Analysis of System-Level Reliability Factors and Implications on Real-time Monitoring Methods for Oxide Breakdown Device Failures

Eric Karl,  David Blaauw,  Dennis Sylvester
University of Michigan


Abstract

Continued technology scaling exacerbates the incidence of degradation and failure in integrated circuits due to mechanisms such as oxide breakdown, negative bias temperature instability and electromigration. This work analyzes the impact of different factors on lifetime distributions for the oxide breakdown effect using a novel monte carlo approach based upon the percolation model and BSIM4. Results of the analysis of oxide failure distributions are used to explore real-time lifetime projection and the use of in-situ monitoring circuits. Under an ideal sensor assumption, the work shows that 500-1000 sensors would be needed to provide lifetime projections with error under 8-10%.