Fast water transport and ionic sieving in ultrathin stacked nanoporous 2D membranes

ABSTRACT

Atomically thin nanoporous 2D membranes, featuring unique sieving characteristics for molecules and ions, have significant potential for seawater desalination. However, they face a common trade-off between permeability and selectivity. Here, we report an ultrathin stacked nanoporous graphene membrane (SNGM) created by layering atomically thin graphene nanomesh. This design achieves highly efficient and selective sieving of water molecules and ions. The SNGMs showcase in-plane nanopores for optimal size-exclusive water input and output, and interlayer 2D nanochannels between adjacent graphene nanomesh membranes for rapid water transport and precise ion/molecular sieving. The resulting SNGMs effectively address the trade-off between water permeability and ion selectivity in conventional desalination membranes, delivering a water permeability of ∼ 1–2 orders of magnitude higher than that of commercial membranes, while maintaining a comparable ion rejection ratio (>95% for NaCl). This advance marks a significant leap forward in adopting 2D nanoporous membranes for desalination technology.

A stacked nanoporous graphene membrane assembled by a few layers of atomically thin graphene nanomesh was constructed, achieving highly efficient sieving of water molecules/ions and resolving the permeability-selectivity trade-off in conventional membranes.

INTRODUCTION

Atomically thin 2D materials have been considered as ideal building blocks for constructing high-performance membranes for molecules and ionic sieving because of their unique physical and chemical stability, and have profound implications for water desalination applications [1–6]. Nanoporous 2D membranes mainly consist of atomically thin monolayer nanoporous membranes and lamellar nanoporous membranes [7–10]. The atomically thin monolayer nanoporous membranes were obtained by introducing nanopores in 2D materials, where the in-plane nanopores play a vital role in molecular sieving [11–14]. On the other hand, the lamellar nanoporous membranes assembled by small 2D nanosheets offer an alternative strategy to exploit the interlayer 2D nanochannels for molecular and ionic transport [15–21]. In either case, these nanoporous 2D membranes suffer a common trade-off between permeability and selectivity, limiting their applications in efficient water desalination [22–26].

The efficient sieving relies on the interaction between molecules/ions and nanochannels in the membranes [27,28]. Atomically thin nanoporous 2D membranes offer ultralow water transport resistance because of their unique single-atomic thickness characteristics [29–31]. Water molecules and ions could transport across the in-plane nanopores in the atomically thin membranes rapidly due to the negligible interaction between molecules/ions and membranes [32–35]. Thus, atomically thin nanoporous 2D membranes generally feature high water permeability and low ion selectivity [36,37]. Lamellar 2D membranes formed by assembling small 2D nanosheets with oxygen-, fluorine-containing functional groups provide tortuous transport pathways for ions, enhancing the interactions between ions and membrane to enable improved ion selectivity [38], but substantially reducing water permeability resulting from the increased tortuosity and interactions [39,40].

Here, we report the design of a stacked nanoporous graphene membrane (SNGM) assembled by a few layers of atomically thin large-area graphene nanomesh (Fig. 1). The constructed SNGMs feature high permeability and selectivity for efficient transport of water molecules while blocking ions. The nanopores in the atomically thin graphene nanomesh provide ultrashort nanochannels for water molecule transport, and the oxygen-containing groups at the edge of nanopores act as spacers to keep interlayer 2D nanochannels open for water molecule transport. The absence of oxygen-containing groups in the plane domains of graphene nanomesh offers a smooth 2D nanochannel for the achievement of decreased interaction between water molecules and 2D nanochannels, thus improving water permeability. With the size exclusion of the in-plane nanopores and the gating effect of the interlayer 2D nanochannels to synergistically tailor the molecular and ionic sieving properties, the SNGMs exhibit a high-water permeability of 77.8 L m^–2 h^–1 bar^–1, and an excellent monovalent salt ions rejection ratio (up to 95.2%). The SNGMs offer an efficient platform for probing the underlying molecules/ions transport behavior in nano-confined spaces, with significant potential to effectively address the common permeability–selectivity trade-off in conventional nanoporous membranes.

Figure 1.

Schematic illustration of the SNGMs for efficient molecular sieving. The dashed box shows the selective sieving mechanism of the SNGMs for molecules and ions.

