Realizing precise ion separation in polyamide nanofiltration membranes via stage control reactions

Abstract

Thin-film composite nanofiltration membranes face a trade-off phenomenon between ion selectivity and permeability due to the structural constraints of single monomers during interfacial polymerization (IP). Inspired by homogeneous precipitation, we decouple disorderly competitive reactions of comonomers through using in situ–generated H⁺ during ultrafast IP processes as equilibrium-shifting inducers for the enamine reaction, thereby regulating the reaction sequence and relative amount of primary/secondary amine monomers. Combining the structural advantage of polyethyleneimine and piperazine monomers, the separation layer had a large free volume, high-density homogeneity, well-tuned nanopores, and tailored charge distribution. The staged-regulated membrane exhibited high water permeance and could even adapt to ion separation in ultrahigh-salt solutions (Mg/Li = 50, 20,000 parts per million), with a notable Mg²⁺/Li⁺ selectivity improvement of more than 1600% over that of the control, directly mixed dual aqueous monomer-prepared membrane. This stage control strategy for precise nanofiltration membrane synthesis can provide extensive flexibility in modulating the IP process for application-specific membrane structure design.

Designed stage control IP can synthesize homogenous, tuned nanopore and tailored charged membranes for precise ion separation.

INTRODUCTION

Nanofiltration (NF) is a cost-effective and environmentally friendly separation method applied in wastewater treatment and valuable resource recovery (1, 2). Numerous studies emphasize the importance of both water-solute separation and solute-solute separation as value-added processes for water treatment (3, 4). A typical example is the extraction of high-value lithium resources from salt-lake brines with high Mg/Li mass ratios (5). Now, thin-film composite (TFC) polyamide (PA) membranes fabricated via interfacial polymerization (IP) have become the gold standard in both market and academic research (6). The separation layer of the TFC membranes can be flexibly personalized according to the actual requirements (7). However, the inherent structural limitations of polymer molecules caused by monomers and rapid and less regulated monomer reaction behavior undoubtedly increase the difficulty of adjusting the PA layer structure. Therefore, TFC PA membranes generally exhibit an inherent “trade-off” effect between flux and selectivity (8, 9). Various methods—such as the introduction of interlayers (10), multiple additives (7), post-treatment (11) and process optimization (12, 13)—have been used to modulate the separation layer structure in a single aqueous phase monomer system. However, the properties and structure of monomers determine that the corresponding membranes have specific application scenarios (14–16). For example, piperazine (PIP)/trimesoyl chloride (TMC)–based and polyethyleneimine (PEI)/ TMC-based TFC membranes are suitable for monovalent/multivalent anion separation and multivalent cation removal, respectively. The molecular structure and properties of polymer film, which are determined by the structure of the monomer, are difficult to change. In certain circumstances, the molecular structure of polymeric membranes can be effectively tuned through the introduction of comonomers, breaking the trade-off phenomenon for efficient separations. Nevertheless, the reactive behavior of multiple monomers is more difficult to control. The inherent differences in monomer properties such as reactivity or diffusion capacity would impede the control over monomer reaction sequence and relative amount, which make it difficult to achieve the desired fine-tuned membrane microstructure necessary for optimal separation performance. Therefore, obtaining a more desirable membrane structure and properties requires a method to reduce the adverse influence of competitive reactions through regulating the dual aqueous phase monomer’s reaction behavior in IP. However, prevailing strategies for regulating monomer behavior whether through intermolecular interactions (e.g., hydrogen bonding and electrostatic forces between substrates/additives and monomers) or interfacial condition alterations (e.g., adding surfactants) cannot achieve control over the reaction behavior of a specific monomer because all molecular components are affected equally. Now, technical challenges persist in manipulating reaction kinetics for dual aqueous phase monomers, and no successful precedents have been reported.

