Raman spectroscopic discrimination of pigments and tempera paint model samples by principal component analysis on first‐derivative spectra

This work explores the application of principal component analysis (PCA) on first‐derivative Raman spectra to investigate historical tempera paint model samples. Various paint model samples were prepared containing pure blue pigments (azurite, lapis lazuli and smalt), pure red pigments (cinnabar, minium and raw Sienna), pure white pigments (lead white, chalk and gypsum), pure egg yolk as binder and tempera model samples obtained by mixing each of the pigments with the binder, and further characterized by Raman spectroscopy. The corresponding Raman spectra were used to apply PCA in order to test whether spectral differences allowed discrimination of samples based on their composition. Multivariate analyses were performed separately on three data matrices, one for each color, namely, white, blue and red, corresponding to the model samples, and all containing the spectral data of the binder model sample. Different pretreatments, that is log and derivative spectra, were performed on the spectra since no pattern distributions were obtained when the original Raman spectra were analyzed. Nevertheless, the multivariate analysis of the original Raman spectra was able to track alterations of sensitive pigments due to laser interaction. Results showed the excellent ability of PCA, when applied to the derived Raman spectra, to discriminate model samples according to their differing compositions in the three groups of model samples tested. This is the first attempt to use this approach in the field of cultural heritage and demonstrates the potential benefits for identifying historical pigments and binders for purposes of conservation and restoration. Copyright © 2010 John Wiley & Sons, Ltd.