Low-Loss Photonic Reservoir Computing with Multimode Photonic Integrated Circuits

Area of Science:

• Photonics
• Integrated Optics
• Reservoir Computing

Background:

• Reservoir computing leverages complex dynamical systems for computation.
• Passive integrated photonics offers a scalable platform for reservoir computing.
• Existing designs face limitations in mode control and signal propagation.

Purpose of the Study:

• To numerically investigate a passive integrated photonics reservoir computing platform.
• To propose and analyze a novel multimodal Y-junction design.
• To enhance reservoir dynamics and computational performance.

Main Methods:

• Numerical simulation of a passive integrated photonics reservoir computing platform.
• Design and analysis of multimodal Y-junctions with tailored adiabaticity.
• Evaluation of combination efficiency and power transfer to reservoir nodes.
• Demonstration on a header recognition task for bit-level processing.

Main Results:

• A novel Y-junction design capturing radiation loss in higher-order modes.
• Increased reservoir dynamics richness through enhanced mode mixing.
• Average combination efficiency improved to 61% from 50% (single-mode).
• Enhanced power delivery to distant reservoir nodes, improving scalability.
• Successful demonstration on a header recognition task.

Conclusions:

• The proposed multimodal Y-junction design enhances integrated photonics reservoir computing.
• The design offers improved efficiency, scalability, and bit-level processing capabilities.
• The CMOS-compatible design is suitable for standard fabrication procedures.