Inspired by the homogeneous precipitation strategy used in inorganic chemistry to control precipitant release via chemical equilibrium shifts, we introduce a stage-regulated IP strategy. Specifically, this approach uses the in situ–generated H⁺ arising from the participation of the first aqueous phase monomer PEI in the IP reaction to trigger the release of the second aqueous monomer, PIP, which is pre-immobilized via an enamine structure, thereby decoupling the reaction stages. The competitive side reactions between PEI and PIP are minimized, and the reaction sequence and relative amounts of dual aqueous phase monomers during the IP process are regulated. Compared with directly mixed dual aqueous phase aqueous monomer-prepared membrane (DMM), staged-regulated membrane (SRM) presents better structures with high free volumes, precisely regulated pore size distributions, controlled charge distribution, and uniform density distributions. Changes in the nanostructure of NF membranes have resulted in extremely high separation accuracy in the field of ion separation (17, 18).

This research develops a promising method to control the reaction sequence and relative amount of dual aqueous phase monomers, thereby avoiding the competitive side reactions to precisely regulate membrane structure and enhance separation performance. By both experiment and computational simulation, revealing the regulatory mechanism. This work not only provides the insights for designing next-generation high-performance NF membranes by combining structural advantages of distinct monomers, but also offers inspiration for the membrane design and application in the currently most challenging separations worldwide.

RESULTS

Stage control process for the monomer’s reaction behavior

By using the reversible reaction of enamines affected by pH value (fig. S1) (19, 20), the reaction sequence and relative amount of PEI and PIP were regulated to prepare TFC NF membranes that combine the structural advantages of both monomers. Specifically, PIP monomers are first prebound with aldehyde-rich oxidized dextran nanoparticles (ODNPs; fig. S2) via an enamine forward reaction under PIP-provided alkaline conditions (pH ≈ 10) (step 1 in Fig. 1A) (21, 22), which can be confirmed from the disappearance of the adsorption peak of −CHO groups at 1731 cm⁻¹ and the appearance of the adsorption peak of C═C at 816 cm⁻¹ (Fig. 1B). Subsequently, the PEI monomer is added to configure an aqueous solution. Because the majority aldehyde sites on ODNPs are occupied by PIP monomers, controlling the ODNPs/PIP ratio could prevent the formation of large aggregates through reaction between aldehyde groups on ODNPs and the amino groups of PEI (fig. S3). In addition, the aqueous solution can remain stable for a long time (fig. S4), which is critical for the IP process, primarily due to the volume and osmotic pressure restriction effects of polymer nanoparticles and the electrostatic repulsion provided by charged properties (fig. S5). The free PEI monomer, as the main reactant, can first react with TMC to form the initial separation layer and release the by-product H⁺; meanwhile, the enamine structure is reversibly degraded as the pH value decreases (step 2 in Fig. 1A and fig. S6) (22). The ultraviolet (UV) absorption peak shift of ODNPs-PIP from 267 nm (C═C) to 286 nm (−CHO) with pH value demonstrates the above process (Fig. 1C). Next, the released PIP monomer, as the main reactant, participates in the main stage of the subsequent IP reaction, while the PEI monomer is immobilized by forming Schiff base polymers with ODNPs (step 3 in Fig. 1A and fig. S7) (23). The aqueous phase UV absorption peak shifts to 330 nm upon pH value reduction, matching the ODNPs-PEI compounds, confirming the enamine-to-Schiff base conversion (Fig. 1D and figs. S8 and S9). In addition, some chemical structure and morphology characterization, including micrograph and x-ray photoelectron spectroscopy, also resolved and confirmed this stage control reactions (figs. S10 and S11). In situ monitoring of the IP interface indicated the necessary pH value changes for the aforementioned processes to occur (fig. S12) (24). It is also demonstrated that the rapid reversibility of the enamine structure occurs at the second-level timescale under acidic conditions (fig. S13), with a duration notably shorter than that of the 30-s IP reaction. This ensures that the reversible reaction can be effectively triggered and completed during the IP process, enabling control over the amount and sequence of monomer reaction. In this view, the regulation of the staged reaction of the dual aqueous phase monomers can be successfully achieved in this study.

Fig. 1. Preparation of NF membranes regulated by a reversible enamine reaction.

(A) Mechanistic schematic diagram of the composite NF membranes prepared from the reversible enamine reaction modulating the reaction order of the PEI and PIP monomers. (B) FTIR spectra evidence for the enamine reversible reaction of PIP monomers with aldehyde groups on oxidized dextran. (C) UV spectroscopy tested the direction of equilibrium shift of PIP monomers with aldehyde groups on oxidized dextran at different pH value. (D) UV spectroscopy determined competitive behavior of the enamine reversible reaction with the Schiff base reaction at different pH value. (E) Comparison of the chemical structure distributions of staged-regulated membranes and directly mixed monomer-prepared membranes.

