Spatial heterodyne Fourier transform (SHFT) spectroscopy is based on simultaneous interferometric measurements implementing linearly increasing optical path differences, hence circumventing the need for mechanical components of traditional Fourier transform spectroscopy schemes. By taking advantage of the high mode confinement of the Siliconon-Insulator (SOI). platform, great interferometric lengths can be implemented in a reduced footprint, hence increasing the resolution of the device. However, as resolution increases, spectrometers become progressively more sensitive to environmental conditions, and new spectral retrieval techniques are required. In this work, we present several software techniques that enhance the operation of high-resolution SHFT micro-spectrometers. Firstly, we present two techniques for mitigating and correcting the effects of temperature drifts, based on a temperature-sensitive calibration and phase errors correction. Both techniques are demonstrated experimentally on a 32 Mach-Zehnder interferometers array fabricated in a Silicon-on-insulator chip with microphotonic spirals of linearly increasing length up to 3.779 cm. This configuration provides a resolution of 17 pm in a compact device footprint of 12 mm2. Secondly, we propose the application of compressive-sensing (CS) techniques to SHFT micro-spectrometers. By assuming spectrum sparsity, an undersampled discrete Fourier interferogram is inverted using l1-norm minimization to retrieve the input spectrum. We demonstrate this principle on a subwavelength-engineered SHFT with 32 MZIs and a 50 pm resolution. Correct retrieval of three sparse input signals was experimentally demonstrated using data from 14 or fewer MZIs and applying common CS reconstruction techniques to this data.
We demonstrate a novel compressive sensing Fourier-transform spectrometer (FTS) for snapshot Raman spectroscopy in a compact format. The on-chip FTS consists of a set of planar-waveguide Mach-Zehnder interferometers (MZIs) arrayed on a photonic chip, effecting a discrete Fourier-transform of the input spectrum. Incoherence between the sampling domain (time), and the spectral domain (frequency) permits compressive sensing retrieval using undersampled interferograms for sparse spectra such as Raman emission. In our fabricated device we retain our chosen bandwidth and resolution while reducing the number of MZIs, e.g. the size of the interferogram, to 1/4th critical sampling. This architecture simultaneously reduces chip footprint and concentrates the interferogram in fewer pixels to improve the signal to noise ratio. Our device collects interferogram samples simultaneously, therefore a time-gated detector may be used to separate Raman peaks from sample fluorescence. A challenge for FTS waveguide spectrometers is to achieve multi-aperture high throughput broadband coupling to a large number of single-mode waveguides. A multi-aperture design allows one to increase the bandwidth and spectral resolution without sacrificing optical throughput. In this device, multi-aperture coupling is achieved using an array of microlenses bonded to the surface of the chip, and aligned with a grid of vertically illuminated waveguide apertures. The microlens array accepts a collimated beam with near 100% fill-factor, and the resulting spherical wavefronts are coupled into the single-mode waveguides using 45& mirrors etched into the waveguide layer via focused ion-beam (FIB). The interferogram from the waveguide outputs is imaged using a CCD, and inverted via l1-norm minimization to correctly retrieve a sparse input spectrum.