Skip to main content
NCBI home page
Nature Communications logoLink to Nature Communications
. 2026 Feb 3;17:2308. doi: 10.1038/s41467-026-69044-5

Overcoming the trade-off in reverse osmosis membranes through homologous matching

Xinyu Shao 1,#, Shiyu Lv 2,#, Xiang Qin 1, Fangchao Cheng 1,3,✉, Dongying Hu 1,✉, Yiqiang Wu 3, Xiaofei Xu 4, Shuangliang Zhao 2,✉
PMCID: PMC12976049  PMID: 41634001

Abstract

The development of cellulose triacetate (CTA)-based reverse osmosis membranes offers a sustainable approach to alleviating the global freshwater crisis, yet overcoming the inherent permeability–selectivity trade-off remains a significant challenge. Herein, we propose a homologous matching strategy to address the trade-off by incorporating carbon dots (M-CDs) derived from m-phenylenediamine (MPD) into the interfacial polymerization process between CTA and polyamide (PA). Systematic characterization and molecular dynamics simulations reveal that M-CDs, which are structurally analogous to the MPD monomer, promote monomer diffusion, regulate cross-linking density, and refine the microstructure of the PA layer. At an optimal M-CDs concentration, the resulting membrane achieves simultaneous enhancements in both salt rejection (99.1% vs. 96.5%) and water flux (18.3 vs. 15.2 L·m-2·h-1), thus surpassing conventional CTA membranes. The incorporation of M-CDs results in a thinner, denser, and more hydrophilic barrier layer with reduced pore size and narrowed distribution, as confirmed by post-annealing structural analysis. Moreover, hydrogen bonding between M-CDs and MPD improves chlorine resistance, maintaining high performance even after exposure to 2000 ppm NaClO solution. Molecular dynamics further illustrate that M-CDs promote water cluster transport while hindering ion penetration, thereby effectively mitigating the trade-off. The innovative use of homologous carbon dots to optimize the CTA–PA interface through structural matching, offering inspiring avenues for developing advanced bio-derived desalination technologies.

Subject terms: Polymers, Pollution remediation, Polymer synthesis

Cellulose triacetate-based reverse osmosis membranes are promising for water purification; overcoming the permeability-selectivity trade-off is challenging. Here, the authors incorporate carbon dots into the membrane to optimize performance.

Introduction

The growing global population and escalating water pollution have exacerbated the imbalance between clean water supply and demand, posing a significant barrier to sustainable development. Given that 97% of the Earth’s water is saline and unfit for direct use, purifying seawater and brackish water has become essential to alleviate freshwater scarcity1–4. Among various purification approaches, membrane-based desalination has gained prominence and currently contributes to approximately 53% of the world’s drinking water production5,6. Compared to alternatives like nanofiltration, forward osmosis, and electrodialysis, pressure-driven membrane processes, such as reverse osmosis (RO), are particularly favored due to their high selectivity, effective purification performance, and operational simplicity7–9. Cellulose triacetate (CTA) has emerged as a promising material for RO membranes, owing to its abundance, biodegradability, and environmental compatibility. However, CTA-based RO membranes still face critical limitations, including the inherent trade-off between salt rejection and water permeability, poor chlorine resistance, and susceptibility to compaction.

Membrane performance is largely governed by structural characteristics such as pore uniformity, thickness, and surface properties10. Recent advances in nanotechnology offer pathways to overcome these limitations by incorporating functional nanomaterials to optimize membrane architecture and enhance performance11–13. Through careful design, the pore structure, layer thickness, hydrophilicity, and surface charge of nanomaterials can be precisely tuned, thereby improving selectivity and antifouling properties14,15. Although several materials, such as MXenes16, metal–organic frameworks17,18, graphitic carbon nitride19, mesoporous hollow nanospheres20, and nanomaterial sublayer21, have demonstrated the potential in enhancing permeate flux, the scalable and defect-free integration of these nanomaterials remains a challenge. In many cases, issues such as nanomaterial aggregation and interfacial defects during membrane fabrication can compromise the salt rejection, preventing the simultaneous achievement of high flux and high rejection that has been reported in ideal, lab-scale demonstrations. Additionally, nanomaterials inherently face challenges, such as poor dispersion, dimensional instability, and inadequate compatibility with composites. These issues can lead to reduced membrane permeability owing to pore blockage by composites or increased membrane thickness, which creates impeded transmission paths. Moreover, the recombination of nanomaterials can create larger pore sizes or free volume within the membrane, thereby increasing the permeability of both solvents and solutes. Attributed to the solution diffusion mechanism, it leads to the higher overall permeability but lower selectivity, facing the trade-off in reverse osmosis membranes22.

Carbon dots (CDs), a zero-dimensional carbon-based nanomaterial with dimensions typically below 10 nm, offer a promising alternative for membrane regulation23. Their tunable surface functional groups, excellent dispersibility, and biocompatibility make them particularly suitable for integration into various membrane types, including nanofiltration (NF), RO, and ultrafiltration membranes (UF)24,25. CDs can be uniformly embedded in polymer matrices to form defect-free selective layers, enhancing water permeability and antifouling performance without compromising structural integrity. Depending on the application, CDs can be strategically incorporated into dense selective layers, supporting layers, or throughout the entire membrane structure26. Previous studies have demonstrated that embedding CDs into polyamide (PA) layers can reduce transport resistance, mitigate concentration polarization, and improve both desalination efficiency and permeability27,28. The high affinity between CDs and PA improves membrane wettability and allows precise control over PA morphology, roughness, and cross-linking density. Amine-rich CDs can further create dense, functional separation layers that are capable of discriminating between monovalent and multivalent ions29. Nonetheless, CDs-enhanced membranes are still subject to challenges such as the permeability–selectivity trade-off, weak interfacial adhesion, and limited long-term stability, particularly when considering environmental sustainability30. In our earlier work, we have employed the mechanistic insights of biomimetic nanochannels of micro-structures on enhancing desalination from molecular simulations, as well as the trade-off relations between permeability and selectivity, facilitating the overcoming of the trade-off in reverse osmosis membranes through homologous matching of additives31–34.

To address these challenges, this study introduces an approach that utilizes the capabilities of CDs to modify dense selective or substrate layers. Particularly, CDs derived from phenylenediamine and wood were used to enhance the properties of the CTA substrate and the phenylenediamine-derived PA selective layer through a homologous matching strategy to achieve high-efficiency desalination. These CDs act as nano-intercalators, facilitating in-situ polymerization of the PA layer on the CTA surface to form a high-performance composite RO membrane. The nanoscale dimensions and abundant functional groups of the CDs help to shorten the water transport pathway and reduce transmembrane resistance by modulating the interfacial polymerization process. Moreover, the modified CDs (M-CDs) serve as enhancers and interfacial bridges between CTA and PA, improving the overall compatibility and performance of the composite membrane. As a result, the optimized membrane achieves a salt rejection rate of 99.1% with a permeate flux of 18.3 L·m−2·h−1. It also exhibits excellent chlorine resistance, with a 16% increase in flux after exposure to 2000 ppm NaClO, and maintains nearly 100% performance stability over 700 minutes of continuous operation. These results demonstrate the superiority of the proposed membrane over existing renewable desalination membranes. By incorporating CDs via a homologous matching strategy, this work highlights the potential of biomass-derived nanomaterials in sustainably enhancing the performance of CTA-based RO membranes for desalination.

Results

Innovative strategy for interfacial layer assembly through homologous matching of CDs and PA derived from PD

To address the limitations of traditional CTA-RO membranes, this study introduces a homologous matching strategy to fabricate a series of CTA-based RO membranes (O-CDs-0.1, M-CDs-0.1, and P-CDs-0.1). Carbon dots (CDs: O-CDs, M-CDs, and P-CDs) derived from different phenylene diamine (PD) precursors were synthesized via a simple hydrothermal method (Fig. 1a). And different CDs were incorporated as nano-intercalations into the corresponding aqueous monomers (OPD, MPD, and PPD) for the in-situ interfacial polymerization of the PA layer on the CTA surface (Fig. 1b). The results revealed that CDs enhanced the low-resistance water transport channels due to their nanoscale size, abundant functional groups, and excellent hydrophilicity (Fig. 1c). Notably, the membrane incorporated with M-CDs (CTA/M-CDs/PA) demonstrated a simultaneous enhancement in permeability (from 15.2 to 18.3 L·m−2·h−1) and salt rejection (from 96.5% to 99.1%), effectively mitigating the inherent trade-off effect (Fig. 1e, f). This superior performance is attributed to the homologous matching between M-CDs and the MPD monomer. Their structural similarity ensured seamless integration at the interface, moderated the polymerization kinetics to form a denser and more uniform PA network, and optimized the incorporation of nanochannels for efficient water transport. This mechanistic insight into the synergistic interaction among the CTA base, the homologous M-CDs, and the PA layer is further supported by molecular dynamics simulations (Fig. 1d).

