Léti and Brewer Science: sharing a common lab and a common goal

With the recent announcement of CEA-Léti and Brewer Science's agreement to create a common lab for further development of 3D stacking processes using Léti’s thin wafer handling processes and Brewer Science, Inc. (BSI) WaferBOND™ materials, I realized although both names popped up frequently in relation to the EMC^3D Consortium, they tend to keep a low profile. An interview was definitely in order to learn more about the roles they play in 3D IC integration developments. Fortunately Andre Rouzaud and Nicolas Sillon of Léti, and Rama Puligadda and Mark Privett of BSI, were available to answer my questions and provide more in-depth information about the collaboration. Here’s what I learned from them.

Léti’s research field in 3D integration is driven by the needs for increasing the density of functions, with an ultimate goal of heterogeneous integration of functions such as RF, logic, memory, MEMS, using a variety of technologies to achieve these processes at a lower cost. Léti develops methods to use short vertical interconnects between stacked dies, specifically involving TSVs. Since a broad spectrum of options must be investigated, it makes sense to form alliances and consortia to help accelerate commercialization of technology.

“The partnership collaborative model we have at Léti is quite special,” explained Rouzaud. “Although open to multi-partner projects (mainly for upstream research), we promote bilateral collaborations with specific identified partners on determined research domains.” Leti’s involvement with EMC3D is an example of a multi-partner project, while the partnership with BSI (also a member of EMC3D) is an example of a bilateral collaboration. In the case of the latter, part of the program involves the establishment of a mid-term common laboratory where a dedicated Léti team works closely with partner assignees (in this case, BSI) in the Léti premises on commonly defined programs that are periodically assessed, adapted, and adjusted through regular Léti/partner steering committees. This flexibility in the common project evolution allows the partner to take into account possible changes in his market environment or in his business strategy.

In the area of temporary bonding and protective materials, Léti is clearly committed with BSI. While the scope of work being done by Leti and BSI goes beyond their affiliation with EMC3D, extending into specific customer projects, Privett noted that information is shared as it relates to the EMC3D goal of developing TSV process at a cost of <$150/wafer. Leti and BSI’s focus in the common lab agreement is in developing standard processes for TSVs. Specifically, they have been working in the area of temporary bonding and protective materials for handling thin wafers. Process development is geared toward either via first or last, die to wafer (D2W) or wafer to wafer (W2W), in the foundry or in the packaging house. Rama says the goal is to be prepared for all scenarios. Although the processes might require slight adaptation, ultimately they can be done either place. “The partner will choose the solution,” she explained. “Temporary bonding and debonding can be used for both chip to wafer and wafer to wafer processes. The silicon still has to go through back grinding and subsequent post-fab processes for either.”

“We feel via last will be adopted before via first, because the first demonstrations with CMOS image sensors have been done with via last, and no adjustments have to be made to front end processes to achieve that.” noted Sillon.

Although the initial common lab agreement between Léti and BSI is set for 3 years, the flexibility of the arrangement allows for it to change as the project evolves. “We are focused together on the roadmap of 3D integration,” noted Rouzaud. “I suspect it will not be done in 3 years.” Privett added that market research(Yole Développment) points to 3D IC integration being in high volume manufacturing by 2015, at which point only 20% of all memory will be manufactured this way (51% wire bonded, 30% discrete). He said that leaves 80% worth of memory still to be addressed. “Hopefully we’ll still be talking about 3D integration in 3 years,” he said. In the mean time, notes Puligadda, “It’s been a wonderful relationship.” – F.v.T.