Discovery of 178 Giant Radio Galaxies in 1059 deg$^2$ of the Rapid ASKAP Continuum Survey at 888 MHz

We report the results of a visual inspection of images of the Rapid ASKAP Continuum Survey (RACS) in search of extended radio galaxies (ERG) that reach or exceed linear sizes on the order of one Megaparsec. We searched a contiguous area of 1059deg$^2$ from RA$_{\rm J}$=20$^h$20$^m$ to 06$^h$20$^m$, and $-50^{\circ}<\rm{Dec}_J<-40^{\circ}$, which is covered by deep multi-band optical images of the Dark Energy Survey (DES), and in which previously only three ERGs larger than 1Mpc had been reported. For over 1800 radio galaxy candidates inspected, our search in optical and infrared images resulted in hosts for 1440 ERG, for which spectroscopic and photometric redshifts from various references were used to convert their largest angular size (LAS) to projected linear size (LLS). This resulted in 178 newly discovered giant radio sources (GRS) with LLS$>$1Mpc, of which 18 exceed 2Mpc and the largest one is 3.4Mpc. Their redshifts range from 0.02 to $\sim$2.0, but only 10 of the 178 new GRS have spectroscopic redshifts. For the 146 host galaxies the median $r$-band magnitude and redshift are 20.9 and 0.64, while for the 32 quasars or candidates these are 19.7 and 0.75. Merging the six most recent large compilations of GRS results in 458 GRS larger than 1Mpc, so we were able to increase this number by $\sim39\%$ to now 636.


Introduction
Giant radio galaxies (GRG) were first discovered in 1974 [1] and originally defined as those for which their projected linear size (LLS) exceeds 1 Mpc, at a time when the Hubble constant was assumed to be 0 = 50 km s −1 Mpc −1 . Today 0 is accepted to be near 70 km s −1 Mpc −1 so that objects of the same angular size now have a linear size smaller by a factor of ∼1.4. Thus the first GRG found (3C 236), originally assigned an LLS of 5.7 Mpc, is now considered to extend over 4.1 Mpc. This change in the adopted value of 0 has led more recent authors to adopt an LLS of 0.7 Mpc as a lower limit for considering a radio galaxy a "giant" one. However, there is no physical reason for adopting either of these thresholds, and in the present paper, we focus on only those larger than 1 Mpc.
In 1983 [2] the first two giant radio quasars (GRQ) were found, [HB89] 1146-037 of LLS=1.06 Mpc, at a redshift of z=0.341, and [HB89] 1429+160 with 1.37 Mpc at z=1.016, and these authors already raised the issue that inverse Compton losses suffered by the synchrotron-emitting electrons by scattering the cosmic microwave background (CMB) photons at 1, and the denser Universe at that cosmic epoch, should not let these GRQ expand to the same extent as in the local Universe. Since then many other GRQs have been found, and currently they constitute about 18% of all known giant radio sources (GRS), see also Sect. 4.1. Unless otherwise noted, here we use the terms GRS, GRG, and GRQ for sources with LLS>1 Mpc.
Despite an intense search for further and larger examples of these rare objects, only two GRGs larger than 3C 236 have been found so far, namely J1420-0545 of 4.7 Mpc [3] and J0931+3204 of 4.3 Mpc [4], the latter being the largest GRQ currently known. A third GRG of supposedly 4.45 Mpc, listed as J1234+5318 in the latest GRS compilation by [5], was originally published by [6], but was later shown to have a lower-redshift host on the basis of a deeper image from the Low Frequency Array arXiv:2111.08807v1 [astro-ph.GA] 16 Nov 2021 from −40 • to −50 • . Knowing that radio galaxies tend to be found up to high redshifts, implying very faint hosts, we further limit ourselves to the high Galactic latitude area from RA=20 ℎ 20 to 06 ℎ 20 (a total of 1059 deg 2 ) as it is covered by DES, providing deep optical , , -band images and photometric redshifts. The literature compilation by [5] lists only three GRS in this area.
Small parts of this area had been surveyed previously in radio, and the corresponding images had been inspected by one of us (H.A.) for the presence of GRS. Firstly, the Australia Telescope ESO Slice Project (ATESP, [32]), covering RA J =22 ℎ 32 -22 ℎ 57 and 23 ℎ 31 -01 ℎ 23 in the range −39.5 • < Dec J < −40.4 • , overlaps for ∼17 deg 2 with our search area. Secondly, a circular area of ∼4.5 deg 2 had been imaged with the Australia Telescope Compact Array (ATCA) at 1.4 GHz as the Phoenix Deep Field (PDS) [33,34] centered on RA J ,Dec J =01 ℎ 13 36 ,−45.7 • . Thirdly, an area of ∼ 40 deg 2 (RA J =20 ℎ 05 -22 ℎ 26 ,−48.1 • < Dec J < −50 • ) overlaps with our search region and has been imaged with ASKAP as part of the EMU Pilot Survey [17] at the same angular resolution as RACS, but to a 1-noise level 10 times lower (∼ 26 Jy beam −1 ). We include in our results only those GRS that can be recognized on RACS images.