Fig. 1. A strategy to address the trade-off in reverse osmosis membranes.

Fig. 1

Open in a new tab

a, b Homologous matching for assembling cellulose triacetate (CTA)/carbon dots (CDs: O-CDs, M-CDs, and P-CDs) derived from o-phenylenediamine (OPD), m-phenylenediamine (MPD), or p-phenylenediamine (PPD) precursors/polyamide (PA) reverse osmosis (RO) membranes. c, d Molecular dynamic simulation. e, f Overcoming the trade-off in this work.

Effects of CDs synthesized from different carbon sources on CTA-based membranes

Eucalyptus powder and PD derivatives (OPD, MPD, and PPD) were used as dual carbon sources to develop sustainable CDs (O-CDs, M-CDs, and P-CDs), with typical nanoscale sizes and abundant functional groups. The synthesis of different wood-derived CDs was confirmed through fluorescence spectroscopy, ultraviolet absorption spectroscopy, and fluorescence behaviors (Fig. S1 and S2)35,36. High-resolution transmission electron microscopy (HRTEM) images and particle size distributions revealed that O-CDs, M-CDs, and P-CDs exhibited uniform particle sizes, with lattice spacings of 1.53 ± 0.21, 2.22 ± 0.56, and 1.88 ± 0.48 nm, respectively (Fig. S3). The small particle size and graphene-like spacing of the synthesized CDs facilitated their effective incorporation into CTA-based composite RO membranes, thereby modifying their internal microstructures and mitigating defects at the nanoparticle-selective separation layer interface37,38. This was further confirmed through scanning electron microscopy (SEM) analysis of CDs-embedded O-CDs-0.1, M-CDs-0.1, and P-CDs-0.1 membranes.

In the presence of CDs, the surfaces of these membranes exhibited irregular and wrinkled pits. This phenomenon can be attributed to the thermal effects resulting from the reaction between amino and acyl chloride groups during the interfacial polymerization of the aqueous phase (PD) and organic phase (trimesoyl chloride, TMC). This reaction caused solvent evaporation, leading to the formation of a nodular and wrinkled morphology during condensation (Fig. 2a). Among the various membranes, the M-CDs-0.1 membrane exhibited a denser and smoother modified surface. This difference may be attributed to the varying diffusion rates of different PD monomer structures in the aqueous solution compared with the organic phase. Cross-sectional SEM images (inserted in Fig. 2a) revealed that M-CDs-0.1 featured a thinner selective layer and a more pronounced fold morphology, enabling M-CDs to facilitate the formation of shorter water channels with reduced resistance across the RO membrane. This was due to the meso-diamine structure of M-CDs nanoparticles, which enables them to function more effectively than O-CDs and P-CDs nanoparticles as mass transport mediums at the interface of aqueous and organic solutions. Moreover, the contact angles of these membranes were assessed (Fig. 2b). Compared with the CTA-RO membrane, the CDs-embedded membranes exhibited significantly lower contact angles, indicating enhanced hydrophilicity. This improvement in hydrophilicity contributed to the increased permeability of the RO membranes. Consequently, the M-CDs-0.1 membrane, assembled with an M-CDs-modified selective layer, achieved a salt rejection rate of up to 99.1% and a permeate flux of 18.32 L·m−2·h−1 (Fig. 2c, d). The superior performance of the M-CDs-0.1 membrane can be ascribed to the homologous matching effects of M-CDs within the CTA/M-CDs/PA RO membrane system. This enhancement was due to the relatively low steric hindrance of MPD and the linear chemical molecular structure of PA molecules synthesized from MPD and TMC (Fig. 2e and S4). Furthermore, the effects of different aqueous monomer types on the chemical structure (Figs. S5 and S6) and separation performance (Fig. S7) of CTA/PA membranes were investigated (Note S1). The results revealed that the PA separation layer formed through the interfacial polymerization reaction between the aqueous monomer (MPD) and organic monomer (TMC) exhibited the highest compatibility with the CTA-based membrane surface. This layer featured optimal separation performance, further supporting the previous findings.

Fig. 2. Effects of CDs derived from different carbon sources on CTA-based membranes.

Fig. 2

Open in a new tab

a Scanning electron microscopy (SEM) images of surfaces and cross-sections of O-CDs-0.1, M-CDs-0.1, and P-CDs-0.1 membranes with different carbon dots (O-CDs, M-CDs, P-CDs) derived from o-phenylenediamine, m-phenylenediamine, and p-phenylenediamine, respectively. b Contact angles of cellulose triacetate reverse osmosis membrane (CTA-RO), and O-CDs-0.1, M-CDs-0.1, and P-CDs-0.1 membranes. c, d Salt rejection and permeate flux data of M-CTA membrane modified by m-phenylenediamine (MPD) monomer, and O-CDs-0.1, M-CDs-0.1, and P-CDs-0.1 membranes. e The homologous matching effects of M-CDs on CTA/M-CDs/PA RO membranes (PA: polyamide, TMC: trimesoyl chloride). Error bars in (b–d) represent standard deviations, n  =  5 independent samples.

Effects of different M-CDs concentrations on CTA-based membranes

The evolution of the membrane’s microstructure with increasing M-CDs content is shown in Fig. 3a. At higher M-CDs concentrations, the membranes develop irregular and fluctuating ridge-valley structures. This alteration may be attributed to the disruption of the cross-linking reaction between MPD and TMC during the interfacial polymerization process, and the presence of M-CDs leads to the formation of additional diffusion channels within the nascent PA network39. Particularly, the increased presence of MPD molecules in the aqueous phase enables deeper penetration into the nascent PA layer, facilitating further reactions with TMC molecules in the organic phase. This process continues until the PA layer reaches a sufficient thickness and density, which inhibits further MPD diffusion and terminates the cross-linking reaction. Cross-sectional SEM images confirm that the membrane modified with 0.05 wt% M-CDs exhibits a relatively thin and dense barrier layer. As the M-CDs concentration increased, the surface thickness of the membrane also increased (Fig. 3a). This indicates that the functionalized hydrophilic groups (−OH and −COOH) on the nanoparticles facilitate molecular transport and the diffusion of MPD to the interface. Despite the thinner permeation-selective layer of the M-CDs-0.05 membrane, the membrane does not achieve the same retention capacity as M-CDs-0.1 owing to its shorter water transport path and lower transmembrane resistance. Among the modified membranes, M-CDs-0.1 exhibits the highest surface roughness, with Ra and Rq values of 10.5 and 14.7, respectively (Figs. S8 and S9), and the low contact angle of 42.6° (Fig. S10). It has been demonstrated that the optimal incorporation of M-CDs can significantly enhance both salt rejection and permeate flux of CTA-based membranes (Fig. 3b).

Fig. 3. Effects of different M-CDs concentrations on CTA-based membranes.

Fig. 3

Open in a new tab

a Scanning electron microscopy (SEM) images of upper surfaces and cross-sections of M-CDs-0.05, M-CDs-0.1, and M-CDs-0.15 membranes loaded with different concentrations of carbon dots (M-CDs) derived from eucalyptus powder and m-phenylenediamine, respectively. b Salt rejection and permeate flux for different membranes. c Variation curves of cross-linking degree as a function of simulation time (PA: polyamide). d Diffusion coefficient of m-phenylenediamine (MPD) molecules in the aqueous phase along the normal direction (z-axis) of the oil-water interface. e Number of MPD molecules aggregated within 5 Å of M-CDs and CDs. f Simulation snapshot of MPD aggregation near M-CDs: blue represents M-CDs, cyan represents MPD molecules. g Average pore size of dry membranes after hydrolysis and annealing treatment. h Full width at half maximum (FWHM) of the dry membrane pore size distribution. Error bars in (b) represent standard deviations, n  =  5 independent samples. Error bars in (d) represent standard deviations, n  =  10 independent samples. Error bars in (e) represent standard deviations, n  =  100 independent samples. Error bars in (g) and (h) represent standard deviations, n  =  10 independent measurements.

