Double charge flips of polyamide membrane by ionic liquid-decoupled bulk and interfacial diffusion for on-demand nanofiltration

Fine design of surface charge properties of polyamide membranes is crucial for selective ionic and molecular sieving. Traditional membranes face limitations due to their inherent negative charge and limited charge modification range. Herein, we report a facile ionic liquid-decoupled bulk/interfacial diffusion strategy to elaborate the double charge flips of polyamide membranes, enabling on-demand transformation from inherently negative to highly positive and near-neutral charges. The key to these flips lies in the meticulous utilization of ionic liquid that decouples intertwined bulk/interfacial diffusion, enhancing interfacial while inhibiting bulk diffusion. These charge-tunable polyamide membranes can be customized for impressive separation performance, for example, profound Cl−/SO42− selectivity above 470 in sulfate recovery, ultrahigh Li+/Mg2+ selectivity up to 68 in lithium extraction, and effective divalent ion removal in pharmaceutical purification, surpassing many reported polyamide nanofiltration membranes. This advancement adds a new dimension to in the design of advanced polymer membranes via interfacial polymerization.

the conventional O/N ratio, and the positive charge should not be high; As the etching time prolongs, the O/N decreases significantly, reaching a minimum of 0.77.It should indeed show positive charge, but the crosslinking degree of the membrane should be very low, failing to form a well crosslinked structure.So how does the author explain it?2.And membranes with different charges exhibit different performance in different fields, which is a conventional membrane performance and not very meaningful.The author should study the universality of this method, such as other additives in to PIP solution.3.The author should focus on studying the first charge reversal process from negative to positive, especially to the first unlabeled "neutrally charged", and the research in the latter interval is obviously does not have much meaning, whether in theory or practical application.That is to say, replacing deionized water with IL at a cost of tens of millions of times higher, to prepare the reported nanofiltration membranes with good or ordinary performance for dye/salt separation.4. According to the XPS analysis (Supplementary Table 1), there are certain IL residues on the membrane surface, which contain lots charged groups.Is the change in separation performance of these membrane related to these residual IL? 5.As shown in Supplementary Table 7, the PA-IL0% membrane, which is a nanofiltration membrane prepared by conventional interfacial polymerization (concluding vacuum assisted), has a MWCO of only 212 Da and exhibits a significant negative charge on its surface (-17.5 mV).However, although it has a high Na2SO4 rejection of 98.5%, it is still significantly lower than the literature results under similar conditions (generally higher than 99%, even compared to the results of author's research group).Although there is only a difference of about 1% in Na2SO4 rejection, this usually results in several times the pure water flux change.So, are there any issues with the author's membrane preparation process?If the author did it intentionally, please provide an explanation.6.The author compared the membrane performance with the results of relevant literature, the latest relevant literature results should be added in Supplementary Figure 18.
Reviewer #3: Remarks to the Author: This work reports positive polyamide membranes by introducing ionic liquid as the co-solvent into the aqueous phase during interfacial polymerization.By altering the content of ionic liquid in the solution, the membrane surface charge can be well tuned from negative to positive and then back to nearly neutral.This leads to interesting performance in separating mono/di valent ions, especially enabling high selectivity of Mg2+/Li+.The authors should shift the focus of the paper to these intriguing results, as in the current format, the rationale of creating positively charged surfaces is unclear.For the paper to be published in Nature Communications, the manuscript should be revised to address the target separation challenge and the specific comments below.1. Fig. 2c shows higher R2NH content for PA-IL-40%, indicating a lower crosslinking degree, which results in faster permeance but higher MWCO, i.e., looser pores (Supplementary Table 6).On the other hand, the rejection of MgCl2 increased for PA-IL-40% membranes as the surface charge became more positive (Supplementary Table 7).The authors should discuss more in details of the completing effect between pore size and surface charge on the solute rejections.2. Interfacial polymerization is a self-limiting process, i.e., the growth of polyamide layer can prohibit or slow down the diffusion rate of amine.How does the reaction rate affect the diffusion of PIP with increasing IL content and thereby affecting the surface charge and membrane thickness?In this context, using 80 g.L-1 of PIP (Line 190 Page 7) as a control experiment is inappropriate as the reaction rate would be so fast to form a dense polyamide layer that could prohibit/slow down further diffusion of PIP, resulting in less positively charged surface (Supplementary Figure 12).3. How could the authors eliminate the interference of surface charge from ionic liquid as significant amount of fluorine was detected in the polyamide layer (Supplementary Table 1)? 4. The estimation of membrane pore size from PEG rejections is inappropriate since the conformation of PEG can be varied in different solvents or in the bulk versus at the membrane surface.Also, their molecular size has a broad distribution, for which using the mean solute size to estimate the membrane pore size could be inaccurate and misleading. 5. Why did the PA-IL-10% membranes perform the best selectivity for Mg2+/Li+ separation (Fig. 5b) but the lowest permeance (Supplementary Table 8)?If the surface charge is the key for cation separations, membrane surfaces with higher positive charge at elevated ionic liquid content (e.g., PA-IL-40% membranes in Fig. 2b and Supplementary Table 6) should exhibit higher selectivity.Please justify.6. Please add the data of LiCl rejection to Fig. 5a and Supplementary Table 7.

Reviewers' Comments:
Reviewer #1: Remarks to the Author: [Note from the Editor: Reviewer #1 was also asked to assess the response given to reviewer #2]: The authors have responded to most questions.However, two questions raised by reviewer 2 has not been well addressed.
Comment 2: The reviewer commented that the O/N ratio of the membrane is only 0.77 and he/thinks the crosslinking degree of the membrane should be very low, failing to form a well-crosslinked structure.The authors replied they had put forward a new approach in another paper (in submission) to calculate the crosslinking degree.But the equations are difficult to understand (I couldn't understand).The authors should provided detailed explanations and equations in this manuscript.It is not reasonable to ask readers to wait for another paper to published before they could understand this paper.
Comment 3: The reviewer said 'And membranes with different charges exhibit different performance in different fields, which is a conventional membrane performance and not very meaningful.The author should study the universality of this method, such as other additives in to PIP solution.'The reviewer asked for the universality of their method.The authors used another ionic liquid ([Emim][BF4]), similar to [Bmim][BF4] that they used in this paper.The reviewers should employ other solvents and summarize the characteristics needed to arrive the 'charge flip' phenomenon.
Reviewer #3: Remarks to the Author: The revision based on the additional experiments and simulation shows convincing evidence and results, which have addressed all the critical comments.Therefore, I would recommend for publication.
[Note from the Editor: Reviewer #3 was also asked to assess the response given to reviewer #2]: Whilst the authors answered most of the comments, the key question still remains as whether the membrane performance in this manuscript were significantly better/or ever better than the existing membranes reported in literature, in the compromise of using pricy ionic liquids.If the different concentrations were selectively chosen for a fair comparison to the membranes made from IL system, what is the performance of the PIP membranes made from "common" conditions (i.e., membranes with >99% Na2SO4 rejections)?Is the selectivity of SO42-/Cl -1 in IL-membranes still better than conventional PIP membranes?In Supplementary Fig. 22, what is the parameter in x-axis at the bottom?If the figure is plotted against permeance, there is barely any advantage of the IL-membranes in this manuscript as compared to the reported membranes in literature, especially membranes made from novel monomers and substrate modification

Point-to-point response to the reviewers' comments
Reviewer #1: Comment 1: The authors have responded to most questions.However, two questions raised by reviewer 2 has not been well addressed.

Response:
We are pleased that the reviewer appreciates our responses.We will provide detailed answers to the two issues raised by the reviewer.
Comment 2: The reviewer commented that the O/N ratio of the membrane is only 0.77 and he/thinks the crosslinking degree of the membrane should be very low, failing to form a well-crosslinked structure.The authors replied they had put forward a new approach in another paper (in submission) to calculate the crosslinking degree.But the equations are difficult to understand (I couldn't understand).The authors should provide detailed explanations and equations in this manuscript.It is not reasonable to ask readers to wait for another paper to published before they could understand this paper.

