NGL Fights Through Economic Adversity

At her presentation at the Lithography Breakfast Forum in July, Janice Golda, assistant director of lithography capital equipment development at Intel Corp. (Santa Clara, Calif.), likened the current push of lithography tools to the increasingly complex contraptions used by Wile E. Coyote in his quest to catch the elusive Road Runner. Using all the tricks of the trade — and coming up with a few more along the way — developers have managed to extend 248 nm lithography further than anyone thought possible, and likely will do the same with 193 nm lithography, continuing to push next-generation lithography (NGL) further out. "Hopefully we can recognize the contraption limits before we go off the cliff," Golda said.

Extreme ultraviolet (EUV) lithography is still considered the frontrunner in the NGL race, but it seems forever destined to be for the "next" generation. According to the current International Technology Roadmap for Semiconductors (ITRS), EUV lithography is targeted for the 45 nm node, and should see volume production in 2009. There is some question, however, whether EUV will be the right answer for 45 nm, or whether 193 nm lithography (wet or dry) will be the more likely choice.

"Everyone's crystal ball for the future is not exactly the best," said Noreen Harned, vice president of product marketing for ASML (Wilton, Conn.). Based on sanity checks with International SEMATECH (ISMT, Austin, Texas) and various lithography directors, however, she expects high-volume EUV lithography to begin in 2009. She also noted that the more likely place to see EUV being used in high volumes is at the 32 nm node.

There is an argument that EUV lithography doesn't really make sense until 2013, given the history of timing in the industry and the need for an acceptable return on investment (ROI). "There are typically six years between the introduction of each new wavelength," said Bob Sell, manager of marketing and strategy development for Corning Inc. (Corning, N.Y.), which makes materials for optical lenses and masks. Because 193 nm lithography was introduced in 2001, it makes sense for 157 nm lithography to be introduced in 2007, rather than 2005, therefore pushing EUV out to 2013, he said. "That way, companies can get their return on investment."

1. Particularly considering the current state of the economy, lithographers will do what they can to stick with existing tools rather than move on to next-generation platforms. (Source: ASML)

The industry has been investing money, time and effort into EUV lithography for years. In an effort spearheaded by Intel, chipmakers joined forces with three U.S. national labs beginning in 1997 to form the EUV LLC, a cooperative agreement designed to further EUV development. The EUV LLC has now wound down its research activities, turning further development over to commercial tool manufacturers. An extension of some of that research is going on at ISMT North in Albany, N.Y., a cooperative with the University at Albany to advance EUV work in mask blanks, resists and EUV extensions.

More recently, Japanese manufacturers have decided they must catch up with chipmakers in the United States and Europe with the formation of their own EUV lithography development initiative. The Extreme Ultraviolet Lithography System Development Association (EUVA), established in June 2002 as a four-year project, includes cooperation from Gigaphoton, Ushio, Komatsu, Canon, Nikon, the National Institute of Advanced Industrial Science and Technology (AIST) and several universities. Although the group has gotten a relatively late start, they plan to develop a beta tool by the end of 2005 (Fig. 2). Table 1 shows what specifications the EUVA is working toward for the beta and production tools.

Nikon Corp. (Tokyo) and Canon Inc. (Tokyo) will collaborate on the beta tool, but ultimately will develop their own production tools. Nikon, Canon and ASML have all been working together in pre-competitive areas of EUV development to some extent. ASML's Harned mentioned the source and the pellicle (or lack thereof) as two key areas in which the tool manufacturers have cooperated. The tool manufacturers exchanged generic throughput models, specifications, etc., to develop a common set of numbers for which source developers could strive, she said. And they are collaborating through SEMI on standards for mask clamping, handling, etc., for EUV. "EUV can't have a pellicle, so you have to have creative and clever ways to deal with mask handling," Harned said.

