Effect of preformed IMC Layer on Electromigration of Solder Capped Cu Pillar Bump Interconnection on an organic substrate

Abstract

The electromigration behavior of 80 μm pitch solder capped Cu pillar bump interconnection on an organic carrier is studied and discussed. Recently the solder capped Cu pillar bump technology has been widely used in mobile applications as a peripheral ultra fine pitch flip chip interconnection technique. The solder capped Cu pillar bumps are formed on Al pads which are commonly used in wirebonding technique. It allows us an easy control of the space between the die and the substrate simply by varying the Cu pillar height. Since the control of the collapse of the solder bumps is not necessary, the technology is called the “C2 (Chip Connection)”. Solder capped Cu pillar bumps are connected to OSP surface treated Cu substrate pads on an organic substrate by reflow with a no-clean process, hence the C2 is a low cost ultra fine pitch flip chip interconnection technology. It is an ideal technology for the systems requiring fine pitch structures. In 2011, the EM tests were performed on 80 μm pitch solder capped Cu pillar bump interconnections and the effects of Ni barrier layers on the Cu pillars and the preformed intermetallic compound (IMC) layers on the EM tests were studied. The EM test conditions of the test vehicles were 7–10 kA/cm2 at 125–170°C. The Cu pillar height was 45 μm and the solder height was 25 μm. The solder composition was Sn-2.5Ag. Aged condition for pre-formed IMCs was 2,000 hours at 150°C. It was shown that the formation of the pre-formed IMC layers and the insertion of Ni barrier layers are effective in reducing the Cu atoms dissolution. In this report, it is studied that which of the IMC layers, Cu3Sn or Cu6Sn5, is more effective in preventing the Cu atom dissolution. The cross-sectional analyses of the joints after the 2000 hours of the test with 7kA/cm2 at 170°C were performed for this purpose. The relationship between the thickness of Cu3Sn IMC layer and the Cu migration is also studied by performing the current stress tests on the joints with controlled Cu3Sn IMC thicknesses. The samples were thermally aged prior to the tests at a higher temperature (200°C) and in a shorter time (10–50 hours) than the previous experiments. The cross-sectional analyses of the Sn-2.5Ag joints without pre-aging consisting mostly of Cu6Sn5, showed a significant Cu dissolution while the Cu dissolution was not detected for the pre-aged joints with thick Cu3Sn layers. A large number of Kirkendall voids were also observed in the joints without pre-aging. The current stress tests on the controlled Cu3Sn joints showed that Cu3Sn layer thickness of more than 1.5 μm is effective in reducing Cu dissolution in the joints.