Friction-differentiated separation of alkali ions through two-dimensional nanochannels based on niobate perovskite

Ai, Xinyu; Zhu, Liu; Cui, FengLu; He, Weijun; Dong, Xinyao; Yang, Yang; Wu, Meiling; Peng, Yong; Wang, FengChao; Jiang, Zhongyi; Zhou, Kai-Ge

doi:10.1038/s41467-026-71579-6

Download PDF

Article
Open access
Published: 16 April 2026

Friction-differentiated separation of alkali ions through two-dimensional nanochannels based on niobate perovskite

Nature Communications volume 17, Article number: 3415 (2026) Cite this article

2581 Accesses
1 Citations
Metrics details

Subjects

Abstract

The separation of alkali ions holds profound fundamental significance and immense application potential, particularly driven by the increasing demand for lithium in the era of sustainable energy. Previous approaches relying on the static factor (e.g., size, charge, solubility, adsorption affinities) prove inadequate in the separation of alkali ions due to their highly similar physicochemical properties. By leveraging the difference of friction-an intrinsic factor that incorporates the effects of ion-media correlation with channel or interlayer species during the migration, this work presents a nanofluidic tribology strategy utilizing a platform based on two-dimensional niobate perovskite nanochannels to achieve alkali ions separation. By generating ~42% difference of transport friction at nanonewton-scale, our channels amplify the separation factor between Li⁺/Na⁺ and K⁺/Na⁺ up to 33.1 ~ 45.9, surpassing the state-of-the-art artificial nanochannels by over an order of magnitude. Tuning the channel structural symmetry from herringbone to zigzag type modulates friction-differentiation, achieving a Li⁺ recognized selectivity (Li⁺/K⁺ ~ 28.0). This work establishes a paradigm for separation science through the lens of friction-differentiation and will inspire more advancing technology to secure strategic alkali ion resources critical for a sustainable future.

Computational screening of sodium solid electrolytes through unsupervised learning

Article Open access 20 September 2024

Ion-dependent structure, dynamics, and allosteric coupling in a non-selective cation channel

Article Open access 28 October 2021

Nano silicated-FeAl₂O₄ functionalized by DL-alaninium nitrate ionic liquid (FeAl₂O₄-SiO₂@[DL-Ala][NO₃]) as versatile promotor for aqua-mediated synthesis of spiro[chromenopyrazole-indene-triones and spiro[chromenopyrazole-indoline-diones

Article Open access 15 July 2024

Introduction

Harnessing the selective ion transport through nanoconfined channels is paramount for advancing applications spanning chemical separations, energy storage, and pharmaceutical technologies¹. Among these applications, the precise discrimination of monovalent cations — particularly alkali ions (Li⁺, Na⁺, K⁺) — has emerged as a critical challenge for securing the strategical metal supply of sustainable energy². Previous selective membranes mostly rely on the difference in the static properties of ions, e.g., size³, dehydration barrier⁴ or Donnan repulsion⁵. However, the alkali ions have the same charge and valence, and moreover the differences in their hydrated ionic radii are within sub-ångström (24–51 pm)⁶, approaching the thermal fluctuations of ion channel width (50–80 pm) (e.g., K channel of streptomyces A (KcsA))⁷. Consequently, the existing approaches to designing a selective membrane fundamentally failed to distinguish the alkali ions with ultrahigh similarity, leading to the selectivity ratios beneath 10 for most of state-of-the-art artificial nanochannels⁸.

A potential yet underexplored strategy lies in differentiating the ion transport friction (Fig. 1a). This intrinsic factor is parallel to the flow direction, and therefore fundamentally governs the transport velocity⁹. By defining the channel width and wall interaction, other orthogonal dynamic interaction arisen from ion-media correlation^2,10,11 with channel or interlayer water collectively influence the ion diffusion, leading to differentiation of transport rates among diverse species¹². Moreover, periodic confinement via a nanochannel can exponentially amplify the mobility differences via diffusion mode transitions¹³. This demands nanochannels with atomic-level periodicity in both spatial confinement and physicochemical interaction with ion flow in order to define the transport friction at nanonewton (nN) scale. The emerging two-dimensional (2D) materials offer a promising platform for constructing such nanochannels, with sub-ångstrom control over confinement dimensions enabling friction impacts on transport¹⁴. Yet existing 2D materials—including graphene oxide and MXene-based channels—fall short due to disordered spatial distributions of functional groups (e.g., carboxyl, hydroxyl) that affect the ion transport via hard-soft acid-base (HSAB) interactions¹⁵. This structural randomness disrupts transport friction, underscoring the need for 2D architectures with atomically ordered functional groups to achieve alkali ion separation driven by friction differentiation.

**Fig. 1: The design of friction-differentiated nanochannels for alkali ion separation.**

To harness this friction-selectivity nexus, we present a nanofluidic tribology strategy using the 2D niobate perovskite K_4-xH_xNb₆O₁₇, a platform where corner/edge-sharing NbO₆ octahedra¹⁶ geometrically define the transport friction landscapes. As a growing family of perovskites, the two prerequisites of friction-differentiation can be fulfilled¹⁴. By assembling niobate perovskite layers with inverse or translational symmetry, niobate perovskite nanochannels exhibit uniform widths of 6–9 Å, matching the Debye length of alkali ions¹⁷. Moreover, two distinct ion channel geometries can be possibly engineered between herringbone and zigzag configurations, providing diverging symmetric HSAB interaction with alkali ions from oxygenate in NbO₆ octahedra (Fig. 1b)¹⁸. The periodicity of channel geometry and HSAB interactions within 2D niobate perovskite enables the precise definition of ion transport friction to uncover the nanofluidic tribology-determined selectivity, offering a pathway to surpassing the alkali ion separation accuracy of state-of-the-art artificial nanochannels.

Results

As the fundamentals of constructing the friction-differentiated channel based on 2D niobate perovskite, the monodisperse building blocks are firstly obtained by a microfluidic liquid-phase exfoliation (MLPE)¹⁹ of K_4-xH_xNb₆O₁₇ crystal (Supplementary Fig. 1a–d). Benefitting from the defect-free MLPE²⁰, over 80% of 2D nanosheets are below 2 nm in thickness, equivalent to a monodisperse bilayer of zigzag assembled niobate perovskite (m-KN) (Fig. 2a, b). Simultaneously, nearly 60% of m-KN is in a lateral size between 1 and 3 μm (Fig. 2c and Supplementary Fig. 1e). In contrast, the niobate perovskite nanosheet not subjected to the MLPE are polydisperse and thicker (Fig. 2b, c, Supplementary Fig. 1f, g), which will be negative for channel fabrication. The transmission electron microscopy (TEM) images of m-KN nanosheet shows an 2D orthorhombic cell along the (010) axis (a = 0.78 nm, c = 0.64 nm), consistent with the zigzag assembled bilayer of octahedral niobate (Fig. 2d inset and Supplementary Fig. 1h, i). Therefore, the m-KN nanosheet in zigzag assembly features monodispersed, thin, and large lateral size, providing an ideal building block with periodic nanoconfinement for the construction of a friction-differentiated ion channel.

