Microporous Polymer Adsorptive Membranes with Unprecedented Processing Capacity for Molecular Separation


 Trade-off between permeability and nanometer-level selectivity is an inherent shortcoming of membrane-based separation of molecules, while most highly porous materials with high adsorption capacity lack solution processability and stability for achieving adsorption-based molecule separation. We hereby report a hydrophilic amidoxime modified polymer of intrinsic microporosity (AOPIM-1) as a membrane adsorption material to selectively adsorb and separate small organic molecules from water with ultrahigh processing capacity. The membrane adsorption capacity for Rhodamine B reaches 26.114 g m−2, 10~1000 times higher than previously reported adsorptive membranes. Meanwhile, the membrane achieves >99.9% removal of various nano-sized organic molecules with water flux 2 orders of magnitude higher than typical pressure-driven membranes of similar rejections. This work confirms the feasibility of microporous polymers for membrane adsorption with unprecedented capacity, and provides the possibility of adsorptive membranes for molecular separation.


Introduction
Molecular separation is an essential component in many human daily activities and multiple industrial, medical, and environmental processes, such as water puri cation, oil and gas re ning, energy generation and storage, pharmaceutical ingredient extraction and puri cation [1][2][3][4][5] . Among various materials and methods, synthetic membranes and membrane-based separation have been extensively studied due to its strong sustainability, scale-up feasibility, and no phase change during separation 1,6 . However, trade-off between membrane permeability and selectivity is an inherent shortcoming of membrane-based molecular separation [7][8] . Taking pressure-driven membrane processes for instance, micro ltration and ultra ltration membranes exhibit high water permeation ux and reject large molecules with high e ciency, but their large pores (0.002-1 µm) are unable to separate nanometer-sized small organic molecules 9 . Processing of these small organic molecules requires membranes of nano ltration level pore size and selectivity, yet the reduction in pore size inevitably results in much lower water permeation ux 10 .
In contrast, adsorption can be a highly selective molecular separation process with speci c physical or chemical interactions between adsorbents and target molecules [11][12][13][14] , but its application is often limited by its low processing rate, high internal diffusion resistance within absorbents, etc [15][16] . Membrane adsorption is a pressure-driven dynamic membrane-based adsorption process, which combines the merit of both adsorption and membrane separation [17][18][19] . Compared to traditional pressure-driven membranes with size sieving as the dominant separation mechanism, adsorptive membranes utilize more speci c membrane-solute interactions like electrostatic interactions, π-π interactions, van der Waals forces, and hydrogen bonding to achieve highly selective and fast separation of small organic molecules [20][21] .
Therefore, they are expected to break through the permeability-selectivity trade-off by simultaneously achieving the selectivity of dense membranes and permeability of porous membranes 18, 22−25 . However, the application of membrane adsorption is hindered by the lack of adsorptive membranes with su cient processing capacity. Adsorptive membranes are usually prepared by post-grafting a nity ligands or adsorptive ller mixing in traditional polymer materials 15,20 . Due to the limited speci c surface area and adsorption sites incorporated, the overall processing capacity of these conventional adsorptive membranes is very low, mostly in the range of 0.01-1 g m −2 . Such membranes could only handle solutions of very low concentration and require frequent cleaning and regeneration, which greatly affects their further development and practical application. Although porous materials such as metal-organic frameworks (MOFs) possess large porosities and rich a nity sites [18][19][26][27][28][29][30] , their poor solution processability hinders their engagement as absorptive membrane separation materials. Besides, most MOFs materials do not have su cient stability in water, especially under acidic or alkaline conditions, resulting in membrane failure during long-term operation 8 .
