Issue Date: July 14, 2008
Chipping In For Chips
THE SEMICONDUCTOR industry has entered an age of chemistry, and there's no better evidence than a series of alliances that IBM has formed over the past year. These alliances, linking a microelectronics technology leader to some of the world's top suppliers of electronic chemicals, demonstrate that silicon is only one element in today's computer chips.
There was a time when semiconductor fabricators could do much of the job of building a chip with just silicon. To create the transistors at the heart of a chip, they started with a thin wafer of elemental silicon and formed a silicon dioxide layer on top by simple thermal oxidation. The SiO2 acted as an insulator between the silicon base and a silicon transistor gate. Chip makers did have to bring in some metal during the lithographic process for drawing integrated circuitry.
Paul Farrar Jr., vice president for process development with IBM's microelectronics business, recalls that it was relatively easy to shrink the size of a computer chip when he started in the industry three decades ago. "We made the gate smaller and used better lithography tools to get the images closer together," he says, referring to the optical aspects of the miniaturization process.
For 40 years, this shrinking, or scaling, has faithfully followed the 1965 prediction made by Intel cofounder Gordon Moore known as Moore's law. He anticipated that the semiconductor industry would double the number of transistors in a computer chip about every two years. But as Farrar notes, in recent years, companies such as Intel and IBM haven't been able to do the job with silicon alone.
"Over the last two or three generations, we could not scale the gate any further," Farrar says. Early in the decade, chip fabricators started augmenting the process. They used germanium—supplied by chemical companies in the form of organometallic precursor compounds—to help stretch silicon-silicon bonds and increase the mobility of electrons that flow through the silicon.
IBM and Intel are in the process of incorporating yet another new element, hafnium, in the transistor gate insulators of their most advanced chips???ones with 45-nm circuit wires. Hafnium oxide has a higher dielectric constant than SiO2 and can stem the electron leakage that was becoming a problem with ultrathin SiO2 insulators. In addition, Farrar says, IBM is experimenting with cobalt and ruthenium compounds to help block electrons from escaping the narrow copper wires in today's computer chips.
In a 2005 talk at a semiconductor manufacturing conference, Farrar made his own Moore-like prediction: that the semiconductor industry would eventually harness upward of 80 elements from the periodic table. The industry isn't there yet, he acknowledges, but more and more elements "are beyond the lab and embedded in the process."
ACCORDING TO Farrar, chemistry's importance is growing at IBM's basic research labs in Yorktown Heights, N.Y.; its process development labs in East Fishkill, N.Y.; and its new R&D center in Albany, N.Y. Some 1,200 IBM staff are involved in semiconductor R&D, and roughly a third of them are in Farrar's process development group. Of his senior technical staff, he says, about 40% are chemists or chemical engineers.
IBM's use of chemistry to advance chip-making isn't new. After all, it was IBM scientists who in the early 1980s invented chemical mechanical planarization (CMP), a semiconductor fabrication technique in which dilute abrasive slurries are applied to the silicon wafer with large polishing disks to smooth and level its surface before applying subsequent layers of circuitry.
IBM's use of partnerships also isn't new. But over the past two years, Farrar says, company executives have made a concerted effort to work more closely with chemical industry partners. He describes an evolution that started in the early 1990s, when the firm began collaborating with other semiconductor companies to solve knotty technological problems. That spawned the formal network of development partners that IBM leads today.
Then, about five years ago, the company opened up more and brought chip fabrication equipment makers such as California's Applied Materials and Japan's Tokyo Electron into the cooperative mix. "If you look at that ecosystem," Farrar says, "you ask yourself, 'What's the next step?' " The answer is chemical suppliers.
One of IBM's early chemistry partnerships was with Shipley, a Marlborough, Mass.-based electronic materials firm later acquired by Rohm and Haas. Rick Hemond, marketing director for microelectronic technologies at Rohm and Haas, recalls when IBM invented the photoresist chemistry used to pattern tiny circuit lines with 248-nm light. In 1994, Shipley and IBM joined up and spent the next six years bringing that chemistry to market.
Fast-forward to February 2008 when IBM and Rohm and Haas announced two partnerships, one for CMP and the other for lithography used during the transistor gate fabrication step known as ion implantation. Both procedures target the semiconductors with 32-nm circuit lines that are expected to hit the market in 2009 or 2010, as well as the subsequent generation of 22-nm chips.
Hemond's boss, James Fahey, is president of Rohm and Haas's microelectronic technologies business and a former IBM executive. Whereas the groundbreaking 1994 partnership focused on photoresists, Fahey says, the ion-implantation pact seeks materials that control the reflection of light used to expose photoresists in the ion-implanting process.
Unwanted reflection has always bedeviled semiconductor manufacturers, Fahey explains, but it is becoming a major stumbling block during the lithographic masking done prior to implanting ions in advanced chips. IBM and Rohm and Haas are researching antireflective coatings that are applied beneath and on top of the photoresist to block errant light waves and keep the implantion zones well defined.
According to Fahey, working with IBM is particularly rewarding for a chemical company because of the chip maker's deep materials knowledge. "IBM is both a materials development company and an integrated circuit company," he says. "Their desire to partner with multiple materials companies around the world is consistent with what they are inside."
Not surprisingly, Fahey and Hemond see little downside to their company's joint development agreements with IBM. Yet they acknowledge that helping IBM is not their only goal. "Our focus with IBM is to solve their problem first—and then bring that solution to the general market," Hemond says. In other words, they want to sell new materials to IBM and eventually sell similar materials to its competitors.
