Contenders vie for next-generation lithography crown
By David Lammers
EE Times 12/03/99
COLORADO SPRINGS, Colo. ? The search for an approach to post-optical lithography has been thrown wide open. At least five serious candidates will duke it out at International Sematech's Next Generation Lithography Workshop, which convenes Monday (Dec. 6), and the field may not be narrowed for some time.
Advances in masks and lenses for 157-nanometer optical lithography mean that a next-generation lithography (NGL) solution may not be needed in commercial production for seven to 10 years. That fact has thrown the door open to a broad field of NGL contenders.
Presentations are expected here next week on: the extreme-ultraviolet solution advanced by Intel Corp.; two flavors of electron-beam projection lithography; the largely European-backed ion-beam projection lithography; X-ray; and a new approach to electron-beam direct write that uses multiple columns.
Projection e-beam's two champions are the Prevail system being developed by IBM Corp. and Nikon Corp., and the Scalpel system backed by Lucent Technologies' Bell Labs and a consortium of commercial sponsors.
"The main focus now is on the 50-nm node," said Gerhard Gross, director of lithography at Sematech. "Last time [in December 1998], the NGL Workshop focus was on 100 nm. The timing is more relaxed for an NGL solution right now, but the requirements are much tighter for 50 nanometers. We have to review all of the NGL technologies again to see if they are OK for this node."
Up to 200 lithography experts are expected to attend this year's workshop, a two-day, invitation-only affair. While the preceding two workshops featured a thumbs-down kind of vote intended to narrow the number of NGL development efforts, the vote at this year's event will serve only to gauge confidence levels in the different solutions.
And the floor is being opened to all of the credible candidates, including a proposal from ETEC Corp. for electron-beam microcolumns that would write directly to a wafer.
"Two years ago, I think we excluded the electron-beam direct-write solution, but now we have the microcolumn proposal, which is focusing on the 50-nm node," said Gross, an assignee to Sematech from Siemens AG. "Things are changing, the 50-nm node is in front of us and we have to include and consider e-beam direct-write again. It is fair for them to talk to us again."
During the past two years, progress in the 157-nm wavelength lithography solution, based on a fluorine (F2) laser, has changed the research agenda considerably. Confidence that a 157-nm solution will be available to print lines down to 70 nm stems from the availability of optical proximity correction (OPC) techniques and phase-shift masks.
Evaluations scheduled
By the first quarter of 2000, Sematech staff will be working with a 157-nm microstepper with a limited field size, supplied by Exitech Ltd. (Cambridge, England). Sematech researchers will begin resist evaluations by the second quarter, Gross said.
"There are lots of contamination problems with 157-nm [lithography]. We have to purge the whole light path with nitrogen," he said. "The lens is made of pure calcium fluoride, but the lens is coated and done. The lens part of the solution has turned out to be relatively straightforward, and it is similar to the lens used in the 193-nm microstepper."
After the workshop an NGL task force will convene, including representatives from the Japanese equipment consortium Selete, the 14 member companies of International Sematech, stepper manufacturers ASML, Canon, Nikon and Silicon Valley Group Lithography, and two Japanese IC manufacturers, Hitachi and Toshiba.
Following that meeting will come a meeting for members of International Sematech only, to recommend the direction of lithography research at Sematech. The Sematech Technology Advisory Group will vote on those recommendations at the end of December.
At the 1998 workshop, participants heard from proponents of the 157-nm solution, including the suppliers of excimer lasers who were "pushing hard in this direction," Gross said.
"There were a lot of questions about the masks, because we only have a few materials that are transparent at this wavelength. At that time, we thought that it has to be calcium fluoride, which would be more or less a showstopper, because that material does not have very good thermal properties," he said. "Then some companies came up with a modified fused silica, and [found] ways to make it transparent. There was [also] progress in the lens materials, and those programs were very important to the success of the 157-nm solution."
By the time of a critical review of the NGL effort six months ago, another major shift in thinking had occurred. With the progress in OPC and phase-shift mask techniques, many experts held that 193-nm scanners with high-numerical-aperture lenses could be used in the 100-nm technology node, due to reach early production in 2004 or 2005.
