The NIKA2 Cosmological Legacy Survey in COSMOS: Final 1.2mm and 2mm source catalogs and redshift distribution of dusty star-forming galaxies

arXiv:2605.00659 · astro-ph.GA · Submitted 2026-05-01 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Astrophysics Radio. Generated commentary on the latest astrophysics papers.

Vera: We're looking at the paper: "The NIKA2 Cosmological Legacy Survey in COSMOS".

Title and authors: Vera: So, we're looking at the NIKA2 Cosmological Legacy Survey in COSMOS, and this paper puts out final source catalogs at both one point two millimeters and two millimeters, plus the redshift distribution for these millimeter-selected dusty star-forming galaxies <ref:2605.00659#pg1,the NIKA2 Cosmological Legacy Survey in COSMOS>.

Jocelyn: It sounds like they are compiling a really deep set of data to investigate the physical properties of these high-redshift DSFGs, while also checking how well their observed redshift distributions match what we predict theoretically.

Subrahmanyan: What this means for us is that we get a chance to test whether the peak epoch of cosmic star formation matches our models, or if those models are being challenged by what the sky actually shows at these high redshifts.

Vera: The paper describes how they achieved deep and homogeneous millimeter coverage in COSMOS, which is really important because it lets them build a robust sample of millimeter-selected DSFGs.

Jocelyn: They found a master catalog with three hundred twenty-three sources that have a signal to noise ratio greater than or equal to three point nine, which gives them about eighty percent purity in their sample <ref:2605.00659#pg3>.

Subrahmanyan: That's the raw number, but I want to know what's important: they identified one hundred four sources detected in both the one point two millimeter and two millimeter bands, nineteen seventy sources only at one point two millimeter, and a final set of twenty-two sources detected only at two millimeter <ref:2605.00659#pg3>.

Vera: And their multi-wavelength identification process was successful for about eighty nine percent of the sample, meaning they confirmed counterparts using interferometric and radio data, leaving only about eleven percent without a confirmed match.

Jocelyn: One of the key findings they highlight is the redshift distribution derived specifically from the one point two millimeter sources, and that distribution has a peak median redshift at two point eight plus or minus zero point one.

Subrahmanyan: That median redshift of two point eight is consistent with the epoch we expect for peak cosmic star formation, which aligns well with SIDES simulations, contrasting with four other galaxy evolution models that show statistical inconsistency.

Vera: That's a big piece of information because it contrasts directly with models like SHARK and the backward-evolution model of Zavala et al. two thousand twenty-one which predict median redshifts closer to two point six.

Jocelyn: And then you have the semi-empirical model from Popping et al. two thousand twenty which predicts a significantly lower median redshift at two point nine, so this NIKA2 measurement really confirms that millimeter selection is an efficient way to probe DSFGs near and beyond the peak epoch of cosmic star formation.

Subrahmanyan: The survey maps cover about one thousand seventy arcmin squared at one point two millimeter and approximately nine hundred twenty arcmin squared at two millimeter, with median noise levels recorded as three hundred fifteen microjanskys at one point two millimeter and ninety-one microjanskys at two millimeter.

Vera: So we have the survey setup and the raw numbers now, but what about how they are trying to make this data even better? The paper outlines some specific improvements they made to their methodology.

Title and authors: Jocelyn: They focused on improving source identification robustness by implementing a refined cross-matching strategy that prioritizes interferometric follow-up data like ALMA and NOEMA over radio observations when the positional uncertainties are high.

Subrahmanyan: That makes sense, because getting those reliable counterparts is crucial for building a clean sample of galaxies to study, especially since they mention this priority order in section four point one regarding identification.

Vera: They also enhanced the flux density estimation by incorporating a Monte Carlo scheme that explicitly models both instrumental noise uncertainty and the scatter in effective flux boosting distributions across different signal to noise ratios, producing corrected fluxes that reflect those combined effects.

Jocelyn: That sounds like a necessary step because you can't just look at the raw flux; you have to account for how much the measurement is affected by both the instrument's noise and how bright the source actually is across different signal to noise ratios.

