Compositional Mapping by Z-Contrast Imaging

Abstract

Quantitative compositional information can be extracted from high-resolution Z-contrast images by comparison of simulated and experimental images. We developed a phenomenological method to determine quantitatively the composition of a material with atomic column spatial resolution directly from the analysis of local integrated intensities of aberration-corrected Z-contrast experimental images [1]. In this work we apply this method to high-resolution aberration-corrected Z-contrast images acquired at 100kV with a dedicated Nion UltraSTEM scanning transmission electron microscope, which is equipped with a Nion aberration-corrector and a Gatan Enfina EELS detector. Specimens for Z-contrast imaging were prepared by mechanical thinning and Ar + ion milling using a Precision Ion Polishing System (PIPS). A beam energy less than 3.5 kV has been selected to reduce amorphisation of the sample. As a final step, the sample was introduced in a Fischione ion mill at 12 o and 0.5 kV to reduce surface damage. The thickness of the analyzed region was determined from the analysis of the corresponding spatially resolved low-loss EELS signal [2].In order to estimate the composition in each column, we follow a procedure previously published [1] in which the first step is to detect pixels with the local intensity maxima associated; for that purpose we apply the Peak Pairs software [3]. Once these maxima intensity pixels are located, it is straightforward, with the help of the image processing software, to measure the intensities integrated within a selected area of the projected unit cell. In order to decide the best integration area, several measurements have been taken using different integration areas. A certain area of integration was finally selected that gives normalized integrated intensity ratios R as shown in equation (1). R values were determined following the procedure described in ref.[1], but taking whole unit cells in the present study. R values show some advantages when they are defined in this way: (a) R depends almost linearly on the composition, (b) R has a very low dependence on specimen thickness over a convenient thickness range, (c) R values are almost unaffected by surrounding dumbbells so that the signal is essentially due to just the atoms contained within the selected atomic column. R i = I column /I substrate (1) 1728