Slow light in photonic crystals

The velocity of light in vacuum, c, is approximately 3 × 108 m s–1, fast enough to make 7.5 round-the-world trips in a single second, and to move a distance of 300 mm in 1 ns. This ultrahigh speed is advantageous for efficient data transmission between two points, whether they are separated on a global scale or on a single chip; however, it also makes control of optical signals in the time domain difficult. Slow light is a technology now being investigated as a means to overcome this problem. In next-generation information networks, path switching of optical packets at network nodes will become very important, and solutions that can perform the task with a high data rate, high throughput, and low power consumption are required. Engineers are now developing photonic routers that exploit all-optical processing to avoid the optical–electronic conversion that introduces a lot of inefficiency. Here, a key device is the optical buffer, a device that temporarily stores and adjusts the timing of optical packets. At present, solutions are based on mechanical variable delay lines and a combination of different delay lines with an optical switch, but these approaches are not ideal owing to their slow response. If the velocity of slow light can be controlled with a response speed much faster than the mechanical method, it could be a solution not only for buffering but also various types of timedomain processing, such as retiming, multiplexing and performing convolution integrals. Control over slow light could also improve the phase control in interferometric modulators and phased-array beam shapers. In addition, slow light offers the opportunity for compressing optical signals and optical energy in space, which reduces the device footprint and enhances light–matter interactions. With enhanced optical gain, absorption and nonlinearities per unit length, numerous optical devices, such as lasers, amplifiers, detectors, absorption modulators and wavelength converters, could be miniaturized. The definition of velocity that is most meaningful in slow-light applications is the group velocity υg, which describes the speed at which a pulse envelope propagates. In general, υg is greatly reduced by a large first-order dispersion arising from an optical resonance within the material or structure. Initially, slow light was generated using extremely strong material dispersion; however, this review discusses dispersion arising from engineered structures, in particular photoniccrystal (PC) waveguides, which offer a promising approach for the on-chip integration of slow-light devices. Photonic crystals are multidimensional periodic structures with a period of the order of the optical wavelength, λ. The research field became active in the late 1970s and 1980s (refs 1–3) with the development of photonic band theory, an optical analogue of electronic band theory, which can be used to compute the dispersion characteristics of light in arbitrary PC structures. The theory predicted the existence of a photonic bandgap (PBG), a frequency band of inhibited optical modes. Since the 1990s, PCs with PBGs have been explored for various device applications4–12. At present, PC slabs, a high-index thin film with a two-dimensional array of air holes surrounded by air cladding, are widely used because of their intrinsic lossless optical confinement and simple fabrication process. The PC waveguide (PCW) consists of a line defect of missing air holes in the PC slab13–23. Light propagates through the defect, confined by total internal reflection in the vertical direction and Bragg reflection, due to the PBG, in the lateral direction. It has been known since 2001 that the strong dispersion in this waveguide generates slow light in the vicinity of the photonic band edge24–35. When discussing a low υg in a PCW, two important optical properties need to be considered: the frequency bandwidth of the effect and higher-order dispersion36. A fundamental limit to the first of these is the delay–bandwidth product (DBP), which affects all approaches to slow light37–39. Although a wide bandwidth is desirable in most applications, it often comes at the price of less delay. The DBP means that the extent to which the group velocity of light is reduced must be balanced with the required bandwidth for the application in mind. Regarding the second issue, the higher–order dispersion that usually occurs in simple slow-light PCWs severely distorts optical signals. This distortion can be eliminated either by combining two PCWs with opposite dispersion characteristics36,40–45, using socalled dispersion-compensated slow-light devices, or by suppressing the higher-order dispersion using so-called zero-dispersion slow-light devices, comprising modified PCWs (refs 46–52) or a coupled-resonator optical waveguide (CROW) based on PC cavities or microrings53–62. It is now possible to slow down short optical pulses using some of these approaches.