Machine-Learning Potential Molecular Dynamics Reveals the Critical Role of Flexibility in Solid-Liquid Nanofluidic Friction.

The confining walls made by 2D materials are often considered solid boundary conditions in studies of fluid transport through nanochannels, while the atomically thin walls inherently exhibit thermal fluctuations at a finite temperature. In this work, we investigate the solid-liquid interfacial friction properties of water confined within flexible nanochannels using machine-learning-potential molecular dynamics. Surprisingly, we find that the friction coefficient (λ) increases with lateral size in the flexible nanochannels, following a linear relationship with 1/, which is absent in rigid channels. For thicker nanochannel walls with a large bending stiffness (), this size dependence weakens. This 1/ scaling resembles the size-dependent thermal fluctuation amplitude of 2D sheets, suggesting that λ can be well controlled by applying the mechanical strain to the channel walls via controlling the phonon modes associated with surface fluctuations. Further analysis reveals that the λ can be decomposed into the contribution from lattice roughness and the thermal-fluctuation-induced ripple of the 2D surface, where the second term scales with . These findings offer insights into the manipulation of nanoscale flow through precise control of local curvature and fluid-solid coupling.