Extragalactic Peaked-Spectrum Radio Sources at Low Frequencies

3 MB·

by  J. R. Callingham

ACCEPTEDTOAPJ2017JAN.04
PreprinttypesetusingLATEXstyleemulateapjv.04/17/13
EXTRAGALACTICPEAKED-SPECTRUMRADIOSOURCESATLOWFREQUENCIES
J.R.CALLINGHAM1,2,3,R.D.EKERS2,B.M.GAENSLER4,1,3,J.L.B.LINE6,3,N.HURLEY-WALKER5,E.M.SADLER1,3,
S.J.TINGAY7,5,P.J.HANCOCK5,3,M.E.BELL2,3,K.S.DWARAKANATH8,B.-Q.FOR9,T.M.O.FRANZEN5,L.HINDSON10,
M.JOHNSTON-HOLLITT10,A.D.KAPIN´SKA9,3,E.LENC1,3,B.MCKINLEY6,3,J.MORGAN5,A.R.OFFRINGA11,P.PROCOPIO6,3,
L.STAVELEY-SMITH9,3,R.B.WAYTH5,3,C.WU9,Q.ZHENG10
1SydneyInstituteforAstronomy(SIfA),SchoolofPhysics,TheUniversityofSydney,NSW2006,Australia
2CSIROAstronomyandSpaceScience(CASS),Marsfield,NSW2122,Australia
3ARCCentreofExcellenceforAll-SkyAstrophysics(CAASTRO)
4DunlapInstituteforAstronomy&Astrophysics,UniversityofToronto,Toronto,ON,M5S3H4,Canada
7 5InternationalCentreforRadioAstronomyResearch(ICRAR),CurtinUniversity,Bentley,WA6102,Australia
1 6SchoolofPhysics,TheUniversityofMelbourne,Parkville,VIC3010,Australia
0 7IstitutoNazionalediAstrofisica(INAF),IstitutodiRadioastronomia,ViaPieroGobetti,Bologna,40129,Italy
2 8RamanResearchInstitute(RRI),Bangalore560080,India
9InternationalCentreforRadioAstronomyResearch(ICRAR),TheUniversityofWesternAustralia,Crawley,WA6009,Australia
n 10SchoolofChemical&PhysicalSciences,VictoriaUniversityofWellington,Wellington6140,NewZealandand
a 11NetherlandsInstituteforRadioAstronomy(ASTRON),Dwingeloo,TheNetherlands
J
AcceptedtoApJ2017Jan.04

0
1 ABSTRACT

We present a sample of 1,483 sources that display spectral peaks between 72MHz and 1.4GHz, selected
]
A from the GaLactic and Extragalactic All-sky Murchison Widefield Array (GLEAM) survey. The GLEAM
surveyisthewidestfractionalbandwidthall-skysurveytodate,idealforidentifyingpeaked-spectrumsources
G
atlowradiofrequencies. Ourpeaked-spectrumsourcesarethelowfrequencyanaloguesofgigahertz-peaked
. spectrum(GPS)andcompact-steepspectrum(CSS)sources,whichhavebeenhypothesizedtobetheprecursors
h
tomassiveradiogalaxies. Oursamplemorethandoublesthenumberofknownpeaked-spectrumcandidates,
p
and95%ofoursamplehaveanewlycharacterizedspectralpeak. WehighlightthatsomeGPSsourcespeaking
-
o above5GHzhavehadmultipleepochsofnuclearactivity,anddemonstratethepossibilityofidentifyinghigh
r redshift (z > 2) galaxies via steep optically thin spectral indices and low observed peak frequencies. The
t
s distributionoftheopticallythickspectralindicesofoursampleisconsistentwithpastGPS/CSSsamplesbut
a withalargedispersion,suggestingthatthespectralpeakisaproductofaninhomogeneousenvironmentthatis
[ individualistic. Wefindnodependenceofobservedpeakfrequencywithredshift,consistentwiththepeaked-

spectrum sample comprising both local CSS sources and high-redshift GPS sources. The 5GHz luminosity
1
v distribution lacks the brightest GPS and CSS sources of previous samples, implying that a convolution of
1 sourceevolutionandredshiftinfluencesthetypeofpeaked-spectrumsourcesidentifiedbelow1GHz. Finally,
7 wediscusssourceswithopticallythickspectralindicesthatexceedthesynchrotronself-absorptionlimit.
7 Keywords: galaxies: active — radiation mechanisms: general — radio continuum: general — radio sources:
2 spectra
0
.
1
1. INTRODUCTION arethoughttohavethelowestpeakfrequencies(cid:46)500MHz,
0
butuntilrecentlylowradiofrequencyobservationshavebeen
7 Gigahertz-peakedspectrum(GPS),compactsteepspectrum
lackingtoconfirmsuchasituation(Fantietal.1990).
1 (CSS),andhighfrequencypeaked(HFP)sourcesareaclass
The argument that GPS, CSS, and HFP sources represent
: ofradio-loudactivegalacticnuclei(AGN)thathavebeenar-
v the first stages of radio-loud AGN evolution was inferred
guedtobetheyoungprecursorstomassiveradio-loudAGN,
i byverylongbaselineinterferometry(VLBI)observationsof
X suchasCentaurusAandCygnusA(Fantietal.1990;O’Dea
thesesources,revealingsmallscalemorphologiesreminiscent
et al. 1991; Dallacasa et al. 2000; Tinti et al. 2005; Kunert-
r
oflargescaleradiolobesofpowerfulradiogalaxies,withtwo
a Bajraszewska et al. 2010). GPS and HFP sources are de-
steep-spectralobessurroundingaflatspectrumcore(Phillips
finedashavingapeakintheirradiospectraandsteepspectral
&Mutel1980;Wilkinsonetal.1994;Stanghellinietal.1997;
slopeseithersideofthepeak. Theyarealsooftenfoundwith
Orienti et al. 2006; An et al. 2012). Additional multi-epoch
smalllinearsizesandlowradiopolarizationfractions(O’Dea
VLBIobservationsmeasuredthemotionofthehotspots,pro-
etal.1991). CSSsourcesarethoughttobearelatedclassthat
vidingindirectevidenceforages(cid:46)105yrs(Owsianik&Con-
hassimilarpropertiestoGPSandHFPsourcesbutpeakfre-
way1998;Polatidis&Conway2003;Gugliuccietal.2005).
quenciesbelowthetraditionalgigahertzselectionfrequencies
The ‘youth’ scenario for these sources is further supported
(Fanti et al. 1990). Hence, the main differentiation between
by high frequency spectral break modeling (Murgia et al.
GPS, CSS and HFP sources is the frequency of the spectral
1999; Orienti et al. 2010) and the discovery of the empiri-
peak and the largest linear size. GPS and HFP sources have
linearsizes(cid:46)1kpcandpeakfrequenciesof 1 5GHz,and calrelationbetweenrestframeturnoverfrequenciesandlin-
(cid:38)5GHz,respectively(O’Dea1998;Dallaca∼sae−tal.2000).In ear size (O’Dea & Baum 1997; O’Dea 1998). This suggests
HFP sources evolve into GPS sources, which in turn evolve
comparison,CSSsourceshavelinearsizesof 1 20kpcand
∼ − into CSS sources, and then finally grow to reach the size of
[email protected] FRIandFRIIradiogalaxies(Fanaroff&Riley1974;Kunert-
2 CALLINGHAMETAL.
Bajraszewskaetal.2010). moderatesamplesizes.