To gain deeper insight into the molecular mechanisms underlying the influence of M-CDs on PA formation, molecular dynamics simulations were conducted (models are shown in Fig. 1d). The cross-linking process between MPD and TMC was simulated in two stages (Fig. 3c): an initial rapid reaction phase with abundant reactants at the interface, followed by a slowdown due to increased steric hindrance and reduced availability of reactive sites, which is consistent with the experimental observations and prior studies40,41. At low concentrations, M-CDs had minimal impact on the cross-linking reaction. However, as the concentrations of M-CDs increased, both the reaction rate and the ultimate degree of cross-linking decreased significantly. In contrast, plain CDs exhibited a weaker effect. Similarly, the diffusion of MPD toward the interface was notably slowed with higher M-CDs concentrations, although CDs also reduced diffusion to a lesser extent (Fig. 3d). Notably, more MPD molecules accumulated around M-CDs than around plain CDs (Fig. 3e–3f), owing to attractive interactions between M-CD-adsorbed MPD and free MPD in the aqueous phase, which further restricted molecular diffusion. Post-annealing analysis of the cross-linked membrane structures revealed average pore sizes in the range of 3–4 Å (Fig. 3g). The smallest pore size (3.12 Å) and the narrowest distribution (smallest FWHM) were achieved at an M-CDs mass fraction of 6.89% (Fig. 3h). These results indicate that an appropriate amount of M-CDs can refine the membrane microstructure, leading to more uniform and smaller pores, which are critical for enhancing separation performance.

Mechanism of enhanced salt rejection and permeation rates of M-CDs in CTA-based membranes

The enhanced salt rejection and permeation capabilities of M-CDs incorporated into CTA-based membranes were systematically investigated through microstructural and physicochemical characterizations. SEM and AFM analyses of the pristine CTA-RO, M-CTA, and M-CDs-0.1 membranes revealed that both modified membranes exhibited a wrinkled PA layer morphology (Fig. 4a and S11). After embedding M-CDs into the intermediate layer between the CTA support and the PA active layer, the surface morphology of M-CTA and M-CDs-0.1 membranes became notably distinct from that of the pristine CTA-RO. Importantly, the incorporation of M-CDs contributed to a more uniform surface microstructure of the PA layer compared to M-CTA.

Fig. 4. Mechanism of enhanced salt rejection and permeation rates of M-CDs in membranes.

Fig. 4

Open in a new tab

a Atomic force microscope (AFM) images of upper surfaces for cellulose triacetate reverse osmosis membrane (CTA-RO), M-CTA membrane modified by m-phenylenediamine (MPD) monomer, and M-CDs-0.1 membrane with carbon dots (M-CDs) derived from eucalyptus powder and m-phenylenediamine. b-c Effect of M-CDs and CDs concentration on water flux and salt rejection rate. d Changes in water content of membranes doped with different amounts of M-CDs or CDs. e Influence of M-CDs concentration on water density distribution inside the membrane (along the pressure gradient direction). f Influence of M-CDs concentration on the membrane’s intrinsic density distribution (along the pressure gradient direction). g Variation of water-water coordination number and Na⁺-water coordination number in the membrane with doping concentration of M-CDs or CDs. h Stable configuration of the cross-linked product after hydrolysis (replacing residual acyl chloride groups) and annealing relaxation; the blue framework represents the carbon-based structure of M-CDs. i Simulated system setup for saltwater separation through the membrane: graphene piston plates on both sides apply pressures P₁ and P₂ to create a transmembrane pressure difference; the cyan transparent area represents water molecules, orange and green spheres indicate Na⁺ and Cl⁻ ions, respectively, and the pink structure corresponds to the carbon skeleton of M-CDs within the membrane. j, k Distribution of M-CDs and CDs inside the membrane. l Enhanced salt rejection and permeation mechanism of M-CDs on CTA-based membranes (PA: polyamide, TMC: trimesoyl chloride). Error bars in (b) and (c) represent standard deviations, n = 2 independent measurements for steady-state flux and n = 20 independent samples for salt rejection. Error bars in (d) and (g) represent standard deviations, n = 100 independent samples.

The observed decrease in contact angles can be attributed to the abundant polar oxygen-containing functional groups of M-CDs and the intrinsic hydrophilicity of amino groups, which is consistent with the chemical structure analysis based on the FTIR spectra (Fig. S12)42. Moreover, the hydrophilic nature of M-CDs facilitated the transport of MPD molecules, enabling their outward dispersion at the interface. However, the formation of hydrogen bonds between MPD and M-CDs molecules interfered with this process, leading to reduced crosslinking in the PA layer. Consequently, a greater number of unreactive acyl chloride groups remained on the surface of the M-CDs-0.1 membrane, along with a higher proportion of unreacted hydrophilic carboxyl and amine groups. These factors contributed to the improved hydrophilicity of the M-CDs-0.1 membrane. The significant increase in surface roughness from 3.18 ± 0.21 (CTA-RO) to 10.5 ± 0.53 (M-CDs-0.1) (Fig. 4a and Table S1) is consistent with the reduction in the contact angle from 68.2° (CTA-RO) to 42.1° (M-CDs-0.1) (Fig. S13). The increase in hydrophilicity and permeability from 15.2 (M-CTA) to 18.3 L·m−2 ·h−1 (M-CDs-0.1) resulted from the incorporation of the CDs-modified CTA matrix and PA layer (Figs. S14 and S15).

Further insights and evidence were gained through molecular dynamics simulations. The hydration swelling of the dry membrane increased its thickness from ~4 nm to ~6 nm. Under an operating pressure of 600 MPa for a 200 mM NaCl solution, the pure PA membrane showed poor permeability and a salt rejection of only 32.09%. With the incorporation of M-CDs, the separation performance of the membrane improved significantly. Concretely, as the M-CDs content increased from 0 to 6.89 wt%, the permeate flux increased from 2.24 to 16.52 × 10⁴ L·m−2·h−1·MPa−1, and the salt rejection rose from 32.09% to 76.55%, effectively overcoming the permeability-selectivity trade-off (Figs. 4b, 4c). However, further increasing the M-CDs concentration led to a decline in performance. In contrast, the unmodified CDs also improved membrane properties, but a clear concentration-dependent trend was not observed.

According to Elimelech’s “solution-friction model”, water transport occurs mainly in the form of clusters, and membrane permeability is governed by pore size, solvent molecular dimensions, and viscosity43. The pure PA membrane after full swelling exhibited a water content of 19.02%, consistent with values reported in the literature (10–23%)44,45 and close to the value of 18.9% reported by Zhang et al.46. The incorporation of M-CDs and CDs increased the membrane water content, enhancing water permeability (Fig. 4d). However, high concentrations of M-CDs reduced water content due to their higher density compared to the PA matrix, which restricted the overall swelling of the membrane. Density distribution analysis along the pressure gradient direction revealed a distinct dense/porous bilayer structure in M-CDs-incorporated membranes (Figs. 4e-4f). This structure originated from the cross-linking reaction between M-CDs and TMC, leading to M-CDs enrichment on the aqueous side and formation of a dense layer that enhanced salt rejection. At low concentrations, M-CDs promoted membrane swelling and formed a dense cross-linked network, improving both permeability and selectivity. At high concentrations, however, the reduced spacing between the M-CDs severely restricted chain mobility, resulting in the formation of rigid, high-density regions and defects, which in turn degraded performance47,48. In contrast, unmodified CDs did not participate in cross-linking and could diffuse freely at the oil–water interface, thus failing to form a dense layer; their distribution within the membrane likely influences separation performance (Fig. 4h–4k).

Further analysis of water–water (O–O) and Na⁺–water (Na⁺–O) coordination numbers showed that the O–O coordination number in the membrane ranged from 5 to 8, consistent with Elimelech’s computational results43. The Na⁺–O coordination number was lower than the O–O value, indicating that Na⁺ ions should undergo partial dehydration before entering membrane channels (Fig. 4g). The incorporation of M-CDs increased the O–O coordination number while decreasing the Na⁺–O value. This observation suggests that water transport pathways were either extended or optimized, while ion transport was further hindered, thereby supporting the proposed mechanism.

