Multivariate analysis of combined Fourier Transform Near-Infrared (FT-NIR) and Raman data sets for improved discrimination of drying oils

This work explores the application of chemometric techniques in the analysis of lipidic paint binders (i.e. drying oils) by means of Raman and near infrared spectroscopy. These materials were widely used by artists throughout history, both individually and in mixtures. We prepared various model samples of the pure binders (linseed, poppy-seed and walnut oils) obtained from different manufacturers. These model samples were left to dry and then characterized by Raman and reflectance near infrared spectroscopy. Multivariate analysis was performed by applying principal component analysis (PCA) on the first derivative of the corresponding Raman spectra (1800-750 cm-1), near infrared spectra (6000-3900 cm-1) and their combination to test whether spectral differences could enable samples to be distinguished on the basis of their composition. The vibrational bands we found most useful to discriminate between the different products we studied are the fundamental ν(C=C) stretching and methylenic stretching and bending combination bands. The results of the multivariate analysis demonstrated the potential of chemometric approaches for characterizing and identifying drying oils, and also for gaining a deeper insight into the ageing process. Comparison with HPLC data was furthermore realized to check the PCA results.