EDP Sciences logo
Subscriber Authentication Point
Search
Menu
Home All issues Volume 697 (May 2025) A&A, 697 (2025) A181 Full HTML
Open Access
Issue
A&A
Volume 697, May 2025
Article Number A181
Number of page(s) 15
Section Extragalactic astronomy
DOI https://doi.org/10.1051/0004-6361/202452869
Published online 20 May 2025

© The Authors 2025

Licence Creative CommonsOpen Access article, published by EDP Sciences, under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

This article is published in open access under the Subscribe to Open model. Subscribe to A&A to support open access publication.

1. Introduction

Over the past few decades, our understanding of galaxy evolution has been greatly enhanced by multiwavelength observations. In particular, the identification of a tight (< 0.3 dex) relation between the stellar masses and the star formation rates (SFRs) of star-forming galaxies (the so-called star-formation main sequence, MS; Daddi et al. 2007; Elbaz et al. 2007; Noeske et al. 2007; Schreiber et al. 2015) at least up to z = 6 (Popesso et al. 2023; Calabrò et al. 2024), with hints of its presence up to z = 9 (Ciesla et al. 2024). This scenario suggests that the bulk of present-day stars formed through a secular process rather than through a series of bursts. Nevertheless, quiescent galaxies (QGs) do not follow this correlation and are located below the MS, indicating a suppressed star-formation for this population. This change in the specific star formation (sSFR ≡ SFR/M*) is not fully understood and corresponds to the clear color bimodality observed between QGs and SFGs, with the latter having typically blue continuum-dominated spectra produced by massive young stars, while the former have redder spectra, dominated by the emission of older, less massive stars (Baldry et al. 2004; Williams et al. 2009; Patel et al. 2012). This color bimodality also seems to be accompanied by a morphological bimodality linked to their morphological evolution (Lee et al. 2013; Osborne et al. 2020; Liu et al. 2019), with SFGs exhibiting disky structures (Sérsic index, n ∼ 1); whereas QGs have more compact morphologies, with a Sérsic index of n ∼ 4 (van der Wel et al. 2014; Ward et al. 2024). This morphological and color bimodality has been shown to arise at high masses (Wuyts et al. 2011; Huertas-Company et al. 2024), suggesting that the rise in this bimodality is more of a mass-driven process, rather than an SFR- or morphology-driven one (Conselice et al. 2024). Additionally, it was shown that the high-mass end log(M*/M⊙) > 10 of the cosmic stellar mass density (SMD) is dominated by QGs for redshifts lower than z = 2 (Brammer et al. 2011; Muzzin et al. 2013; Weaver et al. 2023). As this redshift defines the peak of the SFRD of the Universe (also known as cosmic noon; Madau & Dickinson 2014; Matthews et al. 2021; Leroy et al. 2024) and coincides with a significant fraction of QGs in place at this epoch (Brammer et al. 2011; van der Wel et al. 2014; van Dokkum et al. 2015; Carnall et al. 2023a), it implies that this bimodality between SFGs and QGs is forged early on in its history. Therefore, studying the possible progenitors of massive QGs already in place at z ∼ 2 is essential to our understanding of how SFGs transition to quiescence.

In this context, the combined capabilities of the Spitzer Space Telescope (Spitzer) and the Atacama Large Millimeter/submillimeter Array (ALMA), has shed light on a population of galaxies extremely faint from the rest-frame ultraviolet (UV) up to the optical, making them undetectable, even in the reddest and deepest images of the Hubble Space Telescope (HST; Caputi et al. 2012; Wang et al. 2016, 2019; Franco et al. 2018; Yamaguchi et al. 2019; Williams et al. 2019). This population was seen as a serious candidate for acting as the progenitor of the massive QG population observed at high redshifts (Carnall et al. 2023a), given their significant contribution to the high-mass end (log(M*/M⊙) > 10.3) of the cosmic SFRD and SMD at 3 < z < 6, and their typical halo mass (∼10.13 h−1 M⊙ at z = 4), which aligns with this hypothesis (Wang et al. 2019; Gruppioni et al. 2020). To detect this population, Wang et al. (2019) used the H-dropout selection technique, which consists of searching for sources with very weak detection in the H-band (typically H > 27 mag) and bright detection in the Spitzer/IRAC [4.5 μm]-band (typically [4.5]< 24). Because of these observed properties, this selection resulted in a population of z > 3 massive SFGs, heavily obscured by dust. These galaxies are referred to as “HST-dark”, “H-dropouts”, or “optically dark” galaxies.

To fill the gap between Lyman break galaxies (LBGs; Giavalisco et al. 2004a; de Barros et al. 2014; Arrabal Haro et al. 2020) and H-dropouts at 3 < z < 6, Xiao et al. (2023) extended the HST-dark population to a wider selection (H > 26.5 and [4.5]< 24). This provided a more complete view of SFRD at these redshifts by considering galaxies that are less extreme in terms of dust attenuation and stellar mass. The first studies of this so-called optically-faint galaxy (OFG) population using photometric and ALMA data have located it at the high-mass end of the MS (log(M*/M⊙) > 10). It is characterized by a particularly high dust attenuation (Av > 2), high star formation efficiency (SFE), and a low gas fraction, suggesting a highly efficient conversion of cold gas into stars at work in these systems (Alcalde Pampliega et al. 2019; Gómez-Guijarro et al. 2022, 2023; Xiao et al. 2023, 2024). Nevertheless, Spitzer’s low angular resolution in the NIR have not permitted a thorough morphological study of this population or of galaxies in general at these redshifts.

With the launch of the James Webb Space Telescope (JWST), astronomers now have access to unprecedentedly deep and high angular-resolution near-infrared (NIR) imaging at wavelengths longer than 2 μm. This allows us to spatially resolve the stellar structures of galaxies across cosmic time (Gardner et al. 2006, 2023). While the definition of this population of red, dust-obscured star-forming galaxies, heavily depend on the color criteria used, the first studies using JWST NIR observations have confirmed their high contribution to the high-mass end of the SFRD (∼3.2 × 10−3 M⊙ yr−1 Mpc−3) from z = 3 up to z = 7 (Barrufet et al. 2023) and total SMD, with the latter having been found to be underestimated in the pre-JWST area by ∼20% for 3 < z < 4 and up to ∼40% for 6 < z < 8 (Gottumukkala et al. 2024; Weibel et al. 2024). Additionally, spectroscopic and optical-to-mid-IR (optical/MIR) observations of some of these sources have confirmed their high redshift, massive and dusty properties, while discussing the potential triality of their nature; namely, star-forming (∼70%), quiescent (∼20%), potential active galactic nuclei (AGNs) or young starbursts (∼10%) at 2 < z < 7 (Pérez-González et al. 2023; Barrufet et al. 2025). A few papers have used the resolving power in the NIR to study the internal structures of these galaxies, highlighting their stellar and dust concentrated profiles (Pérez-González et al. 2023; Nelson et al. 2023; Gómez-Guijarro et al. 2023; Smail et al. 2023; Sun et al. 2024). Although these new studies seem to link this population even more strongly to the QGs, their spatially resolved properties are still poorly understood as they only focus on small samples and suffer from completeness bias. We therefore lack morphological evidence that this population corresponds to a key transition phase between SFGs and QGs.

In this work, we study the spatial distribution of different physical quantities for a mass-complete sample at 3 < z < 4, using a resolved SED-fitting procedure. We computed and compared the radial profiles of color-selected QGs, typical blue SFGs (hereafter, referred to as BlueSFGs) and red SFGs (hereafter, RedSFGs) which include, but are not limited to, OFGs (see selection in Fig. 6). In this way we aim to decipher the nature and potential role of the RedSFG population in the morphological transition from BlueSFGs to QGs.

This paper is organized as follows: In Sect. 2, we introduce our data, which includes JWST/NIRCam (Rieke et al. 2003, 2005, 2023a; Beichman et al. 2012) and HST/ACS (Ford et al. 1998; Ryon 2023) observations from the CEERS field, and the associated catalog. We also present our SED fitting analysis, which we use for both resolved and unresolved studies. We take particular care to explain the choice of attenuation law for this work. Section 3 is dedicated to the method used to assemble a mass-complete sample of BlueSFGs, RedSFGs and QGs. In Sect. 4, we analyze the results of this resolved SED fitting method and, in particular, we compare the radial profiles of the physical properties of BlueSFGs, RedSFGs, and QGs. We discuss in Sect. 5 how these results highlight the role of the RedSFG population in galaxy evolution by integrating our results into the broader framework of galaxy evolution. Finally, Sect. 6 summarizes our findings. Throughout this work, we adopt a flat cosmology with [ΩΛ, ΩM]=[0.7, 0.3] and set the Hubble constant to H0 = 70 km s−1. We also quote all magnitudes in the AB system (Oke & Gunn 1983).

2. Observations and data description

2.1. CEERS data

In this study, we used imaging data obtained by the JWST/NIRCam instrument as part of the Cosmic Evolution Early Released Science survey (CEERS; Finkelstein et al. 2017). It has been conducted in the Extended Groth Strip field (EGS) and contains ten pointings, covering a total area of 97 square arcminutes. CEERS provided the first deep extragalactic survey with JWST in seven NIRCam filters (F115W, F150W, F200W, F277W, F356W, F410M, and F444W) with a median 5σ point source depth of 29.20, 29.01, 29.16, 29.15, 29.15, 28.32, and 28.56, respectively. Additionally, we used the F606W and F814W optical-bands from HST observations with the ACS/WFC instrument, having a 5σ point source depth of 28.73 and 28.50, respectively (Davis et al. 2007; Finkelstein et al. 2024).

The JWST/NIRCam data have been reduced following Bagley et al. (2023) and the resulting images were pixel aligned to the HST/ACS images of the Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey (CANDELS; Koekemoer et al. 2011; Grogin et al. 2011). In brief, pointings from June 2022 were processed using the JWST Calibration pipeline v1.7.2 and CRDS pmap 0989, while those from December 2022 were processed using the JWST Calibration pipeline v1.8.5 and CRDS pmap 1023. Then, a customized pipeline developed by the CEERS team incorporating procedures such as 1/f noise subtraction and artifact removal was applied to all pointings. Finally, the mosaics were aligned using astrometry data from Gaia-EDR3 (Gaia Collaboration 2021), with a pixel scale of 0.03 arcsec/pixel, and background subtracted.

2.2. Catalog

We used the photometric catalog built by Gómez-Guijarro et al. (2023), which contains our nine HST/JWST bands of interest (F606W, F814W, F115W, F150W, F200W, F277W, F356W, F410M, and F444W) together with the Canada-France-Hawaii Telescope (CFHT)/MegaCam (Boulade et al. 1998) observations imaging in the EGS field in the u⋆, g′,r′,i′, and z′-bands (Gwyn 2012). This cataloging analysis provided segmentation maps, fluxes, and flux uncertainties for each source across the ten CEERS pointings in both the two HST and seven JWST bands used in this study. To briefly summarize the construction of this catalog, they employed SExtractor (Bertin & Arnouts 1996) following a methodology similar to the one performed in the CANDELS catalog within the EGS field (Stefanon et al. 2017); however, in this case, the F444W-band served as the detection image. Fluxes and their respective errors for each source were measured using Kron elliptical apertures on images that were PSF-matched to the 0.16″ FWHM angular resolution of the F444W-band images. For bands different from the F444W detection band, they corrected to total fluxes and uncertainties using the ratio between Kron fluxes and isophotal fluxes as measured in F444W by SExtractor. They applied an aperture correction for point sources in all bands to account for flux lost outside Kron’s elliptical apertures and scaled the fluxes and uncertainties by the fraction of flux outside the Kron ellipses as measured in the F444W-band PSF. More details about the methodology can be found in Gómez-Guijarro et al. (2023) and Stefanon et al. (2017).

From this photometric catalog, Gómez-Guijarro et al. (2023) estimated the photometric redshift for each source from a SED fitting procedure with the 18 fluxes and corresponding uncertainties from their catalog as input of the EAZy-py code, an updated Python version of the code EAZy (Brammer et al. 2008). To ensure the best photometric redshift determination possible, they employed a set of 13 Flexible stellar population synthesis (FSPS; Conroy & Gunn 2010) templates (corr_sfhz_13; Kokorev et al. 2022) allowing the SED code to span various star-formation histories (SFHs), ages, and dust attenuation, while including the contributions of emission lines to the spectra. When possible thanks to the spectroscopic redshift catalog of Stefanon et al. (2017) or the MOSDEF survey of the EGS field (Kriek et al. 2015), photometric redshifts were replaced by spectroscopic redshifts. Here, we focus our analysis on the 3 < z < 4 redshift range because our aim is to study the potential direct progenitors of QGs at cosmic noon. To ensure a good mass determination, we restricted our analysis to 1755 galaxies with photometric redshifts 3 < z < 4 that are significantly detected (> 5σ) in the F444W and F356W-bands. They defined the parent sample of our study which we have used to build our final and mass-complete samples.