In this work, we report a microporous polymer based adsorptive separation membrane, which has extremely high adsorption capacity and can realize fast and selective molecular separation. The microporous polymer is amidoxime modi ed polymer of intrinsic microporosity (AOPIM-1) polymer, which owns a rigid and contorted three-dimensional structure in its backbone. The ineffective chain packing brings high speci c surface area up to 550 m 2 g −1 and produces interconnected free volumes with a size of less than 2 nm [31][32][33][34] . The amidoxime modi cation endows the polymer with good solution processability and provides abundant adsorption sites for selectively adsorbing charged molecules. Due to its unique chemical and physical structure, it achieves a high-e ciency removal of small organic molecules (>99.9%) with permeating ux 2 orders of magnitude higher than typical nano ltration membranes of similar dye rejections, and such separation performance is maintained throughout multiple adsorption-elution cycles with >98% ux recovery rate (FRR). More importantly, the static adsorption capacity of the membrane material greatly surpasses traditional non-porous polymer adsorbents and comparable to MOF-based adsorbents, while the dynamic processing capacity reaches 26.114 g m −2 , much higher than all the adsorptive membranes reported so far. The unprecedented processing capacity is expected to enable the membrane towards more realistic adsorption-separation application scenarios.

Results And Discussion
Material characterization of AOPIM-1 polymer Polymer of intrinsic microporosity (PIM-1) was synthesized via a polycondensation reaction between TTSBI and TFTPN ( Figure S1) as reported previously 35 , which shows a high speci c surface area (786 m 2 g −1 ). The prepared PIM-1 was further modi ed with hydroxylamine to obtain AOPIM-1 (Figure 1a, Figure   S2). 31,34 Spectroscopic characterizations of AOPIM-1 are provided in Figure S3-4, which signify the successful synthesis of the polymer. Figure 1d shows a three-dimensional view of a modeled amorphous cell of the AOPIM-1 polymer, which presents a highly microporous feature. Its speci c surface area reaches 552.3 m 2 g −1 as deduced from its N 2 adsorption isotherm ( Figure 2a). Moreover, the amidoxime modi ed polymer changes its solubility parameter due to the introduction of polar groups, making it soluble in common casting solvents, such as DMF, NMP, DMSO, etc ( Figure S5). Unlike PIM-1 that could only be dissolved in chloroform and tetrahydrofuran, the great solution processability of AOPIM-1 makes it feasible for the fabrication of asymmetric membranes via the industrially scalable phase inversion approach (Figure 1a). In addition, stability of AOPIM-1 under acidic and alkaline conditions are examined by immersing in acid (pH = 3) and base (pH = 10) for at least 24 h, and the chemical and physical structure of the polymer is found to remain unchanged (Figure 2a and S4). Interestingly, the AOPIM-1 polymer shows a pH-tunable chargeability (Figure 1b). Such tunable chargeability is attributed to the protonation/deprotonation of amidoxime groups, and they could thus be utilized as effective a nity sites with pH-tunability for selective adsorption and separation of oppositely charged molecules (Figure 1c).
The processability, hydrophilicity, high speci c surface area and pH-tunable a nity sites make AOPIM-1 a promising membrane adsorption material for aqueous-based molecular separations, which will be demonstrated in the following sections.
