Intel sees the light, unveils chip industry's first EUV photomask

Hailing the development as a giant step for next-generation lithography, Intel boasts it has delivered the world's first EUV photomask. The giant chipmaker claims the breakthrough will enable the industry to continue Moore's Law scaling on silicon wafers through 2010. Beta tools incorporating the EUV technology will be available by 2003; production systems will hit the market by 2005. Volume production should follow by 2007, Intel asserts.

Two key developments contributed to the breakthrough. The first occurred when the EUV Limited Liability Corp. (EUV LLC), a consortium established to foster EUV development, created a reflective surface by coating a substrate with multiple layers of ultrathin silicon and molybdenum. Because the atmosphere and most materials absorb EUV light, the photomasks have to reflect the 13-nm­wavelength light rather than absorb it. The substrate made by the consortium, of which Intel is a member, has a particularly low rate of thermal expansion.

The second development came when researchers at Intel's Mask Operations and Components Research Center in Santa Clara, CA, demonstrated that existing mask-making technology can be extended to create a 6 x 6 in. EUV substrate within the industry's most advanced standard format for 130-nm processes. "We delivered this industry standard format using our current infrastructure," emphasizes Curt Jackson, advanced resist group leader at the research center. Both Intel and another consortium member, Motorola, had earlier demonstrated the ability to make an EUV mask, but it did not conform to the industry standard.

In addition to Intel and Motorola, the cooperative effort involved consortium members AMD, Micron, and Infineon. Completely funded by the member firms, the research was conducted at the EUV Virtual National Laboratory. VNL comprises three U.S. national laboratories: Lawrence Berkeley, Lawrence Livermore, and Sandia. IBM, which has teamed with Nikon to develop a competing electron-project lithography (EPL) system, decided in mid-March to cover other NGL bets by joining the EUV consortium.

The researchers began work approximately four years ago to develop the mask's reflective capabilities, says David Hwang, engineering manager in the Intel research center. He says the substrate has 80 layers of molybdenum. "It has to be defect free and it takes a long time to develop." The group spent almost nine months developing an e-beam photoresist, he adds.

Intel is touting the fact that the mask's 4x-reduction capability with 200-nm spaces can print a 50-nm feature on a silicon wafer "with existing electron-beam systems and the in-house mask-making system that we have," Jackson crows. Since Intel announced last fall it was developing a 30-nm transistor, the breakthrough has an added benefit. "This is one really key milestone that will enable the printing of that transistor," Jackson points out.

The nature of EUV light makes mask inspection and defect detection a challenge, but one that the industry is up to meeting, Jackson and others believe. Experts note that an EUV lithography mask inspection tool must show enough contrast to differentiate between the absorber metal and both the SiO^₂ and reflective multilayer substrate when a non-EUV wavelength is used.

"EUV mask inspection is different," Jackson emphasizes. "You need a reflective light and much better contrast to differentiate the image. That's the challenge for the future. There is a program in place on EUV mask inspection that has already demonstrated that capability." Multilayer inspection capability, he adds, "is the next step we have to work on."

Both the inspection of the multilayers and the inspection of the pattern itself are key requirements, and, should defects be found, "in both cases after you isolate these defects you want to be able to repair them. There's good news because with existing ion-beam repair technology today it's feasible to repair defects," says Jackson.

KLA-Tencor, the inspection and metrology system manufacturer, has been exploring NGL mask inspection as a participant in a NIST Advanced Technology Program, says Lance Glasser, vice president and general manager of the Rapid division. EUV LLC also participates, so "in that sense we're very well connected," he points out.

One of the advantages of working with the very short EUV wavelength is that no optical proximity correction is needed, Glasser emphasizes. The "simple 4x mask with 200-nm features on it "is not so different from what we're seeing today at KLA-Tencor. In that sense inspection is refreshingly easy. The main thing is you don't have all these OPC 'decorations,' so we don't have to achieve smaller minimum linewidths."

From his perch at a leading inspection tool company, Glasser has high praise for the Intel accomplishment. "The bottom line is that Intel has succeeded in optimizing the absorbers, thicknesses, and materials to make the contrast off those EUV masks just about equal to that of chrome-on-glass masks. So, these masks are quite inspectable from a standpoint of geometry size. We don't anticipate any huge problems because EUV doesn't require the advanced OPC decorations that are common today."

Addressing the business side of the issue, Glasser says it is likely the early EUV mask systems will be used in advanced development. "Those early masks would be inspected on one of our optical inspection tools. Once it's clear that the business is off and running and the timing becomes more clear, then obviously we can talk about whether we can inspect at EUV or not. That's far in the future."

One of the thornier questions to answer, Glasser stresses, is whether deep-UV inspection tools such as the company's Terastor system, which is shipping, and its next-generation Terascan system for 257-nm wavelengths are capable of defect detection at the EUV level. "You can argue that it should be pretty close, but seeing is believing and we just don't have that experience yet. If you think about it, we've never inspected masks at the wavelength they've used. It's always been one optical wavelength over another.

"This is obviously a much bigger stretch," he continues. "Is there a subclass of defects at DUV that are completely difficult to see at EUV? That's unknown for early development. We have a tool that will enable people to bring up the yield and find a lot of stuff. I think we can do our part."

Both ASM Lithography and SVG, which ASML is trying to acquire, are also ready to do their part to bring the breakthrough to the world's fabs in the form of a robust photolithography system, says John Cossins, recently promoted to product manager for EUV at ASML. "Like SVG, we've signed an agreement to commercialize the technology, and we're in the process of putting together a program to do so."

The goal of the program is to meet the broad requirements of EUV LLC's member companies and have a system ready "in the 2004 timeframe," Cossins says. The consortium approached ASML approximately three years ago "and basically asked us if we were interested in bringing this technology to market. We said yes."

Cossins notes that "for a long time, of course, EUV was just another NGL candidate." ASML was keen on both 157-nm lithography and EUV, he adds, pointing out that even though 157-nm "became more and more feasible, it's by no means to the market yet. There are still many hurdles for 157, and that in turn pushed NGL out to the 50-nm node from what was the 70-nm node and not so long ago the 100-nm node."

Cossins says productivity advantages of EUV over EPL's charged-particle lithography make the former more attractive as a "commercial prospect. From a technological standpoint we think the base technology can be exported to below 50 nm and maybe down to 35 nm."

For business reasons ASML wants any EUV-based photolithography system "to be a multigenerational tool," Cossins says before adding a note of caution: "Of course, EUV is not a done deal yet, either."

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit

Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com.

© 2007 Tom Cheyney
All rights reserved.