High-resolution on-chip Fourier transform spectrometer based on cascaded optical switches

Area of Science:

• Photonics and Spectrometry
• Integrated Optics
• Machine Learning Applications

Background:

• Chip-level spectrometers offer stable and cost-effective spectral analysis.
• Miniaturization is crucial for widespread spectral analysis applications.

Purpose of the Study:

• To present a novel silicon on-chip digital Fourier transform spectrometer.
• To demonstrate its capability for accurate spectral reconstruction using machine learning.

Main Methods:

• Design of a spectrometer chip with eight cascaded optical switches and delay waveguides.
• Configuration of 127 Mach-Zehnder interferometers with varied optical path differences.
• Application of a machine-learning regularization method for spectrum reconstruction.

Main Results:

• Successful retrieval of both sparse and broadband optical spectra.
• Negligible reconstruction errors achieved.
• Demonstrated potential for improved spectral resolution with more stages.

Conclusions:

• The chip-level spectrometer is compact, scalable, and offers a high signal-to-noise ratio.
• It is a promising technology for realizing miniaturized spectrometers.
• The integrated approach combined with machine learning enhances spectral analysis capabilities.