Simultaneous voltage-gated control of ion and water transport in Zr₄-Ti₃C₂T_x nanochannel membranes

Abstract

Controlling water and ion transport across nanoconfined channels is essential for natural biological processes and crucial for breakthroughs in diverse scientific and technological fields. Here, we present an efficient voltage-controlled strategy that simultaneously regulates water and ion diffusion by fine-tuning the external voltage applied to a high-conductivity Zr₄-Ti₃C₂T_x nanochannel membrane, which demonstrates high structural stability in aqueous environments. Under positive voltage, ion permeation increased by a factor of 10.18, whereas negative voltage reduced it to 0.17 of its original value. Interestingly, water diffusion exhibited the opposite response, with negative voltage enhancing water transport due to the facilitated rotation motion of nanoconfined water with the increased interfacial hydrogen bonding. This distinct voltage-gated transport behavior provides a potential solution to the longstanding trade-off between permeation and selectivity in membrane separation. In desalination trials, applying negative voltage improved ion rejection from 72.09 % to 98.57 % and doubled water permeation. Additionally, in lithium concentration applications, our approach enabled simultaneous improvements in water permeation and Li⁺ rejection. Our findings open promising pathways for advancements in energy, resource, and environmental applications.

Introduction

Ion transport through transmembrane proteins is essential to key biological functions, such as ATP synthesis and electrical signaling within the nervous system^1,2. The confined nanoscale channels within these proteins are crucial for ion migration. In response to specific environmental stimuli, biological ion channels can discriminate ions with high precision and regulate their flow, a phenomenon known as gating^3,4. For instance, potassium-selective channels allow potassium ions to diffuse thousands of times faster than sodium ions when in the open state^5,6. Understanding how protein nanochannels achieve these complex properties could provide valuable insights for technological breakthroughs across resource, energy, and environmental applications.

With significant advancements in nanotechnology, various artificial nanopores or nanochannels that mimic biological protein channels have been fabricated to investigate ion transport mechanisms under confinement^7,8. To date, an array of anomalous transport phenomena, which deviate from classical continuum expectations, has emerged. These deviations are primarily attributed to the increasing importance of surface effects as the surface-to-volume ratio escalates in extremely confined channels. Specifically, when channel dimensions enter the nanoscale, their sizes become comparable to various relevant length scales associated with surface effects, including molecular length (0.1–1 nm), electrostatic length (1–100 nm), and slip length (0.1–50 nm)^9,10,11,12. For instance, in the context of electrostatic interactions, when the diameter of a nanochannel is smaller than the range of Coulomb forces exerted by charged channel walls, a unipolar ionic distribution arises, characterized by a high enrichment of counter ions^10,11. Huang et al. demonstrated that surface charges can completely govern ionic conductance in these confined channels, which appears to be independent of bulk ionic concentration; this phenomenon is referred to as surface-charge-governed transport¹³. Through careful structural design of the confined nanochannels, this distinctive property facilitates unconventional ion transport behaviors, including ion sieving¹⁴, ion rectification¹⁵, and ion diodes¹⁶.

Recently, two-dimensional (2D) laminar membranes constructed through the parallel stacking of ultrathin nanosheets have garnered significant interest in the field of nanofluidics¹⁷. In these 2D membranes, channels with ångström-scale precision are formed between adjacent nanosheets during stacking, providing a distinctive experimental platform for exploring ion transport mechanisms under extreme confinement¹⁸. Li et al. initially proposed the concept of manipulating ion transport behavior by in situ adjusting the surface potential of graphene-based lamellar membranes using an external electric field¹². Our group further confirmed voltage-gated ion diffusion in confined Ti₃C₂T_x channels, highlighting a more pronounced effect of surface potential on the ion diffusion rates due to the excellent conductivity of Ti₃C₂T_x. Notably, by monitoring structural changes in the Ti₃C₂T_x channels using electrochemical quartz crystal microbalance with dissipation (EQCM-D), we demonstrated that voltage-gated ion diffusion is influenced by both the dynamically changing diameter of the confined Ti₃C₂T_x channels and the ion distribution near the channel surface¹⁹. It is anticipated that a more efficient ionic gating response could be achieved if the nanoconfined structure of Ti₃C₂T_x membrane could be stabilized.

Based on this, a Zr₄-Ti₃C₂T_x layered membrane with stable interlayer spacing was designed to examine its influence on ion and water molecule transport through controlled surface potential (Fig. 1a). When a positive voltage was applied, ion permeability was enhanced, whereas a negative voltage led to inhibition of ion transport. Surprisingly, the diffusion of water molecules exhibited an opposite response: under negative voltage, hydrogen bonds between water molecules and Ti₃C₂T_x at the interface were strengthened, promoting water molecule rotation and enhancing diffusion. To capitalize on these effects, an innovative voltage-gated desalination strategy was implemented. By applying a negative voltage to the membrane surface, the NaCl rejection rate was significantly improved, increasing from 72.09% to 98.57%. At the same time, water permeability was markedly enhanced, with water permeance reaching 0.76 L m⁻²  h⁻¹ bar⁻¹, a 2.05-fold improvement. Additionally, excellent performance was demonstrated in lithium concentration applications, where both water permeability and Li⁺ rejection were enhanced. Compared to reported membranes, significantly higher water permeability was achieved in the Zr₄-Ti₃C₂T_x membrane under external voltage application.

Fig. 1: Characterization of Ti₃C₂T_x nanosheets and Zr₄-Ti₃C₂T_x membranes.

a Schematic illustration of Zr₄ ions intercalated within Ti₃C₂T_x nanochannels, stabilizing the interlayer spacing. b TEM image of a monolayer Ti₃C₂T_x nanosheet. c SEM image of the cross-section of the Zr₄-Ti₃C₂T_x membrane. d Zr 3d XPS spectra with fitting results for the pillared Zr₄-Ti₃C₂T_x membrane. e XRD patterns and corresponding interlayer spacings of the Ti₃C₂Tₓ membrane after immersion in DI water and various 0.2 mol L⁻¹ salt solutions for 2 h. The upper panel shows the (002) diffraction peaks under each condition, while the lower panel summarizes the calculated interlayer spacings. f XRD patterns and corresponding interlayer spacings of the Zr₄-Ti₃C₂Tₓ membrane after immersion in DI water and various 0.2 mol L⁻¹ salt solutions for 2 h. The upper panel shows the (002) diffraction peaks under each condition, while the lower panel summarizes the calculated interlayer spacings. g Comparison of the FWHM of Zr₄-Ti₃C₂T_x and Ti₃C₂T_x membranes. h Changes in the interlayer spacing of the Zr₄-Ti₃C₂T_x membrane as a function of varying Zr₄ ion concentrations. Error bars indicated the standard deviations from three different samples.

Results

Preparation of Zr₄-Ti₃C₂T_x membrane with stable d-spacing

Ti₃C₂T_x nanosheets with high conductivity prepared via the minimum intercalation layer disassembly method (MILD) was used as building blocks for the construction of a voltage-gated ion and water transport system²⁰. Initially, aluminum layers in Ti₃AlC₂ (MAX phase) were selectively etched with LiF and HCl, generating an accordion-like multilayer structure (Supplementary Fig. 1). The disappearance of the (008) diffraction peak at 35° and the left shift of the (002) peak in X-ray diffraction (XRD) pattern indicated that the aluminum elements in Ti₃C₂T_x were successfully etched (Supplementary Fig. 2). Subsequently, Lithium ions were intercalated into the Ti₃C₂T_x interlayers, weakening interlayer interactions and expanding the spacing, which facilitated efficient nanosheet exfoliation. The obtained Ti₃C₂T_x nanosheets colloidal suspension with typical Tyndall effect demonstrated the excellent dispersion in water (Supplementary Fig. 3). Transmission electron microscopy (TEM) revealed that the nanosheets were thin, transparent, and structurally intact (Fig. 2b). Atomic force microscopy (AFM) further confirmed that the nanosheets exhibited a high aspect ratio with a lateral dimension of approximately 1.4 μm and a thickness of about 1.7 nm (Supplementary Fig. 4).

