Large-Scale Electromigration Statistics for Cu Interconnects

Abstract

Even after the successful introduction of Cu-based metallization, the electromigration failure risk has remained one of the important reliability concerns for advanced process technologies. The observation of strong bimodality for the electron up-flow direction in dual-inlaid Cu interconnects has added complexity, but is now widely accepted. More recently, bimodality has been reported also in down-flow electromigration, leading to very short lifetimes due to small, slit-shaped voids under vias. For a more thorough investigation of these early failure phenomena, specific test structures were designed based on the Wheatstone Bridge technique. The use of these structures enabled an increase of the tested sample size past 1.1 million, allowing a direct analysis of electromigration failure mechanisms at the single-digit ppm regime. Results indicate that down-flow electromigration exhibits bimodality at very small percentage levels, not readily identifiable with standard testing methods. The activation energy for the down-flow early failure mechanism was determined to be 0.83 ± 0.01 eV. Within the small error bounds of this large-scale statistical experiment, this value is deemed to be significantly lower than the usually reported activation energy of 0.90 eV for electromigration-induced diffusion along Cu/SiCN interfaces. Due to the advantages of the Wheatstone Bridge technique, we were also able to expand the experimental temperature range down to 150 °C, coming quite close to typical operating conditions up to 125 °C. As a result of the lowered activation energy, we conclude that the down-flow early failure mode may control the chip lifetime at operating conditions. The slit-like character of the early failure void morphology also raises concerns about the validity of the Blech-effect for this mechanism. A very small amount of Cu depletion may cause failure even before a stress gradient is established. We therefore conducted large-scale statistical experiments close to the critical current density-length product (jL)*. The results indicate that even at very small failure percentages, this critical product seems to extrapolate to about 2900 ± 400 A/cm for SiCOH-based dielectrics, close to previously determined (jL)* products of about 3000 ± 500 A/cm for the same technology node and dielectric material, acquired with single link interconnects. More detailed studies are currently ongoing to verify the extrapolation methods at small percentages. Furthermore, the scaling behavior of the early failure population was investigated.