Ion-electron coupling–enabled mechanical ion transistor with ultralow subthreshold swing Authors: Guoyuan Zhang, Linxin Zhai, Xin Peng, Song Luo, Lingfeng Yu, Qiang Chen, Bin Tu, Zhiping Xu, Xiang Zhang, Yahui Xue Published: 2026 ABSTRACT Biological mechanosensitive (BMS) ion channels can activate selective cationic permeation beyond a threshold driving force and switch sharply between on and off states that are essential for life functions. However, it remains a grand challenge to experimentally replicate such biological systems using artificial structures with distinct on-off states for life science, sensing, and energy applications. Here, we report a mechanical ion transistor that exhibits stepwise streaming-current versus pressure response through graphene nanochannels under electrostatic gating. The mechanical ion transistor exhibits an ultralow subthreshold swing (around 4.8 millibars per decade) defined by the minimum required pressure to increase the streaming current by 10 folds. Such a value is orders of magnitude lower than the current state-of-the-art of artificial structures and even sharper than most BMS channels. Our experimental observation of the threshold pressure and the current oscillation at the transition pressure evidences that the mechanical gating effect is attributed to the ionic Coulomb blockade induced by ion-electron coupling. The abrupt threshold transition behavior facilitates the mimicking of ultrasensitive detection of blood pressure pulsation in the olfactory bulb. The discovery here not only enables a better understanding of the functional mechanisms of BMS channels but also offers alternative routes for precise control of ion transport for applications in water filtration, energy harvesting, and neuromorphic sensing. INTRODUCTION Biological mechanosensitive (BMS) ion channels can mechanically gate cationic currents through cellular membranes (1, 2) and enable pivotal biological processes, including touch (3) and sound wave sensing (4), pain signaling (5), and interoception of blood pressure rhythms (6, 7). Artificial channel structures have been developed to replicate the acute mechanosensitivity of biological systems for applications in electrokinetic energy conversion (8–10) and sensory systems (11). Nonlinear coupling of electrical and pressure driving forces has been revealed to enhance the mechanotransduction efficiency in graphene capillaries (12), carbon nanotubes (13), graphene nanopores (14), and monolayer covalent organic frameworks (15). Stretching of elastic membranes can also boost the ion permeation through embedded nanopores (16, 17). However, BMS channels usually require a threshold mechanical force to activate the cationic permeation that eventually reaches saturation and manifest themselves with distinct on and off mechanogating states (18–20), which has rarely been experimentally observed in nanofluidic channels. Contrast conduction and stop bands have been theoretically predicted by an ionic Coulomb blockade (ICB) model for electrically driven ion transport through nanochannels with discrete charges (21, 22) or nanopores (23). It hints at a scenario to realize an on/off switchable ion permeation driven by a mechanical force, although there still lacks irrefutable evidence of experimental observation of ICB (24–27). Within the rapidly evolving paradigm of iontronic (28), the realization of such a mechanically gated ion transistor not only bridges the gap between biological mechanotransduction and artificial iontronic devices but also introduces a paradigm for effectively controlling nanofluidic transport. Here, we report a mechanical ion transistor that exhibits distinct on-off and step-function–like streaming current-pressure characteristics through graphene nanochannels under electrostatic gating. A threshold pressure is observed to initiate an abrupt increase of the streaming current in contrast to the linear response at open circuit conditions. The nanofluidic ion transport exhibits a mechanical on/off ratio reaching two orders of magnitude. We observe that the mechanical ion transistor exhibits an ultralow subthreshold swing (SS; around 4.8 mbar/decade), defined by the minimum required pressure to increase the streaming current by 10 folds. Such a value is comparable to or even lower than that of the most mechanosensitive biological channels. The experimental observation of the threshold pressure and the current oscillation at the transition pressure offers key features for the ICB (27, 29). The streaming current reaches a plateau beyond the threshold instead of continuously linear growth, attributed to the established resistive electric field caused by charge accumulation. Akin to Piezo1/2 channels (1, 2), preferential streaming permeation of Ca2+ over Na+ is also observed. We explore using the mechanical ion transistor to reproduce the ultrasensitive detection of the blood pressure pulsations by mitral