CONCERTO: Millimeter-Wave Spectroscopic Imager

• CONCERTO is a ground-based wide-field Fourier transform spectrometer designed for millimeter-wave astronomy, enabling [CII] intensity mapping and galaxy cluster studies via the Sunyaev–Zel’dovich effect.
• It employs a Martin–Puplett interferometer with two distinct focal plane arrays hosting 4304 kinetic inductance detectors to achieve rapid on-the-fly spectral mapping.
• Key engineering features include robust cryogen-free operation at 60 mK, active vibration mitigation, and real-time tuning for precise spectral calibration and foreground subtraction.

CONCERTO, an acronym for CarbON CII line in post-rEionization and ReionizaTiOn, is a ground-based, wide-field, low-resolution Fourier transform spectrometer for millimeter-wave astronomy installed on the 12-meter Atacama Pathfinder Experiment (APEX) telescope at 5100 m above sea level on Llano de Chajnantor in northern Chile. It operates in the 130–310 GHz transparent atmospheric window, has an instantaneous field of view of 18.6 arcmin, and uses two focal planes hosting a total of 4304 kinetic inductance detectors. Its primary scientific programs are [CII] line intensity mapping of high-redshift star-forming galaxies and spectroscopic studies of galaxy clusters through the Sunyaev–Zel’dovich effect, with additional applications to Galactic and extragalactic millimeter spectral imaging (Fasano et al., 2023, Fasano et al., 2022).

1. Scientific domain and observational rationale

CONCERTO was conceived to address two major science cases. The first is [CII] intensity mapping of star-forming galaxies at high redshift, especially during the post-reionization and reionization epochs. The second is galaxy cluster studies through the Sunyaev–Zel’dovich (SZ) effect, where broad spectral coverage helps separate SZ signal, dust, atmospheric background, and other components. Both applications impose the same instrumental requirements: large mapping speed, multi-frequency capability, and a wide field of view (Fasano et al., 2023).

For [CII], CONCERTO targets the 158 μm fine-structure line redshifted into the millimeter atmospheric windows for $z \gtrsim 5.2$. The observing strategy follows the logic of line intensity mapping: rather than detecting individual galaxies one by one, the instrument measures the aggregate emission from unresolved galaxies and reconstructs the 3-D clustering of line emission from sky position plus frequency. In the commissioning literature, the intended [CII] survey range is described as roughly $4.5, with the broader cosmological aim of constraining the power spectrum of dusty star-forming matter and the evolution of star formation during reionization and post-reionization (Fasano et al., 2022, Catalano et al., 2021).

The cluster program uses the same spectro-imaging capability in a different regime. CONCERTO is designed to measure the SZ spectrum of clusters at redshifts 0.2 to 0.8, with separation of the thermal SZ effect, the kinetic SZ effect, and relativistic corrections to the thermal SZ spectrum. The design paper also emphasizes other science programs, including dust and molecular gas in local and intermediate-redshift galaxies, Galactic star-forming clouds, and CO intensity fluctuations from galaxies at $0.3 (Catalano et al., 2021, collaboration et al., 2020).

A key survey field is COSMOS, chosen because it has deep existing multiwavelength data useful for foreground subtraction and cross-correlation. The targeted field is described as a $72\times72$ arcmin$^2$ patch centered at $\mathrm{RA}=150.12^\circ$, $\mathrm{Dec}=2.21^\circ$, corresponding to the 1.4 deg$^2$ [CII] survey discussed in later calibration and forecasting papers (Fasano et al., 2022, Lundgren et al., 5 Sep 2025).

2. Optical, cryogenic, and detector architecture

CONCERTO is built around a Martin–Puplett interferometer coupled to a large-format cryogenic camera. The warm instrument is divided into a chassis and an optics box. The chassis includes the dilution cryostat or camera, the interferometer, and the readout and control electronics; the optics box includes seven room-temperature mirrors (M5 to M11), two polarizers, and a cold reference with three additional mirrors for the interferometer. The camera itself uses refractive optics with three lenses at 300 K, 4 K, and 0.1 K, together with a third polarizer at 0.1 K (Monfardini et al., 2021).

The focal plane is split by a 45° polarizer into two arrays with the same detector architecture but different filtering chains. The two arrays cover overlapping but non-identical bands: LF 130–270 GHz and HF 195–310 GHz. The projected photometric beam widths are about 35 arcsec in the LF band and 30 arcsec in the HF band. The focal planes lie on opposite sides of the final interferometer polarizer, which splits the transmitted and reflected beams into the two detector arrays (Fasano et al., 2023).

