Abstract

Solder reflow is performed in flip-chip packaging and Surface-Mount Technology (SMT) assembly processes for Ball-Grid Array (BGA) packages. In this process, the two parts (chip-substrate or package-PCB) are heated to a temperature of 200-260C, where the solder melts and forms an interconnection. The solder joints can form different shapes due to both the incoming warpage of the parts and additional CTE mismatch-induced warpage during the attach process. The highly distorted solder shapes result in defects such as solder bridging and non-wet opens. Owing to the large probability for such defects in new package technologies, costly design-of-experiments have to be conducted to establish the process window for maximum assembly yields. Currently, the prevalent techniques for modeling solder reflow, like Surface Evolver or Fluid-Structure Interaction methods, cannot fully capture the interactions between the solid parts and fluid-like molten solder balls, while being efficient enough to render full package simulations computationally feasible. We demonstrate that the novel fully implicit Incompressible Smoothed Particle Galerkin (ISPG) technique implemented in LS-Dyna is very efficient in capturing all relevant physics of solder reflow process. It employs a Lagrangian particle-based solution of the Navier-Stokes equation for molten solder and can be easily coupled with structural physics within one (LS-Dyna) solver. In this technique, the solder shape evolution is governed by the surface tension of solder material and structural boundary conditions, while the contact angle controls the interfacial behavior. We show that ISPG technique can accurately predict solder joint shapes, while interacting with the Copper pad and Solder Mask layer on the substrate and PCB sides. We also demonstrate the capability to simulate large-scale reflow involving hundreds of solder joints where the package and PCB warpage are modeled. This shows great potential in fine-tuning packaging assembly process parameters through guidance from simulation.