In early April the European lithography tool specialist signed a licensing agreement with Integrated Process Systems (ISP) of South Korea. Under the terms of the pact, ASML's thermal systems division will sell ISP's ALD-based equipment to customers outside of Korea. ASML has scheduled a new platform introduction for the technique in 2003.

FILL THE GAP: Dielectrics will drive ALD's commercial success. This forecast predicts a $1-billion global market for four segments by 2006. SOURCE: GENUS; ILLUSTRATIONS BY JAMES SCHLESINGER

Genus in April launched the first ALD product "roadmap" during Semicon Europa. The deposition tool manufacturer, which has been leading the charge to commercialize the technology, predicts that ALD will begin to take off in the second half of 2002.

Meanwhile, ASM International (ASMI), ASML's former parent firm, continues to explore its options in the arena following its 1999 acquisition of Microchemistry, a Finland-based ALD pioneer. Certainly not forgotten, Applied Materials jumped into the market last summer with the introduction of a retrofitted ALD chamber for both metal and dielectric films. Applied's first applications include a tungsten nucleation layer for advanced contacts and a titanium nitride film for liner or electrode uses on tantalum pentoxide–based capacitors.

It's easy to see why ALD has drawn such interest. Although the throughput numbers are still on the modest side, the process benefits of atomic layer deposition technology more than compensate. The technique, which deposits atoms on a wafer a single layer at a time, is pushing the market for ALD-based systems toward the kind of sales growth last seen two years ago.

Werner Rust, Genus's executive vice president of sales and marketing, asserts that traditional deposition methods fail to meet the needs of next-generation device manufacturing in several key areas. These include yields, conformality, process cleanliness, and process extendibility.

Rust says the compound annual growth rate for ALD between 2001 and 2006 will reach 107%. Systems featuring the technology have the potential to grow from approximately $28 million in 2001 to $1 billion in 2006, according to a forecast by VLSI Research that has since been revised upward. Currently, the global market stands at $80 million. The market research firm has said the 2006 figure is based on a combined market value for TFH gap dielectrics, DRAM capacitor dielectrics, and logic gate dielectrics.

The prime driver is the ability of ALD to extend deposition processes to the 90-nm node and below, asserts Rust. The technique should particularly entice DRAM manufacturers because capacitor dielectric films are unable to perform as well as before. Genus has the systems in operation at eight beta sites, says Tom Seidel, the company's CTO. In addition to placing tools at four of the top five DRAM manufacturers, in December the supplier shipped its first 300-mm system to Selete, the Japanese semiconductor research consortium, for its Asuka Project. It's the first 300-mm tool shipped in the industry, Seidel says.

Genus's first successes with the technique have come in the thin-film head arena. Customers such as Read-Rite use ALD alumina thin films for reader gap applications. In a technical paper Read-Rite engineers note that ALD is based on separating precursor gases in proper sequence. The team compared traditional PVD alumina with ALD alumina "at both the film and device levels."

Each cycle on the ALD deposition system consisted of four steps at a deposition rate of 0.95 to 0.98 Å per cycle. A low-temperature heating process of 200°C was used to keep from damaging the reader sensor by overheating it. The substrate was not heated during PVD.

The report points out that the PVD films show a dramatic drop in breakdown voltage (BDV) as the film thickness reaches 300 Å and below. The cause is an increase in pinhole density at the lower thickness. Alumina thin films deposited through ALD, however, show a higher BDV of less than 9.5 MV/cm throughout the film thickness range of 75 to 500 Å. All ALD alumina films were free of pinholes, and both thickness uniformity and control were "excellent," the authors conclude.

HEADROOM: ALD volume TFH production offers high breakdown voltage, pinhole-free films, and "excellent" uniformity compared with PVD, Genus says. SOURCE: VLSI RESEARCH; ILLUSTRATION BY JAMES SCHLESINGER

Though some systems are at thin-film head sites, Seidel notes that the beta site programs—"somewhere between a pilot line and real production"— are valuable sources of technological feedback that will eventually help Genus tailor its tools for chip manufacturing. "It's a good environment because it gives us an entrée to the production of semiconductors with good manufacturing metrics. Other customers are doing capacitors, while still others customers are in the gate area.

