Kwok, a program manager with the Defense Advanced Research Projects Agency, said he plans to have a working nanoscale computer memory by 2004. If so, one square centimeter's worth of the device Kwok envisions could hold more than 10 gigabytes of data, enough for several full-length movies.
Nanotechnology is the ability to manipulate matter on the atomic or molecular scale. In the case of computer memory, staggering storage capacity such as envisioned by Kwok would emerge from using individual molecules as electrical components, or "moletronics." The work is much more specific than even the area of nanoelectronics, which studies advanced uses for polymers and other organic compounds, he said.
"We are focusing on a very specific group of molecules that are small and they 'switch,' so we can build computing circuits with them," Kwok told United Press International at the DARPATECH 2002 conference in Anaheim, Calif., last week. "Since we're so focused, we are able to accelerate the development of this technology from perhaps a 40-year span to a few years to 10 years or so."
The DARPA team is experimenting with moletronics applications conjunction with wires only a few atoms in diameter, Kwok said. The scale makes them difficult to manipulate. At an equivalent diameter of an inch, the wires would be as long as a football field. The most promising technology so far is using the nanowires in an array, instead of linking them end-to-end, he said.
"When we have two-terminal molecules as we do now, you want to have parallel wires on the bottom," Kwok said. "Then ... flow a (single-atom-thick) layer of molecules on top of the wires. That leaves the (storage) molecules sticking up, then you cross them (with perpendicular wires) on top and you have the two terminals connected."
Another nanoscale approach involves structures called quantum dots, which isolates a small number of atoms to achieve unique physical properties, he said. "The quantum dot area is part of the moletronics program. We have one team of researchers at Notre Dame developing that technology. They're making progress and doing excellent work, but they're not at the (nanowire array) maturity level."
The team's 2004 research target involves creating a working memory device that stores 100 billion bits, or gigabits, per square centimeter. The best available hard disk drive -- today's state-of-the-art storage leader -- packs about 15 gigabits into a square inch. Hard drive researchers have demonstrated the ability to store 100 gigabits per square inch, but even that capacity would fall four or five times short of Kwok's nanoscale memory.
Perhaps more relevant for the device's possible uses is a comparison with dynamic random access memory, or DRAM, employed by today's computer processors to hold the information currently being used.
"The difference is that the molecules retain the information a lot longer than the DRAM can," Kwok said. "You don't have the parasitic resistance and corresponding power requirements with the molecules, so you wouldn't need to refresh them as much."
Although the computer chipmaking industry already has mapped a DRAM development strategy for the next decade or so, Kwok said, his team already has built nanoscale devices that exceed the industry's storage density goals for 2008.
Nanomemory's most likely immediate use would be in cache memory to speed up the performance of processors such as Intel's Pentium, Kwok said. Today's top-level consumer processors contain a few hundred thousand bytes of cache, but at a high cost in physical space, increased power consumption and associated heat build-up.
In contrast, Kwok said, nanomemory easily holds hundreds or thousands of times more data and would present several advantages. For example, if chipmakers merely want to maintain performance gains, they easily could employ nanomemory to make smaller, cooler and more power-efficient processors. Using the device's full capabilities would boost performance radically by reducing or eliminating the processor's need to wait for data from long-term storage.
Nanomemory also could make "computer boot-up" an obsolete term. Today, computers must go through a lengthy start-up process that includes retrieving the operating system and other information from a disk drive. Nanodevices easily could hold that much data using only a trickle of electricity, allowing "instant-on" computers, video games and other devices.
Kwok's vision for nanomemory does not end with computers. When moletronics principles become commonplace, the ability to store information could be extended to objects as ordinary as tablecloths, he said. This could revolutionize manufacturing by storing a device's designs in its own materials, so that any factory capable of reading the information could make a perfect copy.
Such wonders will have to wait for more mundane work, however, because nanomemory still has at least one technological hurdle to overcome. For the molecules to store information, an essentially mechanical process is involved that takes about a microsecond, Kwok said. Waiting for this process means data must be "written" into the device at speeds far slower than today's solid-state or hard-drive memory. But Kwok stressed several times his team is focused on improving that performance.
"Once (the information is stored) and I want to go back in and see what state it's in, I can use a vibrational circuit, an oscillator, and interrogate the molecules," Kwok said. "That can be done at about 10 gigahertz."
Today's best processors only run at a few gigahertz, while steady increases in computing speed could approach 10 GHz by the time Kwok hopes to have nanomemory ready for industrialization. If that proves to be the case, the memory would feed the processor data as fast as the circuits could use it.
Soon nanomemory will have to be transferred from the lab to the real world, so computer and electronics makers such as Hewlett-Packard and Motorola are watching the device's promise, Kwok said, and they are involved in the research.
"They want to make sure that for any production line they build, the physics behind it is sound," Kwok said. "All the shock tests and interference tests (necessary for nanomemory's commercial use), believe me they will go through those."