However, the ‘frustration’ hypothesis, which implies that We have entered a new era in radio astronomy where the
these sources are not young but are confined to small spa- limitations of small fractional bandwidth at low radio fre-
tialscalesduetounusuallyhighnuclearplasmadensity(van quencieshavebeenlifted,withtheMurchisonWidefieldAr-
Breugel et al. 1984; Bicknell et al. 1997), has seen a resur- ray (MWA; Tingay et al. 2013), the Giant Metrewave Radio
genceinexplainingthepropertiesoftheGPS,CSS,andHFP Telescope (GMRT; Swarup 1991), and the LOw-Frequency
population (e.g. Peck et al. 1999; Kameno et al. 2000; Marr ARray (LOFAR; van Haarlem et al. 2013) now operational.
et al. 2001; Tingay & de Kool 2003; Orienti & Dallacasa With the all-sky surveys at these facilitates nearing comple-
2008;Marretal.2014;Tingayetal.2015;Callinghametal. tion,suchastheGaLacticandExtragalacticAll-skyMurchi-
2015). The primary reasons that a debate remains about the sonWidefieldArray(GLEAM;Waythetal.2015)survey,the
nature of GPS, CSS, and HFP sources is because there ap- TIFRGMRTSkySurvey(TGSS;Intemaetal.2016),andthe
pearstobeanoverabundanceofthesesourcesrelativetothe LOFARMultifrequencySnapshotSkySurvey(MSSS;Heald
numberoflargeradioAGN(O’Dea&Baum1997;Readhead et al. 2015), astronomers now have unprecedented access to
etal.1996;Snellenetal.2000;An&Baan2012),anddetailed the radio sky below 300MHz. While the first surveys in ra-
spectralandmorphologicalstudiesofindividualsourceshave dioastronomywereconductedatlowfrequencies(e.g.Mills
demonstrated that several of these sources are confined to et al. 1958; Edge et al. 1959), MSSS and the GLEAM sur-
a small spatial scale due to a dense ambient medium and vey represent a significant step forward in the field because
a cessation of AGN activity (e.g. Peck et al. 1999; Orienti they have surveyed the sky with wide fractional bandwidths
et al. 2010; Callingham et al. 2015). It is also possible that and much higher sensitivity. In particular, the GLEAM sur-
boththe‘youth’and‘frustration’scenariosmayapplytothe veyrepresentsthewidestcontinuousfractionalbandwidthall-
GPS,CSS,andHFPpopulation,sincesourceswithintermit- skysurveyeverproduced,withtwentycontemporaneousflux
tent AGN activity may never break through a dense nuclear densitymeasurementsbetween72and231MHzforapproxi-
mediumbutyoungsourceswithconstantAGNactivitycould mately300,000sourcesintheextragalacticcatalogproduced
evolve past the inner region of the host galaxy (An & Baan fromtheGLEAMsurvey(Hurley-Walkeretal.2017).Hence,
2012). the GLEAM extragalactic catalog is a rich dataset to study
OnemethodthatcandeducewhetheraGPS,CSS,orHFP sourcesthathaveaspectralpeakatlowfrequencies.
source is frustrated or young is by identifying whether syn- ThepurposeofthispaperistousetheGLEAMextragalac-
chrotronself-absorption(SSA)orfree-freeabsorption(FFA) ticcatalogtoconstructthelargestsampleofpeaked-spectrum
isresponsiblefortheturnoverintheradiospectrum(e.g.Tin- sources to date, with contemporaneous observations at and
gay & de Kool 2003; Marr et al. 2014; Callingham et al. belowthespectralpeak. Suchastatisticallysignificantsam-
2015). This is because the turnover in a source’s spectrum ple has unparalleled frequency coverage below the turnover
will likely be dominated by FFA when confined to a small of GPS, CSS and HFP sources, providing a database for a
spatialscalebyadensemedium(Bicknelletal.1997;Kuncic comprehensive spectral comparison of the different absorp-
et al. 1998). To successfully discriminate between SSA and tion models, a test of whether sources with low frequency
FFA requires comprehensively sampling the spectrum of the spectral peaks are preferentially found at high redshift, and
source below the turnover, ideally with the observations be- analysisofhowmanypeaked-spectrumsourcesconstitutethe
lowthespectralpeakoccurringsimultaneously(O’Deaetal. widerradio-loudAGNatlowradiofrequencies. Notethatin
1991;O’Dea&Baum1997). Previousstudiesofsamplesof this paper we use the term ‘peaked-spectrum’ to collectively
GPS,CSS,andHFPsourceshaveoftenhadonlyasingleflux refer to GPS, CSS, HFP, and MPS sources, and the terms
densitymeasurementbelowthespectralpeak,andcomposed ‘spectralpeak’and‘spectralturnover’interchangeably.
ofmulti-epochdatawithsparsefrequencysampling,suchthat Therelevantsurveysusedinthesourceselection,thecross-
differentiation between FFA and SSA have been ambiguous matchingroutine,andspectralmodelingprocedureperformed
forsourcesinlargesamples(e.g.O’Dea1998;Snellenetal. are outlined in §2, §3, and §4, respectively. In §5, we dis-
2000; Dallacasa et al. 2000; Snellen et al. 2002; Edwards & cuss the selection criteria implemented to identify peaked-
Tingay 2004; Randall et al. 2011). Other methods of differ- spectrum sources. Comparisons of the identified peaked-
entiating between SSA and FFA, such as spectral variability spectrumsourcestoknownGPS,CSS,andHFPsources,and
(Tingayetal.2015)orthechangeincircularpolarizationover USSsources, arepresentedin§6and§7, respectively. Rel-
the spectral peak (Melrose 1971), have also suffered from evant observed and intrinsic spectral features of the peaked-
havingincomplete,multi-epochdatabelowtheturnover. spectrum samples are outlined in §8. Finally, we introduce
InadditiontoGPS,CSS,andHFPsources,therehavebeen and debate the nature of sources with radio spectra near the
recentstudiesofarelatedclassofradio-loudAGNreferredto limit of SSA in §9. In this paper we adopt the standard
as megahertz-peaked spectrum (MPS) sources (Falcke et al. Lambda Cold Dark Matter (ΛCDM) cosmological model,
2004; Coppejans et al. 2015, 2016). These sources have the with parameters Ω = 0.27, Ω = 0.73, and Hubble con-
M Λ
samespectralshapeasGPS,CSS,andHFPsourcesbuthave stantH =70kms−1Mpc−1(Hinshawetal.2013).
0
an observed turnover frequency below 1GHz. MPS sources
arebelievedtobeacombinationofnearbyCSSsources,and 2. DESCRIPTIONSOFTHESURVEYSUSEDINTHESELECTIONOF
GPSandHFPsourcesathighredshiftsuchthattheturnover PEAKED-SPECTRUMSOURCES
frequencyhasshiftedbelowagigahertzduetocosmological Thesensitivityandfrequencycoverageofthesurveysused
evolution (Coppejans et al. 2015). In particular, Coppejans for selecting peaked-spectrum sources impact the type of
et al. (2015) and Coppejans et al. (2016) have demonstrated sources identified. Most previous studies (e.g. Fanti et al.
that low-radio frequency selection criteria can identify non- 1990; O’Dea et al. 1991; Stanghellini et al. 1998; Snellen
beamed sources located at z > 2 and which appear young etal.1998;Dallacasaetal.2000;Kunert-Bajraszewskaetal.
