Study of Capillary Underfill Filler Separation in Advanced Flip Chip Packages

The key role that underfill materials play in highly reliable, advanced flip chip organic packages has generated an increased focus on their behavior and structure. One such behavior relates to the observation of filler separation from the resin matrix which, to date, has been predominantly attributed to gravity or capillary flow. The phenomenon of silica filler separation is discussed in the context of fine pitch, lead-free solder joints with copper-base (pedestal or pillar) under bump metallization and large die packages. The principle mechanism driving filler separation in these structures was confirmed as a migration of the electrostatically charged filler particles away from the copper regions and towards the solder regions of the interconnect. Based on this finding, various factors that influence the surface of the interconnects or the nature and the mobility of the filler particles during the bond and assembly process were explored. It was found that the oxide states and contact angles of the interconnect surfaces do not appear to impact the degree of filler separation. Within the range explored, average filler particle size is ineffective in changing the separation behavior. On the other hand, lower filler content somewhat increases the extent of separation and is believed to be related to an increase in particle mobility. Assembly process variables with known effects on surface interactions and underfill flow were also studied, revealing no observable shift in the occurrence of filler separation. Finally, and most importantly, a reliability study was conducted to investigate the impact of this phenomenon in a very large die (23 × 23 mm2) flip chip organic package subjected to a high level of thermomechanical stress. Using extended Deep Thermal Cycling to 2000 cycles (as opposed to the standard 1000 cycle criterion), no packaging failures occurred and no signs of interconnect degradation were observed. These results are consistent with finite element modeling of the tested package, which showed that stress changes from filler separation in regions of similar dimensions to those that were experimentally observed were within the limits of model error and typical manufacturing variability.