Low frequency noise statistics for the breakdown characterization of ultra-thin gate oxides

TL;DR

This study analyzes low frequency noise in ultrathin oxide capacitors to identify statistical markers like non-Gaussianity and noise power ratios that can predict oxide reliability and breakdown.

Contribution

It introduces a statistical approach using non-Gaussianity and noise power ratios to assess ultrathin oxide reliability and breakdown potential.

Findings

High peaks in non-Gaussianity indicate new percolation paths.

Noise power ratio correlates with conduction through these paths.

Results are consistent across different oxide thicknesses and qualities.

Abstract

We have investigated the statistics of low frequency noise in the tunneling current of ultrathin oxides (2.5nm-4nm) in metal oxide semiconductor capacitors as a function of the applied voltage stress. The statistical analysis includes (i) non-Gaussianity (nG), which is a measure of the degree of temporal correlation in the noise, and (ii) ratio of integrated noise power to the DC leakage current (R). The occurrence of high peaks in nG indicates the appearance of new percolation paths, and the subsequent conduction through these paths is indicated by R. Our results show that the nG and R characteristics are generic for the oxides of different thickness and growth quality and have the potential, in conjunction with leakage itself, of being used as a prognosticator of oxide reliability.