Introduction

The global issue of water scarcity has become an increasing concern for the United Nations. In response, the United Nations established Sustainable Development Goal 6 in 2015, which focuses on water-related objectives, including Clean Water and Sanitation. One of its key targets is to ensure universal and equitable access to safe and affordable drinking water for all (Target 6.1)1,2. This goal has driven the development of sustainable technologies to produce clean and safe drinking water from unconventional sources, such as wastewater, seawater, and brackish water3,4. Currently, thin-film composite (TFC) polyamide (PA) reverse osmosis (RO) membranes, known for their high water-salt selectivity, have emerged as a leading technology for freshwater production5,6,7. However, these membranes still fall short in removing small (molecular weight (MW) ≤ 150 Da) neutral organic contaminants (SNOCs), which are both highly toxic and prevalent in water and wastewater8.

SNOCs, such as N-nitrosomethylethylamine (NMEA, 88.1 Da, neutral at pH = 7)9, are still capable of traversing commercial RO membranes, leading to unsatisfactory rejection rates (30 to 82%)10,11,12,13. The PA selective layers of RO membranes are critical in controlling solute and water transport, and thus play a key role in determining membrane separation performance14,15. Given that size exclusion is the predominant mechanism for SNOC removal16,17,18, achieving high retention of SNOCs is possible when the PA layers of RO membranes are free of nanosized defects and possess a sufficiently high density as indicated by their molecular weight cut-off (MWCO) or crosslinking degree. However, as the PA layers of RO membranes are typically fabricated through interfacial polymerization (IP) reactions between an amine monomer and an acyl chloride monomer19, the rapid nature of the IP reaction often outpaces the supply of amine monomers, resulting in a PA layer with structural heterogeneity and nanosized defects20,21,22,23. Therefore, an adequate supply of amine monomers and a controllable IP reaction are key to achieving the RO membrane for efficient SNOC removal.

We propose that creating a continuous solid-phase interface via self-assembly of nanomaterials can possibly address the above challenge. Driven by weak noncovalent interactions24,25,26, establishment of a solid-phase interface may be achieved by introducing nanomaterials capable of self-assembling, whereby they spontaneously integrate themselves from a disordered state into well-organized macroscale aggregates at the water/oil interface. Such nanosized solid-phase at the water/n-hexane interface can mediate the shuttling of amine monomers to achieve their pre-enhancement in the organic phase, ensuring the adequate supply of amine monomers. This strategy is expected to control the IP reaction by enhancing the initial reaction kinetics and leveraging the self-limiting nature and nanobubble modulation of the IP process. Hence, an ideal dense PA layer may be obtained to overcome the trade-off between water permeance and SNOC rejection.

In this study, we developed a RO membrane mediated by self-assembly of CdII/L-cysteine nanowires at the water/n-hexane interface, aiming to regulate the IP process. The resultant RO membrane exhibited high rejection of SNOCs (up to 97.9%), featuring excellent water permeance (3.6 ± 0.1 L m−2 h−1 bar−1), overcoming permeance-selectivity trade-off. We unraveled the amphiphilicity and interfacial self-assembly behavior of CdII/L-cysteine nanowires via first-principles calculations and molecular dynamics simulations. Moreover, we systematically investigated the impact of the self-assembly CdII/L-cysteine interface on the morphologies and separation performance of the as-prepared RO membrane. Our findings provide a promising approach for manipulating monomer trans-interface shuttling behavior and customizing the outstanding performance of RO membranes.

