & Streat, M. Removal of chromate by 46. Lu, Y., Chen, T., Chen, X., Qiu, M. & Fan, Y. Fabrication of TiO -doped solvent impregnated resins (SIRs) stabilized by coating and chemical ZrO nanofiltration membranes by using a modified colloidal sol-gel process crosslinking. I. Batch-mode sorption studies. React. Funct. Polym. 59, and its application in simulative radioactive effluent. J. Memb. Sci. 514, 9–14 (2004). 476–486 (2016). 15. Jiang, Z., Karan, S. & Livingston, A. G. Water transport through ultrathin 47. Siew, W. E., Livingston, A. G., Ates, C. & Merschaert, A. Molecular polyamide nanolm fi s used for reverse osmosis. Adv. Mater. 30, 1705973 (2018). separation with an organic solvent nanofiltration cascade–augmenting 16. Cundy, C. S. & Cox, P. A. The hydrothermal synthesis of zeolites: precursors, membrane selectivity with process engineering. Chem. Eng. Sci. 90, intermediates and reaction mechanism. Micropor. Mesopor. Mat. 82, 299–310 (2013). 1–78 (2005). 48. Little, M. A., Chong, S. Y., Schmidtmann, M., Hasell, T. & Cooper, A. I. 17. o Th mas, A. Functional materials: from hard to soft porous frameworks. Guest control of structure in porous organic cages. Chem. Commun. 50, Angew. Chem. Int. Ed. 49, 8328–8344 (2010). 9465–9468 (2014). 18. Furukawa, H., Cordova, K. E., O’Keeffe, M. & Yaghi, O. M. The chemistry 49. Little, M. A. et al. Trapping virtual pores by crystal retro-engineering. and applications of metal-organic frameworks. Science 341, 1230444 (2013). Nat. Chem. 7, 153–159 (2015). 19. Côté, A. P. et al. Porous, crystalline, covalent organic frameworks. Science 50. Jones, J. T. A. et al. Modular and predictable assembly of porous organic 310, 1166–1170 (2005). molecular crystals. Nature 474, 367–371 (2011). 20. Lin, R.-B. et al. Multifunctional porous hydrogen-bonded organic framework 51. Karan, S., Jiang, Z. & Livingston, A. G. Sub–10 nm polyamide nanofilms materials. Chem. Soc. Rev. 48, 1362–1389 (2019). with ultrafast solvent transport for molecular separation. Science 348, 21. Dey, K. et al. Selective molecular separation by interfacially crystallized covalent 1347–1351 (2015). organic framework thin films. J. Am. Chem. Soc. 139, 13083–13091 (2017). 52. Aumann, L. & Morbidelli, M. A continuous multicolumn countercurrent 22. Matsumoto, M. et al. Lewis-acid-catalyzed interfacial polymerization of solvent gradient purification (MCSGP) process. Biotechnol. Bioeng. 98, covalent organic framework films. Chem 4, 308–317 (2018). 1043–1055 (2007). 23. Corcos, A. R. et al. Reducing the pore size of covalent organic frameworks in 53. Loftsson, T., Vogensen, S. B., Brewster, M. E. & Konráosdóttir, F. Effects of thin-film composite membranes enhances solute rejection. ACS Mater. Lett. 1, cyclodextrins on drug delivery through biological membranes. J. Pharm. Sci. 440–446 (2019). 96, 2532–2546 (2007). 24. Zhang, C., Wu, B.-H., Ma, M.-Q., Wang, Z. & Xu, Z.-K. Ultrathin metal/ 54. Kohli, P., Wirtz, M. & Martin, C. R. Nanotube membrane based biosensors. covalent–organic framework membranes towards ultimate separation. Chem. Electroanalysis 16, 9–18 (2004). Soc. Rev. 48, 3811–3841 (2019). 55. O’Brien, D. J. & Craig, J. C. Ethanol production in a continuous 25. Wang, Z., Zhang, S., Chen, Y., Zhang, Z. & Ma, S. Covalent organic fermentation/membrane pervaporation system. Appl. Microbiol. Biotechnol. frameworks for separation applications. Chem. Soc. Rev. 49, 708–735 (2020). 44, 699–704 (1996). 26. Venna, S. R. & Carreon, M. A. Metal organic framework membranes for carbon dioxide separation. Chem. Eng. Sci. 124, 3–19 (2015). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 27. Fenton, J. L., Burke, D. W., Qian, D., Olvera de la Cruz, M. & Dichtel, W. R. published maps and institutional affiliations. Polycrystalline covalent organic framework films act as adsorbents, not membranes. J. Am. Chem. Soc. 143, 1466–1473 (2021). Open Access This article is licensed under a Creative Commons 28. Jones, J. T. A. et al. On–off porosity switching in a molecular organic solid. Attribution 4.0 International License, which permits use, sharing, adap- Angew. Chem. Int. Ed. 50, 749–753 (2011). tation, distribution and reproduction in any medium or format, as long 29. Bera, S. et al. Porosity switching in polymorphic porous organic cages with as you give appropriate credit to the original author(s) and the source, provide a link to exceptional chemical stability. Angew. Chem. Int. Ed. 58, 4243–4247 (2019). the Creative Commons license, and indicate if changes were made. The images or other 30. Song, Q. et al. Porous organic cage thin films and molecular-sieving third party material in this article are included in the article’s Creative Commons license, membranes. Adv. Mater. 