Graphene- All Optical Modulators
Allect Space Systems is working a new class of power-efficient, high-performance all optical modulators using graphene layers that are incorporated on top of silicon waveguides.
Allect wants to transform the AI GPUs’ optical connectivity using more performance and energy-efficient Graphene AOMs.
Design Implications
Based on the literature and simulation studies, here are insights + gaps that our design aims to address:
- Slot waveguide topology is helpful
Slot (and plasmonic slot) waveguides increase the field overlap with graphene, which leads to higher modulation efficiency (dB/µm) and lower pump energy / saturation threshold. Many of the best‐performing designs use slot or strong confinement of the optical mode.
2. Double-layer graphene (DLG) + heterostructures improve performance
Using multiple graphene layers (or graphene-hBN stacks) helps boost absorption (saturable + nonsaturable) and allows tuning of chemical potential more effectively. But more layers also introduces more nonsaturable absorption and possibly slower recovery, so careful trade-offs are needed.
3. Trade-off between insertion loss (IL) and modulation / extinction
Designs that push very high modulation efficiency often suffer from higher IL (especially plasmonic or metal loss). Keeping IL low is essential if the “on” state transmits enough light, especially when saturable absorption is the mechanism (since you need a good signal baseline).
4. Saturation / switching energy is still relatively high in many designs
Even the more advanced slot or plasmonic designs tend to require energies in the ~ fJ to pJ per pulse range. If your goal is “ultra-low power” (e.g. <100 fJ, or even tens of fJ), there is still room to push down.
5. Speed / recovery time is favorable in graphene, but parasitic effects and device length often limit the bandwidth
Graphene’s intrinsic recovery times are ultrafast (hundreds of fs to ps). But in practical devices, the device capacitance, contact resistance, heat dissipation, and the speed of carrier generation / recombination (plus saturation dynamics) can degrade speed.
6. Chemical potential / gating / doping control is important
Enhancing the Fermi level (gating) helps reduce required pump photon energy (Pauli blocking), improves saturation behavior. But achieving good gating (low leakage, stable, reproducible) in a slot geometry with DLG is non-trivial.
7. Non-saturable absorption & parasitic losses are critical
Even if saturable absorption is strong, if a large portion of the absorption is nonsaturable (defects, metal, coupling, scattering) this increases IL and reduces effective modulation. Also, losses in metals (for plasmonic designs) and in coupling to/from the waveguide degrade performance.
Target Specifications
DLG slot-waveguide saturable absorption modulator:
•Modulation efficiency: ≥ 0.2 dB/µm (you mention this was already demonstrated for a DLG modulator) so pushing above 0.3 dB/µm would be great.
•Saturation / switching energy (pump all-optical): aim for ≤100 fJ per pulse (or even tens of fJ if possible), for device lengths under ~20-30 µm.
•Extinction ratio / modulation depth: >10 dB if possible (for good signal contrast), or modularly high depending on application.
•Insertion loss (in “baseline” or “on” state): as low as possible — ideally <1 dB, or <0.5 dB if feasible.
•Recovery time / speed: aim for hundreds of femtoseconds to <1 ps response; or correspondingly high bandwidth (100s of GHz) if in continuous or fast pulsed regime.
•Device footprint / length: keep it compact, probably in the 10–30 µm range.
•Chemical potential gating to allow modulation of absorption strongly (good trade between doping vs mobility / losses).
•Minimize nonsaturable absorption (graphene defects, etc.), metal or contact losses, coupling losses.