Probabilistic effects in thermal cycling failures of high-I/O BGA assemblies

This paper presents a study of the probabilistic effects that act in addition to deterministic (mechanistic) effects to reduce the thermal cycling durability of ball grid array (BGA) interconnects as the component I/O count increases. The mechanistic drivers include increasing thermal expansion mismatch with increasing package size and increasing stress levels with decreasing solder joint size. The most critical joint from this deterministic perspective is usually the one at the outer corner of either the die foot-print or the package foot-print. The probabilistic factors include variabilities in microstructure, interfacial intermetallic layers, joint geometry and void distributions; all of which can place several joints around the critical one at risk, and further reduce the durability of the entire package. Thus, durability can drop as the number of joints in series increase, even if the stress levels do not change. Thus, for large BGAs the mechanistic prediction can overestimate component life. In this paper PBGA-1156 assemblies are subjected to temperature cycling tests and the failure statistics are identified. Using a partitioned daisy-chain design, the durability is found to decrease, as more and more joints are nested together in the critical regions at the package corners. Since all the failed joints do not experience the same thermal expansion mismatch, finite element analysis and fatigue analysis is conducted to normalize all the failure data to a uniform damage level, by quantifying the deterministic (mechanistic) effects. The additional drop in durability with increasing number of joints, after the mechanistic normalization, is attributed to the probabilistic effect of interest. The results suggest that for this example, the probabilistic effects can reduce the deterministic prediction by an order of magnitude, as the number of I/O in the critical region reaches 100.