Viscosity of Binary Liquid Mixtures: A Comparative Analysis of Mixing Rules

Abstract

Modeling the viscosity of binary liquid mixtures (η_m) presents significant challenges due to the complex molecular interactions and the potential for nonmonotonic behavior. A total of 22 viscosity models were evaluated, with ln(η_m) expressed as the sum of an ideal component and a deviation term (δ), using the Arrhenius model as a reference. Based on the extensive experimental data, a classification scheme was developed to categorize η_m and δ into six distinct trend types, thereby aiding in the selection of the most suitable model. For each category, both η_m and δ were computed using all 22 models and compared to experimental results. The Ramírez de Santiago model, along with models incorporating one or more fitting parameters, demonstrated strong performance for monotonic mixtures (type I). For mixtures exhibiting nonmonotonic behavior, polynomial models, such as the Heric–Brewer, Legendre, and Redlich–Kister models, provided the best fit. The optimal polynomial degree was determined through leave-one-out cross-validation. This study offers a systematic approach for selecting the most accurate viscosity model for binary liquid mixtures.