Physicists expose a hidden facet of a familiar phenomenon
While conducting experiments for his physics Ph.D. in the early 1990s, Dan Ralph suddenly found himself in unfamiliar terrain without a compass. Examining nanoscale sandwiches of magnetic and nonmagnetic materials in a Cornell University lab, Ralph discovered that voltages caused by electric currents passing perpendicularly through these layers would sometimes increase abruptly for no apparent reason. "Something kind of drastic was going on," he recalls.
Ralph wrote up the bizarre results as a small part of his doctoral thesis. "I speculated all sorts of things in my thesis. It turns out all of those were wrong," he says.
Although Ralph moved on to postdoctoral studies elsewhere, his graduate adviser wasn't about to let the matter rest. "I didn't have a good idea what was going on except that it was very interesting," recalls Robert A. Buhrman. He urged other students to look at the nanostructures that Ralph had investigated, and he eventually also drew Ralph, now a Cornell physics professor, back into the studies.
Today, as a result of some clever theorizing and years of experimental work at many labs, physicists have a good idea of what was going on. They're now beginning to investigate under what conditions and for what applications magnets can respond to electricity in ways that no one recognized a dozen years ago.
"Ever since the discovery of magnetism, the only way known to change the direction of how a magnet points was to apply a magnetic field," notes William H. Rippard of the National Institute of Standards and Technology (NIST) in Boulder, Colo. Now, he says, research is exploring an "entirely new way" to influence magnetic behavior.
The approach relies on what physicists call the spin-torque, or spin-transfer, effect. In a nutshell, a swarm of electrons can make a magnet's polarity reverse or wobble because each electron has its own intrinsic magnetism, called its spin.
In the past few years, scientists have begun to demonstrate that the new effect could have big commercial payoffs. Some researchers have already harnessed the effect in a prototype magnetic digital memory, which may someday be a contender against, for instance, the flash memory in digital cameras and other electronics. Others have made tiny microwave beacons that can coordinate their signals in a manner reminiscent of crickets and fireflies synchronizing nightly chirrups and blinks. These developments may lead to smaller, faster, and more energy-thrifty devices for data storage, wireless communications, and information processing.
An electron can be thought of as a tiny bar magnet whose north pole can point in any direction. Familiar magnets are objects in which multitudes of electron spins line up in one direction.
Electron spins can exert rotational forces, or torques, on each other, much as arm wrestlers create torques as they push against each other. Basic theoretical and experimental research by several scientists, including Albert Einstein, led physicists to recognize subatomic torque roughly a century ago, notes Mark Covington of Seagate Research in Pittsburgh.
In 1996, practical-minded theorists John C. Slonczewski of IBM T.J. Watson Research Center in Yorktown Heights, N.Y., and Luc Berger of Carnegie Mellon University in Pittsburgh independently proposed a novel twist on the phenomenon: If electron spins are aligned within an electric current flowing through a magnet, they might exert torques large enough to reorient the magnetization of that magnet. The theorists had in mind ultrathin sandwiches of magnetic and nonmagnetic metals similar to the oddly behaving structures that Ralph and other scientists had studied.
Physicists already knew that the spins of electrons in an electric current initially point in random directions, but as the current passes through a magnet, the spins take on the magnet's orientation. Slonczewski and Berger now proposed the reverse effect: A polarized current could force its orientation onto a magnet.
The notion that the electrons in the current might take the leading role in this wrestling match was startling. Further observations reported in 1998 by Maxim Tsoi, now of the University of Texas at Austin, and other physicists in France, Russia, and the United States seemed to bear out the theorists' proposals. "People started to say, 'Hey, maybe that's what's going on,'" Buhrman recalls.
By 1999, experimenters had confirmed the theorists' prediction. Their tests demonstrated that a polarized current that has passed through one magnetized layer and then flows through another can coerce the magnetization of the second layer to swing around to the same direction as the first. Once this alignment occurs, moreover, the polarized current flows more easily through the second layer than it had previously. Similar effects had produced the peculiar voltage hops observed by Ralph in his experiments of years earlier.
In data-storage devices, the magnetization of a layer represents a bit of digital information—a zero or a one. Flipping the magnetization therefore changes the bit's value. In the write heads of hard disk drives and in some other established technologies, bit flipping relies on magnetic fields generated by other magnets, notes Chia-Ling Chien of Johns Hopkins University in Baltimore.
