ELNES separation in spatially-resolved analysis of grain boundaries and interfaces

This paper describes a method to separate faithfully the ELNES belonging to a grain boundary or an interface from that of the bulk in the spatially resolved EELS analysis. The data acquisition is similar to that of the "spatial difference" method but the spectrum separation method is an advancement of the "spatial difference" method by reducing, if not eliminating completely, the subjective nature. This is achieved by regarding the spectrum separation as a procedure to make two vectors orthogonal to each other. The precision of the "orthogonal" separation is enhanced substantially by using numerical filtering of the spectra, which also improves the ability to detect artifacts such as the energy shift effect which can be misinterpreted as a "chemical shift". The operator's subjectiveness hence belongs to the uncertainty of the diagonization, as well as the selection in the analysis of a physical property reflected in the ELNES. Separating ELNES of an hetero-interface can be achieved in two steps following the same principle. Examples are drawn from Si3N4 ceramic materials where the interfacial ELNES are substantially different from those of the matrix. This method can also be applied to "Spectrum-Line" profiles to generate partial ELNES profiles corresponding to the signals from the boundary and the matrix regions, respectively. Comparing with other methods such as the multivariant statistical analysis (MSA), this method has advantage in reducing the noise and in an explicit physical meaning of the interface which enables the direct quantification of interfacial parameters (see Gu, Ultramicroscopy, submitted [1]). (C) 1999 Elsevier Science B.V. All rights reserved.