Results

Nanowire amphiphilicity as the key for self-assembly

The amphiphilic feature, which refers to nanomaterials with both hydrophobic and hydrophilic domains, is a prerequisite for the self-assembly of nanomaterials at the water/oil interface27. Therefore, first-principles calculations were performed to ascertain the amphiphilicity of CdII/L-cysteine nanowires (Supplementary Fig 2). The interaction energies between CdII/L-cysteine nanowires and water/n-hexane molecules were analyzed at a steady state. As depicted in Fig. 1a, the interactions among molecules were implemented by hydrogen bonds. Furthermore, all negative interaction energies (−0.24 eV for R − S*/*H − OH, −5.78 eV for R − S*/*H−hex, −0.50 eV for R − COOH*/*OH2, and −5.75 eV for R − CO*OH/*H−hex, respectively) revealed that these nanowires were favorable for interacting with both water and n-hexane molecules. Specifically, the S atoms and carboxyl groups (−COOH) of the nanowires could generate hydrogen bonds with both water and n-hexane molecules, demonstrating their amphiphilic nature. Notably, the H atom of the −OH in −COOH developed a hydrogen bond with the O atom of the water molecule (R − COOH*/*OH2), while the O atom in the C = O of −COOH rendered a hydrogen bond with the H atom of the n-hexane molecule (R − CO*OH/*H-hex). Although the CdII/L-cysteine nanowires are also abundant in −NH2 groups, first-principles calculations confirmed that these −NH2 groups are exclusively hydrophilic and not amphiphilic (Supplementary Fig. 3).

Fig. 1: Amphiphilicity ascertainment of CdII/L-cysteine nanowires.
figure 1

a The interaction energies between CdII/L-cysteine nanowires and water/n-hexane molecules, involving four conditions (Color code: Cd, pink; S, yellow; O, violet; N, blue; C, green; H, gray): S atom of nanowire interacts with H atom of water molecule (R − S*/*H − OH); S atom of nanowire interacts with H atom of n-hexane molecule (R − S*/*H−hex); H atom of the −OH in −COOH interacts with O atom of water molecule (R − COOH*/*OH2); O atom in the C = O of −COOH interacts with H atom of n-hexane molecule (R − CO*OH/*H−hex). b–e The partial density of states (PDOS) was used to illustrate the bonding mechanism between the involved atoms. Insets are the charge density difference (yellow and green domains stand for electron accumulation and depletion, respectively). f Negative crystal orbital Hamilton population (−COHP) was employed to elucidate the stability of hydrogen bonds, where −COHP > 0 represents the bonding state, while −COHP < 0 denotes the antibonding one. Note: ICOHP refers to the integral of COHP. g Schematic of amphiphilic CdII/L-cysteine nanowires, originating from amphiphilic S atoms and −COOH groups on them. Source data are provided as a Source Data file.

Full size image

To further validate the bonding between the involved atoms from the water, n-hexane, and nanowires, we performed a charge density difference analysis. The finding unveiled that electron transfer occurred when nanowires encountered water/n-hexane molecules, verifying the presence of hydrogen bonds (Fig. 1b–e, inset)28. The bonding mechanism of the nanowires with water/n-hexane molecules was further elucidated by the partial density of states (PDOS) analysis. The s-orbitals of H atoms in the water/n-hexane molecules split into several peaks that overlapped with the p-orbitals of the S atoms in the nanowire molecules at (−8.47, −5.17, and −4.75 eV)/( − 7.17, −6.81, −5.70 and −4.01 eV) (Fig. 1b–c), resulting in a hydrogen bond between H and S atoms29. Additionally, the same behavior was also observed between the s-orbital of the H atom of the −OH in −COOH and the p-orbital of the O atom in the water molecule (Fig. 1d), as well as between the s-orbital of the H atom in n-hexane molecule and the p-orbital of the O atom of C = O in −COOH (Fig. 1e).

We also executed the crystal orbital Hamilton population (COHP) analysis to intuitively reflect the bonding and antibonding states between the involved atoms. Chemically speaking, a negative COHP value denotes bonding states, while a positive value represents anti-bonding states30. To better illustrate the COHP results, the integral of COHP (ICOHP, net bonding contribution) was evaluated to reflect the stability of the hydrogen bonds between the involved atoms31,32. Positive −ICOHP (namely, negative ICOHP) was observed for four conditions (0.25 V for R − S*/*H − OH, 0.05 V for R − S*/*H−hex, 1.13 V for R − COOH*/*OH2, and 0.03 V for R − CO*OH/*H−hex, respectively), further implying that the S atom and −COOH of CdII/L-cysteine nanowire could generate stable hydrogen bonds with both water and n-hexane molecules (Fig. 1f)33. Given that amphiphilic S atoms exist in the backbone of the CdII/L-cysteine nanowires and amphiphilic −COOH groups are free groups on them34, highly symmetrical CdII/L-cysteine nanowires exhibited remarkable amphiphilicity, potentially enabling self-assembly at the water/n-hexane interface (Fig. 1g). Moreover, the self-assembled CdII/L-cysteine solid-phase interface located at the water/n-hexane interface may prove advantageous for amine monomer adsorption and further monomer shuttling & pre-enrichment.

