LIVERMORE, Calif. Once deemed too exotic to ever become practical, extreme-ultraviolet lithography has proved itself in the laboratory but is now battling international technopolitics to show that it's commercially workable as well.
Extreme-ultraviolet, or EUV, technology itself is a "done deal," said one member of a body driving the next-generation chip process. But another conceded that "six zillion engineering challenges" remain before it becomes a fab-line reality. Complicating the efforts of the EUV LLC consortium are a stalled business deal and continued reluctance by Japanese lithographers to join the U.S.-dominated program.
Industry and government leaders met here last week to discuss what Sunlin Chou, general manager of Intel Corp.'s technology and manufacturing group, called "the second stage of the EUV program." In that stage equipment suppliers will take a leading role in developing EUV scanners for the 50-nm technology node. Intel, Motorola, AMD and Micron Technology were early members of the five-year-old EUV LLC. They were joined by Munich-based Infineon Technologies last year and by IBM in February.
Chou, chairman of the EUV LLC, said that "when the story of the EUV program is retold, it will be seen as one of the extraordinary feats in the history of the semiconductor industry and the national labs."
Concerns are growing that the semiconductor industry's traditional scaling the engine of progress known as Moore's Law under which processors double their performance every 18 months may slow in the middle years of this decade. By 2005, a solution is required for commercial production at the 70-nm technology node. Also by then, the 50-nm node will be in development, requiring prototype scanners.
Hopes have centered on 157-nm optical lithography, which uses a fluorine laser. But suppliers of the calcium fluoride crystals needed to make the lens elements have found it difficult to improve yields. One solution: simply grow more of the crystals and make do with the low yields. But that approach causes lens costs to skyrocket, another hurdle on the road to commercial viability.
As Joe Mogab, Motorola's representative to the EUV LLC, said, "The technology itself is a done deal. They have proven that it works. The remaining question is: When will it be ready?"
Sensing opportunity, the EUV program is racing to get its own prototype machines ready by 2004 or earlier, with a targeted throughput of 80 300-mm wafers per hour for the commercial machines expected in 2006 from Dutch lithographer ASML and others.
But complicating that timetable is the politics of technology, including concerns about a foreign company, ASML, acquiring a U.S.-based one, Silicon Valley Group Lithography (San Jose, Calif.), that has ties to national-security operations. And with the other two major lithography vendors Nikon and Canon based in Japan, the EUV implementation effort needs to speed the entry of Japanese companies into the commercialization effort.
In an interview here this week, Intel chief executive Craig Barrett told EE Times that federal agencies gave their blessing to the SVGL acquisition earlier this year. SVGL owns the 98-person Tinsley Laboratories lens-making operation, which supplies optics to the military's spy satellites as well as to the EUV LLC. The Department of Defense joined in that approval, based on agreed-upon export provisions of the Tinsley technology, Barrett said.
But objections from Capitol Hill prompted a 45-day review of the proposed acquisition. Barrett said he now feels "there is growing Congressional support" for allowing the sale to proceed. "We would have very, very serious problems if this stretches out much further," Barrett said, including a potential loss of staff at SVGL during a long limbo period.
As an equipment supplier, ASML is a "licensee" of the EUV technology, with a large engineering staff working here and in Holland to develop a commercial EUV scanner during the next four or five years. But Nikon and Canon have remained outside, working on their own in Japan and as part of an EUV national project that is receiving substantial financial support from Tokyo.
"European and U.S. companies are working here [at the EUV LLC] and I would love to get the Japanese lithography companies committed," Barrett said. In industrial research and development, he said, "Japanese companies like to do their own thing."
With "six zillion engineering challenges" ahead of the EUV effort as it moves to commercialization, Barrett said, he is encouraged that Japanese companies are developing EUV technology, both by themselves and as part of Japan's EUV program. Getting the industry centered on EUV, rather than X-ray or other next-generation lithography candidates, is in the industry's overall interest, Barrett said.
"There is room for both [the EUV LLC and the Japan EUV programs], but I would like to see cooperation when it comes to the implementation of the technology," Barrett said.
Jim Glaze, executive director of the Virtual National Laboratory of scientists from the Lawrence Berkeley, Lawrence Livermore and Sandia national laboratories, suggested that Japan's central government is about to sharply increase its funding of EUV research.
Though the EUV LLC now operates solely with private funding from the member companies, Washington supported the project during the early years. Much of the basic research into the essential technology the silicon and molybdenum coatings of the curved mirrors so that they precisely reflect the 13.5-nm radiation was developed at the national labs.
Talks with Japan
Chuck Gwyn, program director of the EUV LLC, said that Canon in particular had an early interest in EUV and that Nikon also has a team working the technology. "We would like to establish a cooperative effort with them, and we have had early talks," he said. "They are taking part in the Japanese project, and we exchange data with that project. But it is difficult to say if they are interested in joining the EUV LLC as licensees."
At last week's open house, the EUV engineering test stand (ETS) was on display, equipped with the first set of projection optics developed last year. Chou said the ETS printed its first lines, at 100-nm design rules, Jan. 25. Last week, the second set of projection optics was shipped to the Lawrence Berkeley lab for testing and refinement.
Gwyn said the EUV program still has "a lot of work ahead of it," and listed several engineering challenges. Chief among them was the need for a higher-powered source of the EUV radiation.
"The output of the laser plasma source of the ETS is about 10 watts of EUV power, and we need to increase that to more than 100 W in order to achieve our throughput goals. Some people in our program believe we need as much as 150 W," he said.
Creating a commercial mask infrastructure is equally important, Gwyn said. The EUV mask substrate is also a projection optic with patterns created by depositing chrome absorbers on the polished reflective substrate.
"We need mask blanks ready-to-use substrates with multilevel coatings of molybdenum and silicon," Gwyn said. "We need the absorbers and resists to be perfected so that companies like DuPont Photomasks can create commercial masks easily. We have to drive down the rate of defects on the substrates, from about 0.01 defects/cm2 now, by a factor of five."
The scanner prices could range from $15 million to $25 million each, Gwyn said.
At the 70-nm node, said John Carruthers, a retired Intel technology manager who helped start the EUV LLC, it may be possible to use 193-nm scanners for most of the mask layers and EUV lithography for the five or so critical layers that define the gate.
Gwyn said, "It is still our goal to have the EUV scanners ready for major introduction at the 50-nm node. The 157-nm scanners have issues with the calcium fluoride, and they still don't have a resist, and we'll have to see what happens there. For EUV, there may be some opportunities at the 70-nm node, but the big impact will be at the 50-nm node."