**Fig. 2: Construction of friction-differentiated nanochannels based on 2D niobate perovskite with highly crystalline interlayer conduits symmetrically assembled by niobate group.**

Then, to define the channel geometry and symmetry, the 2D m-KN is assembled by vacuum filtration²¹, rendering a transparent membrane stable in water (Fig. 2e inset and Supplementary Fig. 2). Furthermore, no ripples and interlayer gaps can be found on the cross-section view of m-KN membranes, indicating the regulated formation of channels by the organized assembly of nanosheets (Fig. 2e). We characterized the thin film assembled from niobate perovskite nanosheets using TEM (Supplementary Fig. 3a). The selected-area electron diffraction (SAED) pattern of the m-KN assembly aligns well with the simulation results (Supplementary Fig. 3b-d). Therefore, the ion channels within the assembly exhibit herringbone and zigzag symmetries. The small-angle x-ray scattering (SAXS) analysis (Method) on the m-KN membrane demonstrates its long-range order and uniformity (Supplementary Fig. 4). Obtained by wide-angle X-ray scattering (WAXS) spectra (Method)²², Herman orientation factor (f) of (004) plane is up to 0.97 (Fig. 2f), indicating that the membrane overall possesses the periodicity of the ion channel to provide the symmetric niobate unit for the HSAB interaction with alkali ions. Among the existing 2D material-based membranes, the f-value of m-KN demonstrates leading orderliness (Supplementary Fig. 5). In contrast, the membrane fabricated by p-KN is glazing with the various ripples and interlayered gaps across the laminate, resulting in a significant swelling in water (Fig. 2g and Supplementary Fig. 6). Additionally, its Herman orientation factor of (004) plane decreases to 0.82 (Fig. 2h), indicating lower orderliness of the niobate groups, which introduces the fluctuation of the nanoconfinement and leads to poor control of friction-differentiation. The m-KN membrane exhibits more uniform mechanical properties and greater hardness, which aligns with its enhanced orderliness as revealed by WAXS. Additionally, the m-KN membrane demonstrates superior elastic recovery (~65%) compared to the p-KN membrane (~55%) after heavy loading, highlighting the impact of assembly orderliness on its flexibility (Supplementary Fig. 7). From the above analysis, the long-range orderliness of the zigzag and herringbone channels can be confirmed once again, indicating that the laminate built by 2D m-KN provides ordered ångström-scale conduits over long ranges to uncover the differentiation in ion transport friction.

To determine the friction-differentiated selectivity via niobate perovskite nanochannels, the cross-membrane permeation rate of various cations is experimentally monitored in an H-shaped device (Supplementary Fig. 8). For the highly crystalline m-KN membrane, the permeation rates for most ions range from 0.002 to 0.02 mol·m^–2·h^–1 (Fig. 3a), while Li⁺ and K⁺ exhibit much higher permeation rates of 2.02 and 2.85 mol·m^–2·h^–1, respectively, 1–2 orders of magnitude higher than the other cations (Al³⁺, Mg²⁺, Na⁺, Ca²⁺, Ba²⁺) and closed to 88% of the free diffusion (Supplementary Fig. 9). Consequently, the permselectivity of Li⁺/Na⁺ and K⁺/Na⁺ reaches 37.4 and 52.8, respectively, beyond the state-of-the-art records on alkali ions selective artificial nanochannels²³. The 24-hour stability test has been performed on m-KN (Supplementary Fig. 10). The Li⁺ permeation rate and Li⁺/Na⁺ and K⁺/Na⁺ selectivity showed no detectable decline, indicating ultrafast, selective Li⁺ transport and excellent stability of m-KN. Additionally, disturbing the ion channel by the intercalation of poly(styrenesulfonate) (PSS) rapidly decreases the K⁺/Na⁺ selectivity via m-KN membrane, also demonstrating the vital role of channel orderliness (Supplementary Fig. 11 and Supplementary Fig. 12). The anomalous selectivity of m-KN membrane cannot explained by the static properties of alkali ions, such as their hydration or ion radius differences, as seen in previous 2D material-based nanochannels (e.g., graphene oxide (GO), vermiculite, MoS₂ and MXene), which show preferential either Li⁺ or K⁺ prior permeation (Supplementary Fig. 9). Unlike typical mono-bivalent cation separation strategies, its ion diffusion shows no dependence on bare/hydrated cation charge density or dehydration energy (Supplementary Fig. 13a–d). The absorption capacitance does not dominate the selectivity either, since the ion exchange capacitance of niobate perovskite membrane follows the sequence of: Li⁺>Na⁺ ≈ K⁺ (Supplementary Fig. 13e). In contrast, the ion channel order is crucial, as shown by the reduced Li⁺/Na⁺ and K⁺/Na⁺ selectivity in the p-KN membrane (3.3 and 1.7, respectively, Supplementary Fig. 13f inset). Thus, the highly ordered crystalline m-KN membrane exhibits anomalous Li⁺ and K⁺ selectivity, which cannot be simply understood based on the above secondary influence’s vertical to the ion flow. Therefore, the differentiation of friction-a more fundamental factor is essential to be introduced into the following study on the selective ion transport via 2D niobate perovskite channels.

**Fig. 3: The alkali ions permselectivity and friction-differentiation via the nanochannel constructed by m-KN.**

Before elucidating the differentiation of the transport friction, the contributions of the two 2D niobate perovskite channels to the ion transport should be firstly justified, owing to the potential stereoselectivity between herringbone and zigzag conduits. In this respect, the aberration-corrected transmission electron microscopy (ACTEM) and X-ray diffraction (XRD) are utilized to characterize the ion channels in niobate perovskite membranes (Fig. 3b, c). Similar to the bulk phase, the ordered assembly of the NbO₆ layers in zigzag and herringbone symmetry can be observed in the m-KN membrane (Fig. 3b), offering two potential channels with periodic nanoconfinement for the alkali ion transport. XRD analysis reveals the basis for selective ion transport in m-KN membranes. Initial characterization shows key diffraction peaks corresponding to (020), (040), and (120) plane (Fig. 3c)¹⁷. Two distinct (040) peaks at 8.59° and 8.94° reflect herringbone (040)_H and zigzag (040)_Z channel geometries. With Li⁺ or K⁺ permeation, the (040)_H peak (green) partially shifts to 8.22°, indicating herringbone channel expansion while the zigzag channel remains stable (8.90°)¹⁷. Therefore, the herringbone channel in m-KN provides the transport pathway favorable towards Li⁺ and K⁺ and blocks the migration of Na⁺ on the contrary.

