DSA, virtually unheard of just a few years, has quickly emerged, becoming the hot technology at this year's SPIE. Last year, about 10 papers on DSA were presented; this year, there are more than twice that many on the SPIE agenda. The technology promises high resolution, good line width roughness and high pattern fidelity. It works in conjunction with existing optical lithography technology.
"The fact that we can implement DSA with standard lithography techniques and reach high resolution is quite compelling," said Raluca Tiron, R&D advanced lithography process engineer at CEA-Leti.
But manageable defect density with DSA has not been proven, according to Ben Rathsack, a strategic marketing and new business development manager at Tokyo Electron Inc. "The honest answer is: we don't know," Rathsack said in the panel discussion Tuesday.
EUV, which was initially targeted for production in 2005, had been more recently expected to be ready by the 22-nm node, which Intel Corp. is implementing now. Intel now wants to put EUV into production at the 10-nm node, starting in the second half of 2015. Samsung Electronics Co. Ltd. plans to insert EUV into volume manufacturing as soon as 2013.
But EUV has been delayed multiple times, most prominently by the inability to develop a light source with the power and reliability necessary to support throughput that would make EUV economically viable. Cymer Inc., the supplier of EUV light sources to No. 1 lithography vendor ASML Holding NV, and Ushio Inc., another potential supplier, are scheduled to report progress on their light source development at SPIE Thursday.
"I am incredibly impressed by the amount of progress EUV has made," said Chris Bevis, chief technologist of KLA-Tencor Corp. and principal investigator of the firm's REBL e-beam direct write system. "I am not counting on EUV to fail for e-beam direct write to succeed. One size might not fit all. EUV is way ahead of e-beam direct write right now, make no doubt about it."
The delays in EUV and the lingering sense that 193-nm immersion ArF lithography cannot be extended further without expensive multiple patterning—four or even five photomasks required where one used to be sufficient—have left many in the semiconductor industry feeling anxious and pessimistic about the industry's ability to continue reducing the cost per transistor of semiconductors enough to make continued scaling worthwhile.
"I don't sense a panic, but I do sense a concern that at some point the economics won't be as good as they once were," Kalk said. "If you reduce transistor cost by 20 percent per year, everyone's happy. If it's 10 percent, it's a question."