pH-tunable static adsorption feature of AOPIM-1 polymer Two organic dye molecules with similar molecular weight but different charge properties, Methyl Orange (MO, MW = 327, negatively charged) and Methylene Blue (MB, MW = 320, positively charged) are used to evaluate the static adsorption behavior of AOPIM-1. As shown in Figure 2b, the amidoxime modi ed polymer favors the capture of negatively charged MO molecules under acidic conditions while capturing the oppositely charged MB under alkaline conditions, exhibiting a pH-tunable adsorption feature. The adsorption capacity of the polymer is found to be as high as 445.02 mg g −1 (MO, pH = 3.3) and 735.75 mg g −1 (MB, pH = 10.9), respectively, owing to the combination of high speci c surface area and pHtunable chargeability of amidoxime groups within the polymer. And lower equilibrium adsorption capacity is observed under neutral pH conditions where the amidoxime possess minimal chargeability, which also illustrates that the pH-tunable a nity sites make indispensable contribution to the adsorption capacity. As shown in Figure S7, the mixture of MB and MO is green with some precipitation due to the opposite charges of MB and MO molecules 14 . After AOPIM-1 treatment under different pH conditions, one of the colors is completely removed, and the precipitation disappears. The UV-Vis spectra clearly show the disappearance of one of the absorption peaks. The result proves that AOPIM-1 achieves the selective adsorption of MB and MO mixture by adjusting pH environment. In Figure 2c, d, Langmuir adsorption isotherms and pseudo-second-order equation (R 2 > 0.99) are used to t the adsorption of AOPIM-1 on target molecules in different pH environments. Under acidic and alkaline conditions, the maximum adsorption capacity for MO and MB reaches 491.63 mg g −1 and 765.09 mg g −1 , respectively. AOPIM-1 shows a faster adsorption rate, exceeding 50% in 40 min, and reaching equilibrium in 300 min. The obtained adsorption capacity of AOPIM-1 is much higher than traditional non-porous polymer adsorbents and comparable to that of newly developed MOF adsorbents (Table S1)

Adsorption separation of dye molecules by AOPIM-1 membranes
The adsorptive membranes based on AOPIM-1 are constructed via a wet-phase inversion method. The membrane structure is altered through tuning phase inversion conditions including various coagulation bath composition, casting solution concentration, and casting solution composition as summarized in Table S2. With increasing ethanol content in coagulation bath, it could be observed that the structure of the membrane changes from nger-like pores (M1) to sponge-like pores (M3) as demonstrated in the cross-sectional SEM images in Figure 1e and 1f. Meanwhile, the pure water ux of resulted membrane decreases from 1505.7 L m −2 h −1 bar −1 (M1) to 249.8 L m −2 h −1 bar −1 (M3). However, the processibility capacity of M3 increases from 3.7 to 11.4 g m −2 which is 3 times more than that of M1 (Table S2, Figure   S8-10). It should be mentioned that the effective processing capacity of membranes is calibrated under the criterion of 99% rejection ratio of Rhodamine B (RHB) (20 ppm). With increase of the concentration of the casting solution, it could be observed that the membrane thickness (M3-M5 in Table S2) is gradually increased. The increasing in thickness and processing capacity has a linear relationship from M3 11.4 g m −2 (64 µm) to M5 26.1 g m −2 (119 µm). Figure 3e and Table S3 present the comparison of AOPIM-1 membranes with reported adsorptive membranes with respect to their permeation ux and adsorptive capacity, and the dynamic adsorption capacity of the membrane appears to be 10~1000 times higher than previously reported adsorptive membranes. The above experiment results prove that by adjusting the phase inversion process, the membrane structure can be reasonably designed to achieve the purpose of rapidly removing small molecular organic pollutants in aqueous system using dynamic adsorption processes.
The dynamic adsorption separation process was conducted on a dead-end setup using a sponge-like AOPIM-1 membrane (M3). According to zeta potential measurement results, AOPIM-1 membranes exhibit positive charge at acidic pH < 4 while turning into negative charge at alkaline aqueous environment ( Figure 3a). Thus, six types of organic dyes with different sizes and charges were used as target molecules for separation. Negatively charged dyes including Methyl Orange (MO), Congo Red (CR) and (BB) were tested in acidic condition and positive charged dyes including Methyl Blue (MB), RHB and Crystal Violet (CV) ( Figure S11) was tested in alkaline condition. It can be seen from Figure 3b and Table  The adsorptive separation behavior of the AOPIM-1 membrane is further demonstrated in a multi-cycle dynamic membrane adsorption process. It can be seen from Figure 4a that the ux of the membrane decreases with the increasing permeation volume in each cycle, which can be attributed to the accumulation of dyes on the surface and inside of the adsorptive membrane. In the rst seven cycles, 20-30 mL of methanol was adopted as a feed solution to desorb the dye molecules from the membrane for a short time (within 5 minutes). It can be seen in digital photo in Figure 4d that the color of the membrane due to the adsorption of dye disappears after cleaning. At the same time, the ux recovery rate (FRR) in each cycle reaches 98%. After 7 cycles, a long-term desorption treatment (about 2 hours by methanol) on the membrane was performed, and a higher recovery rate is got. The Brunauer-Emmet-Teller (BET) surface area of the AOPIM-1 adsorptive membrane is reduced from 552.3 m 2 g −1 to 415.4 m 2 g −1 after the adsorption test, and it can be easily regenerated to the original level after desorption (Figure 4b).
