The LOFAR LBA Sky Survey I. survey description and preliminary data release
F. de Gasperin, W. L. Williams, P. Best, M. Bruggen, G. Brunetti, V. Cuciti, T. J. Dijkema, M. J. Hardcastle, M. J. Norden, A. Offringa, T. Shimwell, R. van Weeren, D. Bomans, A. Bonafede, A. Botteon, J. R. Callingham, R. Cassano, K. T. Chyzy, K. L. Emig, H. Edler, M. Haverkorn, G. Heald, V. Heesen, M. Iacobelli, H. T. Intema, M. Kadler, K. Malek, M. Mevius, G. Miley, B. Mingo, L. K. Morabito, J. Sabater, R. Morganti, E. Orru, R. Pizzo, I. Prandoni, A. Shulevski, C. Tasse, M. Vaccari, P. Zarka, H. Rottgering
AAstronomy & Astrophysics manuscript no. 2020_LBAsurvey © ESO 2021February 19, 2021
The LOFAR LBA Sky Survey
I. Survey description and preliminary data release
F. de Gasperin , , W. L. Williams , P. Best , M. Brüggen , G. Brunetti , V. Cuciti , T. J. Dijkema , M. J. Hardcastle ,M. J. Norden , A. O ff ringa , T. Shimwell , , R. van Weeren , D. Bomans , A. Bonafede , , A. Botteon ,J. R. Callingham , , R. Cassano , K. T. Chy˙zy , K. L. Emig , (cid:63) , H. Edler , M. Haverkorn , G. Heald , V. Heesen ,M. Iacobelli , H. T. Intema , M. Kadler , K. Małek , M. Mevius , G. Miley , B. Mingo , L. K. Morabito , ,J. Sabater , R. Morganti , , E. Orrú , R. Pizzo , I. Prandoni , A. Shulevski , , C. Tasse , , M. Vaccari , ,P. Zarka , and H. Röttgering Hamburger Sternwarte, Universität Hamburg, Gojenbergsweg 112, D-21029, Hamburg, Germany INAF - Istituto di Radioastronomia, via P. Gobetti 101, 40129, Bologna, Italy Leiden Observatory, Leiden University, P.O.Box 9513, NL-2300 RA, Leiden, The Netherlands Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh, EH9 3HJ, UK ASTRON, the Netherlands Institute for Radio Astronomy, Postbus 2, 7990 AA, Dwingeloo, The Netherlands Centre for Astrophysics Research, University of Hertfordshire, College Lane, Hatfield AL10 9AB, UK Ruhr-Universität Bochum, Universitätsstr 150 / NA7, 44801 Bochum, Germany DIFA - Universitá di Bologna, via Gobetti 93 /
2, I-40129 Bologna, Italy Astronomical Observatory, Jagiellonian University, ul. Orla 171, 30-244, Kraków, Poland National Radio Astronomy Observatory, 520 Edgemont Road, Charlottesville, VA 22903-2475, USA Department of Astrophysics / IMAPP, Radboud University, PO Box 9010, NL-6500 GL Nijmegen, the Netherlands CSIRO Astronomy and Space Science, PO Box 1130, Bentley WA 6102, Australia Institut für Theoretische Physik und Astrophysik, Universität Würzburg, Emil-Fischer-Str. 31, 97074 Würzburg, Germany National Centre for Nuclear Research, ul. Pasteura 7, 02-093, Warsaw, Poland School of Physical Sciences, The Open University, Walton Hall, Milton Keynes MK7 6AA, UK Centre for Extragalactic Astronomy, Department of Physics, Durham University, Durham DH1 3LE, UK Institute for Computational Cosmology, Department of Physics, University of Durham, South Road, Durham DH1 3LE, UK Kapteyn Astronomical Institute, University of Groningen, P.O. Box 800, 9700 AV Groningen, The Netherlands Anton Pannekoek Institute for Astronomy, University of Amsterdam, Postbus 94249, 1090 GE Amsterdam, The Netherlands GEPI&USN, Observatoire de Paris, CNRS, Université Paris Diderot, 5 place Jules Janssen, 92190 Meudon, France Centre for Radio Astronomy Techniques and Technologies, Rhodes University, Grahamstown 6140, South Africa Dep. of Physics & Astronomy, University of the Western Cape, Robert Sobukwe Road, 7535 Bellville, Cape Town, South Africa LESIA, UMR CNRS 8109, Observatoire de Paris, 92195 MEUDON, France
February 19, 2021
ABSTRACT
Context.
The LOw Frequency ARray (LOFAR) is the only radio telescope that is presently capable of high-sensitivity, high-resolution(i.e. < − and < (cid:48)(cid:48) ) observations at ultra-low frequencies ( <
100 MHz). To utilise these capabilities, the LOFAR SurveysKey Science Project is undertaking a large survey to cover the entire northern sky with Low Band Antenna (LBA) observations.
Aims.
The LOFAR LBA Sky Survey (LoLSS) aims to cover the entire northern sky with 3170 pointings in the frequency rangebetween 42 −
66 MHz, at a resolution of 15 (cid:48)(cid:48) and at a sensitivity of 1 mJy beam − (1 σ ). In this work, we outline the survey strategy,the observational status, and the calibration techniques. We also briefly describe several of our scientific motivations and present thepreliminary public data release. Methods.
The preliminary images were produced using a fully automated pipeline aimed at correcting all direction-independente ff ects in the data. Whilst the direction-dependent e ff ects, such as those from the ionosphere, have not yet been corrected, the imagespresented in this work are still ten times more sensitive than previous available surveys at these low frequencies. Results.
The preliminary data release covers 740 deg around the HETDEX spring field region at an angular resolution of 47 (cid:48)(cid:48) witha median noise level of 5 mJy beam − . The images and the catalogue of 25,247 sources have been publicly released. We demonstratethat the system is capable of reaching a root mean square (rms) noise of 1 mJy beam − and an angular resolution of 15 (cid:48)(cid:48) once direction-dependent e ff ects are accounted for. Conclusions.
LoLSS will provide the ultra-low-frequency information for hundreds of thousands of radio sources, providing criticalspectral information and producing a unique data set that can be used for a wide range of science topics, such as the search for highredshift galaxies and quasars, the study of the magnetosphere of exoplanets, and the detection of the oldest populations of cosmic-raysin galaxies, clusters of galaxies, as well as those produced by active galactic nuclei (AGN).
Key words. surveys – catalogs – radio continuum: general – techniques: image processing Article number, page 1 of 19 a r X i v : . [ a s t r o - ph . I M ] F e b & A proofs: manuscript no. 2020_LBAsurvey
1. Introduction
The LOw Frequency ARray (LOFAR; van Haarlem et al. 2013)is a radio interferometric array that operates at very low fre-quencies (10 −
240 MHz), built with the ambition of performinggroundbreaking imaging surveys (Rottgering et al. 2011). Com-pared to existing radio telescopes, LOFAR o ff ers the possibilityof performing transformational high-resolution surveys thanksto the increase in survey speed resulting from its large field ofview (FoV), vast collecting area, and multi-beam capabilities.Two wide-area imaging surveys were designed within its frame-work: 1) LoTSS (LOFAR Two-metre Sky Survey; Shimwellet al. 2017) is a wide area survey at 120–168 MHz that usesthe high band antenna (HBA) system of LOFAR; and 2) LoLSS(LOFAR LBA Sky Survey) is the sibling survey of LoTSS car-ried out in the frequency range 42–66 MHz using the LOFARLow Band Antenna (LBA) system.LoTSS and LoLSS are two wide-area surveys led by the LO-FAR Survey Key Science Project (SKSP; PI: Röttgering). Bothsurveys aim to cover the northern hemisphere. LoTSS has pub-lished its first data release, comprising 424 deg of sky and de-tecting over 320,000 sources (Shimwell et al. 2019). In a selectnumber of regions, where high-quality multi-wavelength datasets are available, the SKSP is also taking longer exposures toachieve significantly higher sensitivities (deep fields). A few ofthem, observed with the HBA system and reaching a noise levelas low as 20 µ Jy beam − , were recently released as part of theLoTSS-deep first data release: Boötes, Lockman, and ELAIS-N1(Tasse et al. 2020; Sabater et al. 2020). In this paper, we focuson the ongoing LOFAR LBA Sky Survey.LOFAR LBA is currently the only instrument capable ofdeep (mJy beam − ), high-resolution (15 (cid:48)(cid:48) ) imaging at frequen-cies below 100 MHz. Even into the SKA era, this capability willremain unique to LOFAR. LoLSS is a long-term project and cur-rently, around 500 deg have been observed at the target integra-tion time per pointing of 8 hrs, whilst data from a further 6700deg are being collected with an initial integration time of 3 hrsper pointing.LoLSS will open a hitherto unexplored spectral window(Fig. 1), addressing one of the original motivations for the con-struction of LOFAR. Compared to other ultra-low frequency sur-veys (VLSSr and GLEAM; Lane et al. 2014; Hurley-Walkeret al. 2017), LoLSS will be 10-100 times more sensitive andwill have an angular resolution that is 5-10 times higher. Forsources with a typical spectral index α ∼ − . S ν ∝ ν α ),LoLSS will be more sensitive than the majority of current andplanned surveys. For sources with ultra-steep spectra ( α < − . ff s at low-frequencies, LoLSS will stand asthe deepest survey available. In the northern hemisphere, whereLoTSS and LoLSS will both cover 2 π steradians, the combina-tion of the two surveys will provide unique insights into the low-frequency spectral index values of a million radio sources.
2. Science cases
LoLSS will investigate low-energy synchrotron radiation with aunique combination of high angular resolution and sensitivity,enabling the study of phenomena such as low-e ffi ciency accel-eration mechanisms and the detection of old cosmic-ray pop-ulations. Studying ‘fossil’ steep-spectrum sources is a require-ment for understanding the nature, evolution, and life cycles of (cid:63) K. L. Emig is a Jansky Fellow of the National Radio AstronomyObservatory Frequency [Hz]0.010.101.0010.00100.00 S e n s i t i v i t y ( ) [ m J y / b ] NVSSGLEAMTGSSVLSSr FIRSTWENSS SUMSSEMUApertif VLASS8CLoLSS-preLoLSS LoTSS E x p l o r e d slope: -0.8slope: -2.3Resolution: Fig. 1.