In comparison, SRM would present heterogeneity, with PIP components as the main body of the whole PA separation layer, whereas the PEI components are located mainly at the bottom (Fig. 1E). DMM has a disordered distribution of PEI and PIP components due to the competitive reaction between both monomers. The shift in the chemical binding energy of the amide bond at different locations in the separation layer of both membranes provides evidence for the abovementioned distribution of monomer structures (figs. S14 and S15) (25). In addition, to further demonstrate the difference between SRM and DMM, without changing the composition of the organic phase, the membranes fabricated using only aqueous phase monomers were named NF-PIP and NF-PEI, respectively, according to the optimal preparation conditions for SRM (fig. S16 and table S1), and successful preparation was demonstrated by infrared spectroscopy (fig. S17). The characterization of the membrane surface constituents and mechanical strength also revealed that the properties of SRM appear to be more similar to those of NF-PIP, indicating that the PIP monomer rather than PEI monomer is the dominant reactant in the late IP stage (figs. S18 to S20). Conversely, direct mixing monomers can lead to the presence of excessive PEI components within the DMM, which is more similar in nature to NF-PEI.

Membrane structure analysis

To investigate the effect of PEI and PIP ratio on the structure of the separation layer, the polymer models of the membrane separation layer constructed from pure PEI, a PEI-to-PIP with series ratio on the main chain, and pure PIP as the monomer were constructed (fig. S21). Figure 2 (A and B) illustrates the free volume distribution and the fractional of various simulated polymers. The polymer model formed by small-molecule alicyclic PIP and aromatic TMC is the most rigid among various models. Such a relatively rigid structure is prone to generate a large amount of free volume within the membrane (26, 27). In contrast, the introduction of PEI notably decreases the fractional free volume (FFV), and only a high PIP content ensures an elevated level favoring high permeability. Higher PIP content also limits the movability of the polymer molecular chains (Fig. 2C and fig. S22), which helps to minimize swelling and maintain pore stability of the membrane material during operation. Moreover, a small amount of PEI can notably regulate the mass density homogenization of polymer (Fig. 2D). For instance, the polymers whatever with 80 and 20% PIP all exhibit nearly identical mass density distributions, both notably outperforming inhomogeneous pure PIP-based polymers in homogeneity. The control of nanoscale polymer homogeneity is key to maximizing water permeability without sacrificing salt selectivity, and a uniform distribution of density facilitates water transport through the membrane (28). In summary, large PIP and little PEI allow polymer structures with high free volumes, restricted chain segment movement and degrees of homogeneity to be obtained, which favors high permeability selectivity similar to the upper SRM component.

Fig. 2. Analysis of the nanofiltration membrane structure.

(A) Polymer models and free volume distribution, (B) FFV, (C) molecular chain diffusion coefficient, and (D) mass density distribution trends for different PIP and PEI ratios. (E) Three-dimensional reconstruction model of the SRM separation layer based on the secondary ion fragment C_xH_yN_z. The samples were prepared by attaching the separation layer to the silicon wafer to remove the polysulfone layer, and the top of the model actually represents the bottom of the separation layer and vice versa. (F) Counting of benzoate secondary ion fragments of SRM and NF-PIP at different detection times. Benzoic acid secondary ion fragments come from the TMC part of the separation layer.

The three-dimensional structure of the SRM PA layer was reverse reconstructed via depth profiling via time-of-flight secondary ion mass spectrometry (Fig. 2E). As expected, SRM exhibits heterogeneity with the upper most 75% being attributed to PIP-dominated polymers and only about bottom 22 to 25% being attributed to PEI-dominated polymers. No delamination was observed for the membranes prepared on the basis of one monomer (fig. S23). The bright yellow color on the bottom side of the separation layer can be attributed to the self-assembled Schiff base polymers of PEI and ODNPs, which corresponds to the actual observed dimensions. It is also expected that there is an abundance of interfacial channels between polyamides and these Schiff base polymers facilitating high permeability (29). The enrichment of the PEI components below the SRM would cause the high positive charge density in this part to favor the divalent cations rejection. In addition, the improvement in the uniformity of the SRM by a small amount of PEI introduction is also demonstrated. On the basis of the benzoic acid–based secondary ion fragments (Fig. 2F), a small change in the curve corresponding to SRM over time indicates greater homogeneity because of the small amount of PEI in the PIP-dominated polymer. The curve corresponding to NF-PIP is in high agreement with the large mass density distribution fluctuations in the molecular simulations, revealing inhomogeneity. In summary, a series of results imply that the enamine reversible reaction indeed mediates the advantageous combination of both monomer structures, which requires modulation of the reaction sequence and relative amount to be realized.

