Commercializing NGL: The Push Forward

To stay on the industry roadmap and proceed with the ever-shrinking feature sizes required for tomorrow's semiconductors, lithographers will eventually have to give up on today's optical techniques. Although photolithography methods have hung on considerably longer than previously imagined, the industry has been preparing for the day when it will have to move on to next-generation lithography (NGL). International SEMATECH (ISMT, Austin, Texas) held its fifth and final NGL Workshop in August, achieving its goal of narrowing the NGL options and making a technology decision. The industry basically confirmed what had been supposed for some time — that it will proceed with both extreme ultraviolet lithography (EUVL) and electron-beam projection lithography (EPL).

"We have held five NGL Workshops and the results since the second one in 12/98 have been the same — pursue both EUVL and EPL," said John Canning, ISMT's NGL program manager. Nikon (Tokyo), Canon (Tokyo) and ASML (Veldhoven, Netherlands) are all working on commercialization of EUVL for the 50 nm node, and Nikon has a program in place for EPL at the 70 nm node as well. "They are all committed to their plans and further workshops by ISMT likely won't change that."

It was originally thought that the industry was going to have to settle on one NGL technology because it could not afford to support more than that. But that tune has changed along with the growing intensity of sticking with Moore's Law. Now, the industry can't afford not to pursue two technologies, Canning said, "since the stakes of not staying on the two-year roadmap are much higher."

	1. This computer illustration shows EUV light as a purple beam within EUVL equipment developed by the three members of the Virtual National Laboratory for EUV LLC. (Source: Sandia National Laboratories)

The cost of pursuing multiple technologies at once will be high, Canning conceded. "But that is the challenge — to find creative means to solve the cost issues." Nikon believes it can handle all three — 157 nm, EPL and EUVL — because it has been working toward EPL for more than a decade, and has positioned itself well via its technology project with IBM. In addition to 157 nm and EUVL, Canon is also pursuing a maskless form of e-beam lithography to release in parallel with EUVL. ASML, on the other hand, dropped its EPL option earlier this year, narrowing its NGL strategy to a combination of very high-NA 157 nm and EUV.

But the fact that the industry has settled fairly definitively on the future of next-generation lithography certainly doesn't mean that all of the technical details have been sorted out. "Until the industry sees fully functional alpha tools and masks, it will be concerned about potential showstoppers," Canning said. "Work continues at global R&D laboratories and suppliers to minimize these risks and build commercial products."

Following the latest NGL Workshop, the NGL Task Force met and updated the critical issues for EUVL and EPL (Table). According to Canning, the key challenges for EUVL include closing the gaps in funding and resources; solving the mask business model; accelerating exposure tool availability; and inserting EUVL at the 50 nm by 2007 for high-volume manufacturing. For EPL, he said, the key challenges are keeping up with the roadmap acceleration by planning and executing aggressively and demonstrating extendibility; solving the critical issues, including building end-user confidence and lowering the risk of decision; mitigating the mask business issues, including ever-rising mask complexity and limited R&D funds; and competing with optical and EUVL, with 157 nm lithography extending to the 65 nm node and EUVL planning for the 45 nm node.

Source: ISMT's NGL Task Force

Although the list looks considerable, lithography players are, by and large, confident that the obstacles can be overcome. What seems less certain is whether they can be overcome in the time dictated by an accelerated roadmap. While the likelihood is "quite high" that the industry will find solutions for remaining technical challenges, "the issues may be either time or cost," Canning said.

Some industry players see EPL positioning as a complementary technology to 157 nm for vias and contacts, Canning said, "and possibly for gate stack lithography if EUV is delayed." Nikon plans to bring EPL along in roughly the same timeframe as 157 nm lithography, Wiesner concurred. While it's very difficult to do contacts with optical technologies, he said, "EPL can do contacts like falling off a log." Meanwhile, 157 nm lithography will be useful at the 70 nm node for one-dimensional lithography such as metalization, gates, lines, etc.

Therefore, EPL will have a strong mix-and-match role, Wiesner said, and may end up being preferable for certain applications. He predicts that, when the masks for F₂ and EPL are both relatively mature, EPL will be cheaper for applications that require a lot of masks, such as ASICs. Unlike F₂, EPL will not likely use such cost-adding techniques as phase shifting or double exposure.

Nonetheless, EPL is not for everybody, and the cost and effort is certainly a barrier for those suppliers not already entrenched in the technology, Wiesner contended. "If you look at what's necessary to develop EPL, it's a massive effort," he said, noting that barriers to entry include getting designers together and transferring technology if you don't already have it. "If we were contemplating a year ago cranking up our EPL effort, it would not be good," he explained. "But Nikon has been working on EPL since the 1980s, with a good investment now underway."