Images used
For our visual inspection we used the RACS images for Stokes I data release 1, available in full resolution from https://www.atnf.csiro.au/research/RACS/RACS_I1. The beam size and shape of the latter images vary with sky position [35], such that the beam major axes range from 15 to 25 . The RACS Stokes I data release 1 images are also provided as "CRACS", convolved to a common resolution of 25 at the URL https://www.atnf.csiro.au/research/RACS/CRACS_I2, which were used to prepare the source catalogue in [35]. Both versions are in J2000 coordinates and pixel intensities are calibrated in Jy beam −1 . For our search for GRS we preferred the full-resolution images to better constrain the most likely host position, but occasionally consulted CRACS to confirm the presence of diffuse emission. The astrometric precision of RACS is better than 1 [18,35] and thus does not affect the reliability of optical identifications. It is rather the angular resolution that causes an increased rms of the source position which amounts to ∼ 2.7 for a faint source of S/N∼ 5 [18].
During a summer internship in 2012 at Univ. of Guanajuato one of us (E.F.J.A.), together with R.F. Maldonado Sánchez [36] had inspected all of the SUMSS images, logging the positions of ∼5000 potentially extended radio sources. Since then, a fraction of these had been followed up by one of us (H.A.) to either optically identify their host, or discard them as unrelated sources. However, the low angular resolution of SUMSS (45 ), together with the limited sensitivity of the optical Digitized Sky Survey [37] often made this task a guesswork, and only the advent of RACS in radio and DES in the optical promised a solution. Thus, in the first round of our search for GRS, we inspected RACS images at a few hundred previously logged positions in our search area. We used A ( [38], http://aladin.u-strasbg.fr/aladin.gml) to display these images, adjusting the contrast such that the noise floor could always be recognized. Whenever the source structure in RACS appeared as two or more unrelated sources, they were discarded, and otherwise the , , -composites of optical images from DES were consulted to find the most likely host, and its position was recorded for later retrieval of complementary data. Over 260 radio sources spotted in SUMSS were thus identified with the help of RACS.
In the second round, we systematically screened the full area of 1059 deg 2 in the full-resolution RACS, marking on it with symbols those sources already identified in the first round, as well as apparently extended sources that had been discarded previously, so as to avoid repeating the identification process for these.
Given that our aim was to find, apart from GRS also those with an appreciable size of LLS 0.5 Mpc, we tried to log all possible sources with a LAS 1.0 , since for our adopted cosmology this is the minimum angular size a standard ruler of 0.5 Mpc would appear if it had a redshift in the range of ∼1.4-1.8 (see e.g. Fig. 30 on p. 1326 of [39] and our − diagram in Section 4.1). We also included, though less complete for LAS 1.5 , bent-tailed radio sources like wide-angle tailed (WAT) and narrow-angle tailed (NAT) sources, since these may serve as indicators of the presence of a cluster of galaxies. We proceeded from RA=20 ℎ 20 eastward by scanning portions of 1 h wide in RA (or ∼ 100 deg 2 at the chosen declination), displaying about 1 deg 2 per screen at a time and logging between 1 and 1.5 candidates per deg 2 , which took about 16 h for the entire 1059 deg 2 . This resulted in a list of approximate centre positions of ∼1330 candidate extended radio sources.

Scrutinizing the Candidates: Host Identification and Radio Morphology
The positions of all the ∼1330 candidates selected in the previous step were then displayed again in A and the likely host was searched on DES DR1 , , −band color composites [22]. DES DR1 has a median delivered point-spread function of 1.12, 0.96, and 0.88 in the , , and -bands, an astrometric precision of 0.151 , and limiting magnitudes of 24.28, 23.95, and 23.34 mag in , , and -bands for objects with S/N=10 and a 2 -diameter aperture. We also used mid-infrared (MIR) images from unWISE [24], colored according to the magnitude difference between the two lowest-wavelength WISE bands, 1 − 2. For sources with an obvious central radio core, finding the host object was easy, both for the edge-darkened Fanaroff-Riley [40] type I (or FR I) sources with a radio brightness peak close to the optical host, and for the edge-brightened FR II sources with a radio core. The problem arises for sources with widely separated radio components, and without an obvious central component in between them. If the central component had no optical or MIR counterpart, the pair of outer components was discarded as a genuine RG. However, if one of the outer components had a convincing host by itself and its radio extent was larger than ∼ 1 , it was recorded as a separate extended RG. If the central component had a clear optical or MIR host, the candidate was discarded if the outer components had an obvious counterpart, unless their radio structure showed clear indications of being connected (e.g. radio bridges or trails pointing at each other) and the optical/MIR objects in the lobes did not show evidence for being active galactic nuclei (AGN). When felt necessary, RACS radio contours were drawn interactively in A and overlaid on DES and/or unWISE images. The unWISE images were especially helpful in suggesting a likely host for sources with widely separated lobes without an obvious radio nucleus and without a prominent galaxy or quasar near their geometrical center. Occasionally the unWISE images helped us to decide between two optical counterparts very close to each other. In cases of lobes with very different radio brightness, a higher probability was assigned to hosts nearer to the brighter lobe (cf. [41]). Whenever the optical hosts were too faint or uncatalogued in DES DR1 we consulted the Dark Energy Spectroscopic Instrument (DESI, [42]) DR9 images at https://www.legacysurvey.org and usually found the corresponding DESI object and its -band magnitude. In case of doubt about the likely host we generally chose the brighter or lower-redshift host, such that the derived LLS should serve as a lower limit.
During this inspection we also measured the LAS for each accepted candidate, and classified their radio morphology into one or more types as listed in Table 1, appended by a "?" symbol to indicate uncertainty. We avoided any systematic overestimate of the LAS; e.g., for FR II sources with bright hotspots we did not measure the LAS between opposite 3-contours, but rather between the centres of the outer half-circles of these contours of each hotspot. Only for rather faint or diffuse lobes we measured out to about the 3-contours (see [46] sect. 2, for a discussion of LAS measures). For bent-tailed sources we measured the LAS along a straight line between the most separate diametrically opposite emission regions of the source, and not along their curved emission ridges. Table 1: List of radio-morphological types we assigned to our sources.