cells. RESULTS Our devices were essentially graphene laminate membranes (GLMs) with precisely controlled atomic-scale channels of interlayer spacing, d (∼0.78 nm), fabricated as previously described (30, 31) through vacuum-filtration-induced self-assembly and subsequent capillary compression (figs. S1 to S5). The presence of π-π stacking in the laminates not only enables the good conductivity of the membrane but also maintains the structure stability even after electrostatic gating (fig. S2, E and F). The membrane device with a thickness of around 6 μm was connected to the working electrode of a potentiostat for the electrical potential control. It was clamped between two counterparts of a pressure electrochemical cell filled with salt solutions of the same concentration and containing Ag/AgCl electrodes for streaming current measurements (Fig. 1A and see fig. S6 for more details). The pressure (ΔP) was controlled by a pump connected to the cell, and the streaming current (Istr) was measured as the component of the current driven by the pressure (12–15). Experimental setup and mechanically gated ion streaming. (A) Schematic of GLM with subnanometer channels clamped between two solution reservoirs containing Ag/AgCl electrodes for streaming current measurements. A pump applied pressure (∆P) across the membrane, and an electrostatic gating voltage (Vg) was applied to control the electrode potential of the GLM. CE, counter electrode; WE, working electrode; RE, reference electrode. (B) Schematics of ionic streaming at the open circuit potential (OCP) (top) and Vg = −0.9 V (bottom). At OCP, the preferential adsorption of cations by π-bonds on graphene surfaces leads to the anion-dominated streaming current. At Vg = −0.9 V, the negatively charged graphene nanochannels are predominantly filled with cations. (C) Streaming current (Istr) response and pressure (∆P) control scheme at OCP and Vg = −0.9 V in 50 mM CaCl2. The pressure ranged from 5 to 50 mbar, and each pressure step was applied for an interval of 15 s and repeated four times. (D) Streaming current response with driving pressure through the electrostatically gated GLM (Vg = −0.9 V). Inset: Open probability (P0) of the mechanosensitive channel of large conductance, MscL, under varied tension [data are adapted from (19)]. (E) Comparison of subthreshold swing (SS) for our mechanical ion transistor with artificial nanochannels in the literature and typical BMS channels. SS represents the minimum required driving pressure to induce a 10-fold increase of the streaming current (fig. S10). CNT, carbon nanotubes; COF, covalent organic frameworks; TRAAK, TWIK-related Arachidonic Acid-stimulated K⁺ channel. To explore the effect of electrode potential on the ionic streaming through our graphene channels, we probed the transport characteristics of aqueous electrolyte purely driven by pressure at the open circuit potential (OCP) and under electrostatic gating control (see the “Streaming current measurements” section). Graphene spontaneously attracts cations on the surface due to π-bond interactions (32) at open circuit conditions (Fig. 1B, top). Before each measurement, overnight soaking of GLM in the testing solution had been carried out to ensure the cation adsorption reached equilibrium. At OCP, using 50 mM CaCl2 as an example, we observe that the streaming current (Istr), dominated by the anions, increases with the driving pressure (ΔP) (Fig. 1C), resulting in a linear current-pressure characteristic (fig. S7A). It indicates that the pressure-induced membrane deformation shows negligible effects on the streaming current in the present pressure range. A similar behavior is also found for 100 mM KCl (fig. S8A). This indicates that the energy barrier for ionic entry inside the GLM barely affects the streaming permeation. Electrostatic gating voltage (Vg) was applied to the GLM to control the graphene surface potential to mimic the charge groups embedded in biological channels (33, 34). For Vg = −0.9 V, the negative charges inside the atomic-scale graphene channels lead to the dominant transport of cations (Fig. 1B, bottom). In this configuration, distinct blocked and open states of the streaming current are unexpectedly observed, in contrast to the linear response at OCP (Fig. 1C). A threshold ∆P is identified to initiate an abrupt increase of Istr. Beyond the critical ∆P, Istr increases only slightly and tends to plateau, manifesting the mechanically activated ion permeation with a step-function–like streaming current-pressure characteristic (fig. S7B), akin to that observed in BMS channels (Fig. 1D and fig. S9). The SS of our mechanical ion transistor, defined by the minimum required pressure to induce a 10-fold increase of Istr, reaches as low as 4.8 mbar/decade. We also probed the device response under Vg = −0.9 V in artificial cerebrospinal fluid (see table S1 for the components) and observed an SS of 6.21 mbar/decade (figs. S9A