Assembling covalent organic framework membranes via phase switching for ultrafast molecular transport

Fabrication of covalent organic framework (COF) membranes for molecular transport has excited highly pragmatic interest as a low energy and cost-effective route for molecular separations. However, currently, most COF membranes are assembled via a one-step procedure in liquid phase(s) by concurrent polymerization and crystallization, which are often accompanied by a loosely packed and less ordered structure. Herein, we propose a two-step procedure via a phase switching strategy, which decouples the polymerization process and the crystallization process to assemble compact and highly crystalline COF membranes. In the pre-assembly step, the mixed monomer solution is casted into a pristine membrane in the liquid phase, along with the completion of polymerization process. In the assembly step, the pristine membrane is transformed into a COF membrane in the vapour phase of solvent and catalyst, along with the completion of crystallization process. Owing to the compact and highly crystalline structure, the resultant COF membranes exhibit an unprecedented permeance (water ≈ 403 L m−2 bar−1 h−1 and acetonitrile ≈ 519 L m−2 bar−1 h−1). Our two-step procedure via phase switching strategy can open up a new avenue to the fabrication of advanced organic crystalline microporous membranes.

an unnatural rebound of the permeability at 18 hours. The authors may wish to explain the mechanism behind increased rejection from 83 % (15 hours) to 99 % while maintaining the permeability at ~ 401 LMH/bar. The optimization table for TFP-TTA is not provided anywhere in the paper.
The authors may wish to explain the mechanism of the high rejection efficiency of Congo Red, which has a side that is small enough (~ 0.73 nm) to permeate through the large pores of TFP-PDA (1.4 nm). Also, the authors may consider giving a more detailed explanation of the low rejection and permeability at 155 ˚C as reference 38 does not seem to provide the information.
In general, this work is not particularly novel and presents a very similar structure as the authors' previous paper. It may not be suitable for Nature Communications.
Reviewer #3: Remarks to the Author: In this work, the authors report a phase-switching strategy to fabricate covalent organic framework (COF) membranes through a two-step procedure. The first step in liquid phase is primarily responsible for polymerization and the second step in vapor phase is primarily responsible for crystallization, which delicately decouples polymerization process and crystallization process. The two-step procedure has some distinct superiority to the commonly used one-step procedure in fabricating defect-free and ultrathin membranes. Using phase-switching strategy, highly crystalline, more compact and continuous membranes were fabricated. Accordingly, ultrafast molecular transport was realized, as manifested by water permeance of 402 L m-2 h-1 bar-1 and acetonitrile permeance of 519 L m-2 h-1 bar-1, which is among the ever-reported highest separation performance for COF membranes. The phase-switching strategy was validated by fabricating two kinds of highly crystalline COF membranes with different pore apertures. The strategy in this work represents a striking breakthrough in the precise construction of advanced COF membranes. This paper is meticulously organized and elegantly written and I read it with great joy. I firmly believe this interesting and innovative work will be highly appreciated by broad scientific communities in chemistry, materials science and membrane technology. Overall, this work is highly suitable for publication in Nature Communications after addressing the following minor revisions. 1. The authors employed the β-ketoenamine-linked Schiff base COF in their work. More explanations may be required for the readers to more clearly understand why the authors preferred this type of COF over the other types? 2. Can the phase-switching strategy be applied to other kinds of COF membranes in principle? 3. Why did the authors choose DMAc as the solvent in the first step? Can the authors outline the major criteria in choosing the solvent? 4. Have the authors tested the rejection performance of dyes in organic solvents? Some mistakes: (1) The scale bar of crosse-section SEM in Fig. 2c, f should be corrected. (2) In line 125, XRD was written twice.