Subrahmanyan: And they also focused on better redshift constraint application by prioritizing photometric redshifts from COSMOS2020 over COSMOSWeb, because that catalog shows a lower normalized median absolute deviation and a lower fraction of catastrophic outliers for the N2CLS sample.

Vera: That choice is smart because it helps mitigate systematic biases observed in the COSMOSWeb data specifically for dusty galaxies, which they discuss in appendix B.

Jocelyn: They also improved their comparison with galaxy evolution models by using SIDES blob catalogs when comparing redshift distributions to theoretical predictions, because that catalog accounts for source blending effects by convolving noiseless maps with the NIKA2 beam.

Subrahmanyan: Using the SIDES blob catalogs ensures that comparisons are made against a physically realistic representation of the observed flux limits, which is important for making those cosmological connections.

Vera: Finally, they improved their assessment of survey completeness and effective area by calculating the completeness function C of x based on the ratio of input flux to instrumental noise level, as defined in equation E point one and illustrated in figure E point one.

Jocelyn: That formula for effective area using equation E point two allows them to get a more precise determination of how faint sources are recovered across non-uniform noise maps.

Subrahmanyan: So, to wrap up this segment, the NIKA2 Cosmological Legacy Survey in COSMOS provides a deep look at dusty star-forming galaxies and their redshift distribution, and these improvements show a focus on getting reliable data for cosmological testing.

Vera: Exactly. The key result is that millimeter selection is an efficient way to probe DSFGs situated near and beyond the peak epoch of cosmic star formation, even with these careful methodological tweaks.

Jocelyn: It really shows how much careful data reduction and cross-matching can refine our understanding of these dusty, high-redshift galaxies.

Subrahmanyan: And for the big picture, this work helps constrain whether our current models for galaxy evolution accurately predict the distribution of star formation activity at those very early times.

The paper's summary: Vera: So, we’ve seen the raw data from this NIKA2 survey in COSMOS—the final catalogs at one point two millimeter and two millimeters, plus the redshift information for those dusty star-forming galaxies.

Jocelyn: Right, it’s about taking all that deep millimeter coverage and putting together a solid picture of what these high-redshift galaxies are actually doing physically.

Vera: The main goal here is to build a comprehensive dataset so we can really investigate the physical characteristics and see how these dusty star-forming galaxies evolve over cosmic time, while also checking those redshift distributions against our existing theoretical predictions.

Jocelyn: They managed to get deep and consistent millimeter coverage in that COSMOS field, which is key because it lets them construct a really robust sample of these millimeter-selected DSFGs.

Vera: They found a master catalog with three hundred twenty-three sources that have a signal to noise ratio greater than three point nine, which gives them about eighty percent purity in their final sample.

Jocelyn: That means they’ve narrowed it down to a very reliable group of galaxies to study, and they accounted for almost ninety percent of those sources using multi-wavelength data from interferometers and radio observations.

Vera: And one of the most important things they found is the redshift distribution derived specifically from the one point two millimeter sources.

Jocelyn: That distribution has a peak median redshift at two point eight plus or minus zero point one, and that result is consistent with our understanding of when cosmic star formation peaked.

Vera: It’s a real contrast to four other galaxy evolution models that show statistical inconsistencies, like the SHARK model and the Zavala et al. backward-evolution model which predict lower median redshifts around two point six.

Jocelyn: And they even compared it to the Popping et al. semi-empirical model which predicts a significantly lower median redshift of two point nine, so this NIKA2 measurement confirms that millimeter selection is an efficient way to probe DSFGs near and beyond that peak epoch of cosmic star formation.

Vera: The survey maps cover about one thousand seventy arcmin squared at one point two millimeter and approximately nine hundred twenty arcmin squared at two millimeter, with the noise levels being around three hundred fifteen microjanskys at one point two millimeter and ninety-one microjanskys at two millimeter.

Jocelyn: The team also detailed some improvements they made to their methods, like refining their source identification by prioritizing interferometric follow-up data over radio when positional uncertainties are high.