Creating these magnetic fields requires bulky components—a disadvantage in computer and other technologies that continue to miniaturize. What's more, because magnetic fields spread unevenly through space, they're tricky to employ, adds Rippard. The use of polarized currents, instead of magnetic fields, keeps the process neatly within the data bits, notes Chien.
This approach probably won't be applicable to most magnetic technologies because it only works at the nanoscale. That's because the intensity of the required currents is extremely high. Flipping one bit, for instance, requires the equivalent of 5 million amperes of current per square centimeter. By contrast, ordinary household wiring carries current densities that reach only a few hundred amperes per square centimeter.
The high-intensity bit-flipping currents generate much heat. That would be a problem for larger devices, but those spanning mere tens to hundreds of nanometers have high ratios of surface area to volume and so can dissipate heat effectively.
Among researchers quick to embrace the spin-torque approach are the developers of magnetic random access memory, or MRAM—which many data-storage specialists regard as the most promising memory chip of the future (SN: 12/18&25/04, p. 389: Available to subscribers at Magnetic Bit Boost: Quantum rewiring for computer memories). Spin torque offers a way to circumvent flaws of conventional MRAM designs, in which currents in pairs of crisscrossing wires generate magnetic fields to flip bits. But fields along single wires can make nearby bits unstable—a problem expected to get worse as memory bits are crammed closer together. On the other hand, spin-torque flipping is predicted to get easier as memory chips shrink.
At the International Electron Devices meeting last month in Washington, D.C., a team of researchers from Sony Corp. in Atsugi, Japan, reported the first prototype magnetic-memory chip composed of an array of spin-torque bits. Because spin-torque switching is more efficient than magnetic field methods, the 4-kilobit chip, dubbed the spin-RAM, uses only one-twentieth as much power to flip a bit as does conventional MRAM, the researchers claim.
"This may be a breakthrough in nonvolatile high-density memory for consumer applications," says Tom Bonifield of Texas Instruments in Dallas.
Flipping is not the only gymnastic trick that magnets can perform when hit by spin-polarized currents. They can also make their magnetization directions twirl like dervishes.
That twirling, known to physicists as precession, is just unconsummated flipping, says Rippard. It takes place when, in the presence of an external magnetic field, spin currents are not large enough to completely overthrow the previous magnetization. Instead they tip the magnetization direction, or arrow, only partway toward a reversed configuration, pushing it out of alignment with the external field. That field then forces the arrow to wobble, or precess, as the shaft of an unsteady top does.
Changing the precession speed is as easy as turning a knob: The stronger the current, the faster the twirl. That tunability could prove handy for microwave generators and detectors, which are used in a range of applications, including collision-avoidance and other radar systems and wireless devices such as cell phones, says Frederick B. Mancoff of Freescale Semiconductor in Chandler, Ariz.
Another advantage of the new devices is how rapidly they can oscillate. Already at 35 billion cycles per second, or gigahertz (GHz), the rates are expected to reach 100 GHz, says Mancoff. Higher-frequency signals carry information faster than lower-frequency ones do. Today's cell phones typically operate at 1 to 2 GHz.
Recent results suggest a way to overcome a potential problem with the oscillators. A lone nano-oscillator produces signal strengths of only trillionths to billionths of a watt. That "won't give you enough signal to be useful," notes Jordan A. Katine of Hitachi San Jose (Calif.) Research Center.
Two teams—one composed of Katine, Rippard, and other NIST researchers and the other including Mancoff and his Freescale colleagues—have tested pairs of spin-torque structures. The groups independently observed that oscillations of structures separated by only a few hundred nanometers became synchronized. This phase locking resembles the synchronization of such physical and biological systems as pendulums swaying, planets orbiting the sun, and insects signaling.
The experiments, described in two reports in the Sept. 15, 2005 Nature, showed that the joint output of a pair of synchronized oscillators is approximately four times as great as the power of a single oscillator. For larger numbers of oscillators in close proximity, the power should rise as the square of their number, the NIST group predicts. In that case, an array of fewer than a dozen nano-oscillators, occupying only a few square micrometers of a chip, in total, could produce sufficiently strong signals for practical applications.
