According to the entropy-scaling principle proposed by Rosenfeld, the shear viscosity─made dimensionless appropriately─is, to good approximation, a univariate function of residual entropy only. We propose models for the shear viscosity for 952 pure substances based on 100,988 experimental data points, which considerably exceeds earlier parametrization studies. As advances over prior work, we propose a revised ansatz function that better captures the low-density conditions. Because experimental data are scarce and unevenly distributed for many substances, leading to considerable data gaps, we propose a parametrization scheme that ensures rather robust extrapolation capabilities of the model, instead of minimizing deviations of the model to experimental data only. A revised “reference” viscosity is applied that shows considerable predictive capabilities for gas-phase viscosities of real substances. Furthermore, we propose a conservative approach for identifying and removing outlying data points. The viscosity is correlated with an average mean absolute relative deviation of 3.1%.
Klenk et al. (Mon,) studied this question.