MULTIVARIATE STATISTICAL-ANALYSIS OF STEM-EDX DATA FROM RADIATION-INDUCED SENSITIZATION IN STAINLESS-STEEL

A series of energy-dispersive X-ray spectra acquired in a field-emissi on gun scanning transmission electron microscope to generate compositi onal profiles across a grain boundary in a neutron-irradiated stainles s steel were subjected to a multivariate statistical analysis (MSA), T he complex profiles generated by conventional spectrum processing were factored by MSA into two independent components of information, one a ssociated with both the pre-irradiation heat treatment and radiation-i nduced regation (RIS) and the other with diffusion prior to precipitat e formation. The two experimental components were separately modelled to deconvolute simple theoretical distributions of composition, which were then recombined to give the full variation in composition across the boundary. A new observation was that the commonly observed RIS enh ancement of Ni and Si did not persist in the boundary plane, the compo sition of which was dominated by pre-irradiation segregation. The pote ntial benefits of MSA for segregation studies are discussed.