WorksPhysics and Astronomy
Euclid: Quick Data Release (Q1) – The evolution of the radio-loud AGN fraction-stellar mass relation to in the Euclid Deep Field North
Abstract
The incidence of radio-loud active galactic nuclei (RLAGN) and their energetic output via AGN feedback are crucial in understanding the evolution of galaxies and the large-scale structure of the Universe. In the local Universe, the incidence of RLAGN has been shown to correlate strongly with host galaxy properties, particularly stellar mass. Using the large-area survey capabilities provided by Euclid and the LOw Frequency ARray (LOFAR), we present the first study of the RLAGN incidence–stellar mass relation to . We perform a stringent analysis accounting for cross-matching reliability and stellar-mass incompleteness. In addition, we test the robustness of our results against AGN and host galaxy classification. This yields a sample of RLAGN in the Euclid Deep Field North, comprising low-excitation radio galaxies (LERGs). We find that the LERG incidence as a function of stellar mass and radio luminosity follows a tight correlation over more than two orders of magnitude in stellar mass to . The slope of the power-law relation decreases from in the local Universe to at , indicating a relative increase in low-mass LERGs at earlier epochs. Separating the LERG population into actively star-forming and quiescent galaxies reveals that quiescent LERGs show a much steeper relation between incidence and stellar mass, with compared to star-forming LERGs that show . However, the relation seems to be independent of whether galaxies are star-forming or quiescent at , where both populations show relatively flat relations with . Our results suggest an evolutionary picture in which the LERG population transitions from cold-mode, merger-dominated accretion at earlier epochs to a scenario in which systems require a massive hot halo to fuel the AGN in the local Universe.
Introduction
It is now well accepted that most massive galaxies host a supermassive black hole (SMBH) at their centre ([78]; [123]), and that there is a tight correlation between the mass of the SMBH and the bulge of galaxies in the local Universe. For example, correlates with the bulge stellar velocity dispersion ([50]; [53]; [121]), and similarly, bulge mass or bulge luminosity shows a tight correlation with . (see [77] for a review). These correlations between SMBHs and their host galaxies suggest co-evolution, often attributed to phases of active galactic nucleus (AGN) activity.
During the AGN phase, the SMBH actively accretes material from its host galaxy and can release vast amounts of energy through radiation and matter outflows. Some AGN additionally launch relativistic, magnetically driven jets of hot plasma, which can deposit substantial amounts of kinetic energy into the surrounding medium ([48]). AGN activity can couple its released energy to the ambient medium and regulate subsequent galaxy growth. This process, commonly referred to as AGN feedback, may suppress star formation in the host galaxy or cooling flows in galaxy groups and clusters, resulting in negative feedback. In some circumstances, however, AGN activity may instead enhance star formation, leading to positive feedback. These processes are therefore thought to play an important role in regulating galaxy evolution ([48]; [111]; [62]).
Some AGN launch relativistic jets that interact with the ambient medium, producing non-thermal synchrotron emission as relativistic charged particles move through magnetic fields. This radio emission is unaffected by dust obscuration, making radio observations a particularly powerful and unique probe of jetted AGN, their environmental interactions, and their role in galaxy evolution. Wide-field radio observations such as the LOw Frequency ARray (LOFAR; [122]) Two Meter Sky Survey (LoTSS; [105]), and the Karl G. Jansky Very Large Array Sky Survey (VLASS; [93]; [80]) enables the radio-source population to be characterised across unprecedented sky areas. The increased depth in deep radio surveys such as the VLA-COSMOS ([110]), the MeerKAT International GHz Tiered Extragalactic Exploration (MIGHTEE; [70]; [58]), and the LOFAR Deep Fields ([119]; [98]; [15]) probe much fainter radio populations, enabling studies of how radio-loud AGN (RLAGN) evolve with redshift ([110]; [130]; [74]), while also probing star-forming galaxies (SFGs) and radio-quiet AGN ([90]; [20]).
Observations of RLAGN in the local Universe () have shown that they broadly separate into two accretion modes associated with different Eddington-ratio regimes: radiatively efficient (‘radiative mode’) and radiatively inefficient (‘jet mode’) AGN (e.g., [63]; [118]; [59]). The latter are thought to host a radiatively inefficient advection-dominated accretion flow (ADAF; [89]), with most of their energy released through twosided collimated jets, whereas radiative-mode AGN emit predominantly electromagnetic radiation from a geometrically thin, optically thick accretion disc (Shakura & Sunyaev [101]). Radiative mode AGN are also observed to drive outflows, but the coupling of these outflows to the surrounding medium is poorly understood (Harrison & Ramos Almeida [62]).
Observationally, the jet and radiative mode populations are commonly classified as low- and high-excitation radio galaxies (LERGs and HERGs, respectively) based on their optical emission-line properties (Heckman & Best [63]). LERGs constitute the majority of the radio-loud AGN population and are typically hosted by massive, redder elliptical galaxies in dense environments, whereas HERGs are less common and preferentially reside in bluer, less massive galaxies with more recent star formation. Note that this classification is not exclusive, such that some HERGs are found in red, elliptical galaxies (e.g., [69, 23, 19]).
The general interpretation is that HERGs tend to accrete efficiently from cold gas by either external (e.g., galaxy mergers) or internal (e.g., bars or instabilities) processes, where a plentiful gas supply can lead to the formation of a standard accretion disc (e.g., [2, 60, 63]). For LERGs, the cooling of hot gas from the halo surrounding massive elliptical galaxies will lead to inefficient accretion (e.g., [8, 63]). LERGs are often connected to negative AGN feedback, which keeps the star formation quenched and disrupts the circum-galactic medium when the jets expand from kpc up to Mpc scales ([52, 91] (Li et al. 2015)).
To understand the significance of radio-AGN feedback in galaxy evolution and AGN activity timescales, we need to study the redshift evolution of the RLAGN population. In the local Universe, one of the strongest observed relations is the dependence of the fraction of galaxies hosting an RLAGN, , on the host galaxy stellar mass, . This relation has been studied extensively and is found to follow a power law for LERGs and for HERGs ([9, 69, 97]). Interestingly, Sabater et al. [97] found that 100% of the most massive galaxies () show radio-AGN activity. In other words, the nuclei of the most massive galaxies are always switched on.
Other works extended this relation to –, using deep VLA and LOFAR surveys, and found a flattening of the power-law relation with redshift ([132, 126, 74]). In addition some works have found that the LERG incidence in SFGs at higher redshifts is more comparable to that of HERGs in the local Universe with . Alternatively, quiescent LERGs show , which can explain the flattening of the RLAGN incidence with redshift ([73, 74]). These results indicate that the standard paradigm of classifying HERGs and LERGs based on their accretion rate may not always hold, suggesting a connection between LERGs and the accretion of cold gas ([74, 83]).
However, reliable classification of HERGs and LERGs suffers from selection bias. Arnaudova et al. [3] compared spectroscopic classifications to state-of-the-art photometric classifications from spectral energy distribution (SED) fitting, and found that these show reasonable overall agreement. Nevertheless, they found significant differences, including identifying 2–5 times more radio-quiet AGN using spectroscopic data. This may impact the separation between HERGs and LERGs, causing their accretion properties to appear less distinct ([130]). Recent results on the X-ray properties of HERGs and LERGs have, however, shown that LERGs tend to have much lower average X-ray luminosities (a proxy for the accretion rate) compared to the HERGs ([76]). Because studies in the local Universe have found a significantly shallower slope for HERGs ([69]), it is crucial to exclude HERGs contamination before claiming a flattening of the – relation for LERGs.
While large radio surveys offer tremendous scientific potential, understanding the physical processes driving radio-source activity requires detailed characterisation of their host galaxies through multi-wavelength observations. Combining wide-field radio surveys with new facilities such as Euclid (Euclid Collaboration: Mellier et al. [41]) will revolutionise our ability to survey wide areas with high sensitivity. These facilities will deliver increased source statistics and detailed host-galaxy measurements to unprecedented depths. Such data enable large statistical studies of radio-source populations and their connection to host-galaxy properties, environments, and redshift.
In this work, we investigate how the incidence of LERGs evolves with , , and radio luminosity. We focus on the Euclid Deep Field North (EDF-N) by combining deep LOFAR radio data (Bondi et al. [15]) with the new optical/near-infrared (NIR) Euclid data (Euclid Collaboration: Aussel et al. 2026). We obtain the sensitivity and host-galaxy characterisation required to probe to and mass-completeness down to . A key advantage of these data is the substantial improvement in completeness provided by Euclid compared to previous analyses of other LOFAR deep fields. This enhanced depth is crucial for investigating the LERG incidence at lower and higher . Consequently, we can extend the – relation beyond the epoch of cosmic noon, at which both the star-formation rate (SFR) density and the SMBH growth rate in the Universe peaked ([102, 82]).
The paper is structured as follows. Section 2 describes the LOFAR and Euclid data in the EDF-N. § 3 presents the methods used to cross-match the LOFAR and Euclid source catalogues, the completeness for our sample, and how to mitigate potential contamination. This is followed by the selection of RLAGN, the differentiation between LERGs and HERGs, and the method to calculate LERG fractions in Sect. 4. In Sect. 5, we present the LERG incidence as a function of and for different radio luminosity limits, and discuss the reliability and interpretation of our results. Finally, Sect. 6 presents our conclusions. In this work, we use a flat CDM cosmology with km s Mpc and , and assume a radio spectral index of , with .
Data
We first describe the LOFAR radio data in the EDF-N. We then present the Euclid data and catalogues, followed by the supplementary data used to obtain physical parameters and spectroscopic redshifts where available.
LOFAR data in the EDF-N
The EDF-N is the fourth LOFAR Deep Field and has a sensitivity similar to that of the other three fields – ELAIS-N1, Lockman Hole, and Boötes – presented in LoTSS Deep Fields DR1 ([119, 98, 10]). The EDF-N was observed with the High Band Antenna (HBA) at 144 MHz for 400 hours. When fully processed at 6″ resolution, the EDF-N will have a central root mean square (RMS) noise of 10 Jy beam. An initial image and a corresponding radio source catalogue in the EDF-N were obtained by Bondi et al. [15] from 72 h of observations. This image has an angular resolution of 6′′ and a central RMS noise of 32 Jy beam.
Bondi et al. [15] extracted the radio source catalogue using the Python Blob Detection and Source Finder (PyBDSF, Mohan & Rafferty [88]), which is a tool to decompose radio interferometry images into a set of islands above a given signal-to-noise ratio (S/N). These islands are fitted using Gaussians, shapelets, or wavelets, giving not only information about the measured fluxes but also about the morphology of radio sources. The resulting catalogue contains sources with a peak flux satisfying S/N 5 in a region of approximately 10 deg centred at RA = 17 h 58 m 55.s2, Dec = 66° 1′ 12.′′0.
In this paper, we make use of the updated radio source positions in the catalogue for the EDF-N as presented by Bisigello et al. [11], hereafter B252. To enhance their cross-matching procedure, B25 improved the positional accuracy of radio sources by using the high-resolution image of the EDF-N made by the International LOFAR Telescope. This image has been produced following the methods described in Sweijen et al. [116] and De Jong et al. [25] using 32 h of the 72 h available. The final high-resolution images cover the inner 6.25 deg region of the EDF-N reaching a central sensitivity of 40 Jy beam and an angular resolution of 1.′′5. The final updated source catalogue comprises 23 309 radio sources B25 and will be used to cross-match with the Euclid data.
