ELECTRIFYING Materials suppliers keep introducing new products to keep up with the needs of semiconductor makers. COLOUR SOURCE INTERNATIONAL PHOTO

Chemical companies are behind many of the changes, by supplying advanced materials that create more powerful semiconductors, lighter batteries, and sharper images on plasma displays.

Progress in the electronics business is not linear, though; it happens by fits and starts. Semiconductors, which make up the largest segment of the electronic materials market, some 10 times larger than the next biggest segment, liquid-crystal display materials, have had an especially rough ride. Worldwide semiconductor sales tripled from 1990 to 1999, and they are likely to grow another third by 2004. But 2001 was a year of global recession that was particularly tough on semiconductor manufacturers. Semiconductor sales dropped about 30% compared with the year before.

And 2002 is not shaping up to be much of an improvement. In the U.S., industry leader Intel reported first-quarter sales of $6.8 billion, up just 2% from a year earlier. Last month, it reported that second-quarter revenues would be $6.2 billion to $6.5 billion, down from the first quarter and down from its earlier expectations.

In Japan and South Korea, semiconductor chip manufacturers are merging their operations or even attempting to sell themselves off to the highest bidder. Seiichi Hasegawa, managing director of JSR Corp.'s electronic materials business, says, "The Japanese semiconductor industry was most severely damaged by the downturn."

For the companies that supply materials to the semiconductor industry, 2001 was also bad, but not as bad as semiconductors. Sales fell 23%, according to the trade association Semiconductor Equipment & Materials International (SEMI). And although semiconductor manufacturers will not enjoy much growth this year, sales of semiconductor materials will expand by more than 11%, the group predicts.

Ian Thackwray, who heads Dow Corning's global electronics business from Lyon, France, says that "2001 was clearly the worst year in history in electronics." He adds that "we're growing in 2002. But I have a question as to how fast that will keep rising and whether or not it is sustainable in the short term." He expects a much better year in 2003.

Similarly, Pierre R. Brondeau, head of the electronic materials business group at Rohm and Haas, said at a conference in late June that his business grew 15% in the second quarter of this year.

Dan Tracy, a senior market analyst at SEMI, says 2000 was somewhat of a fluke year. Sales of semiconductors grew too fast, making 2001's sales drop very difficult to cope with. Ignoring 2000, worldwide semiconductor sales declined only slightly from 1999 to 2001.

From 2003 onward, materials suppliers expect to enjoy renewed prosperity for several reasons. One major factor is the expected recovery in the world economy, Thackwray says. But another reason is the increased sophistication of the semiconductors being produced. SEMI projets that semiconductor sales will grow from $139 billion in 2001 to $216 billion in 2004. Sales of semiconductor materials will grow by a third to $28 billion in the same period.

Also helping materials suppliers is a rising demand for new electronic appliances and portable gadgets. Companies such as Sony and Sharp keep coming up with clever, portable appliances powered by smaller yet longer lasting batteries; these batteries in turn represent a booming opportunity for materials suppliers (see page 25). And the Japan Electronics & Information Technology Industry Association expects that sales of plasma displays, primarily used in flat-screen televisions, will grow 20 times to 3.3 million units between 2001 and 2006.

Still, companies selling materials to semiconductor manufacturers will face considerable challenges in coming years. Competition is fierce in the electronic materials industry, and the resiliency of materials suppliers will be tested as semiconductor manufacturers experiment with new production techniques.

ONE IMPORTANT change is that operators of semiconductor fabrication facilities, or foundries, are progressively migrating toward silicon wafers with a diameter of 300 mm. The current standard wafer--the thin disk on which microcircuits are produced before being assembled into chips--is 200 mm. According to Intel, a 300-mm wafer provides 240% more computer chips than a 200-mm wafer.

However, a semiconductor foundry implementing 300-mm manufacturing makes use of different machines, and often different materials, than a 200-mm plant. One of the world's leading foundries, Taiwan Semiconductor Manufacturing Co., says it spent $500 million simply to test its 300-mm manufacturing process.

Though the wafers on which semiconductors are produced are getting larger, the circuits in these semiconductors are getting smaller. In 2000, the semiconductor industry's leading edge was chips, such as Intel's Pentium 4, with circuit lines 180 nm wide. Today, some of the chips on the market feature 130-nm circuitry, and firms such as Intel say 90-nm circuitry is coming in 2003.

For many years, semiconductor manufacturers have used aluminum as the prime interconnect and silicon dioxide as an insulating material preventing cross talk between the closely spaced aluminum wires. It has been predicted for the past few years that both aluminum and silicon dioxide would have to yield to other materials as the industry moves toward smaller and smaller circuitry.

The prediction has proven true for aluminum, as it is rapidly being replaced by copper. But silicon dioxide is surprisingly resilient. Numerous chemical companies have invested in the development of low dielectric constant, or low-k, alternatives to silicon dioxide, mostly to find that their new offerings are ahead of semiconductor manufacturers' needs or simply not needed. In April, the consulting firm Kline & Co. put out a bulletin called "low-k dielectrics market stalls temporarily."

Another change in the manufacturing process is the introduction of a production step called chemical mechanical polishing or planarization. As manufacturers stack more and more layers of circuitry, they are finding the need to use dilute CMP slurries to smooth each layer out before photolithography and other production steps can take place.