Figure 4l illustrates the enhanced salt rejection and permeation mechanism of M-CDs in CTA-based membranes. During the interfacial polymerization process, M-CDs in the aqueous MPD monomer solution permeated the pores at the CTA-RO membrane interface and reacted with the TMC organic monomers to form M-CDs nanoaggregates. After heat treatment, the irreversible shrinkage of the composite membrane ensured the stable fixation of these nanoaggregates at the interface between the CTA matrix and the PA separation layer. This resulted in the formation of a dense modified PA layer, which significantly enhanced the salt retention capacity of the membrane. Therefore, the M-CDs-0.1 membrane exhibited a salt retention rate of up to 99.1% and a permeate flux of 18.32 L·m−2 ·h−1, outperforming the M-CTA RO membrane. More importantly, M-CDs enhanced the salt rejection rate of the M-CDs-0.1 membrane and maintained the permeate flux, thereby mitigating the typical trade-off between salt rejection and permeate flux observed in conventional CTA-based RO membranes.

Contributions of M-CDs to high chlorine resistance and long-term stability of CTA-based RO membranes

Generally, the interfacial polymerization process for creating a PA layer involves diffusion and reaction steps. First, the monomer MPD, which is embedded within the matrix pores and on the surface, diffuses towards the reaction interface. At this interface, MPD reacts with the TMC monomer in the reaction medium to form a complete PA selective layer. This study adheres to the established reaction mechanism and incorporates M-CDs into the interfacial polymerization process. These M-CDs modulate the reaction rate, thereby influencing the thickness and microstructure of the PA layer. Consequently, the M-CTA membrane exhibits a denser selective layer structure, which effectively rejects most small molecules and monovalent salts. The addition of nanoscale M-CDs with surface electronegativity at the CTA interface positively affects both the interfacial polymerization rate and the formation of the PA layer (Fig. 5a). Figure 5b shows the changes in the surface charges of the CTA-RO and M-CDs-0.1 membranes at a pH of 7.0. The electronegativity observed on the CTA-RO membrane surface is consistent with the findings in the previous literature. In contrast, the M-CDs-0.1 membrane, which incorporates M-CDs within the PA layer, exhibits a significantly higher negative charge. This enhanced negativity can be attributed to the carboxyl and amino groups present in M-CDs. Notably, the high electronegativity from the carboxyl groups generated during the interfacial polymerization reaction leads to a significantly lower contact angle for the M-CDs-0.1 membrane compared with the M-CTA membrane. Additionally, considering the negatively charged Cl− in NaCl solutions, the increased electronegativity facilitates electrostatic repulsion between the salt-rejecting components and the negatively charged ions at the RO membrane filtration interface. Consequently, the M-CDs-0.1 membrane exhibits a superior salt rejection rate during the RO process. Moreover, the increased negative charge on the M-CDs-0.1 membrane surface enhances its ability to attract water molecules through hydrogen bonding, leading to an increase in permeate flux.

Fig. 5. Contributions of M-CDs to high chlorine resistance and long-term stability of membranes.

Fig. 5

Open in a new tab

a Transmission electron microscope (TEM) images and particle size distribution of M-CDs. b Zeta potential measurements of cellulose triacetate reverse osmosis membrane (CTA-RO) and M-CDs-0.1 membrane with carbon dots (M-CDs) derived from eucalyptus powder and m-phenylenediamine. c Contribution of M-CDs to the enhanced desalination efficiency of CTA-based RO membranes. d, e Long-term stability performance of M-CTA membrane modified by m-phenylenediamine (MPD) monomer and M-CDs-0.1 membrane. f Chlorine resistance of M-CTA and M-CDs-0.1 membranes. Error bars in (b) represent standard deviations, n = 4 independent samples. Error bars in (f) represent standard deviations, n = 5 independent samples.

The contributions of M-CDs to enhancing the desalination efficiency of CTA-based RO membranes are illustrated in Fig. 5c. The electrostatic attraction between M-CDs leads to the adsorption of some nanoparticles onto the CTA substrate surface, thereby partially blocking the pores. This blockage hinders the diffusion of MPD, which limits the extent of the interfacial polymerization reaction with TMC. During the interfacial polymerization process, the adsorbed M-CDs occupy space and may react with TMC, competing with the reaction between MPD and TMC. This competition reduces the degree of the interfacial polymerization reaction and hinders the formation of a well-defined PA network structure. With the addition of M-CDs nanoparticles, the surface roughness of the M-CDs-0.1 membrane increases owing to their interaction during the interfacial polymerization process. The abundance of hydrophilic groups on M-CDs nanoparticles expands the reaction zone between the two immiscible phases (aqueous and organic), enabling MPD molecules to penetrate deeper into the organic phase. Consequently, the M-CDs-0.1 membrane exhibits a more nodular and wrinkled morphology, which enhances the network structure of its PA layer. This refined structure enhances permeability and ensures the high retention capabilities of the membrane.

The relationship between permeate flux and salt rejection may result from the complexities of the mass transfer mechanisms and the microstructure of the RO membrane. The unmodified M-CTA membrane exhibited a permeate flux of 15.2 L·m−2 ·h−1 and a salt rejection rate of 96.5% when treating the 3000 ppm NaCl solution at 2 MPa. In contrast, under optimal loading conditions, the performance of M-CDs-0.1 membrane exceeded that of the M-CTA membrane, achieving a permeate flux of 18.3 L·m−2 ·h−1. This indicates a significant increase of 20.4% in permeate flux and a higher retention rate of 99.1% (Fig. 5d, e). This phenomenon can be elucidated through the mass transfer mechanisms across membranes. The mechanisms involve three fundamental steps: the adsorption of water molecules, their diffusion under pressure through the dense structure layer, and subsequent desorption. The SEM analysis indicates that the introduction of M-CDs leads to a rougher surface on the dense layer and a thinner separation layer. These alterations facilitate the efficient transfer of water molecules, thereby enhancing the permeate flux49. Additionally, the increased number of oxygen-containing groups on the membrane surface reduces the water contact angles, thereby enhancing hydrophilicity and further increasing permeate flux. The steric hindrance created by M-CDs limits the penetration of MPD, which delays the formation of the PA layer. Consequently, as the thickness of the active layer decreases, the mass transfer resistance of the membrane significantly decreases, leading to improved permeate flux. However, excessive amounts of M-CDs may result in thicker, denser membranes and agglomeration, which can reduce membrane permeability. Notably, the in situ interfacial polymerization process mainly occurs within the pores of the CTA substrate and the surfactant layer, while post-heat treatment stabilizes the embedding of M-CDs nanoaggregates within these pores. Overall, the strategic surface presence of M-CDs significantly contributes to the partial mitigation of the trade-off in the CTA-based membranes.

The long-term salt interception capability and osmotic stability of CTA-based membranes are crucial factors determining their service life. In this test, the M-CDs-0.1 membrane was continuously operated for 700 min at an operating pressure of 2 MPa, and its permeate flux and salt rejection were systematically recorded (Fig. 5d, e). During these long-term experiments, the M-CDs-0.1 RO membrane exhibited excellent stability owing to the use of CTA-RO as the modified base, which inherently provided high mechanical strength. Furthermore, the excellent hydrophilicity and nanoscale properties of M-CDs created ideal conditions for constructing a highly permeable PA layer. The excellent compatibility of M-CDs maintained the structural integrity of the active layer, while their nanoscale size facilitated a uniform distribution of M-CDs interlayers on the membrane surface. The M-CDs-0.1 RO membrane exhibited excellent performance in both permeate flux and salt rejection. Thus, M-CDs effectively functioned as microstructural water transport channels on the membrane surface, thereby significantly enhancing the desalination performance of CTA-based RO membranes.

Chlorine resistance is a crucial factor in evaluating the stability of RO membranes in practical applications. After treatment with NaClO solution, significant differences were observed between the M-CTA and M-CDs-0.1 RO membranes (Fig. 5f). The permeate flux of the M-CTA RO membrane increased to 1.2 times its original value, while the permeate flux of the M-CDs-0.1 RO membrane increased to 1.16 times its original value. However, the salt rejection rate of the M-CTA RO membrane significantly decreased to 0.95% of its original level. Conversely, the M-CDs-0.1 RO membrane exhibited a slight decrease in salt rejection rate, indicating that the M-CDs modified membrane exhibited superior resistance to chlorine degradation. The lower chlorine resistance of the M-CTA RO membrane can be attributed to the degradation of the amide bond50,51. Particularly, active chlorine atoms underwent substitution reactions with free hydrogen atoms on the amide bond, causing the chlorine atoms to migrate to the ring and bind with PA. This disrupted the cross-linking network structure of the amide and weakened hydrogen bonding, generating more water transport channels and resulting in improved permeability. The incorporation of M-CDs nanoparticles during the formation of the PA layer in the M-CDs-0.1 RO membrane enhanced its chlorine resistance owing to several factors. First, the interaction between the PA layer and M-CDs strengthened hydrogen bonding between the two groups, thereby preventing active chlorine atoms from substituting for free hydrogen. Second, the abundance of hydrophilic groups on the M-CDs surface enhanced the hydrophilicity of M-CDs-0.1, thereby acting as a protective layer for the PA layer52,53. Additionally, zeta potential analysis revealed that M-CDs-0.1 exhibited a higher surface negative charge than RO membranes without M-CDs. This increased negative charge enhanced the Cl− barrier, as electrostatic repulsion between Cl− and the negatively charged surface reduced the likelihood of chlorine reaching the PA surface. This interaction further decomposed Cl− and strengthened the Cl− barrier54,55. Moreover, the presence of M-CDs provided additional sites for active chlorine attack, while the amino groups in the M-CDs nanoparticles served as sacrificial groups. This interaction slowed down the hydrolysis of the PA network and significantly improved the chlorine resistance of M-CDs-0.156,57.

