The electromigration (EM) diffusion mechanisms of electroplated copper (Cu) with tantalum (Ta) barrier/dielectric diffusion barrier SiCxNy and cold/hot two-step sputter-deposited aluminum (Al)–0.5-wt %Cu damascene interconnects with niobium (Nb) liner are examined and compared using the via-EM testing pattern with different linewidths. The interface between Cu and SiCxNy is the dominant diffusion path for the Cu damascene interconnects regardless of the Cu microstructure. An activation energy (Ea) of approximately 0.9±0.03 eV is obtained for the width range of 0.1–6 μm. Therefore, the diffusion mechanism is independent of the Cu microstructure. Regarding the cold/hot two-step sputter-deposited Al–0.5-wt %Cu damascene interconnects with Nb liner, the EM median time to failure (MTF) increases with increasing the linewidth for the Al bamboolike microstructure, indicating that the interface between the Al and Nb liners is the dominant diffusion path. This is probably because a rapid diffusion path along the NbAlx reaction product is formed during the two-step cold/hot sputter deposition at 400 °C. The EM MTF does not increase for more than 4 μm. It is also found that the Ea is approximately 0.9 eV for the Al bamboolike microstructure and that it decreases with increasing linewidth for th- e Al polycrystalline microstructure, meaning that the grain-boundary diffusion is also included and that ratio of the interfacial diffusion and grain-boundary diffusion depends upon the linewidth. This is because the Ea of the grain-boundary diffusion is smaller than that of the interfacial diffusion for Al–0.5-wt %Cu damascene interconnects. The EM diffusion mechanism of the Cu damascene interconnects with Ta barrier/dielectric diffusion barrier SiCxNy is completely different from that of the Al damascene interconnects with Nb liner.