Response:
We thank the review for pointing out this crucial aspect.Generally, the PIP-TMC based polyamide membranes possess a complicated molecular structure comprising crosslinking structures (X), linear structures (Y) and two terminate structures with either amino (Tamino) or carboxyl groups (Tcarboxyl), as depicted in Fig. R1.Conventionally, the calculation of the degree of network crosslinking (DNC) includes only the crosslinking structures (X) and linear structures (Y), yet neglecting the terminal structures of Tamino or Tcarboxyl: Given that each X consists of three O atoms and three N atoms, and each Y has four O atoms and two N atoms, the O/N ratio can be expressed by the following equation: Consequently, DNC can be further expressed as: As a consequent, the membrane's crosslinking degree is evaluated using the O/N ratio on the basis of Equation R3.However, owing to the fact that each Tamino contributes to one N atom and each Tcarboxyl adds four O atoms, these N and O atoms have no contribution to cross-linking degree.Consequently, employing the total N and O atoms ratio from Xray photoelectron spectroscopy (XPS) for DNC calculation lacks precision.This assertion is corroborated by several studies where DNC values of polyamide membranes are negative but still exhibit good separation performance (Adv. Mater. 2018, 30, 1705973;ACS Appl. Mater. Interfaces 2020, 12, 25304-25315;Desalination 2020, 491, 114345;and J. Mater. Chem. A 2020, 8, 3238-3245).
To address this issue, we propose that the concept of the amide group ratio (rN-C=O) provides a more accurate depiction of the crosslinking degree or compactness in polyamide membranes.Here, rN-C=O is defined as the ratio of the number of amide groups (namide) in the polyamide membrane to the total number of functional groups (amide, amino, and carboxyl groups, shown in Fig R2 ): From this equation, we found that a higher rN-C=O means more formation of amide groups, implying a higher crosslinking degree and increased membrane compactness.
Thus, rN-C=O can be calculated from the following equation: Employing this method, we confirmed that the proportion of surface and internal amide bonds in the PA-IL40% membrane is approximately 80.3% and 70.9%, respectively (Supplementary Table 3).These values align with those in the PA-IL0% membrane (about 76%) known for its high crosslinking density.These results demonstrate that the PA-IL40% membrane also possesses a robustly crosslinked structure.Following the reviewer's suggestion, we have detailed analytical method of the amide group ratio in the revised Supplementary Information.Response: We agree with the reviewer that the universality of 'charge flip' phenomenon of polyamide membranes is much meaningful compared to the separation performance.

Supplementary
Despite this, our IL-mediated charge flip approach demonstrates the capability of both positively and negatively charged polyamide membranes to exhibit remarkable monovalent/divalent ion selectivity over the conventional polyamide membranes.For instance, our PA-IL5% membrane delivers an outstanding Cl -/SO4 2-selectivity exceeding 470, and our PA-IL10% achieves a Li + /Mg 2+ selectivity greater than 68, surpassing many reported in existing studies (Supplementary Fig. 22).
Supplementary Fig. 22 Comparison of the Li + /Mg 2+ selectivity and water permeability of PA-IL membranes with that state-of-art positively charged nanofiltration membranes.
This includes membranes synthesized via one-step IP using PEI, novel monomers, or comonomers, as well as membranes fabricated through multi-step IP featuring interlayer modification or post-grafting.
Next, we will answer the comment regarding the universality of the charge flip mechanism.To validate the universality of the charge flip mechanism, we incorporated sodium dodecyl sulfate (SDS) and glycerol as aqueous additives in the interfacial polymerization process, in which the former is capable of facilitating interfacial diffusion via the reduced interfacial tension and the latter can harness its intrinsic viscosity to suppress bulk diffusion.Consequently, we hypothesize that such an integrated strategy can effectively decoupling interfacial and bulk diffusion, similar to our IL-enabled strategy.Briefly, we employed an SDS concentration of 4.5 mM and added glycerol at a volume fraction of 17%, effectively replicating the interfacial tension and viscosity of a 40 v/v% [Bmim][BF4]/water solution.The concentrations of PIP and TMC were maintained at 8 g L -1 and 2.4 g L -1 , consistent with the conditions detailed in our manuscript.Under these specific conditions, the resulting polyamide membrane exhibits a zeta potential of approximately +27.8 mV at pH 6 (PA-Gly17%-SDS in Fig. R3), contrasting with previous studies where polyamide membranes produced using either SDS or glycerol alone are negatively charged (Nat. Commun. 2020, 11, 2015and J. Membr. Sci. 2021, 627, 119142).This finding underscores the importance of decoupled bulk/interfacial diffusion in achieving a charge flip of polyamide membranes.
Taken together, our current experimental and simulation results indicate that reducing interfacial tension and increasing bulk viscosity can achieve a decoupling of interfacial and bulk diffusion of amine monomer, leading to charge flip of polyamide membranes.
This phenomenon demonstrates the universality for both alkyl imidazolium tetrafluoroborate ionic liquid/water solvent systems and the SDS/glycerol additive systems.From a broader perspective, we believe that this work offers an insightful direction for exploiting new charge flip of polyamide membranes by designing interfacial polymerization systems that are capable of reducing interfacial tension and increasing bulk viscosity simultaneously.
Following the reviewer's suggestion, we have included these results and discussions in the revised manuscript.This variation is attributable to the physicochemical properties of [Emim][BF4], exhibiting slightly higher surface tension (57.5 ± 0.3 mN m -1 ) and lower viscosity (1.72 ± 0.02 mPa•s).To further extend universality, we substituted ionic liquids with a combination of sodium dodecyl sulfate (SDS) and glycerol as additives in the interfacial polymerization, in which the former is capable of facilitating interfacial diffusion via the reduced interfacial tension and the latter can harness its intrinsic viscosity to suppress bulk diffusion.This approach also successfully achieves a charge flip from negative to positive on the polyamide membrane (Supplementary Fig. 14).These results strongly validate the decoupled bulk/interfacial diffusion mechanism as a reliable method for modulating the surface charge of PIP-based polyamide membranes." Reviewer #3: Comment 1: The revision based on the additional experiments and simulation shows convincing evidence and results, which have addressed all the critical comments.
Therefore, I would recommend for publication.