Interaction mechanism between CTA, wood-derived CDs, and PA in CTA-based membranes

The chemical functional groups in CDs were confirmed through FTIR spectroscopy (Fig. S12). The observed adsorption peaks at 3502, 1721, and 1670 cm−1 correspond to the tensile vibrations of O–H (hydroxyl), C=O (carbonyl), and N–H (amide) bonds, respectively58. These peaks indicate that the surfaces or edges of the prepared CDs contain oxygen-containing functional groups, as analyzed from a molecular chemistry perspective. FTIR spectra (Fig. S16) of the M-CTA membrane exhibit distinct characteristic peaks at 3410, 1720, and 1380 cm−1, corresponding to the tensile vibrations of O–H, C=O, and C–N functional groups, respectively. The results indicate that a PA layer has been successfully formed on the CTA base surface, in which CDs are embedded. Similarly, the P-CTA and O-CTA membranes exhibit the same chemical structural features (Fig. S5). The incorporation of M-CDs into the PA layer leads to a more pronounced −OH absorption peak, further confirming the integration of M-CDs at the interface between the CTA and PA layers. This observation is consistent with the results from the addition of P-CDs and O-CDs (Fig. S12). The FTIR spectra of M-CTA, O-CTA, and P-CTA membranes exhibit distinct stretching vibration peaks at 1630 and 1560 cm−1, corresponding to C=O and N–H groups. These peaks are indicative of the PA layer, indicating that the introduction of CDs into the membrane does not alter the chemical structure of the PA layer. This observation suggests that the addition of CDs to the aqueous solution does not negatively affect the polymerization process of the PA layer, thereby preserving its integrity and functionality.

Moreover, the FTIR spectrum of the M-CDs-0.1 membrane features a vibration peak at 3430 cm−1, indicating the presence of −OH, C=O, and −NH groups within the membrane. Compared with the M-CTA membrane, the spectrum of the M-CDs-0.1 membrane displays enhanced absorption peaks for C=O, −NH, and −OH groups. This observation indicates that the carboxyl and hydroxyl functional groups from M-CDs are successfully integrated into the M-CDs-0.1 membrane, thereby improving its hydrophilicity and permeability. Furthermore, as the concentration of M-CDs increases, the −OH absorption peak is significantly enhanced, along with a weaker C–C tensile vibration at 1610 cm−1 and a more pronounced C=O tensile vibration at 1710 cm−1. These changes can be attributed to the presence of hydroxyl, carboxyl, and carbonyl groups within the M-CDs nanoparticles (Fig. S17). The presence of these hydrophilic groups is crucial for the successful incorporation of M-CDs into the M-CDs-modified PA layer, contributing to the observed increase in the hydrophilicity of the M-CDs-0.1 surface. However, owing to the relatively low quantity of CDs added to the PA layer, some of the characteristic peaks in the M-CDs spectra may not be easily identifiable and could be masked by the dominant peaks from the CTA and PA molecules.

To further investigate the chemical composition, confirm the successful integration of CDs, and elucidate the interactions among the three components, X-ray diffraction (XRD) and XPS tests were conducted. The XRD patterns (Figs. S18 and S19) reveal a consistent diffraction pattern across all membranes, exhibiting a peak at 2θ = 22.6°. This peak originates from the inherent lattice properties of carbon-based materials, indicating their graphitic carbon nature59. This observation is consistent with the highly disordered structure of carbon, evidenced by a blurred, non-dispersive pattern centered at 2θ = 23.2°, similar to the characteristic diffraction pattern of CTA-II. Furthermore, the overall broadening of the diffraction peaks confirms the successful embedding of M-CDs onto the membrane surface.

The XPS results confirm that the M-CTA RO membrane contains C, O, and N elements, which mainly originate from the PA layer formed through a reaction between MPD and TMC on the membrane surface (Table S2 and Fig. 6a–d). Notably, both M-CDs-0.1 and P-CDs-0.1 membranes exhibit significantly higher oxygen content than M-CTA (Fig. S20). This increase can be attributed to the successful incorporation of CDs, which introduces additional oxygen-containing groups, thereby enhancing membrane hydrophilicity60. The P-CDs-0.1 membrane features a lower total N content than M-CDs-0.1, suggesting a reduced reaction level between P-CDs and the TMC monomer. Additionally, the N1s spectrum (Fig. S21) reveals that P-CDs-0.1 contains a higher −NH2 content, indicating that a greater amount of the MPD aqueous solution is retained on the membrane surface. This phenomenon indicates lower reactivity between P-CDs and MPD, which hinders the formation of a uniform and dense selective layer, resulting in reduced permeability of the RO membrane.

Fig. 6. Interaction mechanism between CTA, wood-derived CDs, and PA in CTA-based membranes.

Fig. 6

Open in a new tab

a–d X-ray photoelectron spectroscopy (XPS) spectra of M-CTA membrane modified by m-phenylenediamine (MPD) monomer, and M-CDs-0.1 membrane with carbon dots (M-CDs) derived from eucalyptus powder and m-phenylenediamine. e Mean squared displacement (MSD) of azimuth shift (CTA cellulose triacetate, PA polyamide, CDs carbon dots). f Interaction energy diagram (CTA cellulose triacetate, PA polyamide, CDs carbon dots).

Furthermore, the introduction of M-CDs and P-CDs leads to a lower C/N ratio for the M-CDs-0.1 and P-CDs-0.1 membranes compared with the M-CTA membrane, indicating a decrease in the degree of cross-linking. This reduction in cross-linking can be attributed to several factors. The nanoaggregates formed during the in situ polymerization of TMC and CDs remain on the membrane surface and infiltrate the membrane voids, further reducing the cross-linking between the organic phase TMC and aqueous phase monomers. Moreover, the M-CDs-0.1 membrane exhibits elevated −NH− content and excess unreacted −NH2 groups. This may result from residual aqueous monomers, as some amino groups in these monomers react with the carbonyl groups of CDs, leading to a decrease in the cross-linking density of the PA layer and promoting the formation of rapid ion transport channels61. Notably, the electrostatic attraction between the negatively charged −OH/−COOH groups in CDs and positively charged amine groups in MPD molecules hinders the migration of MPD toward TMC molecules, thereby reducing the interfacial cross-linking degree. Additionally, the reduced cross-linking of the PA layer may be attributed to the water molecules surrounding the −COOH groups, which hinder the secure attachment of PA macromolecules around CDs62. Consequently, this process leads to the formation of interfacial spaces, resulting in a looser PA structure. Furthermore, the elemental analysis of C1s and O1s spectra (Fig. S21) reveals that the M-CDs-0.1 membrane exhibits a higher total oxygen content (34.52%) and a C=O content (61.14%), consistent with the improved hydrophilicity confirmed by the contact angle analysis.

Exploring the interaction mechanism between CTA, CDs, and PA within CTA-based membranes is crucial and requires an in-depth study. Molecular dynamics simulations of CTA, M-CDs, and PA molecules are conducted (Fig. 6e, f and Figs. S22–S24). M-CDs are embedded at the interface between CTA and PA, which serves as a key connector for facilitating transitions and optimizing the overall structure. The mean squared displacement (MSD) analysis of the azimuthal shift reveals insightful trends. As CTA molecules are isolated, the MSD value is at its lowest, indicating restricted molecular mobility due to their tightly packed arrangement. However, upon the polymerization of PA on the CTA surface, interactions at the interface layer enhance system flexibility, resulting in an increased overall MSD value. Additionally, the incorporation of M-CDs into the PA layer creates additional binding sites, which enhance intermolecular interactions. Consequently, the MSD value decreases and falls between the values observed for the isolated CTA molecules and polymerized PA scenarios. This finding indicates that M-CDs significantly influence the dynamics and interactions within the membrane structure.