Comparison of sensitivity for a number of completed and on-going wide-area radio surveys. The diameters of the grey circles areproportional to the survey resolution as shown in the bottom left corner.Data presented in this paper are labelled as ‘LoLSS-pre’, whilst the fi-nal LoLSS survey is labelled as ‘LoLSS’. For sources with a very steepspectral index ( α (cid:46) − . synchrotron radio sources. LoLSS will also probe processes thatmodify the power-law synchrotron spectra at these extreme fre-quencies, thereby providing new information about certain pro-cesses, such as the absorption by ionised gas and synchrotronself-absorption. LoLSS will thus be a unique diagnostic toolfor studying both the local and the di ff use medium in a varietyof astronomical environments. LoLSS is designed to maximisethe synergy with its sibling survey LoTSS. The combination ofLoLSS (LBA) and LoTSS (HBA) will produce a unique bodyof data for the investigation of radio sources at low frequencies,where several new physical diagnostics are available. Owing to their large luminosities and bright associated emissionlines, active galaxies are among the most distant objects observ-able in the Universe. One of the most e ffi cient techniques forfinding high-redshift radio galaxies (HzRGs) and proto-clustersis to focus on ultra-steep spectrum (USS) radio sources (Miley &De Breuck 2008; Saxena et al. 2018). One of the ultimate goalsof the LOFAR Surveys KSP is to detect >
100 radio galaxiesat z >
6; to enable robust studies of the formation and evolu-tion of high-redshift massive galaxies, black holes, and proto-clusters; and to provide a su ffi cient number of radio sources Article number, page 2 of 19. de Gasperin et al.: LOFAR LBA sky survey I within the Epoch of Reionisation to facilitate H i absorption stud-ies. Combining LoLSS and LoTSS in a large region of the skywill identify USS HzRGs candidates, as well as a set of highlyredshifted GHz-peaked sources (peaking at ∼
100 MHz; Falckeet al. 2004), of which >
30 are expected to be at z > (Smith et al. 2016), prior to optical, infrared, and millimetre-wave follow up. Being dynamically complex and very large magnetised regions,galaxy clusters are important laboratories for studying the contri-bution of particle acceleration and transport to cluster evolution(e.g. Brunetti & Jones 2014). To date, approximately 100 clus-ters are known to contain Mpc-sized, steep-spectrum ( α < − ff use cluster radio sources out to z = α < − .
5; Brunetti et al. 2008). The combination of LBA andHBA data will immediately provide resolved spectral index mea-surements for these sources (see e.g. van Weeren et al. 2012;de Gasperin et al. 2020a, for Abell 2256 and the Toothbrush),whilst high-frequency surveys do not have the required combi-nation of depth, resolution, or coverage to be viable counterpartsto LoTSS. For radio relics, both cosmic-ray acceleration at theshock front and their energy loss processes in the post-shock re-gion are poorly understood and tightly linked to the observationsat ultra-low frequencies (de Gasperin et al. 2020a). LoLSS willthus enable the investigation of the microphysics of cosmic-rayacceleration processes in both radio haloes (turbulence acceler-ation) and radio relics (shock-induced acceleration). Studies ofthese processes are expected to place firm constraints on the the-oretical models (Brunetti & Jones 2014).Furthermore, recent LOFAR observations have discovereddi ff use synchrotron emission from bridges connecting clustersthat are still in a pre-merger phase (Botteon et al. 2018; Govoniet al. 2019; Botteon et al. 2020). Only few clusters are knownto be in such configuration, but data from the eRosita all skysurvey (eRASS) will likely increase their number. These ob-servations demonstrate that relativistic electrons and magneticfields can be generated on very large, cosmological scales thathad never been probed before. These pairs of massive clusterssit in largely overdense regions which result from the collapseof cosmic filaments. The resulting bridges are regions where tur-bulence may amplify magnetic fields and accelerate particles,leading to observable radio emission extending on 3 − α ∼ − . https://ingconfluence.ing.iac.es:8444/confluence//display/WEAV/WEAVE-LOFAR Observations at low frequencies have the ability to traceplasma generated by activity from active galactic nuclei (AGN)that has been mildly re-energised through compression or otherphenomena. Sources of this type can have spectral indices assteep as α = − ffi ciently detect this new population of elusive sources. Thesedetections will enable the study of the interaction of radio galax-ies and tailed sources with the intra-cluster medium (Bliton et al.1998) as well as the new micro-physics involved in the ine ffi cientre-acceleration of cosmic-rays in diluted plasmas (de Gasperinet al. 2017). The study of these sources as a population shedslight upon the long-standing problem of the presence and prop-erties of a seed population of cosmic-ray electrons (CRe) in thedi ff use intra-cluster medium. The existence of such a populationwould mitigate the limitation of some standard cosmic-ray accel-eration theories such as the di ff usive shock acceleration (DSA)of thermal pool electrons (Kang et al. 2014). LoLSS will provide the lowest frequency data-points for a largevariety of radio AGN spectra, ranging from young (few hundredyears) gigahertz-peaked spectrum and compact steep spectrumradio sources to old ( ∼ years) giant Mpc-sized radio galax-ies (e.g. Shulevski et al. 2019; Dabhade et al. 2019). In com-pact objects, this information can be used to distinguish betweenjets which may be ‘frustrated’ and not powerful enough to clearthe medium and propagate outside the host galaxy or a ‘young’scenario, in which the radio AGN may only recently have be-come active (see e.g. Callingham et al. 2017 and O’Dea & Saikia2020 for a recent review). By measuring the properties of thelow-frequency turnover in compact sources and hotspots, we canevaluate the relative importance of synchrotron self-absorption,free-free absorption, or a low-energy cut-o ff (e.g. McKean et al.2016; de Gasperin et al. 2020c). Furthermore, the combinationof LoLSS, LoTSS and higher-frequency surveys such as NVSSor Apertif will enable the study of spectral curvature over a widefrequency band for a large number of sources ( ∼ , , > (cid:48)(cid:48) in the HETDEX region; Hardcastle et al. 2019). TheLBA survey data could also shed some light on the activity cy-cles in the newly-discovered population of low-luminosity FRIIs(Mingo et al. 2019), which are believed to inhabit lower masshosts than their high luminosity counterparts.In the case of blazars (radio-loud AGN whose relativisticallybeamed jets are oriented close to the line of sight), radio-spectralindices are characteristically flat throughout the centimetre bandand even down to the LOFAR HBA band at ∼
150 MHz (Trüst-edt et al. 2014; Mooney et al. 2019). These flat spectra are dueto the superposition of many di ff erent jet emission zones nearthe compact base of the jets of varying size and synchrotronturnover frequency. At su ffi ciently low frequencies, however, theblazar emission is expected to become altered by strong self-absorption. On the other hand an additional component of un-beamed steep-spectrum emission from the optically extendedjets and lobes might start to dominate the source emission. Due Article number, page 3 of 19 & A proofs: manuscript no. 2020_LBAsurvey to their flat spectra, blazars are generally much fainter at MHzfrequencies than unbeamed radio-loud AGN so that their prop-erties in this regime are poorly studied. Massaro et al. (2013);Giroletti et al. (2016); D’Antonio et al. (2019) found that theaverage radio spectrum of large samples of blazars is flat downto tens of MHz, suggesting that their spectra are still dominatedby the beamed core emission even at such ultra-low frequencies.However, previous studies were a ff ected by variability and lim-ited angular resolution, which rendered it impossible to separatethe core and lobe emission of blazars. This will be improved sig-nificantly by LoLSS and follow-up LOFAR LBA observations.Blazars are also an important source class for high-energy as-tronomy and astroparticle physics. Mooney et al. (2019) foundlow-frequency radio counterparts to all gamma-ray sources inthe Fermi Large Area Telescope Third Source Catalog (3FGLAcero et al. 2015) at 150 MHz within LoTSS that are associatedwith known sources at other wavelengths and found source can-didates for unassociated gamma-ray sources within the LoTSSfootprint. Covering the same field, LoLSS opens the opportunityto unveil (possibly new) associations at even lower frequencies.Ultra-low frequency data are also crucial for the study ofremnants of radio galaxies (Brienza et al. 2017; Mahatma et al.2018). Recent investigations have shown that this elusive popu-lation of sources exhibits a range of spectral properties and somestill show the presence of a faint core (Morganti et al. 2020; Ju-rlin et al. 2020). The addition of a very low frequency point fora resolved spectral analysis will constrain the time scale of the‘o ff ’ phase. The ultimate aim is to obtain a census of AGN rem-nants that will provide the rate and duration of the AGN radio-loud phase, allowing for a comprehensive study of triggering andquenching mechanisms and constraining models of the radio ac-tivity in relation to the inter-stellar medium (ISM) and associatedstar formation rates. A key contribution to the quantitative studyof the AGN life cycle will also come from the study of restartedradio galaxies whose identification and temporal evolution dueto plasma ageing will also be possible only through the measure-ment of their low-frequency spectra (Jurlin et al. 2020).LoLSS and LoTSS data, combined with optical, infrared(IR), and millimetre data sets, will also be used to determine theevolution of black hole accretion over cosmic time and to addresscrucial questions related to the nature of the di ff erent accretionprocesses, the role of AGN feedback in galaxy evolution, and itsrelation to the environment. Dramatic examples of such feedbackinclude the giant X-ray cavities seen in the hot atmospheres ofmany cool-core galaxy groups and clusters. These cavities, in-flated by the lobes of the central AGN, represent an enormousinjection of feedback energy. The low-frequency data of thesesystems are critical for constraining the state of the plasma inthe largest cavities, as well as in later phases, when the relativis-tic plasma is e ff ectively mixed by instabilities with the thermalICM. In fact, old (‘ghost’) cavities from earlier generations ofactivity are often only visible at very low frequencies, due toageing e ff ects and large angular scales (e.g. Birzan et al. 2008).These measurements will clarify the AGN duty cycle and theimpact of AGN feedback by refining scaling relations betweenthe radio and feedback power (see Heckman & Best 2014, for areview). Lastly, LoLSS will have the unique potential to revealpossible reservoirs of very old CRe that could explain the oftenobserved discrepancy between the young spectral age of radiogalaxies and the apparently older dynamical age (Heesen et al.2018; Mahatma et al. 2020). LoLSS will give access to the lowest radio frequencies in galax-ies, so that we can study the radio continuum spectrum in un-precedented detail. The main science drivers are: (i) using theradio continuum as an extinction-free star formation tracer ingalaxies; (ii) characterising radio haloes as a mean of studyinggalactic winds; and (iii) investigating the origin and regulationof galactic magnetic fields.Radio continuum emission in galaxies results from twodistinct processes: thermal (free–free) and non-thermal (syn-chrotron) radiation. Both are related to the presence of massivestars, with UV radiation ionising the gas leading to free-freeemission. The same stars end their lives in supernovae, whichare the most likely places for the acceleration of CRe to GeV-energies, which are responsible for the synchrotron emission.The relationship between the radio continuum emission of agalaxy and its star formation rate (SFR), that is, the radio-SFRrelation, is centred on the interplay of star formation and gas,magnetic fields, and CRe (Tabatabaei et al. 2017). At frequenciesbelow 1 GHz the thermal contribution is less than ten percent forthe global spectra, which means that with low frequencies, wecan study the non-thermal radio–SFR relation, which has morecomplex underlying physics. This is particularly the case whengalaxies are not electron calorimeters, meaning that some CReescape via di ff usion and advection in winds. Hence, to make itpossible to exploit the