Interfacial Self-assembly of CdII/L-cysteine nanowires

The successful synthesis of the CdII/L-cysteine nanowires with a high aspect ratio was verified by SEM visualization (Supplementary Fig. 2). A strategy for designing a RO membrane requires CdII/L-cysteine nanowires with self-assembly capability (Fig. 2a). Consequently, molecular dynamics (MD) simulations were implemented to unveil the self-assembly behavior of CdII/L-cysteine nanowires at the water/n-hexane interface. The initially disordered CdII/L-cysteine nanowires in the n-hexane phase rapidly accumulated to the water/n-hexane interface and eventually self-assembled into a continuous interfacial layer, confirming successful self-assembly (Fig. 2b, c). The self-assembly behavior was completed after 30 ns, as evidenced by the coverage rate (CR) of nanowires at the interface increasing with time (from 65.7% at 0 ns to 100% at 30 ns) (Fig. 2d–h). As the self-assembly occurs, an increase in the number of hydrogen bonds between nanowire molecules and a concurrent decrease in those between nanowire and n-hexane molecules highlighted the fact that hydrogen bonding serves as the primary driving force for the self-assembly (Fig. 2i and Supplementary Fig. 4). Moreover, the self-assembly behavior of CdII/L-cysteine nanowires was also corroborated by SEM analysis after depositing the self-assembled resultant on smooth silicon wafers (Supplementary Fig. 5). The distribution of the CdII/L-cysteine nanowires at the water/n-hexane interface evolved from being discrete to generating a continuous film, which agreed with the findings obtained from MD simulations.

Fig. 2: Schematic preparation of the RO membrane and MD simulations to unveil self-assembly behavior of CdII/L-cysteine nanowires.
figure 2

a Preparation procedure of the RO membrane with self-assembled CdII/L-cysteine interface. MD simulations of self-assembly of CdII/L-cysteine nanowires at the water/n-hexane interface: b initial MD models (nanowires immersed randomly in the n-hexane) and c ultimate MD models. Top-view snapshot of the self-assembly process taken at every 10 ns for a total duration of 30 ns: d 0 ns; e 10 ns; f 20 ns; and g 30 ns. h The coverage rate (CR) of nanowires at the water/n-hexane interface over self-assembly time (MD simulation process). i Changes in the number of hydrogen bonds between CdII/L-cysteine nanowires as a function of self-assembly time (red curve is the fitting, fit model: ExpDec2, R2: 90.27%). Error bars in h represent the standard deviations (s.d.) (n = 3) and data are presented as mean values ± s.d. Source data are provided as a Source Data file.

Full size image

Morphology of nanowire-mediated RO membrane

RO membranes were constructed to explore the effect of CdII/L-cysteine nanowires with self-assembly features on membrane physical and chemical properties. The XPS spectra detected the existence of nitrogen elements on the membrane surfaces, indicating the formation of the PA selective layer (Supplementary Fig. 6). Additionally, the surface and cross-sectional morphologies of the RO membranes were visualized (Fig. 3 and Supplementary Fig 7–8). Notably, the surface morphologies of the free-standing RO membranes induced by self-assembled interfaces with different CdII/L-cysteine nanowire concentrations displayed that the RO-0 membrane (control), prepared without a self-assembled CdII/L-cysteine interface, had a relatively smooth surface with only a few discrete nodules (without a “ridge-and-valley” structure, Fig. 3a and Supplementary Fig 7a). Conversely, the “ridge-and-valley” appearance generated in the membrane (RO-0.025, RO-0.05, and RO-0.1) surfaces after introducing self-assembled CdII/L-cysteine interfaces (Fig. 3b-d and Supplementary Fig 7b–d). Close examination of these membrane surfaces revealed that the “ridge-and-valley” structure of the RO-0.025 and RO-0.05 membranes predominantly involved larger elongated protuberances (leaf-like features), whereas that of the RO-0.1 membrane engaged the coexistence of nodular and smaller leaf-like features (Fig. 3b–d). Meanwhile, the presence of the self-assembled CdII/L-cysteine interfaces within the RO membranes (namely RO-0.025, RO-0.05, and RO-0.1) was verified by SEM characterization conducted on their back sides, except for the RO-0 membrane (Fig. 3e–h). Furthermore, quantitative analysis of Cd content revealed an initial increase followed by stabilization (0%, 0.9%, 2.1%, and 2.2% for the RO-0, RO-0.025, RO-0.05, and RO-0.1 membranes, respectively), indicating that a continuous self-assembled CdII/L-cysteine interface was present in the RO-0.05 and RO-0.1 membranes, which was consistent with the SEM results (Fig. 3e-h).

