Euclid-Dark Galaxies Study

• Euclid-dark galaxies are radio-selected, dust-obscured high-redshift systems undetected in Euclid Q1 bands but identified through LOFAR and Spitzer/IRAC observations.
• They exhibit extreme star-formation rates (up to ~2000 M☉/yr) and high stellar masses (~10^11 M☉), marking them as massive starbursts at z~1–5.
• Multiwavelength SED fitting and radio diagnostics reveal a significant AGN contribution (~40%), highlighting a co-evolution of obscured star formation and black-hole activity.

Searching arXiv for the cited Euclid-dark galaxies paper and closely related work. Searching arXiv for ([2607.07296](/papers/2607.07296)) Euclid Quick Data Release (Q1) Searching for radio-selected Euclid-dark galaxies in the EDF-N. I’m looking up the arXiv entry to verify bibliographic details and related context. Euclid-dark galaxies are radio-selected systems in the Euclid Deep Field–North (EDF-N) that are undetected at $>3\sigma$ in all Euclid Q1 broad bands—VIS $\mathcal{I}_E$, NISP $Y_E$, $J_E$, and $H_E$—with no Euclid Q1 catalogue counterpart, yet are detected at 144 MHz with LOFAR HBA and in the Spitzer/IRAC mid-infrared (Collaboration et al., 8 Jul 2026). In the Euclid Q1 study of EDF-N, these objects are defined operationally through the combination of radio selection, Euclid non-detection, and IRAC identification, yielding a population of dust-obscured and/or high-redshift galaxies that are largely invisible to the initial Euclid imaging but remain accessible through radio and mid-infrared data (Collaboration et al., 8 Jul 2026). The available evidence associates them with heavily obscured, massive star-forming systems, often with active galactic nuclei (AGN), at photometric redshifts extending to $z_{\mathrm{ph}}=5.0$ (Collaboration et al., 8 Jul 2026).

1. Definition and observational setting

In the EDF-N analysis, Euclid-dark galaxies are defined as radio-selected sources with no emission at a more than $3\sigma$ level in Euclid Q1 images and no match in the Euclid Q1 catalogue (Collaboration et al., 8 Jul 2026). The Euclid broad-band limits relevant to this definition are given as $5\sigma$ depths of $\mathcal{I}_E\sim25.45$, $Y_E\sim24.6$, $\mathcal{I}_E$0, and $\mathcal{I}_E$1 (Collaboration et al., 8 Jul 2026). By construction, these sources are bright enough at 144 MHz and in Spitzer/IRAC, but absent from Euclid’s first quick release products.

The physical interpretation advanced for their Euclid non-detection is heavy dust obscuration and/or high redshift, specifically that for $\mathcal{I}_E$2 the stellar peak is shifted into the near-infrared, making the systems fainter than the Euclid NISP detection limits (Collaboration et al., 8 Jul 2026). This does not constitute a purely instrumental class: the defining selection is observational, but the inferred population corresponds to dust-enshrouded galaxies whose emitted light is redistributed toward longer wavelengths and whose synchrotron emission remains visible at radio frequencies.

LOFAR and Spitzer/IRAC play complementary roles in establishing the class. LOFAR HBA at 144 MHz, with $\mathcal{I}_E$3 FWHM and rms $\mathcal{I}_E$4 over $\mathcal{I}_E$5, probes synchrotron emission from star formation or AGN without dust attenuation (Collaboration et al., 8 Jul 2026). Spitzer/IRAC at $\mathcal{I}_E$6–$\mathcal{I}_E$7, with $\mathcal{I}_E$8–$\mathcal{I}_E$9 FWHM and depths of $Y_E$0–$Y_E$1 mag, provides counterpart positions for heavily obscured continua at rest $Y_E$2 (Collaboration et al., 8 Jul 2026). This synergy is central: Euclid-dark galaxies are not simply “missing” from Euclid, but recovered through a multiwavelength selection that is specifically sensitive to obscured systems.

2. Sample construction and selection function

The parent catalogue was B25’s LOFAR multi-band catalogue, comprising $Y_E$3 sources with reliable optical/MIR identifications (Collaboration et al., 8 Jul 2026). From this set, the selection isolated “IRAC-only” radio sources that lack HEROES $Y_E$4 counterparts but are detected in IRAC1/2 (Collaboration et al., 8 Jul 2026). The Euclid-dark sample was then constructed through three explicit steps: adopting $Y_E$5 ($Y_E$6 mJy) from LOFAR HBA; requiring no emission $Y_E$7 at the LOFAR or IRAC position in any Euclid Q1 band; and cross-matching LOFAR and IRAC positions to the Euclid Q1 catalogue within $Y_E$8, retaining only sources with zero matches after visual inspection of Q1 images (Collaboration et al., 8 Jul 2026).