and S10A). Such an SS is orders of magnitude lower than the current state-of-the-art of artificial structures and even steeper than those of most BMS channels (Fig. 1E and fig. S10), indicating the ultrahigh mechanical sensitivity of the GLM under electrostatic gating. To gain more insight into the surface charge effect on the mechanically gated ion transport in graphene nanochannels, we investigated the streaming current under different electrostatic gating voltages. The step-function–like variation of Istr with ∆P is observed for both negative and positive Vg (Fig. 2A). The magnitude of the gating voltage enhances the plateau streaming current but does not clearly alter the threshold transition pressure in the measurement accuracy. A plot of the Vg-dependent streaming current at varied driving pressure evidences the distinct on and off states (Fig. 2B), revealing a threshold pressure within (10 and 15 mbar). Such a threshold pressure range maintains quite consistent for a period of 2 weeks (fig. S11). Near the transition pressure, we observe clear streaming current oscillation, e.g., at ∆P = 12 mbar and Vg = 0.6 V (Fig. 2C and fig. S12), indicating that an energy barrier exists to obstruct the ion permeation. Control experiments (fig. S12) exclude these oscillations arise from pump noise, bubble dynamics, or membrane deformation. Beyond the threshold pressure (∆P ≥ 15 mbar), pronounced streaming current increases with the gating voltage, corresponding to the on state. Below the threshold (∆P ≤ 10 mbar), the streaming current is much diminished, indicating the off state. The mechanical on/off ratio of the streaming current, defined by the values at ∆P = 50 and 5 mbar, respectively, generally increases with the gating voltage, and reaches a maximum value of 133 at Vg = −0.9 V (Fig. 2D). The asymmetric switching ratio between opposite gating polarities is attributed to the different permeating ion species. The step-function–like streaming current-pressure characteristics are also observed using KCl solutions with different concentrations (fig. S13). Electrically controlled ionic streaming. (A) Streaming current (Istr) as a function of the pressure (ΔP) under different electrostatic gating voltage (Vg). The electrolyte was 50 mM CaCl2. (B) Istr as a function of Vg under different ΔP. Distinct on and off states are manifested again. (C) Streaming current oscillation near the threshold transition pressure (ΔP = 12 mbar) at Vg = 0.6 V. The streaming current at ΔP = 5 mbar, 10 mbar, and 15 mbar is much smoother. The current oscillation is attributed to the ICB effect and thus termed as Coulomb oscillation (29). (D) Mechanical on/off ratio at different Vg. The on and off streaming currents are defined by values at ΔP = 50 and 5 mbar, respectively. We then probed how the surface charge affects the pressure-driven ion transport in atomic-scale graphene channels. The step-function–like streaming current-pressure characteristics (Fig. 2A) and the current oscillation near the transition pressure (Fig. 2C) indicate that those surface charges exhibit strong Coulomb interactions with permeating ions under extreme physical confinement. The established energy potential well entraps oppositely charged ions and resists the ionic permeation according to the so-called ICB effect (Fig. 3A) (26, 27). The frictional dragging force exerted by the pressure-driven fluid flow enables the entrapped ions to conquer the energy barrier, leading to pronounced stream current beyond a threshold pressure (Fig. 3B). We adopted a one-dimensional ICB model based on the Wien effect theory (22) considering many-body dynamic of ions to interpret the mechanogating effect in electrostatically gated graphene nanochannels (see the “ICB model” section). The independence of the threshold transition pressure on the electrostatic gating voltage (Fig. 2A) indicates that the ionic Coulomb interaction with surface charges (i.e., doped electrons or holes) is discretized (35). The gating voltage increases the number of discrete interaction pairs and thus enhances the plateau streaming current (fig. S14A). The dissociation of attracted cations from paired electrons on the graphene surface (e.g., at Vg = −0.9 V) contributes to the current generation. The ICB model predicts the existence of a threshold pressure, below which the Coulomb attraction dominates the energy barrier for ion transport, indicating the off state, and beyond which pronounced streaming current is observed, corresponding to the on state (fig. S14B). However, the predicted streaming current after activation shows a much more pronounced linear increase than our experimental observation, implying that other rate-controlling mechanisms are involved in the ionic permeation. Mechanisms of mechanically gated ionic streaming and current saturation. (A) Schematic of blocked ionic streaming at Vg = −0.9 V when ∆P is below 10 mbar. The deep Coulomb attraction potential