Fig. 3: Characterization of the free-standing RO membranes induced by self-assembled interfaces with different CdII/L-cysteine nanowire concentrations.
figure 3

a–d SEM micrographs of front surfaces; e–h SEM characterization on back surfaces (the Cd contents are labeled on the corresponding SEM images); i–l AFM images and surface roughness data (Rq, scale bar: −600 ~400 nm).

Full size image

Based on the previous study23,35,36,37, the construction of the “ridge-and-valley” structure of the membranes (RO-0.025, RO-0.05, and RO-0.1) can be attributed to the confinement effect of self-assembled CdII/L-cysteine interfaces against heat dissipation and nanobubble escape across the water/n-hexane interface. Moreover, with increasing the concentration of CdII/L-cysteine nanowires, membrane surface roughness (Rq) gradually increased (from 22.1 nm to 132.0 nm, Fig. 3i–k) and subsequently declined to 66.2 nm (Fig. 3l), which were consistent with SEM results. Notably, the roughness of the RO-0.1 membrane was reduced, although a continuous interface was developed (Fig. 3l), which probably originates from a heightened interaction between CdII/L-cysteine nanowires and m-phenylenediamine (MPD) monomer as a result of the higher nanowire concentration. The incremental interaction curtailed MPD diffusion into the n-hexane phase, consequently diminishing the generation of nanobubbles38.

Performance of the RO membrane

The performance of the as-prepared RO membrane was investigated in terms of water permeance, NaCl rejection, membrane stability, and SNOC rejection. As a control, the RO-0 membrane without a self-assembled CdII/L-cysteine interface exhibited a relatively inferior water permeance of 0.8 ± 0.2 L m−2 h−1 bar−1 (Fig. 4a). In contrast, a substantial improvement in water permeance was observed for the RO-0.025 (2.0 ± 0.3 L m−2 h−1 bar−1), RO-0.05 (3.6 ± 0.1 L m−2 h−1 bar−1), and RO-0.1 (2.7 ± 0.1 L m−2 h−1 bar−1), after introducing self-assembled interfaces with various CdII/L-cysteine nanowire concentrations (Fig. 4a). Specifically, the RO-0.05 membrane displayed the optimum water permeance, outperforming the RO-0 membrane by a factor of 4.5. Among the four membranes, the RO-0.05 membrane with maximum water permeance also exhibited the best retention for NaCl (94.4 ± 1.4% for RO-0, 98.4 ± 0.3% for RO-0.025, 98.9 ± 0.2% for RO-0.05, and 89.1 ± 1.7% for RO-0.1, respectively), deriving from its highest crosslinking degree (95.2%)/lowest MWCO (82.6 Da) (Fig. 4b, Supplementary Fig 9–10, and Supplementary Table 1). Additionally, the RO-0.05 membrane outperformed both laboratory-manufactured and commercial RO membranes by delivering either the optimal water/NaCl selectivity for a given water permeance or the greatest water permeance for a specific water/NaCl selectivity (Fig. 4c). The above findings revealed that the incorporation of self-assembled interfaces of CdII/L-cysteine nanowires can break the trade-off effect between water permeance and selectivity, thereby endowing the membranes with prominent water reuse potential.

