Tricks With Water and Light: 193 nm Extension

1. This advanced binary mask for 193 nm scanners is for use at the 130 nm node. Phase-shift masks and other resolution enhancement techniques should be able to take 193 nm lithography to at least the 65 nm node. (Source: Photronics Inc.)

2. Immersion lithography will likely extend 193 nm lithography to at least the 45 nm node. Shown here is an experimental system, with water placed between the projection lens and the wafer. (Source: Nikon, TEL)

Although subwavelength lithography was once not even considered an option, it's generally accepted these days that a light source of any given wavelength will be able to print features at a resolution one-third its wavelength. Along these lines, the use of 193 nm lithography tools for the 65 nm node is already a foregone conclusion (Fig. 1 ).
 
Not to say that it's a walk in the park. Just what exactly it will take to extend 193 nm (ArF) lithography as far as it's needed, or how far it will manage to go, is still up for debate. The industry has a few tricks up its sleeve, though. At this point, the path to 193 extension includes a whole cadre of resolution enhancement techniques (RETs), including strong phase shifting, assist features and illumination tricks; not to mention the topic of the year — immersion lithography (Fig. 2 ).
 
How far chipmakers are likely to take 193 nm lithography depends in part on how aggressive a set of ground rules they will be able to use, according to Chris Progler, chief scientist at maskmaker Photronics Inc. (Brookfield, Conn.). Few companies will actually do a full device shrink, and they will only really look for 65 nm node devices with 193 nm lithography where they really need it, he said.
 
Progler predicts that there will be a toggling back of ground rules at the 65 nm node. Many of the ground rules can be hit at the 45 nm node, tackled through another round of increases in numerical aperture (NA), or an immersion approach, he said. "Beyond 45, then, I guess I've said this a few times before, and I believe it — I think the notion of nodes is really going to start to diffuse away. The notion of an ITRS-like node every two to three years, honestly, I don't think they'll be sustainable. I think we should rename them hurricanes instead of nodes." The notion of a 70% shrink of all the ground rules is unrealistic, he added.

Try as it might to narrow down the options, the industry is actually faced with a fair number of commonly accepted options as to how to get from node to node. As we move forward, decisions will have to be made. As Craig West, director of applications for DuPont Photomasks Inc. (DPI, Round Rock, Texas), puts it, "You look so far into the future, you've just got to place some forks in the road." At one fork, for example, is 193 wet vs. 157 dry. Another fork shows 157 wet vs. EUV. Other forks ultimately force choices between increasingly complex masks or shorter wavelengths.

But there is hope. "2003 really solidified some of the stretch and perhaps pruned some of the forks in the road with regard to the wavelength choice," West said. "Really, in 2003, what helped all of us is that we looked forward to the next few generations. The 65 nm node can definitely be done with dry exposure 193 nm lithography. And then 45 nm looks to be easily within the capabilities once you add immersion."

Immersion techniques could take 193 nm lithography even further, some believe. "It's not inconceivable that 45 nm is just one of the stops on the path, and could be taken even further than that," said Steve Carlson, senior vice president of technology for Photronics.

Phase-shift masks (PSMs) will be key to furthering the capabilities of 193 nm lithography. "Phase shifting has become a mainstream solution for memory and logic. PSM is at the heart of advanced lithography solutions," DPI's West said. "Many of the layer patterns have easy mapping into phase-shifting masks, particularly the hole patterns for contacts and vias."

Moving forward, PSMs will rely on stronger phase shifting and more complex features. A 6% embedded PSM is relatively standard now, Progler noted. But there are three main types of masks that take phase shifting to the next level: complementary phase shifting, double exposure, and single-exposure techniques that can do low-k₁ imaging.

Complementary phase shifting, in which the primary mask serves as a strong shifter and a second (complementary) mask serves the trim function, is a well-known entity at this point, West said. "The infrastructure has clearly adopted and mastered many of the complementary phase-shift masks," he said. "The one rub about that approach might be at face value that there are two reticles." Each wafer level must be double-printed, he added, which lengthens printing time and presents possible overlay issues.

Complementary PSMs are fairly challenging, Progler said, but they're considered by most to have the highest imaging potential. "However, the two masks make it costly to run."

Double-exposure methods also use two masks, but with a different approach than complementary techniques. While the second mask in complementary phase shifting is what's considered easy, used just to clean up after the first mask, Progler said, the two masks in double-exposure phase shifting look pretty similar to one another. Each of the two masks is individually simpler than the primary mask in complementary techniques.

