IBM Squeezes Magnetic Memory Bit Down To A Dozen Atoms

January 23, 2012 Timothy Prickett Morgan

IBM has revealed one possible contender for data storage that we might see far into the future. In a paper presented by techies from the Almaden Research Center in San Jose, California, Big Blue is showing off that it can encode a bit of data in a dozen iron atoms using antiferromagnetic properties of the atoms.

In the most capacious hard drives used in PCs and servers today, it takes about 1 million atoms to encode a bit of data using disk heads that employ giant magnetoresistance to encode and read data on the platter of a disk drive.

IBM researchers Gerd Binnig and Heinrich Rohrer, working in Big Blue’s Zurich, Switzerland, labs, invented the scanning tunneling microscope (STM) in 1981 to take pictures of atoms. IBMers have spent decades figuring out how to manipulate atoms using modified STM techniques, and these have indeed been employed to create an array of atoms that exploit a special kind of magnetism called antiferromagnetism to control the electronic spin of an array of a dozen iron atoms on a copper nitride substrate to encode a bit.

“The chip industry will continue its pursuit of incremental scaling in semiconductor technology but, as components continue to shrink, the march continues to the inevitable end point: the atom,” explains Andreas Heinrich, the lead investigator into atomic storage at Almaden. “We’re taking the opposite approach and starting with the smallest unit–single atoms–to build computing devices one atom at a time.”

Here’s what an array of eight bits encoding a byte of data looks like under the STM:

And here’s what it might look like if you used IBM’s favorite slogan from founder Thomas Watson, “Think” using 480 atoms:

As I was looking through IBM’s description of the antiferromagnetism breakthrough, I kept thinking, why can’t you just store a bit in a single atom, with the spin one way being a one and a spin the other way being a zero? Maybe it has something to do with how big a cluster of atoms needs to be for the encoded data to be read, or maybe it takes a dozen atoms to encode the magnetism in such a way that it doesn’t degrade over time. As it turns out, if you want to lock in the magnetic encoding, you need a dozen atoms to keep the atoms from wiggling their spins and changing the data. (I am not sure why.)

IBM’s labs in San Jose invented the first magnetic disk, the RAMAC, back in 1956, and it may take many years to figure out how to make an STM disk head to manipulate spin at room temperatures and at the low latency and high-bandwidth that systems of the future will require. In 2010, IBM did figure out how to speed up the STM scanning process by a factor of 100,000. IBM did not speculate on when this research might be commercialized, and quite frankly, I think that is because IBM is not sure if it can be. This particular set of experiments detailed in Science occurred at near absolute zero and an STM is not exactly a small device you can put in your pocket.