Fig. 4: Permeance of the RO membranes induced by self-assembled interfaces with various CdII33/L-cysteine nanowire concentrations.
figure 4

a Pure water permeance and NaCl rejection; b crosslinking degree; c Performance (water permeance and water/NaCl selectivity) comparison of RO-0.05 membrane with literature data (detailed description in Supplementary Table 2); d effect of operation time on water flux and NaCl rejection of the optimal RO-0.05 membrane to evaluate membrane stability; e the removal of RO-0 (control membrane) and RO-0.05 (optimal membrane) for contaminants (including four chlorinated SNOCs and five N-nitrosamines); f Summary of water permeance and water/SNOC selectivity of the RO-0.05 membrane and other RO membranes described in the literature (detailed description in Supplementary Table 3). Error bars represent the s.d. (n = 3). Source data are provided as a Source Data file.

Full size image

To further validate the prospects of the RO-0.05 membrane (optimum performance) for water reuse, we determined its stability under long-term operation, which has been highlighted as one of the major obstacles to its practical applications. Water flux and NaCl rejection were employed as the evaluation parameters to analyze membrane stability. As evidenced by Fig. 4d, the RO-0.05 membrane demonstrated extraordinary stability with neither water flux nor NaCl rejection showing any significant fluctuations over a 48-h filtration period (notably, the membrane flux diminished marginally in the beginning 3 h caused by membrane compaction). Furthermore, the exceptional stability of the RO membrane effectively prevents secondary contamination of the permeate, thereby guaranteeing water safety (Supplementary Fig 11). The rejection measurements of RO-0 and RO-0.05 membranes for the SNOCs (including 1,2-DCE, 1,3-DCP, 1,2-DCP, CF, NDMA, NMEA, NPYR, NDEA, and NPIP) were also implemented. As shown in Fig. 4e, the RO-0.05 membrane presented a remarkable improvement in removal efficiency against all SNOCs in comparison to the RO-0 membrane, which is predominantly attributed to the increment in crosslinking degree/reduction in MWCO. Moreover, the RO-0.05 membrane exceeded membranes reported in the literature for SNOCs (uncharged, MW < 150 Da) rejection, featuring the maximum water/SNOC selectivity (Fig. 4f). Consequently, decently outstanding permeance, excellent solute rejection, and exceptional stability endow the RO-0.05 membrane with great potential for water reuse.

Underlying mechanisms for enhanced RO performance

The enlarged surface area ratio (SAR), resulting from the crumpled surface structures, could contribute to a more effective area for water transport, thereby being responsible for the remarkable water permeance enhancement of RO membrane with self-assembled interfaces of CdII/L-cysteine nanowires (Fig. 5a)23,39. However, the introduction of the self-assembled interface resulted in a remarkable increase in the apparent thickness (~226 nm for the RO-0.05 vs ~23 nm for the RO-0) with negligible change in the intrinsic thickness of the RO membrane (~23 nm for both membranes), demonstrating that membrane thickness variations exert no beneficial effect on water permeance enhancement (Supplementary Fig 12). To further explore the reason for substantial improvement in water permeance of the as-prepared RO membranes, three RO membranes were constructed by implementing conventional IP reaction on the PES, PES-SAN (a PES membrane loaded with self-assembled nanowires), and PES-NSAN (a PES membrane loaded with nanowires but not undergoing self-assembly process) substrates, respectively. Subsequently, we investigated their water permeance and NaCl rejection (Fig. 5b). Among the three RO membranes, the PES-SAN-RO exhibited the maximal water permeance (0.9 ± 0.1 L m−2 h−1 bar−1 for the PES-RO, 2.9 ± 0.2 L m−2 h−1 bar−1 for the PES-SAN-RO, and 1.4 ± 0.3 L m−2 h−1 bar−1 for the PES-NSAN-RO, respectively). Moreover, they also presented comparable SAR (1.30 ± 0.04 for the PES-RO, 1.40 ± 0.09 for the PES-SNA-RO, and 1.38 ± 0.06 for the PES-NSAN-RO, respectively), providing virtually identical water filtration areas (Supplementary Fig 13). These results implied that the incorporation of the continuous interface, constructed by the self-assembly of CdII/L-cysteine nanowires, can eliminate the inherent funnel effect of RO membranes or introduce the gutter effect, resulting in the substantial enhancement in water permeance of the RO membrane40,41. Furthermore, the water permeance of the RO-0.05 (3.6 ± 0.1 L m−2 h−1 bar−1) remarkably outperformed that of the RO-0 (0.8 ± 0.2 L m−2 h−1 bar−1), primarily originating from the synergetic effect of the gutter effect induced by self-assembled CdII/L-cysteine interface and the increased SAR. To evaluate the contribution of the increased SAR and gutter effect induced by the self-assembled CdII/L-cysteine interface to the enhanced water permeance of the RO-0.05 membrane, a systematic comparison of water permeance was conducted among RO-0, RO-0.05, PES-NSAN-RO, and PES-SAN-RO membranes. Quantitative analysis revealed that the total water permeance enhancement factor (ftotal) of the RO-0.05 membrane was 4.5, with the gutter effect contributing an enhancement factor (fgutter) of 2.1 and the increased SAR contributing an additional enhancement factor (fSAR) of 2.4 (Supplementary Fig 14).

