Gas–particle partitioning of semi-volatile organics on organic aerosols using a predictive activity coefficient model: analysis of the effects of parameter choices on model performance

The partitioning of a diverse set of semivolatile organic compounds (SOCs) on a variety of organic aerosols was studied using smog chamber experimental data. Existing data on the partitioning of SOCs on aerosols from wood combustion, diesel combustion, and the α-pinene-O3 reaction was augmented by carrying out smog chamber partitioning experiments on aerosols from meat cooking, and catalyzed and uncatalyzed gasoline engine exhaust. Model compositions for aerosols from meat cooking and gasoline combustion emissions were used to calculate activity coefficients for the SOCs in the organic aerosols and the Pankow absorptive gas/particle partitioning model was used to calculate the partitioning coefficient Kp and quantitate the predictive improvements of using the activity coefficient. The slope of the log Kp vs. log pL0 correlation for partitioning on aerosols from meat cooking improved from −0.81 to −0.94 after incorporation of activity coefficients iγom. A stepwise regression analysis of the partitioning model revealed that for the data set used in this study, partitioning predictions on α-pinene-O3 secondary aerosol and wood combustion aerosol showed statistically significant improvement after incorporation of iγom, which can be attributed to their overall polarity. The partitioning model was sensitive to changes in aerosol composition when updated compositions for α-pinene-O3 aerosol and wood combustion aerosol were used. The octanol–air partitioning coefficientʹs (KOA) effectiveness as a partitioning correlator over a variety of aerosol types was evaluated. The slope of the log Kp−log KOA correlation was not constant over the aerosol types and SOCs used in the study and the use of KOA for partitioning correlations can potentially lead to significant deviations, especially for polar aerosols.