Vera: And they used a Monte Carlo scheme for flux density estimation that accounts for both instrumental noise uncertainty and the scatter in effective flux boosting distributions across different signal to noise ratios.

Jocelyn: They also made a smart choice on their redshift constraints, using photometric redshifts from COSMOS2020 instead of COSMOSWeb because that catalog has a lower fraction of those catastrophic outliers for dusty galaxies.

Vera: These methodological tweaks all point toward getting the most reliable data possible for testing our models and understanding the physical reality of these early star-forming galaxies.

Jocelyn: So, this survey gives us a solid anchor point in cosmic time—the peak epoch—and it shows us how well we can use millimeter observations to find those galaxies.

Vera: It really sets the stage for comparing these results with other large surveys and models that look at galaxy evolution across different scales.

Jocelyn: And it makes us wonder what other observational tools we might need to confirm these redshift distributions in even more detail.

The paper's improvements: Tom: So, we’re looking at how the authors suggest they could make this already solid survey even better for future work and analysis.

Vera: They focused heavily on improving how they identify those sources, specifically by implementing a refined cross-matching strategy that prioritizes interferometric follow-up data like ALMA and NOEMA when the positional uncertainties are high.

Jocelyn: That makes sense, because getting reliable counterparts is crucial for building a clean sample of galaxies to study, especially since they put that priority order in section four point one regarding identification.

Vera: Then they tackled the flux density estimation by incorporating a Monte Carlo scheme that explicitly models both instrumental noise uncertainty and the scatter in effective flux boosting distributions across different signal to noise ratios.

Jocelyn: That means they’re not just looking at one single flux value; they’re modeling the whole mess of how noisy the measurement is, which should lead to more accurate final flux densities.

Vera: They also improved the redshift constraint application by prioritizing photometric redshifts from COSMOS2020 over COSMOSWeb because that catalog shows a lower normalized median absolute deviation and a lower fraction of catastrophic outliers for this Nika2 sample.

Jocelyn: That choice is smart because it helps mitigate those systematic biases we saw in the older COSMOSWeb data when looking at dusty galaxies, which they discuss in appendix B.

Vera: Plus, they enhanced their comparison with galaxy evolution models by using SIDES blob catalogs when comparing redshift distributions to theoretical predictions because that catalog accounts for source blending effects by convolving noiseless maps with the Nika2 beam.

Jocelyn: That’s important because you can’t just compare raw data points; you have to account for how the telescope's resolution smears things out across different sources.

Vera: Finally, they improved their assessment of survey completeness and effective area by calculating a completeness function based on the ratio of input flux to instrumental noise level, which is defined in equation E point one and illustrated in figure E point one.

Jocelyn: That formula for effective area using equation E point two allows them to get a more precise determination of how faint sources are recovered across those non-uniform noise maps.

Vera: So, these improvements show a really rigorous approach to data reduction, moving from the initial detection to a much cleaner final catalog ready for deep cosmological modeling.

Jocelyn: It’s about making sure that when we plug this into our theoretical frameworks later, we aren't biased by noise or bad redshift estimates.

Vera: This level of detail is what allows us to move from just seeing a distribution to actually constraining the physical processes happening in those dusty star-forming galaxies.

Jocelyn: And with these refined tools, we can start asking even deeper questions about how efficiently these galaxies are forming stars at high redshift.

Conclusion: Tom: So we’re wrapping up our look at this paper, "The NIKA2 Cosmological Legacy Survey in COSMOS: Final one point 2mm and 2mm source catalogs and redshift distribution of dusty star-forming galaxies <ref:2605.00659#pg1,source catalogs and redshift distribution of dusty star-forming galaxies>."

Vera: Basically, the authors have delivered a very deep catalog that lets us finally test if our current models for the peak epoch of cosmic star formation are holding up against what we actually see in these high-redshift dusty galaxies.

Jocelyn: They’ve shown that millimeter selection is a really efficient way to probe those galaxies near and beyond the peak, which is a key finding for time-domain radio astronomers like us because it gives us targets with strong physical constraints.

