All-optical routing of digital data in fiber optic telecommunications systems will require optically activated switches capable of directing incoming optical pulses into one of two or more output fibers (see “Toward Optical Packet Switching,” page 84). But not just any optically activated switch will do. The switch must be able to read the header of an optical packet and route the packet into the appropriate fiber. Recently, researchers at Cornell University in Ithaca, N.Y., demonstrated just such a switch.Figure 1. Transmission through the straight waveguide shows a sharp dip at the microring’s resonant wavelength. Images ©OSA. The switch is based on a 5-µm-radius microring resonator coupled to a straight waveguide, fabricated in silicon on a silicon-on-insulator substrate (Figure 1). Light that passes through the straight waveguide will resonate in the microring if its wavelength fits an integral number of times around the ring’s circumference. When light is resonant, it is coupled into the ring, where it circulates repeatedly until it eventually is lost because of scattering and absorption in the ring. Thus, the spectrum of light transmitted through the straight waveguide shows a sharp dip at the ring’s resonant wavelength.However, the ring’s refractive index and, hence, its resonant wavelength can be shifted by free carriers (electrons and holes) generated by two-photon absorption of light passing through the waveguide. Suppose that the light in the waveguide is initially slightly shorter than the resonant wavelength. A limited amount of light will be coupled into the ring, but it will generate some free carriers and shift the resonance to a shorter wavelength. More light will be coupled into the ring, creating more free carriers and shifting the resonant wavelength still shorter. This process ultimately will become stabilized because fewer free carriers are generated when the resonant wavelength shifts below the wavelength of the light.Figure 2. A hysteresis loop in the input-output curves of the microring resonator provides the mechanism for switching. An explanation of the mechanism and of the labeled points on the curves is in the text.This dynamic leads to hysteresis in the input-output transfer function for the device (Figure 2). As the input power increases from zero, it must reach a threshold (P1) before the positive feedback takes effect. But when the input power decreases from a higher level, the positive feedback effect is already active, and the input power must diminish to a lower level (P2) before the feedback becomes ineffective. Thus, there are two stable conditions for the ring with an input of ~6.2 mW: If the 6.2-mW input is preceded by a low input, the transmission is relatively high (A), and if the input is preceded by a high input, the transmission is low (B). The hysteresis provides the basis for optical switching. If an ~6.2-mW pulse is injected into the straight waveguide, it is transmitted relative-ly intact past the microring (Figure 3a). However, if the same pulse is preceded by a brief high-power pulse, it is dramatically attenuated (Figure 3b). Thus, the header of an optical packet could determine the fate of the packet.Figure 3. The hysteresis enables the optical switching capability. The transmission of a pulse through the straight waveguide depends on whether it was preceded by a higher-power pulse.The Cornell researchers performed their experiments with 4-ns pulses to avoid thermal effects in the ring. They have calculated the deleterious thermal effect that would be introduced by optical heating of the microring and have concluded that the bistability would be disabled in the steadystate. They propose that fabricating the ring with an inherent strain could counterbalance the thermal effect and lead to steady-state implementation of the optical switch.