Abstract
Metal (e.g. TiN, Ta2N) inserted hafnia-based high-k gate stacks on Si as well as gallium oxide/gallium gadolinium oxide gate stacks on III-V semiconductors (GaAs, InGaAs, AlInAs) are being investigated for current and future MOSFET applications. The combination of light, medium and heavy elements in multilayers with widths down to a nanometre is a major challenge for high spatial resolution EELS spectrum imaging, especially as the interfaces have significant roughness on this scale. However, it is possible to identify layers significantly thinner than the widths of the profiles that result from beam broadening and/or roughness. To deal with the range of edges present, it is essential to collect edges out to losses of ~3keV. Edges from some elements are difficult to extract but, in suitable case, the ELNES on other edges can be used to give information on such elements. Recording the low loss region of the spectrum as well as the core loss regions at each pixel allows much fuller processing of the data including determination of the absolute numbers of atoms per unit volume. In this way, the distributions of the different phases can be estimated and the overall accuracy is excellent. The results are also in good agreement with earlier work using ELNES.
Export citation and abstract BibTeX RIS