RESULTS AND DISCUSSION

Design and structural characterizations of SNGMs

The SNGMs were formed by sequential stacking of atomically thin graphene nanomesh prepared by a mesoporous template combined with O₂ plasma exposure (Fig. S1). An aberration-corrected scanning transmission electron microscope (STEM) image of the pristine monolayer graphene shows a typical hexagonal honeycomb lattice structure composed of carbon atoms (Fig. S2). By comparison, the STEM image of the SNGM shows the presence of sub-nanometer pores, with an average pore size of ∼5 Å and a pore density of ∼3.0 × 10¹² cm⁻² (Fig. 2a; Figs S3 and S4). This pore size matches well with the optimized sub-nanometer pores for water molecules (∼3 Å) transport while effectively rejecting hydrated ions (>7 Å). The lamellar structure and layer numbers were observed in the high-resolution TEM images of the cross-section of SNGMs (Fig. 2b). The interlayer spacing of 2D nanochannel in the SNGMs is measured to be ∼4 Å, which is similar to previously reported 2D interlayer space that allows fast transport of one monolayer of water molecules [41]. The TEM characterizations suggest that the designed SNGMs possess optimized in-plane nanopores and 2D nanochannels for ultrafast water molecules transport and ion sieving.

Figure 2.

Structural and mechanical performance characterizations. (a) Aberration-corrected STEM image of the monolayer graphene nanomesh. The white dashed circles highlight the nanopores present in the monolayer graphene nanomesh. (b) Cross-sectional high-resolution TEM images of the monolayer graphene nanomesh, and SNGMs assembled by bilayer and trilayer graphene nanomesh. (c) FTIR spectra of pristine chemical vapor deposition graphene (CVDG), monolayer graphene nanomesh, and SNGMs assembled by bilayer and trilayer graphene nanomesh. (d) Load-displacement curves of monolayer graphene nanomesh, and SNGMs assembled by bilayer and trilayer graphene nanomesh. Inset: schematic illustration of the nanoindentation tests on suspended SNGMs. (e) The spatial distribution of the SNGMs at 100 nN force in MD simulations. The data were derived from MD simulations.

Raman spectroscopy was performed to analyze the number of layers and surface defects of the SNGMs (Fig. S5). The number of layers was identified according to the intensity ratio between the G (I_G) and 2D (I_2D) peaks (I_G/I_2D), and the intensity ratio between the D (I_D) and G (I_G) peaks (I_D/I_G) was utilized to investigate the surface defects, respectively. As shown in Fig. S5, with the increase of the layer number from monolayer to trilayer, the I_G/I_2D increases from 0.5 to 1.2 and the I_D/I_G increases from 0.2 to 0.5, suggesting that the sequential stacking of graphene nanomesh increases the defects in the SNGMs. Fourier-transformed infrared (FTIR) spectra and X-ray photoelectron spectra (XPS) characterizations show a strong characteristic absorption peak corresponding to hydroxyl groups (broad peak between ∼3200 and ∼3500 cm^–1) (Fig. 2c, Fig. S6 and Table S1), mainly attributed to oxygen-containing groups at the pore edges of SNGMs after O₂ plasma treatment. The studies suggest that the SNGMs with high-density sub-nanopores and abundant hydroxyl groups were successfully constructed.