Euclid data
Complementing the radio data, we use the first Quick Release of Euclid data (Q1; Euclid Quick Release Q1 [47]). This release comprises the Euclid Deep Fields (EDF-North, EDF-South, and EDF-Fornax) covering a total area of 63.1 deg. All EDFs have been observed in the four Euclid photometric bands, that is, one optical () using the VIS instrument and three NIR bands (, , and ) using the Near Infrared Spectro- and Photometer (NISP; Euclid Collaboration: Jahnke et al. [35]). Euclid provides high angular resolution images with a mean full width at half maximum (FWHM) of 0.′′158 ± 0.′′001 captured by the VIS instrument (Euclid Collaboration: Cropper et al. [32]) for all galaxies down to = 25.5 (Euclid Collaboration: McCracken et al. [40]). This allows for detailed characterisation of galaxy morphologies to large distances (e.g., Euclid Collaboration: La Marca et al. [36]; Euclid Collaboration: Quilley et al. [42]; Euclid Collaboration: Walmsley et al. [46]).
The observations from the VIS and NISP instruments are complemented by ground-based photometry (EXT data) to form the multi-wavelength MER catalogue (Euclid Collaboration: Romelli et al. [43]). Sources are detected using SourceXtractor++ (Bertin et al. [6]) independently for the VIS and an NIR-stack detection mosaic. Thus, the MER catalogue includes only sources detected by Euclid. The VIS_DET column indicates VIS detection: VIS_DET=1; otherwise, the source is detected only in the stacked NIR image (VIS_DET=0).
For this work, we focus on the data in the EDF-N, since this overlaps with the LOFAR deep field. The EDF-N observed by Euclid is centred at RA = 17 h 58 m 55.s9, Dec = 66° 1′ 3.′′7 and covers an area of 22.9 deg, which is over two times larger than the LOFAR pointing in the EDF-N. To obtain Euclid counterparts for the LOFAR source catalogue described in, we retrieved
8 868 897 Euclid sources in the EDF-N from the MER catalogue satisfying the following criteria3:
SPURIOUS_FLAG = 0, to filter out spurious sources;
PHZ_CLASSIFICATION = 1, to exclude objects classified as a star by the PHZ processing function based on their photometry (Euclid Collaboration: Tucci et al. [45]);
DET_QUALITY_FLAG = 0 | 2 | 512, to select sources with good quality photometry, blended sources that have been successfully deblended, or sources within the area of an extended object (Euclid Collaboration: Romelli et al. [43]);
S/N 3.0 for FLUX_DETECTION_TOTAL, to ensure a 3 detection in (VIS_DET = 1) or in the stacked NIR-stacked image (VIS_DET = 0).
Accurately investigating the redshift evolution of RLAGN requires reliable redshift estimates. In addition to the spectroscopic redshifts provided by Euclid (Euclid Collaboration: Le Brun et al. [38]), Q1 benefits from extensive overlap with existing spectroscopic surveys. To maximise the spectroscopic completeness of our sample, we use the catalogue described by Euclid Collaboration: Tucci et al. [45], which matches Euclid MER sources to approximately 40 external spectroscopic surveys. Specifically, we employ an updated version of this catalogue (Euclid Collaboration: Sauder et al. in prep.), comprising 47 466 highly reliable spectroscopic redshifts in the EDF-N with 0.0 5.8, with the majority originating from the Dark Energy Spectroscopic Instrument (DESI) DR1 survey (DESI Collaboration: Abdul Karim et al. [28]).
For host-galaxy photometric redshifts and physical properties, we adopt the parameters derived by Euclid Collaboration: Enia et al. [33], hereafter E26. They used the Euclid pipeline’s nearest-neighbour (NN) algorithm, adding two further bands from Infrared Array Camera (IRAC, Fazio et al. [49]) onboard Spitzer (Werner et al. [127]) and applying more constrained priors to improve the quality of their SED fitting results. For sources with spectroscopic redshifts from DESI, E26 compared their estimates with those in the DESI value-added catalogue (Siudek et al. [107]). They found that the photometric redshifts achieve a normalised median absolute deviation (NMAD) of up to , while and SFR estimates have an NMAD of and dex, respectively.
The reliability of photometric redshifts and physical parameters derived from SED fitting can be significantly affected by the presence of an AGN. In such systems, an a priori unknown fraction of the observed flux originates from the AGN rather than the stellar and interstellar components of the host galaxy, potentially biasing estimates of galaxy properties. To mitigate these effects, Euclid Collaboration: Laloux et al. [37], hereafter L26, derived AGN-specific physical parameters.
First, L26 identified AGN candidates using the range of selection techniques developed for Q1 (Euclid Collaboration: Matamoros Zatarain et al. [39]). The next step was to obtain reliable redshifts for all AGN candidates. Whenever spectroscopic measurements were unavailable, improved redshift estimates were obtained through visual inspection of Euclid NISP slitless spectroscopy (Euclid Collaboration: Fu et al. [34]), machine-learning photometric redshifts for point-like X-ray selected sources (Euclid Collaboration: Roster et al. [44]), and template-fitting photometric redshifts derived with the Bayesian code Prosphoros. For the latter, the luminosity prior and template library were modified to improve redshift estimates for AGN and reduce the rate of catastrophic outliers. Lastly, the resulting physical parameters were obtained through a dedicated SED-fitting campaign using CIGALE (Code Investigating GALaxy Emission; Boquien et al. [16]; Yang et al. [133]). In particular, L26 employed the SKIRTOR AGN module (Stalevski et al. [113, 114]), which models a two-phase dusty torus consisting of dense clumps embedded within a lower-density smooth component. In this work, we adopt the physical parameters from this catalogue for AGN candidates, while using the measurements from E26 for all other sources, as described in Sect. 4.
Cross-matching LOFAR and Euclid source catalogues
In this section, we discuss the likelihood ratio (LR) method that was used to identify Euclid counterparts for the LOFAR sources in the EDF-N. The following subsections then discuss the completeness limit for Euclid galaxies and an estimation of the contamination in our cross-matched sample as a function of .
Likelihood ratio identifications
As will be discussed in more detail in Sect. 3.4, simply cross-matching the LOFAR and Euclid source catalogues using an NN approach will lead to significant contamination of the cross-matched sample. This effect becomes more pronounced towards smaller because the galaxy number density increases for lower masses. Since the RLAGN incidence can reach low values of for (see e.g., Janssen et al. [69]; Williams & Röttgering [132]; Sabater et al. [97]), it is crucial to both mitigate and account for this contamination.
As mentioned in Sect. 2.1, B25 used the LR method, incorporating colour and magnitude, to identify optical/NIR counterparts to the LOFAR EDF-N catalogue. The LR method is a statistical method to identify counterparts in images with widely differing frequency ranges (De Ruiter et al. [26]; Sutherland & Saunders [115]). The LR is the ratio of the probabilities that a galaxy is the true counterpart or an unrelated background object, given its properties. As the Euclid Q1 data were unavailable, B25 could not use them for cross-matching. Here, we apply the LR method for the Euclid data to explore its capabilities in future data releases. We follow the method in appendix B of Best et al. [7], which makes use of the coordinates and the positional uncertainties of both radio and optical sources (for more information, see Richter [96]; De Ruiter et al. [26]).
We avoid using galaxy colour information, as in B25, to avoid introducing priors on galaxy colour or magnitude that could favour particular host-galaxy populations. Although previous studies have shown that LOFAR sources tend to have redder counterparts than the general galaxy population in the local Universe (Williams et al. [131]), recent studies of the LOFAR Deep Fields have also identified radio sources associated with blue galaxies (Best et al. [10]; Igo et al. [65]). Including colour or magnitude information can improve counterpart identification, but may also favour populations represented by the adopted priors. This is particularly relevant given the correlation between galaxy colour and stellar mass (Baldry et al. [5]; Grützbauch et al. [55]), which could bias analyses of host-galaxy properties. A positional-only LR approach avoids these additional photometric priors, and we test its validity in Sect. 3.2.
The LR method uses an error-weighted distance between an optical and a radio source defined as
where and are the differences in right ascension and declination between the radio and optical coordinates, and and are the associated errors. These errors are the quadratic sum of the positional errors for the radio source, the optical source and the astrometric uncertainty or offset between both surveys, . The lower radio resolution makes the radio positional uncertainty dominant. Each radio source has unique positional uncertainties calculated by PyBDSF, which are reported in the radio source catalogue. For the radio source coordinates, we adopt the corrected coordinates obtained by B25 after an astrometric correction. For the Euclid coordinates, we estimate the positional uncertainty as FWHM / S/N = 0\aas@@fstack{\prime\prime}18 / S/N as in Ivison et al. [68]. We conservatively adopt the 99th percentile, yielding \sigma_{\alpha,\mathrm{opt}} = \sigma_{\delta,\mathrm{opt}} = 0\aas@@fstack{\prime\prime}06. Lastly, for the astrometric uncertainty between the two surveys, we use \sigma_{\alpha,\mathrm{ast}} = \sigma_{\delta,\mathrm{ast}} = 0\aas@@fstack{\prime\prime}47, which is the median astrometric radio-optical offset for the EDF-N found by B25. These values account for any remaining astrometric offset for individual sources and were also used by B25, who found them representative of astrometric variations across the EDF-N, facilitating comparison with our results.
Assuming a random distribution of radio sources across the EDF-N, the probability of finding the nearest unrelated object at a separation between and follows a Poissonian background distribution determined by the source surface density . This assumption implicitly neglects spatial variations in the source density and angular clustering, which may be significant in overdense environments such as galaxy clusters. A uniform background density may therefore underestimate chance associations in these regions. Nevertheless, the Poissonian approximation provides a useful baseline for estimating chance associations. In contrast, the distribution of true counterparts is expected to follow a Rayleigh profile set by the positional uncertainties. The LR for each radio-optical association is defined as the ratio of the probability of a true counterpart to that of a random association at a distance between and . For verification purposes, we compared the observed distribution of radial offsets between Euclid and LOFAR sources and find that it is well described by a Rayleigh distribution at separations below , while larger separations are increasingly dominated by chance associations.
Before computing the LR, we account for bright stars that can produce diffraction spikes in optical and NIR images, leading to unreliable photometry. For this purpose, we use the star mask constructed by B25 for the radio source catalogue in the EDF-N. The masking procedure follows the approach outlined in section 3.4.3 of Kondapally et al. [75], which used the Gaia DR2 catalogue to define magnitude-dependent masking radii for individual stars. This results in circular masks with radii ranging from 178″ for stars with to 2″ for stars with , yielding a total masked area of 1.856 deg in the LOFAR EDF-N. After applying the mask, the LOFAR EDF-N field covers an area of 8.73 deg with 19 550 LOFAR sources.
The LR was calculated for every single-component LOFAR source (S_code = S) with a major axis smaller than 15″ following Kondapally et al. [75] and B25. We excluded sources for which PyBDSF was not able to extract an accurate size estimate. This occurs when radio sources are blended, or when visual inspection shows that two lobes classified as separate sources by PyBDSF belong to the same source (Bondi et al. [15]). These selections retain 17 971 out of 19 550 (91.9%) radio sources suitable for LR identification. The matching of the excluded sources will be discussed later in this section. Regarding the Euclid sources, we computed the LR for all galaxies satisfying (see Eq. (1)). Assuming that sources are randomly distributed across the field, means that there is a 3.3% chance that the real optical/NIR counterpart is further away.
