An improved peel stress-based correlation to predict solder joint reliability of lidded flip chip ball grid array packages

Abstract

Board level solder joint reliability of semiconductor packages during thermal cycling continues to be a concern in different applications. Thermo-mechanical simulations using Finite Element Analysis (FEA) have proved to be quite useful in predicting solder joint fatigue life once correlated with physical evaluations. One of the popular approaches to develop a correlation is to calculate the plastic work density increment at critical solder joint from modeling and correlate with reliability data. However, this approach lumps all simulated damage into a scalar parameter, and does not consider stress direction near the cracking interface. Recent literature indicates that the average peel stress may have an impact on cycles to failure for solder ball. Solder balls under compression may delay crack propagation and result in longer fatigue life. This article proposes an empirical correlation incorporating both average peel stress and plastic work density increment. Simulations are performed for Lidded Flip-chip Ball Grid Array (FCBGA) packages of different body sizes, die sizes, and package materials. The new correlation is developed using reliability data from 20 different thermal cycling tests, all using SAC305 Pb-free alloy. The new peel stress-based correlation shows higher coefficient of determination (R2 improved from 0.78 to 0.88) and better accuracy (maximum error in prediction reduced from 40% to 35%) compared to traditional correlation based on plastic work density increment alone.