A comprehensive analysis was performed to assess the interaction energies between CTA, M-CDs, and PA (Fig. 6f). Notably, CTA and M-CDs exhibit the strongest interaction owing to the ability of the hydroxyl groups in CTA to form stable hydrogen bonds with the functional groups on the M-CDs surface. Moreover, Van der Waals forces may contribute to this interaction, particularly owing to the favorable alignment of distances between the surface functional groups of M-CDs and the hydroxyl groups in CTA. In the selective layer, M-CDs and PA exhibit a stronger interaction than that between CTA and M-CDs. This enhanced interaction is due to the strong hydrogen bond donor and acceptor properties of the amide bond in PA, which facilitates the formation of numerous hydrogen bonds with the functional groups on the M-CDs surface. Although CTA and PA molecules can form hydrogen bonds through their hydroxyl and amide groups, the extent and strength of these bonds are limited when compared with those involving M-CDs owing to spatial constraints. Nevertheless, the significance of the hydrogen bonds formed between CTA and PA molecules should not be underestimated. The strong interaction between CTA and M-CDs indicates the significant potential of M-CDs to enhance the overall properties of composite membranes, including high-efficiency desalination, chlorine resistance, and long-term stability. Conversely, the weaker interaction between CTA and PA suggests that the formation of a stable composite structure between CTA and PA without M-CDs is challenging. Therefore, M-CDs can enhance the properties of the membrane and serve as a crucial connector between CTA and PA, thereby optimizing the performance of CTA-based membranes.

Discussion

In this study, a series of CTA-based RO membranes was successfully fabricated through a homologous matching strategy. These membranes were constructed through interfacial layer assembly, which involved the modification of both the CTA substrate and PA selective layer with varying amounts of CDs interlayers (including O-CDs, M-CDs, and P-CDs) derived from different sources of phenylenediamine/wood-derived carbon. Comprehensive morphological and molecular dynamics simulations systematically revealed that M-CDs modulated the interfacial polymerization process between MPD and TMC, thus optimizing the microstructure of the PA layer. At an optimal concentration of 0.1 wt%, M-CDs promoted appropriate MPD diffusion, reduced pore size with a narrow distribution, enhanced surface roughness (Ra = 10.5), and significantly improved hydrophilicity (contact angle = 42.6°). These structural alterations facilitated the formation of efficient water transport channels while imparting high salt rejection (99.1%) and large permeate flux of 18.3 L·m−2 ·h−1, effectively overcoming the permeability–selectivity trade-off in conventional CTA-based RO membranes. Molecular dynamics further elucidated that M-CDs enhance water–water coordination while hindering ion transport, and promote the formation of a dense–porous bilayer structure for improved separation. Additionally, hydrogen bonding between M-CDs, CTA, and PA enhances interfacial stability and chlorine resistance, ensuring long-term performance over 700 minutes. This work demonstrates the efficacy of bio-sourced carbon dots in refining membrane architecture through a homologous matching strategy, providing a sustainable and scalable strategy for advanced desalination technologies.

Methods

Materials and chemical reagents

Cellulose triacetate (CTA, 43–49 wt% acetyl) was purchased from Wengjiang Reagent Co., Ltd. (Guangdong, China). Eucalyptus chips were provided by a wood-working factory (Guangxi, China). Sodium chloride (NaCl, 99 wt%), hexane (≥99.5 wt%), trimesoyl chloride (TMC, 99 wt%), dimethyl sulfoxide (DMSO, 99 wt%), o-phenylenediamine (OPD, AR), m-phenylenediamine (MPD, AR), p-phenylenediamine (PPD, AR), and lauryl sodium sulfate (SDS, AR) were provided by Aladdin Reagent Co., Ltd. (Shanghai, China). Other chemicals used are described in the Supplementary Information.

Preparation of CDs derived from eucalyptus powder and phenylenediamine

First, eucalyptus chips were dried in the oven for 24 h, then grinded and sifted to obtain 300 mesh of eucalyptus powder. 3 g of different carbon sources (OPD, MPD, PPD) were added to the inner lining of a reactor containing 50 ml deionized water, and then 0.3 g of eucalyptus powder was added as another carbon source. After fully stirring for 1 h, the mixture was placed in a high-pressure reactor at 180 °C for 24 h to obtain a brown solid-liquid mixture. The obtained solid-liquid mixture was filtered through a 0.22 μm cellulose filter membrane to filter out the macromolecular particles. To further clarify the filtrate, high-speed centrifugation was performed. Finally, the small molecular precursors were removed by rotating dialysis to obtain pure carbon dots, which were labeled as O-CDs, M-CDs, and P-CDs, matching different carbon sources of OPD, MPD, and PPD, respectively.

Interfacial layer assembly of CTA/wood-derived CDs/PA RO membranes

The CTA-RO membrane was assembled through the phase inversion process, leveraging the non-solvent induced phase separation (NIPS) system, with details provided in the Supplementary Information63. Then, a PA layer was synthesized on the surface of the CTA-RO base membrane via conventional interfacial polymerization reaction. Concretely, the CTA-RO base membrane was immersed in a 2 wt% OPD/MPD/PPD solution containing 0.1 wt% CDs for 5 min, followed by a 3-minute air exposure to remove excess OPD/MPD/PPD solution. After heat treatment in a water bath at 85 °C for 10 min, the treated surface of CTA-RO base membrane was immersed in an n-hexane solution containing 0.15 wt% TMC for 2 min to ensure interfacial polymerization reaction with OPD/MPD/PPD aqueous solution on the surface of CTA-RO base membrane, and then to remove excess TMC solution. The CTA/PA RO membranes prepared by different aqueous solutions were designated as O-CTA, M-CTA, and P-CTA, respectively, to explore the influences of variational aqueous monomers on the performances of the membranes.

Subsequently, in order to modify the PA layer, the synthesized O-CDs, M-CDs, and P-CDs were added to the OPD, MPD, and PPD aqueous solutions at a concentration of 0.1 wt%, respectively. In particular, a series of interfacial layer assemblies of CTA/wood-derived CDs/PA RO membranes was fabricated using the same process outlined above, labeled as O-CDs-0.1, M-CDs-0.1, and P-CDs-0.1, respectively, where “0.1” denotes the concentration of the synthesized CDs in the respective MPD aqueous solution. Moreover, the effect of CDs concentration (0.05 wt%, 0.1 wt%, 0.15 wt%) on the properties of the composite membranes was also further discussed, named as M-CDs-0.05, M-CDs-0.1, and M-CDs-0.15, respectively.

Characterization

The microsize and micromorphology of the CDs were observed using transmission electron microscopy (TEM, FEI Tecnai, USA), and the diameter distribution of the CDs was further analyzed with Image J. The chemical structure was examined by Fourier transform infrared spectroscopy (Nicolet iS 50, Thermo Fisher, USA) to test the FTIR spectra of the samples in the range of 4000–600 cm−1. An X-ray diffractometer (SmartLab 9kw, Nikaku, Japan) was adopted to record the XRD pattern of the sample to analyze the crystal characteristics. The lyophily of the sample was tested by a contact angle device (DSA100, KRUSS, Germany). The surface and cross-sectional micromorphologies of the membranes were characterized using scanning electron microscopy (SEM, Sigma 300, Zeiss, Germany). The surface roughness of membranes was characterized by atomic force microscopy (AFM, Bruker size icon). Molecular dynamics simulations of systems containing CTA, M-CDs, and PA molecules were conducted following a 7-step compression and relaxation scheme, as outlined in Table S3 (detailed in the Supplementary Information). Molecular dynamics simulations of membrane formation and separation processes were performed using the LAMMPS package. System visualization was carried out with VMD and OVITO, while molecular coordinate files were generated using Moltemplate. Crosslinking reactions were modeled using the REACTER protocol. Additional details regarding the simulation methodologies are provided in the Supplementary Information.