2.3. Mass, star formation rate, and attenuation law

Galaxies of the parent sample were then fitted with Cigale. To compute the stellar component of our SED models, we assumed a delayed, exponentially declining star formation history (SFH) such as SFR ∝ 1 τ 2 × exp ( − t / τ ) $ \mathrm{SFR} \propto \frac{1}{\tau^{2}} \times \exp(-t/\tau) $, where τ defines the time at which the SFR peaks, and used the Bruzual & Charlot (2003) (BC03) stellar population synthesis model together with a Chabrier (2003) initial mass function (IMF) and the Charlot & Fall (2000) (CF00) dust attenuation law (see below for more tests on that matter). To decide whether or not letting metallicity free during the fit, we investigated the constraining power of our UV/NIR data on this parameter by using the mock analysis from Cigale similarly to Seillé et al. (2022). Briefly, Cigale uses the best-fit model fluxes and their associated observed photometric errors to run with the same configuration grid, but with flux values randomly chosen within a Gaussian distribution centered of the best-fit value and the photometric error as a standard deviation. Cigale then compares the result from the initial run with this run based on mock SED. Based on this analysis, we concluded that metallicity was not constrained with our data. Therefore, we decided to fix the metallicity to solar (Z = 0.2), as it has been shown to be a good assumption for massive log(M*/M⊙) > 9.6 galaxies, according to the mass-metallicity relation at this redshift (Ma et al. 2016). We also tested the contribution of a nebular continuum with emission lines on the properties derived from our mass-complete sample (defined in Sect. 3.2). We fit the galaxies with and without the nebular module and found no notable effects on the SFR, Av, and mass-weighted age. Nevertheless, we noticed a slight effect on stellar mass, with the inclusion of the nebular continuum underestimating the mass of ∼0.1 dex on average compared to fit with the nebular component for log(M*/M⊙) < 10 galaxies, in agreement with Yuan et al. (2019). Based on these results, and since our main results are drawn from the most massive log(M*/M⊙) > 10 galaxies of our mass-complete sample, the choice of including nebular emission does not impact the results. Then, we decided to not include the nebular emission while fitting.

Since we are looking to study very dusty galaxies, the choice of an attenuation law is crucial. In fact, different laws can lead to different mass and SFR determinations if the dust emission is not considered, which is the case without FIR data (Buat et al. 2019). With Cigale, we can use a Calzetti et al. (2000) (C00), or a Charlot & Fall (2000) (CF00) attenuation-law. In C00, dust attenuation at a given wavelength is measured using the color excess as Aλ = E(B − V)starskλ where kλ defines the reddening curve and k V = A V E ( B − V ) stars = R V $ k_{\mathrm{V}} = \frac{A_{\mathrm{V}}}{E(B-V)_{\mathrm{stars}}} = R_{\mathrm{V}} $. On the other hand, the CF00 assumes that young (< 10 Myr) and old stars do not undergo the same total attenuation. In this model, young stars have not yet left their birth clouds, so their light is attenuated by both the birth cloud and the ISM; whereas for old stars, their radiation is only affected by the dust in the ISM. To model these two attenuating components, CF00 considers two power-law attenuation curves, one for the birth cloud, A λ BC = A V BC ( λ / 0.55 ) n BC $ A_{\lambda}^{\mathrm{BC}} = A_{\mathrm{V}}^{\mathrm{BC}}(\lambda/0.55)^{n^{\mathrm{BC}}} $, and one for the ISM, A λ ISM = A V ISM ( λ / 0.55 ) n ISM $ A_{\lambda}^{\mathrm{ISM}} = A_{\mathrm{V}}^{\mathrm{ISM}}(\lambda/0.55)^{n^{\mathrm{ISM}}} $, with the ratio between the attenuation of the two components controlled by μ = A V ISM ( A V BC + A V ISM ) $ \mu = \frac{A_{\mathrm{V}}^{\mathrm{ISM}}}{(A_{\mathrm{V}}^{\mathrm{BC}}+A_{\mathrm{V}}^{\mathrm{ISM}})} $. Because the power-law indices and μ are not constrained by UV/NIR data (Buat et al. 2012, 2019; Seillé et al. 2022), we opted for and then tested the initial CF00 formulation with nISM = nBC = −0.7 and μ = 0.3.

Firstly, we compared the SFRs inferred by Cigale while using the C00 or CF00 attenuation law on the parent sample. As the left panel of Fig. 1 suggests, C00 and CF00 SFRs are similar for galaxies with low stellar mass (log(M*/M⊙) < 9.5) at these redshifts, while high-mass sources present significantly higher SFRs when using CF00 attenuation law. We therefore investigated the accuracy of the inferred SFR of both attenuation laws for massive galaxies, taking advantage from the GOODS-ALMA 2.0 catalog (Gómez-Guijarro et al. 2022) that contains massive ALMA-detected sources in the Great Observatories Origins Deep Survey South field (GOODS-South; Dickinson et al. 2003; Giavalisco et al. 2004b). This catalog provides total SFRs that take into account unobscured and obscured SFRs, such that SFRtot = SFRUV + SFRIR. In this catalog, SFRIR was obtained by using mid-IR to mm SED-fit of galaxies detected with a peaked significance level of > 3.5σ using ALMA Band 6. They derived the total LIR by integrating the best-fit curve between rest frame 8 and 1000 μm, and subtracted a potential AGN contribution to IR emission. The deduced LIR therefore only takes into account star formation, and is translated into SFRIR using the Kennicutt (1998) conversion. Then, we cross-matched these 88 galaxies from the GOODS-ALMA 2.0 catalog, which are characterized by log(M*median/M⊙)∼10.77 and zmedian = 2.46 to the JWST Advanced Deep Survey (JADES; Eisenstein et al. 2023a,b) initial data release of the GOODS-South field presented in Rieke et al. (2023b) to obtain their UV/NIR data. We used exactly the same nine photometric bands as those in our CEERS catalog to be as close as possible to the data used in our main study. These 88 galaxies are therefore analogs in terms of masses and redshifts to the RedSFGs defined in Sect. 3.3 with log(M*/M⊙) > 10. We performed SED fits on these galaxies and compared the SFRs obtained using UV/NIR data with a C00 or CF00 attenuation law to the total SFRs from the GOODS-ALMA 2.0 catalog. The right panel of Fig. 1 suggests that the median SFR obtained by the CF00 law is in better agreement with the ALMA SFR, with ( SFR UV / NIR CIGALE / SFR tot ) median = 1 . 1 − 4.2 + 3.9 $ (\mathrm{SFR}_{\mathrm{UV/NIR}}^{\mathrm{CIGALE}}/\mathrm{SFR_{tot}})_{\mathrm{median}} = 1.1_{-4.2}^{+3.9} $ while the one from using the C00 law has ( SFR UV / NIR CIGALE / SFR tot ) median = 0 . 57 − 3.7 + 2.7 $ (\mathrm{SFR}_{\mathrm{UV/NIR}}^{\mathrm{CIGALE}}/\mathrm{SFR_{tot}})_{\mathrm{median}} = 0.57_{-3.7}^{+2.7} $. In view of this result, we decided to use the CF00 attenuation law. The Cigale grid used in the rest of the article (including the resolved SED-fitting) can be found in Table 1. It is important to note that the choice between the CF00 and C00 dust attenuation laws was not based on a comparison between the CEERS UV/NIR data in the EGS field and the FIR GOODS-ALMA data, but rather between the latter and the UV/NIR data from the JADES survey. Nevertheless, while CEERS images are shallower than JADES’s in all NIRCam bands (e.g., F444W 5σ point source depth of 29.77 in JADES and 28.56 in CEERS), these cross-matched galaxies are massive, so their SED fits do not suffer severely from the difference in depth. The conclusion would therefore hold if the same experiment were carried out in the CEERS field, but the absence of ALMA observations makes it impossible to reproduce.

Table 1.

Cigale model parameters.

thumbnail Fig. 1.

Measurement comparisons of SFR obtained by SED-fitting with the CF00 and C00 attenuation law for the parent sample of 1755 galaxies at 3 < z < 4 taken from the Gómez-Guijarro et al. (2023) photometric catalog. The inset shows the comparative histograms of the SFR ratio between the UV/NIR SFR obtained via an SED fitting with the CF00 (dark blue) and C00 (light blue) attenuation law and the SFRtot = SFRUV + SFRIR. The UV/NIR data come from JADES and SFRtot from the GOODS-ALMA 2.0 catalog. For each distribution, the median value (filled circle) along with its associated 16% and 84% percentiles (segments) are displayed. The red dotted vertical line indicates the benchmark whereby 100% of the total SFR is retrieved by the UV/NIR SED fit.

3. Determination and classification of the galaxies of the sample

3.1. Completeness

3.1.1. Determination of the completeness in apparent magnitude

The completeness of a catalog is defined as the fraction of sources in a field that are actually detected as a function of their observed magnitudes. Nevertheless, the detectability of a source not only depends on its flux, but also on its surface brightness and, thus, on its morphology. As we aimed to study a mass-complete sample of galaxies at 3 < z < 4 in this work, we evaluated the 90% limiting magnitude in the F444W detection band of our catalog for a large grid of profiles presented in Table 2.

Table 2.

Simulation input parameters.

We randomly injected 100 sources in each of the ten pointings in bins of magnitudes and with various Sérsic profiles across the whole field. We ensured that the injected profiles did not overlap with pre-existing sources as determined by the SExtractor segmentation map, and we avoided placing injected sources near the edges of the pointings or in the vicinity of a bright source. On these images, we ran the detection procedure used to generate the catalog presented in Sect. 2.2 and computed the ratio Nrecovered/Ninjected for each of the 10 fields. We subsequently calculated the mean of these values to generate the final completeness curves of the total field. The results are summarized in Fig. 2 that shows how the 90% limiting magnitude (i.e. the magnitude at which 90% of the sources are recovered) is influenced by the size and brightness profile of the source. It highlights that for galaxies with sizes close to the PSF FWHM of the detection band, the Sérsic index does not have a significant impact on the detection limit. Nevertheless, differences arise at ∼2 × FWHM with higher Sérsic index (i.e., a cuspier surface brightness) profiles detected down to a higher magnitude than lower Sérsic index profiles. On the upper-right corner of this figure, we display different completeness curves of sources injected with ellipticities ranging from e = 0.9 (highly elliptical) to e = 0 (perfect circle) for a fixed n = 1 and R e = 0 . ″ 42 $ R_{\mathrm{e}}=0{{\overset{\prime\prime}{.}}}42 $. These choices of Sérsic index and Re to derive our detection limit are motivated by the broad morphological study of galaxies at redshift z > 3 conducted in Kartaltepe et al. (2023) using JWST/CEERS data in which they extracted Sérsic index and effective radius distributions for their sample. To derive a conservative estimate of the limiting magnitude, we used the results from the injection-recovery method of sources with n = 1 together with the higher Re found in Kartaltepe et al. (2023) ( 3 kpc = 0 . ″ 42 $ \mathrm{3\,kpc} = 0{{\overset{\prime\prime}{.}}}42 $ at z = 3.5, the median redshift of our study), coupled with a null ellipticity. The orthogonal projection of this galaxy size on the n = 1 curve with e = 0 is shown as green solid lines.

thumbnail Fig. 2.

Determination of the F444W limiting magnitude. The dependence of the 90% limiting magnitude on circularized effective radii from 0 . ″ 06 $ 0{{\overset{\prime\prime}{.}}}06 $ to 1 . ″ 2 $ 1{{\overset{\prime\prime}{.}}}2 $ is shown for an exponential profile (n = 1, blue solid line) and a de Vaucouleurs profile (n = 4, red solid line), both with null ellipticity (e = 0). In the top right-hand corner, we show the completeness curves for a Sérsic profile n = 1 with size R e = 0 . ″ 42 $ R_{\mathrm{e}}=0{{\overset{\prime\prime}{.}}}42 $ and variable ellipticity and the projection of the 90% completeness on the e = 0 curve that is the most restrictive one. The grey shaded region depicts the FWHM size of the F444W-band. The green solid lines defines the F444W = 25 limiting magnitude as the projection on the most restrictive, while realistic, galaxy size ( R e = 0 . ″ 42 $ R_{\mathrm{e}}=0{{\overset{\prime\prime}{.}}}42 $) and ellipticity (e = 0) at z = 3.5.

These results highlight that injection-recovering of only point sources will tend to artificially increase the limit in magnitude compared to extended sources, thus introducing potential bias into subsequent completeness-based analysis, which was already found in Stefanon et al. (2017). In the end, we determined a value of mlim, 4.4 μm = 25 as the most conservative estimate for the limiting magnitude in the F444W-band.

