September 12, 2002 11:50 AM PDT
Nanowire or nanotube? Intel looks ahead
- Related Stories
Big ideas, small toolsSeptember 13, 2002
Intel Developer Forum news
Read what visionaries at the confab
expect for wireless, security
On Thursday, at its Developer Forum here, the Santa Clara, Calif.-based chipmaker disclosed a number of technology changes and avenues of research that will direct the future development of its chips.
The company, for instance, confirmed that it is working on a multiple-gate transistor, called the Tri-Gate transistor, that will, if eventually incorporated into commercially released chips, increase the amount of electricity flowing through transistors and microprocessors, and so boost their performance.
Intel also said that it is working with Harvard and other universities on silicon nanowires and carbon nanotubes, two experimental structures made up of, respectively, self-assembling silicon and carbon atoms. After 2010, one of these technologies could begin to replace standard transistors and over time become the building block of chips.
"Right now they both show promise. The determining factor will be to what degree they can be mass-produced," said Sunlin Chou, Intel's senior vice president of technology and manufacturing. "The question is, can you make billions of these on a chip with consistent characteristics?"
Thursday's keynote speeches, which also included a guest appearance by "Star Trek" actor William Shatner, focused on the future of technology. In years to come, as chips continue to get less expensive, more powerful and smaller, computers and wireless connections will be embedded in everything. To prove the point, Intel's chief technical officer, Pat Gelsinger, opened the session decked out like a man who'd had a head-on collision with a Sharper Image catalogue. Riding a Segway scooter, Gelsinger was wearing a baseball hat with an eye monitor, a PDA with a camera, a Bluetooth headset, a universal remote control, and a vest with a server sewn in the back.
"When we get things to silicon, everything gets cheap. When it gets cheap, it gets ubiquitous," Gelsinger said.
So what does this mean on a practical level? By mid-decade, Intel will be able to integrate radios onto ordinary silicon chips, Gelsinger said. As a result, wireless communications will essentially become free. Optical products, which now require thousands of dollars and several engineers to tune, will become cheaper because many of the functions will be integrated into mass-producible silicon chips.
It will also be easy to create sensor networks, where coin-sized "motes" spread over a large area can measure seismic activity, temperature, pressure and other factors. In an experiment at the Great Duck Island Environmental Preserve in Maine, researchers are using sensors to monitor how atmospheric changes can alter animal behavior. Typically, researchers have to make these observations directly, which can disturb animals.
As a part of such efforts, Intel on Thursday released TinyDB and TinyOS, a database and operating system for these sensors. The software is open source and uses little memory or energy.
While Gelsinger spoke of future applications, Chou focused on chip and transistor release schedules. Nanotechnology, the science of manufacturing chips and other products with components measuring less than 100 nanometers (100 billionths of a meter) has already begun, he said. Transistor gates inside current chips measure 70 nanometers.
Next year, the nanotechnology era will begin in earnest with the release of 90-nanometer chips, Chou added. Manufacturing chips with smaller dimensions, however, will require quite a number of technological breakthroughs.
For one thing, lithography, the science of printing circuit patterns onto chips, will have to change. Sometime in the decade, extreme ultraviolet (EUV) lithography will replace conventional lithography. The light wavelengths used in EUV measure only 13 nanometers compared with 193 nanometers for today's lithographic techniques.
"It will be possible to paint much finer lines in the future," Chou said. "We have gone beyond the technologically feasible phase and into the commercial-development phase."
In a similar vein, Intel is experimenting with atomic layer deposition, which lets manufacturers make chips by piling single layers of atoms on top of each other. "It relies on chemical properties to self-assemble," Chou said.
Transistors will also change. Starting with 90-nanometer manufacturing, chips will incorporate strained silicon, an extra layer of silicon and germanium atoms that improves performance. The germanium atoms spread out the silicon atoms and make it easier for electrons to travel. Executives at Amberwave, a leading strained-silicon developer, liken it to running through a forest with fewer trees.
Later, layers of materials for preventing electron leakage, such as high-k dielectrics, will likely be incorporated. Further research, however, is required.
The Tri-Gate transistor is in a similar phase. Current transistors contain a single horizontal gate. By contrast, the Tri-Gate rises up like a microscopic mesa, and allows electrons to flow on the horizontal top and the two vertical sides. The company has manufactured test transistors and will disclose further details at a conference next week in Japan, Chou said.
IBM and chipmaker Advanced Micro Devices are working on dual-gate transistors, both companies stated this week.
Nanotubes and/or nanowires will follow. "These new materials aren't going to come in and sweep away the silicon base. They will come in incrementally," Chou said.
Although Chou reiterated that it's way too early to tell which, if either, alternative will eventually be adopted, they are the two leading contenders. Carbon nanotubes are essentially hollow tubes that resemble a spool of chicken wire. Silicon nanowires, by contrast, are solid.
Shatner, meanwhile, discussed his new book, "I'm Working on That," about how different scientists such as Stephen Hawking have used science fiction to direct their research. Many of the examples in the book focus on ideas found in "Star Trek" that eventually became products. However, few, if any, of the concepts propounded in "T.J. Hooker," Shatner's mid-'80s cop show, are discussed in the book.
Gelsinger told Shatner that his favorite episode of "Star Trek" was "the one where you go back in time." Shatner's Capt. Kirk also apparently meets a woman and struggles with a decision, Gelsinger added.
"That's 17 episodes," Shatner replied.