To uncover the friction-differentiation in the ion transport via herringbone channel, we carried out a molecular dynamic (MD) simulation (Method). By monitoring the trajectories of cations via herringbone channel (Fig. 3d, Source data are provided in Supplementary Data 8-10), Li⁺ mainly transports through the interlayered water layers with a friction less than 0.7 nN (Fig. 3e), similar to previous studies on Li⁺ transport via 2D confined ice^24,25. Meanwhile K⁺ and Li⁺ prefer to transport via the gap between channel wall and interlayered water (Fig. 3d) with the maximum friction of 0.42 and 0.64 nN, respectively, lower than that during the Na⁺ permeation (Fig. 3e). The transport friction exhibits a positive correlation with charge polarization between the media and ions, thus indicates that vertical HSAB interactions assist the ion in overcoming the transport friction barrier (Supplementary Fig. 20). Overall, the interlayered friction of the above cations increases in the order of K⁺˂Li⁺˂Na⁺ (Fig. 3f and Supplementary Fig. 14), consistent with decrease of experimental permeation rate. Notably, a 0.30 nN (42%) decrease of friction between K⁺ and Na⁺ leads to 52.8-fold increase in ion permeation rate of K⁺ over Na⁺, exceeding the linear relationship between diffusion rate and friction decay as described in the classical Stokes-Einstein equation²⁶. Unlike the linear deceleration of transport friction on mass diffusion in bulk solutions, the periodic structure of the nanochannel creates a potential well. Transport friction increases the possibility of ion capture by these wells, converting the diffusion model from underdamped to overdamped¹³. As a consequence, the impact of friction-differentiation at nN scale is amplified in the niobate perovskite nanochannel, offering an opportunity to differentiate the diffusion of alkali ions.

To experimentally evaluate the friction-differentiation in herringbone channel, the nanotribological study on the interface between niobate perovskite nanosheet and salt crystal is carried out by lateral force microscopy (LFM)²⁷. Niobate perovskite nanosheets are coated on the LFM tip (Supplementary Figs. 15, 16) to expose the herringbone channel wall, and alkali halide salt crystals introduce monovalent cations into the interlayer water (Fig. 3g). Measuring the lateral force between the perovskite and salt crystal simulates ion diffusion (Fig. 3h), with the friction obtained representing half of the friction during ion penetration through the herringbone capillary. As shown in the LFM, the friction between ion and channel increases in the order of K⁺˂Li⁺˂Na⁺ (Fig. 3i), consistent with simulation in Fig. 3f. Similarly, LFM tests are performed on vermiculite nanosheets (which also possess periodically arranged oxygen-containing functional groups), the inverse correlation between the alkali ion permeation rate of vermiculite membrane and LFM-quantified friction confirmed the reliability of the method (Supplementary Fig. 17). Therefore, the friction-differentiated herringbone channel in m-KN membrane provides the pathway with low friction for Li⁺ and K⁺ transport, in contrast to the mitigation of Na⁺ dragged by the higher resistance.

With regards to the impact of channel structure on the differentiation of transport friction, the trajectories and friction of cations via the zigzag channel in the m-KN based membrane are also studied by the above MD simulations. Although alkali ions often migrate with hydration shells in wider channels²⁸, the zigzag channel nonvolatile in exfoliation¹⁸ remains anhydrous, providing an ion transport pathway without interlayer water, similar to that in mica²⁹. In contrast to the low-friction alkali ions transport through the herringbone channel, the interlayer friction in the zigzag channel is significantly higher (Supplementary Fig. 14). Consequently, when the herringbone channel is open, ion transport through the zigzag conduit is more difficult in the m-KN membrane, as confirmed by the consistent peak at 8.90° after ion intercalation (Fig. 3c). Notably, the variation of the transport friction in the zigzag channel correlates positively with the HSAB interaction between media and ion (Supplementary Figs. 14 and 20). To leverage the friction-differentiation of zigzag channel, ion flow should be diverted from herringbone to zigzag channel. In MD models, a single graphene layer is intercalated into the herringbone channel (Supplementary Fig. 18). Notably the ion flow is mainly via the gap between graphene and niobate perovskite layer, uprising all transport frictions of alkali ions far beyond 3 nN. Meanwhile, the graphene blockage does not influence the friction-differentiation via the neighboring zigzag channel at nN scale (Fig. 4a). The transport friction increases in the order of Li⁺˂Na⁺≪K⁺ (Fig. 4b), consistent with those in the zigzag channel of m-KN membrane. Moreover, the Li⁺ transport friction in the zigzag channel (3.06 nN) is several nN lower than that of other alkali ions in both channels, indicating a preferential low-resistance pathway for Li⁺ diffusion. The graphene blockage modifies alkali ion friction in the herringbone channel at nN scale, selectively directing Li⁺ to the zigzag channel with lower resistance. Inspired by this model, GO under 7 Å thick is experimentally incorporated with the m-KN nanosheets to assemble membranes (Fig. 4c)³⁰. The nacre-like structure can be found in the hybrid membrane of m-KN and GO (KN-GO) as seen in the cross-section views of ACTEM and SEM, indicating the layer-by-layer assembly (Fig. 4c and Supplementary Fig. 19a, b). With the increase of GO mass ratio, the (004) peak corresponding to herringbone channel gradually decreases (Supplementary Fig. 19c). When the GO mass ratio exceeds 1:3, the diffraction between GO layers emerges at 10.31° and 10.69°, indicating maximum blockage of the herringbone channels.

**Fig. 4: The friction-differentiated alkali ions separation via the zigzag channel in the membrane based on m-KN and GO nacre-like structure.**

Arisen from the blockage of herringbone channel by GO, the friction-differentiated zigzag channel with lithium selectivity can be studied. Firstly, we monitor the Li⁺/K⁺ permselectivity to evaluate the blocking rate of herringbone channel (Fig. 4d), showing the lower friction to both Li⁺ and K⁺ transport on the contrary to the zigzag one (Fig. 4a, b). When the m-KN/GO mass ratio is 3, the Li⁺/K⁺ permselectivity reaches the maximum value (33.3), and the XRD peaks for the herringbone channel disappear, indicating that GO blocks most of the herringbone channel (Supplementary Fig. 19c). In contrast, other mass ratios show lower selectivity (Fig. 4d). Ratios above 3 leave some of herringbone channels are unblocked, showing low K⁺/Li⁺ selectivity similar to the results in m-KN membrane. However, ratios below 3 lead to nonselective ion channels built by spare GO, decreasing Li⁺/K⁺ permselectivity³¹. Therefore, the m-KN/GO mass ratio of 3 enables the construction of a quasi-layered KN-GO structure with blocked herringbone channels, leaving the zigzag channel dominating the ion transport. As a consequence, the nacre-like KN-GO membrane shows a singular lithium permselectivity (Fig. 4e), consistent with the friction-differentiation in zigzag channel (Fig. 4e inset). The characteristic expansion of zigzag channel in the KN-GO membranes after Li⁺ penetration also verifies its low-friction pathway (Fig. 4f). Therefore, directed by the understandings on stereoselective friction-differentiation, the singular recognition on the Li⁺ transport can be achieved by modulating the nacre-like KN-GO membrane.