Response:
We are grateful to the reviewer for his/her constructive feedback, which significantly contributed to the enhancement and refinement of our work.We also appreciate the positive assessment of our efforts and revisions.
Comment 2: Whilst the authors answered most of the comments, the key question still remains as whether the membrane performance in this manuscript were significantly better/or ever better than the existing membranes reported in literature, in the compromise of using pricy ionic liquids.If the different concentrations were selectively chosen for a fair comparison to the membranes made from IL system, what is the performance of the PIP membranes made from "common" conditions (i.e., membranes with >99% Na2SO4 rejections)?Is the selectivity of Cl -/SO4 2-in IL-membranes still better than conventional PIP membranes?
Response: We appreciate the reviewer's insightful comment.We agree with that it is not a fair comparison to the membranes made from different methods using the similar monomer concentrations.In our study, we have optimized the concentrations of PIP and TMC to be 8 g L -1 and 2.4 g L -1 , respectively, which indeed are distinct from the optimal conditions for traditional alkane-water interfacial polymerization.In response to the reviewer's suggestion, we have synthesized polyamide membranes using three typical monomer concentrations in alkane-water interfacial polymerization systems and have compared their selectivity of Cl -/SO4 2-with our PA-IL membranes.Briefly, TMC concentration in hexane was fixed at 1.5 g L -1 , while the concentrations of PIP aqueous solutions were changed from 1.0 g L -1 to 1.5 g L -1 , and 2.0 g L -1 .
The experimental results show that polyamide membranes synthesized using the abovementioned concentrations for conventional interfacial polymerization all exhibit Na2SO4 rejection rates exceeding 99%, with NaCl rejection rates below 30% (Table R2).These membranes demonstrated Cl -/SO4 2-selectivity ranging from 260 to 350, which were lower compared to our IL systems (Table R3, Fig. R4).

Fig. R4
Comparative illustration of Cl -/SO4 2-selectivity in polyamide membranes synthesized from conventional interfacial polymerization (denoted as CIP-PIP followed by PIP concentration in g L -1 ) and those obtained from IL-based interfacial polymerization.

Response:
We thank the reviewer for pointing out this issue.The initial Supplementary Fig. 22 was designed to illustrate only the Li + /Mg 2+ selectivity of polyamide membranes Reviewer #1: Comment 1: The authors have addressed Comment 2 raised by reviewer 2; however, response to Comment 1 is still not convincing; in fact, it is scientifically incorrect.

