The Improvement of the Life Time Performance Estimation for Interconnect Stacks in Realistic Layouts

Abstract

The experimental determination of the intrinsic life time of a metallization stack is determined by the used test structures of metal lines and via-metal test structures. The test structure layout is prepared according to the presumption of the failure mechanisms modelled by Black’s Law. The interconnect reliability test structures are optimized. They have metal tracks which exceed the Blech length and contain dummy lines in a minimal spacing distance to the test line. Interconnects in realistic circuit layouts are different. They contain metal tracks and via connections with different surroundings like neighbour lines and a different density in metal, inter-metal- dielectric and via number. The sum of the overall mass flux divergence of the interconnects which determines the local mass transport and the life time are not only predicted by the current density and temperature. The distribution of mechanical stress influences mass transport as well. The local changes in mechanical stress like stress in chip corners, circuits under pad and via stacks up to thick metal top layers are not considered in the life time calculation based on Black’s Law. The mass flux simulation is a powerful tool to detect critical areas and the reliability of the outcome of process and layout optimizations. But the effort for such simulations is quite high and limited for some exemplary tasks. The paper shows the use of the multiple regression approach with moderation and mediation models. The models allow to detect indirect effects because of existing mediator variables or interactions with potential moderators. The derived models and regression coefficients can be used for more accurate life time predictions, for a reduced simulation effort and investigations on realistic interconnect stacks.