Fig. 2: Ion transport in voltage-gated Zr₄-Ti₃C₂T_x nanochannel membranes.

a Schematic of the experimental setup for investigating ion diffusion modulation under external voltage. b Schematic illustrating ion and water molecule transport modulation under positive voltage. c Profiles of normalized ion permeation rates as a function of Vs measured in LiCl, NaCl, and MgCl₂ electrolyte solutions. d Schematic of ion and water molecule transport modulation under negative voltage. e Normalized Na⁺ permeation rates under voltage modulation in various nanochannel membranes. Data for graphene-based membranes are from ref. ¹². The blue and yellow backgrounds represent negative and positive voltages (vs Ag/AgCl), respectively. f Profiles of normalized Na⁺ permeation rates as a function of negative Vs measured at varying NaCl concentrations. g Investigation of the instantaneous response and hysteresis of ion diffusion under cyclic voltage switching between 0 V and 0.5 V, with alternating voltage duration intervals of 10 min and 90 min. The permeation rate of ions was measured under various voltage conditions using different membranes. Error bars indicated the standard deviations from three different samples.

Under vacuum assistance, Ti₃C₂T_x nanosheets were parallelly stacked layer by layer to form a lamellar nanochannel membrane. To improve the structural stability of Ti₃C₂T_x membrane, tetranuclear hydroxo zirconium ([Zr₄(OH)₈(H₂O)₁₆]⁸⁺, termed as Zr₄) were synthesized by dissolving and aging zirconyl chloride (ZrOCl₂·8H₂O)²¹ (Supplementary Fig. 5). The Ti₃C₂T_x membrane was then immersed in the Zr₄ ions solution, and the Zr₄-Ti₃C₂T_x membrane was obtained after full intercalation. To assess the effects of Zr₄ intercalation treatment on the Ti₃C₂T_x membranes, Fourier transform infrared spectroscopy (FTIR) and X-ray photoelectron spectroscopy (XPS) analyses were performed. For the pristine Ti₃C₂T_x membrane, peaks for -OH stretching at 3400 cm⁻¹ and C = O stretching at 1630 cm⁻¹ were observed in the FTIR spectrum due to the surface oxygen-containing functional groups. After Zr₄ intercalation, distinct Zr=O peaks emerged at 898 cm⁻¹ and 1038 cm⁻¹ in Zr₄-Ti₃C₂T_x membrane, confirming the effective incorporation of Zr₄ ions into the nanochannels (Supplementary Fig. 6)²². Moreover, XPS analysis revealed that the C 1s spectrum of original Ti₃C₂T_x membrane exhibited peaks at 282.05 eV (C-Ti), 284.80 eV (C-C), and 288.65 eV (C = O), while the O 1s spectrum showed peaks at 529.8 eV, 531.2 eV, 532.2 eV and 533.4 eV corresponding to TiO_2-xF_x, C-Ti-O_x, C-Ti-(OH)_x and O = C-OH bonds, respectively (Supplementary Fig. 7). In Zr₄-Ti₃C₂T_x membrane, in addition to the typical peaks observed in Ti₃C₂T_x membrane, the Zr 3d peaks at 183.15 eV and 185.48 eV attributed to the 3d_5/2 and 3d_3/2 orbitals of Zr-OH were also observed, also indicating the successful intercalation of Zr₄ ions into the Ti₃C₂T_x membrane²³ (Fig. 1c and Supplementary Fig. 8).

To further investigate the impact of Zr₄ intercalation on the membrane structure, we performed scanning electron microscopy (SEM) analysis. SEM revealed that slight wrinkles could be observed on both original Ti₃C₂T_x membrane and Zr₄-Ti₃C₂T_x membrane with corrugated structures on their surfaces. Cross-sectional SEM images showed typical layered structure in both membranes, suggesting that little effect of Zr₄ intercalation during the formation of 2D nanochannels^24,25 (Fig.1b and Supplementary Fig. 9, 10). Furthermore, energy-dispersive spectroscopy (EDS) confirmed a homogeneous distribution of Ti, O, and C elements in the Ti₃C₂T_x nanosheets, as well as Zr elements from the pillaring process, suggesting uniform Zr₄ intercalation within the nanochannel membrane (Supplementary Fig. 11). Optical images demonstrated that Zr₄-Ti₃C₂T_x membrane still maintained favorable flexibility, implying that interlayer Zr₄ pillars did not significantly impact the mechanical properties of the original Ti₃C₂T_x membrane (Supplementary Fig. 12).

The structural characteristics of Ti₃C₂T_x membrane before and after Zr₄ ions pillared were analyzed using XRD. Here, the Ti₃C₂T_x membrane was fully immersed in pure water and various chloride salt solutions (0.2 M, 2 h), and after gently wiping the membrane surface, XRD measurements were immediately conducted. The interlayer spacing of Ti₃C₂T_x membrane continuously increased with the prolonged immersion time, eventually stabilizing after approximately 1 h (Supplementary Fig. 13). Furthermore, the size of interlayer channels of the Ti₃C₂T_x membrane exhibited notable changes with different degrees after immersion in different chloride salt solutions. As illustrated in Fig. 1e, the negatively charged surface of the Ti₃C₂T_x nanochannels facilitated the spontaneous intercalation of water molecules and cations from the solution into the nanochannels. The steric hindrance effects generated by the intercalation of different ions and water molecules induced varying degrees of swelling in the nanochannel membrane, leading to changes in the interlayer spacing^24,26. Notably, calculations based on Bragg’s law indicated that Na⁺ showed the most significant impact, and the interlayer spacing increased from 13.4 Å to approximately 16.1 Å after immersion in NaCl solutions.

To suppress this swelling tendency, Zr₄ pillaring was employed to stabilize the interlayer channels of the Ti₃C₂T_x nanochannel membrane, ensuring that the interlayer spacing stabilizing under varying solution environments. As shown in Fig.1f, under the same immersion conditions as Ti₃C₂T_x nanochannel membrane, the interlayer spacing of the Zr₄-Ti₃C₂T_x membrane almost stabilized at approximately 14.1 Å, with fluctuations of less than 0.1 Å in different salt solutions. Moreover, prolonged immersion of the Zr₄-Ti₃C₂T_x membrane in aqueous solutions confirmed its reliable anti-swelling properties (Supplementary Fig. 14, 15). Inductively coupled plasma optical emission spectrometry (ICP-OES) revealed that the concentration of Zr⁴⁺ in the solution remained below the detection limit (<10 ppb) after 24 h of immersion, indicating minimal Zr⁴⁺ leakage from the membrane even after extended soaking. Furthermore, compared to the Ti₃C₂T_x membrane, the interlayer spacing of the Zr₄-Ti₃C₂T_x nanochannels expanded from 13.4 Å to 14.1 Å due to the volume effect of Zr₄. Meanwhile, the full width at half maximum (FWHM) of the Zr₄-Ti₃C₂T_x membrane decreased from 1.62 to 1.09, suggesting that Zr₄ intercalation not only optimized structural stability but also enhanced the alignment of the nanosheets, resulting in a more compact and ordered structure (Fig. 1g).

Zeta potential measurements revealed the opposite charge polarities of Ti₃C₂T_x nanosheets and Zr₄ ions. Specifically, The Ti₃C₂T_x nanosheets demonstrated a negative zeta potential of −29.93 mV in colloidal solutions, primarily due to the deprotonation of surface oxygen-containing functional groups. For the Zr₄ ions, a multinuclear structure formed where each Zr was coordinated by eight bridging hydroxyl groups and four water molecules, resulting in a distorted dodecahedral coordination polyhedron. The zeta potential of the Zr₄ ions was measured at +25.59 mV, affirming their positive charge, in agreement with prior studies^21,27,28 (Supplementary Fig. 16). When intercalating into the Ti₃C₂T_x nanochannels, the steric hindrance caused by the Zr₄ ions also lead to an increase of the interlayer spacing. However, the strong electrostatic attraction between the negatively charged Ti₃C₂T_x surfaces and the positively charged Zr₄ ions resulted in the decrease of the distance between adjacent nanosheets. The final interlayer structure of Zr₄-Ti₃C₂T_x membrane is determined by the balance between van der Waals forces and electrostatic interactions. This pillared layer structure effectively suppressed swelling, and the interlayer spacing remained stable when the membrane was immersed in various salt solutions, indicating that the insertion of cations and water molecules did not significantly alter the nanochannel geometry (Fig. 1a). Furthermore, the introduction of Zr₄ ions led to a slight reduction in the negative zeta potential of the Ti₃C₂T_x membrane, decreasing from −33.2 to −23.2 mV (Supplementary Fig. 17). This reduction was attributed to the strong electrostatic attraction between the positively charged Zr₄ ions and the negatively charged Ti₃C₂T_x surface, which facilitated the firm adsorption of Zr₄ ions onto the Ti₃C₂T_x surface. Consequently, the positively charged Zr₄ ions induced a screening effect, diminishing the surface charge density of the Ti₃C₂T_x membrane. Moreover, the contact angle of water on the original and treated Ti₃C₂T_x membrane was 56.30° and 71.68°, respectively. This increase in contact angle was attributed to the reduced surface potential of the Zr₄-Ti₃C₂T_x membrane, which weakened electrostatic interactions with polar water molecules, thereby decreasing wettability of membrane (Supplementary Fig. 18).

