Interpreting atomic-resolution spectroscopic images

Core-loss electron energy loss spectroscopy is a powerful experimental tool with the potential to provide atomic-resolution information about electronic structure at defects and interfaces in materials and nanostructures. Interpretation, however, is nonintuitive. Comparison of experimental and simulated compositional maps in $\mathrm{La}\mathrm{Mn}{\mathrm{O}}_{3}$ shows good agreement, apart from an overall scaling of image contrast, and shows that the shape and width of spectroscopic images do not show a simple variation with binding energy, as commonly assumed, or with the size of the orbital excited. For the low lying La ${N}_{4,5}$ edge with threshold at around $99\phantom{\rule{0.3em}{0ex}}\mathrm{eV}$, delocalization does not preclude atomic resolution, but reduces the image contrast. The image width remains comparable to that of the much higher lying O $K$ edge with threshold at around $532\phantom{\rule{0.3em}{0ex}}\mathrm{eV}$. Both edges show a volcanolike feature, a dip at the column position not previously seen experimentally. In the case of the O $K$ edge, this represents an experimental verification of nonlocal inelastic scattering effects in electron energy loss spectroscopy imaging. In the case of the ${N}_{4,5}$ edge, the volcanolike feature is due to dynamical channeling and absorption of the probe through the specimen thickness. Simulation is therefore critical to the interpretation of atomic-resolution elemental maps.