radio–SFR relation for distant galaxies atthis frequency, we need to calibrate this relation in nearby galax-ies with known SFRs (e.g. Calistro-Rivera et al. 2017). As a side-e ff ect, we can explore the physical foundation that gives rise tothe relation in the first place, such as the link between magneticfield strength and gas density (e.g. Niklas & Beck 1997). Eventu-ally, LoLSS will detect thousands of galaxies at z < . , provid-ing data that can be used to distinguish between various modelsfor the scarcely explored ultra low-frequency radio-SFR relationand its close corollary, the radio-far-infrared (radio-FIR) corre-lation, down to the frequencies where it may break down due tofree–free absorption. These data will also explore the variationwith galaxy properties (as done at HBA frequencies by Gurkanet al. 2018; Smith et al. 2020), which are essential to constrainif radio data are to be used to probe star formation at higher red-shifts.Low-frequency observations are particularly useful for spa-tially resolved studies of CRe and magnetic fields in nearbygalaxies. The distribution of the radio continuum emission issmoothed with respect to the CRe injection sites near star-forming regions. This can be ascribed to the e ff ects of CRe di ff u-sion, a view that is backed up if we use the radio spectral indexas a proxy for the CRe age (Heesen et al. 2018). However, ra-dio continuum spectra are also shaped by CRe injection, losses,and transport – for instance, in the case of advection in galacticwinds (e.g. Mulcahy et al. 2014). Hence, a fully sampled radiospectrum from the MHz to the GHz regime is necessary for re-liably assessing the age of CRe and also for disentangling thee ff ect of free–free absorption. LoLSS data make it possible todetect the turnover from free-free emission with fairly low emis-sion measures, probing low-density (5 cm − ) warm ionised gaswhich may be prevalent in the mid-plane of galaxies (Mezger1978). Even though one particular statistical study using LO-FAR HBA at 144 MHz hinted that free–free absorption playsa minor role (Chyzy et al. 2018), the contribution from cooler( T < Article number, page 4 of 19. de Gasperin et al.: LOFAR LBA sky survey I ionised gas properties and distinguish its contributions from CRepropagation e ff ects. Furthermore, we are able to explore, for thefirst time, a possible deviation from a power-law cosmic-ray in-jection spectrum at the lowest energies.LoLSS radio-continuum observations could open up a newavenue for studying galactic winds and their relation with thecircum-galactic medium (CGM; see Tumlinson et al. 2017, fora review). Edge-on galaxies show extensive radio haloes, indi-cating the presence of CRe and magnetic fields. By enabling ananalysis of the vertical spectral index profile, LoLSS data can beused to estimate the spectral age of the CRe and, thus, to mea-sure the outflow speed of the wind. LOFAR has allowed someprogress to be made, with radio haloes now detected to muchlarger distances than what was previously possible (Miskolcziet al. 2019). The LOFAR LBA system is likely to detect radiohaloes to even grater distance, thereby providing deeper insightson galaxies interaction with the CGM.Finally, LoLSS could lead to fundamental constraints on thenature of dark matter in dwarf spheroidal galaxies. One of theleading candidates for this are weakly interacting massive par-ticles (WIMPs), which can produce a radio continuum signalannihilating electron-positron pairs. For typical magnetic fieldstrengths, the peak of the signal is expected in the hundred mega-hertz frequency range if the WIMPs are in the mass range of afew GeV. The LOFAR HBA search by Vollmann et al. (2020)has so far provided upper limits, which can possibly improvedwith LOFAR LBA observations, particularly in the lower massrange where HBA observations are less sensitive. Low-frequency observations with LOFAR will open a new areaof discovery space in Galactic science. LoLSS will image a largefraction of the northern Galactic Plane, thereby completing acensus of supernova remnants (SNR). This will enable a searchfor the long-predicted and missing population of the oldest SNR,whose strongly rising low frequency spectra and large angu-lar scales are not visible at higher frequencies (Driessen et al.2018; Hurley-Walker et al. 2019). The combination of LoLSSand LoTSS data will be important in identifying emission fromH ii regions whose morphology is similar to that of SNRs, whilsthaving a flatter spectrum. Additionally, LoLSS data will enablethe measure the low-frequency spectral curvature of supernovaremnants as a diagnostic of shock acceleration and the fore-ground free-free absorption (Arias et al. 2018).LoLSS will provide a map of Galactic non-thermal emis-sion (e.g. Su et al. 2017) and it can also map and characterisethe properties of self absorption by low-density ionised gas thatappears as ‘absorption holes’ against the smoother background.Concurrently, such observations will serve as a proxy to tomo-graphically image the CRe distribution and magnetic field con-figuration throughout the Galaxy (Polderman et al. 2019, 2020).Finally, LoLSS will enable the study of: (i) the role of pulsarwind nebulae in dynamically shaping their environment; (ii) thestar-forming processes in close proximity to very young stellarobjects by detecting their associated thermal and non-thermalemitting radio jets; and (iii) candidate pulsars through their ultra-steep spectra. Radio emission from stars is a key indicator of magnetic ac-tivity and star-planet plasma interactions (Hess & Zarka 2011). Existing studies have mainly focused on cm wavelengths ( ν > >
60% circular polarisa-tion (Vedantham et al. 2020b).The emission characteristics andstellar properties strongly suggest that the low-frequency emis-sion is driven by a star-exoplanet interaction. In parallel, weakradio bursts from the Tau Bootes system that hosts a hot Jupiterhave been tentatively detected in the 14 −
21 MHz range usingLOFAR in beamformed mode (Turner et al. 2020). These dis-coveries herald an unprecedented opportunity to constrain mag-netic activity in main-sequence stars other than the Sun as wellas the impact of the ensuing space-weather on exoplanets, as ex-emplified by the 19 other detections presented by Callingham etal. (under review). Additionally, the recent direct discovery of acold brown dwarf using LoTSS data (Vedantham et al. 2020a)also demonstrates the new potential of deep low-frequency sur-veys in helping us to understand the properties of planetary-scalemagnetic fields outside of the Solar System.Since the detected radio emission is produced via the elec-tron cyclotron maser instability (ECMI), the frequency of emis-sion is directly related to magnetic field strength of either thestar or exoplanet. Therefore, at HBA frequencies, studies arerestricted to a subset of extreme M and ultracool dwarfs thathave strong magnetic fields ( >
50 G). With its lower frequen-cies, LoLSS can begin to probe exoplanets and stars with mag-netic field strengths similar to those found in our Solar System( ∼ ±
15 detections in the complete LoLSS (Calling-ham et al. under review). Hence, LoLSS will play a major rolein characterising the phenomenology of low-frequency emissionof stellar systems and has the potential to dramatically impact onour understanding of the magnetic field properties and environ-ments of other planetary systems around nearby stars.
Continuous, systematic, long-term observations at very low-frequencies will allow for the characterisation of important as-pects of the ionosphere, such as physical parameters of iono-spheric travelling waves, scintillations, and the relation with so-lar cycles (Mevius et al. 2016; Helmboldt & Hurley-Walker2020). All of these are crucial aspects for constraining iono-spheric models. Instruments observing at ultra-low frequency arepowerful tools for deriving the total electron content (TEC) ofthe ionosphere independently from standard observations withsatellite measurements (Lenc et al. 2017; de Gasperin et al.2018b). LoLSS observations will also provide large data setswith which it is possible to study the higher-order e ff ects im-printed on travelling radio-waves as Faraday rotation and theionospheric third-order delay (de Gasperin et al. 2018b). Serendipitous discoveries have always played an important rolein astronomy, particularly with the opening of new spectral
Article number, page 5 of 19 & A proofs: manuscript no. 2020_LBAsurvey
Number of pointings 3170Separation of pointings 2.58 ◦ Integration time (per pointing) 8 hFrequency range 42–66 MHzArray configuration LBA OUTERAngular resolution ∼ (cid:48)(cid:48) Sensitivity ∼ − Time resolution 1 s– after averaging 2 sFrequency resolution 3.052 kHz– after averaging 48.828 kHz
Table 1.
LoLSS observational setup. windows. An example is the transient detected during the ini-tial years of LOFAR observations, whose nature is still unclear(Stewart et al. 2016). LoLSS probes the lowest energy extremeof the electromagnetic spectrum, a regime where exotic radia-tion mechanisms such as plasma oscillations play a role. A po-tentially exciting part of analysing LoLSS will be searching fornew, unexpected classes of objects that are only detectable at orbelow 50 MHz.
3. The LOFAR LBA sky survey
Building on the performance statistics of LOFAR during com-missioning observations, we selected an observing mode forLoLSS that optimises survey speed whilst achieving the desiredangular resolutions of 15 (cid:48)(cid:48) and sensitivity of ∼ − . Asummary of the final observational setup is listed in Table 1.The LOFAR LBA system has the capability of simultane-ously casting multiple beams in di ff erent and arbitrary direc-tions at the expense of reduced observing bandwidth. In orderto maximise the survey speed and to provide an e ffi cient calibra-tion strategy, during each observation we continuously keep onebeam on a calibrator source whilst placing three other beamson three well-separated target fields. The beam on the calibra-tor is used to correct instrumental (direction-independent) ef-fects such as clock delays and bandpass shape (de Gasperin et al.2019); the rationale behind continuously observing the calibratoris that these systematic e ff ects are not constant in time and can bemore easily derived from analysing well-characterised calibratorfields.Ionospheric-induced phase variations are the most problem-atic systematic e ff ect at ultra-low frequencies (Intema et al.2009; Mangum & Wallace 2015; Vedantham & Koopmans 2015;Mevius et al. 2016; de Gasperin et al. 2018b). In order to mitigatethe consequences of poor ionospheric conditions on a particularobservation, we used the following observing strategy: duringeach observation we simultaneously place three beams on threetarget fields for one hour. After an hour, we switch beam loca-tions to a di ff erent set of three targets. We schedule observationsin 8 hour blocks. In total, 24 fields are observed for one hour eachin an 8 hour block. The same process is then repeated eight timesto improve the sensitivity and the u v -coverage of each pointing.In this way, if the ionosphere was particularly problematic dur-ing a particular day, it would have a ff ected only a fraction ofthe data in each field, without compromising the uniform sen-sitivity of the coverage. To prepare the observations, we use ascheduling code that implements this observing strategy, whilstmaximising the u v -coverage so that each field is not observedtwice at hour angles closer than 0.5 hr. We ensure that observa- tions were taken when the Sun is at least 30 ◦ from the targetedfields and their elevation is above 30 ◦ .The total bandwidth available in a single LOFAR observa-tion is 96 MHz (8-bit mode). When divided into four beams thisgives a usable band of 24 MHz. We tuned the frequency cover-age to 42–66 MHz to overlap with the most sensitive region ofthe LBA band taking into account both the sky temperature andthe dipole bandpass (see van Haarlem et al. 2013). To suppressthe e ff ect of strong radio frequency interference (RFI) reflectedby ionospheric layers at frequencies <
