Title :
Low-κ and High-κ breakdown statistics with variability: Clustering model versus reconstruction methodology (Invited)
Author :
Wu, Ernest ; Baozhen Li ; Stathis, James
Author_Institution :
Res. Div., IBM, Essex Junction, VT, USA
fDate :
June 29 2015-July 2 2015
Abstract :
In this work, we first review the characteristics and assumptions of the time-dependent clustering model and the sampling-based reconstruction methodology, two different approaches in dealing with breakdown statistics involving variability for modern BEOL low-κ and FEOL high-κ dielectrics. Then, similarities and differences of these two methodologies will be discussed in comparison with experiments. We show both methodologies can yield comparable results within the validity of their assumptions for dielectrics with small thickness variations. On the other hand, for BEOL low-k dielectrics with large variations, clustering model can be applied in both singlemode non-Weibull and bimodal distributions while the reconstruction methodology cannot be used.
Keywords :
Weibull distribution; electric breakdown; sampling methods; BEOL low-κ dielectrics; FEOL high-κ dielectrics; bimodal distributions; breakdown statistics; sampling-based reconstruction methodology; singlemode non-Weibull distributions; time-dependent clustering model; Correlation; Data models; Dielectric breakdown; Dielectrics; Mathematical model; Reconstruction algorithms; Bimodal; Clustering model; Non-uniform dielectric breakdown; Reliability; TDDB; Variability;
Conference_Titel :
Physical and Failure Analysis of Integrated Circuits (IPFA), 2015 IEEE 22nd International Symposium on the
Conference_Location :
Hsinchu
DOI :
10.1109/IPFA.2015.7224325