Improving the understanding of dielectric breakdown by using LFN technique

Time dependent dielectric breakdown (TDDB) measurements are commonly used to assess the reliability of backend-of-line (BEOL) and middle-of-line (MOL) dielectrics. Dielectric breakdown is one of the main causes of failure for interconnect systems, posing a serious threat to the functionality of electronic components that are expected to work ~ 10 years in nominal operating conditions.

To estimate the lifetime of a dielectric material, an accelerated, but still time consuming, test procedure is normally used: TDDB measurements are performed at higher fields and temperatures compared to nominal operating conditions. TDDB measurements are also a destructive technique since a constant stress voltage is applied until dielectric breakdown occurs. The so obtained the time-to-failure (TTF) is then used to extrapolate the lifetime at nominal operating conditions using one of the several models that have been developed over time based on empirical observations and physical modelling of dielectrics (Power Law Model, E-Model, 1/E Model, etc). Moreover, as the thickness of BEOL and ML dielectrics is being decreased with each new technology node, it has become evident that TDDB are not sufficient to fully understand the mechanisms that lead to dielectric failure. For example, surface defectivity has been found to play a key role in metal drift-induced dielectric breakdown.

The objective of this project is to investigate the potentiality of low-frequency noise (LFN) measurement technique to the studying of dielectric breakdown in thin dielectric films. The LFN methodology relies on studying the statistical distribution of current fluctuations measured in a dielectric under constant stress voltage. LFN is a non-destructive technique and could provide meaningful insights about surface and bulk defectivity as well as the dominant conduction mechanism.