The convolutional neural networks for analysing the micro-cavity array multi-mode quantum frequency comb spectrum features

TL;DR

This paper demonstrates how convolutional neural networks can analyze multi-mode quantum frequency comb spectra from micro-cavities, achieving high accuracy in identifying environmental parameters for sensing applications.

Contribution

It introduces a machine learning approach to automatically extract spectrum features from micro-cavity OFC, enhancing sensitivity and accuracy in environmental parameter detection.

Findings

Single-parameter identification accuracy of 99.5%

Double-parameter identification accuracy of 97.0%

Proposed integrated device for fluid characteristic detection

Abstract

The research on sensing the sensitivity of the light field in the whispering gallery mode (WGM) to the micro-cavity environment has already appeared, which uses the frequency shift of the light field in the WGM or the sensitivity of the resonance peak frequency shift. Multi-mode comb teeth of optical frequency comb(OFC) generated by nonlinear micro-cavity have excellent sensitivity to micro-cavity environment, and they have more sensitivity degrees of freedom compared with WGM light field (the strength of each comb tooth can be influenced by micro-cavity environment). The influence of different substances on the environmental parameters of micro-cavity is complex and nonlinear, so we use machine learning method to automatically extract the spectrum characteristics, the average accuracy of single-parameter identification attains to 99.5%, and the average accuracy of double parameter identification attains to 97.0%. Based on the integration of micro-cavity OFC and wave-guide coupling structure, we propose an set of fluid characteristics detection integrated device in theoretically.