MULTIVARIATE STATISTICAL-ANALYSIS OF FEG-STEM EDX SPECTRA

A multivariate statistical analysis method has been used to examine th e information content of a series of EDX spectra acquired using a FEG- STEM recorded across a segregated grain boundary in a ferritic steel, The spectra have been factored into a set of orthogonal information co mponents (eigenspectra), each of which corresponds to an independent s ource of information, Two meaningful eigenspectra are identified: one is related to the segregation of P and Cr and the other to a combinati on of self-absorption and coherent bremsstrahlung. Spectra are reconst ructed from the meaningful eigenspectra to reduce noise and improve th e sensitivity for detection of trace element segregation.