Digital-image correlation and XFEM based shock-reliability models for leadfree and advanced interconnects

Models for crack initiation and propagation in lead free solders under shock-impact events have been developed using extended finite element method (XFEM). XFEM enables the modeling the solder interconnect without explicitly meshing the crack surface. The crack propagated in crack domain along the solution dependent path with no requirement of re-meshing the model. Interface damage properties of the copper-solder interface have been characterized at high strain rate on a bi-material specimen using high-speed imaging in conjunction with digital image correlation. Damage properties have been used as input to the XFEM models. The XFEM submodels have been developed using node-based submodeling, transient displacement and velocity histories from digital image correlation measurements with high-speed imaging of board assemblies during shock. Four interconnect types have been modeled using XFEM including Sn3Ag0.5Cu, 90Pb10Sn, Cu-Reinforced Columns, and 63Sn37Pb interconnects on ceramic ball-grid arrays. In addition, Sn3Ag0.5Cu on plastic ball-grid arrays have also been modeled. Extended finite element models have been correlated with cohesive-zone models along with experimental results. The board assemblies have been tested at 1,500g and 12,500g. The failed assemblies have been cross-sectioned and the failure modes correlated with model predictions. The predicted failure modes for all four interconnect types correlate well with the observed locations for failure. Previously, the authors have developed explicit finite element models, cohesive-zone models, and global-local models for prediction of transient dynamics and life prediction of electronics [Lall 2004, 2005, 2006
^a-c
, 2007
^a-e
, 2008
^a-d
]. Previous researchers have applied XFEM in various other fields such as concrete, composite materials [Unger 2007, Hettich 2008]. Damage and life prediction of transient dynamics in electronics interconnects using XFEM in conjunction with digital image correlation and explicit submodeling is new.