Based on friction-differentiated transport and stereoselectivity, our niobate-based membranes are utilized to separate the binary mixtures between Li⁺, Na^+, or K⁺ (Method and Supplementary Fig. 21). For m-KN membrane (Fig. 5a), the separation factors of K⁺/Na⁺ and Li⁺/Na⁺ reach 45.9 ± 4 and 33.1 ± 2, respectively, enabling Na⁺ removal. For the KN-GO membrane (mass ratio of m-KN and GO = 3), the Li⁺/K⁺ separation factor is 28.0 ± 3, suitable for potassium removal, consistent with the above friction-differentiated permselectivity. The separation factors are also consistent for larger membranes in a diameter of 10 cm and 25 cm (Supplementary Fig. 22). Unlike most reports that calculate ideal permselectivity, our separation factor for binary mixtures offers a more practical reference for applications. The fluctuation of the above separation factors is under 20% (Fig. 5b), demonstrating excellent stability. Moreover, the ion permeability of our niobate-based membrane is 3–10 times higher than previous 2D material-based counterparts, approaching the ion diffusion level in bulk solution via microporous substrate (Supplementary Fig. 9). Consequently, the fast alkali ion separation surpasses the state-of-art trade-off limit between permeability-selectivity (Fig. 5c, d and Supplementary Fig. 23). The stability of membrane is also evaluated in more harsh conditions, e.g., acidic brines, and only 9.2% variation of permselectivity can be found after 5 cycles of acidic contamination in 120 hours. (Supplementary Fig. 24).

**Fig. 5: The separation of binary and trinary alkali ion mixtures by 2D niobate-based perovskite membranes.**

With above outstanding performance on selectivity and stability, we explored their potential of m-KN and KN-GO membranes in lithium purification from multicomponent mixtures, e.g., brine water. Herein, a simulated brine water containing Li⁺, Na⁺, K⁺, Mg^2+, and Ca²⁺ (the same as Salar de Atacama)^32,33 is used as the feed solution. The commercial nanofiltration membrane (MNF5) is used as the first stage to remove alkali ions, such as Mg²⁺ and Ca²⁺, however the separation among alkali ions cannot be achieved. To leverage the permselectivity of m-KN and KN-GO membranes, three-stages of m-KN membranes are then applied to remove Na⁺ followed by three-stages of KN-GO membranes to remove K⁺. After the above cascadic process, the lithium abundance can be lifted from 3.94 to 95.22% (Fig. 5f and Supplementary Fig. 25). These results demonstrate that our friction-differentiated niobate-based membrane offers a breakthrough beyond the permeability-selectivity trade-off, providing a practical solution for the alkali ions separation.

Discussion

This work has demonstrated a nanotribological strategy rooted in transport friction—a dynamic factor of nanofluidics to achieve alkali ion separation, one order of magnitude superior to the state-of-the-art artificial nanochannels. By leveraging its highly symmetric interlayer space affording HSAB interactions with the alkali ions, a friction-differentiated platform based on 2D niobate perovskite with two stereoscopic symmetric herringbone and zigzag channels, is constructed. By adjusting the channel symmetry, the nanofluidic tribology has been defined in nN scale to confer the friction-differentiation between alkali ions. The low transport friction pathway for K⁺ and Li⁺ emerges in the herringbone channel, acquiring a separation factor for the K⁺/Na⁺ and Li⁺/Na⁺ binary system up to 45.9 and 33.1, respectively. By sealing the herringbone channel with GO in KN-GO composite, the ion transport is directed to the zigzag channel, possessing a Li⁺ preferred pathway with lower friction and enabling a Li⁺/K⁺ binary separation factor of 28.0, in contrast to the Li⁺ and K⁺ preferential penetration through the pristine niobate membrane. Owing to the high selectivity driven by friction-differentiation, a lithium extraction from simulated brine water has been achieved by designing a cascade process combining the niobate membrane and its composite with GO. By breaking through the previous theories unilaterally relying on the static properties of ions, our friction-differentiation strategy can advance the artificial nanochannel design for highly efficient separation membranes and open an avenue towards the sustainable supply of critical metal ions.

Method

Synthesis of 2D Monodisperse Niobate Perovskite (m-KN) Nanosheets

All reagents were used as received without further purification. K₂CO₃, K₂MoO₄, and Nb₂O₅ (A.R., Meryer Co., Ltd.), N-propylamine, and tetrabutylammonium hydroxide (TBAOH, 12.5 wt% in 20 mL H₂O) (Sigma-Aldrich Co., Ltd.). K₄Nb₆O₁₇ crystal was synthesized using a molten salt method. Specifically, K₂CO₃, K₂MoO₄ and Nb₂O₅ (molar ratio is 2:19:3) were mixed evenly via milling. The mixture was placed in an alumina crucible and heated in a muffle furnace at a rate of 42.94 °C/min to 1100 °C, where it was held for 24 hours. Subsequently, the temperature was reduced to 500 °C at a rate of 4.8 °C/min, followed by natural cooling to obtain the K₄Nb₆O₁₇ crystal in centimeter scale (Supplementary Fig. 1b–d) obtained at room temperature. Then, 0.1 g crystal was immersed in 6 M HCl and shaken for 3 days followed by a hydrothermal intercalation with 2 mL n-propylamine and 20 mL DI water in a 100 mL Teflon beaker at 120 °C for 3 days. Likewise, the hydrothermal intercalation of niobate by TBAOH (12.5 wt% in 20 mL H₂O) was carried out to further expand the interlayered space, the p-KN was dispersed in water (Supplementary Fig. 1f), and the unexfoliated one was removed by a centrifugation at 110 x g (Multifuge X1R, ThemoFisher Inc.). In contrast, the microfluidization was carried out by a microfluidizer (NLM 100, IlshinAutoclave Inc.) at 1200 bar for 4 h to obtain the m-KN. Similar centrifugation at 110 x g was applied to remove the unexfoliated ones.

Synthesis of Graphene Oxide (GO)

Graphene oxide (GO) was synthesized from natural flake graphite using a modified Hummers method^34,35. Specifically, flake graphite (25 g, 0.3 mm, Qingdao Jinrilai Graphite Co., Ltd., China) and NaNO₃ (17.5 g, A.R.) were placed in a flask. Concentrated H₂SO₄ (650 mL) was added to the mixture under mechanical stirring in a water bath at room temperature. Subsequently, KMnO₄ (100 g, A.R.) was slowly added to the suspension to prevent a rapid rise in temperature. The mixture was then transferred to an oven and heated at 60 °C for 6 h to ensure complete oxidation. After the reaction, the mixture was poured into ice water, followed by the dropwise addition of H₂O₂ (30 wt%, ~20 mL) until the solution turned bright yellow. The resulting mixture was centrifuged and washed repeatedly with 5% HCl solution and deionized water to remove metal ions and acid residues until the pH reached neutral. Finally, the GO dispersion was obtained by ultrasonication and centrifugation (1000 ~ 11,000 x g) to remove unexfoliated graphite.