2. Through Japanese consortium EUVA, Nikon and Canon are collaborating on development of a beta EUV lithography exposure tool, scheduled for completion by the end of 2005. From there, the companies will develop their own production tools. (Source: EUVA)

The abundance of creative and clever solutions that are needed for various aspects of EUV development are why some naysayers are not convinced EUV lithography will ever really see volume production. "EUV is such a gamble, I think they're going to build all the fabs in Las Vegas," joked Phil Ware, senior fellow, lithography strategy, for Canon USA (Irving, Texas). Japanese suppliers, in particular, have been through this gamble before, developing X-ray lithography systems that ultimately didn't see the light of day. After billions of dollars in development, Canon sold one X-ray lithography tool, Ware noted, in 2000.

But work continues on EUV development, and suppliers have been announcing commercial systems, furthering EUV advancements. XTREME technologies GmbH (Munich, Germany), a joint venture between Lambda Physik AG (Göttingen, Germany) and Jenoptik AG (Jena, Germany), announced its commercially available EUV lithography light source, the XTS 13-35. The light source, which emits a 35 W average power at 1 kHz pulse repetition rate, will be incorporated into microsteppers from Exitech Ltd. (Oxford, England). Exitech is scheduled to install its MS-13 exposure tool for EUV resist evaluation and testing at ISMT North later this year.

3. Carl Zeiss SMT’s commercial illumination system, shown here during final assembly, collects light from the plasma source and ensures uniform illumination of the reticle. (Source: Carl Zeiss SMT)

Carl Zeiss SMT (Oberkochen, Germany) has announced the first commercial illumination system for EUV lithography. The completed illumination system (Fig. 3) allows the use of compact EUV plasma sources.

One thing that has differentiated Nikon from its major competitors is its pursuit of electron projection lithography (EPL) as an NGL technique. Although e-beam technology still has throughput issues, Nikon has made progress in this area, and the tool is expected to have some use in the not-so-distant future. Nikon announced in July that it shipped its first e-beam stepper to R&D consortium Selete (Tsukuba City, Japan).

Although e-beam systems in the past had throughput times of <1 wph, Nikon has increased that to 7-10 wph for 300 mm wafers, with a mass production system expected to achieve 15-20 wph. This is still a far cry from the >100 wph typical of optical lithography tools, but EPL systems are targeted at niche areas that can benefit from a large depth of focus while not suffering from the slow output time. "E-beam was never considered as a standalone NGL tool," said Gene Fuller, principal engineer at Nikon Precision Inc. (Belmont, Calif.). Contact holes are one area for which e-beam technology likely would make sense, he said.

Although research on NGL techniques continues, more of the focus these days is on ways to keep existing tools and technologies going. Given the current economic situation, companies are forced to focus more on the short-term perspective, argued Paul van Attekum, senior vice president of ASML (Veldhoven, Netherlands). "You have to make sure there is still a next generation to come. As a consequence, there is more focus in all companies — and also ASML — on the short term."

But whether the economy is good or bad, chipmakers will always try to stay with what they already have, van Attekum said. "Even when the economics are good, the guy in the fab does not want to change," Harned added.

The industry has been able to sustain 248 nm lithography as the primary tool for much longer than anyone expected. "The longevity of 248 is really surprising the whole industry," van Attekum said.

Because of this, many expect 193 nm lithography to do the same. "People are pushing 248 nm lithography so far that, if we jump through 193 nm lithography so fast, where do you get your return on investment?" Corning's Sell said.

Canon, which has taken up the "Optics Forever" mantra, is, like other manufacturers, exploring new ways to push 193 to its limits. "We even have customers who don't believe ArF is ready yet, and are pushing us to extend KrF further," Ware said. Canon was the first to introduce a tool with a lens numerical aperture of 0.86, although the others have since followed suit. All three toolmakers expect to have reached >0.9 NA by 2004.

Some debate is going on among industry players about Intel's recent announcement that it plans to use 193 nm lithography at the 45 nm node, moving straight on to EUV lithography without stopping at 157. At her presentation at the Lithography Breakfast Forum, Golda reiterated Intel's intent to use ArF (193 nm) at the 45 nm node, clarifying that it would be a dry ArF rather than a liquid immersion form — for which the jury is still out.

