Photonic reservoir computer using speckle in multimode waveguide ring resonators.

Area of Science:

• Photonics
• Optical Computing
• Nonlinear Optics

Background:

• Photonic reservoir computers (RCs) offer a promising avenue for complex computations.
• Existing designs often require high-speed analog-digital conversion or spatial light modulators.
• Multimode (MM) waveguide geometries are less explored for RC architectures.

Purpose of the Study:

• To propose and simulate a novel photonic RC architecture.
• To demonstrate a design that avoids ultra-high-speed analog-digital conversion and spatial light modulators.
• To analyze the performance of a multimode waveguide ring resonator-based RC.

Main Methods:

• Simulated a photonic RC architecture utilizing speckle in a MM waveguide ring resonator.
• Formulated propagation equations for MM ring resonators with optical nonlinearity and feedback into the standard RC form.
• Speckle mixing was employed to perform pseudo-random matrix multiplications.
• Investigated hyperparameters including outcoupling efficiency, nonlinearity saturation intensity, input bias, and waveguide properties.

Main Results:

• The number of waveguide modes directly correlates with the number of effective neurons in the RC.
• A 50-µm planar waveguide ring achieved 206 effective neurons, accurately predicting Mackey-Glass waveforms.
• A 200-µm diameter MM fiber ring achieved 4,238 effective neurons, accurately predicting chaotic solutions of the Kuramoto-Sivashinsky equation.

Conclusions:

• The proposed photonic RC architecture is a viable and efficient method for complex system modeling.
• Speckle in MM waveguide ring resonators offers a simplified yet powerful approach to photonic computing.
• The design is amenable to physical realization using chip-scale devices or discrete components with MM optical fibers.