The concentration of Zr₄ ions played a critical role in anti-swelling property of the Zr₄-Ti₃C₂T_x membrane (Fig. 1h). At lower Zr₄ ions concentrations, the electrostatic attraction between Zr₄ ions and Ti₃C₂T_x nanosheets was insufficient to suppress the swelling induced by cation intercalation, resulting in noticeable swelling phenomenon, similar to that observed in the original Ti₃C₂T_x membrane. As the Zr₄ ions concentration increased from 0.05 M to 0.1 M, the number of Zr₄ pillars in the nanochannels grew, and electrostatic forces between Zr₄ ions and Ti₃C₂T_x nanosheets became the dominant factor regulating interlayer spacing, significantly suppressing swelling behavior. However, at a Zr₄ ions concentration of 0.2 M, swelling of the Zr₄-Ti₃C₂T_x membrane in salt solutions appeared once again, accompanied by a marked increase in the FWHM of the XRD peak. Previous studies on montmorillonite modified with Zr₄ pillars indicated that excessive Zr₄ ions concentrations led to hydrolysis imbalances during preparation, and the by-products would weaken the membrane’s mechanical strength and stability²⁹. Therefore, for subsequent ion transport investigation, the optimal Zr₄ ions concentration was set at 0.1 M.

Voltage-gated control of ion transport in membranes

To realize voltage-gated ion transport, an experimental platform was designed to explore ion transport across the pillared Zr₄-Ti₃C₂T_x membrane with stable nanoconfined channels under a concentration gradient, similar to previous studies. Although Zr₄ intercalation caused a slight reduction in conductivity, the Zr₄-Ti₃C₂T_x membrane retained the inherently high conductivity of Ti₃C₂T_x, reaching 2966.5 S cm⁻¹, which was more than 10 times higher than that of graphene-based membrane (Supplementary Fig. 19). As a result, the Zr₄-Ti₃C₂T_x membrane could be employed as an ideal candidate for constructing voltage-gated ion transport systems. Here, the conductive Zr₄-Ti₃C₂T_x membrane was attached to an annular Pt foil, serving as the working electrode (WE), while a Pt mesh and an Ag/AgCl electrode functioned as the counter electrode (CE) and reference electrode (RE), respectively. During the experiments, equal volumes of salt solution and pure water were placed on opposite sides of the membrane, establishing a concentration gradient that drove ion diffusion. To modulate this process, the surface potential of the Zr₄-Ti₃C₂T_x channels was precisely tuned by applying an external voltage (Vs) via an electrochemical workstation in potentiation mode (Fig. 2a).

Whether or not an external Vs was applied, the Na⁺ concentration increased linearly with time, indicating a concentration gradient-driven transport (Supplementary Fig. 20). However, the diffusion rates, represented by the slope of the curves, changed significantly once Vs was applied. To assess the effect of surface potential on ion permeation rates, external voltages ranging from −0.5 V to 0.5 V were applied to the Zr₄-Ti₃C₂T_x nanochannel membrane as ions with different valences diffused through. As shown in Fig. 2c, for Li⁺, Na⁺, and Mg²⁺, a positive Vs significantly increased the ion permeation rates compared to conditions without Vs. Conversely, when the Vs polarity was reversed, ion permeation rates were effectively suppressed, shifting the ionic transport gate to an “off” state. For Na⁺, the permeation rate increased eightfold, from 6.81 × 10⁻³mol m⁻² h⁻¹ to 5.42 × 10⁻²mol m⁻² h⁻¹ under a positive 0.5 V Vs and decreased by half under a negative −0.5 V Vs. Additionally, Fig. 2c showed that, for all the tested ions, the promotion and suppression of ion flow became more pronounced as the applied voltage increased. These results confirm that ion permeation through the Zr₄-Ti₃C₂T_x nanoconfined channels can be effectively controlled by both the magnitude and polarity of the applied voltage.

Furthermore, while the transport behaviors of Li⁺, Na⁺, and Mg²⁺ ions followed similar trends under the applied voltage, the effect of Vs on Mg²⁺ transport efficiency was less pronounced compared to Li⁺ and Na⁺. Specifically, under +0.5 V Vs, the Mg²⁺ transport rate increased by a factor of 5.79, compared to 6.78 for Li⁺ and 7.97 for Na⁺. Under a −0.5 V Vs, the Mg²⁺ transport rate decreased by 0.65-fold, whereas the Li⁺ and Na⁺ transport rates decreased by 0.53-fold and 0.41-fold, respectively. This disparity is attributed to the higher valence state of Mg²⁺, which induces a stronger electrostatic shielding effect¹⁴. Consequently, the surface potential of the nanochannels has a reduced influence on the charge density of Mg²⁺, leading to its comparatively lower transport efficiency.

The surface potential regulation efficiency of the stabilized lamellar-structured Zr₄-Ti₃C₂T_x membrane for ion transport in extremely confined channels was compared with the original instable Ti₃C₂T_x membrane and graphene-based membrane¹². To quantify the voltage-gating effect, an ionic gating index (η), defined as the ratio of ion permeation rate before and after the external voltage was applied, was introduced. Figure 2e illustrates that although the ion transport behavior in the original Ti₃C₂T_x and graphene channels was influenced by the applied Vs, the Zr₄-Ti₃C₂T_x membrane demonstrated a significantly more distinct response due to its superior structural stability and electrical conductivity. Specifically, the ionic gating index for Na⁺ transport in the Zr₄-Ti₃C₂T_x membrane (η-Na) consistently exceeded 1 under positive Vs and dropped below 1 in the negative Vs range. Moreover, the intensity of this regulatory effect was markedly higher in the Zr₄-Ti₃C₂T_x membrane. For instance, at Vs = +0.5 V and −0.5 V, the η values followed these respective orders: η (positive Vs) for Zr₄-Ti₃C₂T_x (7.97) > Ti₃C₂T_x (2.80) > graphene-based (1.05), and η (negative Vs) for graphene-based (2.60) > Ti₃C₂T_x (2.18) > Zr₄-Ti₃C₂T_x (0.41). In addition, the dependence of ion concentration on the efficiency of ion diffusion modulation was also investigated (Fig. 2f and Supplementary Fig. 21). When the ion concentration in the feed compartment increased, the Na⁺ diffusion rate continued to follow the trend where ion flow accelerated with positive Vs and decelerated with negative Vs. However, we observed that the modulation effect weakened at higher concentrations. A corresponding trend was observed for Mg²⁺, in which the modulation effect decreased with increasing ion concentration (Supplementary Fig. 22).

