Multivariate characterization of polycyclic aromatic hydrocarbons using semi-empirical molecule orbital calculations and physical data

Multivariate characterization of 60 polycyclic aromatic hydrocarbons (PAHs) was performed using data from literature and semi-empirical molecular orbital calculations. This dataset was analyzed by principal component analysis and projections to latent structures by means of partial least square. The PAHs were found to distribute in the chemical domain in several groups mainly governed by the number of aromatic rings and the number of five-membered rings. The multivariate model and training set was used to predict GC retention times on a non-polar column (DB-5). Using a 24 experimental design on the principal components of the chemical characterization model, a test set of PAHs was selected dependent on the distribution in the chemical domain of the PAHs. Such a test set is expected to improve selection of PAHs for future testing and modeling of both biological and chemical responses. Although the model of GC retention times was good, the precision was however not good enough for practical use.