Undoubtedly, the pore sizes of SRM and DMM also vary according to the composition of the two monomers. By testing the selectivity of the composite membrane for a series of neutral molecules (fig. S24), the pore size distribution is fitted as a probability density function (PDF), and all pores with probability densities greater than 0.01 are counted statistically (Fig. 3A). In terms of the most probable pore size (defined as the pore diameter corresponding to the global maximum of PDF curve, D_mp), the pore size of DMM, similar to that of NF-PEI, is predominantly distributed in a smaller range. This can be attributed to the fact that the chain-like aliphatic PEI monomer easily moves and rearranges to form a tight and ordered structure during the posttreatment process. The pore size distribution of SRM is larger and more similar to that of NF-PIP, which is analogous to our previous conclusion that PIP components are the main body of the whole PA layer. It can be observed from the PDF that the pore size range of the membranes prepared with two monomers is much narrower than that of the other membranes, both DMM and SRM, and the latter has the smallest distribution (Fig. 3B). This can be explained by changes in the organization and arrangement of molecular interchain changes in the molecular structure. The disappearance of wide pores means fewer defects, as shown in Fig. 3C, most of the pores in the membrane are smaller than those in conventional divalent ions (e.g., 91% smaller than those in hydrated Mg²⁺, 90% smaller than those in hydrated Ca²⁺, and 88% smaller than those in hydrated SO₄²⁻). Moreover, owing to the easier dehydration process, the steric hindrance of the membrane cannot hinder the penetration of monovalent cations such as Li⁺ or Na⁺ (30). The SRM with properly regulated pore sizes is expected to exhibit high divalent ion rejection rates and considerable monovalent and divalent ion selectivity (31, 32). Although theoretical considerations suggest that DMM featuring a low D_mp and relative narrow pore size range should exhibit enhanced selectivity, experimental observations reveal low divalent salt rejection rates (fig. S25). As mentioned above, the polymer membranes containing extensive PEI structures show strong activity in the simulations; thus, the pore size of the polymer membrane may fluctuate with chain movement during the separation process. This molecular level mobility is corroborated in the macroscopic mechanical properties, where nanoindentation tests show that the DMM and NF-PEI have lower indentation modulus, indicating less resistance to movement of the molecular chain segments (fig. S26). This dynamic pore by the movement of the flexible molecular chains of the NF-PEI and the DMM with higher PEI fractions allow the ions to pass through, resulting in MgCl₂ rejection rate of only ~90%.

Fig. 3. Properties of nanofiltration membranes.

(A) Pore size distribution of the different NF membranes fitted from the rejection curve of neutral molecules. (B) Comparison of the pore size ranges of the NF membranes. (C) Comparison of hydration sizes of representative divalent ions with SRM pore size distribution integrals. The percentage represents how much of the pores are smaller than the hydration ions. (D) Zeta potential of the NF membranes. (E) Areal carboxylate and protonated amine groups densities determined by experiments. (F) Comparison of zeta potentials between front and opposite sides of SRM. (G) Morphological structure characterization of SRM and DMM.

