Data-Sparse Strategies for Inferring Fluid Transport Properties via Molecular Simulation

In an ideal world, any sampling-driven method for simulation (including many molecular simulation methods to study micro- and nano-scale flows) would be carried out in the (approximately) infinite-statistics limit. Practical constraints often prevent the use of "large values of infinity"; in some problems, practical constraints even prevent the collection of infinite statistics for "small values of infinity." In this talk, we discuss two strategies for inferring fluid transport properties from molecular simulations using substantially less data than is traditional, while maintaining useful levels of accuracy. The first strategy is based upon sparse data collection along a spatial degree of freedom; the second strategy is based upon sparse data collection along a temporal degree of freedom. Both methods leverage the framework of entropy scaling, which we will briefly introduce. We validate on, and demonstrate the utility of our results for, large-scale datasets of fluids obtained via molecular-dynamics simulation.