Electromigration of Cu interconnects under AC, pulsed-DC and DC test conditions

Abstract

Electromigration (EM) of a dual damascene, single-via fed test vehicle was measured using DC, AC followed by DC, and three rectangular-wave DC stressing conditions at 598 K. In some of the experiments samples were allowed to cool between stress cycles. Void formation and migration inside Cu interconnects were filmed. We show void formation by AC and void migration and consolidation by DC. We found that neither AC stressing, followed by DC stressing to failure, nor DC coupled with thermal cycling had an effect on the net, DC only, EM performance. All tests, regardless of thermal history and current cycling conditions, resulted in similar DC lifetimes and their distributions. We conclude that for this test structure, only net DC testing time has a significant effect on time to failure. AC stress, thermal history (including cooling to room temperature) and pulsed DC stressing have no effect on electromigration lifetime under DC conditions. This suggests that EM in this test structures depends not on metal microstructure, but only on interface structure. We further conclude that the standard test methodologies, using accelerated DC stress conditions at elevated temperatures, are adequate. However, since only the net DC stress time has any measurable effect on EM lifetime, accelerated testing provides a good predictor for lifetime expectations only for the DC component of continuous operating conditions, but underestimates lifetime expectation under pulsed DC, and variable usage operating conditions by 50% or more, depending on duty cycle. We also conclude that neither AC stress nor DC cycling, and changes to grain structure that may result, vaccinate Cu interconnects against EM failures and thus, do not provide any measurable benefit to net, DC only, EM lifetime.