High-performance and scalable on-chip digital Fourier transform spectroscopy.

Area of Science:

• Photonics and Spectrometry
• Integrated Optics
• Signal Processing

Background:

• Conventional spectrometers are bulky and power-intensive.
• Current on-chip spectrometers have limited spectral channels and signal-to-noise ratios.
• Silicon photonics offers a pathway for miniaturized optical instruments.

Purpose of the Study:

• To develop a transformative on-chip digital Fourier transform spectrometer.
• To overcome limitations in spectral channel count and signal-to-noise ratio of existing designs.
• To leverage machine learning for enhanced spectrum reconstruction.

Main Methods:

• Demonstration of a reconfigurable Mach-Zehnder interferometer for time-domain modulation.
• Fabrication and packaging using industry-standard silicon photonics technology.
• Implementation of machine learning regularization techniques, specifically an 'elastic-D1' regularized regression method.

Main Results:

• Achieved a boosted signal-to-noise ratio via the multiplex advantage.
• Demonstrated unprecedented scalability for exponentially increasing spectral channels.
• Significant noise suppression for broad and narrow spectral features.
• Enhanced spectral resolution beyond the classical Rayleigh criterion.

Conclusions:

• The developed on-chip digital Fourier transform spectrometer offers significant advantages in size, weight, and power.
• The device enables high-resolution spectra acquisition with improved signal-to-noise ratio and scalability.
• Machine learning regularization techniques effectively enhance spectrum reconstruction for various spectral features.