Application of chemometric methods for classification of atmospheric particles based on thin-window electron probe microanalysis data

Conventional single-particle electron probe microanalysis (EPMA) is widely used for evaluating the sources of atmospheric aerosol. The method is capab le of simultaneously detecting the chemical composition and the morphology of each particle. Computer-controlled automatic EPMA allows the analysis of huge numbers of individual particles. Cluster as well as factor analysis a re used for the classification of particles based on the obtained data set. However, the method is not able to detect low-Z elements (C, N, O), theref ore, e.g. organic particles can only be identified by their typical inorgan ic content and high background. Using a thin-window X-ray detector, the cap abilities of EPMA can be extended to determine low-Z elements. The recently developed quantification method based on Monte Carlo simulations is capabl e to evaluate, elemental concentrations in single microscopic particles, in cluding C, N and O. It was shown that also chemical species can be determin ed from the obtained concentrations. Hierarchical and non-hierarchical clus ter analysis, as well as principal component analysis were applied for the classification of particles based on low-Z EPMA data. A mixture of standard particles as well as atmospheric aerosol samples were used to test the cla ssification methods. Different input data (X-ray intensities or elemental c oncentrations) and scaling functions were used for the chemometric methods. Cluster and factor analysis appear to be efficient tools for classificatio n of particles based on low-Z EPMA data. As an example, atmospheric ammoniu m. sulphate and organic sulphur were classified in separate groups, which w as not possible by conventional EPMA. (C) 2001 Elsevier Science B.V. All ri ghts reserved.