To find an appropriate threshold value to select reliable matches with LR , we estimate the completeness , the fraction of real identifications that are accepted, and the reliability , the fraction of accepted identifications that are correct (De Ruiter et al. [26]; Best et al. [7])
Here, and are the probabilities that source is an unrelated background object ‘b’ or a real counterpart ‘id’, respectively, and is the total number of real identifications obtained by summing for all sources. For an object , we need to calculate these posterior probabilities using
where is the a priori probability of finding an optical counterpart for a radio source. To estimate , we calculated the distance between all single-component LOFAR sources and their nearest Euclid galaxy. As discussed above, this showed that the distribution is dominated by a Rayleigh distribution at short distances and the majority (15 960 sources) of the matches are found within a 1.5 arcsec radius. We use this number to estimate the a priori probability to be .
To study the RLAGN incidence down to , where the RLAGN fraction is relatively small, we require stringent cross-matching with minimal contamination. We found that the contamination varies little with . This is caused by the stringent requirement of . We adopt the threshold value for which , resulting in with both the completeness and the reliability being 96.7%. Using this threshold value, we find 13 661 matches with the LR method.
Testing for biases in LR identifcations
In an LR framework that does include magnitude priors and colour information, counterparts at very small separations naturally obtain high LR values, such that the colour term contributes little to the final association probability. Colour information mainly becomes important when multiple candidate counterparts are present at comparable separations, where it preferentially selects sources with colours commonly associated with radio counterparts (e.g., Williams et al. [131]; Kondapally et al. [75]). To evaluate the impact of neglecting colour information, we selected all radio sources with two optical candidate counterparts satisfying , yielding 1 241 objects (6.8% of the sources suitable for the LR analysis). Of these, 896 have detections in both and , allowing us to construct the colour . We then divided the sample into two categories: a ‘clear-match’ class, containing sources with one close candidate () and a second candidate at larger separation (), and an ‘ambiguous’ class, where both candidates satisfy . These classes contain 191 (1.1%) and 148 (0.8%) radio sources, respectively. The colour distribution of the counterparts showed only marginal differences between the two classes. More importantly, these ambiguous systems constitute only a very small fraction of the total sample, indicating that neglecting colour priors does not significantly affect the matching results or the rest of our analysis. We also note that only 10 (0.06%) radio sources have two candidate optical counterparts with , such that we do not expect this to affect our classification significantly.
For the 1579 radio sources (8.1% of the full sample) that are non-compact (7.5%), too large (1.2%), or lacking a reliable size estimate (0.7%), we adopt the optical identifications from B25. They visually inspected sources that could not be identified using the LR method (see section 3.5 of B25). We then cross-matched their optical/IR counterpart candidates with the Euclid catalogue using an NN search radius of 0.7 arcsec.4 This radius was motivated by the separation distribution between Euclid sources and the counterparts identified by B25, which peaks below 0.7 arcsec. At larger separations, match counts increased as expected for random associations because of the larger search area. The NN cross-match adds 1 317 sources to our sample with a median separation of 0.51 arcsec, meaning that they are highly reliable. Ultimately, this results in a fully cross-matched sample in the EDF-N of 14 978 sources, yielding 14 978/19 550 = 76.6% of the LOFAR sample with a counterpart.
Using the LR method with positional information alone could yield more unreliable identifications. Namely, colour priors can help find a reliable counterpart when the distance criterion is ambiguous. However, to prevent or avoid the introduction of potential biases, e.g. through the correlation between galaxy colour and . To verify the cross-matching, we visually inspected a random subset of 100 LR- and 100 NN-matched radio sources.5 Approximately 90% of the LR matches were correctly identified, while all NN matches were confirmed, as expected since they had already been visually inspected by B25. Comparing our LR matches to those of B25 yielded an agreement of 92%. In nearly all discrepant cases, two Euclid sources were separated by less than the 6 arcsec resolution of the radio maps, making it difficult to unambiguously identify the correct counterpart.
Final sample with physical parameters
The last step in creating a cross-matched LOFAR-Euclid source catalogue is to obtain the SFR and for the 14 978 galaxies from the catalogue in E26 or in L26 if they are AGN candidates. The classification of sources into AGN and non-AGN is discussed in more detail in Sect. 4. Since E26 applied additional selections to the Q1 data, and not all sources were detected by IRAC, our full matched sample contains 13 502 sources with physical parameter estimates. Subsequently, we require reliable physical parameters with nnpz_flags = 0 for non-AGN or for the SED fits of the AGN candidates, which reduces the total number of sources to 12 032. Lastly, we impose a cut to retrieve only those sources in the redshift range and satisfying SFR to ensure the reliability of the photometric redshifts and derived physical properties (E26).
These selections result in a total of 10 794 sources in our final sample, comprising 6 161 sources with parameters from E26 and 4 633 sources with parameters from L26. 3 653 sources have physical parameters available in both catalogues, enabling a direct comparison of the derived quantities (see Sect. 5.2 and Appendix C). We apply the same selection criteria to the full sample of 3 877 025 Euclid galaxies in the overlapping area between LOFAR and Euclid in the EDF-N. This resulted in a total of 1 945 312 sources, of which 1 838 361 were drawn from E26 and 106 951 from L26. Note that 3 231 (29.9%) sources in our cross-matched sample and 34 665 (1.8%) sources in the galaxy sample have a spectroscopic redshift.
Using the redshift for a source with flux density and applying the standard radio -correction, we calculate the rest-frame radio luminosity at 144 MHz
where is the luminosity distance and is the typical radio spectral index with a dispersion of 0.2–0.3 (e.g., Böhme et al. [14]). Based on early LOFAR HBA observations in the EDF-N, White et al. [129] found this spectral index to be representative of the radio source population.
Estimating the cross-matching contamination fraction
As a result of Euclid’s excellent sensitivity, the EDF-N has an average galaxy number density of , mostly consisting of low-mass objects. Therefore, a simple NN cross-match between the (lower-resolution) LOFAR and Q1 sources results in many low-mass objects being identified as counterparts, thus yielding relatively high misidentification rates. The LR method described in Sect. 3.1 reduces these rates, but we further estimate the contamination fraction of the cross-matched sample as a function of to account for residual effects.
To obtain the contamination fraction, i.e., the chance of matching with a source that is not the genuine counterpart, as a function of , we again perform the LR matching after separating the Euclid sources into bins using the same LR threshold and maximum value for as used in Sect. 3.1. Next, we calculate the mean of and for every bin, which we define as the true fraction and the contamination fraction , respectively. The results of this analysis can be seen in Figure 1 and show that especially for there is a significant contamination fraction rising from about 5% at to at .
Figure 1. True (green) and contamination (red) fractions for the LOFAR–Euclid cross-matching as a function of obtained with the LR method. For each, we show the interpolated curve (dashed line) and the 68% confidence interval (shaded region).
To verify the true and contamination fractions obtained from the LR matching, we also performed a Monte Carlo-based analysis. For all optical sources in a given bin, we calculate the matching fraction for randomly sampled radio coordinates in the EDF-N using a radius to take into account that sources with a higher S/N have a more well-defined position. We then calculate the matched fraction for the real radio coordinates, such that the difference between the real and random matched fractions yields the contamination fraction. Within uncertainties, the results of this analysis are the same as the results seen in Figure 1 obtained from the LR matching.
Stellar mass completeness for Q1
To prevent any bias on the RLAGN incidence as a function of , we need to quantify the minimum above which Q1 is complete, given its limiting magnitude. We use the method described in Pozzetti et al. [94] and select the faintest 20% of the Euclid galaxy sample in the EDF-N limited to , which is the limit for -detected sources. To verify this limit, we estimated the 90% magnitude completeness limit by fitting a powerlaw distribution to the observed number counts for S/N -detected sources. The 90% magnitude completeness limit, defined as the magnitude at which the observed number counts drop to 90% of the fitted power-law distribution, is , in agreement with the detection limit.
Subsequently, we calculated the stellar mass each galaxy would have if its apparent magnitude were equal to the limiting magnitude given its redshift, , using (Pozzetti et al. [94])
To obtain the completeness limit as a function of redshift, , we divided the sample into redshift bins and determined the 90th percentile of in each bin. Figure 2 shows that our results are similar to the 95% completeness limit obtained in E26. Figure 2 also shows the important improvement of about 1 dex in the completeness limit compared to results obtained for ELAIS-N1, which had the deepest optical/IR data of the three fields in LoTSS Deep Fields DR1 ([30]). Interestingly, Figure 2 shows that Euclid probes down to at , which is about 1.5 dex below the characteristic break in the stellar mass function ([86]).
Figure 2. Stellar mass completeness limit as a function of for Euclid galaxies in the EDF-N. The orange line shows the 90% completeness limit using galaxies with . Red crosses show the 95% completeness limit reported in E26. Purple squares and green triangles show the 90% completeness limits obtained for the ELAIS-N1 field using the and the 4.5 μm IRAC band, respectively (Duncan et al. [30]).
Selecting radio AGN
Having cross-matched LOFAR sources with Euclid in the EDF-N and assessed contamination, we need to separate RLAGN from SFGs. Before proceeding, we clarify the terminology adopted throughout this paper. In the literature, radio-AGN are distinguished from SFGs using a variety of methods, including spectroscopy, extended radio emission, IR colours, and the detection of a radio excess (e.g., [29, 61]). In the following, we classify as radio-AGN sources showing radio emission significantly exceeding that expected from the host-galaxy SFR ([108, 10, 103]). We use the terms ‘radio excess’, RLAGN, and radio-AGN interchangeably throughout. The underlying physical idea is that galaxy radio emission at frequencies GHz is dominated by non-thermal synchrotron radiation produced by highly energetic particles ([21]). These particles are accelerated either in supernova remnants of massive stars ([12, 72]) or by AGN jets and outflows.
At 144 MHz, the emission from SFGs is therefore dominated by synchrotron radiation associated with star formation, leading to a tight correlation between SFR and ([125]). Several studies of this correlation have found very similar parameters for a power-law relation (e.g., [56, 108, 10]). There has been some discussion about whether to include a stellar-mass dependence ([56, 108, 27]). This can be explained by the deeper gravitational potential of more massive galaxies allows them to retain their cosmic ray population for longer timescales, thus resulting in more radio emission for a given SFR. Indeed, previous studies have shown that higher-mass galaxies exhibit higher at fixed SFR, and that including reduces the scatter in the empirical relation (e.g., [95, 108, 85]).
Shenoy et al. [103] explored the dependence of the –SFR relation on several physical parameters, and found that cannot be neglected. Hence, we adopt their relation, i.e.
where is the normalisation and and are power law indices for the SFR and dependence, respectively. We adopt the best-fit parameters found by Shenoy et al. [103] for their photometric data, which are , , and .
To identify radio-excess sources, we must account for the intrinsic scatter that this relation exhibits ([108, 24]). Therefore, we require galaxies to lie at least 0.8 dex (about ) above the mass-dependent –SFR relation, consistent with previous studies ([20, 103]). Because the SFR estimates are subject to considerable uncertainty, we adopt a probabilistic approach. For each source, we draw 5 000 samples of , SFR, , and from Gaussian distributions centred on their median values. The standard deviation of each distribution is set to the mean of the upper and lower uncertainties. These uncertainties are taken either directly from the catalogue or are calculated using the differences between the 84th and 50th percentiles and between the 50th and 16th percentiles. For the sources with a spectroscopic redshift, we adopt a small uncertainty in redshift of . We classify a source as having a radio excess if it exceeds the threshold value in at least 90% of the realisations. This results in 2 748 out of 10 794 sources being classified as radio-excess.