Fabrication of Highly Crystalline Membrane Based on m-KN

m-KN membranes were fabricated by vacuum filtration of corresponding dispersion through Nylon microporous substrate (0.22 μm in pore size, Jinteng Inc.). Similarly, to encapsulate the herringbone channel, the KN-GO membranes were fabricated by modulating the mass ratio between m-KN and GO. Additionally, PSS (M_w = 80k, A.R., Meryer Co. Lt.) was mixed with m-KN to fabricate KN-PSS membrane with disturbed herringbone channels by similar method for KN-GO. The thickness of above membranes were adjusted to 1 μm, and pre-testing vacuum drying at 60 °C was performed overnight to remove the residual water. m-KN nanosheets can fabricate large-area membranes without ripple (Supplementary Fig. 22a, b).

Characterizations on Membranes based on 2D Niobate Perovskite

The thickness and lateral size were in the preparation process of m-KN and p-KN nanosheet were obtained by using the tapping mode of atomic force microscopy (Dimension Icon, AFM, Bruker Inc., Tip: Nanosensors PPP-NCHR with tip force constant of 10-130 N/m). The atomic resolution images of the nanosheet and corresponding membrane were acquired using a field emission transmission electron microscope (Titan Themes Cubed G2 300, FEI Inc.). Additionally, the cross-section of membranes was characterized by a field emission scanning electron microscopy (Apreo S LoVac, Thermo Fisher Scientific Inc.). To evaluate the orderliness of lamellar, WAXS data were collected by Xeuss 2.0 laboratory beamline (Xenocs FR.) equipped with a Pilatus 3 R 300 K detector and Cu-α Source. The intercalation in the ion channel was monitored by the X-ray diffraction spectroscopy (Rigaku Smartlab 9KW). Ion concentration was determined using ion chromatography.

Influence of Nanosheet Thickness on the Orderliness in Crystalline Membrane Based on 2D Niobate Perovskite

WAXS experiments were performed using a Xenocs Xeuss SAXS/WAXS system equipped with a Cu-Kα radiation source. The X-ray beam was oriented either parallel or perpendicular to the plane of the sample sheets. Samples, with 10 mm in both length and width, were positioned at a distance of 163.5 mm from the detector. Scattering patterns were captured using a Pilatus 3 R 300k detector. The degree of alignment of the nanoplatelets within the samples was quantified using Herman’s orientation factor f, which is defined as follows³⁶:

$$f=\frac{3\left\langle {\cos }^{2}\varphi \right\rangle -1}{2}$$

(S1)

and

$$\left\langle {\cos }^{2}\varphi \right\rangle=\frac{{\int }^{\pi /2}_{0}I(\varphi ){\cos }^{2}\varphi \sin \varphi d\varphi }{{\int }^{\pi /2}_{0}I(\varphi )\sin \varphi d\varphi }\varphi$$

(S2)

where I(φ) is the intensity at a certain azimuthal angle, 〈cos²φ〉 is the average value of the square of the cosine of the azimuthal angle for the 040 peak of m-KN sheets. An increase in the value of f corresponds to a heightened degree of in-plane stacking order. The f value of m-KN membrane reaches 0.97 (Fig. 2f), equivalent to the previous crystalline lamellar (Supplementary Fig. 5). However, the membrane constructed by thicker p-KN nanosheet shows a f value lower than 0.82 (Fig. 2h), indicating the poor orderliness in the assembly.

In the SAXS measurements, the X-ray wavelength was 1.5 Å, with a sample-to-detector of 163.5 mm, providing a q-range of 0.2–2 Å^–1. Membrane samples are prepared as free-standing (thickness = 4 μm, diameter = 1.5 cm) and mounted in a vacuum chamber to minimize air scattering.

Membrane Permselectivity and Separation Tests

To evaluate the ideal permselectivity of the niobate perovskite membranes, permeation tests were conducted using a custom-designed H-shaped glass cell. The membrane was sandwiched between a pair of PET masks with a cycle window (~8 mm in diameter) to set the permeation area and sealed by epoxy glue (Stycast 1266, Henkel Loctite Inc.). Initially, the membrane was positioned at the center of a custom-designed H-shaped glass filter (H-cell, see Supplementary Figs. 8 and 21), which consists of a feeding chamber (FC) and a permeation chamber (PC), for the purpose of the permeation test. Subsequently, 0.2 M ion aqueous solution was placed on the side of the H-cell directly in contact with the membrane, while the other side was filled with an equal volume of DI water (resistivity: 18.2 MΩ), solution volumes were equal on the FC and PC sides. Stirring was employed to prevent concentration polarization. The permeation rate ($P$, mol·m^–2·h^–1) of alkali ions could be obtained by monitoring the slope of the ion concentration plotted against the time (t):

$$P=\frac{1}{A\cdot V}\frac{\partial {C}_{{{\rm{alkali}}}}}{\partial t}$$

(S3)

A represents the effective membrane area (m²); V represents the volume of the aqueous solution (L); ${C}_{{{\rm{alkali}}}}$, the concentration of alkali ion, was calculated from a standard curve between ${C}_{{alkali}}$ and the ion conductivity (σ) obtained by SevenMulti^TM dual meter (METTLER TOLEDO Inc.). Noting that the permeation test was carried out after keeping steady for 2 hours to avoid the fluctuation. Then, the ideal permselectivity (${\alpha }_{{{\rm{ideal}}}}$) equals the permeability ratio between two ions of A and B:

$${\alpha }_{{{\rm{ideal}}}}=\frac{{P}_{A}}{{P}_{B}}$$

(S4)

To evaluate the separation factor, separation tests of binary alkali ions mixtures were performed. To maintain the same ionic strength in the FC, a mixture of two alkali chloride aqueous solutions in an individual concentration of 0.1 M act as a feed, and the PC was filled with equivalent HCl to counteract the concentration polarization similar to the function of proton pump in KcsA³⁷. In the binary system, ion chromatography (Eco IC, Metrohm Inc.) was employed to determine the permeation rate following ASTM Standard D6919-17, and the calibration based on a standard sample was carried out before the test, as outlined below:

$${C}_{i}=\frac{{m}_{i}}{{M}_{i}}$$

(S5)

$${P}_{{M}_{i}}=\frac{\left({C}_{2}-{C}_{1}\right)*V}{A*\Delta t}$$

(S6)

${{P}}_{{M}_{i}}$ represents the permeation rate (mol·m^–2·h^–1); ${C}_{i}$ represents the molar concentration (mol·L^–1); ${M}_{i}$ represents the ion mass concentration (ppm), ${C}_{2}$ and ${C}_{1}$ represent the current and the previous; $V$ represents the volume of the aqueous solution (L); A represents the effective membrane area (m²); $\triangle t$ represents the interval of time. The binary system ion selectivity was defined as:

$${\alpha }_{{{\rm{binary}}}}=\frac{{P}_{{M}_{2}}}{{P}_{{M}_{1}}}$$

(S7)

To obtain the separation factors for every type of niobate-based membrane, three individual samples were parallelly examined.

