Yield analysis of nano-crossbar arrays for uniform and clustered defect distributions

Abstract

During the fabrication of nano-crossbar arrays, certain amount of defective elements are introduced to the end product which affect the yield drastically. Current literature regarding the yield analysis of nano-crossbar arrays is very rough and limited to the uniform distribution of defect occurrence with a few exceptions. Since density feature of crossbar architectures is the main attracting point, we perform a detailed yield analysis by considering both uniform and non-uniform defect distributions. Firstly, we briefly explain the present algorithms and their features used in defect tolerant logic mapping. Secondly, we explain different defect distributions and logic function assumptions used in the literature. Thirdly, we formalize an approximate successful mapping probability metric for uniform distributions and determine area overheads. After that, we apply a regional defect density analysis by comparing uniform and clustered defects to formulate a looser upper bound for area overheads regarding clustered distributions. Finally, we conduct extensive experimental simulations with different defect distributions.