A search for radio afterglows from gamma-ray bursts with the Australian Square Kilometre Array Pathfinder
James K. Leung, Tara Murphy, Giancarlo Ghirlanda, David L. Kaplan, Emil Lenc, Dougal Dobie, Julie Banfield, Catherine Hale, Aidan Hotan, David McConnell, Vanessa A. Moss, Joshua Pritchard, Wasim Raja, Adam J. Stewart, Matthew Whiting
MMNRAS , 1–17 (2021) Preprint 4 February 2021 Compiled using MNRAS L A TEX style file v3.0
A search for radio afterglows from gamma-ray bursts with theAustralian Square Kilometre Array Pathfinder
James K. Leung, , , ★ Tara Murphy, , Giancarlo Ghirlanda, David L. Kaplan, Emil Lenc, Dougal Dobie, , , , Julie Banfield, Catherine Hale, Aidan Hotan, David McConnell, Vanessa A. Moss, , Joshua Pritchard, , , Wasim Raja, Adam J. Stewart, and Matthew Whiting Sydney Institute for Astronomy, School of Physics, The University of Sydney, NSW 2006, Australia CSIRO Astronomy and Space Science, PO Box 76, Epping, NSW 1710, Australia ARC Centre of Excellence for Gravitational Wave Discovery (OzGrav), Hawthorn, Victoria, Australia INAF – Osservatorio Astronomico di Brera, Via E. Bianchi 46, I–23807 Merate, Italy Department of Physics, University of Wisconsin-Milwaukee, P.O. Box 413, Milwaukee, WI 53201, USA Centre for Astrophysics and Supercomputing, Swinburne University of Technology, Hawthorn, Victoria, Australia CSIRO Astronomy and Space Science, PO Box 1130, Bentley WA 6102, Australia
Accepted 2021 February 02. Received 2021 January 29; in original form 2020 December 01
ABSTRACT
We present a search for radio afterglows from long gamma-ray bursts using the AustralianSquare Kilometre Array Pathfinder (ASKAP). Our search used the Rapid ASKAP ContinuumSurvey, covering the entire celestial sphere south of declination + ◦ , and three epochs ofthe Variables and Slow Transients Pilot Survey (Phase 1), covering ∼ ,
000 square degreesper epoch. The observations we used from these surveys spanned a nine-month period from2019 April 21 to 2020 January 11. We crossmatched radio sources found in these surveyswith 779 well-localised (to ≤ (cid:48)(cid:48) ) long gamma-ray bursts occurring after 2004 and de-termined whether the associations were more likely afterglow- or host-related through theanalysis of optical images. In our search, we detected one radio afterglow candidate associatedwith GRB 171205A, a local low-luminosity gamma-ray burst with a supernova counterpartSN 2017iuk, in an ASKAP observation 511 days post-burst. We confirmed this detection withfurther observations of the radio afterglow using the Australia Telescope Compact Array at859 days and 884 days post-burst. Combining this data with archival data from early-timeradio observations, we showed the evolution of the radio spectral energy distribution alonecould reveal clear signatures of a wind-like circumburst medium for the burst. Finally, wederived semi-analytical estimates for the microphysical shock parameters of the burst: electronpower-law index 𝑝 = .
84, normalised wind-density parameter 𝐴 ∗ =
3, fractional energy inelectrons 𝜖 𝑒 = .
