Towards a quantitative mechanistic understanding of the thermal cycling of SnAgCu solder joints

Microelectronics manufacturers continue to subject a wide range of products or representative test vehicles to accelerated thermal cycling tests. Most such testing is focused on comparisons, whether among alternatives or to an established requirement. However, more often than commonly recognized such comparisons may not reflect the relative performances in service. In fact, most current models have been shown to fail to account for important systematic trends as well as being inconsistent with our current understanding of the failure rate controlling damage mechanism. An alternative mechanistically justified model for the thermal fatigue life of SnAgCu solder joints has been proposed. Damage and failure occurs by recrystallization of the large Sn grains across the high strain region of the joint, followed by crack growth along the resulting network of high angle grain boundaries. The recrystallization was shown to be the damage rate controlling mechanism, except for extremely high strain assemblies and/or harsh cycling conditions, i.e. if we can predict the recrystallization we can predict the number of cycles to failure. So far the model accounts for a variety of important trends and offers guidance as to the interpretation and generalization of accelerated test results. General extrapolations towards service conditions will, however, require the specific functional dependence of the rate of recrystallization on the stress and the precipitate distributions. Another potential difficulty is that constitutive relations are extracted from single sided creep experiments while the dislocation cell structures built up under cyclic loading are certain to be different. Furthermore, the repeated `annealing´ during the high temperature dwell affects the hardening. Even if these effects could be ignored the ongoing evolution of the constitutive relations would still effectively prevent the extraction of the above-mentioned functional dependence from comparisons between therm- l cycling results and FEM. A special experiment is ongoing in which stresses and strains on the solder joints can be controlled and measured directly, allowing the testing of individual assumptions underlying the proposed model. Preliminary results are presented and compared to results of thermal cycling across different temperature ranges.