The Wall Street Journal
By Don Clark, staff writer
© Copyright 2003 The Wall Street Journal.
For years, computer experts have worried that
But Wednesday, Intel Corp. itself plans to declare that it has removed the biggest roadblock to boosting the performance of chips over the next decade. The company claims in a technical paper that it has successfully replaced materials that have been a vital part of chips for more than 30 years. The change will help prevent electrical current from leaking inside chips, a growing problem as more circuits get packed onto the semiconductors.
Electrical leakage generates heat and consumes power. Unless addressed, the problem is widely expected to slow progress in the electronics industry. Solving it could yield products with hundreds of times the calculating power of today's computers and consumer devices.
Jack Lee, a
But Intel, out of fear of tipping off competitors too early, isn't identifying its new chip materials, which won't be placed into production until at least 2007. As a result, some rivals question Intel's claim.
Luigi Colombo, the head of similar research efforts at Texas Instruments Inc., says his company will release a paper in December showing comparable progress as Intel. He complains that Intel's paper, prepared for a technical conference in
Companies including Motorola Inc. and International Business Machines Corp. also are in the race to solve the electrical-leakage problem. All these efforts are designed to head off an interruption in the pace of progress observed by Mr. Moore: that the number of transistors on a chip doubles every year or two.
Transistors in chips act like switches to pass along signals and help perform calculations, effectively opening and closing when specified amounts of electricity are applied. Shrinking transistors create chips that are faster, consume less power and cost less -- the driver for more powerful computers, consumer devices and equipment for offices and factories. Intel this year introduced a chip containing 410 million transistors, and hopes to crack one billion transistors in several years.
But problems are appearing as parts of transistors shrink to thicknesses of only a few molecules. At such dimensions, a key layer of insulating material that is supposed to block the flow of electricity when a transistor is off begins to let current flow through. It's like a faucet with a bad washer. The leaky chips consume too much electrical power and generate too much heat to be practical for use.
Intel, the dominant maker of microprocessors that power personal computers, estimates that as much as 40% of the electricity consumed by its flagship Pentium 4 line of chips is wasted through leakage. Company officials have joked that chips in a few years will have the heat-generating capabilities of a nuclear power plant.
"It's a very basic problem and a very serious problem that the industry has to come to grips with," says Sunlin Chou, the senior vice president in charge of Intel's technology and manufacturing group.
Solving the problem, most experts agree, requires a wrenching shift away from a major building block for the industry -- silicon dioxide, an insulating material that attaches easily to the silicon foundation of semiconductors. Companies have been searching for new materials -- known in high-tech jargon as "high-k" materials -- that have better ability, at tiny sizes, to alternately serve as an insulator and store electrical current.
Intel, without being specific, says it has found a promising candidate for the insulating layer. Equally important, Intel says it has overcome barriers that have so far prevented the use of high-k materials, which include unpredictable shifts in voltages and slowdowns in the flow of electrons through transistors.
To do that, Intel replaced another mainstay material, polysilicon, which is now used to make the electrodes on each transistor. The company says it has developed metals -- which it also is not identifying -- to make electrodes that work well. Ken David, Intel's director of components research, says the combination of new materials has created test transistors that operate at very high speed, with one-hundredth the amount of current leakage.
"This is very, very encouraging," says Veena Misra, an associate professor of electrical and computer engineering at
Intel plans to gradually disclose the materials as chips using them get closer to production. "We invest a lot and we have a very talented crew," Mr. Chou says. "We don't want to just help our competitors to take advantage of our technology."
TI's Mr. Colombo suspects that Intel is using a material for insulation similar to the one used by his company, a substance called hafnium oxide. TI is experimenting with both metal and polysilicon electrodes, he adds. Motorola has discussed using metals for electrodes such as titanium nitride and tantalum silicon nitride.
Bernard Meyerson, a vice president and chief technologist in IBM's technology group, says the data in Intel's paper are based on dimensions of materials that may not exhibit all the problems that occur when circuitry shrinks. "I'm unable to tell, frankly," whether Intel's paper represents a breakthrough, he says. "There is a tremendous amount of work left to do."
Mr. Chou, Intel's influential head of production technology, stresses that the new materials are "candidates" for use in Intel's manufacturing, but the company could modify its plans. He acknowledged that he faces an internal battle convincing some superiors, such as Chairman Andy Grove, that the high-k material can be churned out in high volumes as reliably as today's chip materials.
"Andy is one of the skeptics and I don't blame him," says Mr. Chou. "He's still going to challenge us to prove it in manufacturing."