Physics of Dielectric Breakdown Revealed by Low Frequency Noise Spectroscopy

Dielectric breakdown is the physical phenomenon wherein a dielectric material loses its insulating properties, due to a prolonged exposure to an applied stress, after which it eventually starts to conduct electricity. It is one of the main failure mechanisms of Back End Of Line (BEOL) dielectrics in interconnect systems, giving rise to several reliability issues for semiconductor devices that are expected to work for at least ten years under operating or use conditions. Traditional dielectric reliability methodologies, such as e.g. Time-Dependent Dielectric Breakdown (TDDB), are based on accelerated testing, where the dielectric is put under higher stress compared to operating conditions, until it breaks down. The associated breakdown voltage as well as the time needed to break down, are used to extrapolate the dielectric lifetime to operating conditions, using an empirical model (e.g. the power law model). Limitations of this conventional approach are that it is time-consuming, destructive, and the acquired physical insight, concerning the physics of dielectric breakdown, is limited. In this work, we therefore explore the potentiality of the Low Frequency Noise (LFN) technique, as an alternative way of assessing the dielectric reliability as well as predicting the dielectric lifetime. In LFN, the statistical distribution of the dielectric leakage current fluctuations, is studied in the frequency domain in the form of a Power Spectral Density (PSD). The PSD is a measure of the trap density of the dielectric, which allows to quickly probe the defectivity of the dielectric, and thus also its lifetime. LFN measurements are non-destructive and fast compared to TDDB, and have the potential to provide additional physical insights into the failure mechanisms of dielectric materials.