Graphene memory would have significant advantages over today's magnetic memory. Bits could be read 30 times faster because electrons move through graphene quickly. Plus, the memory could be denser. Bit areas on hard disks are currently a few tens of nanometers across. At densities of 1 terabit per square inch, they will be about 25 nanometers across, too small to hold their magnetization direction. With graphene, bits could shrink to 10 nanometers or even smaller. In fact, the memory devices would work better with smaller graphene areas. Stanford University researchers have shown that cutting graphene into ribbons a few nanometers wide enhances the difference between its two conductivity states.

The new prototype memory devices, however, are rudimentary. The Singapore researchers take graphene flakes that are 2 micrometers wide and place them on silicon. Then they deposit gold electrodes and add a top layer of the ferroelectric. Özyilmaz says that the device readout time is five times faster than current magnetic memory. The researchers can switch graphene between its two conductivities 100,000 times--practical memory devices go through millions of cycles.

This is not the first attempt at making graphene memory. In an August 2008 IEEE Electron Device Letters paper, researchers at the German nanotechnology company AMO described devices that could switch between two conductivity states using an electric field. "We could cycle 20 to 30 times, but not tens of thousands of times," says physicist Max Lemme, lead author of the paper. Lemme speculates that hydroxyl groups and hydrogen attached to the graphene surface detach when current is applied, changing the sheet's conductivity. Why the graphene sheets nonetheless maintain their conductivity when the power is switched off is not well understood.

Geim, who was involved in the AMO work, says that "when you don't know the mechanism, it's hard to judge whether you can in principle make this mechanism reliable to be reproducible on many devices in an identical manner." With the Singapore researchers' approach, however, "we know the physics behind it and its limitations. With well-known fundamentals behind it, it looks like a very good idea."