3, and fractional energy in magnetic fields 𝜖 𝐵 = . Key words: gamma-ray burst: general – gamma-ray burst: individual (GRB 171205A) –supernova: individual (SN 2017iuk) – radio continuum: general – radio continuum: transients
Long gamma-ray bursts (lGRBs) are produced by ultra-relativistic,narrowly collimated jets, originating from the core collapse ofmassive stars (Woosley 1993; MacFadyen & Woosley 1999).Previous observations of core-collapse supernovae accompanyinglGRBs have provided substantial evidence supporting this progen-itor model; for example, GRB 980425/SN 1998bw (Galama et al.1998) and GRB 030329/SN 2003dh (Stanek et al. 2003). In the stan- ★ E-mail: [email protected] dard fireball model, Fermi acceleration of electrons in the shockfront resulting from jet interaction with the circumburst mediumproduces an afterglow visible across the entire electromagneticspectrum (Sari et al. 1998). In many cases, radio afterglow emissionis detectable at late-time on the order of months to years post-burst,even when the jet has decelerated into the non-relativistic regimeand its expansion has become quasi-spherical (Frail et al. 2000;Chandra & Frail 2012, hereafter CF12).The study of radio afterglows helps to constrain the micro-physics and energetics (calorimetry) of burst events (Frail et al.2000; van der Horst et al. 2008; Granot & van der Horst 2014) and © a r X i v : . [ a s t r o - ph . H E ] F e b J. K. Leung et al. enables the reverse shock to be studied with greater observationallatency compared with shorter wavelengths; for example, unlike theoptical flash that occurs on the timescale of tens of seconds post-burst (Akerlof et al. 1999; Sari & Piran 1999), the radio emissionfrom the reverse shock persists for a few hours to a few days post-burst (Kulkarni et al. 1999; Anderson et al. 2018). Radio afterglowsobserved with diffractive scintillation (Goodman 1997; Waxmanet al. 1998) or Very Long Baseline Interferometry (VLBI; Granotet al. 1999; Taylor et al. 2004) techniques also provide measure-ments of the source size and jet expansion of burst events. However,due to the limited availability of observing time, follow-up obser-vations for radio afterglows have been biased towards known burststhat meet a specific set of selection criteria: for example, burstswith strong X–ray, optical (CF12; Ghirlanda et al. 2013) or grav-itational wave counterparts (Nakar & Piran 2011; Hallinan et al.2017); suspected dark bursts with optical counterparts obscured bydust that could still be detectable at radio wavelengths (Djorgovskiet al. 2001); ultra-long GRBs with prompt emission lasting ≥ , in press ) radio telescope operates at observing frequen-cies ranging from 700 MHz to 1 ,
800 MHz. It has a ∼
30 squaredegrees field-of-view and is capable of reaching ∼ − RMS in one minute of integration time. These capabilities facilitateobservations for multi-epoch unbiased radio surveys, including theRapid ASKAP Continuum Survey (RACS; McConnell et al. 2020)and the first phase of the pilot program for an ASKAP survey forVariables and Slow Transients (VAST; Murphy et al. 2013). Weconducted a search for lGRB radio afterglows using RACS andthree epochs of the VAST Pilot Survey (Phase 1; VAST-P1), withobservations spanning a nine-month period from 2019 April 21 to2020 January 11.The structure of our paper is as follows. We outline the obser-vations and data quality for RACS and each epoch of VAST-P1 usedin our search in §2. In §3, we outline the expected detection ratesof radio afterglows in our search. In §4, we explore the impact ofhost galaxy contamination on the detectability of radio afterglowsin our search (and more generally, in a search for explosive tran-sients). In §5, we describe our search methodology and how wedistinguished host galaxy emission from afterglow emission. In §6,we describe our detection and follow-up of GRB 171205A, an in-teresting source we identified in §5. In §7, we discuss the physical ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ − ◦ − ◦ ◦ +30 ◦ +60 ◦ − ◦ − ◦ ◦ +30 ◦ +60 ◦ ◦ Figure 1.
The VAST-P1 footprint is outlined in red and consists of 113 tiles.The RACS footprint shaded in blue extends between − ◦ ≤ 𝛿 < + ◦ . TheEMU Pilot Survey shaded in orange consists of 10 tiles overlapping with theVAST-P1 and RACS surveys. The sky map is plotted with J2000 equatorialcoordinates in the Mollweide projection and the background diffuse Galacticemission at 887.5 MHz (gray) is modelled from Zheng et al. (2017). interpretation of our GRB 171205A radio observations, the con-sideration of alternative search strategies, and the expected GRByields for comparison surveys (both current and future). Finally, wesummarise the conclusions from our radio afterglow search in §8.We adopt a flat Λ -CDM cosmology with 𝐻 = . − Mpc − , Ω M = .