"The ability to get metrics back in the manufacturing environment. There's no substitute for that," the Genus CTO continues. "You end up getting reports weekly or monthly. You're able to gauge your parameters...." At a Semicon Europa news conference, Seidel said complementary work in metrology is needed to meet the challenges thrown up by the need to measure fractions of an angstrom. He noted that reflectance metrology probably would be required at minimum.

Seidel says a paper presented by Genus in May at the Electrochemical Society Forum in Philadelphia contained "a pretty rigorous analysis" examining variances in thicknesses caused by changes in chamber pressure, timing, or purge. "The overall control is better than 1%," Seidel emphasizes. "That's the inherent capability of the tool." He also notes that Genus's strong CVD technology background means the deposition platforms are battle tested, if you will. "These were shaken out in the CVD arena; maybe only 20% of the components are different for ALD."

PROCESS EXTENDIBILITY: DRAM micrograph at left shows cross section of ~60:1-deep trench capacitor. SEM images at right show Al2O3 thicknesses proving 100% step coverage.

ASML's licensing agreement puts it in the running rather quickly and repositions the thermal division, acquired from Silicon Valley Group, "in a very exciting market," says Jeffrey Kowalski, the division's president. ISP is the only company "that has a tool in production in Asia," and ASML is offering the tool "to our customer base worldwide," the executive notes.

Kowalski says ALD can be adapted to "numerous types of reactors. Whether it's a shower head or a batch reactor or a plasma-enhanced reactor, it's all pretty much the same in ALD." The executive claims that ASML's reactor's unique design gives it a competitive edge in terms of cycle time, productivity, and conformality.

Neither Kowalski nor Seidel seem too put out about throughput rates. "The throughput of these tools is going to have a lot to do with vacuum design and gas delivery kinetics," emphasizes the former ASML executive. "If you have an MOCVD or CVD process, that's two to three minutes, and you might not care what happens in the first three to five seconds." He notes that the proprietary gas delivery and vacuum technology will make ASML's systems competitive.

Seidel says Genus's systems currently operate at a throughput rate of 10 to 15 wafers per hour. "The important issue is that we're simultaneously meeting all the requirements for good chip operations...and high tool availability. You can always add...CVD chemistry to the ALD process, and if you do that you can probably drive throughput up a little bit, but then you're going to hit a wall getting into particle and materials issues. Basically, I think all ALD practitioners will eventually try to add a certain amount of CVD [process] and say, 'Who cares if it gives a reasonable performance, we'll take every extra wafer we can.'"

The good news, according to Seidel, is that a throughput of 10 wafers per hour is "very respectable" for CVD processes. "Your tungsten and your tungsten silicide or your single-wafer modules for comparable wafers have comparable throughputs. So as long as you're not talking about having a huge number of those films in each wafer, your tool set is going to be a modest number of systems. It's always an issue because of the [nature] of thin films."

Risto Puhakka, vice president of operations at VLSI Research, says ALD "gives you the best step coverage you can have." The throughput "depends on how fast you change these gases in the chamber. You can start to cheat on it because you're doing it five times faster." He maintains that hourly output "needs to get a little bit better at 20 wafers per hour," adding that once customers start processing substrates at an hourly rate of 40 to 50, they will "want to get to 100 wafers per hour."

He admits that VLSI Research is in an odd situation because it underestimated the 2001 market share by approximately $24 million. "We estimated the ALD market would be $26.6 million, and it turned out to be $50 million." Puhakka says the updated forecast calls for a "really big jump" in 2004 to $350 million. "Then you can definitely say there's a significant market."

Asked whether the downturn has affected market momentum, Seidel replies, "The answer is yes and no. I hate to give you a double answer. It's not because there's a market downturn. There are resources looking at development-type activities.... If you can make a buck today you don't spend it on development. There are a lot more eyes looking at 'what I can do to make a buck tomorrow.' So the interest level hasn't slowed at all.