3.1.2. Final sample selection

Starting from the parent sample of 1755 galaxies, we carried out an analysis about photometric redshift uncertainties using different SED-fitting codes to verify the quality of our EAzy-based photometric redshifts. To this end, SED fitting on the parent sample was performed with Beagle (Chevallard & Charlot 2016), Cigale (Boquien et al. 2019), Lephare (Arnouts et al. 1999; Ilbert et al. 2006) and HyperZ (Bolzonella et al. 2000) assuming fitting grids as close as possible to the EAZy one. The median redshift values of these four SED codes were then compared with the redshifts determined by EAZy. The typical uncertainty on these photometric redshifts is ∼10% and in the end, we imposed a relative error cut in the redshift determination, such that |zEaZy − zmedian|/zEaZy < 20%. Thus, in our redshift range of interest (3 < z < 4) and above the stellar mass limit defined in Sect. 3.2, the parent sample contains 237 galaxies. Of these, 21 (∼9%) are excluded by this criterion, with only one that would have been identified as RedSFGs in Sect. 3.3. In what follows, we adopt the photometric redshifts from EAZy.

Next, we identified and removed galaxies with potential AGN contamination from our sample, as this population could distort the stellar properties of the resolved SED-fitting and, consequently, its interpretation. When looking for the signature of AGN with photometric data, we need to use different color criteria, as AGN can be of different natures and therefore detectable at different wavelengths. To seek for unobscured AGN, we cross-matched our sample to the AEGIS-XD catalog of Luo et al. (2017) and computed X-ray luminosities in the rest-frame 2−10 keV band using the 0.5−10 keV flux k-corrected using a power-law X-ray spectrum with a spectral index of Γ = 1.4 similarly to Kocevski et al. (2023a). We then imposed the log(LX/L⊙) > 42.5 threshold commonly used to differentiate unobscured AGN from star-forming galaxies (e.g., Luo et al. 2017) and removed 8 out of 216 galaxies. For obscured AGN, we searched for any IR AGN contribution in the FIR catalog of Henry et al. (in prep.) that uses the super-deblending technique from Liu et al. (2018), Jin et al. (2018). We used the rest-frame UV/NIR fluxes from both our HST/ACS and JWST/NIRCam data and added the super-deblended photometry from Spitzer/IRAC 5.8, 8 μm and MIPS 24 μm, Herschel/PACS 100, 160 μm and SPIRE 250, 350, 500 μm and JCMT/SCUBA2 450, 850 μm. We performed a SED-fit using the Code Investigating GALaxy Emission (Cigale; Boquien et al. 2019) and fit the AGN contribution to the SED of galaxies with S/N > 3 in the 8 μm and 24 μm-bands. By defining the AGN fraction, fAGN, as the AGN contribution to the total IR luminosity L IR AGN = f AGN × L IR tot $ L^{\mathrm{AGN}}_{\mathrm{IR}}=f_{\mathrm{AGN}} \times L^{\mathrm{tot}}_{\mathrm{IR}} $, we removed 4 out of 208 galaxies from our sample that have fAGN > 0.2. Finally, as the accretion disk and the jets of AGNs produce synchrotron emission, radio detections have been historically used to detect heavily obscured AGNs. However, due to synchrotron emission from supernova remnants, SFGs can also be bright in the radio band, and this radiation correlates with their rest-frame 42 to 122 μm FIR emission (Helou et al. 1985; Yun et al. 2001; Magnelli et al. 2015) and (more generally) the rest-frame 8 to 1000 μm IR emission (Bell 2003; Delhaize et al. 2017). Therefore, a commonly used method is to identify AGN as galaxies that deviate by more than 3σ from the correlation followed by SFGs. Following Delvecchio et al. (2017), this is achieved by imposing a lower limit on the radio excess parameter, for instance, as in:

r = log ( L 1.4 GHz [ W Hz − 1 ] SFR IR [ M ⊙ yr − 1 ] ) > 22 ( 1 + z ) 0.013 , $$ \begin{aligned} r={\log }\left(\frac{L_{1.4\,\mathrm{GHz}}[\mathrm{W\,Hz^{-1}}]}{\mathrm{{SFR_{IR}}}[M_{\odot }\,\mathrm{yr^{-1}}]}\right) > 22(1+z)^{0.013}, \end{aligned} $$(1)

where L1.4 GHz is the 1.4 GHz rest-frame luminosity and SFRIR is computed from the integrated 8−1000 μm IR luminosities. Here, we use the super-deblended FIR catalog of Henry et al. (in prep.) that provides SFRIR from FIR+submm (observed 100 μm to 1.1 mm) SED fit and L1.4 GHz using the 3 GHz flux density following Wang et al. (2024). We removed 1 out of 204 additional source following this method. We also report the presence of five sources with extremely low luminosity given their stellar mass of log(M*/M⊙)∼10 in our mass-complete sample. Visual inspection of these sources revealed their extremely small size while showing an extremely high dust attenuation ( A v , median ISM ∼ 4 $ A_{\mathrm{v, median}}^{\mathrm{ISM}}\sim 4 $) and SED expected for “little red dots” (Labbe et al. 2025; Kokorev et al. 2024). Indeed, three of them are selected as LRD by the SED’s continuum slope fit from Kocevski et al. (2024) with two being spectroscopically confirmed zspec = 5.08 and zspec = 5.62 highly obscured AGN showing broad emission lines with NIRSpec (CEERS 5760 in Kocevski et al. 2024 and CEERS 82815 in Kocevski et al. 2023b). We therefore decided to exclude them from this analysis. In the end, we removed 18 potential AGN contaminants from the mass-complete sample of 216 galaxies with secure photometric redshifts.

Finally, to ensure reasonable parameter estimates, we excluded poorly fitted galaxies, defined as galaxies with χ reduced 2 > 1.68 $ \chi_{\mathrm{reduced}}^{2} > 1.68 $ as this upper limit corresponds to a significance level of 10% for data with 8 degrees of freedom (9−1 photometric bands). In other words, a significance level of 10% implies that there is a 10% probability that the observed χ reduced 2 $ \chi_{\mathrm{reduced}}^{2} $ value would be as large or larger purely by chance if the model were correct. Therefore, galaxies with χ reduced 2 $ \chi_{\mathrm{reduced}}^{2} $ values above this limit are considered to have unacceptable fits, indicating that the model does not adequately describe the observed photometry for these galaxies. We removed 10 out of 198 additional galaxies based of this criteria in our mass-complete sample derived in Sect. 3.2. In the end, the final sample consists of 1202 galaxies, free from uncertain redshifts, potential AGN contaminants, and poorly fitted objects.

3.2. Determination of the mass completeness

We used the final sample defined above to define the mass completeness and, ultimately, the mass-complete sample that constitutes the working sample of this study.

In order to obtain a mass-complete sample of galaxies detected at 3 < z < 4, the apparent limiting magnitude must be translated into a limiting mass above which we control the fraction of galaxies that could have been missed by our detection method. However, mass determination suffers from systematic uncertainties even when NIR data are available as its determination can be affected by the SED-fitting code used, the prior assumptions made to compute the stellar populations (Gawiser 2009; Lower et al. 2020; Pacifici et al. 2023; Narayanan et al. 2024), the redshift uncertainty, and most importantly the intrinsic variation of the SED shape of each galaxy according to its formation and evolution history. This in turn implies that it is very inaccurate to simply convert the apparent magnitude limit determined in Sect. 3.1.1 into a limiting mass using simply a constant stellar mass-to-light ratio.

We therefore determined the 90% mass-completeness limit empirically following Schreiber et al. (2015), and exploited the relation between the rest-frame L λ rest $ L^{\mathrm{rest}}_{\lambda} $ where λ = 4.4/(1 + z) = 1 ± 0.1 μm in this range of redshift, and the stellar mass inferred by Cigale using the full UV/NIR photometry of each galaxy. As shown in Fig. 3, the mass versus light relation and its corresponding 1σ scatter vary with mass, the later ranging from ∼0.18 dex at low mass (i.e., log(M*/M⊙)∼8.4) to ∼0.58 dex at high mass (i.e., log(M*/M⊙)∼11).

thumbnail Fig. 3.

Lower panel: Distribution in stellar mass and luminosity at rest-frame wavelength 4.4 μm/(1 + z) with 3 ≤ z ≤ 4 for the final sample of 1202 galaxies. The best-fit relation with its 1σ dispersion are displayed in solid and dashed blue lines, respectively. The position of the F444W = 25 limiting magnitudes for z = 3 and z = 4 are shown with two golden lines. Upper panel: Evolution of completeness with stellar mass at the z = 4 luminosity limit. The horizontal golden line showing the 90% completeness level. Consequently, the green vertical line defines the estimated 90% mass completeness limit. Taking into account galaxies with low mass-to-light ratios, our sample is 99% complete for this population to the right of the red dotted line.

This increasing scatter is mainly due to the presence of a population of galaxies with log(M*/M⊙) > 10 that are outliers of the main relation by showing high mass-to-light ratio. It appears that all of these galaxies that deviate by more than 1σ from the main relation are classified as RedSFGs in Sect. 3.3, therefore hinting at their highly dusty nature ( A v , median ISM ∼ 2.2 $ A_{\mathrm{v, median}}^{\mathrm{ISM}}\sim 2.2 $). We incorporated the evolution of the mass-to-light ratio and its scatter in our limiting mass determination by fitting them with a second degree polynomial function. Using the best fit equations, we generated Gaussian distributions of mock galaxies in stellar mass bins and computed the fraction of galaxies that have their luminosities greater than a given limiting luminosity. Because a source with a given flux can have different deduced rest-frame luminosities depending on its redshift, to be as conservative as possible, we imposed the z = 4 limiting luminosity and considered our sample to be complete in mass when completeness exceeded 90%.

The final mass-complete sample comprises 188 galaxies above a limiting mass of log(M*lim/M⊙) = 9.6 and is cleaned of uncertain redshifts, bad fits, and AGN contaminants (see Sect. 3.1.2). It is shown with its original parent sample in Fig. 4. We note that although this 90% completeness threshold is commonly used in the literature, the 10% of galaxies missed down to this mass limit have by definition a particularly high mass-to-light ratio and are in practical associated with RedSFGs defined in the following section. This population constitutes the focus of this study and could be significantly biased against lower mass RedSFGs if we consider this mass limit. We therefore need to impose an even stricter completeness threshold of 99%, which is reached at log(M*/M⊙) = 10 when comparing galaxies’ properties with those of RedSFGs. Thereafter, we qualify the sample with log(M*/M⊙) > 9.6 as complete in mass and we specify when we are comparing subpopulations of galaxies with log(M*/M⊙) > 10 in particular.

thumbnail Fig. 4.

Distribution of the parent and final mass-complete samples in terms of stellar mass and redshift over the studied redshift range. The parent sample is represented as a light green-scale density map. The horizontal red dashed line indicates the limiting mass of log(M*/M⊙) > 9.6 where the parent sample is 90% complete and contains 237 galaxies. Sources encircled in red define the final mass-complete sample of 188 galaxies used in this paper. This sample is clean from objects with uncertain redshifts, potential AGN contaminants or bad SED fits. The redshift and stellar mass distributions of the two samples are shown as marginal histograms of their respective colors.

3.3. Classifications

The goal of this study is to examine the connection between BlueSFGs, RedSFGs, and QGs. To achieve this, the first step was to categorize each galaxy as SFG or QG. We identified the QG population among our mass-complete sample of 188 galaxies by using the union of the widely used UVJ color criteria (Patel et al. 2012; Fumagalli et al. 2014; Fang et al. 2018; Valentino et al. 2020, 2023; Carnall et al. 2023b) with a criteria based on the distance below the MS: ΔMS = SFR/SFRMS where the SFR in computed by Cigale and is compared to the MS from Schreiber et al. (2015). Here, we used the UVJ diagram as defined by Whitaker et al. (2011):

Quiescent = { U − V > 1.2 , V − J < 1.4 , U − V > 0.88 × ( V − J ) + 0.59 , $$ \begin{aligned} \mathrm{Quiescent} = \left\{ \begin{array}{lll} U-V > 1.2, \\ V-J < 1.4, \\ U-V > 0.88 \times (V-J) + 0.59, \\ \end{array} \right. \end{aligned} $$

and defined QGs, sources that enter the UVJ color selection or have ΔMS < 0.6 (∼1.2 dex below the MS). We checked for their IR detection in the catalog of Henry et al. (in prep.) and verified that these sources were not significantly detected (SNR ∼ 1) in observed bands redder than 24 μm. While two of them have SNR ∼ 2 in the 24 μm band, they all have SFR < 100 M⊙ yr−1, placing them below the 0.5 dex scatter of the MS. This way, we ensured that our QG selection selects objects that have sufficiently evolved to be treated apart from SFGs.

