---
title: 'CONCERTO: Millimeter-Wave Spectroscopic Imager'
url: https://www.emergentmind.com/topics/concerto-36a0d9fb-37f0-4c45-a152-6209b2ca25e4
type: topic
---

# CONCERTO: Millimeter-Wave Spectroscopic Imager

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 [2311.04704] [2206.15146].

## 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** [2311.04704].

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 **\(4.5<z<8.5\)**, 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 [2206.15146] [2110.14984].

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 **\(0.3<z<2\)** [2110.14984] [2007.14246].

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 [2206.15146] [2509.04929].

## 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** [2106.14028].

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 [2311.04704].

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 [2106.14028].

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** [2106.14028].

| 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 \(\Delta l\) and optical path difference \(\mathrm{OPD}=2\Delta l\) [2110.14984] [2007.14246].

## 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
\[
I(\delta)\;\rightarrow\;\mathcal{F}\{I(\delta)\}=S(\nu),
\]
where \(I(\delta)\) is interferometric intensity as a function of OPD \(\delta\), and \(S(\nu)\) is the recovered spectrum [2206.15146].

The interferometer scans rapidly enough to keep atmospheric \(1/f\) noise under control. One description gives a **70 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 [2206.15146].

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** [2311.04704] [2106.14028].

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 [2208.07629].

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 \(F_0+F_{\mathrm{mod}}\), **32 samples** at \(F_0-F_{\mathrm{mod}}\), and **1472 samples** at \(F_0\). A full-band sweep establishes the initial resonance comb, a **startup 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 [2206.11554].

## 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 [2311.04704].

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/\(\sqrt{\mathrm{Hz}}\)** in resonance-frequency units, corresponding to **NET ≈ 2.5 mK/\(\sqrt{\mathrm{Hz}}\)/pixel/array**, and to an expected end-to-end sensitivity of **NET ≤ 1 mK·\(\sqrt{s}\) per beam** after combining both arrays and the multi-pixel beam sampling [2106.14028].

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\(^2\)** 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** [2206.15146].

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\(^{1/2}\)** for HF and **95 ± 1 mJy/beam·s\(^{1/2}\)** 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 \(-1/2\) [2406.15572].

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** [2509.04929].

## 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 **\(z\sim6\)** [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 [2204.12827].

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 **\(z=7\)** by a factor of **\(>70\)**, whereas **PCA** achieves only **0.7**. The residual CIB power after arPLS is reduced to levels such as **\(1.1\times10^{-3}\)**, **\(1.5\times10^{-4}\)**, and **\(1.9\times10^{-4}\ \mathrm{Jy^2/sr}\)** at **305**, **253**, and **237 GHz**, respectively [2306.01568].

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 \(r_\sigma=n\sigma\) around these sources, with **\(r_\sigma=3\)** as the most aggressive case studied, produces area losses of about **22%** at **305 GHz**, **29%** at **253 GHz**, and **38%** at **237 GHz** [2306.01568].

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
\[
\mathrm{[CII]}/(\mathrm{residual\ CO+[CI]}) = 62 \pm 32
\]
at **\(z=5.2\)** with **22%** area loss,
\[
5.5 \pm 3.6
\]
at **\(z=6.5\)** with about **29%** area loss, and
\[
2.0 \pm 1.4
\]
at **\(z=7\)**. 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 **\(z \gtrsim 6.5\)** and especially at **\(z>7\)** [2306.01568].

Field-to-field variance adds a second limit to inference. The **SIDES-Uchuu** simulations over **117 deg\(^2\)** show that **CO and [CII] LIM power spectra can vary by up to 50% in 1 deg\(^2\) 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\(^2\)** 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 [2212.02235].

## 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 **\(30\times60\) arcmin\(^2\)**, 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 [2311.04704].

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 **\(-2\)** in the Orion Bar. The same data naturally separate **CO(2–1)** at **230.5 GHz** and **H\(_2\)O** at **183.3 GHz** from the continuum, including a reported **33\(\sigma\)** detection of the water line at the Orion Centre, especially **BN/KL** [2504.20487].

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 [2311.04704].

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 [2311.04704].

Source: https://www.emergentmind.com/topics/concerto-36a0d9fb-37f0-4c45-a152-6209b2ca25e4