In J.R.R. Tolkien’s novel ‘Lord of the Rings’, the Dark Lord Sauron created the “One Ring” as the ultimate weapon to conquer all of Middle-earth. So too it seems that in the world of integrated silicon photonics, the “ring” has become somewhat ubiquitous and powerful. Resonance rings can be made to modulate laser light, act as filters and switches and in some cases even be used as on-chip laser light sources.
Optics are considered to be one of the most viable solutions to the performance limitations of electrical interconnects. Integrated CMOS photonic solutions are arguably one of the most promising approaches for high bandwidth off and on-chip communications. Light modulation is key to any optical interconnection system as it converts electrical data into the optical domain. It is typically realized by changing carrier concentrations (holes and electrons) to affect the refractive index of the waveguide material, which, in turn, is used to modify the propagation velocity of light and the absorption coefficient in the waveguides. Optical modulators can modify phase, amplitude and polarization by thermo-optic, electro-optic, or electro-absorption modulation and they are usually based on interference (Mach-Zehnder interferometers – MZIs), resonance (rings or quantum well resonators) and bandgap absorption (germanium and now graphene-based electro-absorption modulators).
MZIs are probably the most well-known modulators and have played a major role in silicon-photonic based 100 gigabit optical transceivers for data center communication (see www.luxtera.com, www.kotura.com). They work by splitting an optical path into two parallel arms and then changing the index of refraction in one arm to induce a phase shift of the light. The light from the two arms re-unites and interferes either constructively or destructively allowing the light to be modulated. These devices are relatively large (several millimeters) and have energy dissipation of around 1-5 pJ/bit, two orders of magnitude higher than the 2-50 fJ/bit expected for on-chip communications.
Back to “rings”. Resonance-based modulators, are typically made up of silicon wave-guide rings integrated with a PIN junction to enable electronic control of their refractive index. Rings are coupled with linear wave guides data buses. Light from the input waveguide having a wavelength matching the resonance of the ring, will couple into the ring and build up in intensity over multiple round-trips due to constructive interference. This light is then output to a second detector waveguide. If critical coupling is achieved, light of the wave length selected will not propagate past the ring, effectively stopping propagation of that wavelength on the input bus. The PIN junction is used to modulate the ring’s index of refraction enabling it to be used to modulate light on the input bus as well as to act as a switch to move the selected light onto other buses.
The real power of the ring is that it enables wave length division multiplexing (WDM). WDM uses different wavelengths of light to simultaneously send multiple independent data signals down the same waveguide, effectively multiplying bus bandwidth by the number of wavelengths employed. Ring resonators are uniquely suitable for WDM as each resonator interacts only with wavelengths that correspond to its resonant modes. These devices have extremely small footprints (several microns) which results in low power operation as well as permitting integration of thousands of them on a single die.
Dense WDM modulation can be accomplished by cascading microring modulators on the same waveguide. Columbia University experimented with multiple different ring-cascade architectures for a TDM-based bus connecting multiple cores on the same die and showed effective bandwidths of up to 600 Gbps depending on the number of cores sites per switching cluster.
Now the thing that would make Sauron possibly rethink the “ring” as his ultimate weapon, at least for light modulation, is an electro-absorption modulator (EAM); specifically, an Indium-Tin-Oxide (ITO) hybrid plasmonics EAM. This class of transparent conductive oxides have been found to allow for unity index changes which is 3 to 4 orders of magnitude higher compared to classical electro-optical materials, such as Lithium Niobate. George Washington University has shown that when an electrical voltage bias is applied across this device it forms an accumulation layer at the ITO-SiO[SUB]2[/SUB] interface, which increases the ITO’s carrier density and raises its extinction coefficient. They were able to obtain an extinction ratio of –5 and –20 dB for device lengths of 5 and 20μm, respectively. This record-high 1 dB/μm extinction ratio is due to the combination of the hybrid plasmonics mode enhancing the electro-absorption of the ITO and ITO’s ability to change its extinction coefficient by multiple orders of magnitude when applied with an electric field. This change stems from an increase in the carrier density in the ITO film (by a factor of 60) due to the formation of the accumulation layer in the MOS capacitor, which was verified via electrical metrology tests and analytical modeling. In summary they were able to achieve deep sub-λ 3D optical confinement in a single-mode cavity with bandwidths approaching the THz range and power consumption in the atto-joule regime, which is about 3–5 orders of magnitude lower compared to other state-of-the-art devices.
While the “ring” is still the dominant structure for photonic design because of its versatility in filtering and switching, Sauron or any silicon photonics engineer for that matter, would be wise to continuing using them. However, when it comes to modulators, EAMs, and especially those employing hybrid plasmonics, would definitely be worth looking into.