Interconnect Mechanical Performance with Variable UBM and Decreasing Bump Size

Rebecca Wheeling, Ph.D., Jeier Yang, Matt Jordan, N. Scott Bobbitt, Ben White, Mieko Hirabayashi
Sandia National Laboratories, NM, USA

Indium is used to integrate disparate semiconductor materials because of its ability to cold weld and its high ductility, even at cryogenic temperatures. Previous work used a cryogenic focus ion beam (cryo-FIB) and scanning electron microscopy (SEM) to quantify the kinetics if intermetallic growth for 3 separate underbump metallurgies (UBMs) found in literature: Thin Ti/Ni, thin Ti/Ni/Au, and thick Ti/Ni.

Current work seeks to better understand how the indium (In) bump size affects the respective interface kinetics and subsequent mechanical properties. Indium bumps ranging from 4-14 um were aged 1 day at 125 °C (guided by the previous results). Interfacial evolution was characterized and compared using the cryo-FIB technique. Effect on mechanical performance was evaluated by shearing as-fabricated and aged bumps. Atomistic modeling of the interface reactions, relying on density functional theory and molecular dynamics, will complement the metallurgical and mechanical analyses.

Pure indium was selected for this study because it is a commonly used single element interconnect in electronic applications that readily reacts, so it serves as a simpler case for modeling. The basis of this study will be used for board-level SnPb and Pb-free solder interconnects, where continuum modeling dominates current lifetime predictions. The eventual objective is to determine if/when interconnect sizes approach a size scale that requires atomistic considerations to maintain accurate solder behavior predictions.

1. Fabrication method has a distinct effect on bond strength. UBM 3 (TiCuNi) produced the highest strength bumps in the as-fabricated condition. More interfacial failures associated with UBM 1 (TiNi) lowered the peak strength relative to UBM 2 (TiNiAu) and suggest that an effective metallurgical bond may not be forming upon fabrication. The added Au in UBM 2 supports the formation of an IMC stronger than just the product of TiNi bonding, but not brittle enough to induce interfacial failure.
2. Initial diffusion between the TiNi layers may be supporting a solid solution composition rather than an ordered IMC. Continued aging appears to promote IMC formation. Strength implications would be expected between a solid solution alloy and an ordered IMC.
3. We expect the aged bumps to behave much differently than the as-received bumps. This upcoming data will be reported.
4. DFT calculations predict that the IMCs are more brittle than In and also prone to forming vacancy defects. This suggests that IMC formation could compromise the strength of the In-Ni interface, possibly resulting in failure.
5. Bump size between 4 and 14 um do not appear to impact reaction rate, but the smaller bumps contain a larger fraction of the reaction product than the larger bumps. Strength correlations to fraction of IMC are expected.
6. Understanding the application requirements for these In bump arrays will be key in driving the necessary fabrication methods, potential heat treatments post-fab, next assembly processing, etc. in order to produce a desired interfacial microstructure to support the desired mechanical and electrical performance and reliability over time.