The median-times-to-failure (t50's) for straightdual-damascene via-terminated copper interconnect structures, tested underthe same conditions, depend on whether the vias connect down to underlayingleads (metal 2, M2, or via-below structures) or connect up to overlayingleads (metal 1, M1, or via-above structures). Experimental results for avariety of line lengths, widths, and numbers of vias show higher t50's forM2 structures than for analogous M1 structures. It has been shown thatdespite this asymmetry in lifetimes, the electromigration drift velocity isthe same for these two types of structures, suggesting that fatal voidvolumes are different in these two cases. A numerical simulation tool basedon the Korhonen model has been developed and used to simulate the conditionsfor void growth and correlate fatal void sizes with lifetimes. Thesesimulations suggest that the average fatal void size for M2 structures ismore than twice the size of that of M1 structures. This result supports anearlier suggestion that preferential nucleation at the Cu/Si3N4 interface in both M1 and M2 structuresleads to different fatal void sizes, because larger voids are required tospan the line thickness in M2 structures while smaller voids below the baseof vias can cause failures in M1 structures. However, it is also found thatthe fatal void sizes corresponding to the shortest-times-to-failure (STTF's)are similar for M1 and M2, suggesting that the voids that lead to theshortest lifetimes occur at or in the vias in both cases, where a void needonly span the via to cause failure. Correlation of lifetimes and criticalvoid volumes provides a useful tool for distinguishing failuremechanisms.