How To Reach C&ENACS Membership Number
Visit SGI


June 23, 2003
Volume 81, Number 25
CENEAR 81 25 pp. 21-26
ISSN 0009-2347

Page 1 | 2 | 3 | Next Page

Semiconductor materials companies try cooperation of many kinds to solve technological and financial challenges


WIDE VIEW ATMI spends big to develop new materials at its Danbury, Conn., R&D center, then tries to earn its investment back by pricing its products based on their value to customers.
Chemical companies supplying the semiconductor industry have a dual challenge. They must research and develop the new materials that allow firms like Intel and IBM to continue making smaller and faster computer chips. And they must turn a profit as well.

Success at these challenges is not coming easily these days. A fundamental change in the way semiconductors are made is precipitating an unprecedented need for new materials. At the same time, cost-cutting by semiconductor manufacturers is making it more and more difficult for chemical companies to recoup the price of developing these materials.

This cost-cutting became relentless in 2001, when the semiconductor industry crashed after years of double-digit growth. According to the trade group Semiconductor Equipment & Materials International, worldwide sales of silicon wafers, the industry's basic building block, jumped from $5.9 billion in 1999 to $7.5 billion in 2000, then collapsed back to $5.3 billion in 2001. Sales inched up to $5.6 billion in 2002. Wafer shipments in first-quarter 2003 were up 16% over the corresponding 2002 quarter, SEMI says, but only 4% over the preceding quarter.

Chemical companies are responding to the harsher environment by stepping up the pace of cooperation with their semiconductor customers, with makers of chip manufacturing equipment, and even on occasion with other chemical suppliers. They are finding that, while such cooperation may win them the admiration and the business of the semiconductor industry, it doesn't ensure them profits.

The computer industry abides by Moore's law, the 1965 prediction by Intel founder Gordon E. Moore that the number of transistors crammed on a chip would double every 12 months. Moore subsequently refined his prediction to every 24 months, and it has held true ever since.

THE INDUSTRY'S push to keep up with the law with each generation of chip has always led to manufacturing and materials changes, but in recent years these changes have been profound. "There were many generations over which one or maybe two new materials were introduced," says Pam Mattimore, president of Air Products Electronic Chemicals. "Now, we're talking about multiple materials, and they are all piling up on top of each other."

Two fundamental drivers are at the root of many of these changes. One is new generations of photolithography and accompanying photoresist polymers, which together allow chip makers to draw circuit lines on silicon wafers. The other is the change in the metal used to make these circuits from aluminum to copper.

Thomas Baum, vice president of R&D at materials supplier ATMI, says basic drivers like these can ripple through the materials supply world. For example, the shift to copper wiring--now found in only the most advanced chips--greatly reduces power consumption while increasing circuit speed. But copper requires new materials to protect one layer of circuitry from another; hence the industry's shift from titanium nitride to tantalum nitride barrier films, Baum says.

P. Jerry Coder, president of EKC Technology, which has been part of DuPont since November 2002, adds that layers of copper wiring must be smoothed with new families of chemical mechanical planarization (CMP) slurries that are completely different from those used to smooth aluminum.

And the closely spaced copper lines must be insulated from each other with new materials; hence the industry's shift over the past few years from silicon dioxide-based dielectrics to, first, fluorinated silicate glass (FSG) and, now, materials that have a lower dielectric constant, or "k" value. These range from carbon-doped SiO2 materials that are applied by chemical vapor deposition to polymeric materials such as Dow Chemical's SiLK polyphenylene that are applied by spin-on techniques.

"Never before have so many new materials been considered for adoption in the semiconductor industry," Baum says. "It is both a pretty exciting and a pretty overwhelming time."

Part of the chemical industry's challenge is that neither it nor its semiconductor customers have fully worked out how to integrate these new materials into the chip. And as Stephen J. Robinson, president of Rohm and Haas's Shipley Microelectronics business, points out, despite the chip industry's reputation as bold and fast changing, it is actually quite conservative. "The semiconductor industry is very risk-averse," Robinson says. "It doesn't like to take a chance on a new technology that hasn't been proven."

This is definitely the case in the dielectric realm, where chemical companies have been promoting low-k materials for years with little success. The semiconductor industry switched from SiO2 to FSG for chips with 180-nm circuit lines and was expected to migrate to the new low-k materials for 130-nm circuitry. However, when those chips hit the market, they were still FSG-based, with the exception of some high-end SiLK-based chips from IBM and its technology partners.