The mechanical properties of the SNGMs were investigated by nanoindentation in an atomic force microscope (AFM) and molecular dynamics (MD) simulations. The SNGMs were suspended on a SiN substrate with arrays of apertures (diameter 1.0 μm) to construct a micro-testing system (Fig. S7). The typical force–displacement curves of the SNGMs assembled with different layers of graphene nanomesh were recorded by the AFM. As indicated in Fig. 2d, at the same indentation depth, the indentation forces of the SNGMs increase with the increase of stacked layer numbers of graphene nanomesh. Specifically, to achieve an indentation depth of 50 nm, the indentation force for the SNGMs assembled by bilayer graphene nanomesh is 848.6 nN, which is higher than that of monolayer graphene nanomesh. As the stacking layers increased from bilayer to trilayer, the required indentation force for the SNGMs increased by 31.9%. MD simulations reveal that the indentation force at the same indentation depth increases with an increasing number of graphene nanomesh layers (Figs S8 and S9), consistent with the nanoindentation tests (Fig. S10). The increase in force indicated an enhancement of resistance to external loads due to the increased interlayer interactions. For the monolayer graphene nanomesh, the in-plane strain distribution is highly localized at the geometry center of the corresponding layer, which is directly underneath the indenter tip. The localization of in-plane strain distribution is gradually weakened with an increasing number of graphene nanomesh layers (Fig. 2e and Figs S11-S13). Strain analysis revealed that strain shifting away from geometry centers is more prominent with increased interlayer interactions, thus resulting in a reduced strain concentration near the geometry center. These analyses indicate that the interlayer interactions including van der Waals and hydrogen bonding interactions in the SNGMs effectively improve the mechanical strength and structural stability of the SNGMs. The improved structural stability is beneficial to avoid the diminishing of ion selectivity caused by the rupture of SNGMs in the molecules/ions sieving processes.

Transport behavior of water molecules within SNGMs

MD simulations were conducted to explore the transport properties of water molecules within the SNGMs (Fig. 3a, Figs S14 and S15). Figure 3a shows the transmembrane transport pathways of water molecules in the SNGMs. Specifically, water molecules sequentially passed through the in-plane nanopores of the first-layer graphene nanomesh, and then entered the interlayer 2D nanochannels formed between the neighboring graphene nanomesh and finally flowed out of the SNGMs through the in-plane nanopores of the bottom-layer graphene nanomesh. The specific transport behaviors of water molecules in the in-plane nanopores and 2D nanochannels of the SNGMs were further analyzed (Fig. S16). The dynamic hydrogen bonds among water molecules in the bulk and between water molecules and oxygen-containing groups in the SNGMs (Fig. 3b–d) play a critical role in water molecule transport. When the water molecules transport from the bulk water into the in-plane nanopores, the average number of hydroxy bonds among water molecules dropped from 3.4 to 1.5, indicating that there were hydrogen bonds between water molecules and hydroxyl groups at the pore edge of graphene nanomesh (Fig. 3e, Figs S17 and S18). These hydrogen bonds increase the interaction between water molecules and the nanopores and lower the barrier for water molecules to enter the nanopores. This enhanced transport property is in agreement with previous reports that hydroxyl groups can attract water molecules into nanopores [42].

Figure 3.

Water transport behavior investigations. (a) A MD simulation snapshot of the water transport in the SNGMs assembled by trilayer graphene nanomesh. The z-direction indicated by the black arrow is perpendicular to the membrane. (b–d) Schematic illustrating the different hydrogen bonds patterns for water molecules in the bulk state (b), in-plane nanopore (c) and 2D channel (d) inside the SNGMs. (e) Hydrogen bonds distribution among water molecules in the SNGMs assembled by trilayer graphene nanomesh. Inset: schematic illustrating the state of water through the in-plane nanopore and 2D nanochannels.

When water molecules transport from the in-plane nanopores to 2D nanochannels in the SNGMs, the average number of hydroxy bonds among water molecules increases to ∼2.1, suggesting that there are interactions between water molecules within the 2D nanochannels, but remains much smaller than that of bulk water (3.4) (Fig. 3e and Figs S19-S22). Water molecules with a smaller number of hydrogen bonds usually lead to less viscous fluids, as the intermolecular forces are weaker under such conditions, causing less resistance to water flow [43]. This weak interaction allows the fast transport of water molecules (∼3 Å) inside the interlayer 2D nanochannels (interlayer spacing ∼4 Å) in the form of 2D monolayer, which is similar to the molecular superfluidity phenomenon in hydrophobic nano-confined spaces [34,44].

Together, the in-plane nanopore in the SNGMs is favorable for water molecules to enter the membranes through hydrogen bond interactions, and the weak interaction between the 2D nanochannels and water molecules allows ultrafast water permeation. This unique water molecule transport behavior is beneficial for improving the water permeability.