Subrahmanyan: From a theoretical standpoint, this result confirms that the observed redshift peak at two point eight is consistent with SIDES simulations, which is what we want to see from our models of structure formation.

Vera: The implications are that this data helps us narrow down the range where star formation was most intense in the early universe.

Jocelyn: It also sets a benchmark for how well we can use millimeter observations to find these obscured galaxies, which is something we can apply to future surveys like those from Euclid or JWST.

Subrahmanyan: And if our models still disagree with this two point eight median redshift, it tells us that the processes driving galaxy evolution at those early times are different than what we currently think.

Vera: So, for the listener who just wants to know what this means: we have a much more precise picture of where and when these dusty star-forming galaxies were most active.

Jocelyn: And now that we have these tight constraints on their redshift, the next step is figuring out how to follow up those specific sources with deeper spectroscopy or even higher resolution imaging to really nail down those physical properties.

Subrahmanyan: That’s the next big test for cosmology, seeing if the physics of feedback and accretion matches what we see in these high-redshift environments.

ESO · University of Côte d’Azur · School of Physics and Astronomy Cardiff University · Université Paris Cité University Paris-Saclay CEA Institut de Radioastronomie Millimétrique IRAM · Department of Physics University of California Santa Barbara National Observatory of Athens Institute for Astronomy Astrophysics Space Applications and Remote Sensing Department of Astrophysics Astronomy & Mechanics Faculty of Physics University of Athens High Energy Physics Division Argonne National Laboratory

astro-ph.GA

Submitted: 2026-05-01

Updated: 2026-11-17

Comments: A&A in press (accepted 06 October 2026). Data release: https://data.lam.fr/n2cls

DOI: 10.1051/0004-6361/202660707

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 92/100

The gist: The NIKA2 Cosmological Legacy Survey (N2CLS) in the COSMOS field delivers final source catalogs at 1.2 mm and 2 mm, along with the redshift distribution of the millimeter-selected population…

Key concepts

Dusty Star-Forming Galaxies (DSFGs)
These are galaxies undergoing intense star formation that obscure their light due to large amounts of dust. They are selected in millimeter wavelengths because dust emission is strongly detected at these long radio wavelengths, making them key targets for studying early cosmic structure formation.
Redshift Distribution
This describes how the measured distances (redshifts) of the detected galaxies are spread out. The survey found a peak around $z=2.8$, indicating that the majority of these dusty galaxies existed when the universe was relatively young and undergoing rapid star formation.
SIDES Simulations
SIDES simulations are theoretical models used to predict how dark matter halos and baryonic processes evolve over cosmic time. The observed redshift distribution from NIKA2 is compared against these simulations to test whether our understanding of galaxy evolution matches theoretical predictions.

Terminology

Summary

The NIKA2 Cosmological Legacy Survey (N2CLS) in the COSMOS field delivers final source catalogs at 1. p. 0

Where the paper says it

We present the final 1. 2 mm and 2 mm source catalogs and the redshift distribution of the millimeter-selected population from the NIKA2 Cosmological Legacy Survey (N2CLS) in the COSMOS field.

2 mm and 2 mm, along with the redshift distribution of the millimeter-selected population, with the primary objective being to furnish a comprehensive dataset for investigating the physical characteristics and evolutionary trends of high-redshift dusty star-forming galaxies (DSFGs), while simultaneously comparing these observed redshift distributions against theoretical predictions. p. 0
Where the paper says it

2 mm and 2 mm source catalogs and the redshift distribution of the millimeter-selected population from the NIKA2 Cosmological Legacy Survey (N2CLS) in the COSMOS field. Our aim is to provide a comprehensive dataset for studying the physical properties and evolution of high-redshift DSFGs, and to compare the observed redshift distribution with theoretical predictions.

The survey achieves deep and homogeneous millimeter coverage in COSMOS, which facilitates the construction of a robust sample of mm-selected DSFGs. p. 1
Where the paper says it

2007), has transformed the study of DSFGs by enabling robust source identification and redshift estimation for statistically significant samples. COSMOS provides deep and homogeneous coverage from X-ray to radio wavelengths, making it a reference field for studies of galaxy evolution and large-scale structure across cosmic time.