Demand for CMP slurries is rising rapidly. The market, which barely existed a few years ago, is now worth about $300 million per year and growing 20% annually, according to DuPont Air Products Nanomaterials.

Integrating a single new material or equipment change into the chip-making process is complicated. What is particularly challenging for semiconductor manufacturers and their suppliers today is that numerous product and process changes are being introduced all at once. They must find ways to juggle new technology nodes, or the changes required as chip interconnects shrink. Foundry operators, for instance, are attempting to draw 100-nm circuitry on 300-mm wafers while learning how to use new CMP slurries, copper circuitry, and dielectrics.

"The tough part of developing new materials for new technology nodes and new devices is not just the materials' physical properties but also the integration of the materials into the process," Thackwray says. He explains that semiconductor manufacturers try to temper this process by extending the life of materials they know well--silicon dioxide is an example--for as long as possible.

The manufacturing of semiconductors is an immense effort involving hundreds of suppliers of machines, materials, and gases. The independently owned and managed materials suppliers face a bewildering array of choices when channeling the energies of their research and development personnel toward the products that will be needed by semiconductor manufacturers two, three, or five years later.

Other than constantly attending trade shows and liais-ing with customers, equipment and materials suppliers have access to a tool designed to harmonize their R&D efforts with the needs of semiconductor manufacturers. Compiled annually by the consortium International Sematech, the International Technology Roadmap for Semiconductors (ITRS) is a detailed survey of semiconductor manufacturers and their suppliers that tries to predict the future needs of the industry.

"We basically believe in the road map," Hasegawa says, noting that sales of JSR's CMP slurries are basically in line with what ITRS predicted.

But concerning the need to design new materials to insulate interconnects, the survey was overoptimistic in its timing. "If you take the low-k dielectric, in the past few years, the road map has been constantly amended," Thackwray says. "If you were working off the road map five years ago, you were working on something that isn't needed today."

In 1997, the document predicted that new materials with a dielectric constant below 3.0 would be needed at the 180-nm technology node. Scores of companies invested tens of millions of dollars to develop such materials that are today needed in only some of the most advanced semiconductors.

The 2001 edition of the road map predicts that low-k materials will become standard at the 90-nm technology node, which is still at least a year away. So far, semiconductor manufacturers have mostly continued to make use of silicon dioxide specially treated--with silicon tetrafluoride, for example--to lower the dielectric constant.

The low-k example illustrates one of the challenges that electronic materials suppliers face in maintaining their competitiveness. They must constantly make investments in developing materials that will not earn profits until years down the road, if at all. Furthermore, the cost of developing new materials keeps rising. The income statements of both semiconductor manufacturers and materials suppliers indicate their R&D expenditures increased, if they moved at all, in 2001, despite the drop in income.

JSR's Hasegawa says the R&D investment required for new photoresists is as high as the cost of building a new plant where the resists will be manufactured. "Not all the photoresist suppliers have the money to follow the shifts of the technology nodes," he says.

Continuous investment in R&D is essential because the types of semiconductor materials change every two years or so. SEMI estimates that a new generation of photoresist costs $32 million to develop, on average, compared to a photoresist market of about $640 million in 2001 (see page 23).

Partly as a result of ballooning R&D costs, industry consolidation is imminent, Hasegawa maintains. Another reason for consolidation, Thackwray adds, is that electronic manufacturers expect high standards of technical support from their suppliers wherever in the world they have plants. Electronic materials suppliers that lack a global presence find it increasingly tough going.

As a result of the pressure, Hasegawa expects that some materials suppliers will stop developing new products and eventually drop out of some businesses. Mitsubishi Chemical provided an example in 2000 when it sold its photoresist business to Rohm and Haas's Shipley subsidiary.

Dow Corning is actively looking for electronic materials businesses that no longer fit the portfolio of larger companies, Thackwray says. Outright acquisition of electronic materials companies is, however, very difficult, he says, because it tends to be extremely expensive.

INDEED, collaborative agreements seem to be more common in the electronic materials business. Dow Corning, for example, has a supply and marketing agreement with Japan's Toray Industries under which it markets Toray polyimide in North America and Europe. Thackwray says the arrangement combines Toray's manufacturing capabilities in polyimide and Dow Corning's global marketing presence.

Similarly, DuPont and Air Products teamed up last year to produce CMP slurries. DuPont Air Products Nanomaterials combines DuPont's know-how in developing and manufacturing colloidal silica sols with Air Products' experience as a supplier of specialty gases to the semiconductor industry worldwide.

Other times, a materials supplier needs to build support for a new technology among suppliers of other materials that will be affected by the technology. An example is the SiLKnet Alliance, a loose collaboration of materials suppliers and semiconductor equipment producers assembled by Dow Chemical to solve technical challenges posed by implementing Dow's SiLK low-k dielectric.

The unexpectedly slow pace of low-k material adoption by semiconductor manufacturers shows that there are no sure bets in the electronics industry. In coming years, there will likely be more surprises when semiconductor foundry operators select new materials or suppliers. SEMI's Tracy calls this peculiar combination of growth and uncertainty "low visibility." For better or worse, it's a state of affairs to which the industry has become accustomed.

ELECTRONIC CHEMICALS

Chemical & Engineering News
Copyright © 2002 American Chemical Society