In addition to the classification based on a radio excess, we also include all extended radio sources with a major axis size larger than 75 kpc. This value was chosen because an angular diameter distance of 75 kpc spans at least 8\farcs9 on the sky, which is resolvable with the 6\farcsprime\prime resolution. Such extended radio emission is highly likely not powered by star formation, but by synchrotron emission from AGN jets. The full sample contains 710 extended radio sources, of which 300 are already included among the radio excess sources. This means that the full classification yields 3 158 RLAGN out of 10 794 radio sources, of which 1 177 objects have physical parameters from L26 and 1 981 from E26. In Figure 3, we show as a function of the SFR for all matched radio sources in the EDF-N, as well as the –SFR relation. The radio excess criterion of the –SFR relation plus 0.8 dex and the radio sources with an excess are also shown in Figure 3.
Figure 3. Radio luminosity as a function of SFR and for all cross-matched sources: RLAGN with radio excess (orange), extended RLAGN with no radio-excess (red), SFGs (blue). The solid black line shows the mass-dependent –SFR relation from Eq. (6), and the dashed black line shows the radio-excess threshold, using the of our matched sample.
As discussed in Sect. 1, we define as HERGs (LERGs) those RLAGN with evidence (absence of) of a radiative-mode AGN. Generally, a small subset of LOFAR sources can show signs of a radiative-mode AGN (e.g., [30, 73]). As discussed in Sect. 1, spectroscopic data would allow us to more robustly distinguish between these two classes ([3]). Although surveys such as DESI provide spectroscopic observations, they limit the coverage and analysis of radio sources in EDF-N due to optical-based target selection and incompleteness in the optical selection function, especially beyond . More complete spectroscopic observations of radio-selected sources are therefore needed to reliably distinguish between HERGs and LERGs.
Instead, we make use of the AGN fraction obtained from the SED fitting with CIGALE (L26) to identify radiative-mode AGN. This AGN fraction is defined as the ratio of the AGN integrated luminosity to the total luminosity within the wavelength range of 5–20 m [120]. We adopt the method discussed in L26, who found that the AGN fraction itself shows large uncertainties and instead defined a parameter where is the 1 uncertainty on . By comparing with a spectroscopic sample in the EDF-N, they found that requiring to select AGN leads to a relatively small contamination of about 9%6 based on the fraction of non-AGN sources being misclassified as AGN [39 L26].
Using obtained through SED fitting with CIGALE, we classified RLAGN as HERGs for sources with larger than 0.2 and as LERGs otherwise. This selection resulted in a sample containing 353 HERGs and 2 805 LERGs. The resulting number of sources for every selection step discussed above is shown in Table 1. To illustrate the cross-matched catalogue, we show as a function of in Figure 4. We also show the sensitivity limit, assuming a spectral index and using a central point source 5 detection limit of 32 Jy beam.
| Selection step | Number of sources |
| All radio sources in EDF-N | 23 309 |
| Sources outside star mask | 19 550 |
| Cross-matched sources | 14 978 |
| With physical parameters | 13 477 |
| With reliable physical parameters | 10 794 |
| RLAGN | 3 158 |
| LERGs | 2 805 |
| HERGs | 353 |
Table 1. Summary of the number of sources in the LOFAR-Euclid cross-matched sample of the EDF-N for different selection steps.
Figure 4. Radio luminosity as a function of redshift . The colour scheme is the same as in Figure 3. The black line shows the sensitivity limit corresponding to an RMS flux of Jy beam.
Incidence of LERGs
With the cross-matched sample, the RLAGN selection, the completeness, and the cross-matching contamination all in place, we can now calculate the true RLAGN incidence as a function of and in the EDF-N. Since the number of HERGs is too limited for a detailed statistical analysis (see Table 1), we only calculate the LERG incidence in the following sections. To calculate , we follow the method of Janssen et al. [69] to account for the flux-limited nature of the radio and optical samples. Because raw source counts are biased towards intrinsically luminous or lower-redshift objects, we weight each source by . This is the maximum co-moving volume over which a source could be detected given its luminosity and the survey sensitivity limits, following the formalism [99]. This method takes into account the redshift limits from both the radio and optical selection criteria, as well as the limits of the redshift range being analysed.
The maximum co-moving volume in which source would still be detected given the 5 detection limit based on the local radio RMS level is denoted by . For the general galaxy sample within the LOFAR footprint, we determine the maximum co-moving volume of source by using the 3 detection limit for the VIS band. We calculate the volume-corrected as a function of host properties and in different bins, using
where is the number of LERGs in a given and bin and are all galaxies in a given bin. We define as a correction factor to account for the incompleteness of optical identifications for the radio sources. We determine this factor by evaluating the ratio of the number of radio sources with an optical counterpart to the total number of radio sources outside the star mask within a given radio flux bin. We only consider bins that are above the completeness limit discussed in Sect. 3.5. Uncertainties on are estimated assuming Poisson statistics [124], and is only calculated when there are more than five sources in a bin to avoid small-number fluctuations. As discussed in Sect. 3.4, we need to account for the increase in cross-matching contamination as decreases. Therefore, we interpolated shown in Figure 1 and used this to correct , i.e., . Similarly, the total quadratic uncertainty on is defined as the quadratic sum of the Poisson uncertainty on and the uncertainty on .
As discussed above, we first binned all LERGs by redshift to apply a volume correction. For this, we adopted the redshift values . The lower limit was chosen to ensure a large enough cosmological volume is observed with the 8.73 deg sky area. For the local Universe (), the sky area is too small to detect a significant number of sources for an analysis of the LERG incidence with both and . Instead, we will use found by Sabater et al. [97]. This work used the LoTSS DR1 data (Shimwell et al. 2019), which covers in the Hobby-Eberly Telescope Dark Energy Experiment (HETDEX) field, to calculate the RLAGN incidence to . The different radio-AGN selection and higher radio luminosities probed by Sabater et al. [97] do mean that their sample contained a non-negligible sample of HERGs. Nevertheless, the large sky area serves as a good benchmark for the local Universe.
We binned in the range in steps of dex. This step size was chosen because the values in our sample have an average uncertainty of dex. The limits were chosen to be . The lower limit matches the detection limit in the deepest regions of the radio data in the EDF-N at (Bondi et al. [15]). The highest limit ensures at least one bin in every redshift bin, as shown in Figure 4.
Using this binning scheme and the method described above, the resulting are shown in Figure 5. Each panel represents a bin, with fractions plotted at the centres of the bins. As a reference, we also show values from Sabater et al. [97] in the top left panel, as well as the relation for the local Universe (Best et al. [9]). This relation was derived for radio-AGN with radio luminosities at and redshifts . Assuming a spectral index of , this corresponds to . The results reveal a strong correlation between and over a wide redshift range extending beyond the epoch of cosmic noon. Previous studies have found this dependence regardless of the RLAGN selection method (e.g., Williams & Röttgering [132]; Sabater et al. [97]; Kondapally et al. [73, 74]).
Figure 5. Fraction as a function of for different limits and its redshift evolution in the range . The dot-dashed vertical lines indicate the completeness limit for each redshift bin. The black dashed line in each panel shows the relation found in the local Universe (Best et al. [9]). The coloured dashed lines are the power-law fits of for every radio luminosity limit that is detectable in each panel. Top left: Fraction for the local Universe using LoTSS DR1 data (Sabater et al. [97]).
Redshift evolution of the LERG incidence
Qualitatively, Figure 5 shows that the LERG incidence resembles the relation observed in the local Universe to , before flattening towards higher . The fraction at fixed increases with , implying that LERGs are dominated by galaxies spanning a broader range at earlier epochs. For the highest stellar masses, , approaches and exhibits only a modest increase with , implying that the flattening of the power-law relation is driven mainly by the stronger evolution of lower-mass galaxies. Both the increase in normalisation and the flattening are consistent with previous studies (Williams & Röttgering [132]; Wang et al. [126]; Kondapally et al. [74]). Interestingly, Figure 5 also suggests a slight deviation from the approximately linear relation at lower in several bins. Although the current results do not allow us to quantify this deviation robustly, it may indicate a change in the dominant SMBH triggering mechanisms in lower-mass galaxies.
In the redshift range –, there is an indication that almost all the most massive galaxies () with host an LERG, consistent with Sabater et al. [97]. However, the number of galaxies with is small, with about 10 objects per redshift bin. Our requirement of at least five radio sources per bin means most of the values at those masses are not shown. These small-number statistics are more prone to statistical fluctuations, meaning that we would need larger samples to study whether reaches 100% in more bins.
The strong dependence of on may reflect the fuelling mechanisms operating in LERGs (e.g., Allen et al. [2]; Hardcastle et al. [60]). LERGs are generally associated with radiatively inefficient accretion of hot gas, which is expected to be most effective in massive galaxies with substantial hot gaseous haloes. This gas reservoir may explain the higher incidence of LERGs in more massive galaxies. This is consistent with the balance between radio-source heating and radiative cooling losses proposed by Best et al. [8], in which the hot halo persistently fuels the central SMBH, while AGN activity offsets cooling. At lower stellar masses (), galaxies are more frequently disc-dominated and host lower-mass SMBHs, potentially hindering sustained hot-mode accretion. Together, these effects may explain the observed increase of with , although our observations alone cannot establish the origin of this steep dependence.
To quantify the redshift evolution of the – relation, we fit a power law in each bin. For a normalisation at and a power-law index , the relation takes the form
We fit using orthogonal distance regression7 (ODR, Boggs & Rogers [13]), requiring measurements across at least three bins for a given limit. We adopt uncertainties of 0.15 dex in , corresponding to the bin width, and Poissonian uncertainties in . We also fit from Sabater et al. [97] for comparison with the local Universe. The fits are also shown in Figure 5, and the evolution of the fitted power-law parameters with redshift is shown in Figure 6 and Figure 7.
Figure 6. Redshift evolution of normalisation for the powerlaw fits of . Horizontal uncertainties indicate the width of the redshift bins, and the different colours represent different limits. Triangles represent obtained from fitting in the local Universe (Sabater et al. [97]), while the circles show for in the EDF-N.
Figure 7. Same as in Figure 6, but for . The black squares show found by Williams & Röttgering [132] for a combined sample of SDSS-NVSS in the local Universe and VLA-COSMOS for higher redshifts.
Focusing on the results for in Figure 6, we see that the normalisation within a given bin decreases as we set higher limits. This decrease is to be expected because more luminous radio sources are intrinsically rarer. Interestingly, there is a very strong increase in the overall LERG incidence with compared to the local Universe. The overall incidence of LERGs increases for all limits, and especially the most luminous objects show an increase by a factor of up to 20 from the first to the last bin. The incidence of less luminous sources also increases substantially, but the current depth of the radio data cannot definitively
determine whether they will show similar growth to . The increasing normalisation for the most luminous sources has been found before (e.g., Williams & Röttgering [132]), and supports the idea that radio-AGN activity was more prevalent at higher redshifts (e.g., Heckman & Best [63]; Alexander et al. [1]).
