Underfill acceleration factor based on thermal fatigue crack growth rate

A common failure mechanism of organic packages during thermal cycling is underfill cracking, which then leads to solder joint failure. To prevent failure of the package the resistance to fracture and fatigue crack growth in the underfill must be greater than the thermal stress loading. Fracture and fatigue behaviors are conventionally characterized by mechanically induced fracture and fatigue tests despite the fact that fatigue fracture in actual packages is thermally induced during temperature changes. As previously reported [1], we developed a thermally induced fracture testing method in which cracks in fracture samples were grown without applying external mechanical stress but via inducing a coefficient of thermal expansion (CTE) mismatch thermal stress. Preliminary thermal fatigue crack growth rate data [2] were presented including a discussion of thermal fatigue fracture methodology development. In this work, we extend the applicability of the temperature-dependent thermal fatigue fracture behavior to a material-level acceleration factor with the goal to aid material selection and package lifetime prediction. Conventional acceleration factors (AF) are typically measured/predicted at package/component levels. The material-level acceleration factor in this work can be used for material selection criterion and package lifetime prediction to improve package reliability. Underfill cracks were thermally fatigued by applying two temperature spans (100 and 130°C) and the numbers of cycles to fast fracture (Nf) were measured. The thermal stress intensity factor (SIF) was calculated using measured crack length and thermal loading and correlated to Nf. We introduced an equivalent temperature (Teqv) and used it to project AF at temperatures different from the testing temperature. An analytical form of AF was determined at the minimum temperature (Tmin) of 20°C and used to compare the measurement at Tmin- - of 50°C. Good agreement was obtained. The crack growth rates resulting from each thermal cyclic load, da/dN vs. KT, were fit using a Paris law.