FRI
radio brightness is "edge-darkened", fading away with distance from the host FRII classical double with "edge-brightened" outer lobes FRI/II source shows characteristics of both FR I and FR II FRIIncor widely separated double with no obvious radio core at the optical host FRIIcoredom radio nucleus very strong compared to the lobes FRIInaked no evidence for radio trails or bridges: lobes are unresolved hotspots FRIIplume(s) diffuse emission regions displaced sideways from source major axis FRIIrelic lobes slightly diluted FRIIremn lobes very diluted/inflated and of low surface brightness DDRG double-double or "restarted" radio galaxy: an inner and outer pair of lobes hymor hybrid morphology: one arm or lobe is of type FR I, the other type FR II WAT wide-angle tailed RG with outer lobes bent in the same direction, C-shaped) NAT narrow-angle tailed RG: host galaxy is located at one end of radio emission precess Z-or S-symmetry, suggesting precession of the radio jet axis asym length or flux ratio of opposite lobes is near two or more Often the likely optical hosts, or objects superposed on the suspected lobes, appeared stellar, either by visual impression or confirmed by the stellarity index provided in the DES DR1 catalogue. To distinguish between stars and quasars, we made use of both the measurements of parallax and proper motion in the Global Astrometric Interferometer for Astrophysics (Gaia) early data release 3 (Gaia EDR3, [43]) which should be consistent with zero for quasars, as well as their 1 − 2 and 2 − 3 colors in the AllWISE catalogue [23,44]. Based on this information we discarded the likely stars as possible radio galaxy hosts, or else labelled the host as quasar candidate or "Qc" in Table 1. We also took note of any optical peculiarities of the host like stellar/QSO or spiral morphology, presence of optical shells, interactions or perturbations, etc.
Obvious relic-type radio sources in clusters of galaxies, not associated with any single cluster galaxy, as well as nearby late-type spiral galaxies were discarded as radio galaxy candidates. We maintained a few possible candidates for so-called spiral double radio galaxy AGN (SDRAGN, [45]), though all of these had linear sizes 1 Mpc.
Scrutinizing these ∼1330 candidates required ∼ 40 net hours, and led us to discard ∼30% of them, resulting in a list of precise host positions, LAS measures, and a crude radio-morphological classification for ∼1000 extended radio sources, in addition to the ∼260 objects initially suspected on SUMSS images and confirmed by us in RACS.

Culling Complementary Data for the Radio Galaxy Hosts
We first queried the positions of the ∼1300 newly discovered ERG hosts from RACS in the NASA/IPAC Extragalactic Database (NED, ned.ipac.caltech.edu) to find spectroscopic redshifts for 68 of them, mostly among the optically brightest hosts. We then used the VizieR service at the Strasbourg astronomical Data Center (CDS, [47]) at the URL https://vizier.u-strasbg.fr/viz-bin/VizieR to search for host names and complementary data for these hosts. For all hosts we searched names (if available) from the 2MASX catalogue [48] (CDS VII/233), their exact positions, names and -band magnitudes, as well as stellarity index from DES DR1 [22] (CDS I/357), and the WISEA names and magnitudes in all four WISE bands from AllWISE [23] (CDS II/328), as well as W1 and W2 magnitudes from the deeper CatWISE2020 [49] (CDS II/365).