Water molecules’ transport velocity and water permeability

We further evaluated the water molecule transport velocities through the SNGMs by MD simulations. The transport velocities of water molecules through the in-plane nanopore and 2D nanochannels of the SNGMs with different layers of graphene nanomesh were calculated and are plotted in Fig. 4a–c. The transport velocities of water molecules slightly decrease with the increasing graphene nanomesh layers. An average transport velocity of 0.09 m s⁻¹ is achieved through the in-plane nanopores, while the water transport velocity through the 2D nanochannels is considerably faster (0.35 m s⁻¹) (Fig. 4d and Fig. S23). Thus, the water transport through the nanopore dictates the overall transmembrane transport. We have further evaluated the number of water molecules transported through the SNGMs and the corresponding water permeability with a different number of layers stacked graphene nanomesh (Fig. 4e, Figs S24 and S25). Overall, monolayer graphene nanomesh shows a maximum water flow rate of 16 waters ns⁻¹ due to the minimum transport resistance [45,46]. The sequential stacking of graphene nanomesh increases the water molecules transport pathway in the SNGMs and reduces the flow rate (Fig. 4e, inset); the water flow rate still maintains >8 waters ns⁻¹ with three layers of stacked graphene nanomesh.

Figure 4.

Water permeability evaluations. (a–d) Time-averaged water molecules transport velocity in the monolayer graphene nanomesh (a) and SNGMs assembled by bilayer graphene nanomesh (b) and trilayer graphene nanomesh (c and d) (x-z plane). (e) The transmembrane transport number of water molecules and water permeability of the SNGMs assembled with different layers of graphene nanomesh as a function of simulation time (t). Inset: the calculated water flow rate of the SNGMs. (f) Measured water permeability of the SNGMs assembled with different layers of graphene nanomesh. The water permeability was measured with a cross-flow filtration apparatus. (g) XRD showing the shifts of the peak corresponding to the interlayer distance of the 2D nanochannels in the SNGMs before and after 48 hours of tests in a water environment.

To further explore the water transport performance of the SNGMs, we constructed a reverse osmosis (RO) cross-flow filtration apparatus to experimentally determine the water permeability of the membrane (Figs S26-S32). The water permeability of the SNGMs assembled with different layers of graphene nanomesh was maintained at 77.8–117.0 L m^–2 h^–1 bar^–1 for pure water, and the measured values agree well with the theoretically calculated trend (Fig. 4f and Fig. S33). The above results demonstrate that a combination of high-density in-plane nanopores and ultrafast 2D nanochannels in the SNGMs ensures low transport resistance for efficient water molecule transport.

X-ray diffraction (XRD) studies of the SNGMs indicate that interlayer spacing only increases from 3.8 to 4.2 Å (Fig. 4g), suggesting a moderate swelling in the water environment [47], which is favorable for water transport and maintaining the structural stability of the SNGMs. The excellent structural stability of SNGMs is mainly attributed to the decreased number of oxygen-containing groups in the 2D nanochannels that avoids the insertion of water molecules into the interlayer nanochannels and the swelling of the SNGMs. The above results indicate that the constructed SNGMs feature ultrahigh water molecule transport velocity and high water permeability, and could maintain excellent structural stability after long-term operation.

Ionic sieving performance evaluations

A cross-flow RO filtration apparatus was constructed to investigate the molecular and ionic sieving performance of the SNGMs (Fig. 5a, Figs S26 and S27). In the chamber of the RO filtration apparatus, the direction of water permeation direction is perpendicular to the direction of water flow, thus the water will be driven by the external pumping pressure and hydrostatic pressure generated by the gravity of water together to overcome the osmotic pressure and pass through the membrane into the fresh water tank [48–50]. The salt ions rejection performances of the SNGMs assembled with different layers of graphene nanomesh were all determined to be >85% for monovalent/divalent salt ions (Fig. 5b). The measured salt rejections of monolayer membrane are 89.3% for Na⁺ (hydrated diameter ∼7.2 Å) and 86.5% for K⁺ (hydrated diameter ∼6.6 Å), respectively. The slight ion leakage is attributed to the presence of a small fraction of relatively large nanopores in the monolayer graphene nanomesh and intrinsic defects or cracks formed during the graphene growth or transfer processes.

Figure 5.

