Automated background subtraction technique for electron energy-loss spectroscopy and application to semiconductor heterostructures

Electron energy-loss spectroscopy (EELS) has become a stan-dard tool for identification and sometimes also quantificationof elements in materials science. This is important for un-derstanding the chemical and/or structural composition ofprocessed materials. In EELS, the background is often mod-elled using an inverse power-law function. Core-loss ioniza-tion edges are superimposed on top of the dominating back-ground, making it difficult to quantify their intensities. Theinverse power-law has to be modelled for each pre-edge regionof the ionization edges in the spectrum individually ratherthan for the entire spectrum. To achieve this, the prerequisiteis that one knows all core losses possibly present. The aim ofthis study is to automatically detect core-loss edges, model thebackground and extract quantitative elemental maps and pro-files of EELS, based on several EELS spectrum images (EELS SI)without any prior knowledge of the material. The algorithmprovides elemental maps and concentration profiles by makingsmart decisions in selecting pre-edge regions and integrationranges. The results of the quantification for a semiconduc-tor thin film heterostructure show high chemical sensitivity,reasonable group III/V intensity ratios but also quantificationissues when narrow integration windows are used withoutdeconvolution.