Imagine a hard drive that doesn’t spin. In one scheme for increasing computer data storage and speed, an electric current would push magnetic regions along a wire instead of the computer relying on the physical motion of a disk to read data. In the 4 May Physical Review Letters, a team demonstrates that they could push so-called magnetic domain walls at 110 meters per second–100 times faster than ever before–by using nanosecond pulses of electric current. But the bad news is that the walls sometimes move much slower–or not at all–as they become stuck on imperfections in the wire.
The atoms of a magnet act like tiny bar magnets, each with its north pole pointing in the same direction. But a magnetic material can also have many regions, each with a different collective alignment of atoms. These magnetic “domains” are separated by domain walls, thin slices within which the atomic magnets change orientation. Computers store data on a hard disk by creating tiny domains magnetized to the left for a “one” and to the right for a “zero,” for example. Reading the data requires spinning the disk underneath a magnetic sensor. But this retrieval is very slow compared with integrated circuits, where memory is stored electronically. On the other hand, electronic data disappears when switched off, while magnetic storage is more permanent–and about 100 times less expensive.
One could have the best of both worlds if magnetic data could be moved across the magnetic sensor electronically, rather than mechanically. In 2004, Stuart Parkin of IBM Almaden Research Center in San Jose, California, patented the magnetic “racetrack” concept, in which the bits would slide sequentially down a thin magnetic wire that could be tightly coiled inside a chip. In Parkin’s concept, the domain walls represent the bits, rather than the domains between them–“one” for a domain wall; “zero” for no domain wall. The domain wall motion could be driven by spin-polarized currents, in which the moving electrons have their spins aligned. These currents can rotate the atomic bar magnets in the domain wall, creating a sort of domino effect in which the wall tumbles forward. Theory predicts the rate at which domain walls move, but when real domain walls were subjected to pulses of polarized current, experimenters measured velocities 100 to 1000 times slower than expected.
“My feeling is that previous experiments saw slower speeds because they measured for too long,” says Guido Meier of the University of Hamburg. He and his colleagues shortened the length of the current pulses from microseconds to nanoseconds to reduce the chances that a wall would get stuck on imperfections in the crystalline structure during its brief motion.
The team used a single, 3-micron-wide domain wall confined to a thin wire of permalloy, a magnetic material made of iron and nickel that is widely used for disk drives. The researchers tracked the location of the domain wall with 15 nanometer resolution using polarized x-ray images taken before and after the current pulse. They recorded speeds of up to 110 meters per second, just as theory suggested.
However, many of the pulses gave smaller speeds, or no movement at all, when the domain walls hit crystal imperfections. The distance a domain wall could travel unabated was anywhere from zero to about 1 micron.
This is “mixed news for applications,” says Mathias Kläui of the University of Konstanz in Germany. Ideas like the magnetic racetrack will benefit from such high speeds, but the random nature of the domain wall jumps “makes reliable switching a challenge,” he says. Kläui hopes that changing the geometry of the wires in some way might give a more stable flow.