Secondly, we separated the population of red, dust-obscured SFGs (i.e., RedSFGs) from the blue SFGs (i.e., BlueSFGs) by identifying SFGs particularly faint in the rest-frame UV/optical wavelengths. In our redshift range, this intrinsic faintness translates into a faint/extremely faint detection in the optical-to-NIR up to and including the H-band. Previous selection criteria (Wang et al. 2019; Xiao et al. 2023) were based on HST/WFC3 images in the H-band that were shallower than JWST/NIRCam images in the F150W-band. Here, NIRCam data allow us to go a deeper and study sources that are fainter in the observed optical. We therefore decided to adopt

RedSFGs = { F 150 W − F 444 W > 2 , F 444 W < 25 . $$ \begin{aligned} \mathrm{RedSFGs} = \left\{ \begin{array}{ll} F150W - F444W > 2, \\ F444W < 25. \\ \end{array} \right. \end{aligned} $$

Galaxies that do not fit the QGs and RedSFGs criteria are then labeled BlueSFGs. They define the population of more “typical” SFGs and our control sample. The population of RedSFGs described here includes dusty galaxies with strong far-IR emission, generally referred to as dusty star-forming galaxies (DSFGs). However, the color selection made here does not guarantee that all of them are, so we prefer to keep the name RedSFGs to avoid possible confusion. We can see, when looking at the distribution of our sample in the UVJ diagram in Fig. 5, that our RedSFGs criteria selects galaxies that are located apart from the locus of BlueSFGs and preferentially in the DSFG region. In Fig. 6, the H-dropout selection criteria of Wang et al. (2019) is shown by the green dotted triangular region and the Xiao et al. (2023) color selection is represented with a red dotted line. The black solid line marks the ΔMS criteria used in this study. The RedSFGs are shown as red diamonds and lie above this delimiting line. By these criteria, we classified 32 RedSFGs (∼17%), 142 BlueSFGs (∼76%) and 14 QGs (∼7%) in our cleaned mass complete sample of 188 galaxies with 3 < z < 4 defined in Sect. 3.2. We recall that between log(M*/M⊙) = 9.6 and log(M*/M⊙) = 10, our sample can be considered complete for QGs and BlueSFGs but incomplete for RedSFGs. Above log(M*/M⊙) = 10, we can consider our sample complete for BlueSFGs, RedSFGs and QGs.

thumbnail Fig. 5.

Location of BlueSFGs, RedSFGs and QGs at 3 < z < 4 in the UVJ color-color diagram, the purple line delimits the quiescent region as defined in Whitaker et al. (2011).

thumbnail Fig. 6.

F150W − F444W versus F444W color-magnitude distribution for our 3 < z < 4 sample. The markers show the color-selected populations of BlueSFGs, RedSFGs and QGs located above the black solid line defining our RedSFGs selection criteria. As a reference, we indicate the selection criteria for H-dropouts in Wang et al. (2019) (i.e., F150W > 27 mag and F444W < 24 mag) as a green dashed triangular region and the OFG selection from Xiao et al. (2023) (i.e., F150W > 26.5 mag and F444W < 25 mag) as a red dotted line.

4. Results

4.1. Integrated properties

Before studying the resolved properties of BlueSFGs, RedSFGs, and QGs, we compared their integrated properties such as their stellar mass, SFR, and dust attenuation. In Fig. 7, we show the star-forming MS arising from our mass-complete sample. Here, the SFRs are rescaled to the median redshift of the mass-complete sample: zmedian = 3.49. This is done by conserving ΔMS for each galaxy relative to the MS associated with their respective redshift. In this way, the rescaled SFRs represent the overall distribution of the sample in the SFR-stellar mass plane and this ensures that observed differences in SFRs are not simply due to the evolving MS between z = 3 and z = 4 but reflect intrinsic differences in galaxy properties. As expected, the majority of color-selected SFGs lie on the MS within the 1σ uncertainty. In particular, we note that RedSFGs also fall on the MS, although, the lack of mid-IR-to-mm data that are necessary to constrain the SFR of highly dusty galaxies (Buat et al. 2019; Figueira et al. 2022) may induce an underestimation of the SFR for some of them. The location of the QGs in the SFR-stellar mass plane, below the MS, confirms that the redness of these galaxies is not due to dust attenuation as is the case for RedSFGs, but rather to their old stellar population. This is also illustrated in Fig. 8 where QGs are not as dusty ( A v , median ISM ∼ 0.8 $ A_{\mathrm{v,median}}^{\mathrm{ISM}}\sim 0.8 $) as RedSFGs ( A v , median ISM ∼ 2.2 $ A_{\mathrm{v,median}}^{\mathrm{ISM}}\sim 2.2 $) despite their similar stellar mass (log(M*median/M⊙)∼10.5. We note here that despite the higher stellar mass of QGs compared to BlueSFGs (log(M*median/M⊙)∼9.8), their dust attenuation does not evolve with mass – as is the case for BlueSFGs and RedSFGs. They display dust attenuation similar to that of BlueSFGs with log(M*/M⊙)∼10.2. This low A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ demonstrates that despite being compact and cuspy (see Sect. 4.3), these galaxies have stopped funneling gas and dust in their cores, which would have allowed star formation to continue. Instead, they seem to have reached a physical stability. We also witness a clear offset in dust attenuation between BlueSFGs and RedSFGs, with the latter being significantly more attenuated than the first at all stellar masses. Because such a distinction could be due to a bimodal distribution of ellipticities (edge-on galaxies tend to be more attenuated than face-on ones, as the column density of dust and gas along the line of sight is greater), we measured the median ellipticity of RedSFGs on the F444W images. We found the median ellipticity of RedSFGs (emedian = 0.5 ± 0.17) to be similar to the one of log(M*/M⊙) > 10 typical BlueSFGs (emedian = 0.52 ± 0.17), but no trend was found between their ellipticities and dust-obscuration in both cases. This significantly high dust obscuration in RedSFGs is not therefore attributed to a geometrical effect.

thumbnail Fig. 7.

Location of BlueSFGs, RedSFGs, and QGs at 3 < z < 4 in the SFR–M* plane. The Schreiber et al. (2015) MS is displayed as a solid blue line. We represent the 1σ scatter of this MS associated with 0.5 < ΔMS < 2 (∼0.3 dex) as a shaded blue area as well as a wider scatter of 0.33 < ΔMS < 3 (∼0.5 dex) in lighter blue. The ΔMS < 0.6 dex criterion used in combination with the UVJ diagram to define QGs. We rescaled all SFR values to the median redshift of our mass-complete sample (zmedian = 3.51) as explained in the main text. The sample is mass-complete for the three populations above the thick black line.

thumbnail Fig. 8.

Dust attenuation as a function of stellar mass for our final 3 < z < 4 sample. BlueSFGs, RedSFGs and QGs are represented with blue dots, red diamonds and yellow dots with black circle, respectively. Their regression lines and ∼95% uncertainties are shown with a solid, dashed and doted line with a shaded area. We indicate the 99% mass-completeness limit of the sample with a thick black line.

We also studied the relative fractions of BlueSFGs, RedSFGs, and QGs as a function of stellar mass and indicated them with solid and dotted colored lines in Fig. 9. Here, we observe a decrease in the BlueSFGs fraction with increasing stellar mass and an opposite behavior for RedSFGs and QGs that becomes more and more dominant at high mass. At log(M*/M⊙) > 10.4, the BlueSFGs fraction drops below 50%, marking the end of their domination. Consequently, QGs and particularly RedSFGs dominate for log(M*/M⊙) > 10.4. Above log(M*/M⊙) > 10.6, RedSFGs account for over 50% of all galaxies, making them the dominant high-mass population in this redshift range. We note that because of RedSFG incompleteness for log(M*/M⊙) < 10, their fractions might be underestimated in this mass regime. Finally, we find that the number density of the RedSFG sample n ∼ 1.1 × 10−4 Mpc−3 is comparable to the one of compact (Re ∼ 1 kpc) and massive log(M*/M⊙) > 10.5 QGs at z ∼ 2 (van der Wel et al. 2014; van Dokkum et al. 2015). This is an indication that this population could be the direct progenitor of these QGs found at cosmic noon. In the following, we investigate this RedSFG population and its role in the evolution of SFGs toward quiescence by comparing its resolved properties to those of RedSFGs and QGs.

thumbnail Fig. 9.

Mass distribution and relative fractions of the 142 BlueSFGs (blue solid line), 32 RedSFGs (red dashed line), and 14 QGs (yellow dotted line) as a function of stellar mass. The cumulative contribution of RedSFGs and QGs to the total number count as a function of stellar mass is displayed as a black dash-dotted line. The black solid lines show the stellar mass above which > 50% of galaxies are either RedSFGs or QGs. The arrow indicates the mass completeness limit for the RedSFGs population.

4.2. Methodology for obtaining the radial profiles

To obtain spatially resolved stellar population properties for each galaxy in our mass-complete sample, maintaining a high signal-to-noise ratio (S/N) of typically S/N > 3 in the fitted regions was key. This is especially important for bands probing the Balmer break, essential for reliably measuring dust obscuration. Due to higher dust extinction in RedSFGs, compared to BlueSFGs, neither simple pixel-by-pixel binning nor Voronoi binning (Cappellari & Copin 2003) could ensure trustworthy S/N in the UV/optical-bands, resulting in an unfair comparison between the two SFG populations. Consequently, we opted to bin each galaxy into concentric elliptical annuli centered on the brightest pixel value in the F444W-band, with a width corresponding to the size of the F444W PSF ( 0 . ″ 16 $ 0{{\overset{\prime\prime}{.}}}16 $ FWHM). The position angles and ellipticities used for each galaxy were obtained by fitting the corresponding F444W segmentation maps with an ellipse using Photutils (Bradley et al. 2024). Prior to that, we PSF-matched all images to the F444W PSF. We show in Fig. 10 our binning method on RGB, PSF-matched images of a typical BlueSFG, an RedSFG and a QG. This method allowed us to have S/N > 3 in each annulus on the F150W images.

thumbnail Fig. 10.

Illustration of the methodology used to compute radial profiles. From top to bottom, we show an example of BlueSFG, RedSFG, and QG RGB cutouts generated using the F200W, F277W, F356W-bands PSF-matched to the F444W filter. Each cutout has a 2 . ″ × 2 . ″ $ 2{{\overset{\prime\prime}{.}}}\times 2{{\overset{\prime\prime}{.}}} $ angular size and contains a 1 . ″ $ 1{{\overset{\prime\prime}{.}}} $ white bar that defines the scale of the image. The 0.16 . ″ $ 0.16{{\overset{\prime\prime}{.}}} $ angular resolution of the F444W-band is represented by a pink circle. Its radius corresponds to the semi-major axis of annulus 1 as well as the width of each annulus. For each galaxy, five concentric rings with a constant width equal to the FWHM of the F444W-band are displayed. In addition, for each cutout, we present the best model spectra produced by Cigale for annulus 1 (the innermost) and annulus 3. In those, the blue curve is the unattenuated stellar emission.

We computed the total flux in each of the ten photometric bands for each annulus by summing the fluxes on a pixel-by-pixel basis. For the total errors in a given annulus and in a given band, we measured total flux within 20 elliptical apertures in the vicinity of the sources that contained the same number of pixels as the annulus under consideration, and calculated the standard deviation of these measured fluxes. The errors were then assumed to be constant for all pixels belonging to an annulus. Finally, we conducted SED-fitting for each annulus using Cigale using the same grid as in Sect. 2.3 and shown in Table 1. We computed the stellar mass density (Σ⋆) and the star formation rate density (ΣSFR) by dividing the total mass and SFR in the annulus by the total area within each of them.

Figure 10 shows for each studied sub-population, the best fit output by Cigale for the innermost annulus and the third outer annulus (referred to as annulus 1 and 3, respectively). These images and best fits highlight the peculiarity of RedSFGs as being more compact and redder than BlueSFGs and QGs for a given stellar mass. We also acknowledge the high dust obscuration of RedSFGs that significantly impacts the UV and optical part of their SED.