Meanwhile, it can be seen in Figure 4c, the pore size distribution of AOPIM-1 is also restored to the original level after cleaning. As a control group, a negatively charged polyethersulfone ultra ltration membrane prepare by phase inversion was selected to test positively charged organic small molecules. It is found that almost all dye molecules pass through the membrane without noticeable retention ( Figure   S12), which is common for conventional ultra ltration membranes that rejects large molecules and particulate matter (such as proteins, suspended solids, bacteria, viruses, and colloids) but cannot accurately separate small organic molecules 10 . Without necessary speci c surface area and a nity sites for effective adsorption, membranes of conventional polymer materials require further reduced the pore size for separation of small molecules, but it will inevitably lead to a signi cant reduction in water ux and huge energy consumption. The above results reveal that the electrostatic attraction, micropore and membrane pore structure co-govern the dynamic adsorption performance of the membrane.

Adsorptive separation of active pharmaceutical ingredients (APIs)
In the pharmaceutical industry, precise separation of organic molecules such as drugs, proteins, and polysaccharides are indispensable for the production of active pharmaceutical ingredients (APIs) 5,7,56 . For instance, the raw material water extracts of phytochemical drugs, an important category of APIs extracted from natural plants 45 , usually possess a complex composition including polysaccharides (molecular weights usually range from 10,000 to 100, 000 Da), APIs (molecular weight < 2000 Da), and inorganic salts such as sodium chloride. The feasibility of membrane adsorption for the separation of such 3-component systems was evaluated in this section. As schematically illustrated in Figure 5a and Figure S13, a 2-step process is proposed for adsorptive separation of a phytochemical drug water extract using AOPIM-1 membranes. In the rst step, the water extract of natural plants passes through the surface of the membrane, the plant polysaccharides with large molecular weight are trapped on the feed side of the membrane, the active pharmaceutical molecules with small molecular weight are adsorbed within the membrane matrix, and inorganic salts are permeated through the membrane. After the membrane is saturated with API adsorption, the second step uses methanol as the eluent to ow in from the feed side, and the API eluate is obtained on the permeate side.
As a proof of concept, an AOPIM-1 membrane with altered casting conditions (3:1 DMF/1,4-dioxane cosolvent [47][48] ) and tightened membrane pores (water ux of ~121.3 L m −2 h −1 bar −1 , MWCO of ~20 kDa) are prepared, and the MWCO curve of M6 and the pore size distribution are shown in Figure S14. A synthetic water extract (Table S5) is prepared with Dextran T-200 (MW = 20 kDa) used as the target macromolecular polysaccharide, berberine (MW = 336.4 Da) as the target small API molecule, and sodium chloride as the inorganic salt. As can be seen in Figure 5b, during a cross-ow ltration cycle, berberine is continuously adsorbed in the membrane, and the rejection rate of Dextran T-200 is maintained above 90%. At the same time, sodium chloride permeates through the membrane without rejection. At the end of one ltration cycle, the feed solution is changed to methanol to elute the berberine enriched in the membrane. Figure S15 clearly shows the UV-Vis absorption peaks in the stock solution and ltrate before and after processing 50 mL of the synthetic feed. After the ltration, the characteristic absorption peak of berberine disappeared, which proves that the berberine in the feed solution are completely adsorbed in the membrane. At the same time, it can be seen in the illustration that 50 mL of the feed solution is enriched in the membrane and nally eluted by about 5 mL of methanol, achieving a 10-fold enrichment. Compared with other separation methods, the traditional distillation method is energy-intensive and time-consuming, whereas the membrane ltration method is di cult to achieve su cient accuracy and e ciency, and the obtained permeate usually requires further puri cation operations afterwards 57 . The membrane adsorption separation method shows the merit of smallest energy consumption and highest product purity. More importantly, this experiment demonstrates that the ultrahigh processing capacity of AOPIM-1 adsorptive membranes omits the need of frequent cleaning and regeneration, effectively broaden the application feasibility of membrane adsorption from treating trace organic compounds towards mass chemical productions.