due to their small linear size. So far, investigations of MPS 2010)usedsurveysthatobservedtheskyatasinglefrequency
sources have been limited to small sections of the sky and around or above 1GHz. Since the inverted spectrum below
EXTRAGALACTICPEAKED-SPECTRUMRADIOSOURCESATLOWFREQUENCIES 3
thespectralturnoverdistinguishesapeaked-spectrumsource, 104 3C
thelowfrequencysurveyusedforselectiondictatesthetype
MSH
ofsourcesidentified,whilehigherfrequencysurveysareused
PKS
toconfirmtheturnoverandmeasurethespectralslopeinthe
opticallythinregime.Additionally,itisidealifthehigherfre-
CCA MRC 2Jy
quencysurveyshavebettersensitivitiescomparedtothelow
frequencysurveytoensurethattheselectedpeaked-spectrum y) 103 CCA
J
samplehasacompletenesssetonlybythelowfrequencydata. m 4C
The GLEAM extragalactic catalog represents a signifi- (
y TXS
cant advance in selecting peaked-spectrum sources, since it sit VLSSr
is constituted of sources that were contemporaneously sur- en 7C
D
veyedwiththewidestfractionalradiobandwidthtodate,with
twentyfluxdensitymeasurementsbetween72and231MHz. ux 102
With such frequency coverage, selection below and at the Fl
spectral peak can be performed solely using one survey. To ng GLEAM
increasethevalidityofthedetectionofthepeakinthespec- miti 8P7MGNB AT20G
tra of these sources, to remove flat-spectrum sources, and Li WENSS
to measure the slope above the turnover, we also use the TGSS
101
NRAO VLA Sky Survey (NVSS; Condon et al. 1998) and
theSydneyUniversityMolongloSkySurvey(SUMSS;Bock SUMSS
et al. 1999; Mauch et al. 2003). Since the combination of
NVSS
NVSSandSUMSScovertheentireGLEAMsurvey,andare
an order of magnitude more sensitive, this study is sensi- 102 103 104
tivetopeaked-spectrumsourcesthatpeakanywherebetween Frequency (MHz)
72MHz and 843MHz/1.4GHz. Examples of the types of
Figure1. Thedifferentfrequenciesandlimitingsensitivitiesforthemajor
peaked-spectrumsourcesthatthisstudyandpreviousstudies
radiosurveys. TheGLEAMsurveyisshownasablacklineduetoitsvari-
identify are highlighted in Figure 1. Details of the surveys ablelimitingsensitivitiesbetween72and231MHz. Theblue,red,andor-
usedtoselectpeaked-spectrumsourcesinthisstudyarepro- angecurvesrepresenttheSSAspectraofpeaked-spectrumsourcesthatpeak
videdbelow. at200,1000,and10000MHz,respectively.Therefore,thisstudyissensitive
tosuchpeaked-spectrumsourcesportrayedbythebluespectrumbutnotto
2.1. GaLacticandExtragalacticAll-skyMurchisonWidefield previouslyidentifiedpeaked-spectrumsourcesportrayedbytheredororange
spectra. Theotherplottedsurveysnotpreviouslyintroducedareasfollows:
Array(GLEAM)Survey
Mills,Slee,andHill(MSH;Millsetal.1958,1960,1961)survey,Cambridge
The GLEAM survey was formed from observations con- 3C(Edgeetal.1959),4C(Pilkington&Scott1965),and7C(Halesetal.
2007)surveys,Culgooracirculararray(CCA;Slee1995)survey,VLALow-
ducted by the MWA, which surveyed the sky between 72
frequencySkySurveyredux(VLSSr;Laneetal.2014)survey,Texassurvey
and 231MHz from August 2013 to July 2014 (Wayth et al. (TXS;Douglasetal.1996),Parkes(PKS;Ekers1969)survey,MolongloRef-
2015). TheMWAisalowradiofrequencyaperturearraythat erenceCatalogue(MRC;Largeetal.1981,1991),WesterborkNorthernSky
is composed of 128 32-dipole antenna “tiles” spread over a Survey(WENSS;Rengelinketal.1997),2Jysurvey(Wall&Peacock1985),
Parkes-MIT-NRAO(PMN;Gregoryetal.1994;Wrightetal.1994)survey,
10kmareainWesternAustralia(Tingayetal.2013). The
MIT-GreenBank5GHz(87GB;Gregory&Condon1991)survey,andAus-
≈
extragalactic catalog formed from the GLEAM survey con- traliaTelescope20GHz(AT20G;Murphyetal.2010)survey.
sistsof307,455sourcessouthofdeclination+30◦,excluding
Galactic latitudes b < 10◦, the Magellanic clouds, within
9◦ of Centaurus A|,|and a 859 square degree section of the multi-frequency synthesis applied across the instantaneous
skycenteredatRA23handdeclination+15◦. Thepositions bandwidth, and then CLEANed to the first negative CLEAN
of the sources reported are accurate to within 30(cid:48)(cid:48)and the component. Aself-calibrationloopwasthenappliedtoeach
≈
catalog is 90% complete at 0.16Jy (Hurley-Walker et al. of the images. The shallowly CLEANed 30.72MHz band-
≈
2017). The sources in the catalog have twenty flux density width observations were divided into four 7.68-MHz sub-
measurementsbetween72and231MHz,mostlyseparatedby bands and jointly CLEANed, resulting in a RMS of 100 to
7.68MHz. 20mJybeam−1for72to231MHz,respectively. ≈
≈
Whilethedatareductionprocessthatwasperformedtopro- The408MHzMolongloReferenceCatalogue(MRC;Large
ducetheGLEAMextragalacticcatalogisdiscussedindetail etal.1981,1991),scaledtotherespectivefrequency,wasused
inHurley-Walkeretal.(2017),wesummarizethedetailshere tosetaninitialfluxdensityscalefortheimagesandtoapply
consideringtheimportanceoftheGLEAMextragalacticcat- anastrometriccorrection. Thesnapshotsforanobserveddec-
alogtothisstudy. Atsevenindependentdeclinationsettings, lination strip were mosaicked, with each snapshot weighted
the GLEAM survey employed a two-minute “snapshot” ob- by the square of the primary beam response. Due to inac-
serving mode. COTTER (Offringa et al. 2015) was used to curacies in the primary beam model, the remaining declina-
process the visibility data, and any radio frequency interfer- tion dependence in the flux density scale was corrected us-
ence was excised by the AOFLAGGER algorithm (Offringa ingtheVeryLargeArrayLow-FrequencySkySurveyRedux
etal.2012). Forthefiveinstantaneousobservingbandwidths (VLSSr;Laneetal.2014),MRC,andNVSS,andtoplacethe
of 30.72MHz, which are observed approximately two min- survey on the Baars et al. (1977) flux density scale. It is es-
utes apart, an initial sky model was produced by observing timatedthatthefluxdensitycalibrationisinternallyaccurate
bright calibrator sources. WSCLEAN (Offringa et al. 2014) to within 2–3% and accurate to 8-13% when comparing the
wasusedtoperformtheimagingoftheobservations, imple- GLEAMfluxdensitiestoothersurveys(Hurley-Walkeretal.
menting a “robust” parameter of 1.0 (Briggs 1995), which 2017).
−
isclosetouniformweighting. Eachsnapshotobservationhad For each mosaic, a deep wideband image covering 170-
4 CALLINGHAMETAL.
231MHz was formed, with a resolution of 2(cid:48), to provide 17MHz, thus increasing the potential for a sample to be bi-
≈
ahighersignal-to-noiseratioandmoreaccuratesourceposi- asedbyvariablesources.
tionsthanwhatcanbeattainedfromasingle7.68-MHzsub-
band image. The sources in this wideband image were then 3. CROSS-MATCHINGROUTINE
convolved with the appropriate synthesized beam and used The Positional Update and Matching Algorithm (PUMA;
as priors for characterizing the flux density of the sources Line et al. 2016) was used to assess the probability of a
at each of the sub-band frequencies, using the source find- cross-match between the GLEAM extragalactic catalog and
ingandcharacterizationprogramAEGEAN1 v1.9.6(Hancock NVSS/SUMSS. PUMA is an open source cross-matching
etal.2012). software2, specifically designed for matching low-frequency
radio((cid:46)1GHz)catalogsthathavevaryingresolutions. Itim-
2.2. NRAOVLASkySurvey(NVSS) plements a Bayesian positional matching approach that uses
catalogsourcedensity,skycoverage,andpositionalerrorsas
NVSS is a 1.4GHz continuum survey that was conducted
a prior, to calculate the probability of a true match for any
bytheVeryLargeArray(VLA)between1993and1996(Con-
cross-match(Budava´ri&Szalay2008).