The zeta potentials of various composite membranes are determined as shown in Fig. 3D. Undoubtedly, the electrical properties of NF-PEI and NF-PIP are the most positive and negative, respectively. DMM incorporating higher PEI content exhibits enhanced positive surface charge, whereas SRM incorporating less PEI in upper portion demonstrates intermediate zeta potential, positioned between DMM and NF-PIP membranes. The isoelectric point of the SRM membrane is about 5.01, indicating that the membrane surface to exhibit a certain degree of electronegativity as the pH ~6 for the actual test conditions. However, the areal density results for carboxylate groups (produced by ionization of carboxyl groups from hydrolyzed unreacted acyl chloride) and protonated amine groups (produced by protonation of unreacted primary/secondary amines or intrinsic tertiary amines) illustrate that the SRM has instead more positively charged groups internally (Fig. 3E) (18). Considering the structural heterogeneity of the SRM, it is due to the fact that zeta potential only responds to the relative electrical properties of the membrane surface rather than the overall, and the positive charge introduced by PEI is mainly distributed at the bottom of the separation layer. Therefore, the zeta potential at the bottom of the SRM was tested, and as expected, it demonstrated stronger positivity (Fig. 3F). In other words, SRM also has the advantage of the positive electrical properties of PEI-based membranes. During the separation process, the ideal pore size distribution of the SRM is conductive to improve the rejection of divalent cations, whereas the positively charged rich bottom provides further hindrance through the Donnan effect (33). As a result, SRM combines the positive electrical properties of the PEI-based membrane and the molecular chain stability of the PIP-based membrane with its narrow pore size range, resulting in a high rejection rate of MgCl₂ (~99%). The selectivity of conventional NF membranes in high-salinity solutions is notably reduced by Donnan effect attenuation due to counterion adsorption, whereas the heterogeneous structure of the SRM allows for the separation of cations by sieving through the appropriate pore size at the top and then processing based on the positively charged layer at the bottom, which is expected to result in accurate ion separation. In addition, SRM with bigger D_mp and FFV values has a lower monovalent salt rejection rate than NF-PEI and DMM, which facilitates the ion separation efficiency (fig. S27).

The morphology of the NF membrane, including surface shape and thickness, is also an important factor affecting separation performance, and these SRM structures have been adjusted favorably (Fig. 3G and fig. S28). As controls, NF-PIP and NF-PEI exhibited nodular morphologies with thicknesses of 80 to 90 nm and smooth surfaces with thicknesses of ~30 nm, which are similar to the levels reported in the literature and are in line with our experience (fig. S29) (34–36). In the case of DMM, the disordered competitive reaction of PIP and PEI with TMC disrupts the self-limiting effect of IP, and the monomer continues to diffuse and react to form 200-nm-thick and large nodular structures (figs. S30 and S31). In contrast, SRM is not affected by the disordered competitive reaction of the monomer, and the self-limiting effect of IP is enhanced (37). The thickness of the separation layer is ~144 nm. The thickness of the SRM is notably greater than that of NF-PIP and NF-PEI, which aligns with industrial preferences for robust membranes that prioritize long-term structural durability and stability over ultrathin designs (38). The molecular simulation results illustrate that the various substances in the aqueous solution are attracted to each other (fig. S32). Therefore, the appropriately different diffusion coefficients of aqueous phase monomers cause SRM to exhibit a Turing structure of interwoven nanowires, which is different from the nodular structure of other membranes. This rough morphology could provide more permeable spots and allow the membrane to be easily wetted, thus increasing the water permeability (fig. S33 and table S2) (39). In addition, atomic force microscope (AFM) and energy dispersive x-ray spectroscopy (EDS) elemental mapping characterization confirm that the distribution of ODNPs with monomers on the substrate is highly homogeneous in both surface morphology and elemental composition, with no adverse impact on the uniformity of the composite membrane, thereby supporting high performance (figs. S34 to S35).