The Extraction of Lithium from a Multi-component Ion Mixture

We add a lithium extraction experiment with the coexistence of more competition ions such as Na⁺, K⁺, Mg²⁺ and Ca²⁺ in the feed solution to simulate the real brine water of salt lake (Salar de Atacama)^32,33. The input concentrations of Li⁺, Na⁺, K⁺, Ca²⁺, and Mg²⁺ are 0.00046, 0.0928, 0.0205, 0.00173, and 0.8451 M as them in real brine water. To deal with such multicomponent ion mixture, a cascadic multistage separation process by diverse membranes is required. The ion concentration in the permeation side of each stage is measured by ion chromograph and the corresponding abundance of Li⁺ is calculated by Eq. (S8)

$${\varphi }_{{{\rm{Li}}}}=\frac{{C}_{{{\rm{Li}}}}}{\sum {C}_{{M}_{i}}}\times 100$$

(S8)

${C}_{{{\rm{Li}}}}$ represents the molar concentration of lithium; $\sum {C}_{{M}_{i}}$ represents the sum of the molar concentrations of all partial ions; ${\varphi }_{{{\rm{Li}}}}$ represents the abundance of lithium. The commercial nanofiltration membrane (MNF5, DuPont Inc.) is used as the first one stage to remove alkali ions, such as Mg²⁺ and most Ca²⁺, however the separation among alkali ions cannot be achieved. Therefore, the membrane based on niobate perovskite can demonstrate their unique functions. To leverage the permselectivity of m-KN and KN-GO membranes, three-stages of m-KN membranes are applied to remove Na⁺ followed by three-stages of KN-GO membranes to remove K⁺. The lithium abundance during the cascadic separation process is shown in Fig. 5f and Supplementary Table S2.

The Ion Absorption of m-KN Membrane

To evaluate the adsorption effect, the m-KN membrane (6.9 mg) immersed in a 40 mL solution of 0.2 M alkali chloride for 72 hours to achieve adsorption equilibrium. Prior to immersion, the mass concentration of the mixed solution was quantified using ion chromatography, denoted as ${C}_{{M}_{i}\,}$. Following the soaking period, the solution was centrifuged at 11,000 x g for 30 min, and the mass concentration of the supernatant was measured again, denoted as ${C}_{i}.$ The absorbing rate of alkali ions ($\alpha$) was calculated by the following equation:

$$\alpha=\frac{{C}_{{M}_{i}}-{C}_{i}}{{C}_{{M}_{i}}}\times 100\%$$

(S8)

${M}_{i}$ denotes the following three ions: Li⁺, Na⁺, K⁺

The higher value of $\alpha$ indicates stronger absorption of alkali ions with niobate. The lithium and sodium absorption with niobate is significantly higher than that of potassium (Supplementary Fig. 13e), inconsistent with any selectivity of the niobate-based membrane. Therefore, the permselectivity in this work is not determined by the absorption.

Simulation Details on the Interlayered Friction

To identify the model of niobate perovskite-based membrane^18,38, the density functional theory (DFT) was utilized based on projector augmented wave (PAW) method with Plane-Wave basis set using the VASP software (VASP.5.4.4.)³⁹. The niobate laminate with the herringbone channels protonated was applied to simulate the m-KN membrane, and a supercell of 2.6 nm × 4.2 nm × 3.5 nm acts as the ion conduit placed in a box of 2.6 nm × 4.2 nm × 16 nm filled with 3500 water molecules (Fig. 3d and Supplementary Fig. 14a and 18a). Then, all unsaturated oxygen was protonated to form the hydroxyl group in the herringbone channel and conduit terminals. Then, the configuration of all groups correlated to the transport, e.g., the interlayered and terminal oxygenated groups, cations and water were fully optimized, and the convergence condition was set to 10^–2 eV/ Å. Noting that the water molecules prefer to intercalate into the herringbone channel and form a tri-layer to offer the ion transport pathway. Similar DFT optimization was carried out for the model of KN-GO membrane, wherein the herringbone channel was widened from 6 Å to 11 Å to allow the intercalation of a graphene single layer (Fig. 4a and Supplementary Fig. 18a). After relaxation in a box of 2.6 nm × 4.9 nm ×16 nm filled with 4000 H₂O, the water molecules stay out of either zigzag channel or herringbone channel with graphene intercalation in contrast to the herringbone channel in m-KN membrane. Therefore, the type of channel and difference on the interlayered water should be considered in the simulation on the alkali ions selective transport via niobate-based membrane.

To ensemble the effect of channel and interlayered water, molecular dynamic simulations were conducted to map the spatial distributed friction exerted on alkali ions across the aforementioned niobate-based channel. We simplified the driving force on ions due to the concentration gradient by applying a spring with a moving speed of 10^–4 Å/fs and an elastic coefficient of 7 × 10³ nN/Å on the penetrating alkali ions. For ions moving at a constant speed, the friction force can be obtained by the elastic force of the spring. According to Fokker-Planck equation⁴⁰, lower friction results into the higher flux under the same driving gradient, therefore the evaluation of interlayered friction can be used to understand the origin of alkali ions selectivity and assign the transport pathway depending on the type of alkali ions, channels and graphene intercalation. 12-6 Lennard-Jones potential coupled with a polar term was used to model the atomic interactions, which comprehensively considers both van der Waals and Coulombic forces. The cutoff distance for the van der Waals and short-range electrostatic forces was set to be 12 Å. The long-range Coulombic interactions are calculated using particle-particle-particle mesh solver⁴¹, which evaluates Fourier series directly with a desired relative error in forces of 1 × 10⁻⁴. Simulations were performed in the canonical (NVT) ensembles (350 K) using a Nosé–Hoover thermostat. Periodic boundary conditions are applied in the direction perpendicular to ion transport, while non-periodic boundaries are used in the direction parallel to ion transport. The universal force field⁴², AIREBO potential⁴³ and TIP3P model⁴⁴ are used to describe the niobate, graphene and water molecules, respectively. The force field parameters are listed in Supplementary Table S3. All MD simulations were implemented by LAMMPS⁴⁵.

We employed the Vienna Ab-initio Simulation Package (VASP)^46,47 to investigate interfacial electron transfer. Electronic structure calculations were performed using the projector augmented wave (PAW) method within the generalized gradient approximation (GGA) as parameterized by the Perdew-Burke-Ernzerhof (PBE) functional^48,49. A plane-wave energy cutoff of 520 eV was adopted. The geometry optimization was considered convergent when the energy change between electronic steps was below 10⁻⁵ eV and the forces on all atoms were less than 0.02 eV/Å. For both structural optimization and factor calculations, k-point mesh of 2 × 2 × 2 was used. The differential charge density and Bader charge calculation results are shown in Supplementary Fig. 20.

Data availability

All data supporting the findings of this study are available within the article and the Supplementary Information file. Source data are provided with this paper. All the raw data relevant to the study are available from the corresponding author upon request. Source data are provided with this paper.

