CLASSIFICATION OF SYNTHETIC ORGANIC PIGMENTS BY MULTIVARIATE DATA ANALYSIS OF FTIR SPECTRA

For the identification of pure synthetic organic pigmentsFourier Transform Infrared (FTIR) spectroscopy has beenestablished as a powerful analytical technique. Usually theidentification of such pigments by FTIR is processed bycomparing the spectrum of an unknown sample with thespectra present in a database. Those samples are difficultto be identified if they don’t match with one of the librar yspectra, and even class assignments can be performed byexperts’ analysis only.In this paper an approach is presented to facilitate classificationof organic pigments by multivariate statistical analysisof their FTIR spectra. Partial least-squares discriminantanalysis (PLS-DA) was applied to the pigments FTIR spectra,using 11 binar y y-variables for pigment class definition.The threshold of the discriminant variable could be optimizedfor each class. It is shown that application of theclassification models to FTIR spectra of unknown paintsamples gained very similar results as obtained fromexperts.