Optical Switch in Silicon Relies on Photonic Crystal

Breck Hitz

All-optical switching — in which light in a “control” beam switches light in a “signal” beam on or off — is the fundamental building block of such cutting-edge photonics technologies as telecommunications and optical computing. Moreover, optical switches implemented in silicon have the potential advantages of monolithic integration with electronic functionality and of economical CMOS fabrication.

Figure 1. One resonator had a three-point defect and a resonance at 1547.68 nm (left). The other had a four-hole defect and resonances at 1530.47 and 1568.05 nm (right). In both cases, the lattice constant was 420 nm, and the hole size was 230 nm.

Scientists at several laboratories recently demonstrated silicon optical switches based on tiny ring resonators with diameters of tens of microns. Now researchers at NTT Basic Research Laboratories in Atsugi, Japan, have demonstrated a silicon-based optical switch that uses a tiny, submicron resonator housed in a photonic crystal.

The principle of all these switches is the same: Free carriers generated by the control light change the refractive index of a resonator and tune it into or out of resonance with the signal light. The significance of the NTT team’s tiny resonator lies in the fact that the energy required in the control light scales as the volume of the resonator. The smaller the resonator, the smaller the control signal required to switch it. The scientists calculated that several femtojoules of control light in the resonator should be sufficient to switch the signal, although in an experimental demonstration, high coupling losses required several hundred femtojoules applied to the device.

Figure 2. The resonance wavelength shifted to shorter wavelengths withincreasing energy in the probe pulse.

The experimental device was a photonic bandgap resonator with input and output waveguides fabricated in silicon (Figure 1). The researchers fabricated two samples, one with a single resonance at λ = 1547.68 nm, and the other with two resonances, at λ_c = 1530.47 nm and at λ_s = 1568.05 nm. They measured spectral shift as a function of energy by applying picosecond pulses from a PriTel Inc. mode-locked fiber laser to the first sample. The bandwidth of the pulses was much greater than that of the photonic-bandgap resonator, so the pulses essentially provided a white-light probe of the resonator’s transmission (Figure 2).

The investigators attribute the wavelength shift to the plasma effect of free carriers in the silicon created by two-photon absorption of the probe light. The carriers decreased the refractive index of the silicon inside the resonator, shifting the resonance wavelength downward.
To demonstrate optical switching, the researchers used the sample with two resonances. They set the control wavelength to one of the resonances (λ_c) and directed 6.4-ps pulses at this wavelength into the input waveguide. They used the other resonance (λ_s) for the signal and probed the resonator with two continuous-wave signals, one almost exactly on the ls resonance and another tuned to a slightly shorter wavelength.

Figure 3. The transmission of a probe tuned to the resonance wavelength was switched “off” by the control pulse (blue), while a probe tuned to a slightly shorter wavelength than the resonance was switched “on” (red).

When the control pulse generated free carriers and shifted the resonance to a shorter wavelength, the first probe was switched from “on” to “off,” and the second, from “off” to “on” (Figure 3). The switching time was approximately 17 ps in both cases, but the switch-recovery time was slower and different for the two.

The switching speed was faster than the recovery time because the two-photon-absorption process that creates the free carriers is faster than the carriers’ spontaneous relaxation time. The recovery time was different for the two cases because the slope of the resonator’s transmission is steeper near the peak than in the wing. (When the recovery begins, the probe that was switched from “off” to “on” is at the peak of the transmission, and the probe that was switched from “on” to “off” is in the wing.)

Figure 4. “Off”-to-“on” switching performance could be obtained with control pulses containing as little as several hundred femtojoules. Each trace here is artificially offset from its neighbor by 100 ps.

Finally, the scientists measured the strength of the “off”-to-”on” switching as a function of energy in the control pulse (Figure 4). They observed measurable switching at as low as 100 fJ, which they believe is one of the lowest switching energies reported for a passive, all-optical switch.