Probabilistic design approach for cyclic fatigue life prediction of microelectronic interconnects

Deterministic approaches may predict the life of solder joint interconnects used in microelectronic devices far different from the test results and field environment. This difference is caused by uncertainties introduced in finite element and damage modeling with different random variables such as material properties, geometry, damage model constants, and many others. This paper presents a methodology to include these uncertainties in the life prediction approach for solder joint materials. The approach is implemented in predicting the life of ball grid array Pb-free solder joints under thermo-mechanical cyclic loading using the energy partitioning (E-P) damage model. The sensitivity of the finite element results to uncertainties in elastic and inelastic material properties are investigated using this probabilistic approach. In addition final life prediction sensitivity to uncertainties in damage model constants are studied and compared with the effect of uncertainties in material properties. The analyses show that from material properties, creep coefficient and activation energy are two factors that have significant effect in fluctuating the prediction results. Coefficient of thermal expansion was also found to have a strong effect. When compared with the damage model constants, material properties are found to have a negligible effect suggesting a more cautious use of the damage models and constants. Damage model exponents show a more significant effect than damage model coefficients.