28, 2629–2637 (2016). unless indicated otherwise in a credit line to the material. If material is not included in 31. Brutschy, M., Schneider, M. W., Mastalerz, M. & Waldvogel, S. R. Porous the article’s Creative Commons license and your intended use is not permitted by statu- organic cage compounds as highly potent affinity materials for sensing by tory regulation or exceeds the permitted use, you will need to obtain permission directly quartz crystal microbalances. Adv. Mater. 24, 6049–6052 (2012). from the copyright holder. To view a copy of this license, visit http://creativecommons. 32. Bushell, A. F. et al. Nanoporous organic polymer/cage composite membranes. org/licenses/by/4.0/. Angew. Chem. Int. Ed. 52, 1253–1256 (2013). © The Author(s) 2021 NAtuRE MA tERIALs | VOL 21 | ApRIL 2022 | 463–470 | www.nature.com/naturematerials 470 Nature Materials Articles details). The retentate was collected after each measurement. Error bars (s.d.) Methods were calculated by the STDEV.P function using data obtained from at least three Interfacial synthesis of crystalline CC3 films. An aqueous solution of CHDA independent membranes. (0.26 g, 2.24 mmol, 0.8 wt%) in water (32 ml) was carefully layered on top of For the dye rejection measurements, a series of dye feedstock solutions a dichloromethane solution (30 ml) that contained TFB (0.24 g, 1.48 mmol, in different solvents (water, MeOH, acetone and acetonitrile) were prepared 0.8 wt%) and was stored in a glass dish with an inner diameter of 7.4 cm (Fig. 1a). with a dye concentration of 20 ppm using the following dyes: reactive red 120 e in Th terfacial reaction was covered and kept at room temperature (~19–21 °C) −1 −1 −1 −1 (1,470 g mol ), DR (1,373 g mol ), rose bengal (1,018 g mol ), BB (826 g mol ), for between 4 and 96 hours (typically, 24 hours). The continuous crystalline −1 −1 Congo red (697 g mol ), protoporphyrin IX disodium (607 g mol ), acid fuchsin CC3 film that grew at the dichloromethane–water interface was then isolated −1 −1 −1 (585 g mol ), sunset yellow (452 g mol ), methyl orange (327 g mol ), neutral as a free-standing film that could be layered directly onto different substrates, −1 −1 red (289 g mol ) and NP (139 g mol ; Supplementary Table 1 for full details). including glass, steel mesh, carbon tape and silicon wafers. To perform liquid Ultraviolet–visible spectroscopy was used to measure the dye concentration in permeation studies, the CC3 film was transferred onto a PAN support to form the permeate to calculate dye rejection performance. Dye rejection, R (%), of the composite CC3-PAN membranes, which were then soaked in pure water for 1 day membranes was calculated as follows: (Supplementary Figs. 48 and 49; Supplementary Section 1.2 for full experimental ( ) details and Supplementary Fig. 50 for the reaction set-up). Fabrication of PAN R%= 1 −C /C ×100 (3) p f supports via phase inversion is presented in Supplementary Section 1.2. where C and C represent the dye concentrations in the permeate (C ) and feed p f p X-ray diffraction. GIXRD measurements were performed using the I07 beamline (C ). Dye concentrations in the permeate and feed were determined using a Cary at Diamond Light Source in the United Kingdom (wavelength, λ = 0.689 Å), using 5000 ultraviolet/visible/near-infrared spectrometer with the wavelengths specified a vertical (2 + 2)-type diffractometer equipped with a Pilatus 100 K area detector . in Supplementary Table 1. The MWCO was determined by interpolating from Membrane samples were cut into 1 × 2 cm pieces and stuck onto glass supports, the plot of rejection against the molecular weight of the dyes and corresponds to which were then mounted on a hexapod (PI-Micos) to allow independent the molecular weight for which rejection is 90%. During these