20 MHz, the LBA signalpath is taken through a 30-MHz high-pass filter as default.Due to a hardware limitation (which will be removed in afuture upgrade to LOFAR) in each station, only half of the LBAdipoles can be used during a single observation. The choice ofthe dipoles that are used has a large impact on the size and shapeof the main lobe of the primary beam and on the positions andamplitudes of the side lobes. The LOFAR LBA system can beused in four observing modes:LBA INNER: The inner 48 dipoles of the station are used. Thismode gives the largest beam size at the cost of a reducedsensitivity. The calibration of the inner dipoles (the stationcalibration) is less e ff ective than for the outer dipoles dueto mutual coupling and their higher sensitivity to Galacticemission during the station calibration procedure. The e ff ec-tive size of the station is 32 m, which corresponds to a pri-mary beam full width at half maximum (FWHM) of 10 ◦ at54 MHz.LBA OUTER: The outer 48 dipoles of the station are used. Thismode minimises the coupling between dipoles but reducesthe beam size. The e ff ective size of the station is 84 m, pro-viding a primary beam FWHM of 3.8 ◦ at 54 MHz.LBA SPARSE (ODD or EVEN): Half of the dipoles, dis-tributed across the station, are used. At the time of writingthis mode is experimental, but grants an intermediate per-formance between LBA INNER and OUTER, with a sup-pression of the magnitude of the side-lobes compared to thelatter. The e ff ective size of the station is around 65 m, whichprovides a primary beam FWHM of 4.9 ◦ at 54 MHz.Given the better quality of the LBA OUTER station calibrationand the close similarity of the primary beam FWHM with theHBA counterpart (3.96 ◦ at 144 MHz) this observing mode wasused. The LBA OUTER mode results in a primary beam FWHMranging from 4.8 ◦ to 3.1 ◦ for the covered frequency range be-tween 42–66 MHz. The use of the LBA OUTER mode alsoimplies the presence of a non-negligible amount of flux den-sity spilling in from the first side lobe. This e ff ect is partiallycompensated for by the calibration strategy, where sources in thefirst side lobe are imaged and subtracted (see de Gasperin et al.2020a).Since the FWHM of the primary beam of LoTSS and LoLSSis similar, we adopted a joint pointing strategy so that each targetfield is centred on the same coordinates in both surveys. LoLSStherefore has the same pointing scheme as LoTSS (see Fig. 2).The pointing scheme follows a spiral pattern starting from thenorth celestial pole, with positions determined using the Sa ff & Kuijlaars (1997) algorithm. This algorithm attempts to uni-formly distribute points over the surface of a sphere when thereis a large number of pointings. Using the same pointing sep-aration as LoTSS (2.58 ◦ ), the coverage of the entire northernhemisphere requires 3170 pointings. Assuming circular beams,this separation provides a pointing distance of FWHM / / √ Article number, page 6 of 19. de Gasperin et al.: LOFAR LBA sky survey I lowest. The distance between pointings at the mean frequencyis close to FWHM / √ ∼ ∼ (cid:48)(cid:48) at 54 MHz, with a longest baselineof 120 km. LoLSS makes use of CS and RS, whilst IS data werenot recorded to keep the size of the data set manageable . For anexample of the u v -coverage, which by design can be di ff erent foreach pointing, we refer to de Gasperin et al. (2020a). The longestbaseline available for the observations presented in this paperwas approximately 100 km, providing a nominal resolution atmid-band (54 MHz) of 15 (cid:48)(cid:48) .The final aim of LoLSS is to cover the northern sky to adepth of ∼ − . With the LBA system, this requiresaround 8 hrs of integration time at optimal declination, althoughthe final noise is mostly limited by ionospheric conditions andexperiments indicate that in practice, it will range between 1 and1.5 mJy beam − (de Gasperin et al. 2020a). In this preliminaryrelease, where the direction-dependent errors are not corrected,the noise ranges between 4 and 5 mJy beam − .Ionospheric scintillations can make ultra low-frequency ob-servations challenging by decorrelating the signal even on veryshort baselines. Several years of observations of the amplitudesof ionospheric scintillations using LOFAR show that the phe-nomenon is more prevalent from sunset to midnight than dur-ing the daytime (priv. comm. R. Fallows), which broadly fol-lows patterns that have been observed at higher latitudes (Sreeja& Aquino 2014). Therefore, in order to minimise the chancesof ionospheric scintillations, all the observations presented herewere taken during daytime. However, daytime observations havesome drawbacks, predominantly at the low-frequency end ofthe full LBA band (i.e. < −
40 MHz), below the frequencycoverage of LoLSS. Due to solar-induced ionisation, the iono-sphere becomes thicker during the day. This has two main con-sequences: the lower ionospheric layers can reflect man-madeRFI towards the ground, which is typically seen at frequencies <
20 MHz. At the same time, the e ff ect of Faraday rotation isexpected to be larger, because it also depends on the absolutetotal electron content of the ionosphere, which can increase bya factor of 10 in the daytime. Since Faraday rotation has a fre-quency dependency of ν , this systematic e ff ect is dominant atthe lowest-frequency end of our band coverage, where the dif-ferential rotation angle on the longest baselines is typically oneto two radians (de Gasperin et al. 2019).The resolution in time and frequency is chosen to minimisethe e ff ect of time and frequency smearing at the edge of the fieldof view as well as to be able to track typical ionospheric vari-ations, whilst avoiding the compilation of data sets that are toolarge to handle. Data are initially recorded at 1 s / ff ringa et al. 2010)and bright sources removed from far side lobes (de Gasperinet al. 2019). Before they are stored into the LOFAR Long TermArchive , the data are further averaged to 2 sec and 48.828 kHz . The use of IS would have increased the data set size by a factor of ∼
10. A factor of 2 of this would come from more baselines, and a factorof 4 to 8 from the increase in the frequency-time resolution required inorder to account for larger di ff erential ionosphere on the longest base-lines. https://lta.lofar.eu/ This corresponds to 4 channels per Sub Band (SB), where a SB band-width is 195.3125 kHz wide.
The e ff ects of the time and bandwidth smearing due to thisaveraging can be approximated using the equations of Bridle &Schwab (1989). At a distance of 2 ◦ from the phase centre and at15 (cid:48)(cid:48) resolution, the time averaging to 2 sec leads to a time smear-ing that reduces the peak brightness of sources by < ff erential (between stations) TECvalue of 1 TEC unit (TECU; 10 electrons m − ) produces aphase variation of 13 ◦ at 42 MHz. Typical variations within LO-FAR core and remote stations are well within 1 TECU and cantherefore be corrected in each channel without signal loss. Thehighest di ff erential variation that can be tracked within a 2 sectime slot is about 10 mTECU, corresponding to a drift in phaseof ∼ ◦ at 42 MHz.
4. Survey status
Using the survey strategy described above, the full northern skycan be observed in 3170 pointings / × = (see Fig. 2). These observations are concentratedaround the HETDEX spring field and are the focus of this pre-liminary data release. Of the 95 pointings, 19 have only 7 hrs ofusable observation due to various technical problems. One field(P218 +
55) has been observed for 16 hrs and another field is cur-rently missing from the coverage (P227 +
50) but will be addedto the survey in a future release.Archived data includes full Stokes visibilities from all Dutchstations (core and remote) but not from the international sta-tions. The frequency coverage is always 42 – 66 MHz. The dataare also compressed using the Dysco algorithm (O ff ringa 2016).Archived data are already pre-processed to flag RFI before aver-aging and to subtract the e ff ect of Cygnus A and Cassiopeia A ifsome of their radiation was leaking through a far side lobe. Thedata size for an observation of 8 hrs is ∼
100 GB per pointing.The present allocated observing time allows for the cover-age of all fields above 40 ◦ declination. This campaign will cover6700 deg (1035 pointings), that is 33% of the northern sky with3 hrs per pointing, reaching a sensitivity of 2 mJy beam − . Lowerdeclination and full depth are planned for future observing cam-paigns. The LOFAR LBA data have been collected since cycle 0 (2013),which was close to the solar maximum. This led to rapid andstrong variations in the ionospheric properties in the years be-tween 2011-2014, posing a particular challenge for data pro-cessing. From 2014 onwards, solar activity steadily decreased toreach a minimum in 2020. The quality of the LBA data steadilyincreased with decreasing solar activity and we achieved closeto 100% usable data in cycle 8 (2017). Currently, solar activity RA: 11 h to 16 h and Dec: 45 ◦ to 62 ◦ in the region of the Hobby-Eberly Telescope Dark Energy Experiment (HETDEX) Spring Field(Hill et al. 2008). Article number, page 7 of 19 & A proofs: manuscript no. 2020_LBAsurvey
Fig. 2.
Current and planned sky coverage of LoLSS. Each dot is a pointing of the full survey. Red dots are scheduled to be observed by 2022 witha priority on extragalactic fields. The region presented in this paper is coloured in yellow (cycle 8 data), blue (cycle 9 data) and green (cycle 12data). Solid black lines show the position of the Galactic plane with Galactic latitude: − ◦ , 0 ◦ , + ◦ . is close to its minimum, which for this solar cycle had been par-ticularly long (De Toma et al. 2010). If the next solar cycle issimilar to the past one as predicted, the good conditions for lowfrequency observations will continue until around 2022. Solaractivity will then make low-frequency observations challenginguntil 2027.
5. Data reduction
The data reduction of LoLSS is being carried out in a distributedmanner on computing clusters located at the Observatory ofHamburg, the Observatory of Leiden, the Institute of Radio As-tronomy (INAF, Bologna), and the University of Hertfordshire.Synchronisation between the various running jobs is maintainedthrough a centralised database. All computations are carried outin the same environment built within a Singularity containerbased on Ubuntu 20.04.Here we present a preliminary release of LoLSS data, whichwas prepared using the automated Pipeline for LOFAR LBA(PiLL) , that is described in detail in de Gasperin et al. (2019,for the calibrator processing) and in de Gasperin et al. (2020c,for the target processing). PiLL now includes the possibility ofcarrying out full direction-dependent calibration. However, be-cause this is still experimental, in this paper we discuss and re-lease only those data sets that have been processed as far as thedirection-independent calibration. A direction-dependent surveyrelease reaching 1 mJy beam − root mean square (rms) noise and15 (cid:48)(cid:48) resolution will be presented in a forthcoming publication. The pipeline for the data reduction is publicly available at https://github.com/revoltek/LiLF
Cycle Proposal Year N pointings8 LC8_031 2017 479 LC9_016 2017 2412 LC12_017 2019 24
Table 2.
Observations used for the preliminary release.
Depending on the target position, the scheduler code selectsthe closest calibrator from 3C 196, 3C 295, and 3C 380. Here,we summarise the most important calibration steps. Followingde Gasperin et al. (2019), the calibrator data are used to extractthe polarisation alignment, the Faraday rotation (in the directionof the calibrator), the bandpass of each polarisation and phasesolutions, which include the e ff ects of both the clock and iono-spheric delay (in the direction of the calibrator). For each hour ofobservation, the polarisation alignment, bandpass, and phase so-lutions are then transferred to the three simultaneously observedtarget fields. Direction-independent e ff ects are then removed toproduce target phases that now include a di ff erential ionosphericdelay with respect to the calibrator direction.All eight data sets for each target field, which have gener-ally been observed on di ff erent days, are then combined priorto performing the self-calibration procedure. The initial modelfor the self calibration is taken from the combination of TGSS(Intema et al. 2017), NVSS (Condon et al. 1998), WENSS (Ren-gelink et al. 1997), and VLSSr (Lane et al. 2014) and includes aspectral index estimation up to the second order that is used toextrapolate the flux density to LoLSS frequencies. As explainedin de Gasperin et al. (2020c), the first systematic e ff ect to be cali-brated is the ionospheric delay, followed by Faraday rotation andsecond order dipole-beam errors. The latter is the only amplitudecorrection. Since it is a correction on the dipole beam shape, itis constrained to be equal for all stations. It reaches a maximum Article number, page 8 of 19. de Gasperin et al.: LOFAR LBA sky survey I of few per cent level and it is normalised to ensure that the fluxdensity scale is not altered by our initial self-calibration model.Finally, sources in the first side lobe are imaged and removedfrom the data set. The process is repeated twice to improve themodel and the output is a direction-independent calibrated im-age.
