Multi-via electromigration test structures for identification and characterization of different failure mechanisms

Abstract

Experiments on three-terminal 'dotted-I' test structures (copper metal lines with vias at both ends and an additional via at the center) show that the mortality of a single segment not only depends on the values of its current density and length, but also on the stress conditions in the linked segment. The current density in one 25μm long segment was fixed at 0.5MA/cm 2, with electron flow toward the central via. In the other segment, the current magnitude and sign were varied for different test populations, with the current varied from 2.5MA/cm2 to -2.5MA/cm2 with intermediate values including zero. For all cases, some test structures survived for the full 780 hours of testing and some did not. The percent of the lines that failed increased monotonically with an effective jL product defined as the maximum of the sum of the jL products from all paths through the structure. However, some lines with smaller than expected effective jL products failed, and some lines with relatively large effective jL products did not. Simulations of electromigration and electrotnigration-induced failures for all test conditions have been carried out. We find that test conditions leading to extreme values of the effective jL product probe different failure mechanisms than those associated with intermediate effective jL products. It is also shown that the definition of the effective jL product must be modified to account for zero current (inactive) segments that act as reservoirs or sinks. Multi-via test structures in general, and dotted-I test structures specifically, are shown to be versatile tools for identification and characterization of different failure mechanisms and length effects through the use of different test conditions with a single fixed structure. © 2005 Materials Research Society.