308 and Ω Λ = . The observations and quality metrics associated with the twoASKAP surveys used in our search are described in this section.Observations in both surveys were conducted at a central frequencyof 887.5 MHz with a bandwidth of 288 MHz and a typical angularresolution of ∼ (cid:48)(cid:48) . RACS is a large-area survey consisting of 903 tiles covering theentire radio sky south of declination + ◦ . Each tile had an integra-tion time of ∼
15 min and the resulting images had a median RMSnoise of ∼ .
29 mJy beam − . The RACS observations used for oursearch were conducted between 2019 April 21 and 2019 November22. The full set of RACS observations include some tiles that werereobserved, with the final observations occurring on 2020 June 21.These subsequent observations were not used in our search, butwere used in a candidate radio afterglow follow-up discussed in §6.In this paper, RACSe1 refers to the initial observation of a tile andRACSe2 refers to any subsequent observation of the same tile.We used a pre-release version of the RACS data for our radio af-terglow search. As a result, this version of the RACS data did not im-plement some of the improvements to the data quality featured in thepublished data products. These improvements included, for exam-ple, holography corrections for the direction dependent flux-densityscale and techniques to account for variations of the point-spread-function (PSF) size and shape across the ASKAP field-of-view. Amore detailed discussion of these issues and their correspondingimprovements are found in McConnell et al. (2020) and Hale et al.( in prep. ). Please refer to these two papers for comprehensive detailson the survey design, calibration process, data reduction strategy aswell as the data products associated with the published images andsource catalogues. MNRAS000
29 mJy beam − . The RACS observations used for oursearch were conducted between 2019 April 21 and 2019 November22. The full set of RACS observations include some tiles that werereobserved, with the final observations occurring on 2020 June 21.These subsequent observations were not used in our search, butwere used in a candidate radio afterglow follow-up discussed in §6.In this paper, RACSe1 refers to the initial observation of a tile andRACSe2 refers to any subsequent observation of the same tile.We used a pre-release version of the RACS data for our radio af-terglow search. As a result, this version of the RACS data did not im-plement some of the improvements to the data quality featured in thepublished data products. These improvements included, for exam-ple, holography corrections for the direction dependent flux-densityscale and techniques to account for variations of the point-spread-function (PSF) size and shape across the ASKAP field-of-view. Amore detailed discussion of these issues and their correspondingimprovements are found in McConnell et al. (2020) and Hale et al.( in prep. ). Please refer to these two papers for comprehensive detailson the survey design, calibration process, data reduction strategy aswell as the data products associated with the published images andsource catalogues. MNRAS000 , 1–17 (2021) search for GRB afterglows with ASKAP Table 1.
Epochs from the ASKAP surveys, RACS and VAST-P1, were used in our GRB radio afterglow search. Columns 1 through 8 show the epoch number,radio survey used for each epoch, start date of each epoch, solid angle and number of tiles covered by each epoch, median RMS noise of observations in eachepoch, median right ascension offset from ICRF2 source positions for observations in each epoch, median declination offset from ICRF2 source positionsfor observations in each epoch, and the flux-density ratio of sources in each epoch compared against sources in SUMSS. Observations for all epochs wereconducted at the central frequency of 887.5 MHz with a bandwidth of 288 MHz and an angular resolution of ∼ (cid:48)(cid:48) .Epoch − ) RA Offset(arcsec) Dec Offset(arcsec) 𝑆 ASKAP / 𝑆 SUMSS − ◦ ≤ 𝛿 < + ◦ , 903 tiles 0.29 − . ± . + . ± . . ± .
291 VAST-P1 2019 Aug 27 ∼ − . ± . + . ± . . ± .
252 VAST-P1 2019 Oct 28 ∼ + . ± . + . ± . . ± .
308 VAST-P1 2020 Jan 11 ∼ − . ± . + . ± . . ± . Note – Epoch 8 is the third full epoch of the VAST-P1 survey. Several partial epochs between Epoch 2 and 8 were observed, but not used in oursearch. See O’Brien et al. ( in prep. ) for more details.