well resists the ion permeation due to the cation-electron bonding. (B) Schematic of activated ionic streaming at Vg = −0.9 V when ∆P is above 15 mbar. The attraction potential well becomes shallow with enhanced driving pressure, initiating the ion permeation. (C) Schematic of co-ion enrichment and resistive electric field establishment at the ends of highly cation permeable graphene nanochannels at Vg = −0.9 V. The preferential and pronounced permeation of cations through the negatively charged nanochannel leads to the accumulation of anions at the channel entrance and the establishment of a resistive electric potential. (D) Nonlinear streaming current response predicted by the PNP-NS/IC model at high surface charge density (−0.1 C/m2). The inset shows the linear response at a low surface charge density of −0.1 mC/m2 in contrast. To understand the underlying mechanism of streaming current saturation in the on state, we performed finite element (FE) simulations based on a modified Poisson–Nernst–Planck–Navier-Stokes model with ion correlation (PNP-NS/IC) (36–38) (see the “Poisson–Nernst–Planck–Navier-Stokes model with ion correlation” section and figs. S15 to S20). Under electric potential gating (e.g., at Vg = −0.9 V), the predominant filling of cations in atomic-scale graphene channels promotes their effective diffusivity due to the Coulomb repulsion–enabled concerted ion movement (33). Our FE simulations manage to reproduce the nonlinear response of the streaming current at high surface charge density (−0.1 C/m2) and reveal that the ionic streaming through a highly cation-permeable membrane leads to the enrichment of anions at the entrance of the nanochannels under elevated driving pressure (Fig. 3, C and D). The establishment of a resistive electric field due to the concentration polarization at the channel ends suppresses the streaming current enhancement with driving pressure, resulting in the streaming current plateau (fig. S21). We further examined the ionic selectivity in the mechanically gated graphene nanochannels. To reasonably compare the streaming permeability of different ions, the same ionic strength of electrolytes was chosen, that is, 100 mM for monovalent ions and 50 mM for bivalent ions. The step-function–like streaming current-pressure characteristics are observed with all tested solutions, including KCl, NaCl, LiCl, MgCl2, and CaCl2 (Fig. 4A and fig. S22). An approximately constant threshold pressure is observed within the range from 10 to 15 mbar, and the plateau streaming current generally increases with the gating voltage. It is interesting to remark that we observe a selection sequence of Mg2+ > Ca2+ > Li+ > Na+ > K+ at Vg = −0.9 V, which follows the size order of their hydrated shells (Fig. 4B). This contrasts with the scenario driven by concentration gradients (32) or electric fields (39) where larger hydrated ions exhibit higher resistance to transport inside nanochannels. The selection sequence can be explained by the frictional dragging force experienced by the ions in the streaming flow. That is because the water-ion friction coefficient is proportional to the ionic hydration size according to the Einstein relation (40) (see also fig. S23). The preferential permeation of Ca2+ over monovalent ions through our graphene channels is also consistent with the observation in Piezo1/2 channels, which triggers downstream Ca2+ signaling pathways (2). The on/off ratio of mechanical gating follows a similar sequence with the ionic selectivity, and the maximum value is observed for Mg2+ reaching ∼205 (Fig. 4C). Moreover, under a positive gating voltage (e.g., Vg = 0.9 V), we observe consistency in the streaming current magnitude and only slight difference for the bivalent and monovalent salts as the streaming current is dominated by the transport of chloride ions (Fig. 4D). We attribute the pronounced streaming current of Cl− over monovalent alkali metal ions to the stronger cation-π interactions on graphene surface (32). It is remarked that the characteristic step-function–like response persists for other anion-dominated transport such as SO42− (see fig. S22D), which exhibits weaker transport efficiency compared to Cl− due to the stronger ion-ion correlation effects (37). Ionic selectivity of mechanically gated graphene nanochannels. (A) Comparison of streaming current (Istr) as a function of ΔP for different salts (i.e., 50 mM MgCl2, 50 mM CaCl2, 100 mM LiCl, 100 mM NaCl, and 100 mM KCl solution) at a negative gating voltage of Vg = −0.9 V. (B) Relative selectivity of cations with different hydration radius with respective to K+ at Vg = −0.9 V. (C) Comparison of mechanical on/off ratios for different cations under Vg = −0.9 V. (D) Comparison of Istr as a function of ΔP under a positive gating voltage of Vg = 0.9 V for different salts. The streaming current is dominated by Cl−. We also probed the capability of our mechanically gated graphene channels