Evaluation of molecular and ionic sieving performance. (a) Schematic of the SNGMs for molecular and ionic sieving. (b) Rejection ratio of the SNGMs for KCl, NaCl, Na₂SO₄, MgCl₂, boron, rhodamine B (RhB) and tetracycline antibiotic (TC). The error bars indicate the standard deviation of the data acquired from three individual membranes. (c) The calculated water flow rate in NaCl solutions and NaCl rejection of the SNGMs with the pore size of 5 Å in MD simulations. (d) Long-term stability of the SNGMs assembled by trilayer graphene nanomesh conducted under cross-flow conditions. (e) Comparison of the water permeability and ions rejection performance of the SNGMs assembled with trilayer graphene nanomesh with state-of-the-art membranes reported in the literature. GO, graphene oxide; rGO, reduced graphene oxide; MOF, metal–organic framework; COF, covalent–organic framework; CPF, conjugated-polymer framework; Synder NFX and Synder NFW are commercial membranes.

The lamellar structure with stacked layer can effectively avoid the leakage through intrinsic defects or cracks in the monolayer SNGMs membrane. Additionally, the inclusion of interlayer 2D nanochannels in the SNGMs increases the interaction between salt ions and the membrane. The ions rejection performance of the SNGMs gradually increases with the stacking layers of graphene nanomesh. The rejection ratio of the bilayer and trilayer SNGMs exceeds 95.2% and 95.6%, respectively. Monolayer graphene nanomesh and the bilayer SNGMs exhibit a rejection ratio of 79.8% and 83.3% for boron (∼5.0 Å), respectively. Notably, the trilayer SNGMs exhibit a rejection ratio of >90.4% for boron. The measured selectivity of the SNGMs follows the order of Na₂SO₄ > MgCl₂ > NaCl > KCl > boron, which was positively correlated to the hydration diameter of ions. Additionally, the SNGMs show a high rejection ratio for small charged or neutral molecules (99.4% and 99.5% for rhodamine B and tetracycline antibiotic molecules, respectively) with solvated diameters of ∼10 Å (Fig. S34). This excellent sieving performance for organic molecules is difficult to achieve in typical commercial polymer membranes with poor chemical stability. It is worth mentioning that the water permeability of the SNGMs assembled with different layers of graphene nanomesh maintains 66.0–95.3 L m^–2 h^–1 bar^–1 in the salt solution or organic molecular solution compared with that of pure water (77.8–117.0 L m^–2 h^–1 bar^–1) (Figs S35-S37) and that the SNGMs exhibited an extremely low adsorption percentage (<0.5%; Fig. S38). Moreover, the water permeability of the SNGMs only decreased by 33.5% because of the increase of graphene nanomesh stacking layers, suggesting that the proposed lamellar structure improves the ionic sieving selectivity while retaining a high-water permeability.

A MD simulation model was constructed to explore the molecular and ionic sieving mechanism of the SNGMs. Figure 5c and Fig. S39 illustrate the sieving performance of SNGMs stacked by graphene nanomesh with pore sizes of 5 and 8 Å, assuming a fixed interlayer spacing of 4 Å. A 100% NaCl rejection performance is observed in monolayer graphene nanomesh with a pore size of 5 Å that is smaller than the diameter of hydration Na⁺ ions (>7 Å) (Fig. 5c), which reduces to 96.7% in an 8 Å pore size graphene nanomesh. This result indicates that the size exclusion effect of in-plane nanopores can effectively prevent salt ions from passing through the membrane. As the stacking of graphene nanomesh increased from monolayer to bilayer or trilayer, the SNGMs with in-plane pore sizes of 5 or 8 Å all show a prominent salt rejection of 100%, suggesting that the interlayer spacing in stacked 2D membranes can effectively improve salt rejection performance. The water molecule or ion transmembrane transport number decreases with the increase of stacked graphene nanomesh layers, which is consistent with the experimental results (Fig. 4f). The salt rejection showed a slight decrease when the pH was varied from 7 to 3 (the protonation of oxygen-containing groups at low pH values) (Fig. S40), suggesting that the surface-charge effect was involved in the sieving process. The above results indicate that the size exclusion effect of in-plane nanopores and interlayer 2D nanochannels and the gating effect in 2D nanochannels play a synergetic role in the molecular/ion sieving [51–56]. The fundamental reason for this is that the stacking of graphene nanomesh includes enormous 2D nanochannels in the SNGMs, resulting in increased ion transport pathway and interaction.

