The NIKA2 cosmological legacy survey at 2 mm: catalogs, colors, redshift distributions, and implications for deep surveys

Millimeter galaxy surveys are particularly effective in detecting dusty star-forming galaxies at high redshift. While such observations are typically conducted at 1mm, some studies suggest that 2mm may be better suited for selecting sources at even higher redshifts. We use the unprecedented 2mm data from the N2CLS, together with the SIDES simulation, to study and interpret the statistical properties of 2mm-selected galaxies. We use the N2CLS robust sample at 2mm, which contains 25 sources in the deep GOODS-N field and 90 sources in the wide COSMOS. The sources are matched with the N2CLS 1.2mm sources, the ancillary 850um sources, and redshift catalogs to study the colors and redshift distributions. We also produce end-to-end simulations based on SIDES and the observed N2CLS detector timelines to interpret the data. We find a mean S2/S1.2 color of 0.222$\pm$0.008 with a standard deviation of 0.070$\pm$0.010. We measure a mean redshift of $3.6\pm0.3$ in GOODS-N, which is marginally higher than expectations from SIDES ($2.9\pm0.2$) because of an overdensity at $z\sim5.2$, and $3.2\pm0.2$ in COSMOS, which agrees with the $3.2\pm0.2$ predicted by SIDES. We also show that the observed S2/S1.2 colors exhibit a weak dependence with redshift but a large dispersion, which limits its efficiency to select high-z sources. Finally, we studied the nine 2mm sources not detected at 1.2mm, and found that two of them are radiogalaxies, one is a z2 galaxy, and the remaining six are compatible with the expected number of spurious detections. The N2CLS survey shows no evidence for any exotic 2mm-only galaxy population. Using SIDES, we show that 2mm samples have a higher mean redshift compared to 1.2mm because they miss z~2 dusty galaxies. Finally, we discuss the efficiency of single-dish and interferometric blind surveys to build samples of high-z dusty galaxies.

💡 Research Summary

This paper presents a comprehensive analysis of galaxies selected at 2 mm using the final data from the NIKA2 Cosmological Legacy Survey (N2CLS). The authors combine the unprecedented 2 mm observations with the SIDES simulated infrared sky to investigate the statistical properties of 2 mm‑selected sources in two well‑studied extragalactic fields: GOODS‑N (deep, 159 arcmin², rms ≈ 0.047 mJy beam⁻¹) and COSMOS (wide, 1130 arcmin², rms ≈ 0.09 mJy beam⁻¹). By applying a high‑reliability detection threshold (S/N ≥ 4.2, corresponding to 95 % purity), they build robust catalogs containing 25 sources in GOODS‑N and 90 sources in COSMOS.

The 2 mm sources are cross‑matched with the N2CLS 1.2 mm catalog (S/N ≥ 3.0–3.9, 80 % purity) using a 6.5″ radius, yielding counterparts for 24 of the 25 GOODS‑N sources and 82 of the 90 COSMOS sources. The authors also incorporate ancillary 850 µm data, high‑resolution interferometric catalogs (AS2COSMOS, A³COSMOS), and deep radio (JVLA 3 GHz) imaging to pinpoint precise positions. Spectroscopic redshifts are gathered from a variety of public surveys (e.g., AS2COSMOS, COSMOS2020/2025, various ALMA follow‑ups), supplemented by photometric redshifts where spectroscopy is unavailable.

Fluxes are measured via PSF‑fitting photometry on match‑filtered maps, and extensive end‑to‑end simulations based on SIDES are used to correct for flux boosting, filtering, and to estimate uncertainties that include correlated noise and confusion. The resulting mean flux ratio S₂mm/S₁.₂mm is 0.215 ± 0.006 with a standard deviation of 0.056 ± 0.004. The dispersion is largely driven by the different beam sizes (12″ at 1.2 mm vs 18″ at 2 mm) and source blending, as confirmed by the simulations.

Redshift analysis shows a mean ⟨z⟩ = 3.6 ± 0.3 for the GOODS‑N sample and ⟨z⟩ = 3.2 ± 0.2 for COSMOS. The GOODS‑N value is slightly higher than the SIDES prediction (⟨z⟩ = 3.0 ± 0.2) due to an overdensity at z ≈ 5.2, while the COSMOS result matches the SIDES expectation (⟨z⟩ = 3.2 ± 0.2). The authors demonstrate that the S₂mm/S₁.₂mm color exhibits only a weak correlation with redshift; the large scatter (σ ≈ 0.07) severely limits the utility of this color alone for selecting high‑z (z > 5) galaxies.

Nine 2 mm sources lack a 1.2 mm detection. Detailed inspection reveals that two are radio‑loud AGN, one is a z ≈ 2 dusty galaxy, and the remaining six (all in COSMOS) are consistent with the number of spurious detections expected from the simulations. Consequently, the survey provides no evidence for an exotic population of “2 mm‑only” galaxies with extremely cold dust or at very high redshift.

Using SIDES, the authors explain why 2 mm‑selected samples have higher average redshifts than 1.2 mm samples: the longer wavelength preferentially misses the abundant z ≈ 2 “cosmic noon” dusty galaxies whose SEDs peak near 1 mm, while retaining a substantial fraction of higher‑z sources that are observed on the Rayleigh‑Jeans tail. This selection effect naturally raises the mean redshift without invoking new galaxy populations.

The paper also compares the N2CLS results with the Ex‑MORA 2 mm blind survey. After accounting for differences in bandwidth, source blending, and intrinsic source sizes, the fluxes of common sources are consistent. The authors discuss the relative merits of single‑dish (30 m IRAM) versus interferometric (ALMA) blind surveys. Single‑dish 2 mm observations excel at covering large sky areas quickly, providing statistically significant samples of high‑z candidates, but suffer from limited resolution and confusion, which hampers precise redshift determination and source deblending. Interferometric surveys, while covering smaller areas, deliver deep integrations and sub‑arcsecond resolution, allowing accurate identification, redshift measurement, and physical characterization of individual high‑z dusty galaxies.

In conclusion, the N2CLS 2 mm data constitute a valuable, complementary dataset for studying the high‑redshift dusty star‑forming galaxy population. The modest color‑redshift correlation limits the efficiency of color‑based pre‑selection, but the higher mean redshift of 2 mm samples confirms that longer‑wavelength surveys can effectively bias toward earlier cosmic epochs. The authors advocate a hybrid strategy: use wide‑area 2 mm single‑dish surveys to flag high‑z candidates, then follow up with high‑resolution interferometric observations to confirm and characterize the galaxies. This approach maximizes the scientific return of future millimeter surveys aimed at probing the early phases of galaxy formation.