to detect pulsive differential pressures. The BMS channels (most likely Piezo2) of mitral cells in the olfactory bulb of the brain have been revealed to play a critical role in monitoring the blood pressure pulsations by transducing them into neural oscillation signals (Fig. 5A) (7). Note that the normal intracranial pressure ranges from 5 to 15 mmHg (41), which is close to the activation pressure of our graphene channels (i.e., 10 to 15 mbar). We applied sinusoidal pulse pressures at 1 Hz and 2 Hz with an amplitude of 2 mbar and different static values across the GLM to explore the streaming current response. Typically, the GLM at OCP can hardly detect the pulse pressure at both frequencies when using 100 mM KCl as the electrolyte. In contrast, pronounced streaming current oscillations are observed at Vg = −0.9 V under a static pressure of 12 mbar, as confirmed by the fast Fourier transform (Fig. 5B). By scanning the static pressure from 5 to 15 mbar, it is demonstrated that significantly enhanced sensitivity for pulse pressure detection is achieved near the threshold transition pressure (Fig. 5C). Even higher detection sensitivity near the transition pressure is observed when using 50 mM CaCl2 as the electrolyte (Fig. 5D), in consistency with the preferential streaming transport of Ca2+ over K+ through the GLM (Fig. 4). The amplitude of the streaming current response (ΔI) increases with the gating voltage (fig. S24). The above observation hints that the ICB-enabled mechanical ion transistor could significantly enhance the detection sensitivity of differential pressures. We further tested the device response to mechanical stimuli of other waveforms, such as sawtooth and triangular pulses (fig. S25). The device effectively reproduced these nonsinusoidal waveforms, indicating that the device has high mechanical sensitivity and holds potential for practical physiological signal detection. The comparison of our GLM device with existing bioinspired ionic devices in the literature demonstrates its superior performance in sensitivity and responsive pressure (see fig. S26 and table S2). Bionic blood pressure pulsation perception. (A) Schematic of intracranial blood pressure perception by BMS ion channels in brain neurons. The activation of BMS channels (most likely Piezo2) by pressure pulsation produces a mechanically sensitive ionic current to induce nerve activity. (B) Streaming current responses of our mechanically gated graphene nanochannels to 1-Hz (left) and 2-Hz (right) sinusoidal pulsation pressures at a static value of 12 mbar under OCP and Vg = −0.9 V. The average fast Fourier transform (FFT) was used to analyze the streaming current signals. The electrolyte was 100 mM KCl. (C and D) Streaming current amplitudes (ΔI) responding to 1- and 2-Hz pressure pulsations with different static values in 100 mM KCl (C) and 50 mM CaCl2 (D) solutions under OCP and Vg = −0.9 V. DISCUSSION Our mechanical ion transistors can gate ionic streaming effectively and selectively. The step-function–like stream current-pressure characteristic and the current oscillation near threshold transition pressures offer key features for the ICB effect. The streaming current plateau beyond the threshold is attributed to the co-ion enrichment at the entrances of highly permeable and charged graphene nanochannels. The higher frictional dragging force in streaming flow for ions with larger hydration shells enables the preferential transport of bivalent alkali metal ions over monovalent ones, akin to that of Piezo1/2 channels. The abrupt streaming current enhancement near the threshold pressure enables the ultrasensitive detection of pulsive differential pressures. Our experimental system provides an intriguing platform to mimic the behavior of BMS channels that exhibit a threshold mechanical force to activate the ion permeation, which tends to plateau beyond the activation for purposes like the self-protection of cells. This demonstration opens a promising avenue to precisely control fluidic transport under extreme confinement for developing next-generation ion filtration, energy conversion, and biomimetic sensing technologies. MATERIALS AND METHODS Fabrication of GLMs The GLMs were fabricated by vacuum filtration of chemically reduced graphene oxide (GO) aqueous dispersion and subsequent capillary compression (fig. S1) following the previously described procedure (30, 31, 37). For the chemical reduction process (42), the commercially obtained GO aqueous dispersion (10 mg/ml, from Sichuan Kenye Technology Development Co. Ltd.) was diluted to a concentration of 2 mg/ml. This diluted GO solution (10 ml) was mixed with 4 ml of an aqueous vitamin C solution (17.6 mg/ml) in a glass flask. Then, 0.2 ml of ammonia solution (25 to 28 wt %) and 90 ml of ultrapure water were added to achieve a pH of ∼10. The mixed solution was immersed in an oil bath maintained at 95°C for 40 min to obtain black reduced GO (or