Abstract: We present a new conceptual framework for multiscale hydrodynamic modellingof nanofluids that seamlessly couples molecular dynamics (MD) with continuumhydrodynamics using an adaptive, information-driven coupling layer. The method treatslocally resolved MD regions as data-generating microscopes that provide dynamicallyupdated constitutive closures to a continuum solver via machinelearned,reduced-orderoperators and fluctuation-consistent stochastic stress models. We demonstrate the approachthrough synthetic data for nanoparticle-laden water and illustrate improvements in capturingnon-Newtonian effective rheology, interfacial slip, thermal transport anomalies, andBrownian stress over classical continuum models. This result explores the computationallyefficient approach and offers a clear path for uncertainty quantification and experimentalintegration.
A. K. Singh (Wed,) studied this question.