To compare the profiles of the three populations, which span a wide range of stellar mass and SFR, it is necessary to apply a re-normalization in order to remove any dependence of mass and SFR on radial profiles. For example, it is well-known that SFGs follow a mass-size relation (Kormendy & Bender 1996; Shen et al. 2003; Trujillo et al. 2006; Buitrago et al. 2008; Bruce et al. 2012; Ono et al. 2013; van der Wel et al. 2014; Lange et al. 2015; Allen et al. 2017; Dimauro et al. 2019; Mowla et al. 2019; Nedkova et al. 2021; Ward et al. 2024), with the most massive SFGs being more extended, whereas QGs exhibit a distinct sequence in the mass-size plane (Ward et al. 2024) and RedSFGs may also adhere to a different relation due to their pronounced compactness at higher masses compared to BlueSFGs (Gómez-Guijarro et al. 2023). We took this mass effect into account by rescaling the radii to the median mass of our BlueSFGs sample (log(M*median/M⊙)∼9.8) using the Ward et al. (2024) mass-size relation of SFGs. Similarly, we re-scaled the Σ⋆, ΣSFR and A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ profiles to this same median mass using the Schreiber et al. (2015) MS at zmedian = 3.49 and the A v ISM − log ( M * / M ⊙ ) $ A_{\mathrm{v}}^{\mathrm{ISM}}{-}{\log}(M_\ast/M_\odot) $ linear relation of BlueSFGs shown in Fig. 8. This way, we ensured a fair comparison between these three galaxy populations, allowing us to place the QGs and RedSFGs in the context of BlueSFGs. To compare their morphologies, we computed the median of these re-normalized radial profiles for Σ⋆, ΣSFR, A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $, sSFR, and mass-weighted age.

4.3. Radial profiles in Σ⋆, ΣSFR, A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $, sSFR and mass-weighted age

Figure 11 displays the normalized median profiles Σ⋆, ΣSFR, A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $, sSFR and mass-weighted age inferred by our SED fitting procedure. The median Σ⋆ profiles of RedSFGs and QGs are almost identical, leading to similar half-mass radii that are about 37% smaller than those of BlueSFGs. The latter show a shallower increase in Σ⋆ toward the center, indicating a more extensive stellar profile. This indicates that the RedSFG population has developed a bulge characteristic of the QG population.

thumbnail Fig. 11.

From left to right and top to bottom: Stellar mass density (Σ⋆), star formation rate density (ΣSFR), dust attenuation ( A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $), sSFR, and mass-weighted age normalized median radial profiles inferred using Cigale for BlueSFGs (in blue), RedSFGs (in red), and QGs (in yellow). Error bars represent the 68th and 32th quantiles of the distributions in each radial bin. In the Σ⋆ panel we indicate the (normalized) half-mass effective radius for each population with a vertical dashed line.

According to the ΣSFR distribution of QGs, they no longer form stars at all radii, which is not the case for BlueSFGs showing a steady decrease toward their outskirts. On the other hand, the SFR profile of RedSFGs is similar in the core to that of BlueSFGs, but shows a steeper gradient, meaning that these galaxies are still in the process of building up their central stellar bulge.

To compare past star formation, measured by Σ⋆, with current star formation, given by ΣSFR as a function of radius, we measured the sSFR profiles of our galaxies. This gives us indication on ongoing evolution of their structure. The constant sSFR at all radii for BlueSFGs and RedSFGs indicates that both populations will not be changing their structures much in the near future. On the other hand, QGs have suppressed star formation in their cores that may indicate a stabilization of their central regions while the remaining gas is depleted in their outskirts. Since the ΣSFR profile of the QGs seems to be constant, their sSFR profile reflects the Σ⋆ profile. The mass-weighted age profiles of RedSFGs are decaying in their cores similarly to BlueSFGs while being offset to older stellar populations. Nevertheless, this RedSFGs age distribution is not at the QGs level. This indicate that RedSFGs constitute a population of SFGs that have lived long enough to develop an older, and more massive bulge than more typical BlueSFGs.

We see an A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ gradient toward the center of the BlueSFGs and RedSFGs indicating an increasing concentration of dust. However, while this gradient is shallow for BlueSFGs, it is really steep for RedSFGs. On the contrary, QGs do not exhibit any A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ radial gradient, and at all radii have A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ value consistent with very little dust/gas content. The lack of star formation in these QGs is thus mostly due to the lack of gas content for their mass, rather than a low star formation efficiency. These observations of the A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ profile definitively highlight the particularity of RedSFGs as outliers of the star-forming galaxy population with their high attenuation characterized by their high dust concentration.

To study this apparent bimodality between RedSFGs and BlueSFGs as a function of stellar mass, we characterize the slope of the radial profiles presented in Fig. 11 but grouping galaxies in several stellar mass bins. The slopes of these radial profiles are in this analysis simply approximated by a linear fit of the three innermost annuli, and we named CMass, CSFR, C A v ISM $ C_{A_{\mathrm{v}}^{\mathrm{ISM}}} $, CsSFR and CAge the measured slopes that we interpret as the concentrations of these physical quantities. Figure 12 shows that the concentration of stellar mass increases with galaxy mass for BlueSFGs and RedSFGs while QGs are found to have a constant concentration, although probing a relatively narrow mass range. However, we note that the most massive QG exhibit a slightly lower stellar mass concentration which could be interpreted as a sign of recent mergers, possibly dry mergers, which would not increase the stellar mass concentration in the center as in situ star formation does, but rather in their outskirts (Naab et al. 2006; Nipoti et al. 2009). The projected surface density of stars in RedSFGs with log(M*median/M⊙) > 10.5 exhibits the same concentration as QGs, both being higher than the one of BlueSFGs. This suggests that if massive RedSFGs stop forming stars they will end up with the same spatial configuration as QGs. Therefore, from a morphological perspective, RedSFGs could be the progenitors of QGs. In contrast, the most massive BlueSFGs do not show such a concentration of mass, showing that without invoking a violent, major merger event, they feasibly cannot be the progenitors of QGs. However, the dispersion of high-mass BlueSFGs shows that some SFGs (two in this case) have mass profiles similar to those of RedSFGs and QGs while lacking the A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ concentration characteristic of RedSFGs. This could indicate an alternative path to passivity. Nevertheless, the bimodality between BlueSFGs and RedSFGs is very significant when looking at the A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ concentration, with RedSFGs having more concentrated A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ at all masses. This result demonstrates that the BlueSFGs/RedSFGs bimodality is not only linked to stellar mass, but that there is a second-order driver of this bimodality, namely the distribution of dust in galaxies at these redshifts that indicates a compaction event building the bulge in situ.

thumbnail Fig. 12.

From left to right and top to bottom: Σ⋆, ΣSFR, A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $, sSFR, and mass-weighted age concentrations as a function of stellar mass. The different concentrations are determined by measuring the slopes of the three innermost annuli of each curve individually. In each panel, BlueSFGs, RedSFGs, and QG are marked by blue, red and yellow dots, respectively. The medians in stellar mass bins are displayed for each population as diamonds, color-coded by the median dust attenuation in the bin. Uncertainties are represented by shaded regions of the color of the represented population and represent the 68th and 32th quantiles of the distributions in each stellar mass bin. The medians are centered on the median stellar mass of each bin containing at least four galaxies (with the exception of the most massive QG bin, which contains only one galaxy).

We also observe a drop of the SFR, A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $, and sSFR concentrations for the most massive RedSFGs, to the extent that their properties reassemble those of QGs. Coupled with the increased mass-weighted age concentration indicating an older stellar population in the core, this suggests that for these massive RedSFGs with highly concentrated stellar component, an inside-out quenching mechanism is at play.

5. Discussion

We find that above log(M*/M⊙) > 10, where all sub-populations are complete, RedSFGs exhibit steeper radial gradients in stellar mass (Σ⋆) and star formation (ΣSFR) than their parent population of BlueSFGs. In addition, we observe a similar stellar concentration between RedSFGs and QGs, particularly for RedSFGs with log(M*/M⊙) > 10.5. We would add, however, that the Σ⋆ and ΣSFR observed from RedSFGs are located at the upper limit of the 1σ dispersion of BlueSFGs and that these results could therefore become more significant if the number statistic were higher. These trends therefore indicate that this population has and is still assembling a massive bulge in its core, which is not the case for even the most massive BlueSFGs. This bulge growth is probably closely related to the A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ gradient, which is steeper for RedSFGs at all masses compared to BlueSFGs. If we relate the dust obscuration to the gas content of galaxies, our results suggest that RedSFGs have undergone a phase of compaction of their gas content, which has been channeled into their cores, producing a high A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ and SFR gradients. This mechanism therefore allows for compact, MS-like star formation in their cores, particularly around log(M*/M⊙)∼10.5, where the concentration of A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ is highest, which would lead to efficient consumption of the gas into stars and consequently to quenching. We further note that the SFR concentration of RedSFGs derived in this study is likely a lower limit of the true ΣSFR of this population. Indeed, while CF00 seems to better recover the true SFR of massive and dusty galaxies (see Sect. 2.3), the lack of resolved FIR data does not allow for an accurate determination to be made for the SFR distribution (Daddi et al. 2007; Reddy et al. 2012). This could have a particularly high impact at the center of RedSFGs, which is heavily shrouded by dust. The SFR concentration could then be even more pronounced than that observed in this work.

The question of whether or not a compaction phase is necessary to explain the concentration of stellar mass observed in RedSFGs is addressed in Fig. 13. Here, we calculate the evolution of the radial distribution of stars from ΣSFR in each annulus and show that BlueSFGs will not naturally transform into RedSFGs if they maintain their radial ΣSFR profile over the next 300 Myr (blue dotted curve). Instead, a significant compaction event with a steeper ΣSFR profile is required for BlueSFGs to evolve into RedSFGs (blue dashed curve). The duration of this compaction phase, around 300 Myr, corresponds to the mass-weighted median age profile observed at the center of the RedSFGs and QGs and is consistent with the 500 Myr quenching timescale in compact SFGs at z ∼ 2 found in van Dokkum et al. (2015). This suggests that RedSFGs are the result of the evolution of BlueSFGs which experienced a compaction event over a duration of around 300 Myr.

thumbnail Fig. 13.

Median radial profiles of Σ⋆ for BlueSFGs (in solid blue line) and RedSFGs (in solid red line) with their associated uncertainties represented as blue and shaded regions, respectively. The blue dotted curve represent the radial evolution of Σ⋆ for BlueSFGs over the next 300 Myr, derived by applying their median radial profile of star formation surface density (ΣSFR), assumed constant during this period. Additionally, the evolution of the median Σ⋆ radial profile of BlueSFGs over the next 300 Myr considering the ΣSFR profile of RedSFGs is displayed as a dashed blue curve. Each profile is normalised to the Σ⋆ at 4 kpc of RedSFGs. We display the typical median uncertainty of the dotted and dashed profiles by a blue error bar.

Figure 12 (lower left panel) also shows that while most BlueSFGs and RedSFGs do not exhibit any strong radial sSFR gradient (indicating that the stellar structure of these populations will not be modified much by current star formation activity), the most massive RedSFGs break this trend by displaying a similar sSFR profile to the QGs. We interpret this population as being depleted of gas at its center (decrease in its A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ concentration), while continuing to form stars at its periphery. This leads to a decrease in its stellar mass concentration and a converging to that of the QG population.

Finally, the median mass of the RedSFG population (log(M*/M⊙)∼10.5), together with its dominance above this mass, and our observation that the most massive RedSFGs exhibit a typical QG stellar mass profile, point in the direction of a major compaction event that allows for the onset of the in situ growth of a bulge. In this scenario, some high redshift BlueSFGs are experiencing this compaction event after spending some time on the MS and reaching a typical stellar mass of log(M*/M⊙)∼10.5. This major compaction is assumed to be “wet” according to Dekel & Burkert (2014), Zolotov et al. (2015) so that during this phase, angular momentum loss allows cold gas to be efficiently channeled into galaxy cores, producing MS-type star formation with optically thick dust and ISM enrichment. What triggers such a compaction phase is beyond the scope of this article, but scenarios such as a major merger or consecutive minor mergers (Dekel & Cox 2006; Covington et al. 2011), counter-rotating cold gas accretion (Danovich et al. 2015), recycling fountains (Chabanier et al. 2020) or satellite tidal compression (Renaud et al. 2014) are advocated by simulations and facilitate efficient in situ bulge mass growth. The result is a compact mass profile consistent with that of passive galaxies, once all the gas in the galaxies has been depleted and star formation has stopped. To reconcile the high dust content of RedSFGs with the small but still existing amount of dust in z ∼ 2 QGs, it is worth noting that the typical time scale for dust removal in post-starburst galaxies is ∼2 Gyr (Michałowski et al. 2019). Therefore, the most massive RedSFGs could be in this post-starburst phase, removing their dust while keeping a highly concentrated mass profile, then converging toward the properties of QGs at z ∼ 2.

The number density of RedSFGs (n ∼ 1.23 × 10−4 Mpc−3) is of the same order of magnitude as the one of QGs at z ∼ 2 (n ∼ 10−4 Mpc−3) (McLeod et al. 2021). This suggests that RedSFGs at 3 < z < 4 are good candidate for being the progenitor of typical massive QGs at z ∼ 2 (McLeod et al. 2021) and is also consistent with the fact that the bulk of QGs at z ∼ 2 are compact (van der Wel et al. 2014; van Dokkum et al. 2015).