To evaluate whether the voltage-gated ion transport behavior was instantaneous and reversible, we conducted a detailed measurement of ion permeation rate variations over time under an applied voltage. Within the first 30 min of applying +0.5 V, the ion permeation rate exhibited an immediate response, increasing sharply from 7.02 × 10⁻³mol m⁻² h⁻¹ (at 0 V) to 3.3 × 10⁻²mol m⁻² h^–1, corresponding to a gating index of 4.83 (Supplementary Fig. 23a). However, as the voltage application time increased, the ion permeation rate continued to rise, albeit at a diminishing rate. Similarly, while the gating index showed an overall increasing trend, its growth rate gradually slowed. Beyond 2.5 h, voltage-regulated ion transport stabilized, and the gating index plateaued at 7.91 (Supplementary Fig. 23b). These results indicated that voltage-gated ion transport exhibited an immediate response, yet the ion permeation rate also displayed hysteresis relative to voltage application. To further verify this behavior, a variable voltage (0–0.5 V) was continuously applied during the Na⁺ diffusion experiment (Fig. 2g and Supplementary Fig. 24) The results showed that the ion diffusion rate fluctuated rapidly with changes in external voltage. However, the reversibility of the gating index depended on the voltage duration, reaching equilibrium only after prolonged voltage application. This hysteresis in ion permeation rate changes primarily arose from the interplay between the high electrical conductivity of the Zr₄-Ti₃C₂T_x membrane and the finite mobility of ions. Upon voltage application, the membrane surface charge adjusted instantaneously, altering ion species and concentrations within the nanochannels to accommodate the updated charge state. However, due to the limited mobility of ions, their redistribution lagged behind the rapid adjustment of surface charge densitys³⁰. As a result, an equilibration period was required for ion distribution within the confined nanochannels to stabilize, during which the transmembrane ion diffusion rate gradually reached a steady state. In addition, multiple long-term alternating voltage stability tests were carried out at 0 V and 0.5 V with 90-min intervals for 24 h. The results indicated that both the ion flux and the gating index remained stable throughout the testing period, further confirming the long-term reliability and stability of the gating performance (Supplementary Fig. 25).

Experiments were conducted to determine whether the voltage-gated ion transport behavior was influenced by changes in membrane properties during permeation tests. XRD measurements were performed on Zr₄-Ti₃C₂T_x membranes immediately after applying voltages of varying polarities, magnitudes, and durations to investigate the effect of electric field conditions on interlayer spacing (Supplementary Fig. 26). The results indicated that applying $\pm$0.5 V for 1 h led to limited changes in interlayer spacing, which remained at 13.93 Å and 13.97 Å, respectively. A stable lamellar structure with no obvious changes in interlayer spacings was still observed even when the voltage was increased to $\pm$1.0 V for 8 h. Based on these experimental results, we infer that the applied electric field has a limited impact on the interlayer structure of Zr₄-Ti₃C₂T_x membranes. Therefore, ion transport behavior is influenced by changes in surface potential induced by the applied voltage rather than by alterations in nanochannel size. Additionally, high-resolution XPS spectra confirmed that the chemical states of both Ti and Zr remained essentially unchanged, indicating that the Zr₄-Ti₃C₂T_x nanochannel structure was unaffected by variations in membrane surface potential. The negligible presence of TiO₂ in the Ti XPS spectrum following ion permeation experiments further suggested that the applied voltage did not induce membrane oxidation (Supplementary Fig. 27). ICP-OES analysis of the ion permeate compartment showed that the Ti concentration in the solution was below the detection limit, confirming that membrane decomposition was negligible. Furthermore, the constant pH in both the feed and permeate compartments during the diffusion experiments indicated the absence of electrochemical water oxidation or reduction (Supplementary Fig. 28). Based on these analyses, we can rule out that the variations in ion permeation rates under different surface potentials were related to changes in membrane physical structure, chemical composition, or electrochemical reactions during the ion diffusion modulation tests.

The transmembrane ion transport process can be divided into two distinct stages: (1) ion entry from the feed solution into the nanochannels, and (2) ion migration through the nanochannels to the permeate side. A comparison between the nanochannel diameter and the hydrated ion size indicates that the ions selected in this study possess hydrated diameters larger than the width of the Zr₄-Ti₃C₂T_x channels. Recent research has shown that in nanoscale confinements, ions cannot be simply modeled as rigid spheres with constant volume^31,32. Instead, their hydration shells (HSs) may deform or partially shed to facilitate entry into confined spaces. The gradual increase in ionic conductance observed on the permeate side during ion transport experiments supports the occurrence of ion dehydration (Supplementary Fig. 20). In addition to the size-induced dehydration effect, electrostatic interactions also play a crucial role in ion transport. These interactions, however, exert contrasting influences on ion partitioning at the channel entrance and diffusion within the channel. For instance, multivalent cations such as Mg²⁺ are electrostatically attracted to negatively charged nanochannel surfaces, promoting their entry. However, once inside the channel, the same negative surface charges can impede further transport by increasing diffusion resistance due to strong electrostatic interactions. Recent studies analyzing the contributions of ion partitioning and intrachannel diffusion to the total energy barrier have concluded that intrachannel diffusion is the rate-limiting step for salt transport through sub-nanometer pores³³. To further elucidate the role of surface charge in modulating ion transport, both in the presence and absence of an applied electric field, ab-initio calculations based on density functional theory (DFT) were conducted. The solvation states of hydrated ions in confined nanochannels are dictated by the interactions between ions and the Ti₃C₂T_x surface, as well as between ions and water molecules. Using Na⁺ as a model ion, we conducted ab initio molecular dynamics (AIMD) simulations to explore the distribution and HS structure of ions along the channel. The channel wall consists of Ti₃C₂O₂, as oxygen groups are predominant (over 80%) under our fabrication conditions³⁴. The results revealed two solvation states, similar to those found in graphene channels in our previous work (Supplementary Fig. 29)^35,36. One state involves ions residing in a single layer of water (or near the Zr₄-Ti₃C₂T_x surface) with a truncated octahedral structure (first HS, hydration number ≈ 5), termed the L state (Supplementary Fig. 30). The other state features ions embedded in the interlayer of water or within the central water layer, exhibiting a spherical first HS (hydration number ≈ 6), referred to as the M state. According to previous studies, the interaction between cations and Ti₃C₂T_x, with a binding energy (E_b) that exceeds 4 eV, is significantly stronger than that observed in other 2D materials such as GO, BN, and MoS₂^37,38,39. This is attributed to the distinctive polar electrical structure of Ti₃C₂T_x nanosheets, where its surface remains electronegative regardless of the type of terminal groups, while the inner layer exhibits electropositive properties. As a result, in the absence of an applied gate voltage, under the influence of the polar electrical structure, a significant proportion (about 77%) of cations reside near the surface within one layer of water (Supplementary Fig. 31). We also calculated the hydration states of ions under different gate voltages. AIMD simulations also confirm that injecting excess electrons (or removing electrons) in the Ti₃C₂T_x nanosheet alters solvation state of cations due to the changed cation–Ti₃C₂T_x interactions. Specifically, under a positive gate voltage, although the proportion of cations in the L hydration state changed only slightly, their spatial distribution broadened significantly. The FWHM of the L state increased from 0.57 Å (no gate voltage) to 0.73 Å, indicating a wider spread near the channel wall. In contrast, under a negative gate voltage (−V), cations in the M hydration state were nearly absent, and the proportion of L state cations increased markedly from 77% (no gate voltage) to 98% (Supplementary Fig. 31). This redistribution of cations in the EDL provides important evidence that transmembrane cation transport can be effectively regulated by the gating voltage via its influence on cation–Ti₃C₂T_x interactions. Under a negative voltage, strong electrostatic interactions between the negatively charged channel surface and Na⁺ ions promote their localization near the channel wall in an L hydration state, resulting in reduced mobility and slower transport rates. In contrast, under a positive voltage bias, Na⁺ ions are more likely to migrate toward the center of the channel, where they experience weaker electrostatic constraints and exhibit higher transport rates.

Given the calculation of ion diffusion coefficients in the nanochannel directly requires prolonged simulation times (on the order of nanoseconds), which is not feasible with AIMD. Therefore, we assessed the diffusion barrier of Na⁺ using the climbing image-nudged elastic band (CI-NEB) method⁴⁰. Our previous findings indicate that the E_b of ion-Ti₃C₂T_x at three high-symmetry sites: the top of the terminating functional groups (T site), the Ti atom (Ti site), and the C atom (C site), follows the order E_b (C) > E_b (Ti) > E_b (T)³⁷. Consequently, the optimal diffusion pathway is C-Ti-C. The results demonstrate that the diffusion barrier (ΔG₁) increases at negative while decreasing at positive Vs. We also calculated the gating index, which reached a value of 12.1 when injecting 2 electrons (corresponding to an external voltage of −0.9 V), aligning well with experimental observations (Supplementary Fig. 31-32). Thus, the modulation of ion permeation can be effectively explained by the electrified diffusion barrier of ions within the Ti₃C₂T_x nanochannel.