Conclusions
This work has developed a new type of pH-tunable high-capacity adsorptive membranes based on AOPIM-1, in which the high speci c surface area (high adsorption capacity), abundant adsorption sites (adsorption selectivity), reversible adsorbate-adsorbent interaction (fast adsorption/desorption rates), good solubility processability (scale-up feasibility) and hydrophilicity (easy-wet micropores) of AOPIM-1 are fully utilized. The processing capacity and permeability of the membrane is adjusted by manipulating the phase inversion process. While achieving the retention of small molecules, the membrane permeation ux is in line with the level of common ultra ltration, which is 2 orders of magnitude higher than typical nano ltration membranes of similar rejections. The best processing capacity reaches 26.114 g m −2 , which is 10~1000 times higher than the value reported in the literature for existing adsorptive membranes. This newly developed membrane material realizes a dynamic operation process of membrane adsorption-desorption, cleaning, and regeneration with high e ciency, which can be a good supplement to conventional pressure-driven membrane separation processes. Owing to the unprecedented adsorption capacity, the AOPIM-1 membrane exhibits high separation e ciency of actual complex systems and highly guarantees the purity of the product obtained. The development of AOPIM-1 adsorptive membranes thereby broadens the prospects of membrane adsorption for practical applications.

Methods
Fabrication of PIM-1 and AOPIM-1 polymers. Polymer of intrinsic microporosity (PIM-1) was obtained following a previously reported method. Under a nitrogen atmosphere, 3.001 g (15 mmol) tetrauoroterephthalonitrile (TFTPN), 5.106 g (15 mmol) 5,5',6,6'-tetrahydroxy-3,3,3',3'tetramethylspirobisindane (TTSBI) and 30 mL anhydrous DMAc were added into a 100 mL three-necked ask. After the chemicals were completely dissolved, 6.21 g (45 mmol) anhydrous milled K 2 CO 3 was added and the ask was placed into an 160°C oil bath under mechanical stirring. After approximately 3 min, a viscous yellow solution formed, and 20 mL of toluene was added. Several minutes later, a further 20 mL of toluene was added to dilute the solution. Then, the mixture was poured into 300 mL methanol, and an elastic, threadlike, light yellow polymer was observed. The polymer product was dissolved in chloroform and reprecipitated in methanol, and then re uxed in Milli-Q water for 4-5 h and dried at 80°C under vacuum for 48 h.
Hydrophilic amidoxime modi ed PIM-1 (AOPIM-1) was synthesized by dissolving 0.5 g PIM-1 in 30 mL THF and heating to re ux under N 2 . Then, 5.0 mL hydroxyl amine was added dropwise, and the solution was further re uxed for 20 hours. The resulting polymer was precipitated by the addition of ethanol, ltered, washed thoroughly with ethanol and water, and then dried at 110°C for 24 h.
Fabrication of AOPIM-1 membranes. The polymer dope was obtained by dissolving AOPIM-1 in DMF or DMF/1,4-dioxane mixed solvent, and stirred continuously at room temperature overnight to ensure that the polymer dissolves evenly in the solvent. Then the mixture was left at room temperature for 24 h to remove air bubbles. After that, the polymer solution was used to cast lms on a clean glass plate at 25 ºC and 40% relative humidity. For the casting solution with co-solvent, the solvent was allowed to evaporate from the surface of the lm in 20 s to produce a denser selective skin. Next, the glass plate was immersed into a coagulation bath. After 1 h, membranes were transferred to a fresh water bath and kept for 24 h to nish phase separation. Finally, the membranes were immersed in methanol for future use.