Final imaging is performed with WSClean (O ff ringa et al. 2014)with Briggs weighting − . ff ringa& Smirnov 2017). The maximum u v length used for the imag-ing is 4500 λ , which removes the longest baselines that are moreseriously impacted by large direction-dependent ionospheric er-rors. A third order polynomial is used to regularise the spectralshape of detected clean components. The gridding process is per-formed using the Image Domain Gridder algorithm (IDG; VanDer Tol et al. 2018). The implementation of IDG in WSCleanallows for the imaging of visibilities whilst correcting for timeand direction-variable beam e ff ects.The 95 direction-independent calibrated images are com-bined into a single large mosaic. Sources from each image arecross-matched with sources from the FIRST catalogue (Beckeret al. 1995) to correct the astrometry for each pointing indepen-dently (see also Sect. 6.2, for a full description of the process).Only ’isolated’ and ’compact’ sources are used for this cross-matching process. In order to be considered as isolated, a sourceneeds its nearest neighbour to be at a distance of > × (cid:48)(cid:48) . To bedefined as compact, a source needs to have its integrated to peakflux ratio lower than 1.2. The maximum shift applied to a fieldwas of 2.9 (cid:48)(cid:48) , with the majority of the corrections being less than1 (cid:48)(cid:48) . During the process all images are convolved to the minimumcommon circular beam of 47 (cid:48)(cid:48) . Pixels common to more than onepointing are averaged with weights derived by the local primarybeam attenuation combined with the global noise of the point-ing where each pixel belongs. All regions where the attenuationof the primary beam was below 0.3 were discarded during theprocess.
6. Results
Here, we present the images from the preliminary data release,focusing on its sensitivity, astrometric precision, and accuracy,and assessing the uncertainty on the flux density. For the pur-pose of source extraction, we used PyBDSF (Mohan & Ra ff erty2015). The source extractions are made using a 4 σ detectionthreshold on islands and a 5 σ threshold on pixels. To reducefalse positives, we used an adaptive rms box size that increasesthe background rms noise estimation around bright sources. An image showing the local rms noise distribution calculatedwith PyBDSF is presented in Fig. 3. The quality of the imagesvaries significantly across the covered area, with regions withrms noise up to three times higher than others. The generallylower noise in the upper part of the region presented here is likelyrelated to the observing period of the di ff erent fields. The upperregion was observed during cycle 12 (2019) when the solar cyclewas at its minimum, thus reducing the presence of ionosphericdisturbances, whilst the southern part was observed during cy-cles 8 and 9 (2017). The ionospheric irregularities introduce phase errors that canmove sources in a way that is not synchronised across the im-age and with a positional change that is non-negligible com-pared to the synthesised beam. Therefore, the main driver of thenon-uniformity of the rms noise distribution is the presence ofbright sources in combination with the limited dynamic range,caused by the time- and direction-varying ionosphere whichcannot be corrected in a direction-independent calibration. Al-though we limited the length of the baselines to reduce this ef-fect, the sources are still blurred and their peak flux is reduced(see Fig. 4). As expected, the e ff ect is slightly more relevantfor observations taken in 2017 (mean integrated-to-peak fluxdensity ratio: 1.6) than for observations taken in 2019 (meanintegrated-to-peak flux density ratio: 1.5), those that cover thenorthern region of the presented footprint. This e ff ect and thenon-uniformity of the rms noise will be reduced with the fulldirection-dependent calibration.In Fig. 5, we show the histogram of the rms noise across thefield. The histogram includes the edges of the field, where thenoise is higher because of reduced coverage. Most of the coveredregion has a rms noise of ∼ − . The area with a noiseequal or lower than 4 mJy beam − is 222 deg which accountsfor 30% of the presented region. The median rms noise of theentire region is ∼ − . The astrometric accuracy of our observations might be a ff ectedby errors in the initial sky model used for calibration, whichwas formed from a combination of catalogues from surveys atdi ff erent resolutions. These errors might propagate through thephase solutions and introduce systematic errors in the positionof our sources. However, the phase calibration is performed witha reduced number of degrees of freedom (one per antenna pertime slot) thanks to the frequency constraint, which assumes thatphase errors are largely due to TEC-induced delays. Systematicpositional o ff sets are corrected during the mosaicing process (seeSect. 5.1). The way we identify positional o ff sets to correct dur-ing the mosaicing process and the way we assess our final accu-racy are the same and we describe them in detail below.As a reference catalogue we used the FIRST survey, whichhas a systematic positional error of less than 0.1 (cid:48)(cid:48) from theabsolute radio reference frame, which was derived from high-resolution observations of selected calibrators (White et al.1997). To reduce the bias due to erroneous cross-matching wefirst reduce the LoLSS catalogue to isolated sources, namelythose sources with no other detections closer than three times thebeam size (47 (cid:48)(cid:48) ). This brings the number of sources from 25247to 22766 (90%). This process was repeated for the reference cat-alogue, where only sources with no other detections within 47 (cid:48)(cid:48) were selected, reducing the number of sources by 35%. This stepis important to avoid selecting double sources, which are rathercommon. Then, the subset of sources of both LoLSS and theFIRST catalogues are further reduced to include only compactsources. Compact sources are defined as those with a total-fluxto peak-flux ratio less than 1 .
2. This reduces the LoLSS cata-logue to 6000 sources (23%). Finally the two samples are cross-matched with a large maximum distance of 100 (cid:48)(cid:48) . Sources thatare farther apart than 10 × the median absolute deviation (MAD)of the o ff sets are removed in an iterative process. The final num-ber of sources after the cross-match is 2770 (final MAD: 1.1 (cid:48)(cid:48) ).The final mean separation between selected sources in ourcatalogue and FIRST catalogue was found to be − . (cid:48)(cid:48) in RAand − . (cid:48)(cid:48) in Dec with relative standard deviations RA = . (cid:48)(cid:48) Article number, page 9 of 19 & A proofs: manuscript no. 2020_LBAsurvey
Fig. 3.
Rms noise map of the HETDEX region in Jy beam − . The regions with reduced sensitivity are located at the edges of the survey footprintand around bright sources. The location of 3C 295 is marked with a white X, the presence of the bright source increases the local rms substantially. Fig. 4.
Ratio between the integrated flux density and the peak flux den-sity of isolated sources, i.e. sources with no other detections closer than3 × the beam size (47 (cid:48)(cid:48) ), as a function of the distance from the point-ing centre. Unresolved sources should have a value of around unity (redline), with resolved sources having higher values. The binned medians(blue crosses) go from 1.2 to 1.4. Since the majority of the sources inour catalogue is expected to be unresolved at this angular resolution,this is an indication of ionospheric smearing. and Dec = . (cid:48)(cid:48) (see Fig. 6). Given the small global o ff set be-tween our catalogue and FIRST we did not correct for the shift. To calibrate direction-independent e ff ects as well as the band-pass response of the instrument (see Sect. 5.1) we used one ofthe following flux calibrators: 3C 196 (50% of the observations),3C 295 (40% of the observations), and 3C 380 (10% of the ob- C u m u l a t i v e a r e a [ s qd e g ] Fig. 5.
Histogram of the rms noise. The black solid line shows the cumu-lative function. The red dashed lines show that 30% of the survey foot-print (222 deg ) has a rms noise < − , whilst the blue dashedlines show that 50% of the survey footprint (370 deg ) has an rms noise < − . The tail of noisy regions above 8 mJy beam − aredue to the footprint edges and dynamic range limitations due to brightsources. servations). The choice of the calibrator depends on the elevationof the source at the moment of the observation. The flux densityof the calibration models was set according to the low-frequencymodels of Scaife & Heald (2012) and it has a nominal error rang-ing between two and four percent depending on the source used.The LOFAR LBA system is rather simple and stable: two-beam observations, pointing at two flux calibrators simultane-ously, showed that the flux density of one could be recoveredusing the bandpass calibration from the other at the five percentlevel. We can use this value as an estimation of the flux densityaccuracy. Within the presented survey area, there is also 3C 295,whose flux density can be measured at the end of the calibra-tion and imaging process to assess whether it is consistent withthe value given by Scaife & Heald (2012). In the final surveyimage, the integrated flux density of 3C 295 is 130 Jy, against Article number, page 10 of 19. de Gasperin et al.: LOFAR LBA sky survey I
10 5 0 5 10differential RA [arcsec]10.07.55.02.50.02.55.07.510.0 d i ff e r e n t i a l D e c [ a r c s e c ] Fig. 6.
Astrometric accuracy of the sources in the catalogue (see text forthe calculation). The average astrometric o ff sets are RA = − . (cid:48)(cid:48) andDec = − . (cid:48)(cid:48) with relative standard deviations RA = . (cid:48)(cid:48) and Dec = . (cid:48)(cid:48) . The red ellipse traces the standard deviation. an expected flux density of 133 Jy ( ∼
2% error). This can beused to establish an idea of the flux density precision. Adding inquadrature the nominal error on the flux scale (4%) with thesetwo errors provides a global error budget of 7%.To validate this estimation we can compare LoLSS flux den-sities with those from other surveys. This is not trivial as nosurveys of su ffi cient depth to measure the spectral index of ameaningful number of sources in the survey footprint are avail-able at frequencies lower than 54 MHz. This procedure can beattempted using the 8C survey at 38 MHz, although only 230sources from 8C are visible in LoLSS due to the partial overlapof the surveys’ footprints. The alternative approach to validatethe flux level relies on extrapolating the flux densities down toLoLSS frequency from higher frequency surveys.In order to double check the flux density calibration of ourcatalogue, we used data from 8C (38 MHz), VLSSr (74 MHz),LoTSS-DR2 (144 MHz), and NVSS (1400 MHz). For each ofthese surveys, as well as for LoLSS, we restricted the catalogueto isolated sources as described in Sect. 6, using a minimum dis-tance between sources of two times the survey resolution. Eachcatalogue was then cross-matched with the LoLSS catalogue,allowing for a maximum separation of 6 (cid:48)(cid:48) (15 (cid:48)(cid:48) in the case ofVLSSr and 60 (cid:48)(cid:48) for 8C). Because of ionospheric smearing, forLoLSS and LoTSS, we used the integrated value of the flux den-sity. As a first test, we cross-checked the flux density value of3C 295 with that expected from Scaife & Heald (2012). All sur-veys covering 3C 295, except LoTSS, appear to be consistentwithin a few percent with the expected flux, as shown in Table 3.Dynamic range limitations seem to have a ff ected the LoTSS im-age quality in that region.As a next test, we rescaled the flux density of each surveyto the expected value at 54 MHz assuming a flux-independentspectral index of α = − .
78 (de Gasperin et al. 2018a). The stan- LoTSS Data Release 2 will be presented in the forthcoming publica-tion from Shimwell et al. (in prep.).
Fig. 7.