The thickness of EDL can be simply estimated by Debye length, which is approximately 1.36 nm, 0.68 nm, 0.43 nm for electrolyte concentrations of 0.05, 0.2, 0.5 M studied in our work. Comparing the channel diameter of the Zr₄-Ti₃C₂T_x membrane (0.53 nm) with the Debye length, it can be observed that the EDLs at the channel wall surface overlap in our nanofluidic systems. As a result, due to the negatively charged nature of Zr₄-Ti₃C₂T_x membrane, counterions (cations) accumulate within the nanochannels, while co-ions (anions) are repelled from the channels. Consequently, although the cations can migrate from the electrolyte side to the water side driven by the concentration gradient, fewer anions can cross the membrane. However, to maintain charge neutrality, cations on the drain side drag anions through the nanochannels. Due to the strong interaction between Ti₃C₂T_x nanosheets and cations as described above, cation diffusion primarily dominates salt transport. As a result, taking the negative voltage for instance, the dragging effect results in the observed voltage-gated salt transport behavior due to the attenuated cation diffusion. Furthermore, as described above, the gating voltage has a less pronounced effect on Mg²⁺ transport compared to monovalent Li⁺ and Na⁺. Our calculations show that the distribution of water molecules in the first HS of Na⁺ and Li⁺ is broader than that of Mg²⁺, suggesting that Na⁺ and Li⁺ have more flexible HS structures that are more susceptible to changes in surface properties (Supplementary Fig. 33). Due to their rigid HS, Mg²⁺ ions exhibit limited positional mobility within nanochannels under varying gate voltages, in contrast to monovalent ions such as Na⁺ ions. For instance, under a negative voltage, the enhanced surface charge density promotes the accumulation of Na⁺ ions near the channel interface by deforming or partially stripping their HSs, as previously discussed. However, the robust HS of Mg²⁺ ions resists such dehydration-driven displacement, thereby diminishing their response to electrostatic gating. Although the regulatory effect of gate voltage on Mg²⁺ transport is less pronounced, a positive voltage still reduces the electrostatic attraction from the charged channel walls, thereby facilitating the transmembrane migration of Mg²⁺ ions by lowering the energy barrier for intrachannel diffusion.

Voltage-gated control of water transport in membranes

In addition to ions, the effect of applied voltage on water molecule diffusion in the nanochannel membrane was also examined. Unlike the ion diffusion experiments discussed above, sucrose solution and pure water were added to the two reservoirs on either side of the membrane, respectively. The osmotic pressure drives water molecules from the pure water side to the sucrose solution side. As illustrated in Fig. 3a, like voltage-gated ion transport, water transport was also significantly influenced by the applied external voltage. However, totally contrary to voltage-gated ion transport, the water diffusion accelerated under negative voltage and decelerated under positive voltage. To examine the impact of membrane potential on water transport, negative voltages ranging from 0 V to −0.8 V were applied to the Zr₄-Ti₃C₂T_x nanochannels. The results demonstrated that increasingly negative voltages markedly enhanced water transport rates, with higher voltages yielding greater water permeance. For instance, in Zr₄-Ti₃C₂T_x channels with a membrane thickness of 2000 nm, applying −0.8 V Vs increased the water permeance from 0.38 L m⁻² h⁻¹ bar⁻¹ (at 0 V) to 0.69 L m⁻² h⁻¹ bar⁻¹, approximately a twofold increase in the diffusion rate (Fig. 3b).

Fig. 3: Water molecule transport in voltage-gated Zr₄-Ti₃C₂T_x nanochannel membranes.

a Variation in water permeance under different applied voltages. The blue and yellow backgrounds represent negative and positive voltages (vs Ag/AgCl), respectively. b Influence of membrane thickness on voltage-gated water molecule transport behavior. c Effect of osmotic pressure on voltage-gated water transport behavior. d The water in Ti₃C₂T_x nanochannel. The green lines are hydrogen bonds between O atoms on the surface and water. The definition of hydrogen bond can be found in Methods. Here, the result is the snapshot at −V. e The number of Ti₃C₂T_x-water HBs during AIMD simulation. f The MSD curves. The diffusion coefficients of water are in proportion to the slope. g The Free energy profile of water projected onto the Ti₃C₂T_x surface. Here, the red, gray and airy blue dots mean the T, C and Ti sites on Ti₃C₂T_x. The red, gray and blue areas are the position of O, C and Ti sites. h The 1D free energy surface along the path (white dash line) indicated in first figure of (e). Error bars indicated the standard deviations from three different samples.

We also examined the impact of membrane thickness on voltage-gated water transport behavior. The results indicated that as membrane thickness increased, the modulation effect of external voltage on water transport initially strengthened, leading to a significant rise in water permeance. However, with further increases in membrane thickness, the modulation effect weakened, diminishing the improvement in water permeance. This phenomenon primarily arose from the combined influence between the membrane’s transport pathway length and the applied electric field. In thinner membranes (e.g., 1000 nm), the reduced transport distance within the nanochannels enables a naturally higher water permeation rate. Although external voltage promoted water transport, its influence was less significant due to the inherently short pathway, thereby diminishing the voltage-gated effect. In contrast, as membrane thickness increased (e.g., 3000 nm), the extended transport pathway introduced additional resistance, hindering water permeation. While the external voltage effectively regulates water transport in thicker membranes by reinforcing the gating effect, the excessively long pathway increases resistance, ultimately constraining the further enhancement of the gating effect in the 3000 nm membrane.

In addition to membrane thickness, we also explored the influence of osmotic pressure on voltage-gated water transport behavior by varying sucrose concentrations (0.5 M, 0.8 M, and 1.0 M). Keeping all other conditions constant, we observed that as sucrose concentration increased, water permeance gradually rose, indicating a strong dependence of water transport on the osmotic pressure difference. However, the effect of osmotic pressure on voltage-controlled water transport was relatively minor (Fig. 3c). At a sucrose concentration of 0.5 M, applying −0.8 V Vs led to a 1.45-fold increase in water permeance compared to no voltage, while at 0.8 M, the increase slightly diminished to 1.41-fold. Additionally, to ensure the accuracy of the experiment, we measured the sucrose concentration in the solution on the water side after the test to verify whether it permeated through the membrane⁴¹.The results showed that the sucrose concentration on the water side was below the detection limit (<88.6 ppm), indicating that sucrose molecules did not significantly permeate through the membrane⁴². Based on Bragg’s equation, the d-spacing of the Zr₄-Ti₃C₂T_x membrane was determined to be 14.1 Å (Fig. 1f). Given that the thickness of a monolayer Ti₃C₂T_x nanosheet is approximately 8.8 Å, the effective interlayer spacing of the Zr₄- Ti₃C₂T_x nanochannel membrane is estimated to be ~5.3 Å⁴³. Previous literature reports that the hydrated diameter of sucrose is ~14.6 Å, which is significantly larger than the effective interlayer spacing of the membrane⁴⁴. Therefore, although the osmotic pressure difference influenced water transport efficiency, the effect of voltage regulation remained relatively stable under varying osmotic conditions.

By conducting long-time AIMD simulations, we calculated the diffusion coefficient of water using the mean square displacement (MSD) derived from water trajectories via the Einstein relation (Fig. 3f, detailed methodological information can be found in the Methods section). The diffusion coefficient of water confined within the Ti₃C₂T_x channel was calculated to be 0.85 × 10⁻⁹m⁻² s⁻¹, approximately half that of bulk water (Supplementary Table 1). This reduction is attributed to the confinement effect and strong water–Ti₃C₂T_x interactions. Interestingly, the application of electric potential significantly modulates water diffusion. Under a negative potential (−0.5 V), diffusion coefficient increases approximately two-fold relative to the no-field condition, whereas under a positive potential (+0.5 V), diffusion coefficient decreases by about fivefold.

