Portland, Ore. IBM Corp.'s progress in characterizing the magnetic spin of individual atoms and in flipping them from "up" to "down" could lead to molecular-cascade memories, a new type of memory chip that would pack a bit of data in every atom. IBM Fellow Don Eigler's group at Almaden Research Center (San Jose, Calif.) recently demonstrated IBM's new nanoscale characterization method, dubbed "spin-flip spectroscopy."
To study how to switch the spin of individual atoms, IBM constructed a new type of measuring device. It combines a scanning tunneling microscope with a superconducting coil providing a high-strength magnetic field. The whole machine is supercooled to near-absolute zero.
"We invented spin-flip spectroscopy so that we can study how to use magnetic spin for information storage, because at IBM our ultimate goal is the ultimate memory density possible storing bits on individual atoms," said Andreas Heinrich, a researcher in Eigler's lab. "For instance, we demonstrated our molecular-cascade memories two years ago, but at that time we didn't have a way to reset them they just fell over like dominoes, then we had to pick them back up one by one. Now we think we can use magnetism to reset a future version of molecular cascades."
Eigler and Heinrich performed the work with IBM researcher Christopher Lutz and Jay Gupta, an assistant professor at Ohio State University.
With the machine, IBM was able to characterize the precise amount of energy required to flip the spin of an atom from up to down which is usually encoded to mean "1" and "0." The result was 0.0005 electron-volts, some 10,000 times less than the energy of a single molecular bond.
Keep 'em coming
"There is only one machine like it today, but our results have been so promising that we will probably have four more like it built by this time next year. We build-in a 7-Tesla superconducting magnet and cool it to 0.5 Kelvin 10 times colder than our previous molecular-cascade demonstration," said Heinrich.
Next the group wants to characterize a complete albeit very tiny magnetic circuit, so it can perfect the single-atom spin-flip. After that, the group plans to integrate it into a new type of molecular-cascade memory that can serve as a prototype for real memory chips.
"We are still two or three steps away from a molecular-cascade memory that we can reset, but each step is important. Our next step will be to build a nanoscale magnetic structure and study its properties with spin-flip spectroscopy," said Heinrich.
Spintronics is the name being touted for chips that store information on the spin of atoms or electrons. Most demonstrations so far have used large ensembles of many atoms or electrons, rather than individuals, but IBM's intense search focuses on harnessing them one by one.
The first step was to characterize how to switch the spin of an individual atom. To do that, the researchers first had to invent a machine that was sensitive enough. Then the machine had to be built, put in a vacuum, cooled to 0.5 Kelvin. The researchers then had to use the machine to isolate a single manganese atom on a surface of aluminum oxide. And finally, they used the scanning tunneling microscope probe tip to send a current through the atom to measure the energy required to flip the spin of a single manganese atom from up to down.
After initial resistance, once the rising voltage crossed a small threshold the electrical current tunneled from the tip through the manganese atom and then across the aluminum oxide into the substrate. When the voltage was increased further, the current smoothly changed until a second threshold was surpassed and the current spiked. The researchers interpreted the second current spike as the moment when the atom switched spins.
"Once we knew the voltage at which the spin flipped, and the current flow increased significantly, it was a very simple conversion to determine that you need about 0.0005 electron-volts to flip a spin," said Heinrich.
The team also learned that it takes 6 percent more energy to flip the spin of an atom if it is positioned near the edge of a the aluminum oxide insulator as opposed to the middle of the chip. By characterizing such details with its new spin-flip spectroscopy machine, and the four others that are being constructed, IBM will be able to provide important details about single-atom spin-flips that can be used to engineer future nanoscale spintronic memories.
For instance, in IBM's molecular-cascade memory (www.eetimes.com/at/news/ OEG20021024S0047), the carbon end of a carbon monoxide molecule, which is naturally attracted to copper, always sticks to the surface with the carbon end toward the copper. But on single-crystal copper substrates, the lattice where CO attaches is pitched slightly closer than the carbon end of CO, like trying to cram tennis balls into an egg carton. To compensate, the CO spontaneously hopped to adjacent grid sites until it was spaced out over every other copper atom its lowest energy state. But IBM was able to line up these atoms in a staggered 0.25-nanometer grid, so that nudging one at the input causes a chain reaction to perform a preset calculation.
Future experiments with magnetic atoms will attempt to harness the electron's spin as the logic value that cascades from atom to atom, by presetting the cascade into its "starting" position with a magnetic field. Then its preset calculation could be performed by transferring spin from one molecule in the cascade to the next, rather than have the entire molecular cascade "fall down" like dominoes. After such a magnetic molecular cascade performed its preset calculation, it could then be reset by applying the magnetic field to reset it.
"Besides molecular cascades there are other geometries we also want to try out," said Heinrich, although "molecular cascades look the most promising right now. But first we will build and study a small magnetic structure to learn more details about how we can control magnetic spin in individual atoms."
Earlier this year, IBM Almaden scientists described a breakthrough in nanoscale magnetic-resonance imaging in which magnetic-resonance force microscopy was used to directly detect the faint signal from a single electron spin inside a solid.
IBM also separately announced a joint venture with Stanford University called the IBM-Stanford Spintronic Science and Applications Center, or SpinAps. There, the partners will study reconfigurable logic devices, room-temperature superconductors and quantum computers that use the spin of atoms and electrons as logic values.