Assembly and reliability of lead-free flip chips

The small dimensions and the integrated layer structure of a typical underfilled flip chip assembly lead to some rather unique concerns in terms of process windows, assembly yields, and reliability. In driving towards no-Pb soldering, it is important to keep all of these and the consequences for design, materials selection and process optimization in mind. Experimental results for Ball Grid Array (BGA) or SMT components are not readily applicable for flip chips.; This dissertation focuses on understanding the soldering phenomenon for leadfree flip chips. First, the potential lead-free solder alternatives were identified for flip chip applications. Four selected alloys were evaluated for wetting and solder joint collapse under two substantially different conditions, soldering to a blanket Organic Solder Protect (OSP) coated copper surface and to individual contact pads on substrates designed to match the die, respectively. Based on the results from these experiments, the 95.5Sn/3.5Ag/1.0Cu (LF-2) soldering system was considered for a detailed investigation.; LF-2 solder joints reflowed quite readily at moderate peak temperatures and time above the liquidus temperature. Minimum temperatures and times shown to give full collapse were 238°C and 40 seconds when the bumps were dipped in a 2 mil thick film of flux. Unfortunately, wetting and collapse rapidly became sensitive to flux thicknesses below 2 mils, i.e. the process window is narrow. Solder wetting and collapse did not depend on the flux type (for the fluxes considered). Almost complete collapse was achieved on OSP coated copper pads whereas the Sn/Ag/Cu solder does not collapse completely when placed onto Ni/Au pads.; Even though reflow parameters did not affect solder joint collapse, there was a significant difference in the solder joint properties for different reflow parameters. These differences in the solder joint properties resulted in a difference in the mechanical strength of the solder joint. The fatigue resistance of both, encapsulated and non-encapsulated LF-2 joints was significantly lower on Ni/Au-pads than on Cu/OSP. There was faster failure with slower delamination for the assemblies on Ni/Au pads when solder fatigue was the dominant failure mechanism. Local adhesion of the underfill to the chip passivation and the solder surface was dependent on the combination of the flux, the underfill material, the solder alloy and the substrate pad metallurgy. Thermal shock testing of underfilled LF-2 assemblies showed good performance when the thickness of the encapsulant fillet was properly controlled. Very thick and very thin fillets resulted in corner delamination and immediate failure when the delamination reached the solder joints. Fatigue failures were observed when the thickness of the encapsulant fillet was controlled to delay/eliminate corner delamination.