FOOT-LONG Clariant's AZ unit has added a TEL ACT 12 coat-and-develop track, making it the first photoresist company to offer 12-inch wafer processing to customers.
Low-k materials, it turns out, have trouble withstanding subsequent processing steps such as CMP and barrier layer deposition. Intel, the world's largest chip maker, is implementing a low-k carbon-doped oxide for the first time this year in the 90-nm fabrication process that it is debuting at three of its chip facilities. Except for a small amount of SiLK, spin-on materials are still sitting on suppliers' shelves.

This can't be good news for longtime spin-on boosters like Dow and JSR Corp. Other firms, including Air Products, Shipley, and Honeywell, have also invested significant resources to develop spin-on dielectrics with no sales to show for their efforts. As Shipley's Robinson points out, timing is critical to making money in electronic chemicals. "Being early is as bad as being late," he says.

"There were many generations over which one or maybe two new materials were introduced. Now ... they are all piling up on top of each other."

The semiconductor industry's reluctance to change is equally apparent in lithography--the creation of circuit lines by shining light through an etched photomask onto a photoresist. In 2001, the prediction was that deep-UV photoresists--polyhydroxystyrene-based systems activated by 248-nm wavelength light--would be used to make 180- and 130-nm circuit lines and no more. But chip makers have succeeded in extending the technology to 100-nm and even 90-nm lines. Intel, for example, is using 248-nm lithography in its 90-nm chips along with the newer 193-nm lithography, which uses resists generally based on alicyclic-modified methacrylate polymers.

Such extension notwithstanding, photoresist makers like JSR, Shipley, Clariant's AZ Electronic Materials unit, and Arch Chemicals have been hard at work developing yet another generation of photoresist for 157-nm lithography. These resists are generally based on fluorine-containing polymers--some bilayer designs incorporate silicon derivatives--and will be used to etch lines 55 nm wide and lower starting in about 2007.

In May, however, Intel shook the lithography world by revealing that it had scrapped plans to adopt 157 nm because of a technical roadblock. Instead, the firm intends to extend 193-nm lithography all the way through 45-nm lines, then switch to a new technology--extreme ultraviolet lithography, which employs photons with a wavelength around 13 nm--for 32-nm production around 2009.

For Intel, the problem is not with the resists but with the "stepper" equipment that projects light onto the resist. The company has been concerned about the calcium fluoride used to make stepper lenses ever since the discovery that CaF2 causes unwanted birefringence of 157-nm light.

Photoresist makers are pushing ahead despite the Intel news. Two weeks ago, for example, Clariant announced an agreement with the semiconductor maker Infineon Technologies to jointly develop resists for 157-nm lithography.

But Ralph R. Dammel, Clariant's director of technology for 193- and 157-nm products, admits that the photoresist road map has become a lot less clear. In addition to 157-nm lithography and Intel's "extended" 193 nm, the semiconductor industry is also considering immersion technology, in which 193-nm light is projected through water, lowering its effective wavelength to about 132 nm. "Right now the road map is in a state of maximum disarray and maximum uncertainty," Dammel says.

A COMMON THEME in both the dielectric and lithography realms is a fracturing of markets into smaller and smaller niches. This prospect alarms materials suppliers used to putting significant R&D resources into new products and then recouping their investment by selling high volumes at a good profit later on.

For example, EKC's Coder notes that the wrong CMP slurry, photoresist removal chemical, or post-etch cleaning agent can damage the insulating ability of the new low-k dielectrics. "You would hope there would be one clear winner in low-k so you could focus your efforts on it, but in reality there is a multitude of different materials out there, and each one will require different chemical formulations," he says.

Meanwhile, the equipment needed to test new chemicals is getting prohibitively expensive. For example, Arch's Fujifilm-Arch joint venture in Japan just installed a 193-nm scanner at a cost of $15 million. Faced with an uphill battle in getting an adequate return on such an investment, materials suppliers are exploring various forms of cooperation with each other, with equipment suppliers, and with semiconductor makers themselves.

The pharmaceutical and semiconductor industries are similar in their reliance on technology suppliers to help develop their finished products. But while pharmaceutical companies often grant royalties or milestone payments to suppliers of critical technology, the same can't be said of semiconductor companies.

PHOTO FINISH Arch Chemicals researches, develops, manufactures, and tests 248-nm and 193-nm photoresists at its facility in North Kingstown, R.I.
Indeed, materials suppliers report that, if anything, the balance of power is shifting further away from them. "It almost goes in the opposite direction," Clariant's Dammel says of the pharmaceutical comparison. "If we codevelop something, we may owe royalties to our semiconductor industry partner."

