Want to know how to torture a journalist? Invite them to present at a conference but ask them to sign an NDA so they can’t write about it! However, I can say this much: the 3D Program is alive and well at IBM. From IBM Fellow, Subramanian Iyer, I learned some basic truths about scaling and 3D. One of the current challenges being faced where 3D can provide the solution is in the power budget. Communication between chips can take up to a third of the system power. “Semiconductor scaling doesn’t address this issue, but tighter inter-chip connections can,” notes Iyer. “3D does not eliminate the need to scale, it’s an orthogonal element for increasing performance.” He added that the cost per transistor still needs to come down.
I WAS given the green light to talk about the partner vendor’s progress in 3D-related technologies. What was most interesting to me was that next-generation steps are already being taken before 3D ICs have been fully commercialized. Rather than sitting around twiddling their thumbs waiting for the first wave of 3D ICs take off, their R&D divisions are hard at work improving current technologies; many focusing on system enhancements for high volume manufacturing (HVM), solutions to lower cost of ownership, and improve process flows through material advancements.
Noting that 3M is not just about scotch tape and post-it notes (although I truly believe post-it notes might be one of their finest inventions) Mike Bowman updated the JDP partners on the company’s next-generation wafer support system solution for thin wafer handling. This issue continues to be a sticking point (no pun intended) for 3D ICs. 3M has been promoting its wafer stacking solution (WSS) for quite some time, and in Bowman’s opinion it is 3M’s most misunderstood product. WSS is a combined material and equipment solution for temporary bond and debond, and contrary to popular belief, Bowman says this product is well established, with over 50 tools installed, and is qualified for TSV mass production on 300mm wafers.The next-generation WSS is a film-carrier based solution, eliminating the need for a support carrier wafer. A novel material allows for mechanical debond performed at room temperature, eliminating the laser step to ash the adhesive in the current process.
Gary Fleeman, of Advantest, weighed in on test solutions for 3D ICs, noting the added test challenges of TSVs and micropillars such as higher pin counts, finer pitch, lower contact pressure, no drivability, new insertion points in the flow due to die stacks, etc. All this requires a non-conventional test methodology that needs to be flexible, and where yield and contact issues become important. At Sonoscan, Jack Richtsmeier reports that failure analysis and reliability is giving way to production environments. SOI and bonded wafer inspection are target applications for its acoustic microscopy technology.
From Disco Corporation, Tetsukazu Sugiya talked about the company’s work to improve total thickness variation (TTV) during the thinning process. The development of glue and tape planarization methods has added 1-2µm accuracy to the process. He also talked about Taiko wafers, which maintains the original wafer thickness at the edge of the wafer, while thinning the middle, allowing the wafer to hold its rigidity. These wafers offer the potential for carrier-free easy temporary bonding and debonding.
Several companies presented on advancements for TSV fabrication. Bob Forman, of Dow Chemical, talked about addressing the matter of voids that happen during via filling by developing a Cu fill chemistry that fills from the bottom up with “absolutely no voids”. Camelia Rusu, of Lam Research, set the record straight on the importance of smooth sidewalls for TSVs. While some companies prioritize etch rate, and insist that scalloped edges are an unavoidable outcome of DRIE that can be dealt with by conformal coating; Rusu says not only do smooth sidewalls matter for TSV reliability; but they CAN be achieved using the Bosch processes. While there’s some sacrifice to the etch rate, she says the benefit to TSV reliability is worth it. She says this is particularly critical as TSV aspect ratios are expected to increase. She also noted that this is a customer-driven issue. Some customers are satisfied with scallops and would rather not sacrifice the etch rate, while others require the smooth side walls. Lam is happy to customize its processes to suit the needs of its customers.
NEXX Systems, which recently became a wholly-owned subsidiary of Tokyo Electron, was represented by Arthur Keigler, who talked about the additional insight the company can bring to the JDP with IBM because of its expanded tool portfolio, with rounded out offerings for 3D processes. The company is about to launch its next-generation electro-deposition tool, the Stratus Thunder – with productivity enhancements for 3D such as improved chemistry control, and high volume manufacturing needs such as such as increased throughput and cleaning modules. Another tool in the pipeline is the Tornado, a wet-resist strip tool that promises to provide low contact resistance for the seed layer.
It was my job to update all those who gathered for this event on what the world outside of IBM and its JDPs are saying about 3D. As we all know, activities are ramping up all over the place. I decided to showcase what I’ve been gleaning from the headlines and from various events. My comments were characteristically enthusiastic and optimistic; and therefore met with more than a spoonful of skepticism. There was a collective groan of disbelief when I quoted Yole’s prediction that the shipment volume of 3D-IC wafers will reach 10 million units in 2012 (Semicon Singapore, 2012). When I said that Riko Radjocic of Qualcomm declared there were no remaining intrinsic technology hurdles holding up 3D, someone shouted “that’s because Qualcomm is fabless.” And when I reported that TSVs have been found to solve thermal issues and because the stacked chips’ proximity drops power, improves latency and moves heat, and that with regard to FPGAs using 2.5D processes, the reliability is actually quite good, I pretty much lost all credibility. Thermal issues, I was told, depend on the application. While its possible all this is true in low power applications, server applications are another story, and thermal issues are still a concern. I stand corrected. But it made for really great discussions at the networking reception. ~ F.v.T