rGO) nanosheet dispersion with a concentration of ∼0.2 mg/ml. To precisely control the interlayer spacing of the GLM, 100 ml of the rGO nanosheet dispersion (0.2 mg/ml) was vacuum filtered through a mixed cellulose ester (MCE) substrate with a diameter of ∼50 mm and an average pore size of about 0.22 μm. The vacuum was disconnected immediately after all free rGO nanosheet dispersion was gone from the filtrate cake, leaving a wet gel membrane. The gel membrane was then immersed in ultrapure water overnight to remove the remaining ammonia and unreacted vitamin C. Subsequently, the cleaned GLM underwent a 12-hour immersion in 0.1 M sulfuric acid, followed by sandwiching between two glass slides and drying at room temperature for another 12 hours. It was then further dried under an ultrahigh vacuum for 12 hours. Last, the GLM, featuring an interlayer spacing of 0.8 nm, was peeled off from the MCE substrate to achieve free suspension for streaming current measurements. To ensure sufficient mechanical robustness and integrity of the freestanding membranes, the GLM mass loading of all samples was controlled to be around 1.59 mg/cm2. Characterization of GLM To characterize the internal channel structure and stability of the GLM membrane, we used a suite of analytical techniques, including atomic force microscopy (AFM), scanning electron microscopy (SEM), energy-dispersive spectroscopy (EDS), x-ray diffraction (XRD), x-ray photoelectron spectroscopy (XPS), and Raman spectroscopy. AFM (Cypher S, Oxford Instruments, USA) confirmed the monolayer thickness of the rGO flakes in the dispersion (fig. S1, A to C). SEM (Merlin, ZEISS, Germany) observation reveals the thickness of the GLM around 6 μm (fig. S1D). The average interlayer spacing of the GLM was found to be around 0.78 nm by XRD (Smartlab, Rigaku, Japan) examination, close to previous reports (30, 31, 37), and kept stable even after electrostatic gating at Vg = −0.9 and 0.9 V. The additional peak in the XRD spectrometry indicates the existence of π-π stacking between graphene layers that maintains the structural stability (29). The constant peak positions in the Raman (XploRA PLUS, HORIBA, France) spectroscopy of the GLM before and after electrostatic gating at Vg = −0.9 V also confirm the structure stability of the membrane. XPS (AXIS SUPRA+, SHIMADZU/Kratos, Japan) analyses show that the GLM exhibits a C/O ratio around 4.01, whereas the value for original GO is 2.29 (fig. S3). EDS (Merlin, ZEISS, Germany) elemental scanning of the cross sections of the GLM confirms the spontaneous intercalation of monovalent and bivalent metal ions inside the graphene nanochannels by purely soaking the corresponding solutions (fig. S4). A negative gating voltage of Vg = −0.9 V attracts more cations inside the channels, and a positive gating voltage of Vg = 0.9 V brings more anions inside the channels (fig. S5). The high local cation content in fig. S5 reflects the preferential cation entrapment within the subnanometer channels, attributed to the strong cation-π interactions under extreme confinement (43). Streaming current measurements The GLM was sandwiched between two specially designed resin clips, forming a three-layer structure. Each custom clip had a central hole with a radius R = 2 mm, and the alignment between the two holes of the two clips was ensured during assembly. This sandwich structure was clamped between two sealed pressure electrochemical half-cells, with flexible gaskets on either side to ensure sealing integrity (fig. S6). Platinum counter electrodes and Ag/AgCl reference electrodes were placed near the GLM (i.e., the work electrode), constituting a three-electrode system, with electrostatic gating voltage applied via a Keithley 2400 source meter (Keithley Instruments). A high-precision gas pressure pump (OB1, Elveflow) imposed a gas pressure to drive the electrolyte solution streaming through the GLM. The ionic current through the membrane was monitored by using a pair of Ag/AgCl electrodes connected to a Keithley 6487 source meter (Keithley Instruments). The streaming current (Istr) was recorded as the variation component driven by the pressure (ΔP). The Istr was defined as the pressure-driven variation component, obtained by subtracting the zero-pressure baseline from the steady-state plateau current (Istr = Iplateau − Ibaseline), and the Vg was held constant for sufficient time to ensure electrochemical equilibrium before pressure scanning, eliminating interference from Vg-induced potential shifts. Before conducting the ionic streaming experiments, deionized water was filled into the two-compartment electrochemical cell to soak the membrane for 24 hours, thereby removing residual sulfuric acid from the GLM. Then, 50 ml of the test solution was injected into each compartment, allowing for a 12-hour soak to ensure an equilibrium distribution of ions within the channels. Our experiments were primarily focused on simple