Paper EM-FrM7
Band Alignment in Metal/Oxide/Semiconductor Stacks

Friday, October 19, 2007, 10:00 am, Room 612

Understanding band alignment in a Metal/Oxide/Semiconductor (MOS) structure is both a fundamental physical chemistry challenge and a crucial technological issue for nanoelectronics as the correct band offsets across a MOS stack is essential to prevent undesirable leakage currents in device applications. A recurrent problem in the evaluation of the band offsets is the lack of reproducibility between measurements on nominally identical samples and between measurements performed using different experimental methods. In this study, we have performed ultra-violet photoemission and inverse photoemission spectroscopy measurements in the same UHV system on a variety of high-κ dielectric films (HfO₂, Hf_0.8Si_0.2O₂, SiO₂, Al₂O₃, ZrO₂) on different semiconductor substrates (Si, Ge, GaAs). This experimental setup allows the direct determination of band gaps, electron affinities, metal work functions, and band offsets across MOS structures. We have also explored the effect of subsequent metallization of the high-κ film in UHV with either of metals (Ru, Ti, Al) chosen for their very different work function and oxygen affinity. By examining the line shape of the shallow core levels using synchrotron radiation-excited x-ray photoemission spectroscopy, we determined the chemical state of the metal, the oxide and the substrate. Ru and Ti are found to remain metallic upon deposition on all oxide films used in this study. In contrast, strong changes in the chemistry of the entire stack are observed upon Al deposition. An Al₂O₃ interface layer is always present at the Al/Oxide interface, and evidence for reduction of the semiconductor/high-κ interface oxide is also observed. Consistent with the observed level of reactivity, our measured Metal/Oxide band offsets are in good agreement with the values predicted by the interface gap state model for as-deposited Ru and Ti. In the case of Al, owing to an additional Al₂O₃ interface layer between the oxide and the Al, our measured values are in poor agreement with the interface gap state model. This indicates that for reactive metal, the band offset is not well described by this simple model.