donetal.1998). Itcoverstheentireskynorthofadeclination
of 40◦and at a resolution of 45(cid:48)(cid:48). The catalog produced As the surveying telescopes used to create the all-sky
− ≈ catalogs have differing resolutions, multiple matches are
from the survey has a total of 1,810,672 sources, which is
common-place when cross-matching the different catalogs
100%completeabove4mJy. Thepositionsofthesourcesare
accuratetowithin1(cid:48)(cid:48). (seee.g.Carrolletal.2016). Confusedmatchescanmanifest
intwodifferentways: multiplesourcesfromahigherresolu-
2.3. SydneyUniversityMolongloSkySurvey(SUMSS) tioncatalogappeartomatchasinglesourcefromalowerres-
olution catalog, when really only one source truly matches;
SUMSS is a continuum survey designed to have similar
a lower resolution catalog is blending multiple components
frequency, resolution, and sensitivity to NVSS but to cover
togetherandsomultiplesourcesfromahigherresolutioncat-
the sky below the declination limit of NVSS. SUMSS was
alogdotrulymatchasinglelowerresolutionsource.
conductedbytheMolongloObservatorySynthesisTelescope
In this work, the GLEAM extragalactic catalog was used
(MOST; Mills 1981; Robertson 1991) at 843MHz between
as the base catalog, and was individually cross-matched to
1997 and 2003, covering the sky south of a declination of SUMSS and NVSS, with an angular cut-off of 2(cid:48)20(cid:48)(cid:48), which
30◦, excluding Galactic latitudes below 10◦ (Bock et al.
is approximately the full-width half-maximum of the MWA
−
1999;Mauchetal.2003). Theresolutionofthesurveyvaried
beaminthewide-bandimage. Allpossiblematcheswerere-
with declination δ as 45(cid:48)(cid:48) 45(cid:48)(cid:48)cosecδ . The catalog con-
tained,andtheresultscombinedtocreategroupsofpossible
× | |
sists of 211,063 sources, with a limiting peak brightness of
cross-matches to each GLEAM source. Within each group,
6mJybeam−1forsourceswithdeclinationsbelow 50◦,and
the positional probability of a true match was calculated for
10mJybeam−1 for sources with declinations abo−ve 50◦.
each cross-match (a combination of sources that only in-
Positions in the catalog are accurate to within 1(cid:48)(cid:48) 2−(cid:48)(cid:48) for
cludesonesourcefromeachcatalog). Usingthesepositional
sources with flux densities greater than 20mJybeam−−1, and
results, we then selected the sources that PUMA assigned
arealwaysbetterthan10(cid:48)(cid:48).Thesurveyisbelievedtobe100%
as isolated, implying only one source from each catalog
acbo3om0vp◦el.e≈te1a8bmovJyef≈or8smouJrycessoubtehtwoefeanddeecclliinnaattiioonnsooff−−5500◦◦,aanndd lsaoyurwceitshilnay2w(cid:48)2i0th(cid:48)(cid:48)i.nT1h(cid:48)1e0se(cid:48)(cid:48)coafsethsewGerLeEaAccMepstoeduricfeapllosmitaiotcnheodr
− if the positional probability of the cross-match was > 0.99.
Weexcludedcaseswheremultiplesourceswerematchedtoa
2.4. Additionalradiosurveys
GLEAM source, since we are interested in peaked-spectrum
WhileGLEAM,NVSS,andSUMSSweretheonlysurveys sources,whicharedefinedtobeunresolvedattheresolution
used for selecting peaked-spectrum sources, once a peaked- ofthesurveysusedinthisstudy. Additionaldetailsaboutthe
spectrumsourcewasidentifieditwascross-matchedtoother cross-matchingoftheGLEAMsamplearepresentedin§5.
all-sky radio surveys that covered any part of the GLEAM
survey region. This included the 74MHz VLSSr, 408MHz 4. SPECTRALMODELINGPROCEDURE
MRC,andtheAustraliaTelescope20GHz(AT20G;Murphy Selecting and assessing the spectral properties of peaked-
etal.2010)survey. Theseadditionalsurveyswherenotused spectrum sources requires fitting their spectra. The parame-
in any of the following spectral modeling, unless otherwise ter values of various models fit in this study were assessed
explicitlystated,butwillbeshowninanyspectralenergydis- usingtheBayesianmodelinferenceroutineoutlinedinCall-
tributionstohelpidentifyifaspectralfittotheGLEAMand ingham et al. (2015). In summary, a Markov chain Monte
NVSS/SUMSSdataisaccurate. Carlo (MCMC) algorithm was used to sample the posterior
Note that at the time of writing the 150MHz TGSS- probabilitydensityfunctionsofthevariousmodelparameters.
Alternativedatarelease1(TGSS-ADR1;Intemaetal.2016) TheparametervalueswereacceptedwhentheappliedGaus-
was released and undergoing review, including refining the sianlikelihoodfunctionwasmaximizedunderphysicallysen-
uniformity of its flux density scale. The identified peaked- sible uniform priors. The affine-invariant ensemble sampler
spectrum sources were also cross-matched to TGSS-ADR1 ofGoodman&Weare(2010),viathePythonpackageemcee
but TGSS-ADR1 was not used for the selection of peaked- (Foreman-Mackeyetal.2013),wasimplemented.Weutilized
spectrumsourcesdespiteasignificantimprovementinsensi- the simplex algorithm to direct the walkers to the maximum
tivity and resolution compared to the GLEAM survey. This ofthelikelihoodfunction(Nelder&Mead1965).
is largely because TGSS-ADR1 only surveyed the sky at a WhenfittingwithintheGLEAMband,weassumedthatthe
single frequency with a comparatively small bandwidth of flux density measurements were independent and the uncer-
1https://github.com/PaulHancock/Aegean 2https://github.com/JLBLine/PUMA
EXTRAGALACTICPEAKED-SPECTRUMRADIOSOURCESATLOWFREQUENCIES 5
tainties were Gaussian. However, the known correlation be- Inrarecases,thespectrumofasourceisnotwellmodeled
tween the sub-band flux densities within the GLEAM band by a turnover with power-law slopes on either side. To de-
(see §5.4 of Hurley-Walker et al. 2017) had to be mod- scribethespectraofthesecomplexsources,itisassumedthat
eled when GLEAM data were fit simultaneously with other theparticlepopulationproducingthenon-thermalpower-law
surveys, to ensure that any spurious trends present in the spectrumissurroundedbyahomogeneousionizedscreenof
GLEAM flux density measurements did not influence any plasmasuchthat
physicalrelations.Itisnotpossibletocalculatetheexactform
of the covariance matrix that would describe the correlation S =aναe(ν/νp)−2.1. (4)
ν
betweentheGLEAMpoints,butitcanbeapproximatedusing
ThishomogeneousFFAmodelisusedtomodelthespectra
Gaussian processes with a Mate´rn covariance function (Ras-
of such sources because it produces an exponential attenu-
mussen & Williams 2006). The Mate´rn covariance function
ation below the spectral peak, as opposed to the power-law
producesastrongercorrelationbetweenfluxdensitymeasure-
relations described by Equation 3. The FFA model, and the
mentscloserinfrequencyspacethanfurtheraway,asisphys-
spectra of the sources that the model is used to describe, are
ically expected for the GLEAM correlation since it largely
discussedinmoredetailin§9.
arisesfromacomplexinteractionofmulti-frequencyCLEAN,
self-calibration,andside-lobeconfusion(Franzenetal.2016; 5. PEAKED-SPECTRUMSOURCESELECTIONCRITERIA
Hurley-Walkeretal.2017).
A selection criterion that is effective in selecting a partic-
ular type of source, and which is well-defined such that any
4.1. Spectralmodels introducedbiasiseasilyquantified,isrequiredinordertopro-
In this study, the spectra of sources are fit with four dif- duceacompleteandreliablesample. Foridentifyingpeaked-
ferentspectralmodelstohelpselectandcharacterizepeaked- spectrumsources,theselectioncriterioninvolvescharacteriz-
spectrum sources. Firstly, we use the standard non-thermal ing whether the distinguishing feature of a spectral peak oc-
power-lawmodeloftheform: cursinasource’sspectrum.