alkali metal cations, including Li+, Na+, K+, Ca2+, and Mg2+, which were paired with Cl−. Each set of experiments was conducted with a stepwise changing ∆P from −50 to 50 mbar, and each ∆P level was repeated four times with an interval of 15 s. Istr was observed with initial overshot and then quickly equilibrated to a stable value upon applying a ∆P. A similar phenomenon has been observed previously (12–15). The magnitude of Istr shows good reproducibility driven by a repeated constant pressure (Fig. 1C and fig. S7). This consistency indicates that the membrane structure maintained integrity in the pressure measurement range. Given the macroscopic reservoir volume of 50 ml and the relatively tiny permission area of the GLM, we did not observe clear electrolyte level change during the pressure streaming tests. ICB model We adopt a one-dimensional ICB model based on the Wien effect theory, considering many-body dynamics of ions (22) to interpret the mechanogating effect of our graphene nanochannels. The channel has height d, equal length, and width L >> d. In electrostatically gated graphene nanochannels (e.g., at Vg = −0.9 V), the predominant cations (e.g., Ca2+) form tightly bound pairs with discrete surface charges or electrons due to low dielectric permeability in confined space (40). The dissociation of those cation-electron pairs under frictional dragging flow generates a streaming current. The interaction potential between cations and electrons is assumed to follow the exponential function (22)U(x)=fkBTξxTe−∣x∣ξ(1)where xT is a thermal length quantifying the interaction strength, ξ is a characteristic length scale over which the electric field produced by an ion is confined, and f is a parameter characterizing the effective interaction strength. If the permittivity of confined water is the same as in the bulk, ξ has a length scale around 3.5d, and xT ≈ d2/8lB, where lB is the Bjerrum length and equals 0.7 nm in bulk water. Here, kB is the Boltzmann constant and T is the temperature. The ion transport through nanochannels experiences frictions exerted by the wall and water molecules, which can be described by a normalized water-ion friction coefficient (α) (12). Note that the ion permeation through negatively voltage-gated graphene channels is dominated by the cations. The ion velocity (v) can be obtained by analyzing their local balance, and in the absence of an electric fieldv=αKapp(−∂xP)(2)where ∂xP is pressure gradient, Kapp is the hydraulic permeability that depends on the channel height, and Kapp≈bd/2μ, where μ is dynamic viscosity and b is slip length and chosen as 20 nm here. To calculate the ionic current driven by a pressure gradient, we estimate the mean escape time (τ) of a bound K+ from the potential well described by Eq. 1. The analytical solution obtained by solving the Fokker-Planck equation with an absorbing boundary condition is adopted here (22), that isτ(q,Eeq)=1D±∫−∞+∞∫max(0,x)ξlogqEeqxTeq[U(x)−U(y)]+e Eeq(x−y)kBTdydx(3) Here, q = 1 denotes the discrete electronic charge, and Eeq = kBT v/(D qe) is the equivalent electric field that drives the ion motion, where D is diffusion coefficient and e is elementary charge. Assume that the average distance between the discrete cation-electron pairs is λ [∼1 nm at Vg ≈ −1 V (31)], which is assumed to decrease with the electrical gating voltage. When a cation dissociates from the surface charge, the average free movement time (tf) equals λ/DEeq before it binds to the surface charge again. The ion free movement contributes If to the ionic current and, If = 2e/tf for Ca2+. If the total number of cations is Ntot, then the number of free cations (Nf) can be estimated by (22)Nf=tf2 τ(1,Eeq)1+4Ntotτ(1,Eeq)tf−1(4) The mean escape time (τ) generally decreases with the driving pressure, and a zero value of τ indicates that the ion permeation through the electrically gated graphene channels becomes dominated by the free ion diffusion, where Nf equals Ntot. Accordingly, the generated streaming current is obtained byIq(Eeq)=NfIf(5) Equations 4 and 5 were used to fit the experimental data of streaming current through GLM at Vg = −0.9 V using 50 mM CaCl2 as an example. The plot is shown in fig. S6. Poisson–Nernst–Planck–Navier-Stokes model with ion correlation For pressure-driven ion transport in graphene nanochannels, the dielectric permittivity of the confined space and the hydrodynamic slip at the graphene surface need to be considered in FE modeling simulation (COMSOL Multiphysics) (36–38). Simulations of ionic transport through the membrane were carried out using the modified and coupled Poisson-Nernst-Planck (PNP) equations. The laminar flow model was used to simulate the contribution of hydrodynamic terms to the mass transfer process. This model was simplified from the Navier-Stokes equationsρ∂u∂t+ρ(u·∇)u=−∇p+μ∆u(6)ρ∇·u=0(7)where ρ denotes the density of water, u represents the velocity vector, p stands for pressure, and