Ratio of the expected flux density extrapolated from other sur-veys over the flux density as measured in LoLSS as a function of fluxdensity. From top to bottom, the surveys shown are 8C, VLSSr, LoTSSand NVSS. The extrapolated flux density is calculated assuming a spec-tral index α = − .
78 (each source is a black circle). A ratio of 1 (dottedblue line) means perfect extrapolation of the flux density value. Solidlines are detection limits imposed by the survey depth, the vertical lineis due to the LoLSS limit (assumed 1 σ = − ), the diago-nal line is the sensitivity limit of the survey used for comparison. Redcrosses are centered on the binned medians and show the standard de-viations on the y direction and the bin size in the x direction. Greencrosses are the same but assuming a flux-dependent spectral index asfound by de Gasperin et al. (2018a). The dark blue lines show the ex-pected dispersion due to the spectral index distribution. dard deviation of the spectral index distribution is rather large( σ = .
24) and implies a large scatter of the rescaled values,mostly when extrapolating from NVSS data. The results are pre-sented in Fig. 7 (red crosses). Caution must be used when inter-preting these plots as the limited sensitivity of the other surveyscan bias the result, predicting higher than real flux densities forfaint sources. That is the case for VLSSr, as well as for NVSS,
Article number, page 11 of 19 & A proofs: manuscript no. 2020_LBAsurvey
Fig. 8.
Ratio of the expected flux density derived from the combinationof two other surveys over the flux density as measured in LoLSS as afunction of flux density in LoLSS. In this case, the expected flux densityis extrapolated using a spectral index derived from the combination ofthe following surveys: LoTSS-NVSS (top), VLSSr-NVSS (middle), and8C-NVSS (bottom). where the surveys are not deep enough to sample the faint andsteep-spectrum sources present in LoLSS. Given its lower fre-quencies, 8C will instead miss faint, flat spectrum sources. Thediagonal blue lines in Fig. 7 predict these cuto ff levels, which aremore relevant the shallower the reference survey and the largerits frequency distance from 54 MHz (slope of the line). This isnot a problem for LoTSS, where the depth is su ffi cient such thatthe great majority of the sources (up to a spectral index of − − F r ) andthe LoLSS flux densities is F r LoTSS / F LoLSS = .
99 (with a flux-independent spectral index it is F r LoTSS / F LoLSS = . F r / F LoLSS = .
05 in the brightest bin) and for VLSSr matchedsources (with F r VLSSr / F LoLSS = .
91 in the brightest bin).
Fig. 9.
Same as in Fig. 8 but here NVSS and LoTSS are used to predictthe flux density of VLSSr, obtaining a similar level of overprediction.
A way to circumvent the assumption of using a single spec-tral index for di ff erent sources is to extract the spectral indexvalue directly from two surveys and interpolate or extrapolatethe flux density to 54 MHz. In Fig. 8, we show how this ap-proach systematically overestimates, by about 20%, the expectedLoLSS flux density when using NVSS and LoTSS to estimatethe spectral index of the sources. On the other hand, using a sur-vey closer in frequency, such as VLSSr, drastically reduces thee ff ect. This is visible from the second panel of Fig. 8, where theaverage ratio between the extrapolated flux density and that mea-sured in LoLSS is 1.01. Also, the interpolation between 8C andNVSS predicts the LoLSS flux density with an average accuracyof 6%, but based on only 45 sources.As a final cross-check we also tried to predict the flux den-sities of VLSSr sources using LoTSS and NVSS data (Fig. 9).We found that the predicted flux is overestimated by about 25%.This is another way to confirm that the LoLSS and VLSSr fluxscales are in agreement, whilst it shows a disagreement betweenthe LoLSS and LoTSS flux scales. However, this approach hastwo limitations: each source needs to be detected in three sur-veys, reducing the total number of sources, and it relies on theassumption of a pure power law extrapolation. The latter is not agood assumption at 54 MHz, where a number of sources experi-ence a curvature of the spectrum (e.g. de Gasperin et al. 2019).However, the fraction of sources with a curved spectrum is ex-pected to be on the order of ∼ −
30% (Callingham et al. 2017),which should only a ff ect that fraction of sources above the ratio = ff set that we found. One possibility is that LoLSSand VLSSr are both o ff set towards lower flux densities by thesame amount (up to ∼ + NVSS interpolation. Alternatively, the 6C survey, on whichLoTSS flux densities are rescaled (Hardcastle et al. 2020), mightbe o ff set (towards higher flux densities) by ∼ ff set) of6C is estimated to be within ±
5% (Hales et al. 1988), whilst theaccuracy of LoTSS is estimated to be ∼
10% (Shimwell et al.2019). Taking into account these di ff erent tests, we cannot de-rive a more conclusive estimate of the flux density accuracy, butit is reasonable to suggest that assuming a conservative 10% er-ror on the LoLSS flux density scale could be beneficial. Article number, page 12 of 19. de Gasperin et al.: LOFAR LBA sky survey I
Table 3.
Measured flux densities for 3C 295 in various surveys and the expected value following Scaife & Heald (2012).
Survey Frequency Measured Expected Fractional errorname (MHz) flux density (Jy) flux density (Jy) (per cent)LoLSS 54 129.7 133.3 − . − . − . − .
7. Public data release
The data presented in this paper are available online in the jour-nal repository and online in the form of a source catalogue anda mosaic image. The image and catalogue cover a region of 740deg . Of this region, around 500 deg is covered at full depth,whilst the rest is located at the mosaic edges and therefore cov-ered, with a reduced sensitivity. The catalogue contains 25,247 sources. Although we used anadaptive rms box size, a few artefacts around bright sourcesmight still be present, and no attempt has been made to re-move them. The catalogue retains the type of source as derivedby PyBDSF, where it distinguishes isolated compact sources(source_code = ’S’), large complex sources (source_code = ’C’),and sources that are within an island of emission that containsmultiple sources (source_code = ’M’).We note that the catalogue may contain some blendedsources, although the chance of this is low given the sky den-sity. No attempt has been made to correct the PyBDSF catalogueinto physical radio sources (cf. Williams et al. 2019). Further-more, we note that the uncertainties on the source position andon the flux density are derived locally by the source finder fromthe images and do not include the other factors discussed in theprevious Sections. The most conservative approach is to add 2.5 (cid:48)(cid:48) (see Sect. 6.2) in quadrature to the position error and 10% of theflux density in quadrature to the flux error (see Sect. 6.3). Anextract of the catalogue is presented in Table 4.We estimate the completeness of the catalogues followingthe procedure outlined by Heald et al. (2015). For this processwe used the residual mosaic image created after subtracting thesources detected by PyBDSF. This image carries the informa-tion of the distribution of the rms noise of the real mosaic andcan therefore be used to inject fake sources and assess to whatlevel they can be retrieved. We inject a population of 6000 pointsources, randomly distributed, with flux densities ranging be-tween 1 mJy and 10 Jy, and following a number count power-lawdistribution of dNdS ∝ S − . . To simulate ionospheric smearing, thepeak flux density of each source is reduced by 20 percent, whilstits size is increased to preserve the integrated flux density. Wethen attempt the detection of these sources using PyBDSF withthe same parameters used for the catalogue. The process is thenrepeated 50 times to decrease sample noise.We consider a source as detected if it is found to be within25 (cid:48)(cid:48) of its input position and with a recovered flux density that iswithin ten times the error on the recovered flux density from thesimulated value. We found that we have a 50 percent probabilityof detecting sources at 25 mJy and 90 per cent probability of de-tecting sources at 50 mJy. In Fig. 10, we show the completenessover the entire mosaiced region (740 deg ), that is, the fractionof recovered sources above a certain flux density. Our simula- C o m p l e t e n e ss a b o v e f l u x d e n s i t y li m i t F r a c t i o n d e t e c t e d a t a g i v e n f l u x d e n s i t y Fig. 10.
Estimated cumulative completeness of the preliminary data re-lease catalogue (red) and the fraction of simulated sources that are de-tected as a function of flux density (blue), both assuming dNdS ∝ S − . . tions indicate that the catalogue is 50 percent complete over 17mJy and 90 per cent complete over 40 mJy, although we notethat these values for cumulative completeness depend on the as-sumed slope of the input source counts.The mosaic image has about 10 valid pixels, that is the re-gion where at least one primary beam attenuation was higherthan 30%. In the case of pure white noise, with a 5 σ detec-tion limit we expect around 100 false positives. However, thebackground noise of the mosaic image is largely dominated bysystematic e ff ects. To assess the number of false positives, westarted from the mosaic image used to build the catalogue andwe invert its pixel values. Negative pixels due to noise and arte-facts are now positive, whilst its sources are negative. Runningthe source finder with the same parameters used in the originalrun, we evaluated how many artefacts are erroneously consideredlegitimate sources. During this process, we used the same rmsmask produced for the original detection because that evaluationis influenced by the positive pixels. We detected 1055 sources,highly concentrated along the mosaic edges and around brightsources. From this, we conclude that the number of false posi-tive in our catalogue is around 4%.We calculated the Euclidean-normalised di ff erential sourcecounts for the LoLSS catalogue presented here. These are plot-ted in Fig 12. Uncertainties on the final normalised source countswere propagated from the error on the completeness correctionand the Poisson errors (Gehrels 1986) on the raw counts per fluxdensity bin. To account for incompleteness, we used the mea-sured peak intensities to calculate the fractional area of the sur-vey in which each source could be detected, A i . The count ineach flux density bin is then determined as N = (cid:80) / A i . To es-timate an error on this correction, we used the measured uncer- Article number, page 13 of 19 & A proofs: manuscript no. 2020_LBAsurvey
Table 4.