Previous studies have made the assessment of a spectral
S =aνα, (1) peakinradiocolor-colorphasespace(e.g.Sadleretal.2006;
ν
Massardietal.2011),whereradiocolor-colorphasespaceis
wherea,inJy,characterizestheamplitudeofthesynchrotron defined by the spectral index derived between two high fre-
spectrum, α is the synchrotron spectral index, and Sν is the quencies αhigh and the spectral index derived between two
flux density at frequency ν, in MHz. Since the GLEAM lower frequencies α . If α had an opposite sign to
low low
survey has a large fractional bandwidth, and radio sources α , it was assumed that a peak occurs in the frequency
high
are known to show curvature in their observed spectra when rangesomewherebetweenthefrequenciesinwhichα and
low
closelysampled(Blundelletal.1999),wealsofitthecurved α were derived. We also utilize radio color-color space
high
power-lawmodelcharacterizedas foridentifyingsourcesthatpeaksomewherebetweentheend
of the GLEAM frequency coverage and the start of the fre-
S =aναeq(lnν)2, (2) quencycoverageofSUMSS/NVSS.However,duetothelarge
ν
fractionalbandwidthoftheGLEAMsurvey,sourcesthathave
whereqparametrizesthespectralcurvatureandνp =e−α/2q a turnover between 72 and 231MHz could be missed in ra-
is the frequency at which the spectrum peaks. Significant dio color-color phase space since a power-law does not ac-
curvature is represented by values of q > 0.2, and the curatelydescribetheirspectra. Therefore,wewillalsoiden-
| |
spectral curvature flattens towards a standard power-law as tify peaked-spectrum sources through a direct measurement
q approaches zero. While such a parameterization of curva- of their curvature in the GLEAM band. This is outlined be-
turemightnotseemphysicallymotivated, Duffy&Blundell lowin§5.1and§5.2.
(2012)haveshownthatq canbedirectlyrelatedtothemag-
netic field strength, energy density, and electron density of 5.1. Skyarea,resolution,andfluxdensitylimits
lobe-dominatedsources. Beforemakingadistinctionbasedonthespectralproperties
Additionally,thefollowinggenericcurvedmodelwasused of a source, we must first make resolution, cross-matching
to characterize the entire spectrum of a peaked-spectrum and flux density cuts to the GLEAM extragalactic catalog to
source: ensure that a reliable and complete peaked-spectrum sample
withwellunderstoodbiasesisderived. Theselectioncriteria
S (cid:16) (cid:17)(cid:18) ν (cid:19)αthick employedaresummarizedinTable1anddetailedbelow:
S = p 1 e−(ν/νp)αthin−αthick ,
ν (1 e−1) − ν 1. Any source that is resolved in the GLEAM wideband
p
−
(3) image, centered on 200MHz, was eliminated since
whereα andα arethespectralindicesintheoptically peaked-spectrum sources are found to have small spa-
thick thin
thickandopticallythinregimesofthespectrum,respectively. tial scales. The wideband image was used to perform
S isthefluxdensityatthefrequencyν (Snellenetal.1998). thiscutbecauseitachievesthehighestresolutionofthe
p p
When α = 2.5, this model reduces to a homogeneous GLEAMsurveyof 2(cid:48),andallsourcesintheGLEAM
thick
≈
SSAsource. Equation3doesnotassumetheunderlyingab- survey are found within the wideband image. We de-
sorptionmechanismisSSAorFFA,butdoesrequirethatthe termined whether a source was resolved in the wide-
slope of the spectrum above and below the spectral peak be bandimagebythecriterionab/(a b )(cid:54)1.1,where
psf psf
modeledbyapower-law(similartoe.g.Bicknelletal.1997). a,b,a ,andb arethesemi-majorandsemi-minor
psf psf
Notethatthismodelisonlyusedtodescribethespectrumof axesofasourceandthepointspreadfunction,respec-
asource,nottoassesswhetherSSAorFFAisresponsiblefor tively. Whilethisresolutionlimitissignal-to-noisede-
theturnover. pendent, the flux density cut outlined in step 3 below
6 CALLINGHAMETAL.
Table1
Asummaryoftheappliedselectioncriteriaandthenumberofsourcesthatremainedaftereachstageofselection.Italicizednumbersindicatethesubsetof
sourcesselectedfromthepreviousnon-italicizednumber.Thedetailsoftheselectionprocessarediscussedin§5.Withregardtothepeaked-spectrumsamples,
“highfrequency”referstosourceswithaspectralpeakaboveafrequencyof≈180MHz,while“lowfrequency”referstosourceswithaspectralpeakbelowa
frequencyof≈180MHz.bandδrepresentGalacticlatitudeanddeclination,respectively.
Selectionstep Selectioncriterion Numberofsources
0 TotalGLEAMextragalacticcatalog(|b|(cid:62)10◦,δ(cid:54)+30◦) 307,456
1 UnresolvedintheGLEAMwidebandimageab/(a b )(cid:54)1.1 210,365
psf psf
2 δ(cid:62)−80◦ 208,595
3 S (cid:62)0.16Jy 98,329
200MHz,wide
4 Sourceswith8ormoreGLEAMfluxdensity
measurementswithaSNR(cid:62)3 96,698
5 NVSSand/orSUMSScounterpart 96,628
6 Peaked-spectrumselection 1,483
6a Highfrequencysoftsample
α (cid:62)0.1andα (cid:54)−0.5 207
low high
6b Highfrequencyhardsample
α (cid:62)0.1and−0.5<α (cid:54)0 506
low high
6c GPSsample
α (cid:62)0.1andα >0 261
low high
6d Lowfrequencysoftsample
α <0.1,α (cid:54)−0.5,72MHz(cid:54)ν (cid:54)231MHz,q(cid:54)−0.2,and∆q(cid:54)0.2 394
low high p
6e Lowfrequencyhardsample
α <0.1,−0.5<α (cid:54)0,72MHz(cid:54)ν (cid:54)231MHz,q(cid:54)−0.2,and∆q(cid:54)0.2 115
low high p
ensures that this step has not removed potential low suesnearbrightsourcesortheedgeofthesurvey. Ad-
signal-to-noise peaked-spectrum sources. This resolu- ditionally,sourceswithlowSNRwillsufferfromnon-
tioncutreducesthetotalGLEAMextragalacticcatalog Gaussianityintheiruncertainties(Hurley-Walkeretal.
byapproximatelyone-thirdto208,595sources. 2017). Since accurate spectra across the entirety of
theGLEAMbandareneededtoreliablyselectpeaked-
2. Sourcesthatarelocatedwithin10◦ofthesouthcelestial spectrumsources,werequiredthatasub-bandfluxden-
polewereremovedbecauseofgreaterthan80%uncer- sityhaveaSNR(cid:62)3tobeusedinthespectralfitting. If
tainties in the GLEAM flux density scale, and greater asourcewasleftwithsevenorlessGLEAMfluxden-
than 1(cid:48) uncertainties in the GLEAM positions. Such sitymeasurements,whichrepresentslessthanathirdof
large uncertainties are mostly due to blurring of the thetotalGLEAMbandwidth,itwasexcludedfromthe
source resulting from mosaicking many hours of data sample. This step removed 1,739 sources, 1.8% of
with different ionospheric conditions (Hurley-Walker thesourcesfromthepreviousstep. ≈
etal.2017).Thisdecreasedthesamplebyanadditional
1,770sources,or0.8%. Notethatsourcesbetweende- 5. The remaining sources were cross-matched to NVSS
clinationsof 72◦ and 80◦ havelargerGLEAMsys- andSUMSSusingthecross-matchingroutineoutlined
tematic uncer−tainties th−en the rest of the survey to re- in §3. Since NVSS and SUMSS are over two orders
flectthelargeruncertaintyinthefluxdensityscalefor of magnitude more sensitive than the GLEAM survey,
thispartofthesky. 99.93% of the sample have a NVSS/SUMSS coun-
terpart. The 70 sources that do not have a counter-
3. A flux density cut was made to provide a reliable part are retained for follow-up investigations. Note
peaked-spectrum sample and to evaluate its complete- that GLEAM sources that are located between decli-
ness. The GLEAM extragalactic catalog is estimated nationsof 30◦and 40◦havetwocounterpartsdueto
to be 90% complete and 99.98% reliable at 0.16Jy a10◦decli−nationove−rlapbetweenNVSSandSUMSS.
