The prototype is not yet as stable or reliable as commercial computer memory, and building it would require manufacturers to learn to harness materials other than silicon, the workhorse of computing technology. But the scale of the device dwarfs any electronic circuit previously constructed using nanotechnology.
"We're happy the damn thing worked," says chemist James Heath of the California Institute of Technology, whose group built the device. "Our major goal here was never to just make a memory circuit," he says. "It was to develop a manufacturing technique that could work at molecular dimensions."
The device is "a true tour de force," says nano researcher Charles Lieber of Harvard University, who was not part of the study. "He [Heath] has pushed far beyond previous limits of integration density and bit numbers realized previously in the field of molecular electronics."
Researchers are exploring nano-size electronics systems because silicon circuits cannot be packed with wires at increasing densities—yielding higher numbered Pentium processors—forever. Eventually, electrons will start seeping between wires and lithography techniques for stamping out silicon circuits may reach their physical limit.
The Caltech group combined two approaches: molecular electronics (transistors made of molecules) and nanowire crossbars, which are perpendicular junctions of ultrathin wires. To make their device, the team laid down a tightly packed series of 400 parallel silicon wires (separated by just 33 nanometers) and coated them with a layer of barbell-shaped rotaxane molecules. They created a grid of wires by covering the molecule layer with 400 more platinum wires, resulting in groups of molecules sandwiched between each node formed by the crisscrossed wires.
To switch between 0 and 1 the researchers applied a voltage across a group of molecules at a node, which toggled the molecules between two states. The rotaxane molecules each contain a ring around the "handle" of the barbell. A voltage applied across the molecule caused the ring to slide up or down, changing the electrical conductivity of the molecule.
The wires were so crowded that the team could not build conventional electrodes capable of electrifying only two wires at a time (those that define a node); instead they switched the junctions on and off in groups of nine.
One indication that more work needs to be done is that the junctions routinely broke down after being switched more than about 10 times, says Jonathan Green, a physics and chemistry Ph.D. candidate in Heath's lab and first author of the report published online January 24 in Nature. The molecules, he adds, spontaneously flipped back to their previous state after nearly an hour, which is another limitation for a memory device. Commercial flash memory is stable for up to years, he says.
The molecules were also slow to switch between states. Green says that although this time can probably be improved, the speed of such a memory circuit would not come from switching one junction at a time. Instead it would result from switching many junctions at once. "The dominant criticism," he says, "is this is a nice laboratory demonstration, but how will this fit into the real world." His response: "You have to put in the science before you can get the technology."