Nikon has the resources to pursue EPL, 157 and EUVL simultaneously because of what Wiesner calls the "marching army effect." If a company is in sufficiently good shape to support the necessary staff, he said, then the best application of its resources are to keep its EPL specialists working on EPL. Body development (i.e., stages, support structures, environmental controls, software, etc.) for EPL and optical technologies are fairly well synchronized, he said, so the technology is efficiently applied across multiple platforms.

The other aspect of system development involves the guts of the machine — optics, illumination systems, etc. "But people specialized on EPL technology largely don't necessarily have the skills that you'd like to have applied to optics technology, and vice versa," Wiesner said. "As long as the industry has needs and applications for the three technologies and Nikon can support the development for all three, we will. Stopping any one at this point would not significantly accelerate the others."

One difficulty that Nikon may face in getting EPL to market is execution of the machine to schedule, Wiesner said. "We know what the engineering solutions should look like; now we just have to do it," he said. Another significant effort is making sure EPL has its mask infrastructure in place. "That is something that is being paid attention to and must continue to be paid attention to," he said.

Noreen Harned, vice president of the EUV program at ASML (Wilton, Conn.), questions the viability of EPL for commercialization. ASML had been partnering with Applied Materials (Santa Clara, Calif.) to commercialize EPL, but the companies dropped the project earlier this year. "Lack of extensibility (to satisfy multiple generations) was a big reason that we changed our product roadmap," Harned said

The fact that EPL is being supported by only one major tool supplier (while EUV is on the roadmap of all three) is a key indicator that it does not have the support it needs to be a serious contender. "EPL doesn't have the global infrastructure needed for commercialization," she said.

Canon, among others, is exploring maskless (or direct write) e-beam lithography. The NGL Workshop had dropped direct write e-beam lithography several years ago, largely because of considerable throughput issues. However, because of greater bit and circuit densities, as well as forecasts of excessive reticle costs for mask-based NGL techniques, maskless lithography has been making somewhat of a comeback.

One of the leading maskless contenders is Canon's system, dubbed ML2. The company aims to develop ML2 by around the 70 nm node. "Our approach uses a 64 × 64 array of micro e-beam columns working simultaneously on different areas of the field to get estimated throughput rates similar to EPL methods, but without the need for an expensive photomask," explained Phil Ware, senior fellow, lithography strategy, at Canon (Irving, Texas).

In general, e-beam lithography tools have a relatively low throughput rate because of the Coulomb interaction effect. The ML2 system exploits a large field size and high NA to get a higher throughput (Canon is aiming for 20 wph at the 70 nm node), then uses a new correction technique to offset the resulting increased lens aberrations.

One application that Canon hopes to target with this technology is system-on-a-chip (SoC) devices, low-volume products that could be seriously hurt by skyrocketing mask costs. The company's development plan calls for a beta tool by 2005, with production tools ready in the 2007 timeframe.

Nikon's Wiesner, however, questions the feasibility of developing production-ready maskless e-beam lithography within a reasonable timeframe. "The problem is that there's no demonstrated technology and architecture that really gives me any confidence that you can run forward with it." In his estimation, about 10 years of work still lies ahead.

Although Canon's program seems in the best shape, maskless lithography's maturity is nonetheless far behind where EUV or EPL are, Harned agreed. "It's about where those technologies were more than five years ago," she said. Again, she pointed out, the fact that Canon is the only major tool maker with direct write e-beam on its roadmap keeps the technology from being a formidable contender.

Despite the progress made in EUVL, significant challenges remain in both technical performance and cost-effectiveness. Equipment suppliers must strike a balance between the risks of investing too early and seeing industry requirements shift, and leaving development too late to meet industry needs.

	2. This full-field EUV mask produced by Motorola shows patterns in each rectangle that are reduced onto wafers. (Source: Sandia National Laboratories)

All the EUV source options being developed need to be scaled up in terms of power, Harned said, with the laser-produced plasma source developed by TRW's subsidiary CEO (Cutting Edge Optronics) the most mature. Other leading source options come from a joint venture between Philips Lighting and the Fraunhofer Institute, and from Cymer.

