PALO ALTO, Calif., April 29, 2004 — It’s not every day that the 80s new wave act Dead or Alive’s song “You Spin Me Round” highlights a press conference dedicated to topics like pulsed laser deposition and qubit decoherence. But when the underlying theme is spintronics, and one of the people involved says one of the fundamental laws of physics could get turned on its head, perhaps one can forgive the pun.
After all, spintronics, or spin-based electronics, is perhaps the most successful example of nanoengineering outside of the semiconductor business. The giant magnetoresistive (GMR) head, which uses spintronics to greatly increase the readable density of hard disks, ships with practically every hard drive. IBM developed the GMR head, and Monday joined with Stanford University to announce the Spintronic Science and Applications Center, to look for other applications of this emerging nanotechnology.
Spintronics “could be as fundamentally important as the transistor,” said Robert J.T. Morris, director of IBM’s Almaden Research Center. To get to that lofty state, spintronics will have to provide the path past the end of Moore’s Law. He said IBM founded its “SpinAps” Center to help change that “could” to something more definitive.
SpinAps already shows promise in areas like communications because it can double the bandwidth of existing wires. But in the near term its most likely commercial applications are in memory and logic chips. Spintronics allows for nonvolatile solid-state circuits, which means they don’t need power to retain data. The best example today of this kind of memory is flash memory.
IBM, Infineon, Motorola and other companies are developing MRAM, or magnetic random access memory. Much of the early work on MRAM was funded by the Defense Advanced Research Projects Agency (DARPA), starting in 1993, because MRAM is by its nature radiation-resistant, and thus useful in satellites.
MRAM is based on spin electronics. It uses magnetics for memory, meaning that it retains basic data, like instruction sets or whatever a computer was last doing; it gives a system that uses it and “instant-on” capability.
But MRAM “is good at everything,” said Stuart Wolf, DARPA program manager who initially pushed for funding MRAM research. “It’s nonvolatile, high density, high speed, no refresh, low-voltage operation, unlimited read-write endurance.…It might be a universal memory.”
Universal memory is a sort of Holy Grail for the chip business. Semiconductors are mostly memory and logic chips. At one point, the two were combined, even though central processing units (CPUs) still have some memory on them, the systems run faster when they’re combined.
Dean McCarron, analyst at Mercury Research, calls nonvolatile RAM the “Swiss Army knife of memory — it’s for every market.”
McCarron and others note that it will take a long time to get to the point where you can actually replace CPUs with MRAM on a PC, but they say that the technology could take over multiple functions in cell phones within 10 years.
The market will start to find out later this year. Motorola will release a 4-megabit (Mbit) MRAM chip later this year, and Cypress Semiconductor will release a 256-kilobit (Kbit) chip. IBM and Infineon are developing a 16-Mbit memory chip that won’t appear until at least next year.
While MRAM might ultimately replace flash memory in devices like cell phones, digital cameras and music players, the end goal is about remaking chip architecture, said Bob Merritt, an analyst at Semico Research Inc.
“This has the potential to absolutely be a revolution in system design,” he said.
“Everyone knows if you can get the memory as close to processing element as possible, you pick up (device) speed, drop power and it gives you lots of advantages over this von Neumann architecture we have in today’s PC world.”
Merritt says MRAM clearly has the potential to integrate both memory and logic cheaply and efficiently in one chip, achieving the Holy Grail of the chip world. But so do other technologies, such as ferroelectric or FRAM, and Ovonics Unified Memory.
IBM and Stanford did not disclose how much they’ll invest in the research center, which will use existing facilities both within IBM Research and at Stanford. But they will dedicate almost 20 researchers and graduate students to the effort, and IBM’s Almaden Research Center is building a mad scientist’s dream machine, which will combine two magnetrons, a Pulsed Laser Deposition machine and a TEON (Thermal Evaporation of Oxides and Nitrides) device into one system, to help figure out which nanomaterials work best for spintronics.
The center will, appropriately, have at least one mad scientist, IBM’s Stuart Parkin, who is co-directing the center along with Stanford professors James S. Harris and Shoucheng Zhang.
Zhang has published a paper theorizing that spintronics will allow for devices that don’t follow Ohm’s Law (electrical current sent through a device gives off heat). This matters because one of the main problems with today’s processors is the amount of heat they give off.
No spin-doctoring going on there.