Ion separation performance and permeability

The potential application of SRM in ion separation was investigated. As shown in Fig. 4A, the SRM can achieve high rejection rate of divalent Mg²⁺ and ultralow or negative rejection rate of monovalent Na⁺/Li⁺ for both Mg²⁺/Na⁺ and Mg²⁺/Li⁺ mixed solutions of 23:1. This excellent selectivity can be attributed to the optimized distribution of charge and stable and suitable pore size distribution. The narrowed pore size distribution also allows the NF membrane to exhibit excellent anion separation performance. SRM also demonstrated rejection rates of more than 98% for SO₄²⁻ and less than 40% for Cl⁻ against 1:1 SO₄²⁻/Cl⁻ solutions. The separation of Mg²⁺/Li⁺ and SO₄²⁻/Cl⁻ has been recognized as an important method for achieving resource management circularity, such as lithium extraction from saline lakes and zero-liquid discharge strategies for industrial wastewater (40). The separation performance of SRM fully reflects its excellent application potential. In addition, investigating the separation performance at different Mg/Li mass ratios to match some salt-lake brines worldwide is crucial. The individual ion rejection rates and corresponding separation factors of the composite membranes for mixed solutions with different Mg²⁺/Li⁺ ratios and low Li⁺ concentrations [stabilized at 100 parts per million (ppm)] were evaluated, and the results are shown in Fig. 4B and fig. S36. The separation factor gradually decreases as the Mg²⁺ concentration increases but remains high. As envisioned earlier, the separation factor of SRM can still reach ~50 even if the solution has an Mg/Li mass ratio of ~50 and a salt concentration of ~20,000 ppm. Under these harsh conditions, the rejection rate of Mg²⁺ was notably reduced in the other three membranes, and the separation factor of NRM was at least 9.6 times greater than that of the other three control membranes (Fig. 4C and fig. S37). In particular, compared to DMM, which performs even worse than NF-PIP and NF-PEI, the separation factor of SRM is more than 16 times higher. This phenomenon is due to the uncontrolled competitive reactions during the preparation of DMM, resulting in the composite membrane exhibiting compromised properties, where the molecular chain stability is lower than that of NF-PIP and the positive charge is weaker than that of NF-PEI. Both molecular chain stability and positive charge are critical under high-salt conditions. Therefore, the performance of DMM is inferior to that of NF-PIP and NF-PEI. Conventional NF membranes for magnesium-lithium separation require freshwater dilution of high-salinity feed solutions. Salt Lake areas, in turn, often lack freshwater resources. The SRM has the potential to extract lithium with stable performance at high salt concentrations and Mg/Li mass ratios, which can save up to 10 times the amount of fresh water compared with the conventional treatment of 2000 ppm of solutions. The separation factor of SRM can exceed 200 when the solution has an Mg/Li mass ratio of 50 and a salt concentration of 2000 ppm (fig. S38), which is also higher than the other three control membranes. In addition, for a fairer comparison, the effect of fabrication parameters on the performance of DMM is explored, and even under optimal conditions, the overall performance is still far inferior to that of SRM (figs. S39 and S40). This result again suggests the superiority of the SRM prepared by this strategy combining the advantages of the two monomers.

Fig. 4. Separation performance of NF membranes.

(A) Selectivity of SRM for ions in different salt mixtures. (B) Separation factors of SRM with different ratios in cationic mixed salt systems. (C) Mg²⁺/Li⁺ separation factors of four NF membranes for high saline brine at 20,000 ppm. (D and E) Comparison of Mg²⁺/Li⁺and SO₄²⁻/Cl⁻ selectivity for SRM with those reported in the literature for advanced laboratory-prepared membranes or commercial membranes. The literature for permeability comparisons is the same as that for selective comparisons, except that thickness data are not provided. (F) Long-term operational stability of SRM membranes in 72 hours.

The Mg²⁺/Na⁺ separation performance of SRM was further investigated under similar conditions (Fig. 4B and fig. S41). Its excellent separation effect further broadens the application fields of SRM. The selectivity of Mg²⁺/Na⁺ is greater than that of Mg²⁺/Li⁺, which is different from the results obtained by testing single salts (fig. S42). In the system of single salt tests, the difference in relative molecular weights causes the molar concentration of LiCl to be higher than that of NaCl, and a higher Cl⁻ concentration facilitates the passage of Li⁺ through the membrane during cross-flow filtration (41). Consequently, LiCl results in a lower rejection rate and a correspondingly higher separation factor than NaCl does. However, in the mixed salt system, there is no notable difference in the Cl⁻ concentration, so Na⁺ with a smaller hydration radius (table S3) is more likely to pass through the membrane, and the corresponding separation factor is greater. Compared with the reported ion separation NF membranes in the literature (Fig. 4, D and E, and tables S4 and S5), the SRM exhibits a prominent combination of high permeance and selectivity, which demonstrates the great potential of these membranes for practical applications.

