INDUSTRY NEWS

Liquid lithography technology waits for defect data to surface

What's the only thing that can slow down the fastest technological development the semiconductor industry has seen in the past 20 years?

"Defect sources are really the key thing here," asserts Andrew Grenville. "Everybody's concerned about defect sources."

"Defects are number one," says Will Conley.

The two International Sematech assignees are, of course, speaking about immersion lithography (IML), the technology that places water between lens and wafer, thus extending optical lithography's lifetime once again. With a lens designed to offer a numerical aperture greater than one, the technique's high resolution provides smaller features than basic projection lithography at relatively low cost.

ALL WET: Cutaway view of an immersion lithography system. IMAGE COURTESY OF NIKON

Grenville, an assignee from Intel, is the program manager for immersion and 157-nm lithography at the Texas consortium. Conley is 157-nm program manager at the Dan Noble Center of Motorola SPS, spun off and rechristened in mid-February as Freescale Semiconductor. Both spoke in January at a well-attended Sematech workshop on the topic in Los Angeles.

In his presentation on fluids, Grenville pointed out that theoretical and experimental studies show that IML faces no proverbial showstoppers—at present. Nearly all the ducks are in a row for 193-nm-wavelength immersion lithography to take the stage for advanced device manufacturing around 2007. Summarizing basic research, he told conference participants that treated water is the most suitable immersion fluid for use with ArF optical systems, small bubbles have a limited life span, a liquid-film flow method is a promising immersion system, and the technique is ready for evaluation by photoresist suppliers.

In a later interview Grenville reiterated his main theme. "At this point, no showstoppers have been identified for which engineering solutions have not been envisioned. Nanobubbles could be a source of defects that could end up being a real problem. . . but, A, we're looking into figuring out how they would form, and, B, we're figuring out how to investigate them." He adds that another potential problem, thermal aberrations, "appears to be manageable. Suppliers have come up with solutions for that."

Grenville insists that the infrastructure will be there to support the technology. "The message from exposure-tool suppliers and resist suppliers is that they will be ready to meet the timeline. That's assuming, of course, that there aren't any big problems that are uncovered in the meantime. That's part of their strategy, to go ahead with prototype machines later this year and lead into production-ready machines in 2006."

However, until fabs start using the technique, we don't know what we don't know—as Donald Rumsfeld said in another context. "The problem is, until you actually have the scanners operating in a yielding fab, where you can basically do split lots to be able to see whether they generate additional effects, that's really the only way, at the end of the day, that this is going to work," Grenville says.

Conley agrees. Everything looks good "until the tools get in the fab and guys can compare them with their existing dry process where they have lots of yield data," Conley says. "The results will determine whether immersion continues."

Michael Lercel was impressed by what he heard at the conference. "What came out of the immersion workshop were very exciting technical results in regard to imaging performance. It works as advertised," says Lercel, the manager of lithography technology R&D at IBM Microelectronics in Hopewell Junction, NY.

But, he adds quickly, "What I didn't see were any defect levels. Nobody has the data yet. Those are the most difficult types of data to collect. What that requires is having a semimature platform, a cleanroom environment, and enough wafers for data, because this is always a statistical process. It's not surprising that it's the last thing to become available."

Lercel attributes the "amazing amount of progress" since the industry's first IML meeting in December 2002 to the realization that the "people got caught by surprise. It's much more achievable than everybody anticipated."

Still, the IBM manager insists: "If we see some defect data, that would give us more encouragement. The other key aspect, of course, is making sure you don't get them in the first place."

From an imaging standpoint, Lercel freely acknowledges "that this really works, that people have the confidence to run water on and off the wafer and print the wafer. That's been a huge change over the past 12 months."

A cautiously optimistic Moshe Preil, director of applications development for KLA-Tencor's Rapid division, also cites defects as his major concern. "Anyone who has ever done any wet processing on wafers—be it etch, CMP, or resist processing—knows that while such processes can be done cleanly, they need constant care and attention to keep them that way."

Even though the industry has overcome "many seemingly formidable obstacles. . . in remarkably quick order," Preil asserts that lithography equipment manufacturers have not always had process cleanliness uppermost in mind. "Historically, exposure tool suppliers have always treated defects as an afterthought. If the tools had high particle levels in the factory, it was simply assumed that they would be cleaner when they were installed in a customer's fab with cleaner air. Many suppliers never even measured defect levels until the tool was installed at the customer site."

Immersion lithography, though, raises the specter of "process-induced defects," Preil emphasizes. "If the tool is not engineered to run cleanly, no amount of filtering in the customer fab is going to make it clean. The process—not just the materials, but the way the fluid is dispensed onto and removed from the wafer, the way the wafer and the optics interact with the moving air-liquid interface, and the way the fluid is handled at the wafer edges—all [of these] have to be engineered with careful attention to defect-creation mechanisms from the outset.

"So far, no one has shown any data on full-wafer defect levels, so this has to remain a concern," Preil concludes.

Burn Lin, TSMC's technology coordinator for 193-nm immersion, insists that the industry is betting on 193-nm immersion lithography for the next several device generations. The technology is expected to work at both the 65- and 45-nm process nodes. ASML reports that it will ship a manufacturing model, the 1250i, to the Taiwan-based foundry by the end of 2004. Meanwhile, Canon reportedly made two evaluation tools in 2003.

Regarding defects and yields, Soichi Owa, an engineering manager at Nikon, told attendees that the company was waiting to complete its engineering evaluation tool sometime in the third quarter of this year. The full-field immersion exposure system has an 0.85 numerical aperture (NA), Owa says. All three major exposure-tool manufacturers are expected to have commercial 193-nm immersion-ready scanners available by 2006. In addition to Nikon, the scanners from ASML and Canon will have so-called hyper-NA lenses with numerical apertures greater than 1.0.

In addition to defect analysis, Grenville told workshop participants that remaining issues are throughput, water qualification, contamination control, water-boundary coating, and immersion resist. Finding the most effective type of water is a key area that has occupied industry researchers such as Conley and Bruce Smith of the Rochester Institute of Technology. Smith says water doped with phosphates or sulfates may create liquid with a higher refractive index than purified water. At Sematech, Conley and others are using organic materials in the hope of raising the refraction index of water to 1.53. That improvement would further increase the depth of focus.

"Over the years we've spec'd the water content in the photoresist solution, and now we are dipping the entire film in water. So understanding the polymer hydrophobicity or hydrophilicity and engineering it for the best imaging performance is in progress," Conley says. "Moreover, understanding the water solubility of additives such as quenchers and photoacid generators is required, and again their impact on imaging."

In his workshop presentation, Grenville referred to the potential for new types of defects, such as misplaced water droplets or entrainment from water jets. "The more difficult part of the problem right now is that there doesn't even appear to be a defect mechanism we can put our finger on," he said in a post-workshop interview. "It's proving that there are no defect mechanisms," and that's difficult, he points out, because you're trying to prove a negative.

Good-naturedly, Conley remarks, "It doesn't take a rocket scientist to answer all this stuff. It's all engineering. There's no invention involved." Later, he adds, "I'm making it sound simplistic; it's not that simplistic. It's all about just turning the crank and getting the cycle of learning going, understanding what materials can and can't do."

In the end only one major issue matters, Conley insists. "What's the quality of the image? How does it look on the substrate?" —JC