Robust calibration and energy optimization in reconfigurable photonic processors

Area of Science:

• Photonics
• Optical Computing
• Integrated Photonics

Background:

• Reconfigurable photonic processors face performance degradation due to fabrication variations, thermal fluctuations, and signal errors.
• Accurate control of phase and amplitude is critical for high-fidelity photonic operations.

Purpose of the Study:

• To introduce an energy-aware calibration routine for photonic processors.
• To simultaneously enhance computational accuracy and minimize energy consumption.
• To address key obstacles hindering scalable photonic computing.

Main Methods:

• Implemented an output-channel normalization procedure.
• Retuned Mach-Zehnder interferometers and phase shifters by fitting transfer matrices to theoretical models.
• Applied complementary optimizations: global phase offset, voltage-branch selection for SU(2) rotations, and matrix row permutation.

Main Results:

• Achieved a twofold reduction in error during a 4x4 Hadamard-transform test.
• Significantly decreased total electrical power consumption without compromising fidelity.
• Demonstrated a precision, power-conscious calibration technique.

Conclusions:

• The developed calibration routine effectively restores accuracy in reconfigurable photonic processors.
• Energy-aware optimization substantially reduces power usage, making photonic computing more scalable.
• This approach removes a critical barrier for high-performance, large-scale photonic computing systems.