New Hope for Optical Signal Processing

For decades, researchers developing electronics have had enormous success advancing almost any application that has to do with information processing: following Moore’s Law, data density on an electronic chip has doubled every 18 months. Although this exponential growth is likely to continue for a while, inherent physical limitations are expected to prevent it from lasting indefinitely. Some of these limitations are already evident: as electronics in computers are forced to operate at ever higher frequencies, power dissipation and consequent hardware heating are becoming a very serious problem. In nodes of optical telecommunications networks, where data needs to be processed electronically at especially high operational frequencies, the problem is even more significant.

Realizing that electronic signal processing would eventually face a fundamental physical limitation, engineers in the early 1980s explored the possibility of building an optical computer, in which data would be carried by light (photons) instead of by charged carriers (electrons). They didn’t have an easy time. True all-optical signal processing requires a way of influencing light with light itself. That is, one has to use materials with optical properties that can be modified by the presence of a light signal; this can be used to influence another light signal, thereby performing an all-optical signal-processing operation. Unfortunately, these effects tend to be extraordinarily weak, so the proposed optical logic elements of the 1980s were too large; they consumed orders of magnitude too much power to be feasible. People started viewing optical signal processing as impractical.

Now, however, with the limits of electronics looming much closer, engineers have begun turning to optics again. Indeed, it’s likely that data transport between various components of desktop computers (between different parts of the processor, and between the memory and the processor) will soon be performed optically. As the need grows for physical mechanisms that improve our ability to manipulate light, photonic crystals (which were invented in 1987) have emerged as a promising way to meet it.

Photonic crystals are artificially created nanostructured “meta-­materials” whose optical properties vary periodically at the scale of the wavelength of light. Sometimes called “semiconductors for photons,” photonic crystals offer unprece­dented opportunities for molding the flow of light. For example, they have been used to create all-optical switches that are less than a micro­meter in size and an order of magnitude faster than transistors used in commercial electronics. Moreover, photonic-­crystal designs have been proposed that could enable non­linear interaction even between single photons. These materials could thus dramatically change the view that optical interactions are too weak to use for signal processing.

Optical technologies will keep penetrating deeper into electronic designs, and photonic crystals will play a major part in making this possible. Information processing in the near future will thus probably be performed by hybrid electronic and optical designs, with optics taking on an ever more important role.

Marin Soljacic is an assistant professor of physics at MIT and a member of this year’s TR35.