Reliability analysis of group-contribution methods in predicting critical temperatures of organic compounds

This article examines four group-contribution methods (Joback-Reid, Constantinou-Gani, Wilson-Jasperson, and Marrero-Pardillo) with respect to their abilities for calculating critical temperatures (T-c) of organic compounds on the basis of the parameters published in the original papers. To have a reliable data set for this examination, all experimental data of T-c collected in the SOURCE data system at the Thermodynamics Research Center (TRC) were evaluated. This data system contains more than 1900 T-c data points for about 630 organic compounds from which we compiled a selected T-c data set involving 510 compounds with uncertainty less than or equal to 5 K. Upon reviewing the normal boiling point (NBP) values for the same 510 compounds, we selected a NBP data set including 448 compounds with uncertainty less than or equal to 3 K for use in the T-c calculations. We then compared the selected T-c values with the values calculated by the four group-contribution methods on the basis of either selected or estimated NBP data. Particularly, we selected 71 compounds that were not included in the development of the parameters for the four models, and we used them to examine the predictive abilities of the models. The deviations of the calculated T-c data were categorized by compound types.