Evaluation of membrane separation performance

Two key properties of RO membranes, i.e., permeate flux (Jw, L·m−2·h−1) and salt rejection (R, %), were tested using a lab-made cross-flow device and calculated by the corresponding equations (S1-S2). All membranes were measured at least 3 times under identical conditions, including an operating pressure of 2 MPa, NaCl concentration of 3000 ppm, effective membrane area of 15.89 cm2, a test temperature of 25 ± 2.0 °C, and a transverse flux of 1.0 L/min. For the evaluation of chlorine resistance of membranes, the membranes were immersed in the prepared 2000 ppm sodium hypochlorite (NaClO) solution for 2 h, rinsed several times, and then tested for separation performance. The chlorine resistance was assessed based on the permeate flux and NaCl rejection of the RO membranes before and after treatment with the NaClO solution. Additionally, the long-term salt rejection and osmotic stability of the RO membranes were also tested under the same conditions, with the only variation being the duration of the test.

Supplementary information

Supplementary Information (3.1MB, pdf)
Transparent Peer Review file (393.2KB, pdf)

Source data

Source Data (3.3MB, xlsx)

Acknowledgements

S.L.Z. acknowledged the financial support from Guangxi Science and Technology Innovation Platform Program (“Leitai” Action Plan-Guangxi Laboratory Capacity Building) (LT2504240023) and Guangxi Science and Technology Major Program (No. AA23073019). F.C.C. acknowledged the financial support from the National Natural Science Foundation of China (32571972; 32171703), the Natural Science Foundation of Guangxi Province (2024GXNSFFA010010), and the Bagui Youth Top Talent Program of Guangxi (The Major Talent Project of Guangxi Zhuang Autonomous Region). D.Y.H. acknowledged the financial support from Guangxi University’s Doubling Plan for Natural Sciences and Technological Innovation Development (2024BZRC008).

Author contributions

S.L.Z. conceived and guided the project. F.C.C. conceived the idea, coordinated this work, and revised the manuscript. D.Y.H. designed the experiments, wrote the manuscript with assistance and input from all authors, and revised the manuscript. X.Y.S. carried out the partial experiments and wrote the initial manuscript. S.Y.L. supplemented extensive simulations. X.Q. carried out the partial experiments and analyzed the data. X.F.X. conducted the partial simulations. Y.Q.W. revised the manuscript and discussed the results. The final version was approved by all authors before submission.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data that supports the findings of the study are included in the main text and supplementary information file. Source data are provided with this paper or on Figshare (10.6084/m9.figshare.31081366). All data are available from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Xinyu Shao, Shiyu Lv.

Contributor Information

Fangchao Cheng, Email: fangchaocheng@gxu.edu.cn.

Dongying Hu, Email: hdy@gxu.edu.cn.

Shuangliang Zhao, Email: szhao@gxu.edu.cn.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-69044-5.