To explore this further, we calculated the free energy profiles and energy barriers associated with water diffusion under different potentials (Fig. 3g, h)^45,46. The results revealed a very small energy scale within k_BT (1 k_BT ≈ 26 meV), indicating the mobility of water molecules in the Ti₃C₂T_x channel. The free energy landscape showed maxima at oxygen (O) sites, while minima emerged near titanium (Ti) and carbon (C) sites. Furthermore, the free energy profile becomes less corrugated under negative potential but more corrugated under positive potential, compared to the no-field condition. For example, the maximum free energy corrugation is 36 meV without an applied voltage, decreases to 24 meV under negative potential, and increases to 55 meV under positive potential. This reduced corrugation under negative potential directly lowers the diffusion energy barrier (Fig. 3g), facilitating enhanced water mobility. The results described above seemingly contradict conventional expectations, as stronger water–Ti₃C₂T_x interactions under negative potentials would be expected to suppress diffusion, whereas weaker interactions under positive potentials should facilitate it. This anomalous phenomenon suggested that water diffusion within the channel could not be explained solely by water–Ti₃C₂T_x interactions. Instead, the voltage-controlled water transport behavior likely arose from changes in both the spatial distribution and molecular configuration of water molecules induced by the applied external voltage. Taking the negative gate voltage as an example, density distribution calculations indicated that more water molecules accumulated near the interface, with their hydrogen atoms preferentially oriented toward the Ti₃C₂T_x surface (Supplementary Fig. 34). Our results showed that the oxygen atoms on Ti₃C₂T_x could act as hydrogen bonds (HBs) acceptors, forming HBs with the hydrogen atoms in water molecules (Fig. 3d). Notably, we found that the number of interfacial water–Ti₃C₂T_x HBs (n_HB, Fig. 3e) increased under negative Vs while decreasing under positive Vs, indicating that the electric field effectively modulated the hydrophilicity of the Ti₃C₂T_x surface. Specifically, at negative Vs, the injection of excess electrons enhanced water–Ti₃C₂T_x interactions, rendering the surface more hydrophilic, whereas the opposite effect occurred at positive Vs^47,48. This electrified hydrophilicity facilitated water entry into the nanochannels, increasing water content in the membrane and promoting water permeation.

Additionally, transmembrane water transport was directly influenced by changes in water molecule configuration under an applied gating voltage. Previous studies had shown that water reorientation occurred through large-amplitude angular jumps ( ~ 60°), enabling the concerted breaking and formation of hydrogen bonds rather than proceeding through a sequence of small diffusive steps. This transition occurred when two equidistant oxygen atoms surrounded a rotating water molecule, allowing its hydrogen atom to rapidly flip within the plane. Moreover, recent research on water dynamics at electrified graphene interfaces has demonstrated that dangling OH groups in water molecules pointing toward negatively charged graphene reorient rapidly, reducing the average water reorientation time⁴⁹. Based on these findings, we infer that voltage-induced changes in water molecule configuration promote rapid molecular rotation, and this surface-assisted water molecule reorientation facilitates fast water diffusion along the channel wall under extreme confinement.

The water permeability of the pristine Ti₃C₂T_x membrane and the Zr₄-Ti₃C₂T_x membrane was experimentally compared to assess the effect of Zr₄ intercalation (Supplementary Fig. 35). The results showed that the Zr₄-Ti₃C₂T_x membrane exhibited a higher water permeability (0.57 L m⁻² h⁻¹ bar^–1) than the pristine Ti₃C₂T_x membrane (0.36 L m⁻² h⁻¹ bar⁻¹), demonstrating the enhancement of water transport by Zr₄ intercalation. Based on the XRD analysis as described above, in the pristine Ti₃C₂T_x membrane, the randomly distributed channels create a slow and tortuous transport route, impeding water diffusion. The enhanced structural uniformity due to the intercalation of Zr₄ minimized hindrance from local defects and disordered channels, facilitating a more direct and rapid water transport pathway⁵⁰. We also incorporated the Zr₄–water cluster into the simulation model (Supplementary Fig. 36). The results indicate that under negative voltage, the presence of Zr₄ pillars does not affect gating performance, and water diffusion accelerates (−V), consistent with results observed in the absence of Zr₄ pillars (Fig. 3f–h and Supplementary Fig. 37). AIMD simulations reveal that water molecules within the first HS of Zr₄ ions are tightly bound, exhibiting minimal exchange with those in outer hydration layers. This behavior aligns with previous studies where water molecules closely associated with proteins or nanopores displayed restricted mobility due to strong interfacial interactions^51,52. Similar to ion hydration structures, the electric fields generated by the charged surface are effectively screened by polarized water molecules in the first HS. While the strong electric field within this layer tightly binds water molecules, those in the outer layers remain diffusive⁵³.

Desalination and Li⁺ recovery under voltage-gated control

Membranes are preferred over other separation technologies in various applications due to their advantages in simplicity, scalability, energy efficiency, and compact footprint. However, existing artificial membranes face an intrinsic trade-off between permeance and selectivity, which limits separation efficiency⁵⁴. For instance, while optimizing membrane nanostructures can significantly enhance ion selectivity for desalination, achieving a dense nanostructure often sacrifices water permeance, and vice versa. Therefore, the simultaneous enhancement of both salt rejection and water permeance remains a significant challenge. The ion and water transport behavior observed in Zr₄-Ti₃C₂T_x membranes under varying surface potentials suggests that permeation rates of both the ion and water molecule can be effectively modulated by controlling the applied external voltage. This intriguing feature motivates us to develop a voltage-gated confined transport system capable of overcoming the traditional trade-off.

Water desalination application using a forward osmosis (FO) model was primarily selected to explore the potential for improved separation efficiency through voltage-gated transport. For direct comparison with existing literature, we adopted a widely used experimental system in FO desalination studies involving 2D membranes, employing NaCl solution (0.05 M) as the feed and sucrose solution (1 M) as the draw solution^32,55. The desalination performance of the membrane was assessed by measuring both water permeance and salt rejection rate. To evaluate the effect of external voltage on FO performance, we applied voltages ranging from 0 V to −1.0 V across the Zr₄-Ti₃C₂T_x nanochannel (Fig. 4a). The results indicated significant improvements in both water permeance and salt rejection rate with increasing external voltage. Specifically, as the voltage increased from 0 V to −1.0 V, water permeance rose from 0.37 L m⁻² h⁻¹ bar⁻¹ to 0.76 L m⁻² h⁻¹ bar⁻¹, and salt rejection improved from 72.09% to 98.57%. The observed simultaneous regulation of ion and water molecule transport is consistent with the previously discussed results from separate ion and water molecule transport experiments, confirming that the application of negative external voltage not only inhibited ion diffusion but also simultaneously enhanced water permeance (Fig. 2b, d). More significantly, this voltage-gated regulation mechanism is effectively maintained throughout the desalination process, enabling concurrent improvements in both water permeance and salt rejection (Fig. 4c).

Fig. 4: Evaluation of external voltage effect on membrane desalination and lithium concentration performance.

a Water permeance and NaCl rejection rate of Zr₄-Ti₃C₂T_x membranes under varying external voltages (membrane thickness: 2000 nm). b Impact of membrane thickness on voltage-gated desalination performance. c Advantages of voltage-gated confined transport systems in overcoming the traditional trade-off between permeability and selectivity. d Comparison of desalination performance of 2000-nm-thick Zr₄-Ti₃C₂T_x membranes at 22 bar with Ti₃C₂T_x, GO, MoS₂, carbon nanotube (CN), hybrid membranes, and polymeric thin-film composite membranes. Desalination tests were conducted with a 1.0 M sucrose solution; salt concentrations may vary between studies. A direct comparison with literature data is provided in Supplementary Table 2. e Water permeance and lithium recovery rate of Zr₄-Ti₃C₂T_x membranes under different applied voltages. Error bars indicated the standard deviations from three different samples.