≈
based on the wideband image. Therefore, we only in-
vestigatedsources whichhad S (cid:62) 0.16Jy, 5.2. Spectralclassification
200MHz,wide
whereS isthefluxdensityinthewideband Atthisstepintheselectionprocess,itispossibletodefine
200MHz,wide
image. Imposing this flux density limit also guaran- thepositionofthesourcesinaradiocolor-colorphasespace
tees that the sample is only formed from sources with thatischaracterizedfrom72MHzto843MHz/1.4GHz. Us-
signal-to-noise ratios (SNRs) greater than 20, limiting ing the modelling procedure detailed in §4, we fit Equation
the impact of classical confusion on the spectra of the 1 to the twenty GLEAM flux density points to derive the
sources(Murdochetal.1973).Thesamplewasapprox- low frequency spectral index αlow. Similarly, the high fre-
imatelyhalvedto98,329sourcesafterthisfluxdensity quency spectral index αhigh was derived by fitting a power-
cut. law to the SUMSS and/or NVSS flux density point(s) and
to the two GLEAM sub-band flux densities centered on 189
4. The previous flux density cut selects a reliable sample and212MHz.ThesetwoGLEAMsub-bandfrequencieswere
from the wideband image but does not account for lo- chosen because they are near the top of the overall GLEAM
calvariationsinthenoisewithinthesub-bandimages. band and are from two different GLEAM observing bands.
LargevariationsinthelocalRMSnoisewithfrequency Sincesystematicfluctuationsintroducedfromthedatareduc-
in the GLEAM survey are often due to calibration is- tionprocedurearelargelyobservingbandbased(seee.g. Fig-
EXTRAGALACTICPEAKED-SPECTRUMRADIOSOURCESATLOWFREQUENCIES 7
ure 18 of Hurley-Walker et al. 2016), the selection of two 843MHz/1.4GHz.Therefore,sourceswithα (cid:62)0.1
low
sub-band frequencies from different observing bands mini- and α > 0 are also isolated. This sample is re-
high
mizes the impact of any systematic and statistical variations ferred generally to as the GPS sample, and it contains
incalculatingα . 261sources. Asaspectralpeakinthesesourcesisnot
high
The radio color-color phase space from 72MHz to directly detected, they are isolated to largely provide
843MHz/1.4GHzfor96,628sourcesispresentedinFigure spectral coverage below the turnover for known GPS
2, and represents the most populated radio color-color plot sources.
produced to date. As expected from previous spectral index
Sincethespectraofthesourcesarebeingfitacrossalarge
studies at these frequencies (Lane et al. 2014; Mahony et al.
fractionalbandwidthtocalculateα ,anyspectralindexde-
2016; Hurley-Walker et al. 2017), 70% of sources cluster low
around (α ,α ) 0.8 0.2≈, which is located in the rivedfromapower-lawmodelfitcanbeartificiallyflattenedif
low high
≈ − ± asourcedisplaysspectralcurvaturewithintheGLEAMband.
third quadrant of the diagram, as represented by the sym-
This means that sources that display a peak between 72 and
bol Q3 in Figure 2, corresponding to spectra described by
231MHz are shifted towards the third quadrant of Figure 2,
an optically thin synchrotron power-law. The sources in the
andcanbecalculatedtohaveanegativeα ifthecurvature
firstquadranthaveapositivespectralindexthatextendsfrom low
is significant. An example of a negative α being derived
72MHz to 843MHz/1.4GHz, and are likely dominated by low
for a source with significant curvature in the GLEAM band,
GPS or HFP sources that are peaking near or above 1GHz.
such that the source is not located in the second quadrant of
Thefourth quadrantis occupiedby sourcesthatexhibit con-
Figure2,isprovidedinFigure4.
vexspectra,whicharelikelycompositesourceswithasteep-
The curvature model described by Equation 2 was also fit
spectrum power-law component at low frequencies and an
to the GLEAM data to test for any evidence of spectral cur-
inverted component at high frequency, which could indicate
vature between 72 and 231MHz. Approximately 80% of
multipleepochsofAGNactivity.
sources remaining at step 5 of the selection process show
Sourcesthatdisplayconcavespectranearthetop,orabove,
zero or negligible curvature in their spectra covered by the
theGLEAMbandarelocatedinthesecondquadrantofFig-
GLEAM band, with only 20,322 sources having a reliable
ure2,andthusitispossibletoselectpeaked-spectrumsources
value of the curvature parameter q (cid:62) 0.2, which Duffy &
from this region. Note that in the literature, GPS, CSS, and
| |
HFP have been defined as having α (cid:54) 0.5 (O’Dea Blundell (2012) class as significant curvature. The distribu-
high
− tionofqagainstα ispresentedinFigure5forsourceswith
1998).Suchadefinitionwillalsobeappliedtoisolatepeaked- low
∆q (cid:54)0.2,where∆qistheuncertaintyinq. Therequirement
spectrumsourcesinthisstudyforeaseofcomparisontoliter-
aturesamples,butthecontinuousdistributioninα across of∆q (cid:54) 0.2ensuresthemeasurementofq isreliable,andis
high
equivalenttothesignal-to-noisecutmadeinstep3.
α = 0.5 in Figure 2 suggests such a definition is ar-
high
− Due to the impact that curvature has on calculating α ,
bitrary. We have included Figure 3 to help guide the reader low
selectingpeaked-spectrumsourcesinradiocolor-colorphase
in identifying the different areas of Figure 2 used to isolate
space is unreliable for sources that display a peak in the
peaked-spectrum sources, as outlined in the next step of the
GLEAM band. Hence, a source that had α < 0.1, which
selectionprocess: low
was the limit in α used in radio color-color phase space
low
6a. Sources in the second quadrant of Figure 2 have to select peaked-spectrum sources, was also classified as a
a turnover in their spectra somewhere between peaked-spectrum source if a turnover was identified in the
200MHzand843MHz/1.4GHz. Sincetheoriginal GLEAMband.
≈definitionofGPS,CSS,HFPsourcesrequiresα (cid:54) The information used to assess a spectral peak in a radio
high
0.5 (O’Dea 1998), which is shown by the solid blue bandissetbythesignal-to-noiseofthebandwidthavailable,
−line in Figure 2, we use this limit to select peaked- withthemaximumamountofinformationtodetectaturnover
spectrumsourceswithα (cid:54) 0.5. Wealsorequire transpiring when the peak occurs in the middle of the band.
high
α (cid:62) 0.1, instead of α (cid:62)−0, because it is sig- As a peak shifts to the edge of the GLEAM band, the infor-
low low
nificantly more reliable in selecting peaked-spectrum mationtoreliablydetectitdeclinesandisdictatedbythelever
sourcesasthecutatα (cid:62)0.1minimizesthecontami- armclosesttotheedgeoftheband.Forexample,asourcethat
low
nationofflatspectrumsources,andbecausethemedian peaks at the central GLEAM frequency of 151MHz has ten
uncertaintyinα is 0.1. Sourceswithα < 0.1 spectral data points to assess whether the peak is real, while
low low
are discussed below. T≈his sample is referred to as the asourcethatpeaksat85MHzor220MHzonlyhastwodata
highfrequencysoftsampleandcontains207sources. pointstomakethesameassessment.Inparticular,thereliabil-
ityinidentifyingapeakattheedgeofthebandissignificantly
6b. Itispossibleforpeaked-spectrumsourcestoexistinthe
impactedbystatisticalfluctuations.
