Application of Factorisation Methods to Analysis of Elemental Distribution Maps Acquired with a Full-Field XRF Imaging Spectrometer

Abstract

The goal of the work was to investigate the possible application of factor analysis methods for processing X-ray Fluorescence (XRF) data acquired with a full-field XRF spectrometer employing a position-sensitive and energy-dispersive Gas Electron Multiplier (GEM) detector, which provides only limited energy resolution at a level of 18% Full Width at Half Maximum (FWHM) at 5.9 keV. In this article, we present the design and performance of the full-field imaging spectrometer and the results of case studies performed using the developed instrument. The XRF imaging data collected for two historical paintings are presented along with the procedures applied to data calibration and analysis. The maps of elemental distributions were built using three different analysis methods: Region of Interest (ROI), Non-Negative Matrix Factorisation (NMF), and Principal Component Analysis (PCA). The results obtained for these paintings show that the factor analysis methods NMF and PCA provide significant enhancement of selectivity of the elemental analysis in case of limited energy resolution of the spectrometer.