The long-term stability of the SNGMs was investigated under the same cross-flow velocity and hydraulic pressure conditions (Fig. 5d, Figs S41 and S42). The water permeability of the monolayer SNGM shows a slight increase from 98.4 to 120.3 L m^–2 h^–1 bar^–1 after 90 hours of measurement, while the ions rejection ratio decreased by 10.0%. By comparison, the bilayer and trilayer SNGMs retain relatively stable water permeability (69.0–77.3 L m^–2 h^–1 bar^–1) and salt rejection ratio (92.4–95.5%). Moreover, our studies of the trilayer SNGMs show that they maintain stable water permeability and salt rejection, even after several cycles of desalination/drying (Fig. S43). The increased water permeability of the monolayer SNGM is due to the intrinsic defects or cracks in the membrane being enlarged gradually after continuous operation and more ions being allowed to penetrate through the membrane. By contrast, the lamellar structure improves the mechanical strength of the SNGMs and allows the intrinsic defects in atomically thin graphene nanomesh to be shielded, ensuring long-term stability in water permeability and maintenance of the ions rejection performance. Similar water permeability and ions rejection performance were observed in more than 100 SNGMs. The SNGMs exhibit high resistance to bacterial attachment compared with the commercial cellulose triacetate membrane after long-term operation. The excellent anti-biofouling characteristic is mainly attributed to the high hydrophilicity of the SNGMs that hinders the attachment of bacteria (Fig. S44).

Compared with commercial RO membranes, the SNGMs exhibit ∼1–2 orders of magnitude higher water permeability (Fig. 5e and Table S2), while achieving a comparable ion rejection ratio (>95% for NaCl), which is higher than that of most graphene- or graphene oxide-based membranes, metal-organic framework-, covalent–organic framework- or conjugated-polymer framework-based membranes. These results demonstrate that the SNGMs exhibit a combination of water permeability, ion rejection ratio and structural stability, and thus effectively address the common trade-off between water permeability and ion selectivity in conventional desalination membranes.

CONCLUSION

We have constructed an SNGM consisting of a few layers of atomically thin graphene nanomesh featuring excellent molecular and ionic sieving performance. The van der Waals and hydrogen bonding interactions in the SNGMs effectively shield the defects in monolayer graphene nanomesh and result in greatly improved ion/molecular sieving properties and overall structural stability. The hydroxy groups in the in-plane nanopores could attract water molecules into the membranes, and the interlayer 2D nanochannels with smooth surfaces allow the ultrafast transport of water molecules in the form of monolayers, resulting in an optimized water permeability ∼1–2 orders of magnitude higher than that of commercial membranes. The size exclusion effect of in-plane nanopores and the gating effect of 2D nanochannels increase the interaction between ions and SNGMs, achieving precise ion sieving with selectivity of up to 97.2%. The SNGMs design addresses the fundamental permeability–selectivity trade-off in nanoporous membranes, thus promising attractive molecular/ionic separation membranes for the chemical, energy and biomedicine industries.

FUNDING

This work was supported by the National Key R&D Program of China (2021YFA1202400 and 2023YFF1205900), the National Natural Science Foundation of China (22322408, 21925401, 22474096 and 52221001), the Fundamental Research Funds for the Central Universities (2042022rc0004 and 2042024kf1015), the ‘Sharp Knife’ Technology Research Program of Hubei Province (2023BAA002), the Research Funds from the State Key Laboratory of Chemo/Biosensing and Chemometrics of Hunan University (20230755), and the New Cornerstone Science Foundation through the XPLORER PRIZE.

AUTHOR CONTRIBUTIONS

Y.Y., Q.Y. and K.L. proposed the research and supervised the project. J.W. and X.Z. designed and conducted the experiments and analyzed the experimental results. Q.L. and C.M. performed part of the structural characterization of the membranes. Z.Y. and Y.G. conducted simulation analysis on the structural and desalination process of the membranes. J.W. and Y.Y. cowrote the manuscript. Y.Y., Q.Y., K.L., J.W., X.Z. and Z.Y. discussed the results and commented on the manuscript.

Conflict of interest statement

None declared.