Membrane material characterization. The morphology of the as-prepared membranes was observed using a eld emission scanning electron microscope (Hitachi S4800, Japan). Before capturing SEM image, a thin Au layer was sputtered onto the membrane under 10 mA for 2 min (Emitech K550X sputtering). 1 H nuclear magnetic resonance (NMR) spectra were recorded on a Bruker 400 MHz spectrometer using dimethyl sulfoxide-d 6 as a solvent. FTIR spectra of membranes were obtained using a Nicolet 6700 FTIR spectrometer (USA). Nitrogen absorption/desorption measurements were performed on a Quantachrome Autosorb IQ-MP-MP at 77 K. All samples were degassed at 120°C for 12 h before nitrogen absorption measurements were performed. The surface charge of the membrane was determined by streaming potential measurement using a SurPASS 3 electrokinetic analyzer with a atplate measuring cell (10 mm × 20 mm).
Static adsorption behavior of AOPIM-1 polymer. The static adsorption behavior of AOPIM-1 was investigated using dyes with different chargeability (negative, MO; positive, MB) as model solutes. A dry AOPIM-1 membrane coupon (~10 mg) was placed in dye solution with certain concentration (10-500 mg L −1 ) and pH value. The mixture was stirred continuously at room temperature at least 24 h or a certain time. The concentration of dyes was analyzed by a UV-Vis spectrophotometer (PerkinElmer, Inc.). The amount of dye adsorbed by the AOPIM-1, q (mg g −1 ), was determined from the Eq. (1) [49][50] , where q represents the adsorption amount at equilibrium (q e ) or the adsorption amount at time t (q t ). where ΔP (bar), V (L), A (m 2 ), and t (h) represent permeate ux, transmembrane pressure, permeate volume, membrane area and ltration time, respectively. The rejection rate (R (%)) of dyes was calculated from Eq. (5), where C P and C F correspond to the dye concentrations in the permeate and feed solutions, respectively.
The dye concentrations in the permeate and feed solutions were determined using a UV/Vis spectrometer (Biochrom Libra S32). The processing capacity used RHB as feed solution (20 mg L −1 , PH=10) was measured under 0.2 MPa at room temperature and determined by permeation volume when rejection is higher than 99%.
The multi-cycle adsorptive separation performance stability of the AOPIM-1 adsorptive membrane was performed in a dead-end ltration cell with an effective membrane area of 3.14 cm 2 for 8 cycles. For a typical cycle, the RHB solution (20 mg L −1 , pH = 10) was ltrated through the membrane for 500 L m −2 at 0.2 MPa, and the permeate ux (J) was calculated by Eq. (4). Afterwards, the membrane is subjected to a short time ( ve minutes) desorption process by switching the feed solution of the membrane to about 20 ml methanol for cleaning, and the eluent can be obtained at the outlet. After that, fresh feed dye solution was applied for the next cycle.
Adsorptive separation of APIs by AOPIM-1 membranes Separation of a synthetic 3-conponent mixture feed representing inorganic salt, polysaccharides and active pharmaceutical ingredient (API) chemicals in the water extract of natural plants was performed in a lab-scale cross-ow cell ( Figure S13) at room temperature. The feed solution included NaCl, dextran T-200 and berberine with detailed composition listed in Table S5. The concentrations of each substance in the feed and permeate solution were measured by conductometer (FE30K, Mettler Toledo), total organic carbon (TOC) analyzer (Aurora 1030W) and UV-Vis spectrophotometer (PerkinElmer, Inc.), respectively.    reported membranes with respect to their permeation ux and dye rejection/removal e ciency.

Figure 4
Adsorptive separation behavior of AOPIM-1 membrane in a multi-cycle dynamic adsorption-desorption process. (a) Flux and RHB removal rate of the membrane in multiple adsorption-desorption cycles