Chipmakers Immerse Themselves in 193 Wet

Immersion lithography, little more than a parlor trick just a few short years ago, is well on its way to full volume production. Full-field immersion scanners are out in the field, and chipmakers are beginning to come forward with news of their successful application.
 
In an apparent race to be the first to use immersion lithography in a production environment, both TSMC and IBM made separate announcements in December about successful patterning of fully functional chip layers with 193 nm immersion. IBM printed a critical layer of 90 nm microprocessors based on its Power architecture, and TSMC printed a critical layer of 90 nm SRAM chips.

IBM used an ASML TwinScan AT:1150i immersion scanner (with a numerical aperture of 0.75) installed this past summer at Albany NanoTech, where the chipmaker is partnering with several other companies in a New York State technology initiative. The trial involved a production lot of 25 300 mm wafers taken from IBM's fab in East Fishkill, N.Y., to Albany. The immersion system printed one interconnect layer in 90 nm copper technology, then the lot was shipped back to IBM's fab to complete the production process with 193 nm dry lithography.

IBM is not revealing much about the experiment, but they contend that the results were good — showing acceptable control of overlay, critical dimensions and depth of focus. IBM shot the critical layers on the immersion tool in parallel with standard processing, according to Bernard Meyerson, IBM Fellow, and vice president and chief technologist of the Systems and Technology Group, and yield results were comparable.

"It ran the first time, and it ran well. The quality of the result was excellent," Meyerson said. He noted that this use of a full-field immersion tool on commercial chips was significant, compared with the test patterns that have been produced previously. "There is an enormity of difference between that and a precious cargo of microprocessors, running them through the tool, and having them come out on the first run. That's impressive."

TSMC also used ASML's TwinScan AT:1150i, in this case doing the exposure and development of the chip's polysilicon layer at ASML's facility, according to Burn Lin, senior director of TSMC's micropatterning division. Yield, device characteristics, and defect levels were comparable for both dry and wet scanners, according to the company, and the yield-related depth of focus of the immersion scanner was almost twice that of the dry scanner. TSMC plans to give a detailed report at the SPIE Microlithography conference in San Jose at the end of this month, Lin said.

TSMC has ASML's next-generation immersion tool, the XT:1250i (NA=0.85), at its own facilities, and is using it to develop its 65 nm immersion processes, Lin said. "Having the tool in a manufacturing environment does help to clarify a lot of issues. We will address some of them in our presentation at SPIE."

Both IBM and TSMC, despite their announcements, are being fairly tight-lipped about their findings, but both seem pleased with the results. It's part of the tremendously competitive environment that immersion lithography has become. Despite the incredible amount of collaboration that has been done to take immersion from lab curiosity to the next real production tool in just two to three years, chipmakers are now likely to keep whatever gains they can get from splashing water on their wafers under their hats. "There's something there to compete about," noted Andrew Grenville, program manager for immersion strategy at Sematech (Austin, Texas). "It says something about how seriously people are taking this."

Grenville, a Sematech assignee from Intel, said that progress is being made on the top issue for immersion, which is defectivity. Researchers have made amazingly quick work of many of the critical issues identified for immersion lithography since the industry first started getting serious about studying its viability. The list of critical issues continues to change, as problems are solved and new challenges are introduced. Results from the latest poll, taken at the Symposium on Immersion and 157 nm Lithography in August, reduced the list from 10 to just six (with new issues shown in italics):

1. Defect formation, control and characterization.
2. Hyper-NA complexity and reticle polarization constraints, including impact on field size (lens complexity), reduction ratio, RETs and simulation tools.
3. Robust protective layer for bottom element, including laser durability and contamination susceptibility.
4. Availability of high-index immersion fluid with suitable properties.
5. Influence of fluid on resist, resulting in altered process properties (e.g., swelling); necessity/availability of topcoat.
6. Specification and preparation of water, including absorption control, surface tension and wetability, and electrostatic discharge.

Those issues have been reduced through collaborative efforts and data sharing, Grenville noted. "Defectivity couldn't really be proven until we got those tools out in production. We're just starting to see that play out now."

Grenville noted that more work needs to be done on resist interactions — what comes out of the resist, how that impacts the imaging quality of the resists, and how that might impact the bottom element on the lens. But recent findings have shown that contamination may in fact be less of a problem with immersion lithography than it is with dry, he said. It appears that the contaminants are less able to get through the water or, if they do, they don't get out of the water to the lens. "Again, you've got to get real data from the fab," Grenville said, but all the signs are looking positive.

Many experiments have also proven positive for defect formation as well. There is a broad array of things that could cause defects, Grenville said, such as the formation of bubbles when scanning over topography. Experiments have so far shown, however, that fluid flow over the types of topography that would be typical does not generate bubbles, he said.

While these are the kinds of things that give immersion developers confidence, and will better enable them to understand full-field production and to know where to look for problems, industry players have been waiting for the true test — volume production. TSMC's and IBM's recent announcements are the first public proclamations of immersion's viability in real production with working semiconductors.

IMEC (Leuven, Belgium) began its industrial affiliation program on immersion lithography last July, and received ASML's XT:1250i in December. The two-year IMEC program is expected to investigate process-related questions including defectivity and resist-related issues, with the goal of preparing immersion lithography for production by 2007. IMEC's immersion program includes more than 30 participating companies, with chipmakers including Infineon, Intel, Matsushita, Philips, Samsung, STMicroelectronics, Texas Instruments and Sony.

The first-generation immersion tools, with NAs of ~0.75-0.93, take advantage of water's higher refractive index to greatly improve depth of focus (DOF) for lithographers. The next generation, on the other hand, will start to use what are referred to as hyper-NA lenses (NA>1) to improve resolution capabilities. The NA limit for a lens system in a dry tool is <1, and it is only with liquid between the final lens and the wafer that higher NAs can be used.