REVIEWER COMMENTS
Reviewer #1 (Remarks to the Author): In this manuscript, the authors have reported a two-step procedure to assemble COF membrane via a phase switching strategy. Accordingly, two COFs-based membranes were prepared and wellcharacterized. In addition, the obtained TFP-PDA COF membrane exhibited a water permeance of 402 ± 3 L m-2 h-1 bar-1 and a good rejection rate (99%) toward organic dyes (Congo red). Given the interesting applications of porous membranes, I would like to recommend the publication on Nature Communications after addressing the questions stated below: Reply: We would like to thank the reviewer for these highly positive remarks and valuable guidance.
1) The pristine membrane was obtained by evaporation of the solvent at 60 ℃. However, the monomers were dissolved in DMAc, which has a very high boiling point (> 160 ℃). Some explanation should be added here.

Reply:
We would like to thank the reviewer for this valuable guidance. The major reasons for choosing DMAc are the high solubility of amine and aldehyde monomers in DMAc as well as the appropriately slow rate of evaporation for completion of the polymerization in the first step. Due to the monomer's high solubility (TFP 23mg/ml, PDA≈25mg/ml and TTA≈17 mg/ml), a low volume of DMAc is required in the first step. Secondly, due to the relatively high boiling point of DMAc, it evaporates slowly at 60 °C allowing the completion of polymerization. In our experiments, we observed that 2 ml of DMAc requires about 25 minutes for its evaporation at 60 °C. When a low boiling point solvent such as dichloromethane is used, the solvent evaporates very fast leaving no sufficient time for the completion of the polymerization in the first step. We have added this text to the revised supplementary Information file.
"DMAc was chosen in this step due to high solubility of amine and aldehyde monomers in DMAc which reduces reaction volumes (TFP 23mg/ml, PDA≈25mg/ml and TTA≈17 mg/ml) and relatively high boiling point of DMAc, which wins sufficient evaporation time for the completion of the polymerization at this step." 2) It seems that there is a very small deviation between the simulated and experimental PXRD, such as the 100 peak. Is this possible to have slipped AA structure? Reply: We thank the reviewer for this valuable observation. Indeed, there is a slight deviation between the experimental and simulated peaks. This deviation is consistent with previous literature (Angew. Chem. Int. Ed., 2021, 60, 19797-19803). Ideally, an AA stacking of COF layers will result in a membrane with perfectly aligned vertical pores that will afford solvent permeance manifolds higher than what is currently reported. However, in our work and reported literature, there exists a big chance for AB stacking, rendering slight deviation from the simulated pattern. Further exploration is required to achieve perfect AA stacking in future work. We have added relevant discussion on the matter in the revised manuscript.
"is attributed to ππ stacking between two layers (Fig. 4b). The slight deviation from the ideal crystal structure is consistent with the previous literatures 1 ." 3) The authors assumed that the pristine membrane at 60 °C in a reversible state. However, from the FT-IR spectrum, there is no imine bonds, which is different from the presence of amine groups at low temperature. The author should provide more solid evidence to prove their hypothesis.
Otherwise, the mechanism for this two-step procedure needs to be reconsidered.

Reply:
Again, we thank the reviewer for this observation and guidance. The appearance of imine peaks in the FT-IR are masked due to it overlapping with the strong C=C and C=O peaks in the same region of the spectra. The FT-IR peaks of the imine lie in the region of 1615-1630 cm -1 with low intensities (J. Mol. Struct., 2020, 1214, 128150 (2020). However, in the COFs reported in this work, the same region is occupied by the peaks of C=C and C=O (1605-1630 cm -1 ) with higher intensities. The absence of amine bands at ≈3300 cm -1 of the amine monomer and the shifting of carbonyl band from 1643 cm -1 to 1609 cm -1 of the aldehyde monomer can indirectly prove the formation of imine formation. 4) After finishing the first step, the obtained pristine membrane was further being placed at 145 ℃ to finish the linkage rearrangement. However, at this temperature, the stable linkage formed after rearrangement seems not good for the morphology repairing of the membrane. How did the membrane become to be more compact and defect free? This also need more discussions.