References

Shen, J., Liu, G., Han, Y. & Jin, W. Artificial channels for confined mass transport at the sub-nanometre scale. Nat. Rev. Mater. 6, 294–312 (2021).
Article ADS CAS Google Scholar
Razmjou, A. et al. Design principles of ion selective nanostructured membranes for the extraction of lithium ions. Nat. Commun. 10, 5793 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
Hong, S. et al. Precision ion separation via self-assembled channels. Nat. Commun. 15, 3160 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Zhou, X. et al. Intrapore energy barriers govern ion transport and selectivity of desalination membranes. Sci. Adv. 6, eabd9045 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Foo, Z. H. et al. Lithium Concentration from Salt-Lake Brine by Donnan-Enhanced Nanofiltration. Environ. Sci. Technol. 57, 6320–6330 (2023).
Article ADS CAS PubMed Google Scholar
Nightingale, E. R. Jr. Phenomenological Theory of Ion Solvation. Effective Radii of Hydrated Ions. J. Phys. Chem. 63, 1381–1387 (1959).
Article ADS CAS Google Scholar
Marbach, S., Dean, D. S. & Bocquet, L. Transport and dispersion across wiggling nanopores. Nat. Phys. 14, 1108–1113 (2018).
Article CAS Google Scholar
Yu, X. & Ren, W. Ion and Water Transport in 2D Nanofluidic Channels. Adv. Funct. Mater. 34, 2313968 (2024).
Article CAS Google Scholar
Einstein, A. Über die von der molekularkinetischen Theorie der Wärme geforderte Bewegung von in ruhenden Flüssigkeiten suspendierten Teilchen. Ann. Phys. 322, 549–560 (1905).
Article Google Scholar
Oh, S. et al. Differential ion dehydration energetics explains selectivity in the non-canonical lysosomal K⁺ channel TMEM175. Elife 11, e75122 (2022).
Article CAS PubMed PubMed Central Google Scholar
Zhang, J. et al. Mechanism of lithium ion selectivity through membranes: a brief review. Environ. Sci. -Wat. Res. 10, 1305–1318 (2024).
CAS Google Scholar
Ma, M. et al. Water transport inside carbon nanotubes mediated by phonon-induced oscillating friction. Nat. Nanotechnol. 10, 692–695 (2015).
Article ADS CAS PubMed Google Scholar
Li, C., Huang, B., Cao, L. & Li, Z. Molecular diffusion on surfaces in the weak friction limit. J. Appl. Phys. 115, 214906 (2014).
Article ADS Google Scholar
Leng, K. et al. From bulk to molecularly thin hybrid perovskites. Nat. Rev. Mater. 5, 482–500 (2020).
Article ADS CAS Google Scholar
Lee, H. et al. Transformation of K₂Sb₈Q₁₃ and KSb₅Q₈ Bulk Crystals to Sb₂Q₃ (Q = S, Se) Nanofibers by Acid–Base Solution Chemistry. J. Am. Chem. Soc. 145, 15951–15962 (2023).
Article ADS CAS PubMed Google Scholar
Sarahan, M. C. et al. K₄Nb₆O₁₇-derived photocatalysts for hydrogen evolution from water: nanoscrolls versus nanosheets. J. Solid State Chem. 181, 1678–1683 (2008).
Article ADS CAS Google Scholar
Abe, R. et al. Preparation of Ion-Exchangeable Thin Films of Layered Niobate K₄Nb₆O₁₇. Chem. Mater. 10, 1647–1651 (1998).
Article CAS Google Scholar
Bizeto, M. A. et al. Intralamellar structural modifications related to the proton exchanging in K₄Nb₆O₁₇ layered phase. J. Phys. Chem. Solids 71, 560–564 (2010).
Article ADS CAS Google Scholar
Miyamoto, N. & Nakato, T. Liquid crystalline nanosheet colloids with controlled particle size obtained by exfoliating single crystal of layered niobate K₄Nb₆O₁₇. J. Phys. Chem. B 108, 6152–6159 (2004).
Article CAS PubMed Google Scholar
Whitesides, G. M. The origins and the future of microfluidics. Nature 442, 368–373 (2006).
Article ADS CAS PubMed Google Scholar
Abraham, J. et al. Tunable sieving of ions using graphene oxide membranes. Nat. Nanotechnol. 12, 546–550 (2017).
Article ADS CAS PubMed Google Scholar
Yang, J. et al. Water-induced strong isotropic MXene-bridged graphene sheets for electrochemical energy storage. Science 383, 771–777 (2024).
Article ADS CAS PubMed Google Scholar
Huang, L. et al. Shearing Liquid-Crystalline MXene into Lamellar Membranes with Super-Aligned Nanochannels for Ion Sieving. Angew. Chem. Int. Ed. 63, e202314638 (2024).
Article CAS Google Scholar
Han, D. et al. Ultrahigh Lithium Selective Transport in Two-Dimensional Confined Ice. J. Phys. Chem. Lett. 15, 2375–2383 (2024).
Article CAS PubMed Google Scholar
Qiu, H., Xue, M., Shen, C. & Guo, W. Anomalous cation diffusion in salt-doped confined bilayer ice. Nanoscale 10, 8962–8968 (2018).
Article CAS PubMed Google Scholar
Edward, J. T. Molecular volumes and the Stokes-Einstein equation. J. Chem. Educ. 47, 261–270 (1970).
Article CAS Google Scholar
Lizée, M. et al. Anomalous friction of supercooled glycerol on mica. Nat. Commun. 15, 6129 (2024).
Article ADS PubMed PubMed Central Google Scholar
Abdollahzadeh, M. et al. Designing Angstrom-Scale Asymmetric MOF-on-MOF Cavities for High Monovalent Ion Selectivity. Adv. Mater. 34, 2107878 (2022).
Article CAS Google Scholar
Zou, Y.-C. et al. Ion exchange in atomically thin clays and micas. Nat. Mater. 20, 1677–1682 (2021).
Article ADS CAS PubMed Google Scholar
Pacilé, D. et al. Electronic properties and atomic structure of graphene oxide membranes. Carbon 49, 966–972 (2011).
Article Google Scholar
Lv, Y. et al. Tailoring Monovalent Ion Sieving in Graphene-Oxide Membranes with High Flux by Rationally Intercalating Crown Ethers. ACS Appl. Mater. Interfaces 15, 46261–46268 (2023).
Article CAS PubMed Google Scholar
Xiang, J. et al. Satellite-Based Lithium Capacity Monitoring in Salt Lakes: The Atacama Case. Sustainability 17, 5631 (2025).
Article ADS CAS Google Scholar
Munk, L. et al. Lithium Brines: A Global Perspective. Econ. Geol. 18, 339–365 (2016).
Hummers, W. S. Jr. & Offeman, R. E. Preparation of Graphitic Oxide. J. Am. Chem. Soc. 80, 1339–1339 (1958).
Article ADS CAS Google Scholar
Ge, Z. et al. A Universal Method for the Preparation of Dual Network Reduced Graphene Oxide–Ceramic/Metal Foam Materials with Tunable Porosity and Improved Conductivity. Chem. Mater. 30, 8368–8374 (2018).
Article CAS Google Scholar
Wan, S. et al. High-strength scalable graphene sheets by freezing stretch-induced alignment. Nat. Mater. 20, 624–631 (2021).
Article ADS CAS PubMed Google Scholar
Doyle, D. A. et al. The structure of the potassium channel: molecular basis of K⁺ conduction and selectivity. Science 280, 69–77 (1998).
Article ADS CAS PubMed Google Scholar
Kinomura, N. et al. Preparation of a new tungsten(V) phosphate and its polymorph. J. Solid State Chem. 77, 156–161 (1988).
Article ADS CAS Google Scholar
Kresse, G. & Joubert, D. From Ultrasoft Pseudopotentials to the Projector Augmented-Wave Method. Phys. Rev. B 59, 1758 (1999).
Article ADS CAS Google Scholar
Denisov, S. I., Horsthemke, W. & Hänggi, P. Generalized Fokker-Planck equation: Derivation and exact solutions. Eur. Phys. J. B 68, 567–575 (2009).
Article ADS MathSciNet CAS Google Scholar
Kelkar, M. S. et al. Prediction of viscosities and vapor–liquid equilibria for five polyhydric alcohols by molecular simulation. Fluid Phase Equilib. 260, 218–231 (2007).
Article CAS Google Scholar
Rappe, A. K. et al. UFF, a full periodic table force field for molecular mechanics and molecular dynamics simulations. J. Am. Chem. Soc. 114, 10024–10035 (1992).
Article ADS CAS Google Scholar
Stuart, S. J., Tutein, A. B. & Harrison, J. A. A reactive potential for hydrocarbons with intermolecular interactions. J. Chem. Phys. 112, 6472–6486 (2000).
Article ADS CAS Google Scholar
Price, D. J. & Brooks, C. L. 3rd A modified TIP3P water potential for simulation with Ewald summation. J. Chem. Phys. 121, 10096–10103 (2004).
Article ADS CAS PubMed Google Scholar
Plimpton, S. Fast Parallel Algorithms for Short-Range Molecular Dynamics. J. Comput. Phys. 117, 1–19 (1995).
Article ADS CAS Google Scholar
Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 54, 11169–11186 (1996).
Article ADS CAS Google Scholar
Kresse, G. & Furthmüller, J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comput. Mater. Sci. 6, 15–50 (1996).
Article CAS Google Scholar
Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 77, 3865–3868 (1996).
Article ADS CAS PubMed Google Scholar
Jiang, J. et al. Atomic-level insight into super-efficient electrocatalytic oxygen evolution on iron and vanadium co-doped nickel (oxy)hydroxide. Nat. Commun. 9, 2885 (2018).
Article ADS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