Example of entries in the source catalogue. The entire catalogue contains 25,247 sources. The entries in the catalogue include: sourcename, J2000 right ascension (RA), J2000 declination (Dec), peak brightness (S peak ), integrated flux density (S int ), and the uncertainties on all ofthese values. The catalogue also contains the local noise at the position of the source (rms noise), and the type of source (where ‘S’ indicates anisolated source which is fit with a single Gaussian; ‘C’ represents sources that are fit by a single Gaussian but are within an island of emissionthat also contains other sources; and ‘M’ is used for sources which are extended and fitted with multiple Gaussians). Not listed here, but presentin the catalogue, there is also the estimation of the source size, both with and without the e ff ect of beam convolution. The uncertainties on sourcepositions and the flux densities are derived locally by the source finder and are likely underestimated (see text). Source name RA σ RA DEC σ DEC S peak σ S peak S int σ S int rms noise Type( ◦ ) ( (cid:48)(cid:48) ) ( ◦ ) ( (cid:48)(cid:48) ) (mJy / beam) (mJy / beam) (mJy) (mJy) (mJy / beam)LOLpJ110902.0 + + + + + + + + + + + + + + + + + + + + ff erent spectralindices. The LoLSS counts show good agreement with these pre-viously determined counts, with a transition at around 100 mJyof the average spectral index from − . − . The released image reveals the radio sky at 42–66 MHz with adepth of 4–5 mJy beam − and a resolution of 47 (cid:48)(cid:48) (see Fig. 11).Even at a lower angular resolution and reduced sensitivity ascompared to what will be achieved in the full survey, a num-ber of nearby galaxies, radio galaxies, and galaxy clusters showthe presence of resolved di ff use emission in the survey images.In Fig. 13, we present examples of three nearby galaxies, twogalaxy clusters, and faint di ff use emission surrounding an earlytype galaxy, probably hinting at past nuclear activity. We notethat PyBDSF might not correctly associate all emission of thefew very extended sources in the catalogue, as shown in the lastpanel of Fig. 13.The other large-area survey that explores similar frequenciesat comparable resolution of the preliminary release of LoLSSis VLSSr. That survey reaches an average rms map sensitivityof 130 mJy beam − at a resolution of 80 (cid:48)(cid:48) and a frequency of74 MHz. A comparison between VLSSr and LoLSS is presentedin Fig. 14. The number of sources detected by LoLSS is arounda factor of ten higher. However, the fidelity of extended sources as well as the noise level in the vicinity of the brightest sourcesare still compromised by a missing direction-dependent correc-tion. Therefore, we warn the reader of possible larger errors inthe flux density of extended sources compared to what is esti-mated in the paper. The e ff ects of direction-dependent correc-tion will mitigate this problem. We tested the results of such aprocedure, reprocessing a few pointing of LoLSS with the ex-perimental direction-dependent correction strategy outlined inde Gasperin et al. (2020a). The resulting image is presented inthe large panel of Fig. 14, where the final resolution is 15 (cid:48)(cid:48) andthe noise approaches the thermal noise at 1.3 mJy beam − . Com-pared to the direction-independent calibrated image, 30% moresources are detected in the direction-dependent calibrated image,with a superior image fidelity. This illustrates the potential of thefinal release of LoLSS. A full analysis of all the fields repro-cessed with direction-dependent calibration will be presented ina forthcoming paper.Despite the lower angular resolution and lower sensitivity ofthe released images, a number of projects from those describedin Sect. 2 can still be carried out. Examples include the estima-tion of CRe di ff usion in face-on galaxies such as M51 (secondpanel of Fig. 13, Heesen et al., in prep.) or the search for emis-sion from nearby exoplanets. Large-scale sources in dense en-vironments, such as the nearby galaxy cluster Abell 1314 (lastpanel, Fig. 13), point towards the presence of a large amountof di ff use steep spectrum emission; whilst a careful cross-matchwith other radio surveys and high-energy catalogues show thepotential of LoLSS in the characterisation of blazars (Kadler etal in perp.). Article number, page 14 of 19. de Gasperin et al.: LOFAR LBA sky survey I h m h m R i g h t A s c e n s i o n ( J ) Declination (J2000) S u r f a c e b r i g h t n e ss ( m J y b e a m ) Fig. 11.
Mosaic image of the preliminary release of LoLSS, covering the HETDEX spring field region. Beam size: 47 (cid:48)(cid:48) × (cid:48)(cid:48) .Article number, page 15 of 19 & A proofs: manuscript no. 2020_LBAsurvey S (mJy) S . d N / dS ( J y . s r ) de Zotti (scaled = 0.8 )de Zotti (scaled = 0.6 ) RawCorrected
Fig. 12.
Euclidean-normalised di ff erential source counts for LoLSS be-tween 10 mJy and 30 Jy. The open circles show the raw, uncorrectedsource counts, whilst the filled circles show the counts corrected forcompleteness. For comparison, we show the 1.4 GHz source countsfrom various surveys compiled De Zotti et al. (2010), and scaled to54 MHz, assuming a spectral index of -0.8 (in gray) and -0.6 (in black).
8. Summary
In this work, we present the preliminary release of 740 deg (95pointings) of the LOFAR LBA Sky Survey. The data were pro-cessed using the Pipeline for LOFAR LBA (PiLL) code up tothe point where the correction of direction-independent errorsis complete. The final sensitivity of the preliminary release is4–5 mJy beam − at a resolution of 47 (cid:48)(cid:48) . The catalogue that ac-companies the paper contains more than 25,000 radio sources(detected with a 5 σ rms threshold). We used Monte-Carlo simu-lations to assess the completeness of the catalogue and we con-clude that it is 50% complete for sources above a flux densityof 17 mJy and 90% complete for sources above a flux densityof 40 mJy. We evaluated our astrometric accuracy to be within2.5 (cid:48)(cid:48) (1 σ ). We cross-checked the flux density of the sources inour catalogue with other surveys and found good agreement oninterpolated and extrapolated values. We estimate the flux den-sity scale uncertainty to be within 10%. The data presented inthis paper, as well as the final survey products, are available tothe community for scientific exploitation.The final aim of LoLSS, which we have shown here to bean achievable goal, is to cover the entire northern sky in the fre-quency range 42–66 MHz, at a sensitivity of ∼ − anda resolution of 15 (cid:48)(cid:48) at optimal declination. The full survey willrequire 3170 pointings; currently, all pointings above a declina-tion of 40 ◦ are being observed and these observations should becompleted by mid-2022. We plan to further increase the cover-age to a declination of 20 ◦ and, finally, to a declination of 0 ◦ . Thefinal release of the survey will include a full direction-dependenterror correction as demonstrated in Fig. 14 and de Gasperin et al.(2020a).The final release of the survey will facilitate advances acrossa range of astronomical research areas, as described in this work(see Sect. 2). Together with the higher frequency counterpart at144 MHz (LoTSS), LoLSS will allow for the study of more than1 million low-frequency radio spectra, providing unique insightson physical models for galaxies, active nuclei, galaxy clusters,and other fields of research. This experiment represents a uniqueattempt to explore the ultra-low frequency sky at a high angularresolution and depth. Thanks to its optimal combination of reso-lution and sensitivity, the LOFAR LBA Sky Survey will remainunique well into the Square Kilometre Array (SKA) era. Acknowledgements.
LOFAR is the LOw Frequency ARray designed and con-structed by ASTRON. It has observing, data processing, and data storage facil-ities in several countries, which are owned by various parties (each with theirown funding sources), and are collectively operated by the ILT foundation undera joint scientific policy. The ILT resources have benefited from the following re-cent major funding sources: CNRS-INSU, Observatoire de Paris and Universitéd’Orleans, France; BMBF, MIWF-NRW, MPG, Germany; Science FoundationIreland (SFI), Department of Business, Enterprise and Innovation (DBEI), Ire-land; NWO, The Netherlands; The Science and Technology Facilities Council,UK; Ministry of Science and Higher Education, Poland; Istituto Nazionale diAstrofisica (INAF). This research has made use of the University of Hertford-shire high-performance computing facility ( https://uhhpc.herts.ac.uk/ )and the LOFAR-UK compute facility, located at the University of Hertfordshireand supported by STFC [ST / P000096 / / R000905 / / R000972 /
1. IP acknowledges support from INAF under the SKA / CTA
Article number, page 16 of 19. de Gasperin et al.: LOFAR LBA sky survey I h m m m ° ' ' ' ' ' ' pos.eq.ra D e c li n a t i o n ( J ) h m m m ° ' ' ' ' pos.eq.ra h m m m ° ' ' ' ' ' pos.eq.ra D e c li n a t i o n ( J ) h m ° ' ' ' pos.eq.ra h m m ° ' ' ' Right Ascension (J2000) D e c li n a t i o n ( J ) h m m m m ° ' ' ' ° ' Right Ascension (J2000) ) Fig. 13.
Some examples of extended sources in the data release published with this paper. From top-left to bottom-right, the nearby galaxies M101,M51, M106, the galaxy cluster Abell 1550, the emission surrounding the early type galaxy MCG + ff use sources in the ICM in Abell 1314. Contours start at three times the local rms noise. In the last panel,the green dashed lines show the gaussians used by the source finder to model the brightness distribution of the radio sources, the blue regions showthe location and size of the sources, composed by one or more Gaussian components, as they are present in the catalogue.Article number, page 17 of 19 & A proofs: manuscript no. 2020_LBAsurvey h m m m m m m ° ' ° ' ' ' Right Ascension (J2000) D e c li n a t i o n ( J ) ) Fig. 14.