Reply:
We thank the reviewer for this beneficial advice. Generally, the phenomena of dissolution and recrystallization commonly exist in many crystalline materials such as MOFs (Chem. Commun., 2020, With the increase of vapor treatment time, the membranes morphology turned smoother and defectfree. From the SEM observations, the progress is marked with the transformation in Fig. 4g. We suppose that dissolution and recrystallization also occur in this work at high temperature where small COF seeds/grains fuse together to create larger seeds/grains. Such fusion can heal the grainboundary defects during the whole process, yielding defect-free membranes. Moreover, the presence of hydroxyl groups in these COFs can stabilize membranes morphology due to interlayer molecular hydrogen bonding. We have added this description to the revised manuscript. "morphology after 18 h of vapor treatment. The continuous growth of COF membranes indicates that dissolutions and recrystallizations of the COF seeds occurred by the emergence of large seeds due to the fusion of small seeds. These observations are consistent with other crystalline materials such as crystalline salts 2, 3 , zeolites 4-6 , metal-organic frameworks (MOFs) 7, 8 and COFs 9 " 5) Both the thickness of the membranes is about 150 nm. Did the author try to obtain other membrane with different thickness to demonstrate the controllability? The permeance of the membrane may also be improved by optimizing the thickness.

Reply:
We would like to thank the reviewer for this helpful insight. In the initial experiments during this work, we observed that the membranes had two distinct layers, the top layer composed of large particles/threads and the bottom layer composed of defect free membrane, as shown in supplementary Fig. 1 and 2. Irrespective of the top layer's thickness, the bottom layer remained at ≈ 150 nm after 18 h at 145 °C. We optimized the conditions to reduce the top layer by lowering monomer's concentration until no particles were observed. The permeance was tested for each membrane and we observed a similar permeance for all membranes. We deduce that the top layer does not take part in the permeance and only the bottom layer is responsible for the permeance. In this work, we also observed that membranes below 140 nm could not be employed for performance evaluation due to low mechanical strength. We have highlighted this part in the revised manuscript.
"We also observed that thicker membranes were composed of a continuous membrane towards ITO surface, covered by large particles and threads. These particles and threads could be easily removed using adhesive tape before the etching step (Supplementary Fig. 1-3). By optimizing the monomers concentration, ultra-thin membranes were fabricated by avoiding the growth of particles completely." 6) How did the authors test the gas adsorption of the membrane? More details can be put into the experimental section to compare the difference in porosity between the membrane and the COF powder.