In Figure 7, we show the power-law index for each limit and its evolution with . This shows that decreases with for all limits. For the highest radio luminosities, the index is up to and decreases to towards the highest redshifts. This behaviour is consistent with results obtained for the COSMOS field (Williams & Röttgering [132]; Wang et al. [126]) and the LOFAR Deep Fields (Kondapally et al. [73]), where slopes of – are found at . For example, the evolution reported by Williams & Röttgering [132] for sources with is remarkably similar to our results for . They interpreted this flattening as evidence for an increasing contribution from HERG activity or, more generally, cold-mode accretion at higher . While their sample could not be separated into HERGs and LERGs, we excluded HERGs using the CIGALE-derived parameter (see Sect. 4). The persistence of the same trend in our sample therefore implies that this evolution also applies to the LERG population and cannot be attributed solely to an increase in HERG activity. The flattening of the – relation, in particular at lower and where SFGs are expected to dominate the galaxy stellar mass function, suggests the presence of a star-forming LERG population with a potentially different fuelling mechanism (Kondapally et al. [73, 74]). We further investigate this in Sect. 5.2 by studying the redshift evolution of as a function of both and SFR.
As discussed in Sect. 3.3, our EDF-N cross-matched sources have physical parameters from E26 or L26. To test for potential biases from combining the catalogues, we performed two tests. First, for the 947 RLAGN in both catalogues, we compared the probability distribution function (PDF) of the photometric , SFR, and to identify any discrepancies. Second, we calculated as a function of and in the range separately, using from both catalogues to check for any systematic offset. Both tests showed that the physical parameters from both catalogues yield comparable results. More details about these tests can be found in Appendix § 9.
We also repeated the analyses shown in Figure 5, Figure 6, and Figure 7 excluding extended sources without a radio excess. The results were consistent within uncertainties and retained the decreasing trend of with redshift. Finally, we applied a more conservative HERG selection by requiring , which yielded 662 HERGs and 2 496 LERGs. Given its relatively high contamination rate (Euclid Collaboration: Matamoro Zatarain et al. [39]), this threshold provides a conservative test of our HERG selection. The results were again consistent within uncertainties, showing evolution from in the local Universe to at higher redshifts. Together, these tests demonstrate the robustness of our results with respect to our classifications.
Redshift evolution of star-forming and quiescent LERGs
As mentioned in Sect. 5.1, the flattening of the relation between and with for LERGs has been observed in other works as well. For example, Kondapally et al. [73, 74] argued that this can be explained by both an increasing number of active SFGs and an increasing among the star-forming population at fixed towards higher . In particular, Kondapally et al. [73] showed that more quiescent LERGs have a steep power-law dependence with a slope , while star-forming LERGs show out to .
Figure 7 shows that the flattening of continues towards higher redshift down to a power-law index of for the highest radio luminosities. An increasing fraction of star-forming relative to quiescent LERGs would naturally result in a flattening of with . An explanation for this is that the higher gas fractions at higher redshifts (e.g., Genzel et al. [54]; Tacconi et al. [117]) result in an increasing number of SFGs. Moreover, at higher redshifts, galaxies have had less time to undergo the morphological transformation associated with mergers, resulting in a higher prevalence of disc-dominated systems compared to the local Universe (Conselice [22]; Martin et al. [84]).
To test whether there is a difference between star-forming and quiescent LERGs, we split our sample of LERGs into two groups. The sub-classification is based on the idea that while SFGs follow the “main-sequence” of star formation in the SFR– plane (e.g., Whitaker et al. [128]; Speagle et al. [112]; Schreiber et al. [100]), quiescent galaxies lie below this sequence and we can use the specific SFR (sSFR = SFR/) to distinguish between star-forming and quiescent galaxies (Fontana et al. [51]; Pacifici et al. [92]; Merlin et al. [87]). We classify a source as quiescent when it satisfies sSFR , where is the age of the Universe at a certain in Gyr and defines the threshold. This threshold, adopted from Kondapally et al. [73], roughly divides our LERG sample in half and is consistent with traditional UVJ rest-frame colour-selected samples (Pacifici et al. [92]; Carnall et al. [17]). The classification results in 1 535 star-forming (54.7%) and 1 270 quiescent (45.3%) LERGs. To check our selection method, we also used the NUV– and – colours, which were shown to be consistent, as discussed in Appendix B.
Following the procedure adopted earlier for the full LERG sample, we calculate as a function of for different limits in each redshift bin and fit a power law. For quiescent LERGs, the fraction is calculated relative to the quiescent galaxy population, and analogously for star-forming LERGs. Of the 1 945 312 Euclid galaxies within the LOFAR EDF-N footprint, 173 479 are classified as quiescent and 1 771 842 as star-forming based on their sSFR. The power-law fits to the star-forming and quiescent LERG fractions are presented in Appendix C, and the redshift evolution of the fitted normalisation and slope is shown in Figure 8. The power-law normalisations derived from Sabater et al. [97] are included for comparison only, since they did not distinguish between star-forming and quiescent galaxies.
Figure 8. Same as in Figure 6 and Figure 7, but for quiescent and star-forming LERGs. Quiescent galaxies were classified using sSFR with . Left: Results for quiescent LERGs where the upper panel shows the evolution of and the lower panel shows the evolution of . Right: Results of power-law fits for star-forming LERGs.
Despite the larger uncertainties arising from the smaller sample, star-forming LERGs exhibit significantly flatter power-law relations than quiescent LERGs. However, we do not find that the power-law index for quiescent sources remains for the entire redshift range. Out to , the slopes remain consistent with the results in Kondapally et al. [73] for the highest , but the slope decreases and becomes more similar to the star-forming population with at higher redshifts. One caveat is that with the current depth of the LOFAR data in the EDF-N, we cannot probe lower radio luminosities across the entire redshift range as was done in Kondapally et al. [73]. Nevertheless, the similar slopes at higher redshifts suggest that quiescent and star-forming systems may be triggered similarly at those redshifts, at least for the highest radio luminosities.
An explanation for this could be that the large availability of cold gas at high enables very efficient fuelling of the SMBH. However, once the cold gas reservoirs start to deplete and star formation stops, the fuelling process becomes dependent on either gas-rich mergers, which ignite star formation and drive cold gas to the centre of the galaxy, leading to . Alternatively, the fuelling will depend on the ability of the galaxy to accrete hot gas from the halo, which follows . This suggests an evolutionary picture where the higher availability of cold gas and higher merger rates at earlier epochs influence the LERG incidence (Hopkins et al. [64]; Chiaberge et al. [18]).
Nevertheless, our study is currently limited by the lack of spectroscopic data for the full RLAGN sample, which would provide accurate redshifts and reliable classifications of LERGs and HERGs. Combining spectroscopy with deeper radio data across wider areas is needed to obtain enough statistics to test whether the – relation persists to higher redshifts for less luminous radio sources.
Conclusions
In this work, the new Euclid Q1 data in the EDF-N have been combined with LOFAR radio observations at 144 MHz in the same field. The LOFAR sources were cross-matched with optical/NIR Q1 sources using the LR method based only on co-ordinates to obtain a highly reliable sample. We avoid the use of colours to prevent biasing our matching with any priors on galaxy colour. Subsequently, physical parameters for the cross-matched sample were obtained from the catalogues presented in E26 and L26 for galaxies and AGN candidates, respectively. After correcting for incompleteness and cross-matching contamination, sources were classified into SFGs and RLAGN. The latter was done by identifying extended radio emission or a radio excess based on the SFR and using a probabilistic approach.
Finally, we classified all radio-AGN as LERGs or HERGs using the AGN fraction obtained through SED fitting.
The resulting LERG incidence shows a tight correlation over 2 orders of magnitude in across a wide range of redshifts. Our results confirm that LERGs are preferentially associated with the most massive galaxies, as found before by Sabater et al. [97]. Fitting the LERG fractions with a power law showed that the relations flatten with redshift from in the local Universe to at . Moreover, the normalisation of was found to increase with redshift for all probed radio luminosities, with the highest radio luminosities W Hz increasing the most from the local Universe to . Overall, our results show that the LERG population becomes more dominated by less massive sources and that the incidence of LERGs increases significantly with increasing redshift.
One possible reason for the increased incidence at higher redshift could be the greater availability of cold gas, which may enable efficient accretion and thus more activity of the SMBH (Kondapally et al. [73, 74]; Magliocchetti et al. [83]). In addition, the increasing dominance of less massive sources may be related to higher merger rates with increasing redshift. Given that the overall number density of less massive galaxies is higher than for the most massive objects, the RLAGN incidence could consequently be boosted by higher merger rates. In our analysis, we cannot disentangle the effects of the evolving merger rate and the increasing availability of cold gas. The observed flattening of may therefore reflect a combination of these effects, although their relative contributions remain uncertain.
Separating the LERG population into quiescent and star-forming sources suggested an evolutionary picture with particularly distinct behaviour in each class. In agreement with previous work (Kondapally et al. [73, 74]), we find that quiescent LERGs show a significantly steeper dependence with than star-forming LERGs, which show for . However, at the highest redshifts (), both populations show similar relations with . Our interpretation is that the large abundance of cold gas at higher redshifts enables efficient fuelling of the SMBHs. However, as cold gas becomes more scarce at lower redshifts, the fuelling mechanism changes to the accretion of hot gas. This phenomenon is more dependent on the mass of the host galaxy and its halo. In this picture, galaxy mergers can resupply the availability of cold gas to trigger star formation and efficient fuelling of the SMBH for some galaxies. Other sources that do not undergo mergers become quiescent, and their host mass influences how efficiently they can accrete hot gas and thus whether an RLAGN can be triggered.
An interesting avenue to explore would be to check how many of our classified LERGs are part of a known galaxy group or cluster. The RLAGN incidence could be linked to the host galaxy environment, which is often seen as one of the main factors determining the HERG/LERG appearance and RLAGN incidence in general (Heckman & Best [63]; Koulouridis et al. [79]; Hale et al. [57]). Galaxy mergers are also thought to play an important role in triggering AGN and leading to radio activity (Magliocchetti et al. [83]). However, the number of RLAGN in the EDF-N data is too low to estimate the LERG incidence robustly while incorporating these environmental parameters. This would be something to explore further in the first large data release of Euclid (DR1).
Future studies of RLAGNs and their hosts will benefit from combining the wide Euclid survey with LoTSS DR3 (Shimwell et al. [105]) and other large radio surveys. Although LoTSS DR3 is shallower than the EDF-N, and thus less suited to studies extending to , its larger sample will enable more robust statistical analyses at lower redshifts. This will allow studies to consider not only redshift and , but also other properties like the HERG/LERG classification, host-galaxy morphology and environment. Moreover, the upcoming spectroscopic WEAVE-LOFAR survey (Smith et al. [109]) will further improve redshift quality and source classification of LOFAR radio sources. This survey is planned to obtain a million spectra for sources detected across LoTSS using the William Herschel Telescope Enhanced Area Velocity Explorer (WEAVE) multi-object spectrograph (Jin et al. [71]). In addition, the final LoTSS Deep Fields (e.g., De Jong et al. [25]; Shimwell et al. [104]) will provide detailed sub-arcsecond resolution imaging over an estimated 60 deg.8 Comparable angular resolutions will improve cross-matching reliability, enabling direct comparisons with optical surveys such as Euclid and providing a powerful resource for studying the evolution of radio-AGN and AGN feedback.
