Experimental comparison of wafer level reliability (WLR) and packaged electromigration tests

The wafer level reliability (WLR) electromigration (EM) test is a quick test under very highly accelerated conditions using a parametric tester and prober on a full-sized wafer, which can provide cost effective and timely feedback on possible reliability degradation due to process variation/modification, equipment change, or misprocessing (Pierce and Snyder, 1997; McPherson, 1996). Under these highly accelerated conditions, however, there is no guarantee that the failure mechanism is the same as that for less aggressive test or operating conditions. Therefore, the validity of the WLR EM test remains controversial (Lloyd, 1992; Pierce and Brusius, 1993; Liu et al., 1999; Loh et al, 1998). In this study, we deliberately applied various realistic misprocessings on the back-end process that potentially could affect the metal reliability but may not be detectable by parametric test, and investigated the sensitivity of WLR EM, packaged EM, and parametric tests on the process variations and the correlation amongst them.