In the past, NA was progressing and so was the k1 factor. And the product of the two was the resulting resolution improvements," said Boudewijn Sluijk, director of corporate marketing at ASML (Veldhoven, Netherlands). "It's now up to the NA, and it needs to grow much faster than in the past." See Figure 1 for a roadmap and limitations.

1. This chart details the requirements for numerical apertures, with dry production tools reaching their limit by next year. To continue immersion lithography beyond around 2010 will require the use of a fluid with a higher refractive index than water. (Source: ASML)

Sematech will be bringing a 1.3-NA microstepper into its Immersion Technology Center in Austin this coming summer, which will be a "unique capability in the world," Grenville said. "We'll be able to do resist experiments, test out polarization, various illumination schemes."

Last summer, Sematech and Exitech Ltd. (Oxford, UK) announced an agreement to develop the first ultrahigh-NA 193 nm immersion lithography tool. The microexposure tool, the MS-193i, incorporates a 1.3-NA catadioptric, 0.4 mm field, water immersion imaging objective lens developed by Corning Tropel (Fairport, N.Y.). With a 4 kHz, linearly polarized, 193 nm natural bandwidth ArF laser source from Lambda Physik (Göttingen, Germany), the tool is expected to image minimum feature sizes of 70 and 45 nm using binary and phase-shifting masks, respectively.

"It's an exciting capability," Grenville said. "Our immersion program is focused on, first of all, the critical issues list. And now, looking ahead to the true extendability of immersion, that's really our focus."

Although immersion lithography enables higher-NA lenses, though, it doesn't make them any easier to make. There's no doubt that these hyper-NA lithography lenses are the most complex lenses that get built, Grenville said; they drive invention in this whole area. "All of [the lens makers] have stated publicly that they can get to 1.1 NA, and the evolution is possible for 1.3 NA," he said. But the real question is the cost, which scales non-linearly with NA. "But, in principle, it can be built."

Not only is the size of the lens element an issue with hyper-NA lenses, but the catadioptric design required of the lens system poses a whole new set of challenges for the toolmakers as well, according to Sluijk. Today's lens systems are symmetric, composed of two dozen or more lenses in a straight stack. "For the hyper-NA lens design, we will be using mirrors as part of the image-forming elements. Once you use a mirror, you can no longer have it all in one stack because the mirror will block the light," he said. "Actually, a mirror is easier than a lens element. But, from a mechanical point of view, it's a lot easier to build a stack than it is to have a sideport protruding, particularly if you're talking about millilambdas and nanometer accuracy."

Figure 2 shows two examples of catadioptric hyper-NA lens designs from toolmaker Nikon Corp.

2. Two examples of hyper-NA lens systems based on a catadioptric design show the departure from a symmetrical stack to a more complex system with protruding sideports. (Source: Nikon)

ASML aims to have a hyper-NA immersion system available by 2006, which requires that the company be well on its way with the machine design sometime this year. According to its presentation in August, Nikon will be ready to ship its S609 production tool, with an NA≥1, in the second half of this year; and will have an NA≥1.2 production tool ready by the second half of 2006. Canon's development schedule dictates an NA>1.2 production tool ready by 2007.

Current immersion lithography systems use deionized water as the liquid between the optical lens and the wafer, with a refractive index of 1.44. The next step in immersion's progression through the technology nodes will likely be fluids with even higher refractive indexes. The refractive index of the fluid correlates with the NA limits of the lenses. Thus, just as NA~1 is the theoretical limit in dry lithography, where the refractive index of air is 1, the limit with deionized water is NA~1.44. "If you could find a liquid with a refractive index of 1.6 or 1.7, then you could go to a 1.6 or 1.7 higher NA," Sluijk said.

High-index fluids are being explored by several groups, including some promising research at Rochester Institute of Technology's Center for Nanolithography Research.

JSR Corp. (Tokyo) recently announced that it has successfully demonstrated 32 nm line and space patterns with a 193 nm immersion lithography system and its high-index solution based on the company's SOLOnX technology. According to JSR, the fluid offers a highly transparent and low-viscosity solution with a refractive index of 1.64.

The trick will be inventing a fluid that satisfies all the requirements, including having a high index while maintaining a low absorbance, "and then having people coalesce around and accept a certain fluid," Grenville said. "Right now we're still in the discovery phase, and trying to see what that landscape looks like." This coming year is likely to be very interesting, he added, expecting that many results will be presented at SPIE Microlithography this month and the next Immersion Symposium in September.

Higher-index fluids seems to be much easier than the alternative — 157 nm immersion lithography, according to Sluijk. "The window of opportunity (for 157 immersion) is extremely small. If EUV fails, then there might be a window of opportunity for 157 immersion," he said. "That possible window is so small, it's hard to see it. It's also so small that it's hard to imagine the industry making the investment that still needs to be made."

Though not off the table yet, 157 nm immersion lithography will certainly be difficult. "First and foremost, you need an immersion fluid that has sufficiently high transmission and index of refraction, and that hasn't been demonstrated," Grenville said. It's intrinsically harder to have a high-index fluid with 157 nm lithography, and to have the transparency that's required at that wavelength, he said. "From where we sit today, 157 immersion is certainly a long shot. I don't see where it fits in."

Nonetheless, there is work that continues, including at MIT's Lincoln Laboratory, with a DARPA-funded 157 nm immersion microstepper. "It's high-risk research," Grenville said. "There's a place for that, but it's a very different level of discussion than the 193 immersion."

To learn about OPC modeling for immersion lithography, read the web-exclusive feature, " Litho H₂O: OPC Modeling for Immersion Lithography ."