The hygroscopicity of organic aerosol in the atmosphere can be represented by a semi-empirical single parameter, κ. In this work we test possibilities for developing quantitative structure-property relationship (QSPR) models for κ based on chemical similarity. Models were developed in two ways: b... Show moreThe hygroscopicity of organic aerosol in the atmosphere can be represented by a semi-empirical single parameter, κ. In this work we test possibilities for developing quantitative structure-property relationship (QSPR) models for κ based on chemical similarity. Models were developed in two ways: by manually assessing the suitability of several plausible physico-chemical descriptors; and by systematically evaluating hundreds of constitutional (e.g. number of particular atoms, bond types, molecular weight,...), topological, electrostatic, geometrical and quantum-chemical descriptors with the QSPR modelling software CODESSA (COmprehensive DEscriptors for Structural and Statistical Analysis). A set of 74 compounds with measured κ values was taken from the literature and prediction capabilities of the developed models were evaluated by leave-one-out cross-validation procedure. A 5-parameter linear regression model obtained with CODESSA was found to be the most suitable. Among the five descriptors, the two providing the highest contributions to the total variance were found to be (i) the final heat of formation divided by the number of atoms (69 %) and (ii) the ratio of molecular weight and molecular volume (16 %), although other topological and electrostatic descriptors were also of non-negligible importance for prediction of κ. The squared correlation coefficient and the root mean square error of a leave-one-out cross-validation procedure were 0.80 and 0.037, respectively. The results show that quantitative structure-property relationship approaches are useful for modeling κ. Show less