Acknowledgements. SV thanks C. Saulder for compiling the external spectroscopic redshift catalogues for the Q1 data and making this available to the Euclid Consortium. SV acknowledges support from NWO through The Dark Universe Science Collaboration (DUSC): Cosmology with the Euclid mission (OCENW.XL21.XL21.025; Grant ID: https://doi.org/10.61686/NHOQI85483). RK acknowledges support from the Leverhulme Trust through the Leverhulme Early Career Fellowship. LKM is grateful for support from a UKRI FLF [MR/Y020405/1] and LOFAR-UK via STFC [ST/V002406/1]. ALM acknowledges support through the European Space Agency (ESA) Research Fellowship Programme. GJW expresses grateful thanks for the award of an Emeritus Fellowship from The Leverhulme Trust. LOFAR data products were provided by the LOFAR Surveys Key Science project (LSKSP; https://lofar-surveys.org/) and were derived from observations with the International LOFAR Telescope (ILT). LOFAR (Van Haarlem et al. [122]) is the Low Frequency Array designed and constructed by ASTRON. It has observing, data processing, and data storage facilities in several countries, which are owned by various parties (each with their own funding sources), and which are collectively operated by the ILT foundation under a joint scientific policy. The efforts of the LSKSP have benefited from funding from the European Research Council, NOVA, NWO, CNRS-INSU, the SURF Co-operative, the UK Science and Technology Funding Council, and the Jülich Supercomputing Centre. The Euclid Consortium acknowledges the European Space Agency and a number of agencies and institutes that have supported the development of Euclid, in particular the Agenzia Spaziale Italiana, the Austrian Forschungsförderungsgesellschaft funded through BMBMI, the Belgian Science Policy, the Canadian Euclid Consortium, the Deutsches Zentrum für Luft- und Raumfahrt, the DTU Space and the Niels Bohr Institute in Denmark, the French Centre National d’Etudes Spatiales, the Fundação para a Ciência e a Tecnologia, the Hungarian Academy of Sciences, the Ministerio de Ciencia, Innovación y Universidades, the National Aeronautics and Space Administration, the National Astronomical Observatory of Japan, the Netherlands Onderzoekschool Voor Astronomie, the Norwegian Space Agency, the Research Council of Finland, the Romanian Space Agency, the Swiss Space Office (SSO) at the State Secretariat for Education, Research, and Innovation (SERI), and the United Kingdom Space Agency. A complete and detailed list is available on the Euclid web site (www.euclid-ec.org/consortium/community/). This work has made use of the Euclid Quick Release Q1 data from the Euclid mission of the European Space Agency (ESA), 2025, https://doi.org/10.57780/esa-2853f3b. This research makes use of ESA Datalabs (datalabs.esa.int), an initiative by ESA’s Data Science and Archives Division in the Science and Operations Department, Directorate of Science.
References
Authors and affiliations
S. Visser{sup}*1, H. J. A. Rottgering1, L. Wang2, 3, R. Kondapally4, 5, S. Tacchetto6, B. Laloux7, 8, L. K. Morabito4, 5, F. Ricci9, 10, M. Magliocchetti11, A. La Marca12, 1, L. Bisigello13, A. Viitanen14, 15, 10, G. J. White16, 17, B. Altieri18, L. Amendola19, 20, S. Andreon21, N. Auricchio22, C. Baccigalupi23, 24, 25, 26, M. Baldi27, 22, 28, S. Bardelli22, P. Battaglia22, A. Biviano24, 23, E. Branchini29, 30, 21, M. Brescia31, 7, S. Camera32, 33, 34, V. Capobianco34, C. Carbone35, J. Carretero36, 37, M. Castellano10, G. Castignani22, S. Cavuoti7, 38, K. C. Chambers39, A. Cimatti40, C. Colodro-Conde41, G. Congedo42, C. J. Conselice43, L. Conversi44, 18, Y. Copin45, F. Courbin46, 47, 48, H. M. Courtois49, M. Cropper50, H. Degaudenzi14, G. De Lucia24, H. Dole51, F. Dubath14, X. Dupac18, S. Dusini52, M. Fabricius8, 53, M. Farina11, R. Farinelli22, F. Faustini10, S. Ferriol45, M. Frailis24, E. Franceschi22, M. Fumana35, L. Gabarra54, S. Galeotta24, K. George55, B. Gillis42, C. Giocoli22, 28, J. Gracia-Carpio8, A. Grazian13, F. Grupp8, 53, S. Gwyn56, W. G. Hartley14, S. V. H. Haugan57, S. Hemmati58, H. Hoekstra1, W. Holmes59, I. M. Hook60, F. Hormuth61, A. Hornstrup62, 63, K. Jahnke64, M. Jhabvala65, S. Kermiche66, A. Kiessling59, B. Kubik45, K. Kuijken1, M. Kümmel53, M. Kunz67, H. Kurki-Suonio68, 69, A. M. C. Le Brun70, V. Le Brun71, S. Ligori34, P. B. Lilje57, V. Lindholm68, 69, I. Lloro72, G. Mainetti73, O. Mansutti24, O. Marggraf74, M. Martinelli10, 75, N. Martinet71, F. Marulli76, 22, 28, R. J. Massey5, N. Mauri40, 28, E. Medinaceli22, S. Mei77, 78, M. Meneghetti22, 28, E. Merlin13, G. Meylan79, J. J. Mohr55, P. Monaco80, 24, 25, 23, A. Mora81, M. Moresco76, 22, C. Moretti24, 23, 25, L. Moscardini76, 22, 28, C. Neissner82, 37, S.-M. Niemi12, C. Padilla82, S. Paltani14, F. Pasian24, W. J. Percival83, 84, 85, V. Pettorino12, A. Pezzotta21, G. Polenta86, M. Poncet87, L. A. Popa88, L. Pozzetti22, F. Raison8, A. Renzi89, 52, 22, J. Rhodes59, G. Riccio7, I. Risso29, 30, 21, F. Rizzo24, E. Romelli24, M. Roncarelli22, B. Rusholme58, R. Saglia53, 8, Z. Sakr90, 91, 92, D. Sapone93, B. Sartoris8, 24, J. A. Schewtschenko42, M. Schirmer64, P. Schneider74, T. Schrabback6, A. Secroun66, E. Sihvola15, P. Simon74, C. Sirignano89, 52, G. Sirri28, P. Tallada-Crespí36, 37, A. N. Taylor42, I. Tereno94, 95, N. Tessore50, S. Toft96, 97, R. Toledo-Moreo98, 99, F. Torradeflot37, 36, I. Tutusaus100, 101, 91, J. Valiviita68, 69, T. Vassallo24, 55, A. Veropalumbo22, Y. Wang21, 30, 29, J. Weller53, 8, A. Zacchei24, 23, G. Zamorani22, F. M. Zerbi21, E. Zucca22, C. M. Gutierrez41, 102, K. Tanidis103, and D. Scott104 36 Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas (CIEMAT), Avenida Complutense 40, 28040 Madrid, Spain
37 Port d’Informació Científica, Campus UAB, C. Albareda s/n, 08193 Bellaterra (Barcelona), Spain
38 INFN – Sezione di Napoli, Via Cinthia 6, 80126, Napoli, Italy
39 Institute for Astronomy, University of Hawaii, 2680 Woodlawn Drive, Honolulu, HI 96822, USA
40 Dipartimento di Fisica e Astronomia "Augusto Righi" - Alma Mater Studiorum Università di Bologna, Viale Berti Pichat 6/2, 40127 Bologna, Italy
41 Instituto de Astrofísica de Canarias, E-38205 La Laguna, Tenerife, Spain
42 Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ, UK
43 Jodrell Bank Centre for Astrophysics, Department of Physics and Astronomy, University of Manchester, Oxford Road, Manchester M13 9PL, UK
44 European Space Agency/ESRIN, Largo Galileo Galilei 1, 00044 Frascati, Roma, Italy
45 Université Lyon 1, CNRS/IN2P3, IP2I Lyon, Villeurbanne, France
46 Institut de Ciències del Cosmos (ICCUB), Universitat de Barcelona (IEEC-UB), Martí i Franquès 1, 08028 Barcelona, Spain
47 Institució Catalana de Recerca i Estudis Avançats (ICREA), Passeig de Lluís Companys 23, 08010 Barcelona, Spain
48 Institut de Ciencies de l’Espai (IEEC-CSIC), Campus UAB, Carrer de Can Magrans, s/n Cerdanyola del Vallés, 08193 Barcelona, Spain
49 Université Lyon 1, CNRS/IN2P3, IUF, IP2I Lyon, Villeurbanne, France
50 Mullard Space Science Laboratory, University College London, Holmbury St Mary, Dorking, Surrey RH5 6NT, UK
51 Université Paris-Saclay, CNRS, Institut d’astrophysique spatiale, 91405, Orsay, France
52 INFN-Padova, Via Marzolo 8, 35131 Padova, Italy
53 Universitäts-Sternwarte München, Fakultät für Physik, Ludwig-Maximilians-Universität München, Scheinerstr. 1, 81679 München, Germany
54 Department of Physics, University of Oxford, Keble Road, Oxford OX1 3RH, UK
55 University Observatory, LMU Faculty of Physics, Scheinerstr. 1, 81679 Munich, Germany
56 National Research Council, Herzberg Astronomy and Astrophysics Research Centre, 5071 W. Saanich Rd. Victoria, BC, V9E 2E7, Canada
57 Institute of Theoretical Astrophysics, University of Oslo, P.O. Box 1029 Blindern, 0315 Oslo, Norway
58 Caltech/IPAC, 1200 E. California Blvd., Pasadena, CA 91125, USA
59 Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA, 91109, USA
60 School of Physics and Astronomy, Lancaster University, Lancaster, LA1 4YB, UK
61 Felix H. Hormuth Engineering, Goethestr. 17, 69181 Leimen, Germany
62 Technical University of Denmark, Elektrovej 327, 2800 Kgs. Lyngby, Denmark
63 Cosmic Dawn Center (DAWN), Denmark
64 Max-Planck-Institut für Astronomie, Königstuhl 17, 69117 Heidelberg, Germany
65 NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA
66 Aix-Marseille Université, CNRS/IN2P3, CPPM, Marseille, France
67 Université de Genève, Département de Physique Théorique and Centre for Astroparticle Physics, 24 quai Ernest-Ansermet, CH-1211 Genève 4, Switzerland
68 Department of Physics, P.O. Box 64, University of Helsinki, 00014 Helsinki, Finland
69 Helsinki Institute of Physics, Gustaf Hällströmին katu 2, University of Helsinki, 00014 Helsinki, Finland
70 Laboratoire d’etude de l’Univers et des phenomenes eXtrêmes, Observatoire de Paris, Université PSL, Sorbonne Université, CNRS, 92190 Meudon, France
71 Aix-Marseille Université, CNRS, CNES, LAM, Marseille, France
72 SKAO, Jodrell Bank, Lower Withington, Macclesfield SK11 9FT, UK
73 Centre de Calcul de l’IN2P3/CNRS, 21 avenue Pierre de Coubertin 69627 Villeurbanne Cedex, France
74 Universität Bonn, Argelander-Institut für Astronomie, Auf dem Hügel 71, 53121 Bonn, Germany
75 INFN-Sezione di Roma, Piazzale Aldo Moro, 2 - c/o Dipartimento di Fisica, Edificio G. Marconi, 00185 Roma, Italy