We also note that the typical stellar mass (log(M*/M⊙)∼10.5) of RedSFGs corresponds to the critical mass defining the bimodality between BlueSFGs and red QGs in the Local Universe (Baldry et al. 2004). This so-called critical mass was shown to be explained by a wet compaction scenario (Zolotov et al. 2015; Tacchella et al. 2016; Dekel et al. 2019). In this scenario, the wet compaction phase occurs close to this stellar mass, as it is confined by two main physical mechanisms: supernova feedback and virial shock heating of the circumgalactic medium (CGM). On the one hand, for M* < Mcrit, the galaxies’ potential well is not deep enough to retain the gas expelled by supernova feedback or to allow for a significant portion of the gas to heat up, making them particularly effective at suppressing star formation and consequently stopping the gas compaction. On the other hand, for M* > Mcrit, the CGM surrounding the halos is heated by viral shocks that warm the incoming cold gas coming from the intergalactic medium via filaments. The result is a small amount of cold gas, which is not conducive to efficient star formation. It is also suggested that the growth of a super-massive black hole in such a massive system is possible and would guarantee long-term quenching through AGN feedback (Weinberger et al. 2018) once star formation has stopped. According to Pérez-González et al. (2023) and Barrufet et al. (2025) it is possible that ∼10% of RedSFGs contain a highly obscured AGN.

Overall, this scenario seems consistent with our observations of this red population and suggests that RedSFGs can be defined as a phase in galaxy evolution that creates a bimodality between blue and red SFGs. This bimodality in SFGs in place at z = 3 − 4 defines a primeval bimodality that leads to the formation of today’s massive ellipticals.

6. Summary

In this work, we study the transition from SFGs to QGs by investigating the morphology of a color-selected, massive, and dust-obscured population of galaxies, faint in the observed UV/optical, but bright in the NIR – making them particularly red (hence, the nomer RedSFGs used in this work). We focused on this population in the context of a mass-complete sample of 188 galaxies at 3 < z < 4 with log(M*/M⊙) > 9.6, cleaned from potential AGN contaminants, point sources, and galaxies with uncertain photometric redshift determination in the JWST/CEERS field.

Among our mass-complete sample, ∼17% of the galaxies are classified as red RedSFGs, ∼76% are blue BlueSFGs, and ∼7% are classified as QGs. We fit the resolved SED of each galaxy and derived its median Σ⋆, ΣSFR, sSFR, A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $, and mass-weighted age profiles. Our conclusions can be summarized as follows:

  1. RedSFGs and QGs taken together account for over 50% of galaxies at log(M*/M⊙) > 10.5 and this fraction increases with mass. They make up > 80% at log(M*/M⊙)∼11 with a large predominance of RedSFGs for log(M*/M⊙) > 10.6, where they represent 50% of all galaxies.

  2. The radial profiles confirm the compactness of the RedSFGs stellar mass distribution, which is similar to that of the QGs. This suggests that this population is the link in terms of morphological transition between disk-dominated BlueSFGs and bulge-dominated QGs. Their mass-weighted age profiles are also shifted to older ages compared to BlueSFGs, while showing an overall trend similar to them. It confirms that RedSFGs are SFGs on their way to becoming QGs after a certain amount of time on the MS.

  3. The ΣSFR and dust attenuation profiles of RedSFGs also show a high concentration and indicate a gas compaction event. These profiles suggest that we are witnessing the massive bulge growth necessary to explain the compact stellar morphology of RedSFGs. In addition, we tested whether BlueSFGs were able to obtain the Σ⋆ profiles of RedSFGs in less than 300 Myr by applying the ΣSFR profile of RedSFGs to them. We have found that only the application of RedSFGs ΣSFR profile can produce a Σ⋆ distribution similar to that of RedSFGs. This reinforces the need for a compaction event happening in situ to become RedSFGs and, subsequently, QGs. This compaction event then goes on to create a morphological bimodality between BlueSFGs and RedSFGs.

  4. We have investigated the evolution with mass of this bimodality and found that the dust attenuation, Σ⋆, and ΣSFR are more concentrated at all masses in RedSFGs, with a peak of these concentrations for log(M*/M⊙)∼10.5. The fact that the most massive BlueSFGs, which do not show any sign of compaction, do not reach the stellar mass concentration of QGs, proves that increasing stellar mass alone cannot drive a SFG toward quiescence and that compaction is required.

  5. The most massive RedSFGs in our sample, although showing a highly concentrated bulge, strong dust attenuation, and MS-type star formation, also show a significant decrease in ΣSFR, A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $, and sSFR concentrations corresponding to suppressed star formation in their cores which contrasts with lower mass RedSFGs. We interpret these deviations from the general mass trends as a sign that these galaxies are engaged in an inside-out quenching process.

  6. Our results suggest that the RedSFG population is linked to the critical mass log(M*crit/M⊙)∼10.5 defining the bimodality between QGs and SFGs in the Local Universe. This validates the nature of this population as the primeval link between SFGs and QGs.

Our findings can be related to the galaxy evolution scenario suggested in Behroozi et al. (2013), Dekel et al. (2019) by assuming that RedSFGs define a population of SFGs that undergo a compaction phase. They experience the final compaction of their gas content once they reach the critical mass log(M*crit/M⊙)∼10.5, leading to inside-out quenching. This means that the so-called bimodality between local SFGs and QGs originates in a bimodality already existing at z = 3 − 4 between SFGs that have experienced (RedSFGs) or are still experiencing a compaction event and those that have not (BlueSFGs). We suggest that this primeval bimodality takes its seed in relatively rapid compaction events (< 300 Myr) that lead to a short depletion timescale to transform the gas into stars in a concentrated, dust-shrouded region. The dust channeled into the core and produced by star formation activity is trapped in the galaxy’s gravitational potential, producing a steep A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $ gradient characteristic of the RedSFG phase. To put better constrains on the dust and gas content in this obscured population, deep and high-resolution imaging with ALMA would be necessary. Combined with multiwavelength photometry, especially with MIRI, this would allow better stellar mass estimates of these obscured system and shed light on the possible AGN contribution in their cores. Spatially resolved spectroscopy could also be of interest to study the metallicity and age gradients as well as the resolved SFH and velocity profiles of these galaxies. In particular, studies of the kinematic of RedSFG would allow us to measure their three-dimensional stellar distribution and explore their triaxial properties. This three-dimensional information is vital for testing and confirming the in situ bulge formation scenario that we are highlighting in this work. Putting all these informations together, we will be able to lift the veil on the physical process underlying the major compaction phase in the early Universe.

Acknowledgments

MT acknowledges support from CNES. This work was supported by the Programme National Cosmology et Galaxies (PNCG) of CNRS/INSU with INP and IN2P3, co-funded by CEA and CNES. MF acknowledges financial support from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 101148925. MT acknowledges the following open sources softwares used in this work: Numpy (Harris et al. 2020), Astropy (Astropy Collaboration 2013, 2018, 2022), Photutils (Bradley et al. 2024) and SciencePlots (Garrett 2021).

