Effects of Thermal Cycling on Creep of SnAgCu Solder Joints

Abstract

The long-term life of properly optimized high-end microelectronics assemblies is commonly limited by fatigue of the solder joints. The most credible assessments of this rely on modeling-based extrapolation and generalization of accelerated test results. As the validity of extrapolations to long-term service can rarely if ever be verified experimentally, there is a strong need for mechanistically justified models. The models all rely on predictions of stresses, strains, and perhaps functions of those versus time and temperature in cycling. Finite-element-based predictions of these require the assumption of relatively complex constitutive relations capable of predicting inelastic strain rates under continuously varying stresses and temperatures. These are usually extracted from a series of creep or stress relaxation experiments and generalized under the implicit assumption that the inelastic deformation is dominated by dislocation creep. However, the resulting constitutive relations were recently shown to be strongly misleading. It was suggested that this is a result of the inelastic deformation at the stresses and temperatures of primary concern actually being dominated by dislocation core diffusion. This paper lends further support for this picture, showing additional contributions from diffusion on cycling induced dislocation structures to make for ongoing, usually nonmonotonic, variations in deformation rates from cycle to cycle. These effects are certain to have major consequences for modeling. In fact, it has been argued that the combination of wrong constitutive relations with a wrong damage function has led to a partial cancellation of errors across a limited parameter range, but that this cannot be counted on for more detailed predictions of accelerated test results and does not allow for interpretation of such results in terms of even qualitative trends in service.