April 26, 2004 4:01 AM PDT
IBM, Stanford put new spin on chip research
The undertaking, the Spintronics Science and Application Center, will be based at IBM's Almaden Research facility in San Jose, Calif., and Stanford's nearby Palo Alto campus.
Spintronics revolves around precisely controlling the magnetic field emanating from a thin film. Magnetic fields create electrical resistance, and high and low levels of resistance can be assigned a one or zero. By controlling the magnetic field and interpreting the resistance levels at various points on the film, researchers can arrive at digital data.
The metaphorical name of the technology derives from way electrons are said to spin.
Spintronics has actually been around for years. IBM produced disk drive heads, using giant magnetoresistive (GMR) technology, taking advantage of these properties in 1997.
Magnetic random access memory (MRAM) could become the next area where spintronics is incorporated. Ideally, MRAM would be able to store a substantial amount of data, consume little energy, operate at a much faster rate than conventional flash memory--and last forever.
Finding a replacement for flash technology, which is used in cell phones, memory cards in digital cameras and other devices, is an urgent business in the semiconductor market. Demand for flash is growing extremely rapidly.
The basic technology behind flash, however, is getting more difficult to advance, and nearly every major manufacturer is examining alternatives.
Critics and competitors, though, point out that MRAM is far from being the anointed successor. MRAM is unconventional, and IBM has to show it can mass produce the technology cheaply. Intel has also pointed out that the memory cell sizes are large, which makes it difficult to slip into small devices like digital cameras. Others have said that the changes in resistance are too subtle and can lead to data corruption, a contention IBM disputes.
Last June, IBM and Infineon published a paper describing how the companies produced an MRAM chip on the 180-nanometer manufacturing process that held 128 kilobits of data. At the time, the two companies promised to more fully demonstrate MRAM in early 2004 and predicted that MRAM could start to be produced commercially by 2005.
Further out, spintronics principles could be used to flip some kinds of transistors off and on. An Intel spokeswoman said that this is one of the promising areas of long-term research at some universities right now, but it might not hit the stage of commercial implementation until around 2021.
Still, it might come in handy. Researchers at Intel recently published a paper saying that chipmakers in the coming years will likely hit a looming barrier in Moore's Law that could prevent chip designers from gaining performance by shrinking their chips, the engine behind the exponential growth in computer power for more than three decades.
The IBM-Stanford project, also known as SpinAps, was first reported late Sunday by the online edition of The New York Times.