Although the Mach-Zehnder modulator in the previous report boasts a tiny footprint and avoids the necessity of delicately balancing environmental parameters to maintain a resonance, it requires a hefty, multivolt drive signal that cannot easily be supplied from CMOS circuitry monolithically integrated with the modulator itself. Because integrated drive circuitry is critical to the development of on-chip optical interconnects, William M.J. Green and his colleagues at the IBM Thomas J. Watson Research Center in Yorktown Heights, N.Y. — the same group that demonstrated the Mach-Zehnder modulator described in the report preceding this one — have developed a microring modulator whose sensitivity is 25 times greater than that of the most sophisticated previous microring modulators. Their device potentially can be driven by a signal as small as 0.12 V, which could be supplied readily by a CMOS circuit.Figure 1. (a) A conventional microring modulator consists of a ring resonator evanescently coupled to a waveguide. The transmission through the straight waveguide can be modulated by tuning the ring into or out of resonance with the light in the waveguide. When the ring is resonant, light is coupled from the waveguide into the ring, where it dissipates. To distinguish the microring from another microring introduced in (c), the IBM scientists dubbed it the “amplitude resonator.” (b) An alternative to tuning the ring’s resonance is to fabricate it so that it is always resonant and to tune the coupling between the waveguide and the ring. Here a Mach-Zehnder interferometer (yellow highlight)can be adjusted so that the light is coupled either into the output waveguide or into the microring. (c) A second microring, the “phase resonator,” introduces a highly nonlinear phase shift to light traveling in the upper arm of the Mach-Zehnder. The drive voltage necessary to produce on/off modulation is ∼25 times lower than the voltage required to produce the same effect in (b). Images reprinted with permission of Optics Express.A conventional microring modulator consists of a microring resonator adjacent to a straight waveguide (Figure 1a). When the light traveling in the waveguide is not resonant with the ring, the transmission (|t|2) through the waveguide is near unity. However, if the index of the ring is changed by inducing free carriers in it, it can be tuned into resonance with the light in the waveguide. In that case, light is coupled into the ring, and the transmission through the waveguide is significantly reduced. In other words, the transmission through the waveguide is modulated by inducing free carriers in the ring. In practice, a p-i-n junction often is fabricated around the ring to sweep out free carriers and to increase the modulation speed.Figure 2. The scientists fabricated both of the silicon-on-insulator modulators shown in Figure 1c (a) and Figure 1b (b). The red circle indicates the spot that the scientists illuminated with an argon-ion laser to induce free carriers. The inset in (b) is a scanning electron micrograph of the 3-dB multimode interference coupler.One problem with this design is that the evanescent coupling between the ring and the waveguide is very sensitive to fabrication errors. An alternative was developed by Amnon Yariv and his colleagues at California Institute of Technology several years ago. In this design, the ring is fabricated to be in resonance with the light in the waveguide, and the coupling (|κ|2) between the two is modulated with a Mach-Zehnder interferometer (Figure 1b). Because the microring’s resonance enhances the overall sensitivity, the Cal-Tech modulator can be driven with a voltage one-tenth of that required for a similar extinction ratio in a simple Mach-Zehnder.The wrinkle that has been added by the IBM scientists is a further reduction of the required drive signal by adding a second microring to the scheme (Figure 1c). Because this second microring imposes a strongly nonlinear phase shift on light traveling in the upper arm of the Mach-Zehnder, the drive voltage required to produce on/off modulation in the device is reduced by a factor of 25 from the voltage required for on/off modulation of the modulator in Figure 1b. Figure 3. The modulator shown in Figure 2a had an ∼60-ps fall time, but because the scientists made no attempt to sweep the free carriers out of the ring resonator, the rise time was much slower. It has been shown elsewhere that a p-i-n junction fabricated across the ring can reduce the rise time dramatically, and that such an improvement here could lead directly to 10-Gb/s modulation speeds. Experimentally, the IBM scientists fabricated both of the sensitive modulators shown in Figure 1 (Figure 2). They induced free carriers in the microring by illuminating it with light, rather than applying an electrical voltage, and they made no attempt to fabricate a p-i-n junction to sweep out the carriers. As a result, the carriers lingered several nanoseconds after the light was switched off, producing asymmetric modulation in the time domain (Figure 3).In a practical implementation of the scheme, however, it should be fairly straightforward to include a p-i-n junction around the ring and to achieve speeds as high as 10 Gb/s. Even higher speeds might be obtained by compromising the modulator’s sensitivity, and the trade-off could be maximized for a given application.Optics Express, Dec. 10, 2007, pp. 17264-17272.