76 Dipartimento di Fisica e Astronomia "Augusto Righi" - Alma Mater Studiorum Università di Bologna, Via Piero Gobetti 93/2, 40129 Bologna, Italy
77 Université Paris Cité, CNRS, Astroparticule et Cosmologie, 75013 Paris, France
78 CNRS-UBC International Research Laboratory, Centre Pierre Binétruy, IRL2007, CPB-IN2P3, Berkeley, USA
79 Institute of Physics, Laboratory of Astrophysics, Ecole Polytechnique Fédérale de Lausanne (EPFL), Observatoire de Sauverny, 1290 Versoix, Switzerland
80 Dipartimento di Fisica - Sezione di Astronomia, Università di Trieste, Via Tiepolo 11, 34131 Trieste, Italy
81 Telespazio UK S.L. for European Space Agency (ESA), Camino bajo del Castillo, s/n, Urbanizacion Villafranca del Castillo, Villanueva de la Cañada, 28692 Madrid, Spain
82 Institut de Física d’Altes Energies (IFAE), The Barcelona Institute of Science and Technology, Campus UAB, 08193 Bellaterra (Barcelona), Spain
83 Waterloo Centre for Astrophysics, University of Waterloo, Waterloo, Ontario N2L 3G1, Canada
84 Department of Physics and Astronomy, University of Waterloo, Waterloo, Ontario N2L 3G1, Canada
85 Perimeter Institute for Theoretical Physics, Waterloo, Ontario N2L 2Y5, Canada
86 Space Science Data Center, Italian Space Agency (ASI), via del Politecnico snc, 00133 Roma, Italy
87 Centre National d’Etudes Spatiales – Centre spatial de Toulouse, 18 avenue Edouard Belin, 31401 Toulouse Cedex 9, France
88 Institute of Space Science, Str. Atomistilor, nr. 409 Măgurele, Ilfov, 077125, Romania
89 Dipartimento di Fisica e Astronomia "Galilei", Università di Padova, Via Marzolo 8, 35131 Padova, Italy
90 Instituto de Física Teórica UAM-CSIC, Campus de Cantoblanco, 28049 Madrid, Spain
91 Institut de Recherche en Astrophysique et Planétologie (IRAP), Université de Toulouse, CNRS, UPS, CNES, 14 Av. Edouard Belin, 31400 Toulouse, France
92 Université St Joseph, Faculty of Sciences, Beirut, Lebanon
93 Departamento de Física, FCFM, Universidad de Chile, Blanco Encalada 2008, Santiago, Chile
94 Departamento de Física, Faculdade de Ciências, Universidade de Lisboa, Edifício C8, Campo Grande, PT1749-016 Lisboa, Portugal
95 Instituto de Astrofísica e Ciências do Espaço, Faculdade de Ciências, Universidade de Lisboa, Tapada da Ajuda, 1349-018 Lisboa, Portugal
96 Cosmic Dawn Center (DAWN)
{sup}97` Niels Bohr Institute, University of Copenhagen, Jagtvej 128, 2200 Copenhagen, Denmark
{sup}98` Universidad Politécnica de Cartagena, Departamento de Electrónica y Tecnología de Computadoras, Plaza del Hospital 1, 30202 Cartagena, Spain
{sup}99` European University of Technology EUt+, European Union
{sup}100` Institute of Space Sciences (ICE, CSIC), Campus UAB, Carrer de Can Magrans, s/n, 08193 Barcelona, Spain
{sup}101` Institut d’Estudis Espacials de Catalunya (IEEC), Edifici RDIT, Campus UPC, 08860 Castelldefels, Barcelona, Spain
{sup}102` Universidad de La Laguna, Dpto. Astrofísica, E-38206 La Laguna, Tenerife, Spain
{sup}103` Center for Astrophysics and Cosmology, University of Nova Gorica, Nova Gorica, Slovenia
{sup}104` Department of Physics and Astronomy, University of British Columbia, Vancouver, BC V6T 1Z1, Canada
Comparison of LERG fractions for physical parameters obtained with different methods
As discussed in Sect. 5.1, we make use of two catalogues to provide physical properties and photometric redshifts for the Euclid sources. For the analysis in this paper, it is important to verify that this will not introduce bias into the calculated LERG incidence. Therefore, we look at all objects for which we have parameter estimates from both E26 and L26. Of the 10 794 radio sources with reliable physical parameters in the final catalogue discussed in Sect. 3.3, 3 653 are present in both catalogues. The first test is to compare the PDFs of the photometric redshift, SFR, and to check for any discrepancies. Any significant differences would place sources in different redshift or bins, thereby changing the inferred for the two methods of deriving physical properties. This check yielded similar distributions, with the medians and uncertainties close to each other.
(a)
(b)
Figure A.1. Same as in Figure 5, but only for the range . The black line in each panel shows the relation found in the local Universe ([9]). Top: Fraction for the physical parameters derived in E26. Bottom: Fraction for the parameters derived in L26.
(a)
(b)
Figure A.2. Fitted power-law parameters for different limits in the range , where different colours indicate the limits. Top: Power-law index, where circles represent the parameters found for the physical parameters derived in E26 and the squares show the results for the parameters derived in L26 for the same objects. Bottom: Same as in the top panel, for the normalisation.
We also calculate as a function of and in the redshift range for physical parameters from both catalogues. This redshift range was chosen because the number of sources with physical parameters in both catalogues was largest. Within this redshift range, there are 148 LERGs with physical parameters derived in L26. For the same objects, we obtained photometric redshifts and from the catalogue in E26 and calculated for both, as seen in Fig. A.1. Note that this results in slightly fewer objects, 110, with physical parameters in both catalogues, since we required reliable parameters for both catalogues and that the photometric redshifts lie in the same redshift range. Subsequently, we fitted the resulting with a power law (see Eq. (8)) to assess consistency. For the sources with parameters derived in L26, we see a mildly higher normalisation. However, the fitted power-law parameters shown in Fig. A.2 were found to be consistent within uncertainties. This shows that even with a quite limited number of objects, both methods agree, and this test confirms the reliability of the results discussed before.
Figure A.3. NUV– and – colours for all radio sources with a Euclid match in our final sample. Using the colour criteria defined by Ilbert et al. [67], we classify galaxies into star-forming and quiescent and compare this to the same classification using the sSFR for different redshift bins. For the classification using the sSFR, we show sources classified as star-forming in blue and sources classified as quiescent in orange.
Selecting quiescent and star-forming galaxies
As discussed in Sect. 5.2, we split our sample of LERGs into quiescent and star-forming galaxies using the sSFR to investigate if these two populations evolve differently with and . To check our selection method, we use NUV– and – colours to select quiescent galaxies that satisfy ([67 E26])
This combination of criteria is analogous to the UVJ diagram. It has become widely used in recent studies because the NUV– colour is more sensitive to recent star formation, providing a cleaner separation between passive (redder) and star-forming (bluer) galaxies (Arnouts et al. (2013); Ilbert et al. [66]).
In Fig. A.3, we compare the classification of galaxies as quiescent or star-forming based on the sSFR with that obtained using the colour criteria defined in Eq. B.1. Since we require a detection in all three bands (NUV, , ), the comparison includes 9 815 out of 10 794 radio sources with a Euclid counterpart from the final sample (Sect. 3.3). Figure A.3 shows good agreement between the two classification methods, indicating that the adopted sSFR-based classification is robust. Applying the colour selection to the 2 656 detected LERGs yields 864 quiescent (32.5%) and 1 792 star-forming (65.5%) sources, meaning that the colour selection yields relatively more star-forming objects. Although a more stringent threshold of would produce closer agreement with the colour selection, we adopt in Sect. 5.2 to be consistent with Kondapally et al. [73, 74]. We verified that this choice does not significantly affect our results by calculating the LERG incidence for both thresholds.
Fitting fractions of star-forming and quiescent LERGs
As discussed in Sect. 5.2, we separated all LERGs into star-forming and quiescent galaxies and calculated for each class. Subsequently, we used ODR to fit the calculated fractions for each limit in all redshift bins that have at least three values. The fraction for star-forming and quiescent LERGs and the best power-law fits for every limit in all redshift bins are shown in Figs. C.1 and C.2, respectively.
Figure C.1. Same as in Figure 5, but for star-forming LERGs.
Figure C.2. Same as in Figure 5, but for quiescent LERGs.
References
- [1]Alexander, D. M., Hickox, R. C., Aird, J., et al. 2025, New A Rev., 101, 101733
- [2]Allen, S. W., Dunn, R. J. H., Fabian, A. C., et al. 2006, MNRAS, 372, 21
- [3]Arnaudova, M. I., Smith, D. J. B., Hardcastle, M. J., et al. 2025, MNRAS, 542, 2245
- [5]Baldry, I. K., Balogh, M. L., Bower, R. G., et al. 2006, MNRAS, 373, 469
- [6]Bertin, E., Schefer, M., Apostolokos, N., et al. 2020, in ASPC, Vol. 527, Astronomical Data Analysis Software and Systems XXIX, ed. R. Pizzo, E. R. Deul, J. D. Mol, J. de Plaa, & H. Verkouter, 461
- [7]Best, P. N., Arts, J. N., Röttgering, H. J. A., et al. 2003, MNRAS, 346, 627
- [8]Best, P. N., Kaiser, C. R., Heckman, T. M., & Kauffmann, G. 2006, MNRAS, 368, L67
- [9]Best, P. N., Kauffmann, G., Heckman, T. M., et al. 2005, MNRAS, 362, 25
- [10]Best, P. N., Kondapally, R., Williams, W. L., et al. 2023, MNRAS, 523, 1729
- [11]Bisigello, L., Giulietti, M., Prandoni, I., et al. 2025, The Open Journal of Astrophysics, 8, 101