Photometric Redshifts and Conversion from Angular to Linear Size
We used several different sources outside VizieR for photometric redshifts ( ℎ ). For 2MASX galaxies we searched the 2MASS Photometric Redshift catalogue (2MPZ, [50]), and the hosts found in AllWISE [23] were searched in the WISE-Supercosmos photometric redshift catalogue [51] offering 20.4 and 57.9 million objects in the "main" and "reject" catalogues, respectively. Previous experience showed that the ℎ from the "reject" catalogue are mostly sound and were used as well.
The ℎ values offered by DESI DR9 [52] were extracted from the DESI Legacy Survey DR9 through the Notebook Server of the NOIRLab Data Lab Query Interface at https://datalab.noirlab.edu/ query.php. Since DESI DR9 data are stored in different tables, we cross-matched the main photometry table (ls_dr9.tractor) and the one with photometric redshifts (ls_dr9.photo_z) via the objectID searchig a radius of 2 from the host position, and requesting object coordinates, , , -magnitudes and median and mean ℎ . For a few objects the ℎ values from DESI DR9 were obtained directly at viewer.legacysurvey.org.
The DESI DR9 ℎ catalogue also lists the standard deviation (which we call Δ here) of the the ℎ probability distribution function. We found the latter to be Δ ∼0.01 below ℎ ∼0.4, gradually rising to Δ ∼0.05 near ℎ ∼0.8 and to Δ ∼0.15 for ℎ 1.0. For 23 (13%) of our new 178 GRS in Table 1 the Δ values exceed 0.17 and reach up to 0.6, but these objects are predominantly QSO candidates (for which we use additional ℎ estimates, see below) and some very faint galaxies. For 41 GRS up to a redshift of ∼0.6 we had ℎ values from both DESI DR9 and [51]. The difference (DESI DR9 minus [51]) has a mean of 0.03 and a standard deviation of 0.09. Averaging the latter values should further reduce their uncertainty. Although the ℎ values are considerably more uncertain at higher redshift, the LLS is much less sentitive to the redshift for a given LAS at redshifts between 1 and 2 (see our Fig. 6).
It is known that for QSOs it is notoriously difficult to estimate redshifts, and a large fraction of starlike QSO candidates among our hosts have clearly underestimated ℎ values in DES DR1. For some of them we found estimated redshifts in [53], and, in addition, for all our quasar candidates we use MIR colors 1 − 2 and 2 − 3 from AllWISE where available and consulted Fig. 2 of [54] to estimate their redshift. We combined all the redshifts found, and adopted a reasonable average of these if more than one was available, and provide these in Table 1.
In order to find the physically largest objects in our sample, we converted the sources' LAS to their LLS, using a script adapted from the Cosmology Calculator ( [55], http://www.astro.ucla.edu/~wright/ CosmoCalc.html), and based on the adopted redshift ( or ℎ ) and cosmological parameters as listed at the end of Sect. 1. We stress that the LLS is always a lower limit to the 3-D extent of the source and we make no inference about its orientation with respect to the line of sight as done e.g. in [56].

Determination of Radio Flux Densities and Luminosities
Cutouts of at least 2×LAS on a side were obtained for each source from both RACS and CRACS using a script as suggested at http://alasky.u-strasbg.fr/hips-image-services/hips2fits. These were then passed through the breizorro masking program (see https://github.com/ratt-ru/breizorro) which creates a binarized image assigning a value of 1 to every pixel brighter than n , where is the noise level, and values of 0 elsewhere. By trial and error we found that a 3-threshold worked best.
These binarized images were then run through the python matplotlib program which calculates contours around the areas assigned a value of 1, and then colors the boundaries of the ten contours containing the largest areas. These are usually the ones containing the real source confines, plus occasional unrelated sources, but sometimes an isolated central nuclear source could have a contour area smaller than the ten largest.
The contour areas considered to contain the emission from the extended radio source are then selected manually (by clicking the mouse on these) to be run through the source parameter analysis. Usually, we selected between two such areas (e.g. for a source with only two lobes and no radio core) and up to about five such areas (e.g. for a source with a radio core and two separated emission areas for each lobe on opposite sides of the host), but occasionally the presence of a radio core and emission bridge connecting the lobes leads to the source being contained in a single such area.
The pixels making up the individual contour boundaries, along with the x,y positions inside contours that were selected in the previous operation are then parsed to the script that calculates source parameters. It first multiplies the original image by the breizorro mask. For a selected contour it then sums up the brightness values of all pixels inside a contour and finally normalizes by the number of pixels per beam solid angle to obtain the flux density within a particular contour. To figure out which pixels lie inside a contour the python package Shapely (see https://github.com/Toblerity/Shapely) was used. A contour equates to a polygon in Shapely which can generate a MultiPolygon from a collection of polygons. Shapely is also used to find out which contours are to be used in the analysis by determining which contours contain the x,y positions determined by the mouse clicks in the previous step. This allows one to calculate total flux densities, largest angular size, etc., which in fact we use to estimate magnetic field strength and energy densities for the individual lobes based on the assumption of equipartition between particle and magnetic field energies (see [57]). However, for the present paper we limit ourselves to report only the total flux of each source, as well as the position angle of the radio source's major axis.
As a final step we compared the total flux densities obtained from RACS and CRACS to select the most adequate total flux, avoiding as much as possible the contributions of unrelated sources, and these are listed as 888 in Table 1. Rest-frame spectral radio powers at 888 MHz were then obtained via where ( ) is the luminosity distance for the adopted cosmology and (1 + ) −(1+ ) is the -correction to convert from observed to rest frequency and allows for the stretching of the spectrum with respect to the receiver bandwidth. In the absence of measured spectral indices we assume = −0.8, which is slightly steeper than the average spectrum of radio galaxies of = −0.7 [58], given that we are dealing mostly with objects dominated by emission from diffuse radio lobes. We list the decimal logarithm of 888 in Table 1.

Results
Having obtained the LLS for all ∼1440 accepted GRS in our search area, we ranked them in decreasing order of LLS and found ∼210 to exceed a size of 1 Mpc. For these, overlays of radio contours on DES , , −band composites were prepared and scrutinized again. As a consequence, about 30 of these were discarded for being likely separate sources, and 178 objects were accepted as new GRS, of which we mark 14 as likely candidates in Table 1 due to uncertainties in either their LAS, ℎ , or their host ID.
As a sideproduct our list of smaller radio galaxies contains ∼300 ERGs with LLS=0.7-1.0 Mpc, plus ∼400 with LLS=0.5-0.7 Mpc, and 550 smaller ones. All these require further scrutinizing and will be published elsewhere. However, these numbers serve to demonstrate the significant effort necessary to find physically large sources among objects selected purely by their angular size. For example, of the 558 surviving radio galaxies with LAS> 2.0 , only 32% turned out to be GRS larger than 1 Mpc, and of all the 1038 with LAS> 1.4 , only 46% turned out to be ERGs with LLS> 0.7 Mpc. Given that from the candidates originally selected on the RACS images, about 30% had to be discarded upon closer inspection, this implies that to find a single GRS larger than 1 Mpc one has to inspect on average 4.4 candidates larger than 2.0 and to find a single ERG larger than 0.7 Mpc one has to inspect on average 3 candidates larger than 1.4 .

