Common y-intercept and single compound regressions of gas-particle partitioning data vs 1/T

Abstract

Confidence intervals are placed around the log K_p vs 1/T correlation equations obtained using simple linear regressions (SLR) with the gas-particle partitioning data set of Yamasaki et al. [(1982) Env. Sci. Technol.16, 189–194]. The compounds and groups of compounds studied include the polycylic aromatic hydrocarbons phenanthrene + anthracene, me-phenanthrene + me-anthracene, fluoranthene, pyrene, benzo[a]fluorene + benzo[b]fluorene, chrysene + benz[a]anthracene + triphenylene, benzo[b]fluoranthene + benzo[k]fluoranthene, and benzo[a]pyrene + benzo[e]pyrene (note: me = methyl). For any given compound, at equilibrium, the partition coefficient K_p equals (F/TSP)/A where F is the particulate-matter associated concentration (ng m⁻³), A is the gas-phase concentration (ng m⁻³), and TSP is the concentration of particulate matter (μg m⁻³). At temperatures more than 10°C from the mean sampling temperature of 17°C, the confidence intervals are quite wide. Since theory predicts that similar compounds sorbing on the same particulate matter should possess very similar y-intercepts, the data set was also fitted using a special common y-intercept regression (CYIR). For most of the compounds, the CYIR equations fell inside of the SLR 95% confidence intervals. The CYIR y-intercept value is −18.48, and is reasonably close to the type of value that can be predicted for PAH compounds. The set of CYIR regression equations is probably more reliable than the set of SLR equations. For example, the CYIR-derived desorption enthalpies are much more highly correlated with vaporization enthalpies than are the SLR-derived desorption enthalpies. It is recommended that the CYIR approach be considered whenever analysing temperature-dependent gas-particle partitioning data.