"Intel has the right product mix to make that happen. It can pay for the higher-cost masks," Sell said. Because of the complexity of masks with advanced resolution enhancement techniques (RETs), mask sets are expected to cost millions of dollars. "But ASICs and other low-volume devices — can they afford the masks? ... I don't think anybody else thinks they can do that."

For the time being, everyone (including Intel) continues to participate in 157 nm development programs, reserving final judgment until more questions can be answered. "The question of whether 157 will be or not is still far out," van Attekum said. "I think that the jury is still out on that question. I would say for the 65 nm node, it is all relatively clear that we'll do 193. What happens after 65 nm, there is a lot of debate, and a lot of it depends on what devices you're making."

One debate is whether 157 nm lithography could be usurped by a wet form of 193 nm lithography — liquid immersion. Used for years by microscopy, immersion techniques have been a hot topic over the past year or so, with researchers working feverishly to decide, by year's end, whether it makes sense to pursue the technique that effectively reduces the light source wavelength by placing ultrapure water between the final lens element and the wafer. Despite some misgivings, Ware noted several advantages of 193 nm immersion lithography over 157 nm dry lithography:

• Immersion may be the easiest way to reach beyond 65 nm.
• There are no apparent showstoppers.
• There are no CaF₂ issues.
• No nitrogen purging is required.
• Resists are more or less available now, with little tweaking needed.
• Immersion relaxes the k₁ factor.

ArF immersion covers the full range of F₂ dry, and then some, Ware said, adding that immersion lithography could even use binary masks at 65 nm. However, there are still issues to work out before deciding that immersion is indeed the way to go.

The three major toolmakers are expected to announce their intent to pursue or not pursue immersion lithography around the end of this year. "It will take certainly until that time until we hear a preference in the industry," van Attekum said. "Then we'll enter a new phase, where you have a judgment of whether or not you can do it, but you really have to figure out what the issues are."

Nikon, for one, has been playing up the immersion angle, promoting the idea in its booth this year at SEMICON West, and announcing in July its agreement with Tokyo Electron Ltd. (TEL) to jointly develop liquid immersion exposure technology. The companies will verify the basic technologies by the end of the year, aiming to enter mass production as soon thereafter as possible. However, Bernie Wood, Nikon Precision's director of marketing, pointed out, "Although Nikon has been saying a lot about immersion, we have not yet stopped evaluation of F₂."

It's not likely that both immersion and F₂ would be developed intensely at the same time, Fuller said, but it's also not a foregone conclusion that F₂ would be thrown out if the decision is made to pursue immersion. "The issue with all of this, of course, is until you really get deeper into something, you don't really understand all the problems," he said. "So, the idea of completely throwing out a possibility in favor of another one — that's a serious decision."

There are several break points in any given development program, Fuller explained, and early exploration may not require a lot of money or other resources. "When you get to a point where you have hundreds of people on a program, and you're spending many billions of dollars and so on, then you've got to be serious about it, and you've got to be right. Neither immersion nor F₂ has really reached the point where you'd say, absolutely, this is the one and only."

From the supply side, the optimal pursuit also is influenced by a company's position as a supplier of the lithography tools or of the surrounding infrastructure, van Attekum said. On the infrastructure side, 157 nm lithography remains a relatively sticky issue, with development still to be done with resists, masks and particularly pellicles. On the other hand, 193 immersion would require very little change in the way of supporting infrastructure.

From the toolmaker's perspective, however, 157 nm lithography may look like the easier choice. "With 157, I know which problems I have to solve; I know how to solve them," van Attekum said, noting that ASML has a full-field volume product in the market already. "Yes, it is not perfect, but we know what steps we have to take. It's under control; you know what to do." Harned added, "Where the industry is with the knowledge to use 157, we're actually further along than immersion, even though immersion infrastructure is in place."