The detector system uses single-polarization LEKID arrays, microstrip-coupled and fabricated from 20 nm aluminum films deposited on high-resistivity monocrystalline silicon. The substrate thickness is 105 μm for HF and 125 μm for LF. Each array is read out through six lines, for 12 readout lines total, with a multiplexing factor 400 over 1 GHz bandwidth per line (Monfardini et al., 2021).

The cryogenic chain is a fully cryogen-free continuous dilution cryostat with a base temperature of 60 mK and 100% duty cycle. It was designed to tolerate inclinations up to 75°, corresponding to telescope elevations from 15° to 90°. A separate cold-reference cryostat operates near 30 K, with a later upgrade planned to bring it closer to 10 K (Monfardini et al., 2021).

Characteristic	Value	Notes
Telescope	12-meter APEX	Installed in the Cassegrain cabin
Site altitude	5100 m	Llano de Chajnantor, northern Chile
Instantaneous field of view	18.6 arcmin	Wide-field spectro-imaging
Total detectors	4304 KIDs	Two arrays of 2152
LF band	130–270 GHz	Low-frequency focal plane
HF band	195–310 GHz	High-frequency focal plane
Beam widths	35 arcsec / 30 arcsec	LF / HF photometric beams
Data rate	128 MB/s	Full acquisition stream

The relative spectral resolution is described as $R=\nu/\Delta\nu \le 300$, and the absolute resolution is reported as tunable down to 1 GHz in commissioning papers and adjustable, down to about 1.5 GHz in later status summaries. In the original design description, the interferometric resolution obeys

$\Delta \nu = \frac{c}{4\times \Delta l},$

with mirror travel $4.5Catalano et al., 2021, collaboration et al., 2020).

3. Fourier-transform operation and digital acquisition

The core measurement principle is Fourier transform spectroscopy. A linear motor moves a roof mirror to create optical path difference, modulating the incoming radiation and producing an interferogram for each beam in the field of view. The basic relation is stated as

$4.5

where $4.5OPD $4.5Fasano et al., 2022).

The interferometer scans rapidly enough to keep atmospheric $4.570 mm full interferogram path length (forward + backward), completed in about 0.40 s in the faster configuration with 1536 samples or 0.54 s in the slower one with 2048 samples. The KIDs are sampled at 3813 Hz, about 0.26 ms per sample, allowing interferograms to be reconstructed while the telescope continuously scans the sky (Fasano et al., 2022).

This on-the-fly mode is central to the observing concept. The telescope does not stop at each pointing; instead, the dish scans continuously while the interferometer repeatedly records full interferograms. Later status summaries describe CONCERTO as generating more than 16,000 spectra per second during on-the-fly scanning at tens of arcseconds per second, whereas the initial APEX commissioning paper reports more than 20,000 spectra per second during observations. The gain is explicitly attributed not to the efficiency of a single detector but to the combination of a large detector array, spectral multiplexing, and rapid scanning (Fasano et al., 2023, Monfardini et al., 2021).

The readout and control system is organized in five microTCA crates: four crates for KID readout and one crate for control and monitoring. The full system contains 12 KID readout boards, 1 Motor Controller and Martin-Puplett Monitor board (MCMPM), and 1 Cryostat Positioning System board (CPS). For each feedline, a digital comb probes roughly 400 tones across 1 GHz. The FPGA firmware uses 10 band managers, each handling a 100 MHz subband with 40 tone managers, and digital downconversion returns I/Q streams at about 4 kHz, synchronized to interferometer motion (Bourrion et al., 2022).

Because LEKID resonance frequencies drift with sky loading, telescope elevation, and target-to-target conditions, CONCERTO also required a real-time tuning and tracking system. The online algorithm uses a three-point modulation scheme based on LO dithering. For each block of 1536 samples, the readout acquires 32 samples at $4.532 samples at $4.51472 samples at $4.5startup location mode performs a coarse global alignment, a real-time follow mode applies feedline-wide corrections at about 2.4 Hz, and a slower per-detector tuning stage before each scan refines individual tone placement. This tuning infrastructure is essential for stable observations under changing atmospheric conditions (Bounmy et al., 2022).