Likewise, Coder points to a joint-development agreement (JDA) that EKC is currently negotiating with a leading semiconductor firm. In a traditional JDA, he notes, the customer owns the process and the chemical company owns the compositional materials. But in this case, the partner wants more. "We're seeing more and more of the semiconductor companies wanting to participate in the materials intellectual property as well," he says.

But although they struggle with joint development, materials suppliers unanimously concur that the agreements are necessary for success in the semiconductor business.

ATMI President Doug Neugold notes that leading semiconductor companies like Intel, IBM, and Texas Instruments have "always sought out relationships with materials suppliers, and we try to get into as many of those as we can. The majority of the longer-term R&D work we do is geared toward end-user requirements and comes about from discussions we have with them."

Friedrich Herold, vice president of Clariant's AZ unit, explains that his firm enters a wide variety of agreements--ranging from loose ones, in which Clariant merely presents a customer with product samples and gets feedback in return, to well-defined ones like the agreement with Infineon, in which chemistry is discussed in detail and experiments are jointly planned.

MATERIALS COMPANIES in markets such as photoresists pursue individual JDAs even when most of the industry is using a single product. Bruce E. Novich, business director of Arch's microelectronics unit, points out that each customer has its own set of process preferences and fabrication tool setups. "In 193-nm single-layer resists, for example, we have a broad chemical platform, but it is highly engineered to meet the needs of particular companies and, sometimes, particular process lines," he says.

Arch also pursues JDAs for the simple reason that they get the company in on the ground floor with its customers. "Getting in early really sets the market," Novich says. "Even if you have a great product, if you come in late it may not matter."

Although materials companies make extensive use of JDAs with customers, they rarely form them with each other. One prominent example is SiLKnet, a consortium of materials and equipment suppliers organized by Dow to integrate SiLK into the fabrication process.

Air Products' Mattimore argues that DuPont Air Products NanoMaterials, her firm's CMP joint venture with DuPont, is a materials JDA of sorts. At Shipley, Robinson points to his own firm's agreement with DuPont covering fluoropolymers used in 157-nm resists. And Eric Johnson, chief operating officer of JSR Microelectronics, JSR's U.S. unit, notes his firm's just-signed agreement to distribute Honeywell's DUO brand antireflective coating, used in combination with high-end photoresists.

"Never before have so many new materials been considered for adoption in the semiconductor industry."

More common is participation in industry consortia such as International Sematech and the Belgian research organization IMEC. Robinson notes that Shipley recently became an alliance partner in Minatec, a French government-backed consortium that includes semiconductor firms STMicro, Motorola, Philips, and Taiwan Semiconductor Manufacturing Co. One advantage of membership: "We can begin to look at our chemistry in the real world on customer equipment," he says.

According to Robinson, research cooperation on a deep and meaningful level doesn't come naturally in the electronics industry. "Because of the competitive nature of the materials game, the equipment game, and the semiconductor game, it takes huge economic pressure for people to actually do it," he says.

However, Mattimore notes that cooperation is being forced by the increasing interconnectedness of materials in a semiconductor. "A new low-k dielectric creates the need for a new CMP, which creates the need for a new etching gas--it's all interrelated," she says. "Finding a material that looks absolutely fabulous in a lab isn't enough."

Materials suppliers are watching with interest a Japanese effort launched in April called the Consortium for Advanced Semiconductor Materials & Related Technologies, or CASMAT. Most say it is the only materials-related consortium currently in place in the semiconductor industry.

JSR is a CASMAT member, and Johnson points to the firm's low-k dielectric experience to argue that semiconductor materials companies need to be more open to alliances and consortiums that are aimed at solving technical problems. "We've got a good material, but we haven't sold any," he says. "The material itself isn't going to get the job done. We have to be integrated with other parts of the process that haven't really worked together before. It's a really interesting model that needs to mature to enable the adoption of low-k technology."

Page 1 | 2 | 3 | Next Page


Chemical & Engineering News
Copyright © 2003 American Chemical Society

Visit SGI

Visit Eastman

Visit Fluorous Technologies Inc.

Visit ChemSW
Cover Story
Related Stories

Electronic Chemicals
[C&EN, July 15, 2002]

Electronic Chemicals
[C&EN, November 19, 2001]

Related Sites
E-mail this article to a friend
Print this article
E-mail the editor

Home | Table of Contents | Today's Headlines | Business | Government & Policy | Science & Technology |
About C&EN | How To Reach Us | How to Advertise | Editorial Calendar | Email Webmaster

Chemical & Engineering News
Copyright © 2003 American Chemical Society. All rights reserved.
• (202) 872-4600 • (800) 227-5558

CASChemPortChemCenterPubs Page