This work was supported by the National Natural Science Foundation of China (No. 22278302: K.-G.Z., No. 22338011: Z.Y.J., No. 12241203&12388101: F.C.W., No. 12374011: Y.P.), National Key Research and Development Program of China (2022YFC2903504: Y.P.), Haihe Laboratory of Sustainable Chemical Transformations (No. 24HHWCSS00009: K.-G.Z.), and the Fundamental Research Funds for the Central Universities, (No. WK2090000087: F.C.W.). We also thanks Prof. Yongan Yang in Tianjin University for his assistances on characterization.

Author information

These authors contributed equally: Xinyu Ai, Liu Zhu, FengLu Cui.

Authors and Affiliations

Institute of Molecular Plus, Department of Chemistry, Tianjin University, Nankai district, Tianjin, China
Xinyu Ai, Weijun He, Xinyao Dong, Meiling Wu & Kai-Ge Zhou
State Key Laboratory of Synthetic Biology, Tianjin University, Tianjin, China
Xinyu Ai & Zhongyi Jiang
Haihe Laboratory of Sustainable Chemical Transformations, Tianjin, China
Xinyu Ai, Weijun He, Xinyao Dong, Meiling Wu & Kai-Ge Zhou
School of Materials and Energy, Electron Microscopy Centre, Lanzhou University, Lanzhou, China
Liu Zhu & Yong Peng
Department of Modern Mechanics, University of Science and Technology of China, Hefei, China
FengLu Cui & FengChao Wang
Institute of Seawater Desalination and Multipurpose Utilization, Ministry of Natural Resources (Tianjin), Tianjin, China
Yang Yang
School of Chemical Engineering and Technology, Tianjin University, Nankai district, Tianjin, China
Zhongyi Jiang

Authors

Xinyu Ai
View author publications
Search author on:PubMed Google Scholar
Liu Zhu
View author publications
Search author on:PubMed Google Scholar
FengLu Cui
View author publications
Search author on:PubMed Google Scholar
Weijun He
View author publications
Search author on:PubMed Google Scholar
Xinyao Dong
View author publications
Search author on:PubMed Google Scholar
Yang Yang
View author publications
Search author on:PubMed Google Scholar
Meiling Wu
View author publications
Search author on:PubMed Google Scholar
Yong Peng
View author publications
Search author on:PubMed Google Scholar
FengChao Wang
View author publications
Search author on:PubMed Google Scholar
Zhongyi Jiang
View author publications
Search author on:PubMed Google Scholar
Kai-Ge Zhou
View author publications
Search author on:PubMed Google Scholar

Contributions

K.-G.Z. and Z.Y.J. initiated and supervised the project. X.Y.A. performed the experiment and analyzed the data with help from K.-G.Z.; L.Z. and Y.P. carried out the ACTEM; F.L.C. and F.C.W. performed theoretical modelling and simulations. Y.Y. carried out the WAXS. W.J.H. and X.Y.D. helped in ion chromatography test. M.L.W., K.-G.Z., and Z.Y.J. participated in the writing process and provided technical editing of the final manuscript. All authors commented on the final manuscript.

Corresponding authors

Correspondence to FengChao Wang, Zhongyi Jiang or Kai-Ge Zhou.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Wooyoung Shim and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

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

Supplementary information

Supplementary Information (download PDF )

Description of Additional Supplementary Files (download PDF )

Supplementary Data 1 (download XLSX )

Supplementary Data 2 (download TXT )

Supplementary Data 3 (download TXT )

Supplementary Data 4 (download TXT )

Supplementary Data 5 (download TXT )

Supplementary Data 6 (download TXT )

Supplementary Data 7 (download TXT )

Supplementary Data 8 (download TXT )

Supplementary Data 9 (download TXT )

Supplementary Data 10 (download TXT )

Supplementary Data 11 (download TXT )

Supplementary Data 12 (download TXT )

Supplementary Data 13 (download TXT )

Transparent Peer Review file (download PDF )

Source data

Source Data (download XLSX )

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Ai, X., Zhu, L., Cui, F. et al. Friction-differentiated separation of alkali ions through two-dimensional nanochannels based on niobate perovskite. Nat Commun 17, 3415 (2026). https://doi.org/10.1038/s41467-026-71579-6

Download citation

Received: 02 June 2025
Accepted: 22 March 2026
Published: 16 April 2026
Version of record: 16 April 2026
DOI: https://doi.org/10.1038/s41467-026-71579-6