μ is the dynamic viscosity. The total flux of each ion species can be expressed as follows∂ci∂t+∇·Ji+u·∇ci=0(8)Ji=−Di∇ci−zium,iFci∇ϕ−DiNaa3ci∇(cp+cn)1−Naa3(cp+cn)(9)where a=0.66 nm, representing the size of K+ and Cl− ions in KCl aqueous electrolyte; Na is Avogadro’s number; Di,um,i,zi,and ci are the diffusion coefficient, electrophoretic mobility, valence, and concentration distribution of species i, respectively. The last term in Eq. 7, also known as the entropy term, was added as a correction for finite ion size. This ensures that the maximum ion concentration in the reservoir (cp+cn) does not exceed cmax=1/Naa3 (37). It is noted that the electric double layers (EDLs) exhibit oscillatory layer structures within a few nanometers at the electrode-electrolyte interface. Considering that overscreening in the EDL leads to more concentrated co-ion layers, we adopted a PNP model considering ion correlation (PNP/IC) (36). This model accounts for the presence of short-range ion correlations to explain the oscillatory EDL structure. The governing equations areε0ε∇2ϕ−ε0εlc2∇4ϕ=−∑i=iNzieNaci(10)n·(ε0ε∇ϕ)=σs(11)where ε0 is the vacuum permittivity, ε is the relative permittivity of the aqueous solution, σs is the surface charge density of the membrane, and n is the unit normal vector pointing toward the membrane surface. Compared to the traditional Poisson equation, an additional fourth-order Laplacian operator was added to correct for ion-ion correlations and the correlation length lc=0.66 nm (KCl aqueous electrolyte). A two-dimensional domain was constructed for simulation and divided into three parts: Ω1,Ω2, and Ω3. In Ω2, three key parameters were set as follows: graphene channel height (interlayer spacing) d=0.8 nm, surface charge density σs=−0.1 C/m2, and the slip length b=20 nm, as shown in fig. S15. Prior molecular dynamics simulations have shown that the slip length on graphene surfaces is only weakly dependent on the surface charge density that is homogeneously distributed (44). Therefore, the slip length is still selected as 20 nm when the surface charge density changes. Moreover, the ion concentrations at the Γ1 and Γ4 boundaries were set to their bulk values, cp=cn=c0=100 mM, with Γ1 being the pressure inlet boundary (P=ΔP) and Γ4 being the pressure outlet boundary (P=0). At the Γ2 and Γ3 boundaries, impermeable boundary conditions were applied to both concentration and electric field. Considering that strong nanoconfinement would lead to a reduced dielectric constant, we chose the dielectric constant ε in Ω2 equal to 10 while maintaining the dielectric constant in Ω1 and Ω3 at 78.5 (45). Meanwhile, significantly enhanced diffusion coefficient of capacitively charged ions in high density has experimentally observed in electrostatically gated graphene nanochannels due to the Coulomb-interaction induced concerted ion movement (33). Accordingly, we set the cation diffusion coefficient D+=1.957×10−7 m2/s in Ω2 and the anion diffusion coefficient D−=2.032×10−11 m2/s in Ω2, whereas the D+=1.957×10−9 m2/s and the D−=2.032×10−9 m2/s in Ω1 and Ω3. This simulates the situation of highly preferential and pronounced transport of cations through the negative voltage gated graphene nanochannels. The Coulomb repulsion–enabled concerted ion movement, which enhances cation diffusivity, is irrelevant to the driving force (e.g., pressure or concentration gradient) and originates from the highly packed charge density induced by electrostatic gating in subnanometer channels (33). Equations 6 to 11 were coupled to calculate the ion concentration and electrostatic potential with given boundary conditions. We set c0=100 mM, σs=−0.1 C·m−2 for high surface charge density and −0.1 mC·m−2 for low surface charge density. The concentration distribution of K+ and Cl− and the corresponding electrostatic potential were obtained by solving Eqs. 6 to 11. We used the following dimensionless variables to obtain dimensionless results (13)P(adim)=1NAc0kBT∆P≈∆P/2.4 bar(12)ϕ(adim)=ekBTϕ≈ϕ/25 mV(13)J(adim)=Lc0DJ≈J/0.1 mol·(m2·s)−1(14) The results are shown in figs. S16 to S21. Bionic pressure pulsation measurement We used the same apparatus for streaming current measurement to emulate blood pressure pulsations by applying sinusoidal pressure functions. The pneumatic pressure was increased in 1-mbar steps from 5 to 15 mbar, with a dwell time of 120 s at each static pressure plateau. Sinusoidal pressure pulsations with an amplitude of 2 mbar at 1 and 2 Hz were superimposed on the static pressure platform. A pair of Ag/AgCl electrodes was used to record the transmembrane ionic current of the device, capturing streaming current responses corresponding to the pressure pulsations. The amplitude of the pulsatile streaming current response was extracted as a metric of detection sensitivity of pulse differential pressures, which shows dependence on the static pressure and electrostatic gating voltage (fig. S24).