Locally Enhanced Progressive Degradation in Thin Oxides: A Calibration Effort

Abstract

Extensive research has been conducted in the scope of Time Dependent Dielectric Breakdown (TDDB), a stringent reliability phenomenon in advanced CMOS technologies, mainly focusing on the physics and statistics behind soft and hard breakdown events. However, the wear-out phase following the percolating path formation, a progressive degradation in terms of defect accumulation until device failure, lacks substantial knowledge and has not been thoroughly modelled. This work addresses the modelling of this wear-out phenomenon: firstly, by carefully calibrating soft breakdown events matching experimental observations by means of the well-established percolation model and the analytical algorithm for breakdown times calculation; secondly, by simulating their progressive degradation in terms of defects accumulation with a Monte-Carlo implemented Markov Model with Gaussian distribution of the activation energy for defect formation, using the tuned parameters from the calibration procedure. Complete replicas of TDDB events are thus simulated, reproducing the voltage dependence of the wear-out times in the Weibull profile. Observations of early failure and time-out events highlight the need of an interdependent generation of energy distributions for excitation/relaxation processes.