The SRM exhibits a high permeance of 16.3 liter m⁻² hour⁻¹ bar⁻¹, which is about twofold greater than that of the DMM (7.9 liter m⁻² hour⁻¹ bar⁻¹) and surpasses all other controlled membranes notably (fig. S43). This can be largely attributed to the optimized structure of the SRM separation layer combining a large free volume and mass density uniformity in the upper layer and the inclusion of nanochannels coming from between the Schiff base polymer and the polyamide. In addition, the increased hydrophilicity of the surface and the increased effective filtration area also contribute positively to water molecule transport. To visualize this superiority of SRM, the permeability (defined as the product of permeance and thickness, which intuitively reflects the intrinsic transport efficiency of materials) is used as a comparison metric, and the SRM can show the extremely high performance (~2347 liter·m⁻²·hour⁻¹·bar⁻¹·nm), which surpasses state-of-the-art ion separation NF membranes (fig. S44 and table S6). This high permeability allows SRM to present excellent permeance even at a slightly higher thickness.

Undeniably, the introduction of nanomaterials also has a positive impact on the permeance improvement, but the effective combination of the two monomers is undeniably an important factor. Under similar conditions, the ultrathin nanoparticle-containing membranes based on individual monomer preparation are less permeable than SRM (fig. S45).

Operational stability is critical in practical applications, so the stability of the staged-regulated NF membranes was evaluated with 2000 ppm of MgCl₂. The performance of the SRM at a stable operation pressure (0.6 MPa for 72 hours) is shown in Fig. 4F. The permeance and rejection rate remain almost unchanged throughout the process, indicating its tremendous potential for long-term operation. Furthermore, the contamination resistance of the SRM membrane was evaluated via the use of dextran (a polysaccharide) as a contaminant (fig. S46). On the basis of the hydrophilic and homogeneous properties of the SRM, the SRM also exhibited a slow rate of flux decline and a notable rate of flux recovery (42).

DISCUSSION

In summary, we present a strategy for the staged control of the reaction behavior of dual aqueous phase monomers in IP. By modulating the reaction sequence and relative concentration of PIP and PEI monomers, we have prepared the SRM that combines the structural advantages of both monomers. The method uses the reaction by-product H⁺ as a trigger without the need for external stimulation. No complex equipment retrofits indicate its potentials to be directly integrated into legacy IP production lines. Experimental and molecular simulation results reveals that the SRM exhibits uniform mass density, high FFV, optimized pore size distribution, and a heterogeneous charge distribution characterized by a concentrated positive charge at the bottom. The well-tuned nanopores can selectively reject divalent cations at the initial separation stage, reducing ion concentration reaching the enriched positively charged bottom. This enables the separation layer bottom to still maintain strong Donnan repulsion in high-salinity application environments with minimized counter-ion adsorption, thereby achieving highly selective separation. Consequently, the SRM displays outstanding ion selectivity with Mg²⁺/Li⁺ separation factors of about 50 at 20,000 ppm and about 200 at 2000 ppm. Besides, the SRM also exhibits excellent permeability primarily due to the large free volume characteristics and structural homogeneity. These above results demonstrate that our proposed enamine reversible reaction for modulating the reaction behavior of two monomers is an effective and reliable way to address the challenge of competitive side reactions and enable precise structural tailoring of TFC membranes. The strategy will inspire the design of high-performance functional membranes that combine the structural advantages of distinct monomers into tailored nanostructures for addressing the most challenging separation problems.

MATERIALS AND METHODS

Fabrication of TFC NF membranes

The TFC NF membranes were prepared by IP as follows. The ultrafiltration substrate is cut into the appropriate size (15 cm by 15 cm) and fixed with a polymethylmethacrylate plate frame and rubber ring. The properties of the substrate are shown in fig. S47. Subsequently, 25 ml of the aqueous solution is poured onto the membrane surface and held for 2 min, followed by pouring and again using an air knife to remove the residual aqueous solution from the membrane surface. Last, 15 ml of the organic phase solution is poured onto the membrane surface to initiate the IP reaction for 30 s. After the reaction is completed, the membrane surface is washed with hexane for 15 s and treated in an oven at 60°C for 120 s. The prepared NF membranes were stored in deionized water before testing. For the SRM, the configuration of the aqueous solution is critical. The ODNPs must first fully react with the PIP monomer, and the PEI monomer can only be added after the color of the enamine solution is stable and no longer changes (fig. S48).

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S50

Tables S1 to S7

References