Mapping fluid structuration to flow enhancement in nanofluidic channels

Fluid flow in miniature devices is often characterized by a boundary slip at the wall, as opposed to the classical paradigm of a no-slip boundary condition. While the traditional mathematical description of fluid flow as expressed by the differential forms of mass and momentum conservation equations may still suffice in explaining the resulting flow physics, one inevitable challenge against a correct quantitative depiction of the flow velocities from such considerations remains in ascertaining the correct slip velocity at the wall in accordance with the complex and convoluted interplay of exclusive interfacial phenomena over molecular scales. Here, we report an analytic engine that applies combined physics-based and data-driven modeling to arrive at a quantitative depiction of the interfacial slip via a molecular-dynamics-trained machine learning algorithm premised on fluid structuration at the wall. The resulting mapping of the system parameters to a single signature data that bridges the molecular and continuum descriptions is envisaged to be a preferred computationally inexpensive route as opposed to expensive multi-scale or molecular simulations that may otherwise be inadequate to resolve the flow features over experimentally tractable physical scales.

Automated summary

Fluid flow in miniature devices often exhibits boundary slip at the wall, contrary to the classical no-slip boundary condition. A new analytic engine has been developed that combines physics-based and data-driven modeling to quantitatively depict interfacial slip using a molecular-dynamics-trained machine learning algorithm. This approach maps system parameters to a single signature data, bridging the gap between molecular and continuum descriptions and providing a computationally inexpensive alternative to costly multi-scale or molecular simulations for resolving flow features at experimentally tractable physical scales.