Austin, Texas Those congregating in San Jose, Calif., today for the 2005 SPIE Microlithography conference will find an altered landscape in semiconductor patterning, with imprint lithography finding fertile terrain and maskless lithography facing a funding drought.
Imprint lithography's progress is evident in a Molecular Imprints Inc. system to be unveiled today at the Society of Photo-Optical Instrumentation Engineers' lithography meeting. The Imprio 250 system, priced at about $5 million to $6 million, is capable of sub-50-nm lines and spaces, with 10-nm alignment, albeit at throughput of only a few wafers per hour, according to MII, based here. With integrated wafer handling and much-improved alignment, imprint systems, which use ultraviolet light to harden polymers within a template, appear poised for acceptance.
On the downside, the Defense Advanced Research Projects Agency, which has funded much of the most important lithography research since 1991, confirmed to EE Times that its Advanced Lithography Program will end in September, when the federal 2005 fiscal year comes to a close. That will shut the public-funding spigot on 10 U.S.-based programs that are struggling to advance maskless lithography (ML2)
in both its multiple-electron-beam and mirror-array forms.
Darpa's change of course may give government-backed efforts in Japan and Europe an advantage, sources said.
But the news wasn't unexpected. "While it is true that the Advanced Lithography Program ends in fiscal year 2005, that has been Darpa's intent for a number of years," a spokesman wrote in a statement to EE Times. "Since February 2001, Darpa has indicated its intent to end its investment with the FY 2005 funding."
The statement adds that "the agency remains attentive to possible lithography approaches that could serve the unique needs for low-volume, high-performance circuits for the Department of Defense."
Other parts of the DOD budget continue to support lithography research. The Naval Air Warfare Center, for example, is funding a $34 million program to develop a soft X-ray light source and mask technology for nanolithography and optical instruments. The program backs efforts at JMar Technologies Inc. (San Diego) and elsewhere.
Darpa has invested about $1 billion in lithography research, most of it at the bleeding edge, over a 14-year period. Led by Darpa lithography manager David Patterson, who retired last year, the agency's programs helped put imprint lithography on the map and funded early demonstrations of immersion 193-nanometer litho for the 45-nm process node.
With some $15 million in ML2 funding this fiscal year, Darpa is supporting maskless programs at several universities and national labs, as well as putting $5.6 million into a program that counts ASML, Lucent, and Cypress Semiconductor subsidiary Silicon Light Machines as participants. Lucent and Silicon Light Machines are working on separate litho-friendly forms of digitally controlled micromirrors, the optical modulators that have achieved success in the flat-panel TV market.
Total Darpa spending is at the $3 billion level for this year, and Congress is considering a similar budget for fiscal 2006. The Darpa spokesman said the agency is looking at increasing funding for "3-D microsystems that bring together sensors, actuators and power management," as well as wide-bandgap ICs for both RF and high-power applications. The agency will discuss its research directions at a meeting Aug. 9-11 in Anaheim, Calif.
While Darpa has supported advanced research, the National Institute of Standards and Technology has funded midstage development efforts at Molecular Imprints and elsewhere. MII expects its Darpa funding to end this year; but the company recently received word that the second year of its NIST contract, which includes support for partners KLA-Tencor, Photronics, Motorola Labs and others, will continue to be funded, with NIST allocating $17.5 million over three years.
"Darpa has supported some of the R [research]; NIST has funded some of the D [development]. Then it is up to Molecular Imprints to handle commercialization," said Michael Falcon, an MII business development manager.
The goal of the NIST program led by MII is to develop alignment capabilities that would extend imprint lithography to the semiconductor industry, where multiple mask layers must be precisely aligned. Using an approach acquired from the Massachusetts Institute of Technology, MII claims it has reached alignment precision of 10 nm good enough to penetrate the semiconductor-patterning market at the 45-nm node.
What bothers the chip industry, an LSI Logic manager said, is that maskless lithography systems have very low throughput. In an era of cost pressures, companies expect 100 wafers per hour or more from scanners that can cost $20 million to $30 million. Maskless systems have the attraction of taking expensive mask sets out of the equation, but at throughputs that in the initial stages would be in the range of one to five wafers per hour. With FPGAs, structured ASICs and direct-write electron-beam systems as alternatives, maskless approaches face daunting challenges getting a foothold in the market.
Speaking at a conference here last year, outgoing ASML CEO Stuart MacIntosh said the company's joint venture with Micronic Laser Systems AB to create a mirror-based maskless system appealed to the technologists at several foundries. "But the people who make the decisions at those companies were never willing to write a check," MacIntosh said.
ASML will likely wind down its maskless research programs this year after Darpa support ends, a spokesman said.
Canon Inc., which has worked with Hitachi Ltd. on a closely guarded multiple-electron-beam ML2 system, also has put maskless on the back burner. "All of the available development resources at Canon have been put into immersion," a Canon spokesman said.
The difficulty of getting a foothold in the market with very low-throughput systems is one reason Darpa's decision to end its funding is critical. Daniel Herr, who manages the lithography-related programs at Semiconductor Research Corp. (Research Triangle Park, N.C.), has created time lines that show that many lithography technologies take 30 years to develop, measured from the first paper presented at an often-obscure academic conference to the year when the technology gets established in commercial form.
Walt Trybula, a Sematech senior Fellow who has organized several industrywide maskless meetings, said the industry needs a cost-effective way to prototype. Mask sets for 65 nm are expected to cost $3 million, or much more if respins are necessary for critical layers. But maskless systems are unlikely to be ready even for the 45-nm node. Trybula acknowledged that while immersion 193-nm lithography seemed to overcome its technical challenges quickly and fairly painlessly, maskless lithography faces much higher technical hurdles.
"There is not the same groundswell of support for maskless that there was for immersion lithography," he said.
Srini Raghvendra, director of business development at the design-for-manufacturing unit of Synopsys Inc., said his customers plan to use "dry" 193-nm scanners for the 65-nm node and switch to immersion systems for the 45-nm generation.
"A lot of the load going forward will be borne" by optical proximity correction techniques and phase shift masks, he said. "We are developing immersion-ready lithography models that accurately reflect the physics of the latest chemistry in the resists." The approach will be demonstrated on the SPIE show floor.