secondquadrantabovethelimitofα = 0.5. Such
high Therefore, if the spectral peak of a source is measured at
−
sourceshavewiderspectralwidths,orhigherfrequency
datapointN ,whichisabovethecentralfrequencydatapoint
p
peaks, than those peaked-spectrum sources identified
N ,thenadetectionofaspectralturnoverisconsideredreli-
c
in step 6a. Therefore, we also select peaked-spectrum
ableif
sources with α (cid:62) 0.1 and 0.5 < α (cid:54) 0, and
low high
−
refertothiscollectionofsourcesasthehighfrequency (cid:32) (cid:33)1/2
hardsample.Suchasamplemaybemorecontaminated N(cid:80)H σ2
by variable flat-spectrum sources than the soft sample i
duetotheshallowerdependenceonα . Therearea ∆q (cid:54)0.2 i=Np , (5a)
totalof506sourcesinthishighfrequenhicgyhhardsample. − N(cid:80)H S
i
6c. Sources located in quadrant 1 of Figure 2 could i=Np
be GPS, HFP, and CSS sources that peak above where σ and S are the local rms noise and flux density in
i i
8 CALLINGHAMETAL.
1
0.8
0.6
0.4
0.2 0.2 0.4 0.6 0.8 1
2.0
Q4 Q1
0 500 1000 1500 2000
1.0
0.0
h
g
hi
α
-1.0
-2.0
Q3 Q2
-3.0
2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0
− − − − α
low
Figure2. Radiocolor-colordiagramforthe96,628GLEAMsourcesthatremainafterstepfiveoftheselectionprocessthatisdetailedinTable1. αlow is
derivedfromthetwentyGLEAMdatapointsbetween72and231MHz.αhighwascalculatedbetweentheGLEAMfluxdensitymeasurementscenteredon189
and212MHzandNVSSand/orSUMSS.Contoursandadensitymapareplottedfortheregionsurrounding(αlow,αhigh)≈−0.8duetothelargenumberof
points. Thecolorinthedensitymapconveysthenumberofsourceseachcoloredpixel,correspondingtothevaluesofthecolorbaratthetopleftoftheplot.
Thecontourlevelsrepresent100,500,1000,and2000sources,respectively. Atαhigh=−0.5,thehorizontalsolidbluelinedelineatesthelimitbelowwhich
sourcesareidentifiedaspeaked-spectrumintheliterature. Theverticalsolidbluelineatαlow=0.1markstheseparationbetweenthehighandlowfrequency
peaked-spectrumsamples. Thedashedredlinesrepresentspectralindicesofzeroandtheone-to-onerelationofαlowandαhigh. Showningrayatthecorners
oftheplotaremockspectraofthesourcesforthatquadrant.Individualerrorbarsarenotplottedtoavoidconfusion,butthemedianerrorbarsizeforthesample
isillustratedatthetopleftofthefigure.Thefirst,second,third,fourthquadrantsdiscussedinthetextarelabeledbyQ1,Q2,Q3,andQ4,respectively.Thetwo
histogramstothetopandtotherightofthediagramrepresenttheone-dimensionaldistributionsofαlowandαhigh,normalizedbythemaximumvalueofthe
distribution.Themedianvalue,andstandarddeviation,ofαlowandαhighare−0.81±0.19and−0.82±0.21,respectively.Themedianvaluesofαlowand
αhighareshownbydashedblacklines.Over-plottedonthehistograms,inred,areGaussianfitstothedistributions.
each subband i, respectively. N represents the highest fre- peaks at 151MHz is below 1% when ∆q = 0.2. This is
H
quencydatapointintheband. IfN (cid:54)N thentheinverseof equivalenttoasignal-to-noisecutof30inthewidebandim-
p c
thesumofthepreviousequationoccurs age. Note that the functional form of Equations 5a and 5b
assumesthatthenoisebetweenthesub-bandsisindependent.
(cid:32) (cid:33)1/2 WhilethisisnotthecasefortheGLEAMdataatlowfrequen-
N(cid:80)p σ2 ciesduetoconfusion,sinceweonlytesthighsignal-to-noise
i
sources for a spectral peak, the impact of confusion is min-
∆q (cid:54)0.2 i=NL , (5b) imized on the spectrum and the noise can be approximated
− N(cid:80)p S as Gaussian and independent (Franzen et al. 2016; Hurley-
i
Walkeretal.2017).
i=NL
The distribution of the curvature parameter, q, against the
whereNListhelowestfrequencydatapointintheband. The frequencyofthepeakintheGLEAMband,νp,forthesources
second terms of Equations 5a and 5b represent the combi- remaining after step 5 of the selection process, is presented
nation of the signal-to-noise ratios in each of the sub-bands in Figure 6. The accumulation of black data points towards
above or below the spectral peak. The magnitude of these low q and ν is a function of noise, particularly in the high-
p
termsdecreaseasthenumberofpointsavailabletoassessthe est sub-band in GLEAM. A concave spectrum is considered
reliabilityofaturnoverincreases. Wechosetosubtractthese significantifq (cid:54) 0.2(Duffy&Blundell2012).
terms from 0.2 as the number of false detections of spectral −
EXTRAGALACTICPEAKED-SPECTRUMRADIOSOURCESATLOWFREQUENCIES 9
2.0 2.0
GPSsample 0 200 400 600 8001000
1.0 (6c) 1.5
1.0
0.0 Highfrequency
αhigh (6e) hardsample(6b) terq 0.5
e
-1.0 m
a
Lowfrequency Highfrequency r
a
softsample(6d) softsample(6a) p 0.0
-2.0 re
u
t
a
v 0.5
r−
u
-3.0 C
2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0
− − − − αlow
1.0
Figure3. Aschematicoftheradiocolor-colorplotofFigure2. Thecol- −
oredareasrepresenttheregionsinwhichpeaked-spectrumsourceswerese-
lectedforthedifferentsamplesoutlinedinstep6oftheselectionprocess.
1.5
Thegreen,blue,yellow,purple,andmaroonsectionsaretheareasinwhich −
thehighfrequencysoft(step6a),highfrequencyhard(step6b),GPS(step
6c),lowfrequencysoft(step6d),andlowfrequencyhard(step6e)samples
wereselected,respectively. Notethathighfrequencyhereimpliesthepeak 2.0
− 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0
ofthesourcegenerallyoccursabovefrequenciesof≈180MHz. Theareas − − − − α
highlightedforthelowfrequencysamplesareonlyindicativeand≈5%of low
sourcesoutsidetheseregionsarelocatedabovetheone-to-oneline.
Figure5. Thedistributionofthecurvatureparameterqagainstαlowforthe
96,628sourcesremainingafterstep5oftheselectionprocess.Onlysources
with∆q(cid:54)0.3areplotted.Theverticalandhorizontalreddashedlinesrep-
resentthemedianofαlowandthecut-offusedtoindicatesignificantconcave
curvatureq=−0.2,respectively.Thecolorofthedensitymapcorresponds
tothenumberofsourcesineachpixel,assetbythecolorbarinthetopright-
GLEAM J233343-305753 handcorner. Thecontourlevelsareat50,150,450,and850sources. Peak-
spectrumsourcesthatareselectedinstep6areplottedinred,withsources
5 abovethebluelineαlow = 0.1mostlyidentifiedbasedontheirpower-law
spectralproperties.Sourcesbelowthislinewereselectedbasedonaspectral
peakintheGLEAMband.Theroughdiagonaledgeofthepeaked-spectrum
4 sourcedistributionaroundαlow ≈ −0.5andq ≈ −0.5isduetothere-
quirementofmoresignificantcurvatureintheGLEAMbandtoselectlower
signal-to-noisesources.Themedianuncertaintiesforqandαlowareshown
atthebottom-leftcornerofplot.
3
)
y
J Therefore,weselectpeaked-spectrumsourcesbasedonthe
(
y detection of a peak in the GLEAM band, and also separate
t
nsi 2 intosoftandhardsamplesdependingontheirvalueofαhigh:
e
D
6d. A source is added to the low frequency soft sam-
ux ple if it is found with αlow < 0.1, αhigh (cid:54) 0.5,
Fl 72MHz(cid:54)ν (cid:54)231MHz, q (cid:54) 0.2, and ∆q i−s less
p
−
than that set by Equations 5a and 5b. Such a sample
hasasizeof394.
