Chrastil revisited for supercritical carbon dioxide

Chrastil's semi-empirical model has been widely used, since its introduction in 1982, to fit solubility data in supercritical carbon dioxide. However, it has been used with different sources for carbon dioxide's density, with different objective functions and different numerical algorithms to determine the three Chrastil coefficients. These differences lead to an unnecessary variance in fitting performance and Chrastil coefficient values. The result is a reduction in the perceived quality of Chrastil's model. Also, calculated solubilities based on published Chrastil coefficients using a different source for carbon dioxide's density risks introducing considerable incremental error. This work uses Span and Wagner's Fundamental Equation of State and a two-pass Nelder Mead algorithm to minimize the absolute average relative deviation (AARD) for 701 binary data sets. The median AARD is just 8.8% while just 5.2% of the 701 systems have an AARD greater than 30%.

Automated summary

Chrastil's semi-empirical model has been widely used for fitting solubility data in supercritical carbon dioxide since 1982. Differences in using different sources for carbon dioxide's density, objective functions, and numerical algorithms have led to unnecessary variance in fitting performance and Chrastil coefficient values. This study using Span and Wagner's Fundamental Equation of State and a Nelder Mead algorithm achieved a median AARD of 8.8% for 701 binary data sets, with only 5.2% of systems having an AARD greater than 30%.