A NONLINEAR FRACTURE MECHANICS PERSPECTIVE ON SOLDER JOINT FAILURE: GOING BEYOND THE COFFIN-MANSON EQUATION D. Bhate, G. Subbarayan School of Mechanical Engineering, Purdue University, West Lafayette, IN 47907-2088 Phone: (765) 494-9770 Fax: (765) 494-0539 Email: ganeshs@purdue.edu ABSTRACT Predicting the fatigue life of solder interconnections is a challenge due to the complex nonlinear behavior of solder alloys and the load history. Long experience with Sn-Pb solder alloys together with empirical fatigue life models such as the Coffin-Manson rule have helped us identify reliable choices among package design alternatives. However, for the currently popular Pb-free choice of SnAgCu solder joints, designing accelerated thermal cycling tests and estimating the fatigue life are challenged by the significantly different creep behavior relative to Sn-Pb alloys. In this paper, a hybrid fatigue modeling approach inspired by nonlinear fracture mechanics is developed to predict the crack trajectory and fatigue life of a solder interconnection subjected to both isothermal accelerated thermal and anisothermal power cycling conditions. The model is shown to be similar to well accepted cohesive zone models in its approach and application and is anticipated to be computationally more efficient in a finite element setting. The approach goes beyond empirical modeling in accurately predicting crack trajectories. It is argued that such non-empirical models that capture the physics of material degradation and failure can form the basis for determining meaningful Pb-free solder environmental testing conditions as well as the acceleration factors relative to field use. KEY WORDS: Lead-free solder, acceleration factors, cohesive zone modeling NOMENCLATURE D disturbance measure T traction z material constant w dissipated energy S entropy t time T temperature Greek symbols η characteristic life β shape parameter ξ D equivalent inelastic strain ξ c material constant δ separation δ c characteristic separation σ c cohesive stress φ(δ) area under the traction separation curve from origin to instantaneous separation (δ) δ max separation when unloading commences ξ internal variable such as the plastic strain trajectory φ free energy α i factors such as environmental effects INTRODUCTION Solder joint fatigue has been the subject of a great deal of study. The reason for this is easily understood on examining the nature of the problem: complex and non-standardized packaging geometries, variable environmental conditions that impose strong thermomechanical effects and perhaps most importantly, the complex creep and rate dependent viscoplastic material behavior of the solder alloys themselves. The drive towards lead-free solders has only served to add more complications to this problem on account of their unique creep properties [1] and microstructure [2]. Despite these complexities, the microelectronic packaging industry primarily relies on the tried and trusted technique of imposing controlled thermal cycling on electronic packages in an experimental environment, estimating Weibull characteristic lives and shape parameters and relating this to empirical rules such as the Coffin-Manson rule and its variants [3]. It is common knowledge that this technique does not deal with the physics of the problem at hand: as such, predictions of fatigue life for different material and geometry systems is not feasible unless thermal cycling tests are repeated for the various electronic packages under consideration. Solder fatigue failure is at its essence, a fatigue crack growth problem. It is therefore natural that a non-empirical understanding of this problem can only result from adopting a fracture mechanics approach. While linear elastic fracture mechanics (LEFM) does provide approaches such as the Paris law [4] that deal with fatigue crack growth, the assumptions made in these approaches are almost always not valid for studying crack growth in solder interconnections. This is primarily due to the fact that typical fatigue failures in solder involve large cracks (relative to pad size) and large scale yielding, both of which invalidate the use of LEFM [4]. One of the reasons LEFM breaks down for this class of problems is because it does not explicitly deal with the specific nature of material degradation in the vicinity of the crack. It has been shown that fatigue crack propagation is the end result of accumulating degradation in front of the crack tip (a region called as the process zone) either in the form of microcracking 1220 0-7803-9524-7/06/$20.00/©2006 IEEE