References

  • 1.Kocher, J. D. & Menon, A. K. Addressing global water stress using desalination and atmospheric water harvesting: A thermodynamic and technoeconomic perspective. Energy Environ. Sci.16, 4983–4993 (2023). [Google Scholar]
  • 2.Foglia, F., Frick, B., Nania, M., Livingston, A. G. & Cabral, J. T. Multimodal confined water dynamics in reverse osmosis polyamide membranes. Nat. Commun.13, 2809 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Sapkota, B. et al. High permeability sub-nanometre sieve composite MoS2 membranes. Nat. Commun.11, 2747 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Hou, Y. et al. A super liquid-repellent hierarchical porous membrane for enhanced membrane distillation. Nat. Commun.14, 6886 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Qasim, M., Badrelzaman, M., Darwish, N. N., Darwish, N. A. & Hilal, N. Reverse osmosis desalination: A state-of-the-art review. Desalination459, 59–104 (2019). [Google Scholar]
  • 6.Chowdhury, M. R., Steffes, J., Huey, B. D. & McCutcheon, J. R. 3D printed polyamide membranes for desalination. Science361, 682–686 (2018). [DOI] [PubMed] [Google Scholar]
  • 7.Li, D., Yan, Y. & Wang, H. Recent advances in polymer and polymer composite membranes for reverse and forward osmosis processes. Prog. Polym. Sci.61, 104–155 (2016). [Google Scholar]
  • 8.Li, Z., Siddiqi, A., Anadon, L. D. & Narayanamurti, V. Towards sustainability in water-energy nexus: Ocean energy for seawater desalination. Renew. Sust. Energ. Rev.82, 3833–3847 (2018). [Google Scholar]
  • 9.Liu, W. et al. Pressure-driven membrane desalination. Nat. Rev. Methods Prim.4, 1–22 (2024). [Google Scholar]
  • 10.Lu, D. et al. An ultrahigh-flux nanoporous graphene membrane for sustainable seawater desalination using low-grade heat. Adv. Mater.34, 2109718 (2022). [DOI] [PubMed] [Google Scholar]
  • 11.Cao, S. et al. Cellulose nanomaterials in interfacial evaporators for desalination: A “natural” choice. Adv. Mater.33, 2000922 (2021). [DOI] [PubMed] [Google Scholar]
  • 12.Jiang, Z., Karan, S. & Livingston, A. G. Water transport through ultrathin polyamide nanofilms used for reverse osmosis. Adv. Mater.30, 1705973 (2018). [DOI] [PubMed] [Google Scholar]
  • 13.Werber, J. R., Osuji, C. O. & Elimelech, M. Materials for next-generation desalination and water purification membranes. Nat. Rev. Mater.1, 1–15 (2016). [Google Scholar]
  • 14.Ramanathan, A. A., Aqra, M. W. & Al-Rawajfeh, A. E. Recent advances in 2D nanopores for desalination. Environ. Chem. Lett.16, 1217–1231 (2018). [Google Scholar]
  • 15.Dai, R., Li, J. & Wang, Z. Constructing interlayer to tailor structure and performance of thin-film composite polyamide membranes: A review. Adv. Colloid Interface282, 102204 (2020). [DOI] [PubMed] [Google Scholar]
  • 16.Ihsanullah, I. Potential of MXenes in water desalination: Current status and perspectives. Nano-Micro Lett.12, 72 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Dutta, S. et al. Metal–organic frameworks for water desalination. Adv. Funct. Mater. 34, 2304790.
  • 18.Peng, Y., Li, W. & Zhu, H. Amphiphilic MOF nanoflakes for ultraselective polyamide membranes. Matter5, 1350–1352 (2022). [Google Scholar]
  • 19.Gao, X. et al. Highly permeable and antifouling reverse osmosis membranes with acidified graphitic carbon nitride nanosheets as nanofillers. J. Mater. Chem. A5, 19875–19883 (2017). [Google Scholar]
  • 20.Dong, X. et al. Mesoporous hollow nanospheres with amino groups for reverse osmosis membranes with enhanced permeability. J. Membr. Sci.657, 120637 (2022). [Google Scholar]
  • 21.Hailemariam, R. H. et al. Reverse osmosis membrane fabrication and modification technologies and future trends: A review. Adv. Colloid Interface276, 102100 (2020). [DOI] [PubMed] [Google Scholar]
  • 22.Park, H. B., Kamcev, J., Robeson, L. M., Elimelech, M. & Freeman, B. D. Maximizing the right stuff: The trade-off between membrane permeability and selectivity. Science356, eaab0530 (2017). [DOI] [PubMed] [Google Scholar]
  • 23.Liu, H. et al. A review of carbon dots in synthesis strategy. Coord. Chem. Rev.498, 215468 (2024). [Google Scholar]
  • 24.Li, S. et al. The development of carbon dots: From the perspective of materials chemistry. Mater. Today51, 188–207 (2021). [Google Scholar]
  • 25.Wareing, T. C., Gentile, P. & Phan, A. N. Biomass-based carbon dots: Current development and future perspectives. ACS Nano15, 15471–15501 (2021). [DOI] [PubMed] [Google Scholar]
  • 26.Petukhov, D. I. & Johnson, D. J. Membrane modification with carbon nanomaterials for fouling mitigation: A review. Adv. Colloid Interface327, 103140 (2024). [DOI] [PubMed] [Google Scholar]
  • 27.Yang, Z., Guo, H., Yao, Z., Mei, Y. & Tang, C. Y. Hydrophilic silver nanoparticles induce selective nanochannels in thin film nanocomposite polyamide membranes. Environ. Sci. Technol.53, 5301–5308 (2019). [DOI] [PubMed] [Google Scholar]
  • 28.Li, Y. et al. Graphene quantum dot engineered ultrathin loose polyamide nanofilms for high-performance nanofiltration. J. Mater. Chem. A8, 23930–23938 (2020). [Google Scholar]
  • 29.Peng, H. et al. Phosphonium modification leads to ultrapermeable antibacterial polyamide composite membranes with unreduced thickness. Adv. Mater.32, 2001383 (2020). [DOI] [PubMed] [Google Scholar]
  • 30.Mazhari, R., Bide, Y., Hosseini, S. S. & Shokrollahzadeh, S. Modification of polyacrylonitrile TFC-FO membrane by biowaste-derived hydrophilic N-doped carbon quantum dots for enhanced water desalination performance. Desalination565, 116888 (2023). [Google Scholar]
  • 31.Qin, X., Feng, X. & Hu, D. From asymmetry to symmetry: Biomimetic multi-channel structures for enhanced desalination using co-dissolving porogens. J. Membr. Sci.685, 121914 (2023). [Google Scholar]
  • 32.Song, X. et al. Bioinspired humic acid-based membranes for water desalination: Mechanistic insights from molecular simulations. Desalination571, 117092 (2024). [Google Scholar]
  • 33.Tao, J., Song, X., Bao, B., Zhao, S. & Liu, H. The role of surface wettability on water transport through membranes. Chem. Eng. Sci.219, 115602 (2020). [Google Scholar]
  • 34.Zhu, Z., Lv, S., Gao, Q., Zhao, S. & Lu, X. An analytical model for evaluating fluid flux across carbon-based membrane. J. Membr. Sci.644, 120157 (2022). [Google Scholar]
  • 35.Ma, Y., Wu, L., Ren, X., Zhang, Y. & Lu, S. Toward kilogram-scale preparation of full-color carbon dots by simply stirring at room temperature in air. Adv. Funct. Mater.33, 2305867 (2023). [Google Scholar]
  • 36.Xu, W., Zeng, F., Han, Q. & Peng, Z. Recent advancements of solid-state emissive carbon dots: A review. Coord. Chem. Rev.498, 215469 (2024). [Google Scholar]
  • 37.Lim, S. Y., Shen, W. & Gao, Z. Carbon quantum dots and their applications. Chem. Soc. Rev.44, 362–381 (2014). [DOI] [PubMed] [Google Scholar]
  • 38.Xue, Y.-R. et al. Harmonic amide bond density as a game-changer for deciphering the crosslinking puzzle of polyamide. Nat. Commun.15, 1539 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Yang, W.-J. et al. Carbon quantum dots (CQDs) nanofiltration membranes towards efficient biogas slurry valorization. Chem. Eng. J.385, 123993 (2020). [Google Scholar]
  • 40.Zheng, S. Z., Gissinger, J., Hsiao, B. S. & Wei, T. Interfacial polymerization of aromatic polyamide reverse osmosis membranes. ACS Appl. Mater. Interfaces16, 65677–65686 (2024). [DOI] [PubMed] [Google Scholar]
  • 41.Li, W. Y., Liu, X., Li, Z., Fane, A. G. & Deng, B. L. Unraveling the film-formation kinetics of interfacial polymerization via low coherence interferometry. AICHE J.66, e16863 (2020). [Google Scholar]
  • 42.Hu, X. et al. Versatile, aqueous soluble C2N quantum dots with enriched active edges and oxygenated groups. J. Am. Chem. Soc.142, 4621–4630 (2020). [DOI] [PubMed] [Google Scholar]
  • 43.Wang, L. et al. Water transport in reverse osmosis membranes is governed by pore flow, not a solution-diffusion mechanism. Sci. Adv.9, 8488 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Kotelyanskii, M. J., Wagner, N. J. & Paulaitis, M. E. Atomistic simulation of water and salt transport in the reverse osmosis membrane FT-30. J. Membr. Sci.139, 1–16 (1998). [Google Scholar]
  • 45.Zhang, X. J., Cahill, D. G., Coronell, O. & Mariñas, B. J. Absorption of water in the active layer of reverse osmosis membranes. J. Membr. Sci.331, 143–151 (2009). [Google Scholar]
  • 46.Zhang, C. et al. Molecular dynamics insights into water transport mechanisms in polyamide membranes: influence of cross-linking degree. J. Phys. Chem. B.129, 1697–1706 (2025). [DOI] [PubMed] [Google Scholar]
  • 47.Khoo, Y. S. et al. Eco-friendly surface modification approach to develop thin film nanocomposite membrane with improved desalination and antifouling properties. J. Adv. Res.36, 39–49 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Lau, W. J. et al. A review on polyamide thin film nanocomposite (TFN) membranes: History, applications, challenges and approaches. Water Res.80, 306–324 (2015). [DOI] [PubMed] [Google Scholar]
  • 49.Sarkar, P., Modak, S. & Karan, S. Ultraselective and highly permeable polyamide nanofilms for ionic and molecular nanofiltration. Adv. Funct. Mater.31, 2007054 (2021). [Google Scholar]
  • 50.Stolov, M. & Freger, V. Degradation of polyamide membranes exposed to chlorine: An impedance spectroscopy study. Environ. Sci. Technol.53, 2618–2625 (2019). [DOI] [PubMed] [Google Scholar]
  • 51.Verbeke, R., Gómez, V. & Vankelecom, I. F. J. Chlorine-resistance of reverse osmosis (RO) polyamide membranes. Prog. Polym. Sci.72, 1–15 (2017). [Google Scholar]
  • 52.Ji, Y.-L. et al. Bio-inspired fabrication of high perm-selectivity and anti-fouling membranes based on zwitterionic polyelectrolyte nanoparticles. J. Mater. Chem. A.4, 4224–4231 (2016). [Google Scholar]
  • 53.Yao, Y. et al. High performance polyester reverse osmosis desalination membrane with chlorine resistance. Nat. Sustain.4, 138–146 (2021). [Google Scholar]
  • 54.Yao, Y. et al. More resilient polyester membranes for high-performance reverse osmosis desalination. Science384, 333–338 (2024). [DOI] [PubMed] [Google Scholar]
  • 55.Gohil, J. M. & Suresh, A. K. Chlorine attack on reverse osmosis membranes: Mechanisms and mitigation strategies. J. Membr. Sci.541, 108–126 (2017). [Google Scholar]
  • 56.Wu, D. et al. Effects of chlorine exposure on nanofiltration performance of polyamide membranes. J. Membr. Sci.487, 256–270 (2015). [Google Scholar]
  • 57.Shao, F. et al. Layer-by-layer self-assembly TiO2 and graphene oxide on polyamide reverse osmosis membranes with improved membrane durability. Desalination423, 21–29 (2017). [Google Scholar]
  • 58.Guo, C. et al. Amino-rich carbon quantum dots ultrathin nanofiltration membranes by double “one-step” methods: Breaking through trade-off among separation, permeation and stability. Chem. Eng. J.404, 127144 (2021). [Google Scholar]
  • 59.Zhu, J. et al. Graphene quantum dot inlaid carbon nanofibers: Revealing the edge activity for ultrahigh rate pseudocapacitive energy storage. Energy Storage Mater.47, 158–166 (2022). [Google Scholar]
  • 60.Kolangare, I. M., Isloor, A. M., Inamuddin, Asiri, A. M. & Ismail, A. F. Improved desalination by polyamide membranes containing hydrophilic glutamine and glycine. Environ. Chem. Lett.17, 1053–1059 (2019). [Google Scholar]
  • 61.Akther, N. et al. Effect of graphene oxide quantum dots on the interfacial polymerization of a thin-film nanocomposite forward osmosis membrane: An experimental and molecular dynamics study. J. Membr. Sci.630, 119309 (2021). [Google Scholar]
  • 62.Xu, S. et al. Novel graphene quantum dots (GQDs)-incorporated thin film composite (TFC) membranes for forward osmosis (FO) desalination. Desalination451, 219–230 (2019). [Google Scholar]
  • 63.Dsilva Winfred Rufuss, D., Kapoor, V., Arulvel, S. & Davies, P. A. Advances in forward osmosis (FO) technology for enhanced efficiency and output: A critical review. J. Clean. Prod.356, 131769 (2022). [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information (3.1MB, pdf)
Transparent Peer Review file (393.2KB, pdf)
Source Data (3.3MB, xlsx)

Data Availability Statement

The data that supports the findings of the study are included in the main text and supplementary information file. Source data are provided with this paper or on Figshare (10.6084/m9.figshare.31081366). All data are available from the corresponding author upon request. Source data are provided with this paper.

Articles from Nature Communications are provided here courtesy of Nature Publishing Group

RESOURCES