measurements, the alignment with six degrees of freedom during the data collection (Supplementary volume and the concentration of the permeate and the retentate were measured, Fig. 51a). The measurements were conducted by moving the detector while and the mass balance of the feed solution could be calculated as follows: maintaining a fixed sample position. The grazing incidence angle is set at 2°. Data collection was performed at room temperature using in-plane (over the 2θ range C ×V = C ×V +C ×V (4) p p r r f f 3–40°, 0.50° step size) and out-of-plane (over the 2θ range 2–40°, 0.25° step size) measurement geometries, and GIXRD scans were processed in DAWN 2 (ref. ). where C , C and C are the dye concentrations in parts per million (grams per f p r GIXRD patterns were refined by Pawley refinement through TOPAS Academic . litre) of the feed, permeate and retentate, respectively; V , V and V represent the f p r High-resolution synchrotron PXRD data were collected using the I11 beamline at volume of the feed, permeate and retentate in litres, respectively. Typically, 0.2 l Diamond Light Source (λ = 0.827 Å). The full PXRD details are presented in the of the feed solution was added into the cell, then 0.1 l permeate was collected and Supplementary Information. 0.1 l retentate was left in the cell. Reversible filtration tests, membrane absorption For the in situ GIXRD measurements performed on solvated samples, tests, long-term operation information, membrane stability tests, water and pieces of Mylar film were used to cover the membrane surface with a thin layer MeOH feedstock mixture separation experiments and graded sieving experiments of solvent (water, MeOH, acetone and acetonitrile) during the GIXRD scans for the ternary system are presented in Supplementary Section 1.4. The set-up of (Supplementary Fig. 51b). To investigate the reversible transformation between a commercial bench-scale dead-end stirred filtration unit with transparent cells CC3α-PAN and CC3γ′-PAN, a membrane sample was removed from water (a 50 ml transparent Merck Millipore Amicon dead-end stirred cell connected to without drying and covered with 1.0 ml of a MeOH solvent layer before recording an 800 ml Merck Millipore Amicon RC800 reservoir) is shown in Supplementary the GIXRD data (Supplementary Fig. 35). To more closely mimic the reversible Fig. 23. Reversible filtration measurement data in water and MeOH are shown in membrane separation experiment where the feedstock was cycled between Supplementary Tables 5–7, and dye rejection measurement data in a water and water and MeOH, a CC3-PAN sample was removed from water without drying, MeOH mixture are shown in Supplementary Table 8. soaked in 100 ml MeOH for 1 minute and covered with a thin layer of MeOH (1.0 ml) before the GIXRD measurement. The same process was repeated with Data availability the identical CC3-PAN sample using water or MeOH (Fig. 4e). For the in situ Source data are provided with this paper. In addition, source data are deposited in measurements performed using solvent vapours, nitrogen gas was bubbled −1 the University of Liverpool Research Data Catalogue (https://doi.org/10.17638/ through a 2 l bottle that contained the organic solvent at a flow rate of 10 l min . datacat.liverpool.ac.uk/1512). Further details can be obtained from the authors The ‘wet gas’ generated during this process was then continually flowed over the upon request. membrane sample during the full measurement and Mylar film was used to seal the sample environment. References Separation measurements. Solvent permeance and dye rejection measurements 56. Nicklin, C., Arnold, T., Rawle, J. & Warne, A. Diamond beamline I07: a were performed using a Sterlitech HP4750 dead-end membrane filtration system beamline for surface and interface diffraction. J. Synchrotron Radiat. 