Magnetic Bit Boost: Quantum rewiring for computer memories

A quantum-mechanical memory component that might replace the electronic memories used for decades in computers and other gadgets has come closer to practicality, thanks to improvements achieved by research teams in the United States and Japan.

Current electronic devices rely mainly on two types of on-chip memory—static random access memory and dynamic random access memory, which is more compact. These memories can be accessed quickly, but they’re volatile—shutting off power erases the data. Nonvolatile memory, such as hard disks, takes longer to access.

Although other types of nonvolatile memory are increasingly available—for instance, the flash memory in a digital camera—such storage options typically cost more per bit and hold less data than disks do or have slow access times.

Now, scientists led by Stuart S.P. Parkin of the IBM Research Division in San Jose, Calif., and an independent team headed by Shinji Yuasa of the National Institute of Advanced Industrial Science and Technology in Tsukuba, Japan, have tripled the key measure of performance of a memory component called a magnetic tunneling junction (SN: 4/3/99, p. 223).

Such quantum-mechanical junctions are the foundation of a fledgling memory technology called magnetic random access memory, or MRAM. It’s nonvolatile yet has access speeds that rival those of conventional memories.

Two companies are already about to ship chips of MRAM. In those chips, the key performance measure, called tunneling magnetoresistance, stands at about 70 percent. By contrast, the improvements described in back-to-back reports in the December Nature Materials boost the components’ magnetoresistance to more than 200 percent.

“These are definitely important results,” comments Robert A. Buhrman of Cornell University. “They could make a big difference in making MRAM a success.”

Potential uses for the souped-up junctions go beyond electronic memory. They include extraordinarily sensitive read heads for hard disk drives and components of so-called spintronics circuits, which aim to exploit electrons’ magnetic properties (SN: 7/17/04, p. 37: Available to subscribers at Feel the Force: Magnetic probe finds lone electron).

Another application could be switches that open or close to reroute information in a computer circuit. “One might even imagine a computer that instantly reconfigures itself to solve a given problem in the most efficient way,” say William H. Butler and Arunava Gupta of the University of Alabama at Tuscaloosa, in a commentary in the journal issue carrying the new reports.

A magnetic tunneling junction resembles a sandwich in which the bread consists of thin layers of magnetic materials, such as iron or cobalt, and the filling is an insulating layer of, say, aluminum oxide. While the insulator blocks electron flow, a quirk of quantum mechanics called tunneling enables some electrons to pass through the barrier. If the magnetic fields of the outer layers point in opposite directions, the barrier becomes less penetrable than if those fields point the same way. Changing the field direction on a magnetic layer therefore alters the rate of electron flow. A high flow might represent a 1 and a low flow a 0.

In both junctions reported in Nature Materials, the improved magnetoresistance resulted from swapping a crystalline inner layer of magnesium oxide for the usual amorphous material. However, the IBM version appears closer to commercial reality. Parkin and his coworkers used less-orderly magnetic outer layers to produce a sandwich compatible with existing chip-making methods.

By contrast, Yuasa and his colleagues grew all layers of their junction as perfect crystals. That also achieved high magnetoresistance, but the fabrication method “is not good for mass-production,” Yuasa admits.

The Japanese institute has announced that it too is developing a mass-producible magnetic tunneling junction.