by Griff Resor, Resor Associates
March 13, 2007 – Kicking off the 20th annual Conference on Optical Microlithography last week during SPIE, TSMC’s lithography guru Burn Lin wove a tale that began as a history lesson, but shifted to a modern message tackling today’s top lithography challenges. The common thread throughout: “The devil is in the mask.”
Lin, senior director of TSMC’s micropatterning division, reminded the audience that he presented the concept of DUV in 1975 at the 3-Beams conference. He then turned back the clock a decade and described emulsion masks and contact printing. This kind of lithography served the early IC business well. He compared this stage of IC technology to the Wright Brothers’ first airplane.
As the IC industry progressed, mask defects and cost threatened progress. “The devil was in the mask,” Lin said. 1X projection lithography provided the solution, which he compared this to a robust propeller airplane.
Soon the IC industry ran into mask tolerance issues. “The devil was in the mask,” Lin observed again. So the industry shifted to 10X, 5X and eventually 4X reduction steppers. This was compared to the stage of advanced propeller planes. The shift to reduction optics created the “mask makers holiday,” he said, and making masks got easier for a while.
Reduction tools then shifted from stepping to scanning to work around limits in the optics. This change was equivalent to the first jet planes, Lin contended. Step and scan extended optics to 248, 193, and 193 immersion.
We are now at the jumbo jet stage, Lin believes. Once again industry is up against the limit of optical printing. Mask Error Factor is often 4X. And once again, he repeated, “The devil is in the mask.”
Today TSMC is using 193 immersion lithography on two layers of its 65nm node process and on 12 to 18 layers of its 45 nm node process. Defect levels are at single digits per cm2, equivalent to dry lithography. Clearly 193 immersion is in production at the 45nm node.
Lin presented his wish list for improved optical tools. Today’s top coat and 193i resists compromise resist performance to make immersion imaging work. He pushed the audience to consider new tradeoffs in design. For next generation fluids he expects that a full liquid bath will be needed. He’d like to eliminate top coats and to see a “hyper contrast” resist so that two exposures can be printed into the resist with double exposure in a single pass. He put up his machine design concept for a two-mask-one-lens machine. Lin is not happy with double exposure, but recognizes that it will be needed to push 193 immersion to the 32nm node. He compares double exposure to the supersonic airliner. It worked but ultimately failed because its technology was too expensive.
Lin went on to examine EUV. He doesn’t see a workable solution. Mask flatness is 10X worse than needed. Resist, source power, reliability, and cost are all show stoppers. He compared EUV to the Space Shuttle. Only two were built – like the two Alpha tools for EUV.
Lin turned finally to his wish for a maskless technology. “The devil is in the mask, he said, “so lets get rid of the mask!” TSMC’s foundry model depends on low mask costs. He reviewed efforts to build maskless wafer writing tools at Multi-Beam Systems, KLA-Tencor, and Mapper. He is currently working with the Mapper project. TSMC wants a multi-beam direct write tool in 2009 to write first level contact holes at the 22nm node, or in 2011, when EUV says they will be ready. He compared multi-beam direct write to the helicopter, a very flexible tool.
In the end the audience got a very thorough review of modern litho alternatives to support TSMC’s business model. Lin is clearly concerned about the cost of extending optics to the 32nm node. He’s even more concerned about the favored choice for the 22nm node, namely EUV. He made a pitch for E-beam direct write using a multi-beam approach.
Often Lin has recognized the best path forward before others. His review of litho history made this point. He may be right again.