The master catalog comprises 323 sources detected with a signal-to-noise ratio (S/N) at least 3. p. 5

Where the paper says it

The master catalog comprises 323 sources with S/N ≥ 3. 9 (≥ 80% purity), of which 104 appear in both bands, 197 are detected only at 1.

9, corresponding to approximately 80% purity. p. 6
Where the paper says it

The gray solid line indicates the S/N cut of 3. 9, corresponding to 80% purity.

This catalog includes 104 sources detected in both the 1. p. 0
Where the paper says it

, SHARK and FLAMINGO). The N2CLS master catalog includes 323 sources detected at ≥80% purity, with 104 sources detected in both bands, 197 only at 1.

2 mm and 2 mm bands, 197 sources detected exclusively at 1. p. 0
Where the paper says it

, SHARK and FLAMINGO). The N2CLS master catalog includes 323 sources detected at ≥80% purity, with 104 sources detected in both bands, 197 only at 1.

2 mm, and a final set of 22 sources detected only at 2 mm. p. 1
Where the paper says it

For COSMOS, only data obtained prior to May 2021 were used. The final data set presented here covers exactly the same area but has an integration time 2.

Multi-wavelength identifications using interferometric and radio data successfully secure counterparts for about about89% of the sample, leaving only about11% without a confirmed counterpart. p. 0
Where the paper says it

9, ensuring ≥80% purity. Multi-wavelength counterparts are identified using high-resolution interferometric (sub-)mm data (NOEMA, ALMA) and radio observations (VLA, MeerKAT).

A key finding concerns the redshift distribution derived from the 1. p. 1

Where the paper says it

2026), and the confusion noise limits of the IRAM 30 m telescope derived from the GOODS-N deep survey observations (Ponthieu et al. This paper presents the final data reduction of the NIKA2 observations on the COSMOS field obtained as part of the N2CLS, together with the associated source catalogs, identification products and the resulting redshift distribution.

2 mm sources, which exhibits a peak median redshift of z = 2. p. 0
Where the paper says it

1, consistent with the epoch of peak cosmic star formation. In the total sample, we lack redshift for ∼ 2% of the identified galaxies, plus 34 sources for which no accurate positional proxy is available, preventing the identification of a multi-wavelength counterpart.

8 plus or minus 0. p. 0
Where the paper says it

1, consistent with the epoch of peak cosmic star formation. In the total sample, we lack redshift for ∼ 2% of the identified galaxies, plus 34 sources for which no accurate positional proxy is available, preventing the identification of a multi-wavelength counterpart.

1. This result is consistent with the epoch of peak cosmic star formation and aligns well with SIDES simulations, contrasting with four other galaxy evolution models that show statistical inconsistency. p. 0
Where the paper says it

We identify 66 galaxies at z > 4 (19 with spectroscopic redshift). The observed redshift distribution agrees well with the SIDES simulations, while four other galaxy evolution models are statistically inconsistent with the data.

Specifically, while models like SHARK and the backward-evolution model of Zavala et al. p. 0
Where the paper says it

The redshift distribution is compared to predictions from the SIDES simulations and other galaxy evolution models (e. , SHARK and FLAMINGO).

(2021) predict median redshifts closer to z about 2. p. 0
Where the paper says it

2020; Zavala et al. 2021; Traina et al.

6, the semi-empirical model of Popping et al. p. 1
Where the paper says it

2017; Lagos et al. 2018; Popping et al.

(2020) predicts a significantly lower median redshift of z about 1. p. 7
Where the paper says it

2020): This catalog is based on ALMA Band-7 follow-up observations of the 160 brightest 850-µm sources (S850µm ≳ 8 mJy) in the main region of the SCUBA2-COSMOS survey (S2COSMOS; Simpson et al. This region reaches a median sensitivity of 1.

9. The NIKA2 measurements confirm that millimeter selection is an efficient method for probing DSFGs situated near and beyond the peak epoch of cosmic star formation. p. 0
Where the paper says it

2 mm sources peaks at 2. 1, consistent with the epoch of peak cosmic star formation.