The List of 178 new and 3 known GRS in the Search Area
In what follows we shall limit ourselves to the sample of 178 new and 3 previously published GRS larger than 1 Mpc. Their sky distribution is plotted in Figure 1. In about 20 cases the host could not be determined with certainty, and either the brightest in optical or MIR was chosen, such that its redshift and linear size can be considered a lower limit. The basic properties of these 181 GRS are listed in Table 1, the columns of which are as follows: (1) name of the GRS derived from truncation of the sexagesimal equatorial J2000 coordinates of the optical host, appended with a "C" if the GRS is considered a candidate; for each object we indicate in which survey it was noticed first: P = previously published (3 objects), A = ATESP (3 objects), E = EMU Pilot Survey (8 objects), S = SUMSS (45 objects), R = RACS (122 objects); (2,3) RA and Dec (J2000) of the optical host in decimal degrees; (4) the largest angular size of the radio emission; (5) the adopted redshift and type: s for spectroscopic, p for photometric, and e for estimated, with question marks indicating uncertainty due to inconsistent ℎ values found, see also Sect. 2.4; (6) references used to select the adopted redshift: 1 [ [69]; (7) largest linear size of the radio emission; (8,9) name and type of the host object, where G=galaxy, GP=galaxy pair, GQ=galaxy or QSO, Q=QSO, Qc=QSO candidate, with a "?" sign indicating uncertainty; (10) -band magnitude from DES; (11) integrated flux density from RACS or CRACS (see Sect. 2.5); (12) decimal logarithm of spectral radio power at 888 MHz; (13) position angle of the major axis of the radio source, measured from North through East; (14) radio morphology according to Table 1. 150 120 90 60 30 0 330 300 270 240 210 0°7 5°D EC (J2000) Figure 1. All-sky Aitoff projection of the distribution of our newly discovered GRS in red, together with the GRS from six major published GRS lists, namely in black from [5] in orange from [15], in dark blue from [59], in dark green from [60], in dark magenta from [16], and in dark red from [46]. The thick blue line shows the area within | | ≤ 2 • from the Galactic plane.
Host redshifts range from 0.02 to ∼2.0, but only 10 of the 178 new GRS have spectroscopic redshifts. For the 146 host galaxies the median -band magnitude and redshift are 20.9 and 0.64, while for the 32 QSOs or candidates these are 19.7 and 0.75. The lowest total radio flux densities are 5.0 mJy with 10 objects fainter than 10 mJy. The lobe brightness of various GRS barely exceeds 1 mJy beam −1 . We compare the properties of our new sample with those of six previous large compilations of GRS in Sect. 4.1.