We also investigated the FO desalination performance of Zr₄-Ti₃C₂T_x membranes with varying thicknesses, both before and after applying external voltage (Fig. 4b and Supplementary Fig. 38). In the absence of an applied voltage, reducing membrane thickness resulted in an increase in water permeance but a decrease in salt rejection efficiency. For instance, when the Zr₄-Ti₃C₂T_x membrane thickness decreased from 3000 nm to 1000 nm, water permeance increased from 0.18 L m⁻² h⁻¹ bar⁻¹ to 0.54 L m⁻² h⁻¹ bar⁻¹, while the salt rejection rate decreased from 74.03% to 65.27%, illustrating the typical trade-off effect. Upon applying a voltage of −1.0 V to the Zr₄-Ti₃C₂T_x membrane, the negative surface potential effectively suppressed ion transport, significantly reducing salt flux and increasing the salt rejection rate to 99.19%. Remarkably, despite the enhanced repulsion of salt ions, water permeance also increased, doubling from 0.18 to 0.47 L m⁻² h⁻¹ bar⁻¹. In this study, the Zr₄-Ti₃C₂T_x membrane exhibits significant salt rejection and rapid water permeability, exceeding the performance limits of conventional two-dimensional membranes. More importantly, this result highlights its potential to address the long-standing challenge of trade-off effect in selectivity and permeability in membrane separation technologies. To further verify the practical applicability, long-term desalination stability tests were performed. The Zr₄-Ti₃C₂T_x membrane maintained a high and stable salt rejection rate over extended testing periods, demonstrating reliable long-term desalination performance and superior selectivity (Supplementary Fig. 39). The Zr₄-Ti₃C₂T_x membrane exhibited significantly higher water permeance under applied voltage compared to the reported membranes (Fig. 4d). It should be noted that the thermodynamic energy requirement for water electrolysis is 1.23 V (vs. SHE = 1.033 V vs. Ag/AgCl), where the oxygen and hydrogen gases are generated at the anode and cathode^56,57. Although increasing the applied voltage may further enhance the desalination rate, to ensure experimental accuracy and stability, we limited the maximum voltage in the forward osmosis desalination experiments to −1 V.

We further examined the ion-selective transport behavior of the Zr₄-Ti₃C₂T_x membrane in a mixed salt system under an applied voltage (50). The experiments were conducted using a solution containing 0.05 M LiCl and 0.05 M MgCl₂. The results indicated that, consistent with the single-salt system, Li⁺ and Mg²⁺ transport could be modulated by the applied voltage, with the gating index closely matching that of the single-salt system. Specifically, at +0.5 V, the permeation rates of Li⁺ and Mg²⁺ reached 4.38 × 10⁻¹mol m⁻² h⁻¹ and 1.27 × 10⁻³mol m⁻² h⁻¹, corresponding to 5.14-fold and 3.45-fold increases, respectively, compared to the no-voltage condition. Notably, the differential enhancement in ion permeation under voltage gating further improved Li⁺/Mg²⁺ selectivity, increasing the selectivity coefficient from 1.71 to 3.45.

In addition to desalination studies, we further extended our investigation to explore the potential of FO in solution concentration processes. Specifically, we focused on lithium resource enrichment, a representative application in salt lake lithium extraction. Given that post-treatment of draw solutions is widely recognized as a major challenge in FO applications, MgCl₂ solution—one of the main constituents of salt lake brine in western China—was selected as a natural draw solution, as the diluted brine produced during the FO process can be returned to salt ponds and conveniently reconcentrated through solar evaporation⁵⁸. The results demonstrated that both water permeance and lithium rejection rate also distinctly increased with rising external voltage. Specifically, when the voltage increased from 0 V to −1.0 V, water permeance rose from 0.14 L m⁻² h⁻¹ bar⁻¹ to 0.39 L m⁻² h⁻¹ bar⁻¹, while the lithium rejection rate improved from 72.11% to 96.54% (Fig. 4e). This trend aligns with observations from desalination experiments, further confirming the effectiveness of negative voltage in enhancing water permeance and selective rejection. It should be noted that reverse osmosis (RO) is the dominant technology for large-scale seawater and brackish water desalination due to its maturity and widespread implementation⁵⁹. However, RO typically requires high operating pressures, posing significant challenges in combining it with the voltage-gated system. In particular, under high-pressure conditions, maintaining the sealing integrity of the three-electrode system while ensuring close contact between the working Pt electrode and the membrane is very difficult. In future research, more effort will be dedicated to the experimental setup to investigate the impact of voltage gating on reverse osmosis separation, advancing the application of pressure-controlled transport technology in practical separation processes.

Discussion

In summary, we have introduced an innovative voltage-gated desalination strategy that effectively enhances both water permeability and salt rejection through the application of external voltage to regulate ion and water molecule transport within high-conductivity Zr₄-Ti₃C₂T_x nanomembranes. From a membrane design perspective, Zr₄ ions were intercalated into Ti₃C₂T_x layered membranes, where the strong electrostatic interactions between Zr₄ ions and Ti₃C₂T_x substantially improved the stability of the interlayer structure. This interaction effectively mitigated membrane swelling in aqueous solutions, ensuring high stability across diverse aqueous environments. Additionally, applying external voltage to the Zr₄-Ti₃C₂T_x membrane enabled precise control over ion and water molecule transport: positive voltage substantially increased ion permeability, while negative voltage inhibited ion transport and enhanced water molecule permeability. Notably, negative voltage facilitated water molecule diffusion by strengthening hydrogen bonds at the Ti₃C₂T_x membrane–water interface, promoting rotational dynamics of water molecules. In desalination experiments, the application of negative voltage raised the ion rejection rate from 72.09% to 98.57% and doubled water permeation. Moreover, this approach demonstrated excellent performance in lithium concentration applications, significantly enhancing both water permeability and lithium-ion rejection rates. This advancement effectively addresses the longstanding trade-off between selectivity and permeability in membrane technologies, providing a robust foundation for the development of next-generation membrane technologies in seawater desalination, resource recovery, and ion sieving.

Methods

Materials and reagents

Ti₃AlC₂ powder (400 mesh) was obtained from Jilin Technology Co., Ltd, China. Hydrochloric acid (HCl, AR, 10 M) was sourced from Shanghai Aladdin Reagent Co., Ltd, China. Sucrose, lithium fluoride (LiF, AR, ≥99%), zirconium chloride (ZrOCl₂·8H₂O, AR, ≥99%), anhydrous sodium chloride (NaCl AR, ≥99.5%), lithium chloride (LiCl, AR, ≥99%), and anhydrous magnesium chloride (MgCl₂ AR, ≥98%) were purchased from Tianjin Kemiou Chemical Reagent Co., Ltd, China. All chemicals and reagents were of analytical grade and used without further purification.

Characterization

The interlayer spacing of Ti₃C₂T_x nanochannels was analyzed using X-ray diffraction (XRD, Ultima IV, Japan) with a Cu Kα radiation source (λ = 1.54181 Å, 40 kV, 30 mA, and 6°/min). The thickness and surface morphology of the nanosheets were characterized using atomic force microscopy (AFM, Bruker Multimode 8) in tapping mode. Transmission electron microscopy (TEM, JEOL JEM-F200, Japan) was employed to examine the micro-morphology of the nanosheets. Scanning electron microscopy (SEM, Zeiss Gemini SEM 300, Germany) provided insights into the morphology and size of the Ti₃C₂T_x nanosheets. Inductively coupled plasma (ICP) was utilized to determine ion concentrations in the solution (ICP-5000, China). X-ray photoelectron spectroscopy (XPS) was used to analyze the chemical composition of the membrane surface with monochromated Al Kα radiation (Thermos Fisher Scientific, Germany). The colloidal solution zeta potential was measured using a Zetasizer (Malvern ZS90, UK). Original and pillared Ti₃C₂T_x membrane surface zeta potential measurements were carried out using an electrokinetic analyzer (SurPASS 3, AntonPaar, Austrian). Membranes were cut into 1 × 2 cm squares and taped on the measuring cell using a double-sided tape. The conductivity was measured at room temperature using a multifunction digital four-probe tester (ST-2258C, China). The distance between probes was 1.0 mm. The range of voltage used was from 0 to 4 V. The sucrose concentration was determined using the refractive index measured at room temperature with an Abbe refractometer (2WAJ, Japan) and converted based on the refractive index–sucrose concentration table provided by ICUMSA (International Commission for Uniform Methods of Sugar Analysis).

Synthesis of the Ti₃C₂T_x nanosheets

Ti₃C₂T_x nanosheets were prepared by selectively etching the precursor Ti₃AlC₂ as reported before. First, LiF (1.0 g) was added to HCl (20 mL 9 M), then the mixture was continuously stirred at 30 g for 20 min in water bath at 45 °C until the LiF was completely dissolved. Subsequently, Ti₃AlC₂ (1.0 g) precursor was gradually added into the etchant solution and reacted for 24 h under stirring. After the reaction was completed, the mixture was centrifuged at 1370 g for 5 min and washed by water until the expanded clay-like sediment appeared. The clay which contained multiple layers of Ti₃C₂T_x was exfoliated to the monolayer by manual shaking. The colloidal solution of single-layer Ti₃C₂T_x nanosheets was obtained after centrifugation at 1790 g for 1 h.