Fig. 5: In-depth investigation into the mechanism responsible for optimizing membrane separation performance.
figure 5

a The surface area ratio (SAR) of the RO membranes was analyzed by AFM. b Water permeance and NaCl rejection of three conventional RO membranes: PES-RO fabricated on the PES substrate, PES-SAN-RO prepared on the PES substrate loaded with self-assembled nanowires, and PES-NSAN-RO constructed on the PES substrate loaded with nanowires but not performing the self-assembly process, respectively. c Water permeance and NaCl rejection of the RO-0.05 and NMPD (no MPD pre-enrichment) membranes developed via free-standing IP reaction. d–g Schematic illustration of nanowire-assisted MPD shuttling and MPD pre-enrichment for the construction of a dense PA layer. Error bars represent the s.d. (n = 3). Source data are provided as a Source Data file.

Full size image

Researchers have verified that improving the supply of amine monomers in the organic phase (the reaction zone) can benefit the IP reaction, thus leading to a denser PA selective layer characterized by elevated solute rejection37,42,43. The continuous monitoring of the MPD diffusion behavior revealed a notable increase in MPD concentration in the n-hexane phase when the self-assembled CdII/L-cysteine interface was present at the water/n-hexane interface (Supplementary Fig 15). Seeking to clarify this mechanism, we explored the functional groups and the zeta potential of CdII/L-cysteine nanowires at pH = 7 (water conditions) and pH = 9 (MPD/water conditions). Strong hydrogen bonds between CdII/L-cysteine nanowires (−NH2 and −COOH groups) and MPD molecules (−NH2), together with the coordination interaction between Cd centers of CdII/L-cysteine nanowires and −NH2 in MPD molecules, can facilitate the adsorption of MPD to the self-assembled CdII/L-cysteine interface (Supplementary Fig 16a). Additionally, the electrostatic attraction between the negatively charged CdII/L-cysteine nanowires and the positively charged MPD molecules can further enhance the efficient attachment of MPD to the interface (Supplementary Fig 16b). Given that the self-assembled CdII/L-cysteine interface is located at the water/n-hexane interface, with portions in both the aqueous and n-hexane phases, we conjectured that the self-assembled CdII/L-cysteine interface where MPD molecules are adsorbed will mediate MPD shuttling and subsequently achieve MPD pre-enrichment in the organic phase, which is the predominant contributor to the increased membrane selectivity.

To validate this deduction, we prepared the RO-0.05 and NMPD (no MPD pre-enrichment: n-hexane containing CdII/L-cysteine nanowires was pre-poured onto the surface of the pure water) membranes. Subsequently, the water permeance and NaCl rejection of the two membranes were examined. As expected, the result illustrated that compared with the RO-0.05 membrane (98.9 ± 0.2%), a dramatic decrease in NaCl rejection rate of the NMPD membrane (11.2 ± 2.2%) accompanied by a relatively smooth PA layer and a water permeance of 21.3 ± 3.2 L m−2 h−1 bar−1 (Fig. 5c and Supplementary Fig 17). This result renders compelling evidence that the pre-enrichment of MPD, which can ensure sufficient amine monomers for IP reaction, is the principal reason for strengthening the membrane selectivity. The relevant mechanisms are schematized in Fig. 5d-g.