Indeed, Claus Poppe, director of BASF's electronic materials business unit, points out that IBM and its partners reap a dual benefit from materials advances because IBM is both a semiconductor manufacturer and a major technology licensor. "When IBM develops new technology, they usually make it available to their development partners," Poppe says. "And if that new technology is successful, it's also to our advantage because IBM's customers will need our materials."
It was a little more than a year ago that BASF and IBM announced an agreement to jointly research and develop electronic materials at IBM's Yorktown Heights laboratories and BASF's Ludwigshafen, Germany, headquarters. Poppe declines to provide details on the agreement, although he says research results will be published soon.
According to Poppe, BASF has similar research agreements with other major semiconductor manufacturers. "This is a business strategy of BASF—to provide solutions for the industry," he says. "You can't do that simply sitting in your own lab." However, because most IBM competitors are in the business of selling computer chips—and not licensing technology—they are less open about the development alliances they form.
Terry A. Francis, chief technologist for the Matheson Tri-Gas subsidiary of the Japanese industrial gases company Taiyo Nippon Sanso, agrees with Poppe that the chance to serve the larger market is critical to making joint development agreements economically sound. And the fact that IBM technology can flow immediately to joint development partners such as Infineon, Chartered, and Toshiba makes working with IBM that much more lucrative, Francis points out.
IN APRIL, Matheson and IBM announced a four-year agreement to develop new materials and processes for 32-nm circuitry and beyond. One goal of the agreement, Francis says, is to come up with new organometallic precursor gases and gas delivery systems that work at the lower temperatures demanded by germanium and other exotic semiconductor materials.
Francis is frank in explaining why Matheson seeks partners such as IBM. "We really don't want to go out and buy all the equipment to test the chemicals and hire all the Ph.D.s," he says. "So we look for situations to work both with the equipment suppliers and the device makers."
He contrasts working with IBM to working with Intel, another U.S. leader in advanced semiconductors. "When you work with Intel, you only work with Intel," Francis says. "They are a very good customer and we like working with them, but whatever you do with Intel will only be fanned out in Intel."
Francis conjures up an image of the atmosphere at Albany Nanotech, part of the University at Albany's College of Nanoscale Science & Engineering, where IBM has set up labs and brought in partners from the academic, semiconductor, equipment, and materials communities. "You have a kind of a chemical stew," he says. "You get interesting reactions just by being there and walking down the hallways. You get academic innovators who don't know what you can't do. It's a catalytic reaction that makes business sense."
Hemond also waxes about the close relationship between IBM and Rohm and Haas researchers, who drive back and forth between their respective labs in East Fishkill and Marlborough with ideas and data in hand. But IBM isn't some benevolent corporation promoting research creativity for the public good.
"We determined a really long time ago that we are not the only smart people in the universe," says Bernie Meyerson, chief technology officer for IBM's systems and technology group and its vice president for strategic alliances. Echoing his colleague Farrar, Meyerson says IBM seeks to create an "ecosystem" where numerous companies can collaborate on precompetitive problems that no single company can afford to tackle on its own. "And then you go off and beat each other to death in the marketplace," he jokes.
Affordability aside, Meyerson says IBM and its competitors and suppliers need to work together because materials start to act funny at the near-atomic scale that will be found in 32-nm- and 22-nm-generation chips. "We have reached the point that materials no longer exhibit exactly macroscopic behavior," he says. "Because we are approaching the quantum mechanical limits of some of these materials and devices, we need to test them at the native dimensions they will be used at."
AS THE COMPUTER industry's complementary metal-oxide semiconductor, or CMOS, chip technology approaches atomic dimensions, it's clear that, after 40 years, its days are numbered. To that end, IBM and Japan's JSR signed an agreement in December to conduct exploratory materials research on completely new technologies.
Eric R. Johnson, president of JSR Micro, says his company has run joint development programs with IBM for more than 10 years. The new collaboration, in contrast, is focused on basic research and is more open-ended in nature. Under the jointly funded agreement, JSR employees will work at IBM's Almaden Research Center in San Jose, Calif., on everything from the extreme ultraviolet lithography expected to be needed beyond 22-nm chips to new imprint lithography techniques and to molecule-scale self-assembly of computer circuitry.
Johnson is excited about the deal because, as he says, "IBM is one of the last of our customers with pure research capabilities." Yet it's also a full-fledged chip manufacturer with the development, device integration, and validation capabilities to test new materials in real-world conditions. "Without integration and validation data, the materials are basically academic," Johnson notes.
He's also happy that his scientists will experience a world-class basic research environment. "JSR does good applied research, but we don't have a lot of mad scientists throwing stuff against the wall," he says. Still, this kind of research costs money, and what Johnson hopes eventually to get out of it are contracts to sell new materials to IBM and its technology partners.
At IBM, Farrar and Meyerson want something from their materials alliances as well. Farrar says it's getting harder and harder to pattern and etch circuitry in increasingly tiny electronic devices. "We're starting to be challenged in how small we can go with conventional lithography," he says. "In the end, we're going to go to a chemical world where materials self-assemble into the patterns they need to be."
For that, Farrar and Meyerson will be looking for a few good chemical companies to help them out.
- Chemical & Engineering News
- ISSN 0009-2347
- Copyright © American Chemical Society