Raman correlation spectroscopy: A method studying physical properties of polystyrene by the means of multivariate analysis

Raman spectroscopy was used to determine physical properties of various grades of polystyrene, such as the Vicat temperature, the Izod impact and the Melt Flow Index. A number of 255 measurements were acquired on 195 samples of polystyrene pellets. The objective of this study is the correlation of these Raman results with those obtained by conventional techniques. Raman data endured different pre-treatments, such as correction of baseline, area normalization, selection of variables or multiple scatter correction. First, to discriminate the different families of polystyrene, Principal Component Analysis was applied to Raman spectra. Then, on the basis of Partial Least Square regression with random cross validation, the correlation between Raman spectrum of polystyrene and its physical properties was studied and discussed in the case of each pre-processing, in terms of correlation factor, the root-mean-square-error of calibration, the root-mean-square-error of validation and the root-mean-square-error of prediction, the bias and the number of latent variables. For example, in the specific case of the Vicat temperature, the best correlation factor is close to 0.98, and the root-mean-square-error of prediction is 0.85 degrees C with three latent variables which are correlated with the Raman spectrum of raw materials. (C) 2013 Published by Elsevier B.V.