Fig. 1. Overview of the proposed on-chip spectrometer. (a) Conceptual illustration of the spectrometer. The on-chip spectrometer consists of a tunable directional coupler integrated with a comb-drive actuator. The incident light propagates through the directional coupler and forms an interferogram at the output. The unknown spectrum is reconstructed from the measured interferogram by computational reconstruction techniques. (b) Cross-section view of the directional coupler. The comb-drive actuator can modulate the waveguide to achieve in-plane displacement so that the coupling gap breaks the lithography resolution limitation and achieves strong coupling. The strong coupling improves the OPD and thus the reconstruction resolution. (c) The comb-drive actuator can achieve a large enough in-plane tuning displacement to bypass the limitation of the BOX layer thickness when expanding the operation wavelengths to longer wavelength bands.

Fig. 2. Device structure. (a) Top view of the on-chip spectrometer fabricated on an SOI wafer, consisting of two suspended silicon waveguides integrated with an array of five comb-drive actuators. (b) Zoom-in view of a single comb-drive actuator. (c) Isometric view of the device. The movable parts are formed by locally removing the BOX layer beneath.

Fig. 3. Modulation of the directional coupler. (a) Coupling gap fluctuation along the waveguide direction under different applied bias voltages. Insets provide zoom-in views at the second actuator. (b) Coupling gap as a function of applied bias voltage. Insets show modal field plots of the directional coupler in initial and final states, respectively.

Fig. 4. Single-wavelength spectrum reconstruction. (a) Normalized calibration matrix of the spectrometer. The wavelength is swept with a 0.1 nm resolution, and 53 steps of DC bias voltage are gradually applied to the comb-drive actuator. (b) Measured interferograms at several different laser wavelengths. (c) Corresponding reconstructed spectra from the interferograms shown in (b).

Fig. 5. Double-wavelength spectrum reconstruction. (a) Reconstruction of double-wavelength spectra with varying wavelength spacing. (b) Zoom-in view of the reference and reconstructed spectra with wavelength spacing of 0.1 nm. (c) Zoom-in view of the spectra with 0.2 nm resolution.

Fig. 6. Comparison of reported on-chip FT spectrometers. The performance is compared with the focus on (a) bandwidth-to-resolution ratio (BRR) versus power consumption; (b) resolution-power product (RPP) versus resolution; (c) resolution-footprint product (RFP) versus resolution. BRR, RPP, and RFP are calculated using Rayleigh resolution, while EBRR, ERPP, and ERFP are calculated using enhanced resolution.

Fig. 7. MIR spectrum reconstruction. (a) Reconstruction of double-wavelength spectra with different wavelength spacings. (b) Zoom-in view of the reference and reconstructed spectra with wavelength spacing of 1 nm. (c) Zoom-in view of the spectra with 1.5 nm resolution. (d) Spectral resolution as a function of tuning range.

Fig. 8. Polynomial fitting of effective index difference Δn. (a) Fitting of Δn in the NIR waveband. (b) Relation between Δn and coupling gap under linear and logarithmic coordinates when the incident wavelength is 1.55 μm. (c) Fitting of Δn in the MIR waveband. (d) Relation between Δn and coupling gap under linear and logarithmic coordinates when the incident wavelength is 4.16 μm.

Fig. 9. Spectral resolution at the Rayleigh criterion for different coupling lengths in the (a) NIR spectrometer and (b) MIR spectrometer.

Fig. 10. Power consumption of the NIR spectrometer driven by the comb-drive actuator. (a) Average tuning power required to reach the corresponding applied voltages. (b) Static power required to hold at the corresponding voltages.