Identification of chemical components in XPS spectra and images using multivariate statistical analysis methods

Abstract

A variety of data analysis methods can be used to enhance the information obtained from a measurement, or to simplify extraction of significant components from large data sets. Much work is needed to improve the quantification and interpretation of XPS spectra and images from complex organics. Multivariate analysis (MVA) is increasingly used for applications in electron spectroscopy to aid the analyst in interpreting the vast amount of information yielded by spectroscopic techniques. In general, the goals of MVA are to determine the number of components present, identify the chemical components, and quantify component concentrations in the mixture. Principal component analysis (PCA) is frequently used to determine the number of mathematical components which describe the data set. These mathematical components must then be related to chemically meaningful components. Various approaches to solve rotational ambiguities of spectral resolution, including local rank method (EFA), pure variables method (Simplisma) and multivariate curve resolution (MCR), are tested in the determination of chemical components from XPS data. Limitations associated with the resolution of a single matrix are shown to be partially or completely overcome when several related matrices are treated together. The test data sets contain XPS images or spectra acquired from blends of poly(vinyl chloride), PVC, and poly(methyl methacrylate), PMMA. The PVC degrades rapidly upon exposure to the X-ray beam. Spectra and images from the blend, acquired as a function of time, provide the multi-dimensional data sets for algorithm evaluation. In addition to spectral resolution, multivariate image analysis methods, such as principal component analysis, are used to extract maps of the pure components from an images-to-spectra data set.