23, (Supplementary Fig. 22). We also used a commercial bench-scale 50 ml transparent 1245–1253 (2016). Merck Millipore Amicon dead-end stirred cell, which was connected to an 800 ml 57. Basham, M. et al. Data analysis workbench (DAWN). J. Synchrotron Radiat. Merck Millipore Amicon RC800 reservoir, to visualize the filtration process 22, 853–858 (2015). (Supplementary Fig. 23). During these measurements, the feedstocks were kept 58. Coelho, A. TOPAS Academic v.5 (Coelho Software, 2012). under a 10 bar nitrogen pressure (3 bar for Merck Millipore Amicon cells) at room temperature, and the feedstock was continually stirred using a stirring bar rotating Acknowledgements at 400 r.p.m. The Hansen solubility parameter (δ) and the physical properties of the A.I.C., M.A.L., M.E.B., Y.W. and A.H. acknowledge the Engineering and Physical organic solvents (Supplementary Table 2) were used to investigate the relationships Sciences Research Council (EP/N004884/1). A.I.C. and A.H. acknowledge the between pure solvent permeances and the combined solvent properties. Leverhulme Trust via the Leverhulme Research Centre for Functional Materials Design −2 −1 Flux J (l m h ) was calculated according to the following equation: for funding. A.H. acknowledges the China Scholarship Council for a studentship and the Royal Society of Chemistry for a Researcher Mobility Grant (M19–2442). Z.J. and J = ΔV/ (A ×Δt) (1) A.G.L. acknowledge the Engineering and Physical Sciences Research Council for funding where ∆V is the volume of permeate collected in litres in a given amount of time, A (EP/R018847/1). A.I.C., M.A.L., A.H., Y.W., H.H. and J.R acknowledge Diamond Light is the membrane surface area in square metres and ∆t is the time in hours between Source for access to beamlines I07 (SI24359) and I11 (CY23666) that contributed to the the start and end of the measurement. results presented here. We thank M. Liu for advice about the cage synthesis conditions, −2 −1 −1 Solvent permeance P (l m h bar ) was calculated according to the following H. Chen for assistance during the I07 measurements, P. Cui and H. Gao for assistance equation: during the I11 measurements, Y. Li for fitting the logical curve (Fig. 5a), R. Clowes for instrument support and setting up the filtration cell, H. Yang and T. Mitra for assistance P = ΔV/ (A ×Δt ×p) (2) collecting SEM images and K. Arnold for collecting the FIB-SEM images. where ∆V is the volume of permeate collected in litres in a given amount of Author contributions time, A is the membrane surface area in square metres, ∆t is the time in hours between the start and end of the measurement and p is the transmembrane A.I.C. and A.G.L. conceived the project. A.H. and M.A.L. conceived the graded molecular sieving part of the study and performed the associated measurements. pressure. To calculate solvent permeance, typically, 0.2 l of pure solvent or dye A.H. and Z.J. prepared the membranes. A.H. led the characterization and membrane feedstock (20 ppm dye concentration) was added to the feedstock tank. The cell was then pressurized to 10 bar under nitrogen. The solvent permeate was performance tests. Z.J. synthesized the PAN membrane support and conducted the AFM then calculated based on the amount of time it took ~0.1 l of pure solvent or dye measurements. M.E.B. helped with the cage synthesis. H.H. and J.R. designed and set up the in situ GIXRD system and helped with these measurements. A.H., M.A.L. and feedstock to flow through the membrane (Supplementary Tables 3 and 4 for full NAtuRE MA tERIALs | www.nature.com/naturematerials Nature Materials Articles Y.W. contributed to the GIXRD measurements and processed the X-ray data. A.G.L. Additional information provided facility support and insight into advanced membrane science. Z.J. designed the Supplementary information The online version contains supplementary material interfacial synthesis method for the membrane. M.A.L. cosupervised the project with available at https://doi.org/10.1038/s41563-021-01168-z. A.I.C. and A.G.L. and wrote the manuscript with A.H. and Z.J., with contributions from Correspondence and requests for materials should be addressed to all coauthors. All of the authors participated in the discussion of data and commented Andrew G. Livingston or Andrew I. Cooper. on the manuscript. Peer review information Nature Materials thanks Michael Mastalerz and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Competing interests Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing interests. NAtuRE MA tERIALs | www.nature.com/naturematerials

Journal

Nature MaterialsSpringer Journals

Published: Apr 1, 2022

There are no references for this article.