Search for a command to run...
Extremely high-density and fine-pitch applications, such as artificial intelligence and high-performance computing, require advanced interconnect techniques e.g., Cu-Cu thermal compression bonding and bump-less direct CuCu hybrid bonding. The CuCu joints provide lower electrical resistivity, allow extremely fine-pitch, and result in lower electromigration as compared to solder-based C4 and C2 interconnects. In recent years, hybrid bonding has gained a lot of traction from applications like high bandwidth memory, CMOS image sensors, 3D system-on-chip, primarily due to its ability to achieve the interconnect size and pitch down to 1 micron or even below. However, hybrid bonding technique incurs higher infrastructure, equipment, and material processing costs, thus limiting its implementation to only semiconductor foundries. Cu-Cu TCB on the other hand provides a viable economical alternative to achieve high-density, fine-pitch interconnects while taking advantage of the existing infrastructure in current OSATS ecosystem. KS and collaboration partners have published serval papers highlighting the fundamental understanding of Cu-Cu bonding process including material preparation, processing, and equipment-related challenges. For example, surface roughness is key parameter defining the bonding pressure required to flatten asperities at the mating interfaces. Co-planarity of both chip and substrate is also extremely critical since part of the applied pressure is utilized to bring two entities in intimate contact. A high-yield Cu-Cu TCB process can be successfully performed in a Class-1000 clean room, which is impossible for a competitive hybrid bonding type process. Cu-Cu TCB is less sensitive to foreign material interference although we do emphasize to keep the bonding environment as clean as possible since minimizing foreign interference will always guarantee higher yields. Most of KS prior research is dedicated to the surface morphology that only includes the Cu-pillars protruding out of the chip surface. This structural configuration ensures intimate contact of Cu-pillars to the Cu-pads and further allows the Cu-pillars to slightly deform under pressure and thus ensuring a reliable bond. This configuration, however, surfers from the fact that there is now a tiny gap between the chip and substrate surface, which at least requires environmental protection or at most complete underfilling. There is still no consensus whether underfill is necessary for the Cu-Cu TCB assemblies. For example, the chiplets used for heterogenous integration typically involve 50% or more coverage of the Cu-pillars over the entire chip area. This should provide enough mechanical strength so that no reinforcement from underfill is required as well as very good thermal conductivity. However, environmental protection of the pillars is still a huge concern since interconnects can corrode over time. Underfilling a tiny gap is challenging and currently there is no way of reliably filling it. CVD type coatings such as ALD-Al2O3 are promising. However, the deposition process is very slow and adds costs to the process. Recently, we have been focusing on pre-applied organic underfil materials that would circumvent the tiny gap issue. In this work we will highlight some key results. Another area we are focusing is to utilize the current hybrid-type test vehicle construction where Cu is slightly dished i.e., a few nm below the oxide surface. Exposing both chip and substrate to TCB-like temperatures would force the copper to protrude slightly out of surface and thus making an intimate contact between copper surfaces and forming a copper first hybrid bond. Cu-Cu TCB takes advantage of the novel KS flux-less TCB technology, which uses in-situ FA vapor to remove native oxides on the copper surface. In-situ oxide removal is a key enabler for the successful implementation of the Cu-Cu TCB.