References

  1. Alcalde Pampliega, B., Pérez-González, P. G., Barro, G., et al. 2019, ApJ, 876, 135 [NASA ADS] [CrossRef] [Google Scholar]
  2. Allen, R. J., Kacprzak, G. G., Glazebrook, K., et al. 2017, ApJ, 834, L11 [NASA ADS] [CrossRef] [Google Scholar]
  3. Arnouts, S., Cristiani, S., Moscardini, L., et al. 1999, MNRAS, 310, 540 [Google Scholar]
  4. Arrabal Haro, P., Rodríguez Espinosa, J. M., Muñoz-Tuñón, C., et al. 2020, MNRAS, 495, 1807 [NASA ADS] [CrossRef] [Google Scholar]
  5. Astropy Collaboration (Robitaille, T. P., et al.) 2013, A&A, 558, A33 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  6. Astropy Collaboration (Price-Whelan, A. M., et al.) 2018, AJ, 156, 123 [Google Scholar]
  7. Astropy Collaboration (Price-Whelan, A. M., et al.) 2022, ApJ, 935, 167 [NASA ADS] [CrossRef] [Google Scholar]
  8. Bagley, M. B., Finkelstein, S. L., Koekemoer, A. M., et al. 2023, ApJ, 946, L12 [NASA ADS] [CrossRef] [Google Scholar]
  9. Baldry, I. K., Glazebrook, K., Brinkmann, J., et al. 2004, ApJ, 600, 681 [Google Scholar]
  10. Barrufet, L., Oesch, P. A., Weibel, A., et al. 2023, MNRAS, 522, 449 [NASA ADS] [CrossRef] [Google Scholar]
  11. Barrufet, L., Oesch, P., Marques-Chaves, R., et al. 2025, MNRAS, 537, 3453 [Google Scholar]
  12. Behroozi, P. S., Wechsler, R. H., & Conroy, C. 2013, ApJ, 770, 57 [NASA ADS] [CrossRef] [Google Scholar]
  13. Beichman, C. A., Rieke, M., Eisenstein, D., et al. 2012, SPIE Conf. Ser., 8442, 84422N [NASA ADS] [Google Scholar]
  14. Bell, E. F. 2003, ApJ, 586, 794 [Google Scholar]
  15. Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  16. Bolzonella, M., Miralles, J. M., & Pelló, R. 2000, A&A, 363, 476 [NASA ADS] [Google Scholar]
  17. Boquien, M., Burgarella, D., Roehlly, Y., et al. 2019, A&A, 622, A103 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  18. Boulade, O., Vigroux, L. G., Charlot, X., et al. 1998, SPIE Conf. Ser., 3355, 614 [Google Scholar]
  19. Bradley, L., Sipőcz, B., Robitaille, T., et al. 2024, https://doi.org/10.5281/zenodo.12585239 [Google Scholar]
  20. Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008, ApJ, 686, 1503 [Google Scholar]
  21. Brammer, G. B., Whitaker, K. E., van Dokkum, P. G., et al. 2011, ApJ, 739, 24 [NASA ADS] [CrossRef] [Google Scholar]
  22. Bruce, V. A., Dunlop, J. S., Cirasuolo, M., et al. 2012, MNRAS, 427, 1666 [NASA ADS] [CrossRef] [Google Scholar]
  23. Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000 [NASA ADS] [CrossRef] [Google Scholar]
  24. Buat, V., Noll, S., Burgarella, D., et al. 2012, A&A, 545, A141 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  25. Buat, V., Ciesla, L., Boquien, M., Małek, K., & Burgarella, D. 2019, A&A, 632, A79 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  26. Buitrago, F., Trujillo, I., Conselice, C. J., et al. 2008, ApJ, 687, L61 [Google Scholar]
  27. Calabrò, A., Pentericci, L., Santini, P., et al. 2024, A&A, 690, A290 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  28. Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682 [NASA ADS] [CrossRef] [Google Scholar]
  29. Cappellari, M., & Copin, Y. 2003, MNRAS, 342, 345 [Google Scholar]
  30. Caputi, K. I., Dunlop, J. S., McLure, R. J., et al. 2012, ApJ, 750, L20 [NASA ADS] [CrossRef] [Google Scholar]
  31. Carnall, A. C., McLeod, D. J., McLure, R. J., et al. 2023a, MNRAS, 520, 3974 [NASA ADS] [CrossRef] [Google Scholar]
  32. Carnall, A. C., McLure, R. J., Dunlop, J. S., et al. 2023b, Nature, 619, 716 [NASA ADS] [CrossRef] [Google Scholar]
  33. Chabanier, S., Bournaud, F., Dubois, Y., et al. 2020, A&A, 643, L8 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  34. Chabrier, G. 2003, PASP, 115, 763 [Google Scholar]
  35. Charlot, S., & Fall, S. M. 2000, ApJ, 539, 718 [Google Scholar]
  36. Chevallard, J., & Charlot, S. 2016, MNRAS, 462, 1415 [NASA ADS] [CrossRef] [Google Scholar]
  37. Ciesla, L., Elbaz, D., Ilbert, O., et al. 2024, A&A, 686, A128 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  38. Conroy, C., & Gunn, J. E. 2010, ApJ, 712, 833 [Google Scholar]
  39. Conselice, C. J., Basham, J. T. F., Bettaney, D. O., et al. 2024, MNRAS, 531, 4857 [Google Scholar]
  40. Covington, M. D., Primack, J. R., Porter, L. A., et al. 2011, MNRAS, 415, 3135 [CrossRef] [Google Scholar]
  41. Daddi, E., Dickinson, M., Morrison, G., et al. 2007, ApJ, 670, 156 [NASA ADS] [CrossRef] [Google Scholar]
  42. Danovich, M., Dekel, A., Hahn, O., Ceverino, D., & Primack, J. 2015, MNRAS, 449, 2087 [Google Scholar]
  43. Davis, M., Guhathakurta, P., Konidaris, N. P., et al. 2007, ApJ, 660, L1 [NASA ADS] [CrossRef] [Google Scholar]
  44. de Barros, S., Schaerer, D., & Stark, D. P. 2014, A&A, 563, A81 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  45. Dekel, A., & Burkert, A. 2014, MNRAS, 438, 1870 [Google Scholar]
  46. Dekel, A., & Cox, T. J. 2006, MNRAS, 370, 1445 [CrossRef] [Google Scholar]
  47. Dekel, A., Lapiner, S., & Dubois, Y. 2019, ArXiv e-prints [arXiv:1904.08431] [Google Scholar]
  48. Delhaize, J., Smolčić, V., Delvecchio, I., et al. 2017, A&A, 602, A4 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  49. Delvecchio, I., Smolčić, V., Zamorani, G., et al. 2017, A&A, 602, A3 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  50. Dimauro, P., Huertas-Company, M., Daddi, E., et al. 2019, MNRAS, 489, 4135 [NASA ADS] [CrossRef] [Google Scholar]
  51. Eisenstein, D. J., Johnson, B. D., Robertson, B., et al. 2023a, ArXiv e-prints [arXiv:2310.12340] [Google Scholar]
  52. Eisenstein, D. J., Willott, C., Alberts, S., et al. 2023b, ArXiv e-prints [arXiv:2306.02465] [Google Scholar]
  53. Elbaz, D., Daddi, E., Le Borgne, D., et al. 2007, A&A, 468, 33 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  54. Fang, J. J., Faber, S. M., Koo, D. C., et al. 2018, ApJ, 858, 100 [NASA ADS] [CrossRef] [Google Scholar]
  55. Figueira, M., Pollo, A., Małek, K., et al. 2022, A&A, 667, A29 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  56. Finkelstein, S. L., Dickinson, M., Ferguson, H. C., et al. 2017, The Cosmic Evolution Early Release Science (CEERS) Survey, JWST Proposal ID 1345. Cycle 0 Early Release Science [Google Scholar]
  57. Finkelstein, S. L., Leung, G. C. K., Bagley, M. B., et al. 2024, ApJ, 969, L2 [NASA ADS] [CrossRef] [Google Scholar]
  58. Ford, H. C., Bartko, F., Bely, P. Y., et al. 1998, SPIE Conf. Ser., 3356, 234 [Google Scholar]
  59. Franco, M., Elbaz, D., Béthermin, M., et al. 2018, A&A, 620, A152 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  60. Fumagalli, M., Labbé, I., Patel, S. G., et al. 2014, ApJ, 796, 35 [Google Scholar]
  61. Gaia Collaboration (Brown, A. G. A., et al.) 2021, A&A, 649, A1 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  62. Gardner, J. P., Mather, J. C., Clampin, M., et al. 2006, Space Sci. Rev., 123, 485 [Google Scholar]
  63. Gardner, J. P., Mather, J. C., Abbott, R., et al. 2023, PASP, 135, 068001 [NASA ADS] [CrossRef] [Google Scholar]
  64. Garrett, J. D. 2021, http://doi.org/10.5281/zenodo.4106649 [Google Scholar]
  65. Gawiser, E. 2009, New Astron. Rev., 53, 50 [Google Scholar]
  66. Giavalisco, M., Dickinson, M., Ferguson, H. C., et al. 2004a, ApJ, 600, L103 [NASA ADS] [CrossRef] [Google Scholar]
  67. Giavalisco, M., Ferguson, H. C., Koekemoer, A. M., et al. 2004b, ApJ, 600, L93 [NASA ADS] [CrossRef] [Google Scholar]
  68. Gómez-Guijarro, C., Elbaz, D., Xiao, M., et al. 2022, A&A, 658, A43 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  69. Gómez-Guijarro, C., Magnelli, B., Elbaz, D., et al. 2023, A&A, 677, A34 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  70. Dickinson, M., Giavalisco, M., & GOODS Team 2003, in The Mass of Galaxies at Low and High Redshift, eds. R. Bender, & A. Renzini, 324 [Google Scholar]
  71. Gottumukkala, R., Barrufet, L., Oesch, P. A., et al. 2024, MNRAS, 530, 966 [NASA ADS] [CrossRef] [Google Scholar]
  72. Grogin, N. A., Kocevski, D. D., Faber, S. M., et al. 2011, ApJS, 197, 35 [NASA ADS] [CrossRef] [Google Scholar]
  73. Gruppioni, C., Béthermin, M., Loiacono, F., et al. 2020, A&A, 643, A8 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  74. Gwyn, S. D. J. 2012, AJ, 143, 38 [Google Scholar]
  75. Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357 [NASA ADS] [CrossRef] [Google Scholar]
  76. Helou, G., Soifer, B. T., & Rowan-Robinson, M. 1985, ApJ, 298, L7 [Google Scholar]
  77. Huertas-Company, M., Iyer, K. G., Angeloudi, E., et al. 2024, A&A, 685, A48 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  78. Ilbert, O., Arnouts, S., McCracken, H. J., et al. 2006, A&A, 457, 841 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  79. Jin, S., Daddi, E., Liu, D., et al. 2018, ApJ, 864, 56 [Google Scholar]
  80. Kartaltepe, J. S., Rose, C., Vanderhoof, B. N., et al. 2023, ApJ, 946, L15 [NASA ADS] [CrossRef] [Google Scholar]
  81. Kennicutt, R. C. 1998, ApJ, 498, 541 [NASA ADS] [CrossRef] [Google Scholar]
  82. Kocevski, D. D., Barro, G., McGrath, E. J., et al. 2023a, ApJ, 946, L14 [NASA ADS] [CrossRef] [Google Scholar]
  83. Kocevski, D. D., Onoue, M., Inayoshi, K., et al. 2023b, ApJ, 954, L4 [NASA ADS] [CrossRef] [Google Scholar]
  84. Kocevski, D. D., Finkelstein, S. L., Barro, G., et al. 2024, ArXiv e-prints [arXiv:2404.03576] [Google Scholar]
  85. Koekemoer, A. M., Faber, S. M., Ferguson, H. C., et al. 2011, ApJS, 197, 36 [NASA ADS] [CrossRef] [Google Scholar]
  86. Kokorev, V., Brammer, G., Fujimoto, S., et al. 2022, ApJS, 263, 38 [NASA ADS] [CrossRef] [Google Scholar]
  87. Kokorev, V., Caputi, K. I., Greene, J. E., et al. 2024, ApJ, 968, 38 [NASA ADS] [CrossRef] [Google Scholar]
  88. Kormendy, J., & Bender, R. 1996, ApJ, 464, L119 [NASA ADS] [CrossRef] [Google Scholar]
  89. Kriek, M., Shapley, A. E., Reddy, N. A., et al. 2015, ApJS, 218, 15 [NASA ADS] [CrossRef] [Google Scholar]
  90. Labbe, I., Greene, J. E., Bezanson, R., et al. 2025, ApJ, 978, 92 [NASA ADS] [CrossRef] [Google Scholar]
  91. Lange, R., Driver, S. P., Robotham, A. S. G., et al. 2015, MNRAS, 447, 2603 [CrossRef] [Google Scholar]
  92. Lee, B., Giavalisco, M., Williams, C. C., et al. 2013, ApJ, 774, 47 [NASA ADS] [CrossRef] [Google Scholar]
  93. Leroy, L., Elbaz, D., Magnelli, B., et al. 2024, A&A, 691, A248 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  94. Liu, D., Daddi, E., Dickinson, M., et al. 2018, ApJ, 853, 172 [Google Scholar]
  95. Liu, C., Hao, L., Wang, H., & Yang, X. 2019, ApJ, 878, 69 [NASA ADS] [CrossRef] [Google Scholar]
  96. Lower, S., Narayanan, D., Leja, J., et al. 2020, ApJ, 904, 33 [NASA ADS] [CrossRef] [Google Scholar]
  97. Luo, B., Brandt, W. N., Xue, Y. Q., et al. 2017, ApJS, 228, 2 [Google Scholar]
  98. Ma, X., Hopkins, P. F., Faucher-Giguère, C.-A., et al. 2016, MNRAS, 456, 2140 [NASA ADS] [CrossRef] [Google Scholar]
  99. Madau, P., & Dickinson, M. 2014, ARA&A, 52, 415 [Google Scholar]
  100. Magnelli, B., Ivison, R. J., Lutz, D., et al. 2015, A&A, 573, A45 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  101. Matthews, A. M., Condon, J. J., Cotton, W. D., & Mauch, T. 2021, ApJ, 914, 126 [CrossRef] [Google Scholar]
  102. McLeod, D. J., McLure, R. J., Dunlop, J. S., et al. 2021, MNRAS, 503, 4413 [NASA ADS] [CrossRef] [Google Scholar]
  103. Michałowski, M. J., Hjorth, J., Gall, C., et al. 2019, A&A, 632, A43 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  104. Mowla, L. A., van Dokkum, P., Brammer, G. B., et al. 2019, ApJ, 880, 57 [NASA ADS] [CrossRef] [Google Scholar]
  105. Muzzin, A., Marchesini, D., Stefanon, M., et al. 2013, ApJ, 777, 18 [NASA ADS] [CrossRef] [Google Scholar]
  106. Naab, T., Khochfar, S., & Burkert, A. 2006, ApJ, 636, L81 [NASA ADS] [CrossRef] [Google Scholar]
  107. Narayanan, D., Lower, S., Torrey, P., et al. 2024, ApJ, 961, 73 [NASA ADS] [CrossRef] [Google Scholar]
  108. Nedkova, K. V., Häußler, B., Marchesini, D., et al. 2021, MNRAS, 506, 928 [NASA ADS] [CrossRef] [Google Scholar]
  109. Nelson, E. J., Suess, K. A., Bezanson, R., et al. 2023, ApJ, 948, L18 [NASA ADS] [CrossRef] [Google Scholar]
  110. Nipoti, C., Treu, T., & Bolton, A. S. 2009, ApJ, 703, 1531 [Google Scholar]
  111. Noeske, K. G., Weiner, B. J., Faber, S. M., et al. 2007, ApJ, 660, L43 [CrossRef] [Google Scholar]
  112. Oke, J. B., & Gunn, J. E. 1983, ApJ, 266, 713 [NASA ADS] [CrossRef] [Google Scholar]
  113. Ono, Y., Ouchi, M., Curtis-Lake, E., et al. 2013, ApJ, 777, 155 [NASA ADS] [CrossRef] [Google Scholar]
  114. Osborne, C., Salim, S., Damjanov, I., et al. 2020, ApJ, 902, 77 [NASA ADS] [CrossRef] [Google Scholar]
  115. Pacifici, C., Iyer, K. G., Mobasher, B., et al. 2023, ApJ, 944, 141 [NASA ADS] [CrossRef] [Google Scholar]
  116. Patel, S. G., Holden, B. P., Kelson, D. D., et al. 2012, ApJ, 748, L27 [NASA ADS] [CrossRef] [Google Scholar]
  117. Pérez-González, P. G., Barro, G., Annunziatella, M., et al. 2023, ApJ, 946, L16 [CrossRef] [Google Scholar]
  118. Popesso, P., Concas, A., Cresci, G., et al. 2023, MNRAS, 519, 1526 [Google Scholar]
  119. Reddy, N., Dickinson, M., Elbaz, D., et al. 2012, ApJ, 744, 154 [NASA ADS] [CrossRef] [Google Scholar]
  120. Renaud, F., Bournaud, F., Kraljic, K., & Duc, P. A. 2014, MNRAS, 442, L33 [NASA ADS] [CrossRef] [Google Scholar]
  121. Rieke, M. J., Baum, S. A., Beichman, C. A., et al. 2003, SPIE Conf. Ser., 4850, 478 [NASA ADS] [Google Scholar]
  122. Rieke, M. J., Kelly, D., & Horner, S. 2005, SPIE Conf. Ser., 5904, 1 [NASA ADS] [Google Scholar]
  123. Rieke, M. J., Kelly, D. M., Misselt, K., et al. 2023a, PASP, 135, 028001 [CrossRef] [Google Scholar]
  124. Rieke, M. J., Robertson, B., Tacchella, S., et al. 2023b, ApJS, 269, 16 [NASA ADS] [CrossRef] [Google Scholar]
  125. Ryon, J. E. 2023, ACS Instrument Handbook for Cycle 31 v. 22.0 (Baltimore: STScI) [Google Scholar]
  126. Schreiber, C., Pannella, M., Elbaz, D., et al. 2015, A&A, 575, A74 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  127. Seillé, L. M., Buat, V., Haddad, W., et al. 2022, A&A, 665, A137 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  128. Shen, S., Mo, H. J., White, S. D. M., et al. 2003, MNRAS, 343, 978 [NASA ADS] [CrossRef] [Google Scholar]
  129. Smail, I., Dudzevičiūtė, U., Gurwell, M., et al. 2023, ApJ, 958, 36 [NASA ADS] [CrossRef] [Google Scholar]
  130. Stefanon, M., Yan, H., Mobasher, B., et al. 2017, ApJS, 229, 32 [NASA ADS] [CrossRef] [Google Scholar]
  131. Sun, F., Helton, J. M., Egami, E., et al. 2024, ApJ, 961, 69 [NASA ADS] [CrossRef] [Google Scholar]
  132. Tacchella, S., Dekel, A., Carollo, C. M., et al. 2016, MNRAS, 458, 242 [NASA ADS] [CrossRef] [Google Scholar]
  133. Trujillo, I., Förster Schreiber, N. M., Rudnick, G., et al. 2006, ApJ, 650, 18 [Google Scholar]
  134. Valentino, F., Tanaka, M., Davidzon, I., et al. 2020, ApJ, 889, 93 [Google Scholar]
  135. Valentino, F., Brammer, G., Gould, K. M. L., et al. 2023, ApJ, 947, 20 [NASA ADS] [CrossRef] [Google Scholar]
  136. van der Wel, A., Franx, M., van Dokkum, P. G., et al. 2014, ApJ, 788, 28 [Google Scholar]
  137. van Dokkum, P. G., Nelson, E. J., Franx, M., et al. 2015, ApJ, 813, 23 [NASA ADS] [CrossRef] [Google Scholar]
  138. Wang, T., Elbaz, D., Schreiber, C., et al. 2016, ApJ, 816, 84 [Google Scholar]
  139. Wang, T., Schreiber, C., Elbaz, D., et al. 2019, Nature, 572, 211 [Google Scholar]
  140. Wang, W., Yuan, Z., Yu, H., & Mao, J. 2024, A&A, 683, A174 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  141. Ward, E., de la Vega, A., Mobasher, B., et al. 2024, ApJ, 962, 176 [NASA ADS] [CrossRef] [Google Scholar]
  142. Weaver, J. R., Davidzon, I., Toft, S., et al. 2023, A&A, 677, A184 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  143. Weibel, A., Oesch, P. A., Barrufet, L., et al. 2024, MNRAS, 533, 1808 [NASA ADS] [CrossRef] [Google Scholar]
  144. Weinberger, R., Springel, V., Pakmor, R., et al. 2018, MNRAS, 479, 4056 [Google Scholar]
  145. Whitaker, K. E., Labbé, I., van Dokkum, P. G., et al. 2011, ApJ, 735, 86 [Google Scholar]
  146. Williams, R. J., Quadri, R. F., Franx, M., van Dokkum, P., & Labbé, I. 2009, ApJ, 691, 1879 [NASA ADS] [CrossRef] [Google Scholar]
  147. Williams, C. C., Labbe, I., Spilker, J., et al. 2019, ApJ, 884, 154 [Google Scholar]
  148. Wuyts, S., Förster Schreiber, N. M., van der Wel, A., et al. 2011, ApJ, 742, 96 [NASA ADS] [CrossRef] [Google Scholar]
  149. Xiao, M. Y., Elbaz, D., Gómez-Guijarro, C., et al. 2023, A&A, 672, A18 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  150. Xiao, M., Oesch, P. A., Elbaz, D., et al. 2024, Nature, 635, 311 [NASA ADS] [CrossRef] [Google Scholar]
  151. Yamaguchi, Y., Kohno, K., Hatsukade, B., et al. 2019, ApJ, 878, 73 [Google Scholar]
  152. Yuan, F.-T., Burgarella, D., Corre, D., et al. 2019, A&A, 631, A123 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  153. Yun, M. S., Reddy, N. A., & Condon, J. J. 2001, ApJ, 554, 803 [Google Scholar]
  154. Zolotov, A., Dekel, A., Mandelker, N., et al. 2015, MNRAS, 450, 2327 [Google Scholar]