This procedure yields 166 Euclid-dark galaxies over $Y_E$9, corresponding to a surface density of $J_E$0 (Collaboration et al., 8 Jul 2026). The study also defined an “isolated” sub-sample to reduce blending and contamination. In that refinement, $J_E$1 LOFAR/IRAC2 contours were required not to overlap any $J_E$2 Q1 or IRAC neighbour, leaving 88 isolated Euclid-dark galaxies (Collaboration et al., 8 Jul 2026).

The distinction between the full and isolated samples is methodologically important. The 166-source set captures the broader incidence of Euclid-dark systems under the survey selection, whereas the 88-source isolated sample is optimized for cleaner photometry and more reliable physical inference. A plausible implication is that conclusions regarding redshift, stellar mass, and star-formation rate are intentionally anchored to the subset least affected by source confusion.

3. Multiwavelength characterization and SED inference

The photometric basis for physical characterization spans the UV to radio. The study used CFHT $J_E$3; HSC $J_E$4; Euclid VIS $J_E$5; NIR imaging from CFHT, HSC, and Euclid $J_E$6, $J_E$7, $J_E$8; Spitzer IRAC1–4; WISE W3/W4; MIPS $J_E$9; AKARI FIS; Herschel PACS/SPIRE; JCMT SCUBA-2 at $H_E$0; and LOFAR low- and high-resolution data (Collaboration et al., 8 Jul 2026). For non-detections in the optical/NIR, the analysis adopted $H_E$1 upper limits from empty apertures, with analogous upper-limit handling in the MIR/FIR (Collaboration et al., 8 Jul 2026). This treatment is crucial for a class defined by non-detection in Euclid itself, because the upper limits constrain the degree of obscuration and the location of the stellar SED peak.

SED fitting was performed with CIGALE v2022, described as an energy-balance code (Collaboration et al., 8 Jul 2026). The adopted ingredients were Bruzual and Charlot (2003) stellar populations, a Chabrier IMF, and a delayed-$H_E$2 star-formation history of the form

$H_E$3

Dust attenuation followed the Charlot and Fall two-component model with power-law slope $H_E$4; dust emission was modeled with Dale et al. (2014) templates; AGN emission used Fritz et al. (2006) torus models via an AGN fraction $H_E$5; and the radio domain included star-formation and AGN components via $H_E$6 and $H_E$7 parameters (Collaboration et al., 8 Jul 2026). The Bayesian outputs were photometric redshift, $H_E$8, SFR, $H_E$9, $z_{\mathrm{ph}}=5.0$0, $z_{\mathrm{ph}}=5.0$1, and radio luminosities (Collaboration et al., 8 Jul 2026).

For the isolated sample with FIR detections, comprising 26 sources, the inferred photometric redshifts span $z_{\mathrm{ph}}=5.0$2 with median $z_{\mathrm{ph}}=5.0$3; the mean stellar mass is $z_{\mathrm{ph}}=5.0$4; the mean star-formation rate is $z_{\mathrm{ph}}=5.0$5; and the mean dust attenuation is $z_{\mathrm{ph}}=5.0$6 mag (Collaboration et al., 8 Jul 2026). These values place the population among the most massive and dust-obscured star-forming systems identified in wide-area radio-selected samples. The combination of high $z_{\mathrm{ph}}=5.0$7, large stellar masses, and substantial infrared luminosities supports the interpretation that Euclid-dark selection isolates galaxies in which the rest-frame UV/optical continuum is strongly suppressed.

4. Radio diagnostics and AGN identification

The study used two radio-based AGN diagnostics: brightness temperature from sub-arcsecond imaging and radio excess relative to the infrared-radio correlation (IRRC) (Collaboration et al., 8 Jul 2026). For the first, recent International LOFAR Telescope imaging at $z_{\mathrm{ph}}=5.0$8 resolution and rms $z_{\mathrm{ph}}=5.0$9 was used to estimate brightness temperatures according to

$3\sigma$0

The adopted starburst limit from Condon (1991) was

$3\sigma$1

with $3\sigma$2 and $3\sigma$3, implying $3\sigma$4 (Collaboration et al., 8 Jul 2026). Sources with $3\sigma$5 were interpreted as hosting AGN-powered compact radio cores.

The second diagnostic was the IRRC, parameterized as

$3\sigma$6

The analysis adopted the Delhaize et al. (2017) redshift evolution

$3\sigma$7

A source was classified as radio-excess when $3\sigma$8 dex, corresponding to $3\sigma$9 (Collaboration et al., 8 Jul 2026). Within the 53 isolated sources covered by ILT imaging, approximately 34% show radio excess; when brightness-temperature-selected AGN are added, the AGN fraction rises to approximately 40% (Collaboration et al., 8 Jul 2026).

These diagnostics are complementary rather than redundant. Brightness temperature isolates compact high-surface-brightness radio cores, whereas IRRC offsets identify systems with radio output exceeding that expected from star formation alone. The resulting $5\sigma$0 AGN incidence indicates that Euclid-dark selection is particularly effective at finding galaxies in which obscured star formation and black-hole accretion coexist.