We performed an independent quality control check on theastrometric accuracy and flux-density scale to the aforementionedpapers since the version of the data used for our search differs fromthe published data products. For our quality control analysis specif-ically, we used the Selavy (Whiting & Humphreys 2012) sourcefinder with default settings to extract sources from the RACS imagesand filtered for sources that were (a) isolated ( ≥ (cid:48)(cid:48) from nearestneighbour), (b) with non-extended PSF (major axis length ≤ (cid:48)(cid:48) with a major-to-minor axis ratio ≤ ≥
10. The reason we chose a non-extended PSF criterion overa conventional point-source criterion (e.g. peak-to-integrated flux-density ratio) was due to aforementioned PSF variations affectingmeasurements of the integrated flux-density. This set of criteriaensured only isolated sources with reasonably compact PSFs andhigh signal-to-noise were used for the quality control analysis inthis section; we did not apply these criteria to filter sources for ourradio afterglow searches in subsequent sections of this paper.Using this subset of sources, we determined the astrometricoffset of RACS sources from the position of sources from the secondrealisation of the International Celestial Reference Frame (ICRF2;Ma et al. 2009); the offset is − . (cid:48)(cid:48) ± . (cid:48)(cid:48) + . (cid:48)(cid:48) ± . (cid:48)(cid:48) 𝑆 RACS / 𝑆 SUMSS = . ± .
29, assuming a spectral index 𝛼 of − . 𝑆 𝜈 ∝ 𝜈 𝛼 . We have thus independently verified the reliability ofthe astrometry and flux-density scale for our early data productsused in our scientific analysis. Table 1 summarises the datasets from the three VAST-P1 epochsused in our search. VAST-P1 covers 113 tiles using the same tilingfootprint as RACS; see Figure 1 for the sky coverage of both sur-veys. The integration time per tile per epoch was ∼
12 min and theresulting images for each epoch had a median RMS noise varyingfrom 0 .
26 to 0 .
29 mJy beam − . Some epochs have a few ( ≤
5) tilesmissing due to various issues (e.g. solar interference, observationfailures, etc.) discussed in O’Brien et al. ( in prep. ). All observationswere processed using standard procedures in the ASKAPsoft pack-age (Cornwell et al. 2011; Guzman et al. 2019) and sources wereextracted with the Selavy (Whiting & Humphreys 2012) sourcefinder. The full details of the observing strategy as well as the cal-ibration, reduction and source extraction procedures are describedin O’Brien et al. ( in prep. ). Similar to before, we used a pre-release version of the VAST-P1data, and hence, performed an independent quality control check.Applying the same procedure from the RACS quality control analy-sis, we found the astrometry of each VAST-P1 epoch to be typicallyaccurate to < (cid:48)(cid:48) in both right ascension and declination, while theflux-density scale was typically calibrated to within 6 per cent fromthat of SUMSS with a RMS scatter ≤
30 per cent. The astrometricaccuracy and flux-density scale for each epoch are listed in Table 1.We have thus demonstrated that the data quality for VAST-P1 earlydata products was comparable to the RACS early data products;both were sufficient for the analysis described in subsequent sec-tions, involving analyses of positional offsets and source variability.