4. Commissioning, systematics, and calibration framework

The project received an ERC Advanced Grant beginning in January 2019. The instrument was shipped on 1 March 2021, installation at APEX began on 6 April 2021, cryogenic cooling started on 10 April 2021, and the nominal focal-plane temperature was reached on 12 April 2021. The first extended-source observation, on the Crab Nebula, was obtained on 2 May 2021. Commissioning continued until the end of June 2021, and regular science operations began in July 2021. Remote observations from France began shortly after installation. The instrument was dismounted from APEX in May 2023 and returned to the Néel Institute in Grenoble, where it had been designed and assembled (Fasano et al., 2023).

Early on-sky tests verified that all readout lines were connected and that more than 90% of the designed pixels showed a resonance. The internal quality factors improved from approximately 10k on the sky simulator to approximately 17k on sky at APEX, consistent with a coupling quality factor of about 15k. Initial continuum noise measurements on blank sky yielded 2.5 Hz/$0.3 in resonance-frequency units, corresponding to NET ≈ 2.5 mK/$0.3, and to an expected end-to-end sensitivity of NET ≤ 1 mK·$0.3 after combining both arrays and the multi-pixel beam sampling (Monfardini et al., 2021).

A major commissioning difficulty was beamsplitter vibration induced by interferometer motion. The moving roof mirror generated airflow and excited a resonant oscillation of the central beamsplitter membrane near 47.25 Hz. The beamsplitter is described as a 480 × 800 mm$0.3 elliptic polyimide membrane, 50 μm thick, with 50 μm copper wires on a 100 μm pitch. The adopted mitigation combined a laser sensor installed in December 2021 to monitor membrane motion with two active speakers added in April 2022 to generate counter-waves. The mirror full-stroke rate was also reduced from 2.5 Hz to 1.9 Hz. The active system reduced beamsplitter vibration by more than a factor of 7 peak-to-peak (Fasano et al., 2022).

Calibration and performance assessment were later consolidated into an instrument-model framework. In continuum mode, the effective beam full widths at half maximum are 31.9 ± 0.6 arcsec for HF and 34.4 ± 1.0 arcsec for LF; the main beam is slightly elongated, with mean eccentricity 0.46; and two error beams at about 65 arcsec and 130 arcsec yield a main beam efficiency of about 0.52. Calibration with Uranus gives 19.5 ± 0.6 Hz/Jy for HF and 25.6 ± 0.9 Hz/Jy for LF, with point-source continuum uncertainties of 3.0% and 3.4%, respectively. On the COSMOS field, the measured continuum NEFDs are 115 ± 2 mJy/beam·s$0.3 for HF and 95 ± 1 mJy/beam·s$0.3 for LF at mean PWV 0.81 mm and elevation 55.7°; the map RMS follows the expected inverse-square-root law with slopes close to $0.3Hu et al., 2024).

A separate forward-modeling program was developed for spectroscopic calibration. The model simulates raw interferograms from atmosphere, reference source, and stray light, and is used to determine zero path difference (ZPD), relative detector response, and absolute spectral brightness calibration. The ZPD must be known accurately because, for a Gaussian line at 300 GHz, a 0.035 mm ZPD error causes about a 10% loss in recovered amplitude. The calibration pipeline reduced residual ZPD errors to below 0.02 mm, and the absolute spectral-brightness calibration derived from science data agrees between the airmass and emissivity methods at the 9% level for LF and 10% for HF, excluding the least reliable low-frequency range below 170 GHz (Lundgren et al., 5 Sep 2025).

5. Survey methodology, foregrounds, and inference

The principal [CII] analysis problem is not interferogram formation but foreground separation. Realistic CONCERTO-like simulations were therefore built from the SIDES extragalactic sky model, extended to include CO, [CII], and [CI] lines. These cubes span 125–305 GHz with 1 GHz frequency channels and 5 arcsec pixels. The simulations show that the continuum is the brightest astrophysical component, by a factor of roughly 3 to 100 depending on frequency, and that the predicted $0.3 [CII] power spectrum varies by more than two orders of magnitude depending on the assumed relation between star formation rate and [CII] luminosity and on the assumed high-redshift star-formation history (Bethermin et al., 2022).

