Influence of thermal heating effect on pulsed DC electromigration result analysis

Abstract

Pulsed DC electromigration (EM) tests up to 1 MHz have been performed on single level Al-0.5%Cu metallization. The results are in good agreement with an average current model when the Joule heating has been taken into account with an original thermal model. The developed thermal model gives good approximation T pdc of the average temperature needed for the exploitation of pulsed electromigration tests, as a function of the stress parameters; frequency, duty cycle and current density. The difference between T pdc and T eff, deduced from the mean temporal value of the atomic flux, is less than 0.4% for our test conditions. Thus, median time to failure (MTF) increase in the MHz region previously reported by some authors, appears as a thermal effect. The activation energy calculated for 10 kHz is equal to the d.c. value, suggesting the same failure mechanisms in both cases. Furthermore, scanning electron microscope (SEM) observations showed very large metal accumulations and hillocks which were due to an easy interface diffusion. These accumulations are larger in the PDC case than in d.c. experiments, and their size has been correlated to the experimental resistance decrease.