Electrons are the trusty foot soldiers of electronics, dutifully carrying the electric charges that enable our everyday activities, from making mobile phone calls to listening to music to—ahem—reading online news stories.

But there is more to the humble electron than just shunting electric charge from one place to another. Electrons also have a property known as spin; an electron can "spin up" or "spin down," pointing like a tiny magnetic compass needle in one of two directions. Because that spin can be manipulated, and because the two spin states correspond neatly to the 0s and 1s of digital bits, researchers in the relatively young field of spintronics have been working to encode information in the spin of electrons, thus taking fuller advantage of the particle's natural data-carrying ability.

But electron spins tend to have quite brief usable lifetimes; even in optimal conditions—in pure samples held just a hair above absolute zero—the information encoded on an electron spin is lost on timescales of seconds, if that. Now researchers at the University of Utah, The Florida State University, University College London and the University of Sydney in Australia report a way to extend that informational lifetime to more than 100 seconds by encoding an electron's spin onto the much longer-lived spin of an atomic nucleus, which can then be read out electronically. The study appears in the December 17 Science.

Like many spintronics researchers, University of Sydney physicist Dane McCamey and his colleagues targeted electrons of phosphorus atoms trapped in silicon. Phosphorus-doped silicon is a promising medium for spintronics because each phosphorus atom "donates" an extra electron that orbits rather freely and hence is open to manipulation in the silicon crystal. And silicon, which already forms the backbone of conventional information-processing devices, is a medium that would allow easy interfacing with extant electronics. "We want to develop a memory element for spintronics that is also compatible with conventional silicon-based electronics," McCamey says.

The group did not have to look far to find its spintronic memory, because although the spin of a phosphorus donor electron has a short lifetime, the spin of the phosphorus nucleus is rather robust. "The nuclear spin is much more immune to the environment," McCamey says.

Starting with an ensemble of spin-down nuclei, the researchers used a specially tuned radio-frequency pulse to make a sort of logic gate: if the electron's spin is down, the nucleus remains unaffected; if the electron's spin is up, the nuclear spin is flipped up as well. The result is that the electron spins and nuclear spins match—the spin information has effectively been encoded in a longer-lived nuclear memory. So even if the electron itself loses its memory, its spin contents have been preserved via proxy.

The information stored in the phosphorus nuclei can then be read electronically. With another pulse, the researchers selectively flip the electron spins of only those nuclei whose spins are oriented in one particular direction. And that flip registers as a change in current passing through the device—when the spin of the phosphorus electrons is opposed to the that of the conduction electrons in the silicon sample, the phosphorus donors can capture a conduction electron and cause a measurable dip in the current. If the spin of the phosphorus electron and the conduction electron are the same, the latter cannot be captured and proceeds through the silicon unabated. By tracking the changes in current, the researchers can infer the pointing of the nuclear spins.

At the moment, long-term storage has only been demonstrated for large numbers of phosphorus donors—there may be 10 billion phosphorus atoms in the device used for the new study—but McCamey says the technique should apply to the smaller scales needed for practical spintronics. "We like to use big samples to do things first," he says. "But I don't see any physical reason why we can't do this with much, much smaller ensembles of atoms."