The survey maps at 1. p. 0
Where the paper says it

Astronomy & Astrophysics manuscript no. main ©ESO 2026 May 4, 2026 The NIKA2 Cosmological Legacy Survey in COSMOS: Final 1.

2 mm and 2 mm cover approximately 1070 arcmin squared, with median noise levels recorded as 315 mu Jy at 1. p. 0
Where the paper says it

We analyze deep NIKA2 observations at 1. 2 mm and 2 mm, covering ∼1070 arcmin2 with median noise levels of 315 µJy and 91µJy, respectively.

2 mm and 91 mu Jy at 2 mm.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved source identification robustness by implementing a refined cross-matching strategy that prioritizes interferometric follow-up data (like ALMA and NOEMA) over radio observations when positional uncertainties are high. This allows the system to identify reliable counterparts more accurately, as stated in Section 4.1 regarding the priority order of identification.

  2. Enhanced flux density estimation by incorporating a Monte Carlo scheme that explicitly models both instrumental noise uncertainty and the scatter in effective flux boosting distributions across different Signal-to-Noise ratios (S/N). This produces corrected flux densities that reflect these combined effects, as detailed in Section 3.

  3. Better redshift constraint application by prioritizing photometric redshifts from COSMOS2020 over COSMOSWeb, as this catalog shows a lower normalized median absolute deviation and a lower fraction of catastrophic outliers for the N2CLS sample. This choice mitigates systematic biases observed in the COSMOSWeb data for dusty galaxies, as discussed in Appendix B.

  4. More accurate comparison with galaxy evolution models by using SIDES blob catalogs when comparing redshift distributions to theoretical predictions, as this catalog accounts for source blending effects by convolving noiseless maps with the NIKA2 beam. This ensures that comparisons are made against a physically realistic representation of the observed flux limits.

  5. Improved assessment of survey completeness and effective area by calculating the completeness function C(x) based on the ratio of input flux to instrumental noise level, as defined in Equation E.1 and illustrated in Figure E.1. This allows for a more precise determination of how faint sources are recovered across non-uniform noise maps using the formula for effective area (Equation E.2).

Abstract

We present the final 1.2mm and 2mm source catalogs and the redshift distribution of the mm-selected population from the NIKA2 Cosmological Legacy Survey (N2CLS) in the COSMOS field. Our aim is to provide a comprehensive dataset for studying the physical properties and evolution of high-redshift DSFGs. N2CLS covers 1070 arcmin2 with median noise levels of 315 μ Jy and 91 μ Jy at 1.2 and 2mm, respectively. Sources are extracted with a S/N threshold of 3.9, ensuring >80% purity. Multi-wavelength counterparts are identified using high-resolution interferometric (sub-)mm data (NOEMA, ALMA) and radio observations (VLA, MeerKAT). Redshifts are compiled from spectroscopic and photometric catalogs (e.g., COSMOSWeb). The N2CLS master catalog includes 323 sources detected at >80% purity, with 104 sources detected in both bands, 197 only at 1.2mm, and 22 only at 2mm. Multi-wavelength identifications are secured for 89% of the sample. The redshift distribution of 1.2mm sources peaks at 2.8 plus or minus 0.1, consistent with the epoch of peak cosmic star formation. In the total sample, we lack redshift for 2% of the identified galaxies, plus 34 sources for which no accurate positional proxy is available, preventing the identification of a multi-wavelength counterpart. We identify 66 galaxies at z>4. The observed redshift distribution agrees well with the SIDES simulations, while four other galaxy evolution models are statistically inconsistent with the data. N2CLS is the largest contiguous deep survey to date with this depth and homogeneous coverage. This homogeneous coverage is important, as 25% of N2CLS sources lack a SCUBA2 850 micron counterpart, likely because the strongly non-uniform noise distribution of the SCUBA2 map results in lower sensitivity in parts of the field. The released data products provide a legacy dataset for studying dust-obscured galaxy evolution.

Sources

Related papers