If photonic circuits are ever to compete with their electronic counterparts, strong confinement of light waves coupled with low propagation losses is needed. A new class of waveguides offers both.
Miniaturized circuits that use light to carry digital information would be inherently faster than conventional electronic circuits, and have a capacity thousands of times greater. But there's a snag: the development of practical, small photonic components is impeded by the diffraction limit — the fact that light will spread out on passing through any region narrower than its wavelength. On page 508 of this issue1, Bozhevolnyi et al. flag a new route around this obstacle. They present the first components that guide and manipulate light in the form of so-called channel plasmon– polaritons. These guide the light along the bottom of sub-wavelength V-shaped grooves, milled in a metal film, without high propagation losses.
Channel plasmon–polaritons are young members of an extended family known as the surface plasmons. These are electromagnetic waves that originate in the collective excitation of free electrons at the interface of a metal and an insulating dielectric, such as air. Surface plasmons remain tightly bound to the interface: a plasmon of an optical wavelength — between about 400 and 750 nanometres — penetrates around 10 nm into the metal and decays over a few hundred nanometres in the dielectric.
Surface plasmons thus concentrate light in a volume less than its wavelength across. They can also be used to transmit electromagnetic signals: for the near-infrared wavelengths around 1.5 micrometres, typically used in telecommunications, the propagation length of a plasmon at a planar gold–air interface is about a millimetre, and therefore long enough to connect two devices on a chip optically. The use of surface plasmons is also compatible with available planar electronics technology, also offering the possibility of transporting optical signals and electrical current on the same substrate.
But to create miniature photonic circuits, surface plasmons have to be confined not just in the direction perpendicular to the interface, but also in the plane of the inter-face, so that they can propagate efficiently through narrow metal strips. This has proved problematic: when a propagating plasmon is squeezed from the sides, its propagation length is severely reduced2. The several strategies deployed to circumvent this problem all represented an imperfect trade-off between sub-wavelength lateral confinement and propagation length.
In short, a new actor was needed: enter the channel plasmon–polariton (CPP). The fundamental idea of guiding light along the bottom of milled V-grooves in a planar metal surface was proposed3 15 years ago and subsequently refined4. Compared with other plasmons, these channel modes are laterally strongly confined and suffer only low propagation loss. Extensive numerical simulations5 have confirmed this, and also showed that so-called single-mode operation — meaning that, at a given wavelength, energy (information) is transmitted at one speed — can be attained in such waveguides simply by adjusting the depth of the grooves. (Single-mode operation is advantageous where optical interconnectors are used, because in multi-mode operation, light can jump between modes at a junction, provoking a distortion in the shape of the light pulse.) That finding was followed by the prediction last year6 that transmission of light through a sharp 90° bend in a CPP waveguide was possible almost without loss. Waveguides created in dielectric materials with a ‘photonic band-gap’ (that is, in which light in a certain wavelength range cannot propagate through the bulk structure) could do the same job, but only at the price of a much larger device.
So much for the theory. On the experimental side, too, things have begun to evolve rapidly. Just eight months ago, Bozhevolnyi and colleagues reported7 the experimental achievement of CPP propagation along a straight, V-groove waveguide drilled in a gold film using focused ion-beam milling. Working at telecommunication wavelengths between 1.4 and 1.6 µm, they used a near-field optical microscope to build up a picture of the propagation all along the waveguide. They thus showed that the light was confined to a width of around 1 µm that was less than its wavelength, and had a propagation length of 90–250 µm, depending on the exact wavelength.
In their latest contribution1, Bozhevolnyi et al. fabricate the first CPP-based optical components. The first of these is a ‘Y-splitter’, a junction in which two straight waveguides are connected to a third over a distance of only 5 µm, just over three times the light's wavelength. This feature is of paramount importance for the implementation of miniature optical circuits on a chip.
As a proof of principle, the authors also demonstrate very high performance for a Mach–Zehnder interferometer (in effect, two Y-splitters fork-on-fork that can split and then reunify a light beam) and a functional ring resonator. This latter component (see Fig. 3 on page 510) can — according to the phase difference of the light waves entering and leaving the ring, and therefore the degree of constructive or destructive interference between the two — act as a wavelength filter for the transmitted light.
The fabrication process exploits current planar technology: milling grooves onto a metal film with a focused ion beam is similar to drawing lines on a paper with a pencil. The possibilities for such techniques are huge, but there are still some problems. One is how to get external light into a CPP waveguide and extract it at the other end. Coupling to a standard single-mode optical fibre, as practised by Bozhevolnyi and colleagues1,7, might produce large losses, and other alternatives should be tested. Nevertheless, the successes already achieved in a very short period of experimental research using channel plasmons to mould the flow of light hint at bright prospects ahead.