In work reported in the Aug. 5, 2005 Physical Review Letters, the NIST team demonstrated that nano-oscillators synchronize with incoming microwaves, suggesting that the devices might be suitable for making directional receivers and transmitters that can pick up or radiate microwave energy in chosen orientations. The technology "could be a way toward nanoscale wireless communications—for instance, from one chip to another chip in a computer," Rippard says. Chips or cards with fewer wires may be simpler to make, denser, and faster than conventional wired circuits, he adds.
Scientists are considering where, beyond MRAM and oscillators, this new form of electromagnetic muscle is leading.
One possibility is a more exotic form of spin-torque-assisted memory. In a U.S. patent issued in late 2004, MRAM pioneer Stuart S.P. Parkin of IBM Almaden Research Center in San Jose, Calif., suggests adorning semiconductor chip surfaces with vertical U-shaped wires, up to 20 micrometers tall, made of a magnetic material. Each wire could store data bits as neighboring blips of magnetization, arranged like beads on a necklace.
A spin-polarized current passing through the wire could shunt the beadlike memory regions back and forth past a read-write head at the U's base. Such a system could achieve data-transfer and storage capabilities that might eventually challenge hard disks, Parkin proposes.
Spin torque might even have a mechanical realization, contends Pritiraj Mohanty of Boston University. He and his colleagues are building a nanoscale I-shaped mechanical balance expected to tilt in response to spin-polarized currents and give a polarized electrical response to being tilted.
Such a "spin battery" could prove very important, Mohanty claims, because a new, electron spin–based approach to information processing, called spintronics, lacks reliable ways to generate spin-polarized currents on demand.
In the Sept. 2, 2005 Physical Review Letters, theorists in Russia, China, and Sweden calculated how much spin current a nanoscale I-beam might detect and generate. "You can get a huge effect—big enough to think about practical devices," Mohanty says.
Now that engineers of the nanoworld can make flows of torque move in and out of structures at will, many more unexpected twists on technology are bound to follow.
One scientist's technological advance can be another's nuisance
Despite the technological promise of spin-torque effects, their impacts on nanostructures are not always welcome. In the hard disk–drive industry, for instance, scientists have identified spin torque as a looming threat. In particular, it's expected to pose a problem for the detectors, known as read heads, that recognize the zeroes and ones of digital data by sensing the orientations of magnetic bits on hard disks.
Each detector consists of stacks of magnetic and nonmagnetic metals. It works by allowing one of its magnetic layers to align with the magnetization of one of the bits on the hard disk. That alignment results in a telltale electrical resistance in the stack.
To deal with the more densely packed disks expected in the next few years, prototype next-generation read heads have to be smaller than those in use today. So, electric-current densities in the prototypes have crossed into the range at which the spin-torque effect is showing up, says Jordan A. Katine of Hitachi San Jose (Calif.) Research Center.
Unabated, the changes to the orientation of the magnetic layer in the head that result from uncontrolled spin currents are now becoming as large as those from the underlying data's magnetic fields are. Consequently, the spurious spin torque–caused signals are swamping the legitimate bit readings. "We don't like it when that happens," Katine adds.
Researchers are investigating ways to cancel the torque without suppressing resistance changes in the head that are necessary to decode stored information. One potential fix would include adding to the read head extra magnetic layers designed to generate electrons polarized oppositely to those that cause the torque. But, Katine notes, "finding the best ways [to counteract spin torque] is still a very active area of research."
Department of Physics
Carnegie Mellon University
Pittsburgh, PA 15213
Robert A. Buhrman
School of Applied and Engineering Physics
212 Clark Hall
Ithaca, NY 14853
Department of Physics and Astronomy
Johns Hopkins University
3400 North Charles Street
Baltimore, MD 21218
1251 Waterfront Place
Pittsburgh, PA 15222
Jordan A. Katine
Hitachi San Jose Research Center
San Jose, CA 95120
Frederick B. Mancoff
Freescale Semiconductor, Inc.
1300 North Alma School Road
Chandler, AZ 85224
Department of Physics
590 Commonwealth Avenue
Boston, MA 02215
Daniel C. Ralph
Department of Physics
536 Clark Hall
Ithaca, NY 14853
William H. Rippard
Electromagnetic Technology Division
National Institute of Standards and Technology
Boulder, CO 80305
John C. Slonczewski
IBM J.T. Watson Research Center
P.O. Box 218
Yorktown Heights, NY 10598