Thumbnail Image
Publication or External Link
Manoharan, Subramani
McCluskey, Patrick
Copper (Cu) wire bonds have become the dominant wire material used in microelectronic packages, having replaced gold (Au) in the majority of applications. Cost saving has been the key factor to drive this transition in wire bond material, although there are other advantages to Cu such as better electrical and thermal conductivity, reduced wire sweep during transfer molding and most importantly slower intermetallic compound (IMC) formation with Al (bond pad). Although IMC layers are much thinner than for Au-Al bonded joints, growth of second phase, Cu9Al4, due to exposure to high temperature leads to interfacial separation, which is exacerbated under thermal cycling condition ultimately leading to failure of the joint. Part I of this dissertation aims at addressing the effect of combined loading (thermal aging and cycling) on the reliability of Cu wire bonded devices using a unique long dwell thermal cycling profile that accelerates growth of different IMC phases (CuAl2 and Cu9Al4) and accelerates failure due to CTE mismatch between epoxy mold compound, die and Cu wire bond. Unlike many of the studies presented in literature, the test vehicle in this study are made of commercial off-the-shelf (COTS) parts, where a multitude of factors vary from one another, such as wire diameter, wire bond and bond pad characteristics, etc., the combination of which play a significant role in the life time of these devices and is not fully captured by first-principal models. Hence, a data-based life estimation method is developed, to aid in part selection based on initial bond characteristics. Critical parameters of wire bond that contribute to reliability are identified, the most significant of which is Al bond pad thickness, which controls the growth of IMC and influences time for Cu9Al4 IMC phase formation. Second part of this work is focused entirely on the Al bond pad thickness. Part II-A focuses on the qualitative comparison of pad thickness effect on the quality of initially formed bond through use of bond shear analysis and the effect of bond interface aging on bond shear analysis. Test vehicle consists of three pad thicknesses namely, 0.5 µm, 1 µm and 4 µm, over which Cu wirebonds with four different thermosonic bond recipes are made. Results from Part II-A provide guidelines for bond comparison using bond shear analysis. Part II-B focuses on the effect of bond pad thickness on the reliability of Cu wire bonds under isothermal aging at 175°C and 200°C for 1000 hours and 650 hours respectively. Test vehicle in this study consists of 0.675 µm and 3 µm pad thickness on silicon die in 20 leaded 5x5 QFN package. Wire bonds with one thermosonic bonding recipe are made on all the 90 packages used in the study. Electrical resistance and cross-sectional analysis are used to derive failure times, which is in turn used to build empirical relationship between pad thickness and time to failure. Result from this study shows longer time to failure for wire bonds on 3 µm pad compared to 0.675 µm pad due to delay in Cu9Al4 formation.