- [12]Blumenthal, G. R. & Gould, R. J. 1970, Reviews of Modern Physics, 42, 237
- [13]Boggs, P. T. & Rogers, J. 1990, Contem. Math., 112, 186
- [14]Böhme, L., Schwarz, D. J., de Gasperin, F., Röttgering, H. J. A., & Williams, W. L. 2023, A&A, 674, A189
- [15]Bondi, M., Scaramella, R., Zamorani, G., et al. 2024, A&A, 683, A179
- [16]Boquien, M., Burgarella, D., Roehlly, Y., et al. 2019, A&A, 622, A103
- [17]Carnall, A. C., Walker, S., McLure, R. J., et al. 2020, MNRAS, 496, 695
- [18]Chiaberge, M., Gilli, R., Lotz, J. M., & Norman, C. 2015, ApJ, 806, 147
- [19]Chilufya, J., Hardcastle, M. J., Pierce, J. C. S., et al. 2025, MNRAS, 539, 463
- [20]Cochrane, R. K., Kondapally, R., Best, P. N., et al. 2023, MNRAS, 523, 6082
- [21]Condon, J. J. 1992, ARA&A, 30, 575
- [22]Conselice, C. J. 2014, ARA&A, 52, 291
- [23]Croston, J. H., Hardcastle, M. J., Mingo, B., et al. 2019, A&A, 622, A10
- [24]Das, S., Smith, D. J., Haskell, P., et al. 2024, MNRAS, 531, 977
- [25]De Jong, J. M., Van Weeren, R. J., Sweijen, F., et al. 2024, A&A, 689, A80
- [26]De Ruiter, H. R., Willis, A. G., & Arp, H. C. 1977, A&A, 28, 211
- [27]Delvecchio, I., Daddi, E., Sargent, M. T., et al. 2021, A&A, 647, A123
- [28]DESI Collaboration: Abdul Karim, M., Adame, A. G., Aguado, D., et al. 2026, AJ, 171, 285
- [29]Drake, A. B., Smith, D. J. B., Hardcastle, M. J., et al. 2024, MNRAS, 534, 1107
- [30]Duncan, K. J., Kondapally, R., Brown, M. J., et al. 2021, A&A, 648, A4
- [32]Euclid Collaboration: Cropper, M., Al-Bahlawan, A., Amiaux, J., et al. 2025, A&A, 697, A2
- [33]Euclid Collaboration: Enia, A., Pozzetti, L., Bolzonella, M., et al. 2026, A&A, 711, A14
- [34]Euclid Collaboration: Fu, Y., Bouwens, R., Caputi, K. I., et al. 2026, A&A, 711, A285
- [35]Euclid Collaboration: Jahnke, K., Gillard, W., Schirmer, M., et al. 2025, A&A, 697, A3
- [36]Euclid Collaboration: La Marca, A., Wang, L., Margalef-Bentabol, B., et al. 2026, A&A, 711, A17
- [37]Euclid Collaboration: Laloux, B., Bongiorno, A., Salvato, M., et al. 2026, A&A, 712, A182
- [38]Euclid Collaboration: Le Brun, V., Bethermin, M., Moresco, M., et al. 2026, A&A, 711, A7
- [39]Euclid Collaboration: Matamoro Zatarain, T., Fotopoulou, S., Ricci, F., et al. 2026, A&A, 711, A20
- [40]Euclid Collaboration: McCracken, H. J., Benson, K., Dolding, C., et al. 2026, A&A, 711, A2
- [41]Euclid Collaboration: Mellier, Y., Abdurro’uf, Acevedo Barroso, J., et al. 2025, A&A, 697, A1
- [42]Euclid Collaboration: Quilley, L., Damjanov, I., de Lapparent, V., et al. 2026, A&A, 711, A8
- [43]Euclid Collaboration: Romelli, E., Kümmel, M., Dole, H., et al. 2026, A&A, 711, A4
- [44]Euclid Collaboration: Roster, W., Salvato, M., Buchner, J., et al. 2026, A&A, 711, A16
- [45]Euclid Collaboration: Tucci, M., Paltani, S., Hartley, W. G., et al. 2026, A&A, 711, A5
- [46]Euclid Collaboration: Walmsley, M., Huertas-Company, M., Quilley, L., et al. 2026, A&A, 711, A9
- [47]Euclid Quick Release Q1. 2025, https://doi.org/10.57780/esa-2853f3b
- [48]Fabian, A. C. 2012, ARA&A, 50, 455
- [49]Fazio, G. G., Hora, J. L., Allen, L. E., et al. 2004, ApJS, 154, 10
- [50]Ferrarese, L. & Merritt, D. 2000, ApJ, 539, L9
- [51]Fontana, A., Santini, P., Grazian, A., et al. 2009, A&A, 501, 15
- [52]Gaspari, M., Tombesi, F., & Cappi, M. 2020, Nature Astronomy, 4, 10
- [53]Gebhardt, K., Bender, R., Bower, G., et al. 2000, ApJ, 539, L13
- [54]Genzel, R., Tacconi, L. J., Lutz, D., et al. 2015, ApJ, 800, 20
- [55]Grützbauch, R., Chuter, R. W., Conselice, C. J., et al. 2011, MNRAS, 412, 2361
- [56]Gürkan, G., Hardcastle, M. J., Smith, D. J., et al. 2018, MNRAS, 475, 3010
- [57]Hale, C. L., Best, P. N., Duncan, K. J., et al. 2025a, MNRAS, 544, 1323
- [58]Hale, C. L., Heywood, I., Jarvis, M. J., et al. 2025b, MNRAS, 536, 2187
- [59]Hardcastle, M. J. & Croston, J. H. 2020, New A Rev., 88, 101539
- [60]Hardcastle, M. J., Evans, D. A., & Croston, J. H. 2007, MNRAS, 376, 1849
- [61]Hardcastle, M. J., Pierce, J. C. S., Duncan, K. J., et al. 2025, MNRAS, 539, 1856
- [62]Harrison, C. M. & Ramos Almeida, C. 2024, Galaxies, 12, 17
- [63]Heckman, T. M. & Best, P. N. 2014, ARA&A, 52, 589
- [64]Hopkins, P. F., Hernquist, L., Cox, T. J., et al. 2006, ApJS, 163, 1
- [65]Igo, Z., Merloni, A., Hoang, D., et al. 2024, A&A, 686, A43
- [66]Ilbert, O., McCracken, H. J., Le Fèvre, O., et al. 2013, A&A, 556, A55
- [67]Ilbert, O., Salvato, M., Le Floc’h, E., et al. 2010, ApJ, 709, 644
- [68]Ivison, R. J., Greve, T. R., Dunlop, J. S., et al. 2007, MNRAS, 380, 199
- [69]Janssen, R. M. J., Röttgering, H. J. A., Best, P. N., & Brinchmann, J. 2012, A&A, 541, A62
- [70]Jarvis, M., Taylor, R., Agudo, I., et al. 2016, in MeerKAT Science: On the Pathway to the SKA, 6
- [71]Jin, S., Trager, S. C., Dalton, G. B., et al. 2024, MNRAS, 530, 2688
- [72]Kennicutt, Jr., R. C. 1998, ApJ, 498, 541
- [73]Kondapally, R., Best, P. N., Cochrane, R. K., et al. 2022, MNRAS, 513, 3742
- [74]Kondapally, R., Best, P. N., Duncan, K. J., et al. 2025, MNRAS, 536, 554
- [75]Kondapally, R., Best, P. N., Hardcastle, M. J., et al. 2021, A&A, 646, A3
- [76]Kondapally, R., Holc, T., Aird, J., et al. 2026, arXiv e-prints, arXiv:2609.04130arxiv.org/abs/2609.04130
- [77]Kormendy, J. & Ho, L. C. 2013, ARA&A, 51, 511
- [78]Kormendy, J. & Richstone, D. 1995, ARA&A, 33, 581
- [79]Koulouridis, E., Gkini, A., & Drigga, E. 2024, A&A, 684, A111
- [80]Lacy, M., Baum, S. A., Chandler, C. J., et al. 2020, PASP, 132, 035001
- [82]Madau, P. & Dickinson, M. 2014, ARA&A, 52, 415
- [83]Magliocchetti, M., La Marca, A., Bisigello, L., et al. 2026, A&A, 711, A40
- [84]Martin, G., Kaviraj, S., Devriendt, J. E. G., Dubois, Y., & Pichon, C. 2018, MNRAS, 480, 2266
- [85]McCheyne, I., Oliver, S., Sargent, M., et al. 2022, A&A, 662, A100
- [86]McLeod, D. J., McLure, R. J., Dunlop, J. S., et al. 2021, MNRAS, 503, 4413
- [87]Merlin, E., Fontana, A., Castellano, M., et al. 2018, MNRAS, 473, 2098
- [88]Mohan, N. & Rafferty, D. 2015, Astrophysics Source Code Library, record ascl:1502.007
- [89]Narayan, R. & Yi, I. 1994, ApJ, 428, L13
- [90]Novak, M., Smolčić, V., Delhaize, J., et al. 2017, A&A, 602, A5
- [91]Oei, M. S., Hardcastle, M. J., Timmerman, R., et al. 2024, Nature, 633, 537
- [92]Pacifici, C., Kassin, S. A., Weiner, B. J., et al. 2016, ApJ, 832, 79
- [93]Perley, R. A., Chandler, C. J., Butler, B. J., & Wrobel, J. M. 2011, ApJ, 739, L1
- [94]Pozzetti, L., Bolzonella, M., Zucca, E., et al. 2010, A&A, 523, A13
- [95]Read, S. C., Smith, D. J., Gürkan, G., et al. 2018, MNRAS, 480, 5625
- [96]Richter, G. A. 1975, Astron. Nachr., 296, 65
- [97]Sabater, J., Best, P. N., Hardcastle, M. J., et al. 2019, A&A, 622, A17
- [98]Sabater, J., Best, P. N., Tasse, C., et al. 2021, A&A, 648, A2
- [99]Schmidt, M. 1968, ApJ, 151, 393
- [100]Schreiber, C., Pannella, M., Elbaz, D., et al. 2015, A&A, 575, A74
- [101]Shakura, N. I. & Sunyaev, R. A. 1973, A&A, 24, 337
- [102]Shankar, F., Weinberg, D. H., & Miralda-Escudé, J. 2009, ApJ, 690, 20
- [103]Shenoy, S., Smith, D. J. B., Biddle, S. K., et al. 2026, MNRAS, 546, stag137
- [104]Shimwell, T. W., Hale, C. L., Best, P. N., et al. 2025, A&A, 695, A80
- [105]Shimwell, T. W., Hardcastle, M. J., Tasse, C., et al. 2026, A&A, 707, A198
- [107]Siudek, M., Pucha, R., Meczua, M., et al. 2024, A&A, 691, A308
- [108]Smith, D. J., Haskell, P., Gürkan, G., et al. 2021, A&A, 648, A6
- [109]Smith, D. J. B., Best, P. N., Duncan, K. J., et al. 2016, in SF2A-2016: Proceedings of the Annual meeting of the French Society of Astronomy and Astrophysics, ed. C. Reylé, J. Richard, L. Cambrésy, M. Deleuil, E. PÉcontal, L. Tresse, & I. Vauglin, 271–280
- [110]Smolčić, V., Delvecchio, I., Zamorani, G., et al. 2017, A&A, 602, A2
- [111]Somerville, R. S. & Davé, R. 2015, ARA&A, 53, 51
- [112]Speagle, J. S., Steinhardt, C. L., Capak, P. L., & Silverman, J. D. 2014, ApJS, 214, 15
- [113]Stalevski, M., Fritz, J., Baes, M., Nakos, T., & Popović, L. Č. 2012, MNRAS, 420, 2756
- [114]Stalevski, M., Ricci, C., Ueda, Y., et al. 2016, MNRAS, 458, 2288
- [115]Sutherland, W. & Saunders, W. 1992, MNRAS, 259, 413
- [116]Sweijen, F., van Weeren, R. J., Röttgering, H. J., et al. 2022, Nature Astronomy, 6, 350
- [117]Tacconi, L. J., Genzel, R., Saintonge, A., et al. 2018, ApJ, 853, 179
- [118]Tadhunter, C. 2016, A&AR, 24, 10
- [119]Tasse, C., Shimwell, T., Hardcastle, M. J., et al. 2021, A&A, 648, A1
- [120]Thorne, J. E., Robotham, A. S., Davies, L. J., et al. 2022, MNRAS, 509, 4940
- [121]Tremaine, S., Gebhardt, K., Bender, R., et al. 2002, ApJ, 574, 740
- [122]Van Haarlem, M. P., Wise, M. W., Gunst, A. W., et al. 2013, A&A, 556, A2
- [123]Volonteri, M. 2012, Science, 337, 544
- [124]Wall, J. V. & Jenkins, C. R. 2003, Practical Statistics for Astronomers, Vol. 3 (Cambridge University Press)
- [125]Wang, L., Gao, F., Duncan, K. J., et al. 2019, A&A, 631, A109
- [126]Wang, Y., Wang, T., Liu, D., et al. 2024, A&A, 685, A79
- [127]Werner, M. W., Roellig, T. L., Low, F. J., et al. 2004, ApJS, 154, 1
- [128]Whitaker, K. E., van Dokkum, P. G., Brammer, G., & Franx, M. 2012, ApJ, 754, L29
- [129]White, G. J., Barrufet, L., Serjeant, S., et al. 2024, MNRAS, 535, 1406
- [130]Whittam, I. H., Jarvis, M. J., Hale, C. L., et al. 2022, MNRAS, 516, 245
- [131]Williams, W. L., Hardcastle, M. J., Best, P. N., et al. 2019, A&A, 622, A2
- [132]Williams, W. L. & Röttgering, H. J. A. 2015, MNRAS, 450, 1538
- [133]Yang, G., Bouquien, M., Brandt, W. N., et al. 2022, ApJ, 927, 192