Discussion

The current study elucidated that the amphiphilicity of CdII/L-cysteine nanowires facilitates their spontaneous self-assembly at the water/n-hexane interface, impelled by hydrogen bonds. The MPD pre-enrichment in the organic phase, achieved through MPD shuttling mediated by a self-assembled CdII/L-cysteine interface, ensured a sufficient supply of MPD monomers. This accelerated IP reaction kinetics, and thereby emerged as the primary contributor to the improved elimination efficiency of small and neutral organic contaminants (MW < 150 Da). The performance test of three conventional RO membranes revealed that the remarkably intensified membrane water permeance originated from the gutter effect induced by the self-assembled CdII/L-cysteine interface, coupled with the increased SAR derived from the construction of a crumpled structure that was attributed to the confinement effect of the CdII/L-cysteine interface against heat dissipation and nanobubble escape. Consequently, the simultaneous enhancement in membrane permeance and selectivity endows the self-assembly adjusted IP strategy with considerable potential for future membrane design and optimization. Moreover, the facile synthesis protocol of the CdII/L-cysteine nanowires under mild conditions, coupled with their container shape/dimension-independent self-assembly capability, provides significant potential for the scalable production and application of these RO membranes. However, further optimization of the transfer protocol (freestanding PA layers depositing to substrates) remains necessary.

Methods

General

Materials and chemicals used in this work have been shown in the Supplementary Methods.

Synthesis of nanowires

CdII/L-cysteine nanowires were synthesized in accordance with the following protocol34: specifically, L-cysteine (0.025 mol L−1) and Cd(ClO4)2·6H2O (0.01 mol L−1) were dissolved in 15 mL deionized (DI) water by vigorously stirring. The as-prepared reaction solution (pH: ~8.0, adjusted by NaOH) was then transferred immediately into an oven maintained at a constant temperature of 37 °C. Finally, the white powders were collected after a 120-h reaction, centrifugation, DI water washing, and vacuum-drying at 60 °C. Chemicals were provided in Supportting inforfamition.

Preparation of nanowire-regulated free-standing RO membranes

The free-standing RO membranes regulated by CdII/L-cysteine nanowires were constructed at a water/oil interface without a substrate. Specifically, a 110 mL MPD aqueous solution (2.0 wt%) was poured into a customized container and then allowed to stand until a stable liquid surface was achieved. Subsequently, a 2 mL n-hexane suspension containing various contents of CdII/L-cysteine nanowires (0.025, 0.05, and 0.1 wt%, respectively) was carefully spread onto the surface of the MPD solution to construct a continuous nanowire interface after self-assembling for 8 min (optimal self-assembly duration, Supplementary Fig 18). Afterward, the IP reaction was triggered by adding a 0.1 wt% TMC/n-hexane solution (5 mL) and conducted for 1 min. Finally, the nanowire interface and fabricated PA layer were transferred to the PES membrane surface. The as-prepared membranes were labeled as RO-0.025, RO-0.05, and RO-0.1, respectively, which was consistent with the nanowire content. The RO membrane prepared using the procedure described above, except for the addition of nanowires, was denoted as RO-0.

Preparation of no MPD pre-enrichment TFC membrane

A membrane with no MPD pre-enrichment in the organic phase was fabricated using the following procedures. Briefly, 100 mL DI water was poured into a customized container, and then a 2 mL n-hexane dispersion containing 0.05 wt% CdII/L-cysteine nanowires was carefully spread onto the liquid surface to construct a continuous nanowire interface after self-assembling for 8 min. Subsequently, a 10 mL MPD solution with a high concentration was injected into the DI water to generate a 2.0 wt% MPD solution, followed by reacting with a 0.1 wt% TMC/n-hexane solution for 1 min. Finally, the PA selective layer was transferred onto the PES substrate and thoroughly rinsed with n-hexane. The resultant membrane was named as NMPD.

Preparation of conventional TFC RO membrane

The conventional TFC RO membranes were fabricated by executing the IP reaction directly on three substrates, including PES, PES-SAN, and PES-NSAN. The PES-SAN substrate was fabricated by loading the self-assembled nanowires (2 mL of 0.05 wt% nanowires/n-hexane dispersion was implemented for an 8-min self-assembly process) onto the PES substrate using vacuum filtration, whereas the PES-NSAN substrate referred to a PES substrate loaded with nanowires but not undergoing self-assembly process (2 mL of 0.05 wt% nanowires/n-hexane dispersion was rapidly filtered onto the PES substrate). Subsequently, the three substrates were exposed to a 2.0 wt% MPD solution for 8 min before removing the excess MPD solution. Then, the MPD-saturated substrates were immersed in a 0.1 wt% TMC/n-hexane solution for l min. Finally, the resulting membranes were gently rinsed with n-hexane and labeled as PES-RO, PES-SAN-RO, and PES-NSAN-RO, respectively.

