Moldability is a crucial aspect of flip chip technology. It is an increasing challenge to ensure moldability with rapid advances in flip chip technology such as decreasing bump pitch and stand-off height, especially when commercial molded underfill (MUF) is used and, in particular, during panel level molding. One key challenge faced is severe void entrapment beneath the die. Typically, large DOE matrix experiments are used to address this issue, which require significant time and process resources. 3D flow simulation can be used to optimize the process to reduce defects with a smaller number of actual runs. By correlating theoretical and experimental phenomena, flow simulation enhances the understanding of the complex fluid dynamics during the molding process. 3D flow simulation can assist in widening the process window, which is limited by the inherent machine and material challenges. This can be achieved by prediction of the effect of varying design, material, and process parameters on melt front behavior and void locations. 3D mold flow simulation using Moldex3D V10 is used to optimize the MUF transfer molding on selected flip chip devices.
This paper proposes and verifies a systematic flow simulation methodology designed to save computational resources by using a three step analysis. The initial step, simplified panel level simulation, is to optimize the process parameters to obtain a balanced melt front. Next, on the package level, we studied the effect of various parameters. This analysis provides a prediction of the void location and an insight into the appropriate parameters to minimize the void problem. The optimized parameters from the preliminary simulation were used as guidelines. For the second step, a full validation was conducted. A complete full panel-level flow model was built, where the process and design parameters adopted in the actual molding were implemented. The actual void location and size from the experiment were captured by scanning acoustic microscope (SAM) machine and parallel lapping (p-lapping). Short shots were also obtained to study the melt front behavior. The panel mold filling simulations showed good correlation with the experimental short shots and actual void locations. The prediction capability is further enhanced by zooming in to the column level, and this enhanced model was able to predict the other lower risk voids away from the main problem areas. This was correlated with actual CSAM data and p-lapping.
The 3D flow simulation enhances the understanding of causes of flow imbalance, void signature, void formation, and the effect of varying bump height, die thickness, mold cap thickness, gate height, die orientation, transfer profile, and mold temperature as potential enhancement measures. With a successful correlation between simulation and process data as shown in this paper, we have demonstrated that mold flow simulation is a reliable tool to effectively reduce the design-to-implementation cycle time, identifying potential key problems during actual fabrication and potential solutions to reduce defects.