Reply:
We thank the reviewer again for this valuable insight and advice. The porosity measurements were done by first fabricating several membranes and grinding them into powders. We have included the following relevant descriptions into the experimental section of the revised supplementary information file.
"N 2 adsorption/desorption isotherms were recorded on PS2-1055-B gas adsorption analyzers at 77 K using a liquid nitrogen bath. Several membranes were first fabricated and then grinded for porosity evaluation. All the membrane samples were degassed at 120 °C for 6 h under vacuum before N 2 analysis." Reviewer #2 (Remarks to the Author): The paper by Khan et al. reports a 2-step procedure named "phase switching strategy" to achieve greater crystallization. This strategy allows the amine and aldehyde monomers to be cast on solid support and form membranes through solvent evaporation at 60 ˚C. The crystallization was delayed to the second step, where a high temperature (145 ˚C) was applied to initiate linkage rearrangement to form a COF membrane.
Hence, the authors may want to clarify the differences between the present method and the reported ones. Reply: Firstly, we would like to thank this reviewer for the important and valuable input. The strategy/method presented in this work has some distinct innovative features which have never been Furthermore, the methods for synthesizing the COF powders cannot be directly and easily applied to COF membranes because molecular separation membranes require the homogeneous dispersion and appropriate linkage/connection of the powdered materials to avoid the formation of defects, which are not a major concern in powder synthesis. The films fabricated at room temperature in the  The authors may wish to explain the mechanism of the high rejection efficiency of Congo Red, which has a side that is small enough (~ 0.73 nm) to permeate through the large pores of TFP-PDA (1.4 nm). Reply: We thank the reviewer for this excellent and helpful advice. Congo red is a charged dye (anionic) with dimensions of 0.73 x 2.26 nm. In polar solvents such as water, it is important to consider the hydrodynamic radii of the solvated charged species which includes its solvation shell, thereby the larger dimension (2.26 nm) may be responsible for their high rejection rates but it could very well also be that the 0.73 nm dimension in solution has a much larger hydrodynamic cross section that cannot permeate through the pores. Additionally, dyes may also aggregate in water which substantially increases their overall dimensions. To make sure that this selectivity comes from the pores of COF membranes, we used PPH-IX dye with dimension of 1.45 x 1.54 nm in organic solvent and observed a rejection of more than 99%.
Also, the authors may consider giving a more detailed explanation of the low rejection and permeability at 155 ˚C as reference 38 does not seem to provide the information. Reply: We thank the reviewer for pointing out this issue. We optimized the temperature in the second step for COF membrane fabrication and observed that membranes crystallized at 155 °C exhibited poor permeance. We assume that COF particles grow in the pores which have also been observed by other groups as well (Angew. Chem. Int. Ed., 2018, 57, 4083-4087), eventually yielding membranes with low permeance. Secondly, high temperature treatment often rapidly leads to a kinetically stable state, but with poor orderliness. To confirm this, we characterized TFP-PDA membranes crystallized at 155 C through XRD and BET. As expected, the crystallinity and surface area. of membranes crystallized at 155 C are lower compared to those of crystallized at 145 C. We felt justified in assuming that growth of COF particles in the pores and low crystallinity are the main causes of low permeance for membranes prepared at 155 C. Therefore, a temperature of 145 °C was determined as the optimum temperature with a rational balance between permeance and rejection.
We have revised the manuscript accordingly.
"At 155 °C, the low flux may arise from the partial blockage of pores by intergrown COF particles, also reported in previous literatures 10 . Secondly, treatment at high temperature leads to kinetically stable product rapidly, but with poor orderliness. To confirm this, we characterized TFP-PDA membranes crystallized at 155 °C through XRD and BET. Indeed, the crystallinity and surface area of membranes crystallized at 155 °C is lower than those of membranes crystallized at 145 °C. We assume that growth of COF particles in the pores and low crystallinity at 155 °C render membranes low permeance (Supplementary Fig. 8)." We also included an additional figure in the Supplementary Information file.
Supplementary Fig. 8. XRD and BET surface area of TFP-PDA membranes pre-assembled at In general, this work is not particularly novel and presents a very similar structure as the authors' previous paper. It may not be suitable for Nature Communications.
We thank the reviewer for presenting us with this valuable opinion of our work. We would like to mention however, that the fabrication of COF powders and the fabrication of COF membranes differ significantly, as explained in our reply to this reviewer's first remark/question.
The motivation of this work was based on decoupling the polymerization and crystallization steps of the process and replacing liquids with vapors during crystallization as an efficient alternative to existing methods. We would like to clarify that the idea in this work is very innovative and as we have pointed out in our rebuttal, unprecedented, aiming to provide a generic, efficient, facile and broad strategy to fabricate COF membranes. We designed a phase-switching (switching from liquid phase to vapor phase) strategy to decouple polymerization and crystallization steps for the first-time during powders which are very difficult, even impossible in some cases, to process into continuous membranes. In this work, we present an alternative strategy to fabricate highly crystalline COF membranes through a two-step (phase-switching) procedure, free of all the specific restrictions for amine and aldehyde monomers that were present in all of the mentioned references by the reviewer.
Reviewer #3 (Remarks to the Author): In this work, the authors report a phase-switching strategy to fabricate covalent organic framework (COF) membranes through a two-step procedure. The first step in liquid phase is primarily responsible for polymerization and the second step in vapor phase is primarily responsible for crystallization, which delicately decouples polymerization process and crystallization process. The two-step procedure has some distinct superiority to the commonly used one-step procedure in fabricating defect-free and ultrathin membranes. Using phase-switching strategy, highly crystalline, more compact and continuous membranes were fabricated. Accordingly, ultrafast molecular transport was realized, as manifested by water permeance of 402 L m-2 h-1 bar-1 and acetonitrile permeance of 519 L m-2 h-1 bar-1, which is among the ever-reported highest separation performance for COF membranes. The phase-switching strategy was validated by fabricating two kinds of highly crystalline COF membranes with different pore apertures. The strategy in this work represents a striking breakthrough in the precise construction of advanced COF membranes. This paper is meticulously organized and elegantly written and I read it with great joy. I firmly believe this interesting and innovative work will be highly appreciated by broad scientific communities in chemistry, materials science and membrane technology. Overall, this work is highly suitable for publication in Nature Communications after addressing the following minor revisions. Reply: We thank the reviewer for these highly positive remarks and valuable guidance.
1. The authors employed the β-ketoenamine-linked Schiff base COF in their work. More explanations may be required for the readers to more clearly understand why the authors preferred this type of COF over the other types? Reply: We thank the reviewer for this valuable guidance. In this work β-ketoenamine-linked Schiff base COFs were chosen primarily due to two reasons: (I) the initial reversible Schiff base reaction yields an enol which leads to crystalline arrangements which is subsequently converted to a ketone-enol Some COFs such as boronated COFs are highly sensitive to even small amounts of moisture, hindering their application in aqueous environments. To better convey this, we have updated the text in the revised manuscript accordingly.
"The initial reversible Schiff base reaction yields an enol form leading to crystalline arrangement which is subsequently converted to keto-enol tautomer form. Moreover, these COFs exhibit excellent chemical stability under harsh acidic and basic conditions 11,12 , unlike the boronated COFs, which are highly sensitive to even small amount of moisture and thus difficult to be utilized in aqueous environments 13 ." 2. Can the phase-switching strategy be applied to other kinds of COF membranes in principle? Reply: We thank the reviewer for this valuable guidance. We fabricated COF-LZU1 membranes based on imine linkage to further confirm its feasibility. The membranes crystallinity was confirmed by XRD and the completion of reaction was verified by FT-IR. We have modified the manuscript and Supplementary Information files accordingly.
"The reliability of our two-step procedure via phase-switching was further confirmed by the fabrication of COF-LZU1 membranes with defect-free surface and high crystallinity as evident from the XRD data with characteristic peaks similar to the literature 14 . The completion of reaction was confirmed through FT-IR with the disappearance of peaks from the initial monomers and the formation of imine bonding ( Supplementary Fig. 9). Moreover, the method section was also modified accordingly, as shown below.

