Abstract
We have examined several different types of test structures which are commonly used today for assessing electromigration reliability. Compared to lifetime measurements using test structures designed with the stress current fed through vias, it is concluded that conventional NIST test structures are inadequate for lifetime determination because they do not incorporate "drift-velocity-type" failures which can occur when current flows between different interconnect levels. Failure to incorporate the "drift-velocity-type" failures in the test structure scheme can lead to dramatic lifetime overestimation especially for small linewidths with bamboo-like microstructures. For Kelvin-type test structures designed to have voltage tapping on the current feeding vias, we have found that the design for the voltage tapping can significantly affect the lifetime measurements. The length of the voltage tap lead needs to be sufficiently larger than the characteristic thermal diffusion length of the migrating atoms at the test conditions for accurate lifetime measurements. Our experimental results also showed that the measured lifetimes were independent of the number of vias used for stress current feeding. This indicates that the effect of current crowding in and around the vias is not a primary factor for determining lifetimes under low stress conditions as chosen in the experiments. This experimental finding agrees well with simulations based on a vacancy model.

This publication has 10 references indexed in Scilit: