A New Kind of Nanophotonic Waveguide

A new kind of nanophotonic waveguide has been created at MIT, overcoming several long-standing design obstacles. The resultant device might lead to single-photon, broadband and more compact optical transistors, switches, memories, and time-delay devices needed for optical computing and telecommunications.

If photonics is to keep up with electronics in the effort to produce smaller, faster, less-power-hungry circuitry, then photon manipulation will have to be carried out over scales of space, time, and energy hundreds or thousands of times smaller than is possible now. One or two of these parameters (space, time, energy) at a time have been reduced, but until now it has been hard to achieve all three simultaneously. John Joannopoulos and his MIT colleagues have succeeded in the following way. To process a photonic signal, they encrypt it into light waves supported on the interface between a metal substrate and a layer of insulating material. These waves, called surface plasmons, can have a propagation wavelength much smaller than the free-space optical wavelength. This achieves one of the desired reductions: with a shorter wavelength the spatial dimension of the device can be smaller.

Furthermore, a subwavelength plasmon is also a very slow electromagnetic wave. Such a slower-moving wave spends more time "feeling" the nonlinear properties of the device materials, and is therefore typified by a lower device-operational-energy scale, thus achieving another of the desired reductions. Finally, by stacking up several insulator layers, the slow plasmon waves occupy a surprisingly large frequency bandwidth. Since the superposition of waves at a variety of frequencies can add up to a pulse that is very short in the time domain, the third of the desired scale reductions is thereby achieved.

Reducing energy loss is another great virtue of the MIT device. The plasmons are guided around on the photonic chip by corrugations on the nano-scale. In plasmonic devices the corrugations have usually been in the metal layer; this has always led to intractable propagation losses. However, in the MIT device they reside in the insulator layer; this, it turns out, allows for a drastic reduction of the losses by cooling.
(Karalis et al., Physical Review Letters, 5 August; contact Aristeidis Karalis, aristos@mit.edu)