Abhishek Deshpande, Qian Jiang,and Abhijit Dasgupta
Center for Advanced Life Cycle Engineering (CALCE), University of Maryland, College Park, MD, USA
Functional solder joints experience multiaxial stresses (tensile and shear) as they undergo a combination operational loads such as temperature cycling and out-of-plane PWB flexure/warpage. Research groups often use finite element simulations to quantify stresses and strains in the critical solder joint, by modeling the solder joints as homogenous isotropic volumes. Solder strain (or other similar damage metrics, such as work density) is used to construct fatigue durability curves from fatigue test data and to predict failures under life-cycle loading conditions.In reality, each SAC solder joint consists of few highly anisotropic grains and is neither homogeneous nor isotropic. As a result, homogeneous, isotropic finite element models erroneously misrepresent the true material behavior and neglect the resulting stress concentrations at grain boundaries and triple corners between mis-oriented grains and IMC interfaces. Due to piece-to-piece variability in the grain structure of tested joints, a simple homogeneous isotropic representation leads to significant piece-to-piece uncertainty in predicting the strain levels (and hence the fatigue durability) of each solder joint at any given level of applied loads. A typical approach for dealing with this variability is to test a large number of samples at each loading level and use confidence intervals to determine the statistical variability. However, such a process is resource-intensive and time-consuming as temperature cycling tests can take few months to complete.Therefore, this study aims to quantify the role of the grain structure on the variability in fatigue durability predictions, based on a simulation-based ‘virtualtesting’ alternative. In addition, this study also draws attention to the limitations of modeling solder joints as homogenous isotropic volumes. The approach consists of parametric, grain-scale, anisotropic FEA simulations. Findings of this study can enable more accurate ‘digital twins’ and empower engineers to obtain more accurate, faster and cheaper a-priori estimates about stochastic fatigue reliability predictions.