5. Host-galaxy properties and starburst character

Among the 88 isolated Euclid-dark galaxies, 26 are detected in the far-infrared (Collaboration et al., 8 Jul 2026). For this FIR-detected subset, the infrared luminosity distribution has $5\sigma$1, placing the objects in the ULIRG regime (Collaboration et al., 8 Jul 2026). Using the Kennicutt (2012) calibration,

$5\sigma$2

the corresponding mean is $5\sigma$3, or approximately $5\sigma$4 (Collaboration et al., 8 Jul 2026). These systems therefore occupy the extreme high-SFR tail of dusty galaxy populations selected over wide survey areas.

The same FIR-detected sources lie approximately $5\sigma$5 dex above the star-forming main sequence at $5\sigma$6, using Popesso et al. (2023), and are therefore consistent with starburst systems (Collaboration et al., 8 Jul 2026). The paper further notes that comparison to RS-NIR dark samples shows similar high masses but even more extreme SFRs, attributed to the shallower FIR depth (Collaboration et al., 8 Jul 2026). This suggests that the Euclid-dark selection, at least in the present data release, preferentially isolates the most luminous obscured star-forming galaxies rather than a complete census of all dusty massive systems.

The astrophysical picture emerging from these measurements is that Euclid-dark galaxies occupy a regime of large dust columns, large stellar masses, and very high instantaneous star-formation rates. In the language of the study, they trace heavily obscured ($5\sigma$7 mag), massive ($5\sigma$8), high-SFR ($5\sigma$9–$\mathcal{I}_E\sim25.45$0) systems at $\mathcal{I}_E\sim25.45$1–5, often with radio power that can rival or exceed the contribution from star formation (Collaboration et al., 8 Jul 2026).

6. Stacking analysis, global trends, and survey-scale implications

A median stacking analysis was performed on the 53 isolated sources with ILT coverage, split into radio-excess and IRRC-consistent subsamples (Collaboration et al., 8 Jul 2026). Median photometry was extracted in UV–optical data from Euclid, Subaru, and CFHT; in IRAC, WISE, Herschel, SCUBA-2; and in LOFAR low-resolution images, and the resulting median SEDs were fitted with CIGALE to derive $\mathcal{I}_E\sim25.45$2, $\mathcal{I}_E\sim25.45$3, $\mathcal{I}_E\sim25.45$4, $\mathcal{I}_E\sim25.45$5, SFR, $\mathcal{I}_E\sim25.45$6, $\mathcal{I}_E\sim25.45$7, and $\mathcal{I}_E\sim25.45$8 (Collaboration et al., 8 Jul 2026).

The non-radio-excess stack yielded $\mathcal{I}_E\sim25.45$9, $Y_E\sim24.6$0, $Y_E\sim24.6$1, $Y_E\sim24.6$2, $Y_E\sim24.6$3, and $Y_E\sim24.6$4 (Collaboration et al., 8 Jul 2026). The radio-excess stack yielded $Y_E\sim24.6$5, $Y_E\sim24.6$6, $Y_E\sim24.6$7, $Y_E\sim24.6$8, $Y_E\sim24.6$9, and $\mathcal{I}_E$00 (Collaboration et al., 8 Jul 2026). Both stacks lie above the main sequence, by approximately $\mathcal{I}_E$01 dex for the non-excess stack and approximately $\mathcal{I}_E$02 dex for the excess stack (Collaboration et al., 8 Jul 2026).

Stack	Derived properties
Non-radio-excess	$\mathcal{I}_E$03, $\mathcal{I}_E$04, $\mathcal{I}_E$05, $\mathcal{I}_E$06, $\mathcal{I}_E$07, $\mathcal{I}_E$08
Radio-excess	$\mathcal{I}_E$09, $\mathcal{I}_E$10, $\mathcal{I}_E$11, $\mathcal{I}_E$12, $\mathcal{I}_E$13, $\mathcal{I}_E$14

The principal result of the stacking analysis is that the two subsamples share similar global stellar and dust properties and differ primarily in their radio emission (Collaboration et al., 8 Jul 2026). This constrains a common misconception that radio-excess Euclid-dark galaxies must necessarily be hosted by a fundamentally distinct galaxy population. The stacked results instead indicate broadly comparable host-galaxy characteristics, with radio diagnostics revealing different levels of AGN contribution superposed on similarly obscured massive systems.

At the survey level, the study argues that Euclid’s $\mathcal{I}_E$15 LOFAR-covered area reveals a lower surface density of Euclid-dark galaxies than smaller deep fields such as COSMOS and GOODS-N, but a higher fraction of AGN-luminous, rarer systems, reflecting the trade-off between area and depth (Collaboration et al., 8 Jul 2026). Future Euclid releases, with approximately 11 visits in EDF-N and approximately 1.3 mag greater depth, together with JWST/MIRI follow-up, are expected to refine redshifts, improve SED coverage, and enable more robust estimates of obscured star-formation-rate density and obscured AGN demographics at cosmic noon (Collaboration et al., 8 Jul 2026). This suggests that the current Euclid-dark sample is an initial wide-area census of a transient co-evolutionary phase rather than the final characterization of the population.