INDUSTRY NEWS

Atomic-layer deposition maps out its place in the nanoscale world

David Back's employer required little convincing when it came time to sign on the dotted line for a major atomic-layer deposition (ALD) project. Atomic-layer deposition "is an essential part of our DRAM technology roadmap," he states flatly.

ENABLING THE ROADMAP: ALD research tool at the University of Albany-SUNY's Center of Excellence in Nanotechnology. PHOTO BY VINNY GIORDANO, COURTESY OF ALBANY-SUNY'S CENTER OF EXCELLENCE IN NANOTECHNOLOGY

Back will oversee Infineon Technologies' research work with deposition tool manufacturer Genus and the University of Albany-SUNY in the three-year, $12-million effort to develop processes for metal electrode and high-k dielectric materials. "We see ALD as really important for the roadmap. There are a number of enabling technologies, and this is one of them."

Launched in late January, the partnership hosted by Albany-SUNY's Center of Excellence in Nanoelectronics is just one of many recent business and research developments that reflect the torrid pace of ALD's development. Sales of deposition tools in this market segment are expected to increase from approximately $100 million to around $1.35 billion in 2008, according to VLSI Research.

"I think there are a half-dozen or so suppliers now, and that only validates the very high potential business enterprise there," notes Tom Seidel, chief technical officer for Genus. "VLSI continues to forecast total equipment revenues approaching a billion dollars toward the end of the decade. . . . There can be some push-outs; there can be some pull-ins, but there's no question that the technology's time has come."

A roundup of other recent ALD-related developments shows the technology's increasing momentum:

• Novellus Systems purchased Angstron Systems. The OEM plans to integrate Angstron's ion-induced ALD technology in its Inova xT platform for copper barrier and seed deposition.

• Researchers at IMEC have used materials other than the standard polysilicon to make high-k dielectrics and metal gates of equivalent oxide thickness values smaller than one nanometer. Working with International Sematech, the Leuven, Belgium–based consortium used titanium nitride or tantalum nitride gates and hafnium oxide dielectric material for metal gates with an equivalent-oxide thickness of >0.8 nm. The researchers employed atomic-layer CVD to deposit the hafnium oxide. The high-k metal-gate devices demonstrated high conductance and low leakage rates, IMEC says.

• ASM International has signed an agreement to purchase Genitech, a South Korean supplier of plasma-enhanced ALD systems for metal and dielectric deposition. The acquisition adds Genitech's Superfill CVD tool for catalytic-enhanced MOCVD copper layers to ASM's ALD-related technology portfolio.

• Tegal bought the assets of Simplus Systems, a small vendor specializing in nanolayer deposition (NLD) and MOCVD technology. The Simplus NLD cluster tool is designed for manufacturing DRAM and logic devices in barrier, copper seed, and high-k dielectric processes. Tegal claims NLD technology has a faster throughput than its ALD counterparts.

• KLA-Tencor's electron-beam metrology system, the MetriX 100, is said to be the first noncontact tool for measuring the composition and thickness of metal film in production settings, including ultrathin barriers processed using ALD. The in-line tool slots into 90- and 65-nm process technologies. Its multiple x-ray detectors and tunable E-beam column will help chipmakers to control advanced copper processes, according to the supplier.

ALD's acceptance rate varies, depending on the device in production. Giant magnetoresistive (GMR) thin-film head makers were early adopters of the deposition technology, while DRAM capacitor applications have been the latest to gain commercial traction. Genus says it recently sold its first StrataGem ALD system to a large Taiwanese foundry, where it will be used to make DRAMs at the 110-, 90-, and 70-nm technology nodes. The unnamed chipmaker is the third company to buy the vendor's ALD tools for advanced memory processing.

"The status of the market today is that the DRAM guys are probably moving faster than the logic guys in terms of embracing ALD," notes Subrata Chatterji, vice president of the ALD business unit for Aviza Technology. Samsung has been using ALD tools for 110-nm devices, while Infineon plans to start employing ALD at the 90-nm node sometime in 2005, he adds.

Aviza received a U.S. patent in April for a multilayer composite dielectric film concept. The film combines high-k metal oxide and metal silicate to improve compatibility with polyelectrode and silicon substrates for transistors at the 65-nm node.

The industry continues to mull over the matter of materials selection. Some film winners have appeared though, asserts Dana Tribula. The managing director of group marketing for Applied Materials' copper, PVD, and integrated systems calls the matter of materials "not an open question. Clearly, for contact-metal lines, tungsten is a no-brainer. For the barrier layer, tantalum nitride is the primary choice at the 65-nm technology node because it's the closest thing to the same material used on the PVD side. There's been a lot of learning about ALD films in terms of how to integrate them and what their barrier properties are like.

"The barrier-seed combination is fairly key in terms of determining reliability, and electromigration and stress migration [are key] on the interconnect side. That's been the biggest challenge in finalizing the integration scheme."

Asked about the level of acceptance, Tribula says Applied has shipped 12 of its Endura or Endura 2 systems, the equipment manufacturer's two ALD mainstays. "It's pretty much everywhere for 65-nm technology evaluation. [The technology] is essentially competitive with the existing PVD process."