Region of the LOFAR LBA sky survey re-imaged using direction-dependent calibration (rms noise: 1.3 mJy beam − - beam: 15 (cid:48)(cid:48) × (cid:48)(cid:48) ).Top right: Same field without direction-dependent calibration (as presented in this paper; rms noise: 3 mJy beam − - beam: 46 (cid:48)(cid:48) × (cid:48)(cid:48) ). Green andblue regions show the location of the identified sources as described in 13. Bottom right: Same field in the VLSSr survey (rms noise: 73 mJy beam − - beam: 80 (cid:48)(cid:48) × (cid:48)(cid:48) ). PRIN “FORECaST” and the PRIN MAIN STREAM “SAuROS” projects. ABacknowledges support from ERC Stg DRANOEL n. 714245 and MIUR FAREgrant “SMS”. IP and MV acknowledge support from the Italian Ministry of For-eign A ff airs and International Cooperation (MAECI Grant Number ZA18GR02)and the South African Department of Science and Technology’s National Re-search Foundation (DST-NRF Grant Number 113121) as part of the ISARP RA-DIOSKY2020 Joint Research Scheme. References
Acero, F., Ackermann, M., Ajello, M., et al. 2015, Astrophys. Journal, Suppl.Ser., 218, 23Arias, M., Vink, J., de Gasperin, F., et al. 2018, A&A, 612, A110Becker, R. H., White, R. L., & Helfand, D. J. 1995, Astrophys. J., 450, 559Birzan, L., McNamara, B., Nulsen, P., Carilli, C., & Wise, M. W. 2008, Astro-phys. J., 686, 859Bliton, M., Rizza, E., Burns, J. O., Owen, F. N., & Ledlow, M. J. 1998, MNRAS,301, 609Bock, D. C.-J., Large, M. I., & Sadler, E. M. 1999, Astron. J., 117, 1578Botteon, A., Shimwell, T. W., Bonafede, A., et al. 2018, MNRAS, 478, 885 Botteon, A., van Weeren, R. J., Brunetti, G., et al. 2020, Mon. Not. R. Astron.Soc. Lett., 000, 1Bridle, A. H. & Schwab, F. R. 1989, in Synth. Imaging Radio Astron., Vol. 6,247Brienza, M., Godfrey, L., Morganti, R., et al. 2017, A&A, 606, A98Brüggen, M., Reiprich, T., Bulbul, E., et al. 2020, A&ABrunetti, G., Giacintucci, S., Cassano, R., et al. 2008, Nature, 455, 944Brunetti, G. & Jones, T. W. 2014, Int. J. Mod. Phys. D, 23, 30007Brunetti, G. & Vazza, F. 2020, Phys. Rev. Lett., 124, 051101Calistro-Rivera, G., Williams, W. L., Hardcastle, M. J., et al. 2017, MNRAS,469, 3468Callingham, J. R., Ekers, R. D., Gaensler, B. M., et al. 2017, Astrophys. J., 836Callingham, J. R., Tuthill, P. G., Pope, B. J., et al. 2019, Anisotropic winds in aWolf–Rayet binary identify a potential gamma-ray burst progenitorCassano, R., Brunetti, G., Rottgering, H. J. A., & Bruggen, M. 2010, A&A, 509,68Chyzy, K. T., Jurusik, W., Piotrowska, J., et al. 2018, A&A, 619, A36Condon, J. J., Cotton, W. D., Greisen, E. W., et al. 1998, Astron. J., 8065, 1693Dabhade, P., Röttgering, H. J., Bagchi, J., et al. 2019, Giant radio galaxies in theLOFAR Two-metre Sky Survey-ID’Antonio, D., Giroletti, M., Giovannini, G., & Maini, A. 2019, MNRAS, 490,5798
Article number, page 18 of 19. de Gasperin et al.: LOFAR LBA sky survey I de Gasperin, F., Brunetti, G., Brüggen, M., et al. 2020a, A&A, 642, A85de Gasperin, F., Dijkema, T. J., Drabent, A., et al. 2019, A&A, 5, A5de Gasperin, F., Intema, H. T., & Frail, D. A. 2018a, MNRAS, 474, 5008de Gasperin, F., Intema, H. T., Shimwell, T. W., et al. 2017, Sci. Adv., 3,e1701634de Gasperin, F., Lazio, T. J. W., & Knapp, M. 2020b, A&A, 644, A157de Gasperin, F., Mevius, M., Ra ff erty, D. A., Intema, H. T., & Fallows, R. A.2018b, A&A, 615, A179de Gasperin, F., Vink, J., McKean, J. P., et al. 2020c, A&A, 635, A150De Toma, G., Gibson, S., Emery, B., & Kozyra, J. 2010, AIP Conf. Proc., 1216,667De Zotti, G., Massardi, M., Negrello, M., & Wall, J. 2010, Radio and millimetercontinuum surveys and their astrophysical implicationsDriessen, L. N., Domˇcek, V., Vink, J., et al. 2018, Astrophys. J., 860, 133Emig, K. L., Salas, P., De Gasperin, F., et al. 2020, A&A, 634, 138Falcke, H., Körding, E., & Nagar, N. M. 2004, New Astron. Rev., 48, 1157Gehrels, N. 1986, Astrophys. J., 303, 336Giroletti, M., Massaro, F., D’Abrusco, R., et al. 2016, A&A, 588, 141Govoni, F., Orrù, E., Bonafede, A., et al. 2019, Science (80-. )., 364, 981Gurkan, G., Hardcastle, M. J., Smith, D. J. B., et al. 2018, MNRAS, 475, 3010Hales, S. E. G., Baldwin, J. E., & Warner, P. J. 1988, MNRAS, 234, 919Hardcastle, M., Shimwell, T., Tasse, C., et al. 2020, eprint arXiv:2011.08294Hardcastle, M. J., Williams, W. L., Best, P. N., et al. 2019, A&A, 622, A12Heald, G. H., Pizzo, R. F., Orrú, E., et al. 2015, A&A, 582, A123Heckman, T. & Best, P. 2014, Annu. Rev. Astron. Astrophys., 52, 589Heesen, V., Croston, J. H., Morganti, R., et al. 2018, MNRAS, 474, 5049Helmboldt, J. F. & Hurley-Walker, N. 2020, Radio Sci., 55Hess, S. L. & Zarka, P. 2011, A&A, 531, 29Hill, G. J., Gebhardt, K., Komatsu, E., et al. 2008, Panor. Views Galaxy Form.Evol., 399, 115Hurley-Walker, N., Callingham, J. R., Hancock, P. J., et al. 2017, MNRAS, 464,1146Hurley-Walker, N., Filipovi´c, M. D., Gaensler, B. M., et al. 2019, Publ. Astron.Soc. Aust., 36, e045Intema, H. T., Jagannathan, P., Mooley, K. P., & Frail, D. A. 2017, A&A, 598,A78Intema, H. T., van der Tol, S., Cotton, W. D., et al. 2009, A&A, 501, 1185Israel, F. P. & Mahoney, M. J. 1990, Astrophys. J., 352, 30Jurlin, N., Morganti, R., Brienza, M., et al. 2020, A&A, 638, A34Kang, H., Petrosian, V., Ryu, D., & Jones, T. W. 2014, Astrophys. J., 788Lacy, M., Baum, S. A., Chandler, C. J., et al. 2020, Publ. Astron. Soc. Pacific,132, 035001Lane, W. M., Cotton, W. D., van Velzen, S., et al. 2014, MNRAS, 440, 327Lenc, E., Anderson, C. S., Barry, N., et al. 2017, Publ. Astron. Soc. Aust., 34,e040Lynch, C. R., Lenc, E., Kaplan, D. L., Murphy, T., & Anderson, G. E. 2017,Astrophys. J., 836, L30Mahatma, V. H., Hardcastle, M. J., Croston, J. H., et al. 2020, Investigating thespectral age problem with powerful radio galaxiesMahatma, V. H., Hardcastle, M. J., Williams, W. L., et al. 2018, MNRAS, 475,4557Mandal, S., Intema, H. T., van Weeren, R. J., et al. 2020, A&A, 634, A4Mangum, J. G. & Wallace, P. 2015, Publ. Astron. Soc. Pacific, 127, 74Massaro, F., Giroletti, M., Paggi, A., et al. 2013, Astrophys. Journal, Suppl. Ser.,208, 15McKean, J. P., Godfrey, L. E. H., Vegetti, S., et al. 2016, MNRAS, 463, 3143Mevius, M., van der Tol, S., Pandey, V. N., et al. 2016, Radio Sci., 51, 927Mezger, P. G. 1978, A&A, 70, 565Miley, G. & De Breuck, C. 2008, Astron. Astrophys. Rev., 15, 67Mingo, B., Croston, J. H., Hardcastle, M. J., et al. 2019, MNRAS, 488, 2701Miskolczi, A., Heesen, V., Horellou, C., et al. 2019, A&A, 622, 9Mohan, N. & Ra ff erty, D. 2015, Astrophys. Source Code Libr., 1502.007Mooney, S., Quinn, J., Callingham, J. R., et al. 2019, Blazars in the LOFARTwo-Metre Sky Survey first data releaseMorganti, R., Oosterloo, T. A., Brienza, M., et al. 2020, arXiv:2011.08239Mulcahy, D. D., Horne ff er, A., Beck, R., et al. 2014, A&A, 568, A74Niklas, S. & Beck, R. 1997, A&A, 320, 54Norris, R. P., Hopkins, A. M., Afonso, J., et al. 2011, Publ. Astron. Soc. Aust.,28, 215O’Dea, C. P. & Saikia, D. J. 2020, Compact steep-spectrum and peaked-spectrumradio sourcesO ff ringa, A. R. 2016, A&A, 595, A99O ff ringa, A. R., De Bruyn, A. G., Zaroubi, S., & Biehl, M. 2010, in Proc. Sci.,Vol. 107, 2089O ff ringa, A. R., McKinley, B., Hurley-Walker, N., et al. 2014, MNRAS, 444, 606O ff ringa, A. R. & Smirnov, O. 2017, MNRAS, 471, 301Polderman, I. M., Haverkorn, M., & Ja ff e, T. R. 2020, eprint arXiv:2002.09204Polderman, I. M., Haverkorn, M., Ja ff e, T. R., & Alves, M. I. 2019, A&A, 621,A127Rees, N. 1990, MNRAS, 244, 233 Rengelink, R. B., Tang, Y., de Bruyn, a. G., et al. 1997, Astron. Astrophys.Suppl., 124, 259Roger, R. S., Costain, C. H., & Bridle, A. H. 1973, Astron. J., 78, 1030Rottgering, H., Afonso, J., Barthel, P., et al. 2011, J. Astrophys. Astron., 32, 557Sabater, J., Best, P. N., Tasse, C., et al. 2020, A&ASa ff , E. B. & Kuijlaars, A. B. 1997, Math. Intell., 19, 5Saxena, A., Marinello, M., Overzier, R. A., et al. 2018, MNRAS, 480, 2733Saxena, A., Röttgering, H. J. A., & Rigby, E. E. 2017, MNRAS, 469, 4083Scaife, A. M. M. & Heald, G. H. 2012, MNRAS, 423, 30Shimwell, T. W., Röttgering, H. J. A., Best, P. N., et al. 2017, A&A, 598, A104Shimwell, T. W., Tasse, C., Hardcastle, M. J., et al. 2019, A&A, 622, A1Shulevski, A., Barthel, P. D., Morganti, R., et al. 2019, LOFAR first look at thegiant radio galaxy 3C 236Smith, D. J. B., Best, P. N., Duncan, K. J., et al. 2016, SF2A-2016 Proc. Annu.Meet. French Soc. Astron. Astrophys., 271Smith, D. J. B., Haskell, P., Gürkan, G., et al. 2020, A&ASreeja, V. & Aquino, M. 2014, J. Atmos. Solar-Terrestrial Phys., 120, 96Stewart, A. J., Fender, R. P., Broderick, J. W., et al. 2016, MNRAS, 456, 2321Su, H., Hurley-Walker, N., Jackson, C. A., et al. 2017, MNRAS, 465, 3163Tabatabaei, F. S., Schinnerer, E., Krause, M., et al. 2017, Astrophys. J., 836, 185Tasse, C., Shimwell, T., Hardcastle, M., et al. 2020, A&ATrüstedt, J., Kadler, M., Brüggen, M., et al. 2014, in Proc. Sci., 43Tumlinson, J., Peeples, M. S., & Werk, J. K. 2017, The Circumgalactic MediumTurner, J. D., Zarka, P., Grießmeier, J.-M., et al. 2020, eprint arXiv:2012.07926Van Der Tol, S., Veenboer, B., & O ff ringa, A. R. 2018, A&A, 616, 27van Haarlem, M. P., Wise, M. W., Gunst, a. W., et al. 2013, A&A, 556, A2van Weeren, R. J., de Gasperin, F., Akamatsu, H., et al. 2019, Space Sci. Rev.,215van Weeren, R. J., Röttgering, H. J. A., Ra ff erty, D. A., et al. 2012, A&A, 543,A43Vedantham, H. K., Callingham, J. R., Shimwell, T. W., et al. 2020a, Astrophys.J., 903, L33Vedantham, H. K., Callingham, J. R., Shimwell, T. W., et al. 2020b, Nat. Astron.,1Vedantham, H. K. & Koopmans, L. V. 2015, MNRAS, 453, 925Vollmann, M., Heesen, V., Shimwell, T. W., et al. 2020, MNRAS, 496, 2663White, R. L., Becker, R. H., Helfand, D. J., & Gregg, M. D. 1997, Astrophys. J.,475, 479Williams, W. L., Hardcastle, M. J., Best, P. N., et al. 2019, A&A, 622, A2erty, D. A., et al. 2012, A&A, 543,A43Vedantham, H. K., Callingham, J. R., Shimwell, T. W., et al. 2020a, Astrophys.J., 903, L33Vedantham, H. K., Callingham, J. R., Shimwell, T. W., et al. 2020b, Nat. Astron.,1Vedantham, H. K. & Koopmans, L. V. 2015, MNRAS, 453, 925Vollmann, M., Heesen, V., Shimwell, T. W., et al. 2020, MNRAS, 496, 2663White, R. L., Becker, R. H., Helfand, D. J., & Gregg, M. D. 1997, Astrophys. J.,475, 479Williams, W. L., Hardcastle, M. J., Best, P. N., et al. 2019, A&A, 622, A2