Line-to-Line TDDB Modeling: LER Specs for Sub-20-nm Pitch Interconnects

We present a variability-aware predictive modeling for line-to-line time-dependent dielectric breakdown (TDDB), which is based on a dynamic percolation model (DPM). Our DPM focuses on intrinsic breakdown and accounts for the impact of the process-induced geometrical variations. For the intrinsic breakdown, the generation of defects in the dielectric and the formation of percolation paths between metal lines is modeled dynamically. As for process-induced geometry variations, we account for line-edge roughness (LER) and die-to-die spacing variations. Finally, we calibrate our DPM-to-TDDB measurements from low-
${k}$
planar capacitors (p-caps) of various thicknesses and apply them to derive predictions for sub-20-nm interconnects. We report variability specs for meeting ten-year lifetime at operating conditions. We show that the LER autocorrelation length must be considered for deriving realistic predictions, which would, otherwise, be too conservative if LER distributions were assumed to be fully Gaussian. For current state-of-the-art characteristics of LER and die-to-die spacing variations, we predict that line-to-line spacings as narrow as 6 nm can be reliable for the considered low-
${k}$
material when extreme ultraviolet lithography (EUV) spacer-assisted multipatterning is used as a litho-patterning process.