The Variety of Source Morphologies, Sizes and Radio Luminosities
Within the classification scheme we used (see Table 1) we find that the vast majority of the newly found GRS are FR II (166 or 93%), of which 30 do not have a radio core detected (i.e. S 888, 0.5 mJy), four have virtually unresolved lobes (tagged as "naked" Table 1. List of 181 giant radio sources in our search area (see text for column descriptions). The table will also be available at http://cdsarc.u-strasbg.fr/viz-bin/cat/J/other/Galax/9.xx  Table 1), and one (J0023-4732) shares both of the latter properties. Apart from these clear FR IIs we found 11 (6.2%) mixed types I/II, and only a single one (J0422-4518C) is of type FR I and classified as a candidate WAT with very bent, diffuse, and low surface brightness lobes. We tagged 21 of the GRS as "remnant" and 34 as "relic" sources which refers to the degree of diffuseness of the lobes. These will be discussed in Sect. 3.2.4. Since we use the tag "relic" only in connection with the tag FR II, there should not be any confusion with "relic" radio sources found in clusters of galaxies and not associated with individual host galaxies. Nine FR IIs were labelled as "asymmetric" implying that the most likely host was found more than ∼2 times closer to one of the outer lobes than the other. This is not unusual given that the ratio of the larger to smaller distance of hotspots of FR IIs to their host can reach and exceed five [41]. Apart from the known double-double GRS (J0116-4722) only one other noteworthy example (J2240-4724) of this type of source was found.
No clear examples of hybrid morphology sources (see e.g. [70]) were found except for possibly J0138-4116 and J0433-4948, and no single example of the so-called "Odd radio circles" (ORC, [71]) was found in our inspection of RACS. While the original FR classification [40] suggested a clear-cut separation of these types in radio luminosity, it was later shown that the dividing radio luminosity not only increases with optical luminosity [72,73] but also that there is a large overlap in radio luminosities between these types (e.g. [74,75]). In our new sample of GRS we can see only a marginal trend in the median 888 from 25.0 for the 12 FR I or FR I/II sources, 25.7 for the 51 FRIIrelic or remnant sources, and 26.0 for the 76 clear FR II sources. We defer any discussion of this so-called "Radio-HR" [76] or Owen-Ledlow diagram [73] to a later paper, including smaller sources with a more appreciable fraction of FR I types.
As noted in Sect. 2.5 our algorithm for flux integration includes an estimate of equipartition parameters, assuming = 1 for the proton-to-electron energy ratio, a filling factor of 1, a uniform source magnetic field in the plane of the sky, and = −0.8 for the source's spectral index [57]. For the source lobes, excluding the core emission, we find magnetic fields, , of ∼0.9-3.0 G (0.09-0.3 nT) independent of using RACS or CRACS, and minimum energy densities in particles and fields, , of ∼1.4-17×10 −14 J m −3 for RACS, and 0.8-14×10 −14 J m −3 for CRACS. The minimum total energies range from 1.4×10 50 J to 1.7×10 54 J for RACS and from 2×10 50 J to 2.2×10 54 J for CRACS. The median values are comparable to those found for 3C 236 by [1]. We defer a more detailed discussion of this to a later paper.
In what follows we select a few examples of the variety of GRS we found by means of overlays of CRACS contours on , , -band composites from DES DR1. We recommend viewing these images at an amplified scale to appreciate the optical host and its environment in more detail. there is only one GRS with both > 1 and LLS>2 Mpc, the QSO FBQS J0204-0944 at =1.004 with LLS=2.1 Mpc, making our new find of J0500-4242 the largest GRQ known at > 1. We classify the latter as "FR IInaked" since the outer lobes are barely resolved with little or no indication of a radio tail or bridge pointing towards the host. Future higher-resolution and sensitivity radio observations should show some extended structure around these hotspots. Both these GRS are shown in Fig. 2. WATs tend to occur exclusively in clusters of galaxies and are predominantly hosted by the brightest cluster member. Although they are generally considered of FR I type, they appear as either jet-dominated FR I types or lobe-dominated FR II types or a mixture thereof. There are no clear-cut examples of giant WATs in our sample, and the object that comes closest to this morphology is J0532-4118 at ℎ =0.72 with LLS∼1.2 Mpc, hosted by the = 20.9 mag galaxy DES J053249.90-411856.8 (right panel of Fig.  3). The DES image shows indications of a distant cluster, in fact, listed as ID 4063500118 by [78], and as WaZP DES YR1 J053249.9-411856 in [79], with the GRG host being the brightest member in -band and the third-brightest in -band (the brightest in is located 33 N of the WAT host). WAT type sources are not expected to grow to sizes larger than about a Mpc, since they tend to be located near the density peak of the intracluster medium of their host cluster. Thus, J0532-4118 may be exceptional, also for its high redshift. It is difficult to identify the largest WAT reported in literature given that this morphological type is rarely mentioned in publications on GRS, but e.g. J2233+1315 hosted by the =15.2 mag, ℎ =0.093 galaxy 2MASX J22330133+1315019 which is the brightest of cluster WHL J223301.3+131503 extends over 14.7 or ∼1.7 Mpc [80] 1 , or else the FR I source J1049+5513 listed in [84] and hosted by 2MASX J10490732+5513153 at =0.1262 can be seen to extend over 11.5 or 1.56 Mpc in the Lockman Hole LoTSS Deep Field [85]. Since the radio morphology of the latter is quite likely foreshortened by projection, surveys like LoTSS or EMU should reveal even larger WATs that are oriented closer to the plane of the sky.

Giant Radio Sources associated with Clusters of Galaxies
Although the visual classification of the environment of the GRS hosts in our sample is beyond the scope of this work, we cross-matched the 181 GRS of Table 1 with major cluster catalogues and found the following coincidences.
The host galaxy of J0406-4544 is the brightest and dominant member of cluster WHY J040607.9-454451 at ℎ =0.315 [86]. The radio source is a symmetric and straight FR II of projected size 1.8 Mpc, with its radio axis oriented perpendicular to the optical major axis of the outer halo of the host galaxy, but ∼ 25 • off from perpendicular for the innermost isophotes, suggesting the presence of isophote twist. The cluster is rather poor and scattered, and as yet unreported as X-ray emitter.
A cross-match of our 181 GRG hosts with 2.52 million member galaxies of 60542 clusters of the WaZP cluster catalogue [79] yielded 14 matches, of which two are unlikely cluster members, seven are brightest members (one is the above-mentioned J0532-4118), and another five are lower-ranked members. The six other brightest cluster members (J0020-4625, J0105-4505, J0213-4744, J0429-4517, J0508-4737, and J2250-4751) are all hosts of fairly straight FR II type GRS, which is considered rare. All five lower-ranked cluster member GRS hosts are also FR IIs, and special mention deserves the GRS J0150-4507, which is the 2nd-brightest member of the very rich and filamentary cluster WaZP DES YR1 J015035.3-451112 at ℎ ∼0.3, also detected via the Syunyaev-Zeldovich (SZ) effect as PSZ2 G273.69-68.38 and SPT-CL J0150-4511. The GRS host lies in a filament ∼ 7 (∼1.9 Mpc) NW of the cluster centre, with its radio source axis oriented perpendicular to the galaxy filament, similar to the trend seen by [26,27].
In conclusion, we find 13 of 181 GRGs (∼ 7%) to be the brightest cluster members and 9 (∼ 5%) are lower-ranked members of which J0320-4515 had already been reported as a member of Abell S0345. Similar fractions of GRS in clusters were reported by [15,59].