Preparation of the tetranuclear hydroxo zirconium solution

The Zr pillaring solution was synthesized by dissolving ZrOCl₂·8H₂O in deionized water. The solution was then aged and refluxed at 60 °C for 10 h in a water bath⁶⁰. To prevent further polymerization and aging, the resulting solution was immediately diluted to 0.1 M and used for subsequent membrane processing.

Fabrication of the Zr₄-Ti₃C₂T_x membranes

Diluted titanium carbide nanosheets were deposited onto a polyvinylidene fluoride (PVDF) membrane with a pore size of 0.22 μm under vacuum, yielding a two-dimensional original Ti₃C₂T_x membrane. The dried original Ti₃C₂T_x membrane was subsequently immersed in a pillaring solution for 2 h to facilitate the thorough intercalation of Zr₄ ions within the Ti₃C₂T_x nanochannels. Following this process, the Zr₄-Ti₃C₂T_x membrane was rigorously rinsed with deionized water to eliminate any residual unbound Zr₄ ions. The dried Zr₄-Ti₃C₂T_x membrane was then stored in a vacuum-sealed environment to preserve its integrity for subsequent voltage-gated ion and water transport tests. The membrane thickness could be effectively controlled by varying the amount of titanium carbide nanosheets deposited onto the substrate.

Ion penetration measurement

In this study, a homemade U-shaped device was employed to investigate voltage-gated ion transport behavior. The Zr₄-Ti₃C₂T_x membrane, secured with a circular platinum ring, was clamped between two Teflon gaskets, connecting the feed cell and the penetration cell. A three-electrode system was established, comprising a working electrode (pillared membrane), a reference electrode (Ag/AgCl), and a counter electrode (platinum net), allowing precise adjustment of the membrane surface potential. Both the reference and counter electrodes were immersed in the feed cell. For the ion permeation test, equal volumes (120 mL) of a salt solution and pure water were simultaneously introduced into the feed and permeate compartments to establish a concentration gradient that drives ion diffusion. A conductivity meter was placed in the permeate compartment to monitor ion concentration changes in real time. To mitigate concentration polarization effects at the membrane surface, two peristaltic pumps were utilized to exchange the solutions in the cells every 5 min. Ion concentrations during the experiment were calculated based on the relationship between the conductivity of the various salt solutions and their respective ion concentrations.

Desalination experiment

The forward osmosis desalination experiment was conducted using the same ion permeation test platform. The membrane with an effective area of 0.50 cm² was mounted in the module, separating two compartments with the membrane surface facing the feed side. Equal volumes (120 ml) of 0.05 M sodium chloride solution and 1.0 M sucrose solution were used as the feed and draw solutions, respectively. Water molecules from the NaCl solution were drawn into the sucrose chamber. The osmotic pressure (π) can be calculated using the van’t Hoff equation:

$$\pi={iMRT}$$

(1)

where π is the osmotic pressure, i is the ionization constant of the solute, M is the molar concentration of the solution, R is the gas constant (0.08206 L atm mol⁻¹ K⁻¹), and T is the absolute temperature (Kelvin) at room temperature.

Peristaltic pumps were employed to minimize the impact of concentration polarization on water permeation. The water permeance of the membrane was calculated using the following equation:

$${R}_{w}=\frac{{V}_{T1}-{V}_{T0}}{{S}_{m}{{\cdot }}\Delta T{{\cdot {{{\rm{bar}}}}}}}$$

(2)

where R_w is the water permeance of the Zr₄-Ti₃C₂T_x membrane, V_T1 and V_T0 are the volume of the feed side at the time points T₁ and T₀, respectively. ∆T is the time interval between two samples, S_m is the effective contact area between the membrane and the solution on both sides. To evaluate the diffusion performance of sucrose as a draw solution, the feed solution was replaced with ultrapure water under identical conditions. The sucrose content was determined from the refractive index of the solution measured using an Abbe refractometer⁶¹.

During the water flux experiment, the concentration of salt ions in the solution was measured by ICP-OES. The total molar amount of Na⁺ permeating through the Zr₄-Ti₃C₂T_x membrane in unit time was then calculated. The rejection rate Rs can be calculated according to the equation as follows:

$${R}_{s}=1-\frac{({C}_{p}+\Delta {C}_{p}) \times ({V}_{p}+\Delta {V}_{p})-{C}_{p}{V}_{p}}{{C}_{f}\Delta V}$$

(3)

where R_s is the rejection rate of the Zr₄-Ti₃C₂T_x membrane to Na⁺, C_p and C_f are the initial concentration of Na⁺ in the permeate side and in the feed side, respectively. ∆V_p was the concentrations of Na⁺ changed when the volume of the solution in the permeate side changed, V_p is the initial volume of the solution in the permeate side, and ∆V is the solution volume change in the permeate side.

DFT calculation

The DFT calculations were carried out using Vienna ab initio Simulation Package⁶² (VASP 5.4). The Perdew-Burke-Ernzerhof parameterization of generalized gradient approximation (GGA) is used for the exchange-correlation functional⁶³. The projector augmented wave (PAW) method is used for the ion–electron interaction⁶⁴. The kinetic energy cutoff is 500 eV. The semi-empirical Grimme’s DFT-D3 method with zero damping is used for the corrections of van der Waals interaction⁶⁵. For the structure optimization of the primitive cell of Ti₃C₂O₂, the Brillouin zone integration was performed using 17 × 17 × 1 Monk horst–Pack k-point sampling. For geometry relaxation, the convergence criteria for the total energy and Hellman-Feynman forces were 10–4 eV and 0.01 eV/Å, respectively. The lattice constant of the optimized structure of Ti₃C₂T_x is 3.024 Å, which is in good agreement with the experimental value of 3.057 Å⁶⁶. The vacuum layer thickness is set to 30 Å. The binding energy of ions on different sites of the Ti₃C₂T_x surface can be found in our previous work³⁷. The climbing image nudged elastic band (CI-NEB) method is used to calculate the diffusion barrier of ions⁴⁰. We constructed a 3×3×1 supercell of Ti₃C₂T_x with a single ion on the surface. In all CI-NEB calculations, five images are used among the initial (C site) and final state (C site). The fixed charge calculation was performed in this work to simulate the response under external electric potential as previous work⁶⁷. We can obtain Ti₃C₂T_x with about $\pm 0.9$ and $\pm 0.5$ V Vs electrode potential by changing $\mp \,2$ and $\mp \,1$ electrons in the system (Fig. 3c, d), respectively.

AIMD calculation

AIMD simulations were performed to study the structure of water and hydrated ions in the Ti₃C₂T_x nanochannel. We used an orthorhombic supercell (12.09 × 10.47 × 16.20 Å) of the Ti₃C₂O₂ monolayer, with interlayer spacing consistent with the XRD results (Fig. 1f). Periodic boundary conditions were applied in all directions. The system contains 32 water molecules (or 31 water molecules and one Na⁺ ion), corresponding to a water density of 1.0 g/m³. The initial structures were from the classical MD simulation as in previous work³⁷. The C and Ti atoms in Ti₃C₂T_x are fixed during the simulations. Simulations were carried out using the Nose–Hoover thermostat with a damping constant of 1 ps. In order to reproduce the structure of bulk water at ambient conditions (298 K, 1 atm), the higher temperature of 390 K was used to well reproduce the structure of liquid water^68,69. The time step used to integrate the Newtonian equations is 1.0 fs, with each simulation running for over 60 ps. The fixed charge method was performed to simulate the response under external electric potential. We were able to achieve Ti₃C₂T_x with about $\pm 0.5$ V electrode potential by changing $\mp \,3$ electrons in the system. The self-diffusion coefficient of water was calculated from the MSD extracted from simulated ion trajectories by using the Einstein relation in 2D, $D= < {|{{{\bf{r}}}}\left(t\right)-{{{\bf{r}}}}(0)|}^{2} > /4t$, where r(t) is the instantaneous position of O atoms projected to the x-y plane at time t, $< \ldots >$ is the thermodynamic ensemble average. Time averaging and statistical error analysis were performed based on 3 independent MD runs. The data from the first 10 ps were discarded for analysis. The geometry-based criterion used to analyze the HB network includes a distance between the oxygen atoms of both molecules smaller than 3.6 Å, and an angle defined within the dimer geometry (∠O-O-H) smaller than 30°⁷⁰.