The complexities are certainly considerably more than they will be for ArF immersion, but the idea still lingers that F₂ lithography could emerge in a wet form as well. If that does happen, it could push out even further the timeline for EUV, which is driven not only by capability, but by need. "If there's no need because, say, we can do F₂ immersion, then it's going to get pushed out," Fuller said. "On the other hand, one of the things that pushed out F₂ was the capability to build good calcium fluoride. So there's always that balance."

Part of the reason Intel figures on being able to extend dry ArF lithography through the 45 nm node is because of its interpretation of the ITRS. Traditionally, a technology node on the roadmap is dictated by the half-pitch of a circuit design. Although those are admittedly loose guides, Intel's roadmap shows a significant difference in numbers (Table 2 ). At the 45 nm node, Intel's logic chips are expected to have a half-pitch of ~75 nm.

"It's almost a marketing kind of thing to decide which node you're going to call it. The customers expect that you're going to go from 130 to 90 and so on," said Nikon's Fuller, who was involved in lithography at Texas Instruments. "My observation looking at quite a few roadmaps of different companies now for the last few years that I have access to is that everybody does it a little differently. When you look at the real internal product roadmaps, and what feature size they're going to have, they're all a little bit different, and they all reference the same ITRS."

"What is at this moment a node? There are no nodes anymore," van Attekum argued, noting that ASML has customers and applications with half-pitches every 5 nm, or even every 2 nm. Ultimately, lithographers are more interested in a realistic, optimized solution than they are in a particular node. "If you talk about the 45 nm node, then you have quite a different set of requirements if this is for a memory manufacturer than that same label for a microprocessor manufacturer."

The whole node concept may just diffuse away after 65 nm, mentioned Chris Progler, chief scientist at maskmaker Photronics (Allen, Texas), at the Lithography Breakfast Forum. Chipmakers will likely be more interested in fine-tuning cost vs. performance rather than trying to keep pace with the node system.

Although Fuller supports the concept of a roadmap, it's used more as a guide than a driver. "It's not the actual process plan of any given company," he said. "The only suggestion is the one that everybody says — that we're just stretching the truth a little bit when we talk about these nodes. I don't think it has to change, necessarily, but, if you ask anybody...they'll all acknowledge that the roadmap is a little bit out of whack right now."

Another complaint with the current roadmap is the number of possible solutions listed for possible NGL development. System and material suppliers have been stretched so thin in NGL development that the last thing they want to see on the roadmap is another candidate for further development.

"Ten years ago, there was relatively little debate that we had to go from i-line to 248, and from 248 to 193," van Attekum said. "Nowadays, that is much more uncertain." For anyone in the industry to try to develop all the various options in parallel would be impossible. "For the industry, it would be best if we could agree that this will be the solution. But there's not a clear optimum for the total industry."

The Potential Solutions chart on the roadmap is not intended to be a definitive list of the upcoming lithography technologies, Fuller noted. "You have to demonstrate that you have a viable concept, and that there's some reasonable expectation that, given the right amount of funding, it could be developed into a manufacturing-worthy tool. And that's about all it takes to get it on that Potential Solutions chart," he explained. "Some of the ideas will just never happen; some of them will become mainstream. The interesting part is, if you go back a few years and look at what everybody thought was mainstream, it varies."

By trying to rule out the use of 157 nm lithography, Intel may at least be helping the industry make a clear decision, van Attekum said. "But it's optimized for Intel. The big question is, Will that be accepted by others? So far there's not a clear reaction."

What is likely to happen in more than one aspect of lithography development is the divergence of the industry. Depending on the types of devices being made, some chipmakers may pursue further extension of current tools, willing to absorb the necessary mask costs, while others may opt for a shorter wavelength, even if that means a platform change. This will require that resources be split even further among the suppliers.

Given the current economic situation, let alone the prospect of having to develop several different technologies simultaneously, suppliers often voice concern about where the resources are going to come from. "It would be nice if there were more sources to share the expense," ASML's van Attekum said. "On the other hand, this is not research for universities. You don't do it because you want to do research."