Photonic reservoir computing with a silica microsphere cavity

Area of Science:

• Photonics
• Nonlinear Optics
• Information Processing

Background:

• Reservoir computing (RC) is a computational paradigm utilizing a fixed nonlinear system (reservoir) to process time-varying signals.
• Photonic implementations of RC offer potential advantages in speed and energy efficiency.
• Silica microsphere cavities support high-Q whispering gallery modes, enabling strong light-matter interactions and nonlinear effects.

Purpose of the Study:

• To experimentally demonstrate a novel photonic reservoir computing system.
• To utilize a passive silica microsphere cavity as the reservoir.
• To evaluate the system's performance on benchmark tasks without feedback loops.

Main Methods:

• Fabrication and characterization of a passive silica microsphere cavity.
• Coupling non-return-to-zero and multiple-level signals into the microsphere.
• Exploiting whispering gallery mode interference for nonlinear response.
• Operating the microsphere in an over-coupled state for enhanced memory.
• Evaluating generalization, chaotic time series prediction, and channel equalization.

Main Results:

• Consistent nonlinear response observed for different signal types.
• Achieved a correlation coefficient of 0.923 for generalization.
• Obtained a Normalized Mean Squared Error (NMSE) of 0.06 for chaotic time series prediction.
• Attained a Symbol Error Rate (SER) of 0.02 at 12 dB Signal-to-Noise Ratio (SNR) for channel equalization.
• Demonstrated that higher quality factor microspheres offer greater memory capacity.

Conclusions:

• The passive silica microsphere cavity functions effectively as a photonic reservoir.
• The system achieves high performance on complex computational tasks without requiring delayed feedback.
• This approach provides a pathway towards low-consumption, integrated photonic reservoir computing systems.