The speed of light
- 29 November, 2000 12:01
As the limitations of miniaturisation appear to have been reached for today's electronic computers, researchers are trying to push beyond them by substituting light for electrical voltages in computer components.
"What we are accomplishing in the lab today will result in the development of super-fast, super-miniaturised, super-lightweight and lower-cost optical computing and optical communication devices and systems," says Donald Frazier, a senior scientist for physical chemistry at NASA's George C Marshall Space Flight Center.
Here, scientists are working on solving a variety of problems that must be overcome before digital optical computing can be realised.
But will these superfast, super-small digital optical computers have general applications? Or is that the wrong question to ask?
Just as an earlier era saw superhighways built with more capacity than anyone imagined would be needed - and where traffic now idles for miles in smoggy jam-ups - once digital optical computers are built, applications will follow.
Using light instead of electrical voltages to perform computations and communications, digital optical computers are said to promise switching speeds and parallelism that will swamp the capacity of today's massively parallel computers and could eventually put that kind of computational power on desktops, if not in handheld devices.
They'll be in satellites managing the ever-expanding demands of communications, and they'll be aboard long-term space flights, says physicist Hossin Abdeldayem, also at NASA's Marshall Space Flight Center. The need for the power of optical computing is already being met through programs that do complex modelling, such as those used for weather prediction, he says.
Computations that would take 11 years with conventional electronic computers will take only one hour with optical computers, Abdeldayem adds.
The need for this kind of speed already exists, he says, though it will take another 10 years before computers using all-optical digital technology are on the market. Abdeldayem says he's worried that Japan and Europe are investing more heavily than the US in the research that needs to be done.
The first stage in the movement toward an all-optical world will probably be hybrid electro-optical computers. On-chip miniature lasers and detectors controlled by electronics can already be manufactured. The optics will handle the communications, with light traveling along fibres or films; the advantage over electrical communications is that light waves don't generate cross-talk or require insulation. Frequencies can be multiplexed to easily achieve parallelism.
In the world of massively parallel computers, designers are working on free-space backplanes in which optical signals make the connections. Unlike electrical signals, light signals can cross paths without affecting the information that's received at their destinations, and information can also be multiplexed, with possibly as many as 1000 separate channels in a single pulse.
In communications, optical switching devices have already passed the proof-of-concept phase. It's no longer necessary to translate between optical and electrical signals at every switch. For the burgeoning Internet, this optical processing means terabit speeds are possible.
But the components that are crucial for all-optical digital computers are still in the design phase. Logic gates and bistable devices (or flip-flops) which work without the intervention of electronics, haven't been perfected yet.
Achieving the nonlinear behavior needed for all-optical logic gates and bistable devices still requires a great deal of energy. The amount of power required in the laser pulses needed for optical computing, while feasible in the laboratory, isn't a possibility for a miniaturised computer.
Scientists trying to solve these problems are concentrating on organic materials, some of which exhibit strong binary - rather than linear - transitions and fast switching speeds. The switching speeds are important because a computer can operate no faster than the switching speed of its underlying substrates.
Groups at Brown University and IBM's Almaden Research Center have reported achieving 100 picosecond rates. At the Marshall Space Flight Center, Abdeldayem is using organic films only one micron thick, driven by laser pulses, to run at pico- and femtosecond rates.
Key logic gates
Abdeldayem has designed an "and" gate, one of the basic logic gates used by computers, and he reports being close to producing a "nand" gate, which would be even more significant. All of the Boolean logic used by computers can be created out of nand gates.
Another issue also remains: which aspect of light would be the best solution for creating the ones and zeros that are the lifeblood of computer logic? Abdeldayem's organic switches provide an "on" or "off" transmission state. But so far, researchers haven't identified a particular organic material that works best.
Some scientists have suggested that it might be better to use the direction of the polarisation of light for ones and zeros, with polarisation in one direction meaning "one" and polarisation in the other direction meaning "zero". They argue that then both states would involve a flow of energy, so the result of one operation could easily cascade down to the next logic gate in a sequence.
Long-term memory is also an active field of research. Scientists are looking at storing data in holograms or crystals.
Big applications that are already pushing the limits of today's capacity will probably be the first to benefit when all-optical digital computing is finally realised. What else will follow when optical computing matures as a technology is anyone's guess. Who could have predicted 50km-long traffic jams and the Web?