Sporadic Degradation in Board Level Drop Reliability - Those Aren't All Kirkendall Voids!

Abstract

The sporadic occurrence and growth of voids within the Cu3Sn intermetallic layer formed when soldering to copper is the cause of growing concern for microelectronics manufacturers. Overlooked for decades, and commonly underestimated when observed, there are reasons why the phenomenon is becoming a threat to ever more products. In fact, while rarely severe the problem has been known to cost individual companies anywhere from tens to hundreds of millions of dollars. The voiding has been shown to depend on some yet to be identified property of the plated Cu, rather than the solder, or materials, or the assembly process. Effective screening processes have been developed that would have prevented all the losses alluded to above, but there is clearly a long term need for us to learn how to control and preferably prevent the problem. We are busily researching potential remedies, but a safe resolution to this highly variable problem may well require a fundamental understanding of the underlying mechanisms. The phenomenon is commonly referred to as Kirkendall voiding, albeit without any arguments or data to support this assumption. In the present work we argue, however, that there must be more to it than a simple 'Kirkendall effect'. Along the way we present data showing Cu to be by far the dominant diffusing species in the Cu3Sn layer, a result which is important for the interpretation of experimental void observations. A review of the literature revealed only a single marker experiment contradicting this conclusion, and we argue that the 'markers' in that experiment could not have been at the Kirkendall plane.