"Within half a year, people started to believe that 157-nm would be available for the 70-nm node, which is why we now define the 157-nm solution as being available for 70-nm production," Gross said.
Not everyone is ready to cede that territory to optical scanners, however.
"The Scalpel program is aimed at getting tools in the hands of our members for at least part of the 100-nm generation," said Lloyd Harriott, director of advanced lithography research at Bell Labs. "It varies company to company, but we want to be in volume sales for 70-nm-and-below production. That means timing our beta tools for use in late 2002 and 2003, and production tools in 2004."
Asked about competition with 157-nm scanners, Harriott said that Scalpel will overlap with 157-nm tools in the 100-nm generation.
The Scalpel development effort, centered at Bell Labs, will involve staff from both ASML and Applied Materials Corp., the two equipment vendors that have formed eLith, a corporation intended to bring Scalpel scanners to the commercial marketplace. In addition, engineers from the IC vendors backing the Scalpel approach will also contribute to the development program, and their companies will receive the early prototype machines.
At Bell Labs, a second-generation "alpha" machine is being designed now that will incorporate a different writing strategy ? that is, the way the beam exposes the wafer ? than the first Scalpel. In the first-generation systems, a single electron beam scans the field in narrow stripes, which are "stitched" together in a time-consuming edge-matching step. To get to a higher throughput of about two dozen 300-mm wafers per hour, the Bell Labs team is developing beam deflectors that scan the beam in a wider swath.
Harriott said beam deflectors "are not a new thing, they are used in scanning microscopes. But we came up with ways to make them more efficient. The beam deflectors, and a new stage system, are the biggest differences in our second-generation alpha machine."
Using the alpha design, eLith engineers will develop the beta machines and send early beta systems out to the IC makers, including Lucent, Motorola, Samsung and Texas Instruments, among others.
Harriott said that while optical systems may have higher throughput at the 70-nm node, the important metric is "the cost of printing a wafer. For Scalpel, we use a simple binary mask, with no OPC, and so the mask costs are projected to be much lower. The advantage is there even for DRAMs, which is why Samsung is very interested in Scalpel."
Masks can be expensive, but since DRAM makers can use the same mask for several years for a single DRAM generation, it's worth the price, said Pascal Didier, senior vice president of worldwide customer operations for Cymer Inc. (San Diego), which claims to own 90 percent of the excimer laser market. Logic vendors, on the other hand, will spend more up front on the new generation of lithography tools but have the advantage of better accuracy to push the process shrink forward sooner.
"For the last couple of years, the logic guys have been in the control seat, driving the shift in lithography," Didier said. "In logic, vendors don't mix and match deep UV and I-line. For the DRAM, its 50-50."
At Semicon Japan, Cymer presented a paper describing the operational characteristics of a molecular F2 laser optimized for microlithography, which it claims is an encouraging feasibility study supporting the development of 157-nm lithography. After a checkered history, F2 lasers are now being seriously considered for coming onstream by 2005 to 2006 , he said.
Part of the reason is that advances in calcium fluoride masks allow 20 to 30 percent of the light to pass through, sufficient to transmit to the wafer and react with the photoresist, said Didier.
Tool questions
One of the biggest challenges of the 157-nm node will be determining the right tool design. Currently, there are two methods: refractive, which uses mirrors, and catadioptric, based on both mirrors and lenses. According to Didier, Canon, Nikon and ASML support the refractive approach, while Silicon Valley Group Lithography backs the catadioptric. Another hurdle is developing a cost-effective photoresist.
F2 can use binary masks down to the 70-nm node, so optical techniques such as phase-shift mask or OPC won't be needed. That begs the question of whether 157-nm deep-UV technology can be pushed further. "The question is, if you can employ some of these optical tricks on 157 nm, can you get it down to, say, 30 nm?" Didier asked.
With chip companies pursuing their pet lithography projects with an almost religious zeal, the debate over next-generation techniques may rage on for two to three years, Didier said. Any one of the NGL contenders is going to cost $6 billion and $9 billion in R&D, he added. But with the coming of 157-nm optical lithography, NGL won't be needed until about 2010, he said. |