All Tables

Table 1.

Cigale model parameters.

In the text
Table 2.

Simulation input parameters.

In the text

All Figures

thumbnail Fig. 1.

Measurement comparisons of SFR obtained by SED-fitting with the CF00 and C00 attenuation law for the parent sample of 1755 galaxies at 3 < z < 4 taken from the Gómez-Guijarro et al. (2023) photometric catalog. The inset shows the comparative histograms of the SFR ratio between the UV/NIR SFR obtained via an SED fitting with the CF00 (dark blue) and C00 (light blue) attenuation law and the SFRtot = SFRUV + SFRIR. The UV/NIR data come from JADES and SFRtot from the GOODS-ALMA 2.0 catalog. For each distribution, the median value (filled circle) along with its associated 16% and 84% percentiles (segments) are displayed. The red dotted vertical line indicates the benchmark whereby 100% of the total SFR is retrieved by the UV/NIR SED fit.
In the text
thumbnail Fig. 2.

Determination of the F444W limiting magnitude. The dependence of the 90% limiting magnitude on circularized effective radii from 0 . ″ 06 $ 0{{\overset{\prime\prime}{.}}}06 $ to 1 . ″ 2 $ 1{{\overset{\prime\prime}{.}}}2 $ is shown for an exponential profile (n = 1, blue solid line) and a de Vaucouleurs profile (n = 4, red solid line), both with null ellipticity (e = 0). In the top right-hand corner, we show the completeness curves for a Sérsic profile n = 1 with size R e = 0 . ″ 42 $ R_{\mathrm{e}}=0{{\overset{\prime\prime}{.}}}42 $ and variable ellipticity and the projection of the 90% completeness on the e = 0 curve that is the most restrictive one. The grey shaded region depicts the FWHM size of the F444W-band. The green solid lines defines the F444W = 25 limiting magnitude as the projection on the most restrictive, while realistic, galaxy size ( R e = 0 . ″ 42 $ R_{\mathrm{e}}=0{{\overset{\prime\prime}{.}}}42 $) and ellipticity (e = 0) at z = 3.5.
In the text
thumbnail Fig. 3.

Lower panel: Distribution in stellar mass and luminosity at rest-frame wavelength 4.4 μm/(1 + z) with 3 ≤ z ≤ 4 for the final sample of 1202 galaxies. The best-fit relation with its 1σ dispersion are displayed in solid and dashed blue lines, respectively. The position of the F444W = 25 limiting magnitudes for z = 3 and z = 4 are shown with two golden lines. Upper panel: Evolution of completeness with stellar mass at the z = 4 luminosity limit. The horizontal golden line showing the 90% completeness level. Consequently, the green vertical line defines the estimated 90% mass completeness limit. Taking into account galaxies with low mass-to-light ratios, our sample is 99% complete for this population to the right of the red dotted line.
In the text
thumbnail Fig. 4.

Distribution of the parent and final mass-complete samples in terms of stellar mass and redshift over the studied redshift range. The parent sample is represented as a light green-scale density map. The horizontal red dashed line indicates the limiting mass of log(M*/M⊙) > 9.6 where the parent sample is 90% complete and contains 237 galaxies. Sources encircled in red define the final mass-complete sample of 188 galaxies used in this paper. This sample is clean from objects with uncertain redshifts, potential AGN contaminants or bad SED fits. The redshift and stellar mass distributions of the two samples are shown as marginal histograms of their respective colors.
In the text
thumbnail Fig. 5.

Location of BlueSFGs, RedSFGs and QGs at 3 < z < 4 in the UVJ color-color diagram, the purple line delimits the quiescent region as defined in Whitaker et al. (2011).
In the text
thumbnail Fig. 6.

F150W − F444W versus F444W color-magnitude distribution for our 3 < z < 4 sample. The markers show the color-selected populations of BlueSFGs, RedSFGs and QGs located above the black solid line defining our RedSFGs selection criteria. As a reference, we indicate the selection criteria for H-dropouts in Wang et al. (2019) (i.e., F150W > 27 mag and F444W < 24 mag) as a green dashed triangular region and the OFG selection from Xiao et al. (2023) (i.e., F150W > 26.5 mag and F444W < 25 mag) as a red dotted line.
In the text
thumbnail Fig. 7.

Location of BlueSFGs, RedSFGs, and QGs at 3 < z < 4 in the SFR–M* plane. The Schreiber et al. (2015) MS is displayed as a solid blue line. We represent the 1σ scatter of this MS associated with 0.5 < ΔMS < 2 (∼0.3 dex) as a shaded blue area as well as a wider scatter of 0.33 < ΔMS < 3 (∼0.5 dex) in lighter blue. The ΔMS < 0.6 dex criterion used in combination with the UVJ diagram to define QGs. We rescaled all SFR values to the median redshift of our mass-complete sample (zmedian = 3.51) as explained in the main text. The sample is mass-complete for the three populations above the thick black line.
In the text
thumbnail Fig. 8.

Dust attenuation as a function of stellar mass for our final 3 < z < 4 sample. BlueSFGs, RedSFGs and QGs are represented with blue dots, red diamonds and yellow dots with black circle, respectively. Their regression lines and ∼95% uncertainties are shown with a solid, dashed and doted line with a shaded area. We indicate the 99% mass-completeness limit of the sample with a thick black line.
In the text
thumbnail Fig. 9.

Mass distribution and relative fractions of the 142 BlueSFGs (blue solid line), 32 RedSFGs (red dashed line), and 14 QGs (yellow dotted line) as a function of stellar mass. The cumulative contribution of RedSFGs and QGs to the total number count as a function of stellar mass is displayed as a black dash-dotted line. The black solid lines show the stellar mass above which > 50% of galaxies are either RedSFGs or QGs. The arrow indicates the mass completeness limit for the RedSFGs population.
In the text
thumbnail Fig. 10.

Illustration of the methodology used to compute radial profiles. From top to bottom, we show an example of BlueSFG, RedSFG, and QG RGB cutouts generated using the F200W, F277W, F356W-bands PSF-matched to the F444W filter. Each cutout has a 2 . ″ × 2 . ″ $ 2{{\overset{\prime\prime}{.}}}\times 2{{\overset{\prime\prime}{.}}} $ angular size and contains a 1 . ″ $ 1{{\overset{\prime\prime}{.}}} $ white bar that defines the scale of the image. The 0.16 . ″ $ 0.16{{\overset{\prime\prime}{.}}} $ angular resolution of the F444W-band is represented by a pink circle. Its radius corresponds to the semi-major axis of annulus 1 as well as the width of each annulus. For each galaxy, five concentric rings with a constant width equal to the FWHM of the F444W-band are displayed. In addition, for each cutout, we present the best model spectra produced by Cigale for annulus 1 (the innermost) and annulus 3. In those, the blue curve is the unattenuated stellar emission.
In the text
thumbnail Fig. 11.

From left to right and top to bottom: Stellar mass density (Σ⋆), star formation rate density (ΣSFR), dust attenuation ( A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $), sSFR, and mass-weighted age normalized median radial profiles inferred using Cigale for BlueSFGs (in blue), RedSFGs (in red), and QGs (in yellow). Error bars represent the 68th and 32th quantiles of the distributions in each radial bin. In the Σ⋆ panel we indicate the (normalized) half-mass effective radius for each population with a vertical dashed line.
In the text
thumbnail Fig. 12.

From left to right and top to bottom: Σ⋆, ΣSFR, A v ISM $ A_{\mathrm{v}}^{\mathrm{ISM}} $, sSFR, and mass-weighted age concentrations as a function of stellar mass. The different concentrations are determined by measuring the slopes of the three innermost annuli of each curve individually. In each panel, BlueSFGs, RedSFGs, and QG are marked by blue, red and yellow dots, respectively. The medians in stellar mass bins are displayed for each population as diamonds, color-coded by the median dust attenuation in the bin. Uncertainties are represented by shaded regions of the color of the represented population and represent the 68th and 32th quantiles of the distributions in each stellar mass bin. The medians are centered on the median stellar mass of each bin containing at least four galaxies (with the exception of the most massive QG bin, which contains only one galaxy).
In the text
thumbnail Fig. 13.

Median radial profiles of Σ⋆ for BlueSFGs (in solid blue line) and RedSFGs (in solid red line) with their associated uncertainties represented as blue and shaded regions, respectively. The blue dotted curve represent the radial evolution of Σ⋆ for BlueSFGs over the next 300 Myr, derived by applying their median radial profile of star formation surface density (ΣSFR), assumed constant during this period. Additionally, the evolution of the median Σ⋆ radial profile of BlueSFGs over the next 300 Myr considering the ΣSFR profile of RedSFGs is displayed as a dashed blue curve. Each profile is normalised to the Σ⋆ at 4 kpc of RedSFGs. We display the typical median uncertainty of the dotted and dashed profiles by a blue error bar.
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.