"Methods
Pre-assembly step in liquid phase toward pristine membranes "DMAc was chosen in this step due to high solubility of amine and aldehyde monomers in DMAc which reduces reaction volumes (TFP 23mg/ml, PDA≈25mg/ml and TTA≈17 mg/ml) and relatively high boiling point of DMAc, which wins sufficient evaporation time for the completion of the polymerization at this step." 4. Have the authors tested the rejection performance of dyes in organic solvents? Reply: We would like to thank the reviewer for another valuable question. We have tested the dye, PPH-IX, in organic solvent to make sure that the membrane's pores are broadly suitable for dye rejections.
The text in the manuscript was revised accordingly.
"Dyes can form aggregates in aqueous solution at high concentrations 15,16 , therefore, an ethanolic solution of PPH-IX was used to confirm the broad applicability of COF membranes for the dyes rejection 17 ." Some mistakes: (1) The scale bar of crosse-section SEM in Fig. 2c, f should be corrected. Reply: We express thanks to the reviewer for catching this mistake and apologize for this oversight. We have corrected the mistake in the revised manuscript. (2) In line 125, XRD was written twice. Reply: We appreciate the reviewer's watchful eye in catching this error. We have corrected the error in the revised manuscript.
"via phase switching was evaluated through FTIR, XRD, X-ray photoelectron spectroscopy (XPS)," In addition, we would like to make some revisions on the author affliations and acknowledgements, as shown below.
In the author affiliations part, we added "Haihe Laboratory of Sustainable Chemical Transformations, Tianjin 300192, China" due to its financial support and removed "Joint School of National University of Singapore and Tianjin University, International Campus of Tianjin University, Binhai New City, Fuzhou 350207, China" and changed the order of "Zhejiang Institute of Tianjin University, Ningbo, Zhejiang 315201, China", according to their contributions.