Low-noise Fourier Transform Spectroscopy Enabled by Superconducting On-Chip Filterbank Spectrometers

Historically employed spectroscopic architectures used for large field of view mapping spectroscopy in millimetere and sub-millimetre astronomy suffer from significant drawbacks. On-chip filterbank spectrometers are a promising technology in this respect; however, they must overcome an orders-of-magnitude increase in detector counts, efficiency loss due to dielectric properties, and stringent fabrication tolerances that currently limit scaling to resolutions of order 1000 over a large array. We propose coupling a medium-resolution Fourier transform spectrometer to a low-resolution filterbank spectrometer focal plane, which serves as a post-dispersion element. In this arrangement, medium resolution imaging spectroscopy is provided by the Fourier transform spectrometer, while the low resolution filterbank spectrometer serves to decrease the photon noise inherent in typical broadband Fourier transform spectrometer measurements by over an order of magnitude. This is achieved while maintaining the excellent imaging advantages of both architectures. We present predicted mapping speeds for a filterbank-dispersed Fourier transform spectrometer from a ground-based site and a balloon-borne platform. We also demonstrate the potential that an instrument of this type has for an R1000 line intensity mapping experiment using the James Clerk Maxwell Telescope as an example platform. We demonstrate that a filterbank-dispersed Fourier transform spectrometer would be capable of R1000 measurements of CO power spectra with a signal-to-noise ratio of 10–100 with surveys of $10^5$–$10^6$ spectrometer hours.

💡 Research Summary

The paper introduces a novel instrument concept that combines a medium‑resolution Fourier Transform Spectrometer (FTS) with a low‑resolution on‑chip filter‑bank spectrometer (FBS) acting as a post‑dispersion element. Traditional spectroscopic approaches for large‑field‑of‑view (FoV) mapping in the millimetre and sub‑millimetre bands either rely on grating spectrometers, which suffer from severe throughput loss due to slit apertures and are difficult to scale to resolving powers R ≈ 1000, or on Fourier Transform Spectrometers, which provide high throughput and broad instantaneous bandwidth but suffer from a multiplexing penalty: the photon noise on each detector scales with the full optical bandwidth, dramatically degrading sensitivity for broadband observations.

On‑chip filter‑bank spectrometers, based on superconducting resonators terminated in microwave kinetic inductance detectors (MKIDs), can provide compact spectral dispersion directly at the focal plane. However, achieving R ≈ 1000 with FBS requires thousands of MKIDs per spatial pixel, imposes stringent dielectric loss constraints (loss tangent tan δ), and demands sub‑50 nm lithographic tolerances—requirements that are not yet feasible for large arrays.

The authors propose to let the FTS define the overall spectral resolution (R ≈ 1000) while a low‑resolution (R ≈ 200) FBS placed after the interferometer splits the broadband output into many narrower channels. Each channel receives only a fraction (≈1/R_FBS) of the total photon flux, reducing the photon‑noise limited NEP by roughly the square root of this factor—more than an order of magnitude improvement. Because the FBS operates at low resolution, its efficiency loss due to dielectric absorption is modest; with realistic oversampling the net optical efficiency can reach ~70 %. The combined system therefore retains the high throughput advantage of the FTS (Θ_I = 2π A_I R) while mitigating its multiplexing penalty.

Key advantages of the “post‑dispersed FTS” (FBDFTS) architecture are:

1. Detector count reduction – The required MKID count drops from several thousand per pixel (for a pure FBS) to a few hundred for the combined system, compatible with existing read‑out electronics.
2. Relaxed fabrication tolerances – Low‑resolution filter banks need only ~50 nm tolerance, achievable with current lithography, unlike the sub‑10 nm precision demanded for high‑R FBS.
3. Polarisation handling – The FBS can be duplicated for each polarisation at the focal plane, avoiding the polarisation‑dependent losses typical of bulk gratings.
4. Compact optics – The filter bank occupies a few centimetres, far smaller than a metre‑scale grating required for R ≈ 1000, simplifying the instrument layout and preserving the FTS’s imaging capability.
5. Scalability – The concept can be implemented on existing platforms (e.g., a 15 m ground‑based telescope at Maunakea, or a balloon‑borne telescope similar to BLAST) with modest modifications.

Performance simulations were carried out for two representative platforms: a ground‑based observatory (CCAT‑prime/FYST) and a balloon‑borne telescope. Using a realistic atmospheric model (25 % percentile weather at Maunakea) and assuming a 50 % efficient FTS coupled to a 70 % efficient R = 200 filter bank, the authors calculate mapping speeds. The FBDFTS achieves mapping speeds an order of magnitude faster than a conventional broadband FTS and comparable to low‑resolution instruments, while delivering R ≈ 1000 spectral resolution.

A concrete science case is presented for line‑intensity mapping (LIM) of CO rotational lines with the James Clerk Maxwell Telescope (JCMT). The authors show that with 10⁵–10⁶ spectrometer‑hours of observation, the FBDFTS can measure the CO power spectrum with signal‑to‑noise ratios between 10 and 100 at R ≈ 1000, outperforming existing LIM experiments (e.g., SPT‑SLIM, CONCERTO) which are limited to R ≈ 200–300.

The paper also discusses practical implementation details: the Mach‑Zehnder FTS design naturally provides two output ports; placing identical filter‑bank focal planes at both ports can recover near‑100 % optical efficiency. The required MKID read‑out bandwidth, thermal loading, and cryogenic considerations are within the capabilities demonstrated by recent MKID arrays.

In summary, the proposed post‑dispersed FTS architecture elegantly solves the long‑standing trade‑off between spectral resolution, detector count, and optical efficiency for large‑FoV mm/sub‑mm spectroscopy. By delegating the high‑resolution requirement to the interferometer and using a modest‑resolution on‑chip filter bank to reduce photon noise, the system offers a realistic path toward R ≈ 1000, wide‑band, imaging spectrometers suitable for next‑generation facilities such as AtLAST, CCAT‑prime, and balloon‑borne LIM missions. The concept leverages mature technologies (FTS scanning mechanisms, MKID detectors, superconducting resonators) and therefore presents a low‑risk, high‑impact solution that could be realized within the next decade.