3. This chromeless phase mask shows a minimum etched quartz feature of 165 nm (mask dimension). This design leaves some patches of chrome on the quartz, with a minimum patch size of 350 × 270 nm. (Source: LSI Logic, DuPont Photomasks)

Leaky chrome

The third contender is chromeless mask lithography, a single-mask, single-exposure technique (Fig. 3 ). It relies on complex masks that typically require multiple writes to create, but its single exposure is an important benefit. United Microelectronics Corp. (UMC, Hsinchu, Taiwan) announced in early December that it had successfully used such chromeless PSMs to produce functional customer chips at the 90 nm node. The foundry said it was evaluating the use of the technology for the 65 nm node and beyond, along with other RETs.

Chromeless Phase Lithography (CPL) is a version of the technique from ASML MaskTools (Santa Clara, Calif.) that is getting a lot of attention. "The greatest benefit in CPL is the potential of a single mask (of many complex subresolution patterns) for more rapidly printing critical layers without a decomposed, multimask method like in complementary designs," West said. "The balance point in this decision will probably come down to software tool solutions and scanner tool efficiencies. Many of the key mask technologies are already in production or a development priority for any of the etched quartz concepts. And a strong shifter solution will always be the preferred lithographic solution, regardless of the type or flavor."

CPL uses chromeless features on the mask to define patterns that have nearly 100% transmission and are phase shifted by 180°. Phase shifting is induced through etched quartz on standard chrome-on-quartz substrates. CPL has all the benefits of complementary phase shifting, but with the added benefit of one mask, according to Photronics' Carlson. "In fact, it provides one of the more extreme extension opportunities."

"You may get somewhat better resolution perhaps with double exposure," said Dinesh Bettadapur, president and CEO of ASML MaskTools. "But in terms of what you gain, CPL is a much stronger value proposition." The marginal resolution gap is on the order of a few percent, according to Carlson, and it's only visible on certain patterns.

Also gaining a lot of attention is a mask method that simply takes the current 6% embedded PSM and raises transmission. "We call it variable transmission because we're still trying to sort out the optimum transmission," Progler said, noting that some masks may have 40% transmission while others have 9%, for example. "To some extent, the type of layer you're doing drives the transmission optimum."

Increasing the transmission rate — essentially creating more "leaky" chrome — gives the mask a stronger phase effect, Progler explained. The light that gets through the dark part of the mask interferes with the light part of the mask, creating phase shifting. There are trade-offs, however. "When you go to higher transmission, you get artifacts that can be difficult to fix in the layout."

Other techniques besides phase shifting also play into resolution enhancement. "Many of the emerging concepts take this pattern partitioning concept further with X and Y patterns selectively printed using dipole illumination," West said. "The use of subresolution assist features (SRAFs) is a given as lithographers work to expand limited process windows while managing variable pattern densities or pitches, especially for logic."

It is the combination of so many of these techniques that creates the real solutions for advanced technology nodes, West noted. "Many of the concepts are merged to derive the best benefits. Complementary PSMs are fully dependent on model-based OPC (optical proximity correction). OPC requirements drive the use of SRAFs on embedded PSMs."

The cornerstone of the new PSMs has been the ability to quickly convert designs from their plain vanilla format into their phase-shift applications, West said. "The design part of each one of those strategies is a critical developmental area in and of itself. Historically, as I look back, one of the things that paced the introduction of alternating aperture phase-shift masks was the design structure, more than the ability to fabricate those."

EDA software tools have evolved and matured, and IC designers and software developers are beginning to work more closely with maskmakers and lithographers. Photronics and ASML MaskTools announced a strategic alliance in December to develop a production-ready, maskmaking infrastructure for CPL technology. Photronics will help to develop a CPL manufacturing flow that's optimized for its process. In return, it will receive an R&D license for CPL patents related to maskmaking, wafer imaging and software implementation.

With increasing complexity and the abundance of mask flavors being explored, masks — and those buying them — will undoubtedly take a hit in terms of cost, yield and defectivity. "The more complex the mask and the less demand there is for it, the more expensive it will be," Progler said. "The fewer masks that are built, the lower the yield will be."

"Clearly, complexity from newer nodes, increasingly interdependent RET solutions, and the fact that the mask is an active element in image formation will continue to drive costs," West said. "Costs for tools, R&D to reach higher standards, and the constant demand for service where capacity and capability join are the key drivers."

Defects will also continue to be more critical. "They're a primary part of both our technology planning and our business planning, out of necessity," West said. "Some level of defectivity can be absorbed in the wafer, and in fact it's almost expected. And yet the reticle, because it's the master, it really has to be held to a higher standard."

"Resolution is the No. 1 requirement," West added, noting that DPI's R&D center is working on designs with <50 nm reticle features. "Each new node will require smaller, finer, more accurate features. Unless the reduction ratio of the scanners change, we plan for lots of pressure on fidelity factors to support the extremely small error budgets — sharing is not always easy. These incremental challenges are on a steep learning curve given the expected timing to rapid solutions. To stay with a common wavelength for more nodes, even with the innovation of immersion, is a fortunate feat for the industry to leverage 193 nm know-how and infrastructure."

