Detection of electron energy-loss edge shifts and fine structure variations at grain boundaries and interfaces

Series of electron energy-loss spectra (EELS) and X-ray spectra (EDX) were acquired across a grain boundary in a stainless steel. Cr segregation was detected with both methods and Mo was detected using EDX. However, when the spectral series were subjected to multivariate statistical analysis only edge intensities, without edge shifts or major shape changes, were detected in the Cr and Fe-L edges. The sensitivity of the analysis to edge shifts was then investigated by deliberately shifting one of the EELS spectra by different amounts. An eigenvector associated with the imposed shift was easily detected even when the shift was just one channel (0.3 eV). It was estimated that the sensitivity to edge shifts was as low as 0.1 eV for the experiment. The use of higher dispersion and improved counting statistics should improve the sensitivity of the multivariate analysis to a few tens of millivolts and provide the means for revealing real edge shifts and shape changes in the presence of artefacts and instrumental instability.