It's the semiconductor industry's nature "to stick with the old" and wait for the new processes to prove themselves, Tribula points out. To some extent, the industry can "design around . . . this line-resistance issue inherent in the PVD process." The work-around eliminates the need for the ALD process, she says. Nevertheless, Applied is "seeing ALD deposition on the customer roadmaps. That development and others, such as the Novellus acquisition of Angstron, are further positive signs. "That's another validation of the fact that this is something that's coming."

Seidel of Genus says some film choices are still under consideration. "Applied says the films are settled: In contact metals, it's tungsten; in barriers, it's tantalum nitride. In barrier they [Applied] know the business very well, and in interconnect they [also] know the business very well. Those are the materials of choice."

However, he insists, "people are still interested in tungsten nitride and in alternatives to tantalum nitride." The overall integration level of tantalum nitride, which IBM began using several years ago to the point it became a de facto standard, accounts for its widespread acceptance. Chipmakers are understandably reluctant to change materials that are well integrated in the process, he notes.

Risto Puhakka, vice president of operations at VLSI Research, says implementing ALD in production settings still raises yield issues. Both high-k dielectric and barrier applications require low defect densities. "The other big issue is how this new material is going to behave. There's a reason [the industry] has focused on silicon dioxide. So there are a lot of challenges."

How much will those challenges affect the rate of ALD's growth? Puhakka says the market in 2003 grew $50 million from the previous year to $100 million, prompting "quite a bit of debate" and raising the question of whether the rate will double again this year. The answer, he says, "is hugely dependent on how ALD is penetrating in different applications." The technology has made good headway in the capacitor business, he says. However, ALD has yet to take off in device processes such as barrier metal deposition or high-k dielectrics. In the case of the latter, the technological development hasn't reached its cruising altitude. "People are commonly expecting high-k dielectrics at the 45-nm node, so it's a little bit early to be ramping up for that yet."

Will developments in metrology keep pace with ALD's ascent? Murali Narasimhan, senior director of marketing for the films and surface technology division of KLA-Tencor, sees both pluses and minuses. The standard practice for measuring thin films—optical metrology—may not work with ALD films. The films "tend to be transparent, even if they are metallic, because they are so thin." Ellipsometry is one option, but the films must be approximately 100 Å thick.

"The disadvantage of that technique, if you have an unknown film with no idea of the refractive index, is that you can have errors in the thickness measurement," explains Narasimhan. "You have to independently measure the refractive index with a research-grade ellipsometer."

On the other hand, because "so many variables can change the composition and refractive index . . . optical measurement becomes a little subjective." Overcoming this drawback requires multiple measurements grouped together. As a fairly new method, the E-beam technique that KLA-Tencor is marketing as a production-ready system may currently trail behind the measurement precision of established optical techniques. In addition, with the E-beam technology, "you could have some modification of the film."

However, the overwhelming E-beam advantage, Narasimhan points out, "is that you get the composition and thickness independently." This is important because the thickness composition "often determines the electrical functionality of the film." In ALD, tantalum nitride barriers are judged on how well they keep copper at bay, and that "depends on the composition, so you want to monitor that." Other examples are DRAM capacitor dielectric materials, such as hafnium oxide or aluminum oxide. The critical point in this instance is the storage capacity of the DRAM capacitor, "which depends on the dielectric constant of the film, which in turn depends on its composition." This in turn requires tight monitoring of the film thickness and composition.

With optical techniques, "you're looking at changes in the optical properties of materials as a result of the change in composition. That's a big second-order effect. In the E-beam x-ray technique, you are directly stimulating atoms and counting x-rays of the number of atoms of a given species. Fabs can choose to do one or the other. It's a little early to tell which is going to win.

"We see changes in the composition of ALD films as a function of substrate type, film thickness, and process conditions, and this may require in-line monitoring of composition."

KLA-Tencor recommends that customers consider using in-line monitoring in a high-throughput, low-cost-of-ownership model with the optical technique and a "higher-sampling engineering analysis process development" option with the E-beam metrology option, according to Narasimhan. A 20-µm E-beam spot, advanced pattern recognition, and fab automation are features that allow the system to be used for production-line monitoring of ALD, if required, the company says.

Throughput will play a critical role in ALD's eventual level of acceptance. Genus's Seidel says 20 wafers per hour is standard. "It's possible to use ALD for films as thick as several hundred angstroms and still have decent throughputs. A lot of processes are run with less than 10 wafers per hour and their benefits are sufficient. . . .We expect to see productivity pick up and not be an issue a year or two from now."

Tribula of Applied Materials says the Endura machines have achieved a rate of 30 wafers per hour on 10-Å film on processes "limited for this particular chamber." Because ALD is new, "it just takes a while to prove to the customer that it does work—that there's enough comfort level from a process integration standpoint."

As a DRAM producer, Infineon is particularly cost conscious, Back says, hence, the semiconductor manufacturer's decision to participate in the partnership with Genus and Albany-SUNY. "Every time we introduce a new material system there's always a tremendous amount of work and know-how required. What we want to do is expand the overall base of know-how. Our customers want to see a functioning product without incurring additional qualification and risk."—JC

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit