A physics-based deconstruction of the percolation model of oxide breakdown

Abstract

The percolation model has been very successful to explain the gate oxide breakdown statistics and to establish a quantitative relation between the breakdown and the generation of defects. However, this simple geometric picture reaches its limit when the oxide thickness is scaled down and becomes comparable to the “size” of the involved defects. Moreover, the geometric definition of the breakdown as a path of interconnected defects is conceptually far from the measurement criteria used for breakdown detection during ultra-thin oxide stress experiments. In this work, we present a new version of the percolation model which is based on coupling a trap-assisted tunneling model with a statistical description of the position of the defects in the oxide. The model provides a description of the breakdown statistics as a function of the current level used for breakdown detection (Δ I BD ) and is much closer to the actual physics of the breakdown. Comparison with breakdown data demonstrates that the new model not only reproduces all the important results of the standard percolation picture but provides successful explanation for other experimental observations such as the change of the breakdown distribution with Δ I BD .