SANTA CLARA, Calif. The race to develop a viable next-generation lithography accelerated at the SPIE Microlithography conference here this week, as backers of the extreme-ultraviolet and electron-beam-projection lithography techniques jockeyed for inside position.
Intel-led consortium EUV LLC of Livermore, Calif. announced that its engineering test stand had successfully been used to print lines at 100-nm spacings, proving the concept of extreme-ultraviolet lithography (EUV). EUV gained further momentum when Motorola Inc. said it was dropping out of the group of companies backing e-beam projection lithography (EPL) to concentrate its resources on extreme UV.
The industry will need early EUV systems by 2004 for development of 50-nm devices, said Peter Silverman, director of lithography capital equipment development at Intel. He urged the industry to accelerate funding for development of the EUV infrastructure so that beta tools can be made ready in time.
The IBM-Nikon team that is developing an e-beam projection lithography test stand also vied for attention at the conference, organized by the Society of Photometric and Instrumentation Engineers (SPIE). IBM Fellow Hans Pfeiffer, manager of electron-beam technology at that company, said an IBM-developed Prevail e-beam optical system is now integrated into Nikon's scanner column hardware in Japan.
The partners expect to have an alpha column fully integrated this year and to show a "proof of lithography" set of images by the fourth quarter.
The IBM-Nikon team plans to introduce its first production tool in mid 2004, said Tekeshi Yamaguchi, a general manager at Nikon Corp. of Tokyo. The goal is to have the EPL system ready for the 70-nm technology node in 2004, using EPL for the critical layers alongside the 157-nm F2 lithography tools that are also in development.
Nonetheless, both NGL approaches still face enormous challenges, both in the systems themselves and the masks required, SPIE speakers said.
The throughput of the EPL approach is limited, Yamaguchi said. For "noncomplementary" images that can be printed in one scan, the EPL technology can achieve throughput of 20 wafers per hour or higher. If two passes are needed, for complementary, "doughnut hole" images, then throughput is limited to about 10 wafers per hour, he said.
"There is no perfect solution," Yamaguchi said, noting that Nikon is pursuing 157-nm scanners, the EPL system and even EUV technology as part of a national project in Japan.
"Nikon believes that EPL can be extended to the 35-nm generation, with large process margins and acceptable cost of ownership. Depending on the device type, EPL is a good solution," he concluded.
Intel has been on a 22-month cycle in bringing a more advanced process technology to its fabs ever since the 0.5-micron process generation, said Silverman. For that to continue, he said, EUV technology will be needed for process development about 18 months earlier than the planned commercial introduction. The 50-nm node is expected to debut in 2007 or even earlier, if the production tools can be hurried into place.
"The company who crashes into the wall first will be the winner," Silverman said, drawing laughter from the SPIE attendees.
Because the initial 157-nm scanners will be introduced with high-numerical-aperture lenses, it will be difficult to improve on the 157-nm technology after its introduction, he said.
Though many believe the industry will figure out ways to extend the 157-nm tools to the 50-nm process generation, Silverman expressed doubts.
"By 2004, we believe both the 193- and 157-nm technologies will be running out of steam. We will need something better. We need EUV, to ride it down the curve. When do we need it? By 2004 but it will take a lot of money to have it ready in four years.
"On the standard time scale, the EUV technology should be ready by 2005," Silverman said. "But we hope the EUV LLC will work with the suppliers to reduce the development time by six months or so."
The engineering test stand (ETS) that printed its first images 10 days ago was based on a 1,500-watt laser plasma source that generates an EUV beam with 10 W of power. More than 50 W of power will be needed by the time the EUV system reaches production, Silverman said.
Eyes on the optics
Boosting the power of the source and improving the reflectivity of the optics are essential to achieving the throughput of 70 to 100 300-mm wafers per hour that will be required for mass-production of DRAMs, microprocessors and other high-volume products. More than 120 scientists and engineers from the Virtual National Laboratory encompassing the Lawrence Livermore, Sandia and Lawrence Berkeley national labs are working alongside about 40 staffers from the member companies of the EUV LLC.
Consortium director Chuck Gwyn said that by next month, the engineering test stand will add a scanning capability, moving from the small, "static" (i.e., non-scanning) images shown at this week's SPIE conference to 24 by 32-mm field sizes on a scanning platform. Gwyn predicted that the improved ETS will be able to print 70-nm lines and spaces at full field sizes by the end of this year.
Also by year's end, the consortium expects to fit the ETS with new projection optics: a set of mirrors that reflect the 13.5-nm-wavelength EUV radiation through the reticle and down to the wafer level. That improved set of mirrors is being evaluated now at the synchrotron at Lawrence Berkeley labs, where precise interferometry technologies have been developed for the EUV development project.
At this week's conference, scientists from the national labs reported that the current reflectivity rate is several points below the 70 percent goal and described plans to improve the uniformity of the coating layers.
The mirrors are rotated in a magnetron and coated with 40 layers of molybdenum and 40 layers of silica, creating a surface that achieves a nearly 70 percent reflectivity ratio. That technology is being transferred to the Tinsley division of the Silicon Valley Group, acquired last month by the Netherlands' ASM Lithography. In Europe, ASML is also working with its long-time optics supplier, Karl Zeiss.
Since much of the basic intellectual property in the EUV LLC was developed in the national labs and thus was funded by U.S. taxpayers, the ASML takeover of SVG creates a difficult situation for backers of the project. Tinsley's EUV optics are believed to be superior at this point, one source said. But how ASML divides the EUV commercialization between its U.S. and German optical suppliers will be an important issue as Washington debates the ASML-SVG merger, the source said.
Scouting funds, friends
Meanwhile, both the EUV consortium and the EPL effort are scrambling for financial support, as well as for customers that will pledge to buy the early systems.
The EUV LLC plans to host an open house in Livermore in mid April. At a similar event last year, Gwyn announced that Infineon Technologies had entered the consortium, joining Intel, Advanced Micro Devices Inc. and Motorola as the major backers. There is speculation that this year's event may include an announcement of IBM's participation.
At the International Electron Devices Meeting (IEDM) last December, one IBM source said that IBM had decided to join the EUV but was bargaining for better financial terms. Gwyn and IBM officials declined to comment at SPIE on whether IBM would join the EUV LLC anytime soon.