Racetrack Memory Charges Ahead

From MIT Technology Review, racetrack memory is making experimental progress towards surpassing flash memory.

Racetrack memory stores data in vertical nanowires, it can theoretically pack 100 times as much data into the same area as a flash-chip transistor, and at the same cost. There are no mechanical parts, so it could prove more reliable than a hard drive. Racetrack memory is fast, like the dynamic random-access memory (DRAM) used to hold frequently accessed data in computers, yet it can store information even when the power is off.

When Parkin first proposed racetrack memory, in 2003, “people thought it was a great idea that would never work,” he says. Before last April, no one had been able to shift the magnetic domains along the wire without disturbing their orientations. However, in a paper published that month in Science, Parkin’s team showed that a spin-polarized current would preserve the original magnetic pattern.

The Science paper proved that the concept of racetrack memory is sound, although at the time, the researchers had moved only three bits of data down a nanowire. Last December, Parkin’s team successfully moved six bits along the wire. He hopes to reach 10 bits soon, which he says would make racetrack memory competitive with flash storage. If his team can manage 100 bits, racetrack could replace hard drives.

In one implementation of racetrack memory, information is stored on a U-shaped nanowire as a pattern of magnetic regions with different polarities. Applying a spin-polarized current causes the magnetic pattern to speed along the nanowire; the data can be moved in either direction, depending on the direction of the current. A separate nanowire perpendicular to the U-shaped “racetrack” writes data by changing the polarity of the magnetic regions. A second device at the base of the track reads the data. Data can be written and read in less than a nanosecond. Racetrack memory using hundreds of millions of nanowires would have the potential to store vast amounts of data.