Nobody is saying that immersion lithography will be easy. But, given the alternatives (157 nm lithography, for example), it's a calculated risk that most players seem eager to take. "If the alternative's really impossible, then you stick with the almost impossible," said Mordy Rothschild, leader of the Submicrometer Technology Group at MIT Lincoln Laboratory (Lexington, Mass.). The Lincoln Lab group, the first to work on 193 nm and then 157 nm lithography, started expanding into immersion lithography for both wavelengths about two or three years ago, Rothschild said.

Despite misgivings about bubbles or possible water-photoresist interactions, for example, immersion lithography is looking like the safest bet. The technique places a liquid (most likely water, in the case of 193 nm) between the final projection lens and the wafer. Because of the higher refractive index (1.44 vs. 1), projection lenses can be made with a higher NA, effectively reducing the 193 nm wavelength to 134 nm and improving resolution.

There are a few general schemes for introducing the water to the system: shower designs, in which the water is deposited on the wafer, then taken away; bathtub designs, in which the entire wafer is immersed in a bath that moves with the wafer stage; and swimming pool designs, where the entire stage is under water. The industry is gravitating toward the shower concept, and in fact both Nikon's and ASML's first designs are variations on that theme. Both companies have announced agreements with TEL for immersion development, with TEL providing its expertise in liquid delivery, for one.

Nikon's concept is shown roughly in Figure 4 . "Basically, the scheme is you have a nozzle on one side, you put water under the lens, and you suck it up from the other side," said Gene Fuller, principal engineer at Nikon Precision Inc. (Belmont, Calif.). Although there is more to it than that, tool manufacturers are holding the details close to their chests. Fuller conceded, though, "It's not going to look strange or unusual to anybody."

4. A shower system for immersion lithography delivers the liquid onto the wafer from one side, then takes it away from the other. (Source: Nikon)

Although the bathtub concept makes the edges of the wafer easier to expose, it adds too much weight to the stage, which hurts throughput. "Just adding 1-2 liters of water for immersion causes you to lose productivity," said Boudewijn Sluijk, director of product marketing for ASML (Veldhoven, Netherlands). "So now we're doing a shower head concept." This scheme still allows the wafer edges to be exposed, he added. "The most challenging thing was to get the water containment under the lens."

Sluijk declined to give specific details of ASML's liquid delivery system. "Let me just say that, if you put a drop of water on the table, you can move that drop of water by blowing on it," he said. "So you can imagine that there are ways to move that water across a wafer."

However the water is introduced, one major concern with immersion lithography has been the possible existence of bubbles, which can pose a major threat to image integrity. In principle, bubbles can be generated from two or three different sources, Rothschild said. Under normal conditions, water has air in it, and those air molecules could form bubbles. A simple solution is to degas the water, and there are commercially available systems to do this.

"But there are still risks," Rothschild said. "What happens when the water is in contact with the wafer?" As the water fills the gap where the ambient air was previously, there is potential for air to get trapped at the interface. "I think work needs to still be done and will be done to make sure this doesn't occur."

"Finally...forget about air," Rothschild continued. "Bubbles could be made from outgassing from the photoresist." Every photoresist outgasses some, he noted, and researchers have been studying its effects on the lens. "Liquid would actually help protect the lens, but, depending on the properties and the rate at which the outgassing occurs, it may form bubbles."

Another concern related to the water is what it does to the materials that it's in contact with — namely, the last element of the optical train, and the photoresist. Various groups have been studying what happens when the water's in contact with the photoresist, with concerns about contamination. At first glance, at least, it appears that there is not much interaction, Rothschild said. "With photoresists, there may be a little bit less of a concern at this point only because the photoresist is not exposed to the water for too long."

The lens, on the other hand, faces a continuous flow of water, which could cause damage. "Calcium fluoride is not really hydroscopic, but it can be sensitive enough to water over time, especially if you radiate that lens with a laser," Rothschild said. "Calcium fluoride gets roughened by water, even within a matter of hours." To combat this issue, several organizations are working on a protective layer solution, he added.

Researchers continue to assure the industry that immersion lithography does not appear to have any showstoppers. But what "showstopper" typically means is that it violates conservation of energy, Fuller noted. "The real showstoppers are all the niggling things that don't get done, or the money doesn't come."

