AUSTIN, Texas With 157-nanometer optical lithography having "missed the window" for the upcoming 65-nm process node, International Sematech has all but slammed the door on the decade-long 157-nm development effort.
The consortium said last week that it is opening an Immersion Technology Center (iTC) here to focus on 193-nm immersion lithography, with an eye toward possibly extending the technology to the 32-nm process node. Sematech has also contracted with Exitech Ltd. for spring delivery of a 193-nm immersion scanner.
The announcements follow Sematech's closure late last month of its 157-nm Resist Test Center. Among the related programs that have been halted, said Sematech lithography director Giang Dao, are an effort to develop a pellicle that could withstand 157-nm radiation and one to promote the supply of higher-quality calcium fluoride lens materials.
The Sematech decision is "part of a resource allocation issue that the industry has been [facing] for a long time," said Risto Puhakka, vice president at VLSI Research Inc. (San Jose, Calif.). The only unfinished business for 157 nm, Puhakka said, is "an official declaration that 157 is dead and I'm not willing to say that just yet. But the industry now has one clear path: 193 immersion."
The industry had set its sights on 157-nm lithography for the 65-nm node, which is expected to ramp late next year, but for various technical reasons 157 "missed the window," Dao said. The presumptive choice for 65 nm in its early phase is now 193-nm dry lithography, with 193 immersion expected to be ready later in the 65-nm node. Extreme-ultraviolet technology is a candidate for the 32-nm node, but some industry watchers now say EUV may not be ready in time at an acceptable cost. That means 193 immersion could extend past the 45-nm process generation and into 32 nm.
The iTC program will be supported in part by funds from the state of Texas and will operate as part of the Advanced Materials Research Center. The iTC is expected to spend about $15 million over the next two to four years, targeting the use of 193 immersion tools at the 45-nm and possibly the 32-nm node.
Research into how resists degrade when exposed to water will be contracted out to several universities. A Sematech-sponsored program at MIT Lincoln Labs will pursue the coatings needed to protect the bottom lens element from the corrosive effects of submersion. Studies will also focus on how to counter light polarization effects and how bubbles in the water might affect exposures.
Texas Instruments Inc. lithography research manager Shane Palmer will move here next week to head a study group that will consider the challenges of using 193-nm immersion technology for the 32-nm process node, which is expected to move into manufacturing at the end of the decade.
Dao said that if fluids with a higher refractive index than water can be used together with high-numerical-aperture catadioptric lenses (i.e., lenses combining both refractive optic elements and reflective elements) and various reticle enhancement techniques, 193-nm scanners could indeed be extended to the 32-nm node. "The study led by Palmer will determine if it is feasible or not," Dao said.
The scanner to be supplied by Exitech (Cambridge, England) will have a small field of exposure, of 0.4 mm2. It will feature a catadioptric "hyper-NA" lens, developed by Corning Tropel (Fairport, N.Y.), with a numerical aperture of 1.3 far higher than the 0.85-NA lenses now used in 193-nm scanners.
The 157 effort, based on a fluorine laser, ran into multiple technical problems, starting with the realization several years ago that the calcium fluoride crystals needed to build the lenses had intrinsic birefringence properties. Once immersion 193 became the industry's choice for the latter half of the 65-nm node and beyond, infrastructure investments for calcium-fluoride crystal growth dried up.
Also, the organic (plastic) pellicles now used to protect the masks couldn't withstand the 157-nm radiation. And finding a resist that would work well proved particularly nettlesome, despite several years of effort.
Early this year, it became apparent that the major scanner vendors, including ASM Lithography, Canon and Nikon, had stopped development of the 157-nm lithography platforms in favor of immersion 193. ASML and Canon shipped several 157 systems apiece to research centers in Europe and Japan, but no further development of those platforms is planned.
Still as VLSI Research's Puhakka suggested, no one is quite ready to say 157 has breathed its last. Japanese research consortium Semiconductor Leading Edge Technologies Inc. continues to study the approach, as does MIT's Lincoln Laboratories. A Sematech spokesman said the 157 research work completed thus far at the consortium will be carefully archived in case some future need arises to restart the 157-nm program.
One line of thought, expressed at last January's Sematech-organized Litho Forum, is that if immersion fluids could be found that would work well with the 157-nm wavelength light, then a "wet" 157-nm system, with a 115-nm wavelength, could be used at the 32-nm node.
Coming up with a wet 157 solution, however, would first depend on development of a workable dry 157 platform and that would take billions more in development funding. The biggest obstacle is obtaining sufficient supplies of the extremely pure calcium fluoride needed for the hyper-NA 157 lenses. No company will make the investments in calcium fluoride furnaces with such slim prospects of a payback, one source said last week.
"An immersion 157 system would be good for one node, at best," Sematech's Dao said.
"With 157 out of the picture, the question now becomes, How long will 193 immersion last?" said Phil Ware, senior fellow at Canon Inc.'s lithography division. With a film of water between the lens and the wafer, the refractive index of water, at 1.43, narrows the 193-nm light to 134 nm. If fluids can be perfected with refractive indices of 1.5 to 1.65, Ware said, "we might be able to reach the 32-nm node with 193-nm scanners. But it would have to be with a very aggressive k-factor, or even double exposures."
The k-factor is an amalgamation of the resist quality, optical enhancement techniques, and other elements in the lithography platform. Double exposures, as the name implies, would use two exposures to create lines and spaces, a procedure that would cut throughput sharply, Ware noted.
If refractive, full-field lenses with numerical apertures in the 1.0 to 1.2 range could be built for 193 immersion scanners, they would almost surely be huge, with diameters nearly double today's 193-nm lenses, Ware said.
"Catadioptric designs buy you one generation; then you are back on the path of huge lenses," he said.
Klaus-Dieter Rinnen, director of equipment research at Gartner Dataquest Inc., said the Sematech program is acutely needed because the scanner industry faces huge R&D costs. Immersion 193, EUV and maskless lithography all require enormous sums before they can be used on the factory floor. The makers of exposure tools had combined revenues of $5.4 billion in 2003 on shipments of 694 scanners, Dataquest estimates.
Now that 157 appears to have gotten "squeezed out," Rinnen said, the "expectation is that engineers will have to figure out a way to use 193-nm wavelengths to create 32-nm chips." It should be noted, he added, that the 32-nm node described by the International Technology Roadmap for Semiconductors, when put into practice, is likely to use a half-pitch of 45 nm, with gate lengths much smaller than the 32-nm width.
"It sounds rather ambitious," Rinnen said, to use wet 193 scanners for 32-nm chips. But 248-nm scanners were used for 130-nm devices, he noted.
Indeed, the immersion trend could conceivably extend to 248-nm scanners, though scanner companies are wary of such a development. By using immersion with the relatively low-cost 248-nm tools and resists, companies could save money compared with 193-nm dry tools, the Dataquest analyst said. A gallon of 193-nm resist costs about $5,000 to $6,000 now, compared with $1,000 to $1,500 for a gallon of 248 resist. Mask costs also are much lower.
"If I was a manufacturer of scanners, the last thing I would want to do is sabotage my 193-nm business with a 248-nm immersion offering. But there are cost-of-manufacturing arguments that indicate such a development could happen," Rinnen said.