Even when a high-power source can be perfected, it will likely lead to concerns about wall plug efficiency, Wiesner said. "If you have to crank up a huge power drain for each tool, that's a consideration. What do you do with the waste heat?" Although these are just engineering issues, he said, they are on a much more intense accuracy scale when dealing with the 13.5 nm wavelength of EUV because precision must be maintained at atomic dimensions. "Trying to hold structures and surfaces at those dimensions will be difficult."

Reticle blank defect density is another troublesome issue, Canon's Ware noted, and another one that tool vendors have little control over. The prospects of having defect-free masks in the short term are not there, according to Wiesner. Of course, the key concern is whether a defect will actually print on the wafer. Unfortunately, studies have shown that, on a production EUV machine, very tiny defects will print.

From the mask maker's perspective, however, NGL does not fundamentally change their objective, the bottom line being to pattern masks, said Tom Blake, vice president of marketing at DuPont Photomasks Inc. (DPI, Round Rock, Texas). "The mask substrates will be different than today's quartz, so there is ongoing development for new blank technologies." As a result, the mask makers will simply need to optimize their process technologies around the new materials.

However, the tools that the mask makers will use to locate and identify reticle defects still need to be developed, Blake noted. Regardless of the lithography technique, the sectors that need more visibility are inspection and metrology, Wiesner said. "I think there's enough effort with respect to repair — there's a lot of good work going on there — but there's still some worries with inspection and metrology. What kind of commercial tool will you have available to do it?"

With the latest acceleration, the industry is geared toward a 2007 timeframe for EUVL's introduction into volume manufacturing. But, because the solutions to many of the challenges that the technology faces are so unknown, it is difficult to know whether 2007 is even a realistic target. That's why Nikon, for one, is focused more on getting to a proof of concept, Wiesner said, rather than cranking up a full-scale EUV development program.

"It's a commonly held belief in the semiconductor industry that, if you can measure it, you can fix it," Wiesner said. "I believe it can be done, but can it be done by 2007? I don't know. That's why we're in proof of concept for EUV. These questions have to be answered to spend real development money."

Having an EUVL tool available in the 2007 timeframe may be possible, Ware said. "But it is still unclear if a complete production-worthy EUV process will be available in that timeframe (meaning one with an affordable cost of ownership). It is impossible to know for sure at this point in time."

ASML's Harned noted that all the elements seem to be aligned for 2007. "All the major players have EUV on their roadmap, and all are talking about beta tools by around 2005, and volume production by 2007," she said. "And the top 10-20 chip manufacturers around the world, they all show EUV being used in their roadmaps. The demand and the interest seem to be there."

ASML is working on an alpha tool for a full-field EUV scanning system, Harned said, targeting its completion for late 2003. It's a first- and second-level tool, capable of creating test chips, albeit at a very low throughput. "There's a lot we believe we can gain by going through the process of tool build and integration, to help achieve performance and better understand the issues," she said, adding that an alpha tool will help reduce the risks involved with commercializing the technology.

Whether it's EUV or any other lithography technology, it must meet imaging performance criteria, be cost-effective, and be supported with a worldwide infrastructure, Harned said. "These are the challenges." Of course, simultaneously, manufacturers must work to meet demands in such areas as throughput, tight overlay, tool and mask costs, ease of use, extensibility, etc., she added.

"All NGL methods have significant hurdles, not the least of which are the various flavors of reticles," Ware said. "But if you look at the lithography tables in the ... ITRS 2001 update, you will find most of the requirements for the years beyond 2005 are listed in red boxes. This means that there is no known solution at the moment. Invention is required to meet these requirements."

All the players can really do is keep working toward commercialization, establishing the infrastructure needed for success. Although ISMT has completed its series of NGL Workshops, it will continue to fund and support EUVL and EPL. "With EUVL as the preferred solution for both the 45 nm and 32 nm nodes, ISMT plans to take a very active role in commercializing EUVL," Canning said. "Through our prior funding of critical issue projects at the EUV LLC, we have identified several areas where we can accelerate the technology: mask blank supply, mask blank deposition, and inspection tools; mask inspection, repairs and manufacturing; resist development through the extension of our very successful Resist Test Center; and EUV materials lifetime evaluation; industry coordination of EUV source metrology, and reticle protection and standards issues."

Canning added, "For EPL, we plan to collaborate with Selete to evaluate and drive the Nikon EPL technology for application on contact levels at 70 nm in 2005. We will concentrate on developing solutions for the stencil mask process."

Meanwhile, the industry will stretch 193 nm lithography as long as it can. "The message is that we're going to push ArF as far as it goes, certainly through the 90-100 nm node, with very high-NA machines," Wiesner said.