Response:
We appreciate the reviewer's acknowledgment of our efforts in addressing Comment 2. Following the reviewer's suggestion, we have made a more scientifically robust and comprehensible clarification to address the concerns raised.
Comment 2: The authors may misunderstand the traditional method of correlating XPS results with crosslink degree.Usually, the method tries to push to two limits: fully crosslinked, and fully linear structure.In the fully crosslinked state, due to the large molecular weight (which is validated by the insolubility in solvents), the contribution of terminal groups to the overall elementary distribution is negligible.
Response: In the traditional method, the degree of crosslinking in polyamide networks is assessed by considering both fully crosslinked and fully linear structures, as illustrated in Equation R1: where X and Y denote the crosslinked and linear structures within polyamide, respectively (Desalination 2011, 278, 387-396;Science, 2015Science, , 348, 1347Science, -1351)).This traditional view hypothesizes that the "fully crosslinked state" exhibits a minimal impact of terminal groups on the elemental composition of polyamide membranes.
However, in actual polyamide membranes, the "fully crosslinked state" is more theoretical concept than an achievable reality.Achieving a completely crosslinked structure in polyamide membranes is nearly impossible due to inherent challenges, such as the hydrolysis of acyl chloride monomers and the steric hindrances.This assertion is supported by numerous studies reporting O/N ratios greater than 1.3, suggesting a crosslinking degree under 60% (Table R1).Notably, some studies report O/N ratios exceeding 2, which, in theory, would indicate a negative degree of crosslinking.Despite this, these polyamide membranes exhibit superior performance, with water flux rates exceeding 30 L m -2 h -1 bar -1 and Na2SO4 rejection rate above 99% (J.Mater.Chem. A 2020, 8, 3238-3245).These findings manifest that actual polyamide membranes significantly deviate from being fully crosslinked, challenging the traditional method's effectiveness in correlating the degree of crosslinking with membrane compactness and separation efficiency.Moreover, considering that current characterization methods may not definitively ascertain whether carboxylic groups in polyamide network are integrated within linear structures or positioned terminally, we have confirmed the significant presence of carboxylic terminal groups through molecular dynamics simulations.For instance, we constructed a model of a PIP-TMC polyamide membrane with a reaction degree of 95% using the modeling approach from our previous work (Angew. Chem. Int. Ed. 2021, 60, 14636).According to traditional calculations, the degree of crosslinking of this model stands at 93.6%, yet terminal groups account for 8.5% of the total functional groups and 83.6% of all unreacted groups.This underscores that even highly crosslinked polyamide networks may harbor a substantial number of terminal groups.
More importantly, given that amines are difunctional monomers, unreacted amine groups can only exist as terminal groups, unable to contribute to the formation of linear structures.This is fully validated by our IL-enabled polyamide membranes exhibiting pronounced positive charges derived from these terminal amine groups.
Taken together, these results all strongly demonstrate that the contribution of terminal groups to the crosslinking degree of polyamide should not be negligible.

Fig. R1
Fig. R1 Schematic representation of all potential molecular structures in PIP-TMC polyamide membrane.Here, X, Y, Tamino, and Tcarboxyl represent the crosslinking structure, linear structure, amino-terminated structure, and carboxyl-terminated structure, respectively.

Fig. R2
Fig. R2 Schematic representation of all potential functional group structures in polyamide and their corresponding N and O atom counts.

Comment 3 :
The reviewer said 'And membranes with different charges exhibit different performance in different fields, which is a conventional membrane performance and not very meaningful.The author should study the universality of this method, such as other additives in to PIP solution.'The reviewer asked for the universality of their method.The authors used another ionic liquid ([Emim][BF4]), similar to [Bmim][BF4] that they used in this paper.The reviewers should employ other solvents and summarize the characteristics needed to arrive the 'charge flip' phenomenon.

Fig. R1
Fig. R1 Full-atom molecular model of a PIP-TMC membrane under the reaction degree of 95%.The amine and carboxylic terminal groups are represented with a bold balland-stick model, where carbon, oxygen, nitrogen, and hydrogen are depicted in cyan, red, blue, and white, respectively.

Table 3
Elemental profile and the deconvoluted amount of amide groups in the inner PA-IL40% membrane.

Table R1 .
Some reported O/N ratios, degree of crosslinking, and separation performance of polyamide membranes.