Modeling and Analysis of Phase Instability in Photonic Processor

As photonic processors transition toward large-scale quantum and classical information processing, maintaining phase stability across multi-port input signals has become a critical bottleneck. Environmental factors, such as mechanical vibrations in fiber-optic cables and thermal drifts, introduce stochastic phase noise that disrupts the interference patterns required for reliable computation. This study investigates how to model, quantify, and mitigate these phase instabilities in an 8-mode reconfigurable photonic processor.

The researchers employed two primary methodologies to analyze phase noise. First, they developed a theoretical model based on Brownian motion, where phase fluctuations are treated as a cumulative random walk process with variance increasing linearly over time. Second, they implemented an experimental setup using a mesh of Mach-Zehnder Interferometers (MZIs) and tunable phase shifters. By configuring the processor to map two-mode input signals to four-mode outputs, they were able to extract the relative phase difference between inputs using sine and cosine power measurements. This allowed for continuous phase reconstruction without the limitations of traditional phase-wrapping intervals.

Findings

The study found that the Brownian random walk model provides a robust statistical framework for predicting phase noise in photonic processors. Experimental data collected over 35 trials showed a high degree of agreement with the simulated distributions, with a discrepancy of approximately 9%. Furthermore, the researchers applied a Discrete Fourier Transform (DFT) to both experimental and simulated data, confirming that the observed noise characteristics align with the expected spectral features of Brownian motion. The authors also demonstrated that this model can be effectively utilized for input phase correction through self-feedback control mechanisms within the processor.