"If there's anything I would worry about, it's more like the unknown," Rothschild said. "We don't yet know what we really don't know." Although he is confident that immersion lithography will proceed on schedule, he said that he'd feel better when the industry has had a chance to run tests on moving systems, due out later this year. "The truth is, there are no working step-and-scan systems," he said, noting that the small number of test systems being used now are using static water. With working step-and-scan systems, researchers should be able to find more definitive answers.

Despite the possible unknowns, immersion lithography falls more in the arena of holding on to the knowns. A key attraction is that it can be built on existing platforms and infrastructures. "Clearly listing in the positives would be the continuity in the materials strategies," DPI's West said, including attenuated embedded materials, soft pellicles, and the evolution of resists rather than the invention of new materials. "Immersion lithography keeps the infrastructure in place and intact."

As far as the system itself is concerned, the wafer handling must necessarily change to introduce liquid to the system, and software will need to be adapted. But much of the system remains unchanged, including most of the illumination system, and the reticle stage. The first evaluation systems to come from the tool manufacturers will look almost identical to their latest dry releases. "All three of the major companies are working along the same kind of idea. And that's to use their existing base to build an immersion tool," Fuller said, noting that Nikon's first immersion tool is based on its NSR-S307E, an ArF system with an NA of 0.85, which was recently released and had its first shipment made.

The three leading lithography tool suppliers — Nikon, ASML and Canon — said last summer that they would let the industry know by December or January whether they would pursue the production of immersion tools. In early December, both Nikon and ASML announced their commitment, as well as their schedules for shipping their pre-production and production systems. At publication time, Canon had not yet made an official declaration of its intent, but began reporting in November that it was "seriously evaluating" 193 nm immersion lithography, considering it a potential successor to its super high-NA dry technology.

Nikon expects to complete its engineering evaluation model, based on the S307E, in the second half of this year, with a pre-production model scheduled for 2005. The pre-production version will be based on the S307E's successor, with an NA of 0.92. Production tools, with an NA>1.0, should begin selling in 2006.

ASML said it expects to deliver its first pre-production immersion tool in the third quarter of this year. Based on its recently released Twinscan XT:1250, it will also have an NA of 0.85. Coinciding with this announcement was ASML's announcement that it had received its first order for the XT:1250i, from foundry Taiwan Semiconductor Manufacturing Co. (TSMC, Hsinchu, Taiwan).

In late January, IMEC (Leuven, Belgium) announced that it would extend its collaboration with ASML to immersion lithography, with plans to install ASML's XT:1250i in the fourth quarter of this year. IMEC also said that it will launch an industrial affiliation program (IIAP) on 193 nm immersion lithography that will run in parallel with its 157 nm lithography IIAP.

A big reason immersion is so appealing is because it could be put into production in a fairly short timeframe, Fuller noted. It historically takes five to six years from the first full-field test tool before a new technology or wavelength really takes off, he said. "Immersion can come on a lot quicker than that six years. Assuming 2004 is the first year we really get early prototype tools...we'll be seeing fairly substantial production use in three years at the most."

The tool suppliers are working first on systems with existing NA levels (0.85). Rather than go immediately for improved resolution, they are taking advantage of immersion's greater depth of focus (DOF), thereby relaxing the k₁ factor and the need for such complex masks. Figure 5 demonstrates ASML's early immersion results, showing improved DOF for wet systems.

5. Keeping NA the same for immersion systems as it is for dry systems (in this case 0.75) does not improve resolution, but does increase the depth of focus, which relaxes the k1 factor. (Source: ASML)

"This is in fact what TSMC has been interested in for some time, not to get the NA up to 1.3," Nikon's Fuller said. As a foundry, where many designs are being run at low volumes, TSMC is undoubtedly feeling the strain of skyrocketing mask costs caused by increasing mask complexity. No wonder, then, that TSMC has been such a leading proponent of immersion lithography.

Immersion lithography is a very good trend for maskmakers and chipmakers alike, Progler said. "We don't want to make all our masks to be impossible to build," he said, noting that there needs to be a healthy mix of simple and complex masks to keep the industry moving forward. "I think it's good for us if that healthy mix continues. In absence of immersion coming on, that healthy mix is hard to see."

Eventually, though, the complex masks will prevail. The most advanced IC makers will do what they've typically done, Progler said. They'll adopt immersion lithography, then still try to throw all the RETs they can at it, trying to separate themselves from the pack. Ultimately, strong phase shifting will be just as prevalent an extension of 193 immersion as it will be to extend 193 dry.

For the latest updates on immersion lithography, join us for a FREE live webcast March 10, 12-1 p.m. CST. The session will include presentations from Burn Lin of TSMC, Mordy Rothschild of MIT Lincoln Lab, and others to be announced. These will be followed by a Q&A period, moderated by SI's Managing Editor Aaron Hand. Register today at www.semiconductor.net/webcasts .