Graphene‑Sheet‑Driven Nanopump: Harnessing Brownian and Directed Motion to Generate Net Water Flux through Carbon Nanotube Channels

Background

Seawater desalination via distillation is energy‑intensive, while reverse osmosis (RO) remains the industry standard yet still demands high pressures to drive water through semipermeable membranes [1]. To address this, researchers are exploring nanomaterials such as carbon nanotubes (CNTs) that promise high selectivity, low resistance, and near‑thermodynamic performance [6–10]. However, conventional membrane improvements alone cannot overcome the energy ceiling; an alternative driving mechanism is required [11].

In a CNT connecting two reservoirs, water molecules enter spontaneously from both ends due to Brownian motion. Because the two influxes cancel, no net flow results. Conventional strategies enhance one side’s entry via pressure gradients, temperature differences, or electric fields [12–16]. Instead, we propose weakening the opposing side’s competitiveness by placing a mobile graphite sheet on one membrane surface, thereby breaking the symmetry without external pressure.

Controlling nanofluidic transport is critical for diverse applications—from energy storage to biosensing [17–23]. Our design introduces a simple, tunable element: a small graphite sheet that can either diffuse thermally or be driven unidirectionally. MD simulations confirm that the Brownian mode yields a net flux similar to aquaporin channels (≈1.8 ns⁻¹) [24–25], while the driven mode can increase flux by an order of magnitude, depending on the applied force. This approach paves the way for energy‑efficient, high‑throughput water pumps.

Model and Simulation Method

The simulation cell (Fig. 1) comprises a (6,6) CNT (2.56 nm long, 0.81 nm diameter) sandwiched between two parallel graphite sheets (5.1 × 5.1 nm²). A 272‑atom graphite nanoshift (the “small sheet”) adheres to the upper membrane due to strong C–C Lennard‑Jones interactions, maintaining an average separation of 0.34 nm. In the Brownian mode, the sheet’s temperature is varied between 100 and 500 K while water remains at 300 K. For the driven mode, each carbon atom experiences an additional acceleration of 0.1 nm ps⁻² (≈2 pN) along the x‑axis.

Water is modeled with TIP3P, and all interactions use established Lennard‑Jones parameters [6]. Periodic boundary conditions guarantee continuous flow. The system is equilibrated at constant volume and temperature (NVT) for 125 ns using GROMACS 4.6.5 (2 fs timestep), with the last 120 ns used for analysis. Two independent runs confirm statistical robustness.

Snapshot of the simulation system: a CNT connecting two water reservoirs, separated by two graphite sheets (sage green, 5.1 × 5.1 nm²). A small blue graphite sheet sits on the upper membrane, creating a nanometer‑scale pump.

Results and Discussion

Brownian Motion of the Graphite Sheet

We first quantified the net flux induced by the thermally diffusing sheet. Using the convention of upflux ( +z ) and downflux ( –z ), the flow (sum) and flux (difference) are defined as in previous studies [31,32]. Fig. 2 shows that, regardless of sheet temperature, the system consistently exhibits a down‑to‑up net flux of ~2 water molecules ns⁻¹—comparable to the 1.8 ns⁻¹ measured in aquaporin channels [24,25]. The total flow remains essentially temperature‑independent, indicating that the sheet’s Brownian motion primarily weakens the top side’s competitive entry rather than heating the water.

Water flux and flow versus sheet temperature (error bars from two independent runs).

Analysis of translocation times and occupancy (Fig. 3) reveals that the average residence time of water within the CNT remains ~10 ps, with ~10 molecules occupying the channel—consistent with the single‑file structure. Density and hydrogen‑bond (H‑bond) profiles (Fig. 4) confirm a four‑fold density increase inside the CNT and a reduction in H‑bond number as water enters the confined space. Dipole orientation (Fig. 5) shows two symmetric peaks (20°–40° and 140°–160°), underscoring the highly ordered chain.

Translocation time τ and occupancy ⟨N⟩ versus sheet temperature.

Density ρ and H‑bond number along the z‑axis at different sheet temperatures (ρ₀ = 1.0 g cm⁻³).

Probability distribution of the average dipole orientation inside the CNT for various sheet temperatures.

Unidirectional Motion of the Graphite Sheet

When the sheet is propelled by a constant force, both flow and flux rise sharply (Fig. 6a). The sheet velocity (Vₓ) scales linearly with the driving force, while the effective friction coefficient ξ decreases (Fig. 6b), confirming that the sheet’s drag on surrounding water weakens the top side’s competitiveness. At forces above 1.6 pN, the flux plateaus at ~16 ns⁻¹—an eight‑fold increase over the Brownian case.

a Flux, flow, and efficiency η versus driving force. b Sheet velocity Vₓ and friction coefficient ξ versus force.

Translocation time and occupancy (Fig. 7) exhibit a linear decline and increase, respectively, with force—reflecting the accelerated passage of water molecules under the sheet’s drag. Density, H‑bond, and dipole profiles (Fig. 8) remain largely unchanged, indicating that the single‑file chain is preserved even under driven conditions.

Translocation time τ and occupancy ⟨N⟩ versus driving force.

a Axial density and H‑bond number along z for different forces. b Dipole orientation distribution under varying forces.

Additional Discussion

The sheet’s initial placement on the membrane (fixed at 0.34 nm) guarantees asymmetric interaction; the sheet–CNT distance remains constant for Brownian motion but increases with force in the driven case (Fig. 9). This behavior explains the observed flux enhancement. If the sheet were placed in bulk water, the symmetry would be restored and the bias transport would vanish.

Mean sheet‑membrane and sheet‑CNT distances for Brownian (a) and driven (b) motions.

Temperature analysis (Fig. 10) shows that the sheet’s average temperature follows the target setpoint, while the surrounding water remains at 300 K—despite the thermostat’s suppression of heat exchange, collisions from the moving sheet subtly perturb local water dynamics, facilitating the bias flow.

Mean temperatures of sheet and water versus target sheet temperature.

Conclusions

We have demonstrated a low‑energy nanofluidic pump that leverages the spontaneous permeability of CNTs and a mobile graphite sheet to generate continuous net water flow. Brownian motion of the sheet yields a stable flux of ~2 ns⁻¹, while a modest driving force can boost flux to ~16 ns⁻¹—an order‑of‑magnitude improvement. The driven mode also shortens translocation time and increases occupancy linearly with force. These findings provide a practical route to enhance RO membrane performance by harnessing natural water permeability, opening new avenues for scalable, energy‑efficient desalination technologies.

Abbreviations

CNT

Carbon nanotube

MD

Molecular dynamics

RO

Reverse osmosis