Understanding decoupling mechanisms of liquid-mixture transport properties through regression analysis with structural perturbation

Abstract

Optimization of liquid thermophysical properties is important for engineering applications; often achieved by mixing two or more liquids. An important issue is that properties tend to be coupled, which can be problematic if improvement of one property is accompanied by deterioration in another. Therefore, optimization is typically a compromise between properties, and it could be enhanced if they could be decoupled. Such decoupling however first requires an understanding of the common and distinct physical origins of each thermophysical property of interest. Here, we introduce a new approach to gain such understanding, combining molecular-simulation-based structural perturbation with regularized statistical analysis. Considering viscosity and thermal conductivity of a water–glycol mixture as a test-case, we identify the role of structure on each property, and highlight the important role that hydrogen bonding can play in such decoupling.