Molecular dynamics simulations

The self-assembly behavior of CdII/L-cysteine nanowires at the water/oil interface was visualized by MD simulations. The computational details were exhibited as follows. The molecular structures of water, n-hexane, and CdII/L-cysteine were constructed by GaussView before optimizing them using Gaussian (Supplementary Fig 19). The system, which comprises two layers (a water layer with 9997 water molecules as well as an oil layer with 200 CdII/L-cysteine molecules and 1657 n-hexane molecules), was placed in a box of dimensions 9.4 × 9.4 × 8.4 nm3 after energy minimization. The whole MD process was conducted at a constant temperature of 298 K and a pressure of 1 bar. Subsequently, 100 ns explicit MD simulations were implemented for this system using the GROMACS software package (version 5.0) and trajectories were sampled every 10 ns. Analysis and visualization of MD trajectories were facilitated using Visual molecular dynamics (VMD) software (version 1.9.2).

First-principles calculations

To investigate the hydrophilic and oleophilic nature of the CdII/L-cysteine nanowires, the first-principles calculations based on the plane-wave basis set approach, including the interaction energy, charge density difference, PDOS, and COHP, were performed using the Vienna ab initio simulation package (VASP, version 5.4.4)30,44,45. The projector augmented wave (PAW) method was deployed to depict the ion–core electron interactions. The valence electrons were represented with a plane wave basis set featuring an energy cutoff of 450 eV. Electronic exchange and correlation were described with the Perdew–Burke–Ernzerhof (PBE) functional. The DFT-D3 protocol was employed to consider the van der Waals interaction. The convergence criteria for the self-consistent electronic structure and geometry were established at 10−5 eV and 0.05 eV/Å, respectively.

Separation performance examination

A lab-scale cross-flow filtration apparatus equipped with three stainless filtration units (filtration area: 6.3 cm2) was employed to determine the RO membrane performance. After pressurizing at 16 bar to achieve a constant flux, each membrane coupon was tested for pure water permeance and selectivity towards NaCl (2000 ppm), neutral molecular probes (40 ppm), and SNOCs (200 µg L−1) at 25.0 ± 0.5 °C. The physicochemical properties of the SNOCs and neutral molecular probes were summarized in Supporting Fig. 1. The concentrations of NaCl and neutral molecular probes in both the feed and permeate solutions were quantified by a conductivity meter (DDSJ-308F, INESA instrument) and a total organic carbon analyzer (TOC-L CPH, Shimadzu, Japan), respectively. Additionally, the concentrations of four chlorinated SNOCs were examined employing a gas chromatograph-tandem mass spectrometer (GC-MS/MS, Agilent Technologies), whereas the concentrations of five N-nitrosamines were analyzed using high-performance liquid chromatography-tandem quadrupole mass spectrometry (UPLC-MS/MS, Thermo TSQ Quantum). The evaluation of water permeance (A, L m−2 h−1 bar−1) solute rejection rate (R, %), and solvent/solute selectivity (A/B, bar−1) was performed based on the subsequent equation46,47,48,49:

$$A=\frac{\varDelta V}{{A}_{{{{\rm{m}}}}}\times \varDelta t\times \varDelta P}$$
(1)
$$R=\frac{{C}_{{{{\rm{f}}}}}-{C}_{{{{\rm{P}}}}}}{{C}_{{{{\rm{f}}}}}}\times 100\%$$
(2)
$$\frac{A}{B}=\frac{R}{(1-R)\times \varDelta P}$$
(3)

where ΔV (L) is the cumulative permeate volume that has been collected after passing through the effective filtration area Am (m2) over the permeation duration of Δt (h) when being subjected to an applied trans-membrane pressure (∆P, bar). Cf and Cp are the solute concentrations in the feed and permeate solutions, respectively. B (L m−2 h−1) is the solute permeance.