Foreground removal in the mock COSMOS survey is separated into two problems: dust-continuum subtraction and line-interloper mitigation. For the continuum, the standard PCA approach and the arPLS baseline-removal method were compared. The simulations show that arPLS suppresses the residual continuum to a sub-dominant level of the [CII] signal at $0.3 by a factor of $0.3, whereas PCA achieves only 0.7. The residual CIB power after arPLS is reduced to levels such as $72\times72$0, $72\times72$1, and $72\times72$2 at 305, 253, and 237 GHz, respectively (Cuyck et al., 2023).

For interloping lines, especially CO and [CI], the adopted strategy is masking based on external COSMOS catalogues, using stellar mass as the practical proxy for line brightness. In the simulations, the most massive galaxies, though only 22–25% of objects, produce 99.7% of the CO power. Masking circles of radius $72\times72$3 around these sources, with $72\times72$4 as the most aggressive case studied, produces area losses of about 22% at 305 GHz, 29% at 253 GHz, and 38% at 237 GHz (Cuyck et al., 2023).

After continuum subtraction and masking, the limiting contaminant becomes the residual population of faint, unmasked interlopers. For a CONCERTO-like survey, the resulting power ratios are forecast to be

$72\times72$5

at $72\times72$6 with 22% area loss,

$72\times72$7

at $72\times72$8 with about 29% area loss, and

$72\times72$9

at $^2$0. The conclusion of these forecasts is explicit: dust continuum is not the limiting foreground if arPLS is used, whereas residual CO and [CI] dominate the error budget at $^2$1 and especially at $^2$2 (Cuyck et al., 2023).

Field-to-field variance adds a second limit to inference. The SIDES-Uchuu simulations over 117 deg$^2$3 show that CO and [CII] LIM power spectra can vary by up to 50% in 1 deg$^2$4 fields, and that Poisson variance alone can underestimate the total variance by up to 80% for luminosity functions. For CONCERTO’s roughly 1.4 deg$^2$5 survey scale, this implies that the first [CII] and CO power-spectrum constraints must incorporate non-Poisson field variance, especially for the high-redshift [CII] slices (Gkogkou et al., 2022).

6. Observational results, operational record, and status

The first large-area spectral-imaging demonstration was the Orion Nebula. A preliminary status paper presented 2 hours of observations in 12 scans over $^2$6 arcmin$^2$7, using an interferometer stroke of 30 mm, calibration from Uranus emission, and a preliminary cross-check against the Planck 217 GHz map. The spectral maps reveal Orion BN/KL, Orion South, Orion Bar, and the north-south main filament. In those first maps, the Orion Bar is dominated by free-free emission, while other regions are dominated by Galactic dust emission. The demonstrated spectral coverage extends over roughly 140 to 310 GHz, excluding the water-vapor line near 180 GHz, and establishes that CONCERTO can produce large-area spectral cubes rather than only broadband images (Fasano et al., 2023).

A later Orion analysis characterized the instrument’s intermediate-resolution spectro-imaging performance more fully. The practical spectral resolution used there was 6 GHz, sampled at 3 GHz, and the measured sensitivity was 200 mK in one second, for one beam and a 6 GHz frequency width, over an 18 arcmin field of view. The Orion spectra are described as a mixture of dust and free-free emission, with dust emissivity index 1.3 to 2.0, a positive continuum spectral index around 1 in the Orion Centre, and a negative index around $^2$8 in the Orion Bar. The same data naturally separate CO(2–1) at 230.5 GHz and H$^2$9O at 183.3 GHz from the continuum, including a reported 33$\mathrm{RA}=150.12^\circ$0 detection of the water line at the Orion Centre, especially BN/KL (Désert et al., 29 Apr 2025).

Operationally, by the time of dismounting CONCERTO had accumulated about 793 hours in the [CII] intensity-mapping program and 465 hours across 11 programs: 6 SZ, 3 interstellar medium, and 2 evolved stars. The collaboration stored 50 days of data, corresponding to about 174 TB of raw data before compression (Fasano et al., 2023).

The project status after hardware operations shifted from commissioning to data exploitation. The later instrument papers identify the main remaining analysis tasks as OPD fine reconstruction, FTS reference characterization, KID off-resonance response, and atmospheric emission contamination modeling, together with a dedicated instrument model for systematic effects. In that sense, CONCERTO occupies a specific place in millimeter instrumentation: it is documented as a pioneering mm-wave spectroscopic imager, based on a wide-field multiplexed FTS with thousands of KIDs, that has already demonstrated spectral-cube production on sky and has moved into the calibration and inference stage required for its [CII], SZ, and millimeter spectral-imaging programs (Fasano et al., 2023).