1
6e. Forsourcestobeaddedtothelowfrequencyhardsam-
0.9
ple,itisrequiredthatα < 0.1, 0.5 < α (cid:54) 0,
0.8 72MHz(cid:54)ν (cid:54)231MHlzo,wq (cid:54) 0−.2, and ∆hqighis less
p
7080 100 200 300 400 500600 800 1000 thanthatsetbyEquations5aand5−b.Similartothehigh
Frequency (MHz) frequencyhardsampleselectedinstep6b,thissample
ismorelikelytocontainvariableflat-spectrumsources.
Figure4. Spectral energy distribution of GLEAM J233343-305753 from
72MHzto1.4GHz. Theredcircles, bluesquare, greenleftward-pointing Thereareatotalof115sourcesinthissample.
triangle,brownrightward-pointingtriangle,andnavydownward-pointingtri-
anglerepresentdatapointsfromGLEAM,TGSS-ADR1,MRC,SUMSS,and Thehardandsoftsamplesselectedinthiswayaretogether
NVSS,respectively. Theplotinsetinthetop-rightcorneristhecolor-color referred to as the low frequency peaked-spectrum sample to
diagramofFigure2,withthepositionofGLEAMJ233343-305753marked differentiatefromthesamplesselectedsolelyonradiocolor-
byabluecircle.Evidently,αlowwascalculatedtobenegativeforthissource colorphasespace,whicharereferredtoasthehighfrequency
duetothecurvatureintheGLEAMdata.Thefitofthegenericcurvedspectral
modelofEquation3toonlytheGLEAM,SUMSS,andNVSSfluxdensity peaked-spectrumsample. Thelocationsofmorethan95%of
pointsisshownbytheblackcurve.Theorangelinesrepresentthepower-law the low frequency peaked-spectrum samples in radio color-
fitsfromwhichαhighandαlowhavebeenderived. colorphasespacearealsodisplayedinFigure3.
Theselectionprocessalsoidentifiedtwoknownpulsarsas
peaked-spectrum sources, PSR J0630-2834 and PSR J1645-
10 CALLINGHAMETAL.
0.0 menttoproduceareliablecollectionofsourcesthatpeakbe-
low 72MHz, rather than a complete sample. The properties
of these sources are also provided online in the same format
0.2
− asthetablepresentedinAppendixA.
0.4
− 5.4. Fluxdensityvariabilityandblazarcontamination
q
Thereliabilityofproceedingstudiesinidentifyingpeaked-
er 0.6
t − spectrumsourcesusingmulti-epochsurveydataweresignif-
e
m icantly impacted by radio source variability (e.g. Dallacasa
a
ar −0.8 et al. 2000; Snellen et al. 2002). Previously, the often sin-
p
gle low frequency data point used to justify a spectral peak
e
ur 1.0 was provided by an observation of the source when it was
t −
a inalessactivephasecomparedtowhenhighfrequencydata
v
ur weretaken. Hence,variablesourcesinpreviousmulti-epoch
C 1.2
− studies could masquerade as peaked-spectrum sources even
though the intrinsic spectrum of the source was flat (Tinti
−1.4 et al. 2005). Since GLEAM surveyed the sky in four ob-
servingbands,withatwo-minutecadencebetweenobserving
1.6 bands,ourestimateofthespectralpeakandtheslopebelowa
−
turnoverisnotimpactedbyvariability. Whilehigh-frequency
80 100 120 140 160 180 200 220 studieshaveobservedsourcessuchasblazarsvaryingonhour
Peak frequency ν
p to month time-scales, the radio sky below 1GHz has been
Figure6. Thedistributionofthecurvatureparameterqagainstthefrequency shown to be significantly less variable (McGilchrist & Riley
ofthepeakintheGLEAMbandνp forthe96,628sourcesremainingaf- 1990; Lazio et al. 2010; Bell et al. 2014; Rowlinson et al.
ter step 5 of the selection process. All sources with ∆q (cid:54) 0.3 are plot- 2016).
tedinblacktoprovideanindicationtheimpactthenoisehasonselecting
Itispossiblevariabilitywillimpactourdeterminationofthe
peaked-spectrum sources. Peaked-spectrum sources selected in step 6 are
over-plottedinred,withallthesourceswithνpbelow≈180MHzselected slopeabovetheturnoversincedatafromSUMSSandNVSS
on the basis of a peak in the GLEAM band. The concentration of black were utilized. However, one strength of having such a well
pointstowardlowνpisduetonoisewithintheGLEAMband.Sourcesabove sampled low-frequency survey is that the defining feature of
≈180MHzarelargelyidentifiedaspeaked-spectrumfromtheirpositionin
a peaked-spectrum source, whether the source has peak or a
radiocolor-colorphasespace.Thehorizontalreddashedlinecorrespondsto
limitofqbelowwhichcurvatureintheGLEAMbandwasconsideredsig- positivespectralslope,ischaracterizedcompletelybythelow
nificant.Themedianuncertaintiesforqandνpareplottedatthebottom-left frequency data. Hence, any high frequency variability will
cornerofdiagram. onlymovethesourcebetweenthehardandsoftsamples,en-
suringthesourcewillstillbeidentifiedasapeaked-spectrum
0317 (Manchester et al. 2005). These two sources were re- source.
movedfromthepeaked-spectrumsamples,andweexpectthe As mentioned above, blazars are known contaminants
contamination of pulsars in the total peaked-spectrum sam- in peaked-spectrum selections because of their variability.
ple to be less than 1% based on the total number of pulsars Whileouridentificationofapeaked-spectrumsourcemaynot
detectedbytheGLEAMsurvey(Belletal.2016). besignificantlyimpactedbyvariability,thesamplewillcon-
Hence,atotalof1,483extragalacticpeaked-spectrumcan- tain some blazars. This is because a peak can occur in the
didates were selected from 96,628 sources. Note that all of spectrum of a blazar if it is observed during an AGN flare,
thepeaked-spectrumcandidateswereclassedasisolatedby suchthataSSAcomponentdominatestheemissionspectrum
thecross-matchingroutine,asdetailedin§3.Thespectraofa (Torniainen et al. 2007), or due to a variation in the beam-
sourcefromeachofthesamplesdiscussedinstep6ofthese- ing angle. The two samples most likely to be contaminated
lectionprocessarepresentedinFigure7. Intheappendix,the with blazars are the hard and GPS samples, since the defini-
spectra for all peaked-spectrum sources selected are plotted. tion of these two samples include sources with αhigh 0.
∼
The tables providing the characteristics for the high and low Bycomparingthereportednatureofasourceintheliterature
frequencypeaked-spectrumsamples,andtheGPSsample,in (e.g.Massaroetal.2015),weestimatethatblazarsrepresent
the style of the table presented in Appendix A, are available 10%ofsourcesinthehardandGPSsamples,and<3%of
≈
online. sources in the soft samples. Future high resolution imaging,
inparticularwithLOFAR,andlowfrequencymulti-epochob-
5.3. Sourcespeakingbelow72MHz servationsfromtheMWATransientSurvey(MWATS;Bellet
al., inprep.) survey, andthesecondyearofGLEAMsurvey
It is also possible to identify sources that peak below
data,willisolatethesourcesinthesamplespresentedthatare
72MHz on the basis of significant curvature in high signal-
jet-dominated.
to-noisespectra. Anexampleofasourcethatisbeginningto
turn over, but peaks below the lowest GLEAM frequency, is
5.5. Obtainingredshifts
shown in Figure 8. However, since we do not detect a peak
inthespectrum, thesesourcesarenotusedinanyofthefol- We obtained redshift information for the peaked-spectrum
lowing analysis but are presented to encourage observations candidates selected in §5.2 by cross-matching our sample
below72MHztoconfirmthespectralturnover. Thereare36 to previous targeted optical observations of GPS, CSS, and
sourcesidentifiedwithq < 0.2,aturnoverbelow72MHz, HFP sources (O’Dea et al. 1991; Fanti et al. 1990; Labiano
−
and a signal-to-noise greater than 100 in the GLEAM wide- et al. 2007; de Vries et al. 2007; Holt et al. 2008), radio-
band image. We apply such a high signal-to-noise require- opticalstudiesthatcombinedthelargespectroscopicsurveys