Microdisk and microring resonators have been widely studied because they can perform many of the basic functions required in photonic circuits and because they are much smaller than conventional components. In virtually all of the studies to date, however, the coupling between the resonator and its input/output waveguides has been fixed.Recently, researchers at the University of California, Los Angeles, demonstrated a variable coupling scheme. Such a device -- a tiny, wavelength-selecting switch -- can be used to build tunable optical delays, dynamic dispersion compensators and reconfigurable add/drop multiplexers.Figure 1. When the coupling waveguides are straight, no light is coupled from the waveguide into the microdisk (a). Coupling is accomplished when an electrostatic force pulls the waveguides into near-contact with the disk (b).The device is a 20-µm-diameter microdisk between two waveguides, fabricated on a silicon-on-insulator wafer (Figure 1a). The suspended portions of the waveguides are 0.35 µm wide, 0.6 µm thick and 150 µm long. When a voltage is applied to the electrodes between the waveguides, the waveguides are pulled inward toward the microdisk by electrostatic force (Figure 1b).In the absence of a voltage on the electrodes, there is a 1.4-µm gap between the waveguides and the disk, and there is virtually no coupling between them; any light injected into a waveguide passes through the waveguide and exits on the other side. When the waveguides are pulled to within ~0.1 µm of the microdisk, almost all the light in the waveguide is evanescently coupled into the disk and transferred to the other waveguide (Figure 2). The gap between the waveguides and the microdisk can be continuously reduced from 1.4 to 0 µm without any mechanical instability.The researchers patterned the device with one photomask. Laterally coupled microdisks like this typically require electron-beam lithography to ensure the precision of the disk-waveguide spacing. In this case, the spacing was adjustable, and standard optical lithography provided sufficient precision.Figure 2. The calculated coupling between the waveguides and the microdisk depends on the distance between them. The "through port" can be considered the near waveguide in Figure 1, and the "drop port" the other. ©2005 IEEE.A microdisk resonator is a wavelength-selecting device because only a resonant wavelength -- that is, one that can fit exactly an integral number of times around the circumference of the disk -- can be coupled into it. The calculation illustrated in Figure 2 assumes that the light in the waveguides is resonant with the disk. Any other wavelength would show no coupling as the gap diminished from 0.6 to 0 µm.In their experimental demonstration, the researchers activated only the input waveguide. When they injected relatively broadband spontaneous emission from an erbium-doped fiber amplifier into the waveguide, they observed 100 percent transmission when no voltage was applied to the microelectromechanical-systems-based electrodes (Figure 3). As they increased the voltage (and reduced the gap between waveguide and disk), the disk's wavelength selectivity be-gan to appear in the spectrum of the transmitted light, until the peak resonance was attenuated by 9 dB. Figure 3. As the gap between the waveguide and the microdisk shrinks, spectral features appear in the light transmitted through the waveguide. When the waveguide is pulled to within 90 nm of the disk (by the electrostatic force caused by 61 V applied to the electrodes), light at the peak disk resonance is attenuated by 9 dB. ©2005 IEEE.The multiple peaks that appear in Figure 3 are due not to azimuthal modes around the circumference of the disk, but to the oscillation of transverse modes in the vertical direction. The researchers believe that these could be eliminated by reducing the thickness of the microdisk. They also predict that power transfer could be further increased by enhancing the Q-factor of the microdisk.