Remnant Radio Sources
There is as yet no consensus about the criteria to classify radio galaxies as of "remnant" type [87][88][89], especially about whether to consider the presence of a radio core. Here we have made a crude attempt to assign the tag "remnant" to objects with very diffuse and little collimated lobes, and "relic" to objects with moderate indications of the latter, suggesting the absence, or at least relative faintness, of any hotspots in these lobes. Note that these tags were given as a proxy for the suspected age of the lobes, independently of the presence of a radio core, because (a) the presence of a radio core does not seem to have any relation with the age of the lobes and (b) could be an indication of restarting activity after a previous cycle of "feeding" a pair of lobes with relativistic particles. Thus the absence of a radio core is by no means an indicator of the final "death" of the radio activity. Fig. 4 shows two examples: at left is J0133-4655, hosted by the =20.3 mag galaxy DES J013320.80-465501.4 at ℎ =0.65 with LLS=1.04 Mpc. The galaxy near the centre of the South (S) lobe is barely detected in WISE and is considered a chance superposition. The fact that the S lobe is closer to the host and ∼1.4 times more luminous than the North lobe, is consistent with statistical expectation [41]. At right is J0150-4634 (DES J015045. 33-463459.8) at ℎ =0.88, of 1.15 Mpc. The extended emission near the host is oriented at an angle of ∼ 30 • from the axis connecting the outer lobes, suggesting the central engine is both precessing, as well as restarting its radio activity.

Discussion
The three previously reported GRS in our search area are J0320-4515 of 1.8 Mpc [90], and J0116-4722 and J0213-4744 of 1.8 and 1.3 Mpc, respectively, published by [91]. Their redshifts range from 0.063 to 0.146 and their total 843-MHz flux densities from SUMSS range from 2.1 to 6.7 Jy. Our new sample of GRS at high Galactic latitude (| | 25 • ) between Dec −40 • and −50 • has hosts in the redshift range from 0.021 to ∼2.0 with total 888-MHz flux densities as low as 5.0 mJy and a median of ∼40 mJy.

Comparison of our Sample with previously published GRS
The literature compilation of GRGs by [5] contains 223 objects larger than 1 Mpc, and after that four further large samples of GRGs and one of GRQs were published. These contain (after correction of their LAS measures and exclusion of a few spurious objects consisting of separate sources) 68, 48, 18 and 25 GRS from [15,16,59,60] and 76 GRQs from [46]. This makes a total of 458 GRS published, neglecting here a few others that were reported in much smaller numbers in various papers, sometimes not even mentioning their GRS nature. Our new list of 178 GRS thus not only increases this number by ∼39% from 458 to 636, but is also the largest yet published individual list of newly found GRS. Table 2 compares the six published samples (first six rows) with our new sample (last row). Columns are (1) the surveys used and reference, (2) their observing frequency, (3) the number of GRS >1 Mpc published, (4) the median LAS, (5) lowest and median total flux density of the GRS, (6) lowest and median decimal logarithm of the radio spectral power at the frequency listed in col. 2, (7) median redshift of the hosts, (8) the fraction of hosts with spectroscopic redshift, and the fraction of hosts that are QSOs or candidates, and (9) the median LLS. For comparison, for a spectral index of = −0.8, 150 is larger by 0.62 and 1400 is lower by 0.15 than our 888 .  Table 2 shows, not surprisingly, that the literature compilation by [5] has the largest median LAS and LLS, while ours is the one with the second-smallest median LAS, only surpassed by [46] who performed a dedicated search for extended radio emission from spectroscopic QSOs from SDSS DR14Q. The latter is also the sample with the highest median host redshifts, followed by our present sample. The sample with the lowest minimum and median flux density is from ASKAP [16] with its much higher sensitivity. In Fig. 5 we present the distribution of spectral radio power at 888 MHz from our new GRS sample drawn from RACS, as function of LLS in the left panel, and as function of host redshift in the right panel. The left panel, also called the − diagram is similar to that in Fig. 5 of [5] except for our sample being smaller, such that any trend is washed out by the large dispersion in P 888 . The right panel shows a clear trend of P 888 rising with redshift for galaxies, likely due to Malmquist bias. The panel is similar to that of Fig. 4 of [15] except for the fact that our sample fills in a lack of galaxies at ∼1 which is due to the deeper DES DR1 images used by us compared to the SDSS images used by [15]. The dedicated search for GRQs among ∼526,000 SDSS DR14 QSOs by [46] had shown a weak trend for radio power to increase with redshift for GRQs with z 0.5 out to z∼2.5. Such a trend cannot be seen in our much smaller sample due to the lack of known spectroscopic QSOs in our search area.
In Fig. 6 we show the location of our newly found 178 GRS in the − diagram, compared to the previously published 458 GRS, with reference lines drawn for "standard rulers" of four different sizes. The fact that no GRS larger than 5 Mpc has yet been found suggests this may be a physical limit. The diagram also shows that GRS become much more numerous for smaller LLS, but for LLS<1 Mpc this has not been quantified in literature as yet. Fig. 7 shows the very rapid decrease of the number of GRS larger than a given LLS as function of LLS. The data in Fig. 7 actually show a flatter slope of this relation of about −3.3 below ∼2 Mpc, while above that size the slope is steeper than −4. Below 1 Mpc this slope becomes flatter: the number of ∼300 GRS found by us with LLS=0.7-1.0 Mpc, plus ∼400 of 0.5-0.7 Mpc, suggest slopes of −2.8 and −1.8, respectively, for these latter size ranges. For the literature compilation by [5] with 136 GRS of 0.7< LLS < 1.0 Mpc the slope in this LLS range is much flatter (−1.4), indicating that little attention had been paid to these smaller sources in literature prior to this compilation.

Multi-wavelength Counterparts of GRS hosts from RACS
The use of the deep multi-band optical images from the Dark Energy Survey (DES DR1 [22]) brings a clear improvement over the use of previous shallower optical surveys like the Digitized Sky Survey (DSS2, [37]) or SkyMapper (SMSS) DR1 [21]. In fact, only 86 (43%) of the 178 hosts of the new GRGs are listed in the USNO B1.0 catalogue drawn from the DSS images [92], and only 35 (19.6%) are found in the SMSS DR1 catalogue. In contrast, 175 of the 178 new GRG hosts are found in the DES DR1 catalogue [22] and of the remaining three, J0600-4908 and J2331-4928 are listed in the DESI DR9 catalogue ( [42], https://www.legacysurvey.org), and J0131-4901 is recognized on DESI DR9 images but not catalogued, so we estimated its redshift.
On the other hand, the WISE-based object catalogues are also very efficient in detecting GRG hosts: 163 (91.6%) of the 178 new GRG hosts are detected in AllWISE [23], and 169 (96%) are detected in CatWISE2020 [49].

Comparison of RACS with SUMSS
The motivation for our effort to identify over 1400 extended radio sources on RACS was its three times better angular resolution of ∼ 15 compared to that of SUMSS (45 ). However, we found that for sources with very large LAS of 15 and very diffuse and low surface brightness lobes, SUMSS often traces these sources with higher fidelity. In Fig. 8 we show one example of this by comparing the CRACS and SUMSS images of J0138-4231, the nearby galaxy NGC 641, where the RACS image is affected by streaks parallel to the inner jets and the SW lobe of the source, possibly due to the fact that ASKAP has a shortest baseline of 22 m [14] and RACS was observed with a short exposure time of 15 min, while SUMSS [19] required 12-h integrations for the East-West array with virtually zero minimum spacings. We added contours from GLEAM [77] on top of the SUMSS image, confirming the reality of the lobes. Another example for which we found the SUMSS image of better quality than RACS is J2226-4316, a 13.8 wide GRS of LLS=1.4 Mpc hosted by the =0.0931 galaxy 2MASX J22263358-4316356. The good sensitivity of SUMSS for low surface brightness extended sources is also borne out by the fact that 45 (25%) of the 178 newly found GRS were first seen on SUMSS images, only that the SUMSS angular resolution did not allow their secure optical identification.

Summary and Future Prospects
We described the results of our search for radio sources extended by more than ∼ 1.4 on images of the RACS 888-MHz survey in a contiguous high-Galactic latitude (| | > 12.5 • ) area of 1059 deg 2 between declinations of −50 • and −40 • . A subsequent search for the optical host galaxies or quasars on color images of the Dark Energy Survey DR1 resulted in the identification of ∼1000 such sources, and further use of available spectroscopic or photometric redshifts revealed that 178 of these are giant radio sources (GRS) larger than 1 Mpc in projection and previously unreported in literature. We thus increased the number of known GRS by ∼39% from 458 to 636. To our knowledge we present the largest sample of newly discovered GRS exceeding 1 Mpc in a single publication. We also demonstrated that the depth and angular resolution of RACS, together with deep optical/IR images, is perfectly suited for the identification of large RGs.
About 18% of the 178 new GRS are hosted by QSOs or QSO candidates which is similar to previously published GRS lists (except for the dedicated search for GRQs by [46]). At least 10% of all GRS are hosted by brightest or other cluster galaxies. This fraction is a lower limit, since over half of our GRS lie at distances where cluster catalogues are still very incomplete.
Finding 178 GRS in 1059 deg 2 corresponds to a density of ∼ 0.17 deg −2 in this area and confirms that RACS images have the potential of revealing ∼5000 GRS over its entire survey area (Dec< +41 • ). However, regions of low Galactic latitude as well as regions not covered by the deepest optical surveys will yield a lower identification rate. On the other hand, the fact that RACS is being repeated at two higher frequencies (∼1.2 and ∼1.6 GHz) will increase its sensitivity and potential for finding GRS.
We expect to double the present sample of GRS by inspecting the sky region from RA=20 ℎ to 6 ℎ 20 , and −65 • < Dec < −50 • , immediately south of that studied here, and also covered by DES, and similar in size to the one just studied. Taken together, this promises a sample of over 350 GRS, plus ∼600 sources with LLS=0.7-1 Mpc. However, the amount of work necessary to find new GRS is clearly impractical and calls for machine learning (ML) algorithms to find good candidates more efficiently. Samples like the one presented in this work may serve as training sets for such algorithms.