We constrained the expected afterglow detection rate for on-axis af-terglows in our search using archival data of previous observations.In this work, we only consider the lGRB afterglow detections asshort gamma-ray burst (sGRB) afterglow detections require flux-density sensitivities well below the thresholds of the surveys we usein this work (e.g. Fong et al. 2015). To determine the expected lGRBafterglow detection rate, we used the CF12 catalogue, consisting of304 on-axis lGRB afterglows observed by radio telescopes over a 14year span from 1997 to 2011. Radio emission was detected from 95of these lGRBs, and of these, only 64 had known redshifts and fit-ted maximum flux-density measurements. In this paper, ‘maximumflux density’ refers to the flux density of a radio afterglow at thelight curve peak for the observing frequency of interest. Some otherpapers in the literature (e.g. CF12) refer to this as the ‘peak fluxdensity,’ but we choose to use different terminology in our paper toavoid any confusion with standard radio astronomy terminology.Figure 2 shows the scaled maximum flux densities against red-shift and luminosity distance for radio-detected afterglows from thesample as well as the 5 𝜎 flux-density threshold for various surveysconsidered in this work. Note that while the threshold of the Evolu-tionary Map of the Universe (EMU; Norris et al. 2011) Pilot Surveyis lower than the RACS/VAST-P1 threshold, it was unsuitable forour search due to the limited sky coverage (see Figure 1). We scaledthe fitted maximum flux density of the lowest frequency observationfor each GRB in CF12 to 887.5 MHz, which is the central frequencyof RACS and VAST-P1 observations. Two different spectral indiceswere applied for the scaling, 𝛼 = / 𝛼 =
2, correspondingto different scenarios for the self-absorption frequency, 𝜈 sa . In boththese scenarios, we considered the typical afterglow spectrum inthe slow cooling regime as described in Granot & Sari (2002, here-after GS02). In the 𝛼 = / MNRAS , 1–17 (2021)
J. K. Leung et al. z M a x i m u m F l u x D e n s i t y ( m J y ) EMU-PilotRACS / VAST-P1SKA-Shallow G R B G R B G R B A P R E S E N T D AY ( z = , D L = ) with k-correctionwithout k-correction lGRB (SN), α = lGRB (no SN), α = lGRB (all), α = 2 Luminosity Distance (Gpc)
Figure 2.
The fitted maximum flux density scaled to 887.5 MHz as a function of 1 + 𝑧 (bottom axis) and luminosity distance 𝐷 𝐿 (top axis) for 64 radioafterglows featured in CF12. The maximum flux density refers to the flux density of a radio afterglow at the light curve peak. Each burst is scaled with twodifferent spectral indices, 𝛼 = (self-absorption is not significant) and 𝛼 = 𝛼 = scenario, the 14 afterglows with supernova association are plotted with red star markers, while those without are plotted with blue squaremarkers. For the 𝛼 = 𝐿 𝜈 = × erg s − Hz − , are shown for two separate cases: (i) the relation governed by the typical inverse squaredecrease of the maximum flux density with luminosity distance is shown with the orange dashed line, (ii) the relation accounting for negative 𝑘 -correctioneffects leading to the plateauing effect at cosmological distances is shown with the purple dash-dotted line. For a more detailed discussion of the scaling andthe dependence of maximum flux density on redshift, see §3. The gray horizontal lines show the 5 𝜎 -limit of different radio surveys considered in this work.The region shaded in red above 1 .
50 mJy and below 𝑧 = . 𝐷 𝐿 = assumed to be above 𝜈 sa and below the synchrotron frequency corre-sponding to the peak, 𝜈 m , from which we scaled; this corresponds topower-law segment D in GS02. In the 𝛼 = 𝜈 sa , corresponding to power-lawsegment B in GS02. In this scenario, the maximum flux density atthe frequency we scaled from could either be due to the passageof 𝜈 m (in an ISM environment) or 𝜈 sa (in a wind environment); inboth cases, the appropriate spectral scaling is still 𝛼 =
2. Predictingthe temporal evolution of the flux density at 𝜈 m as it moves towardsour observing frequency requires further modelling for each burst;we consider this beyond the scope of our work and argue that thesmall difference between the frequency we scaled from (typically1.43 GHz or 8.46 GHz) and our observing frequency would not re-sult in substantial changes to the flux density at 𝜈 m as it movestowards our observing frequency. We therefore only used the twospectral indices to provide a range on the expected maximum fluxdensity for each burst when scaled to 887.5 MHz and noted that thisrange should be shifted upwards slightly if the temporal evolutionof the synchrotron peak was also factored in.Considering the 𝛼 = / 𝜎 -level during the 14 year span covered by CF12. From archival data, the duration, or length of time the radio af-terglow would be detectable, for GRB 980425 and GRB 030329,would be approximately 250 days (Li & Chevalier 1999) and 150days (Frail et al. 2005), respectively. The implied fractional time aradio afterglow from a known GRB is detectable in the sky at theRACS/VAST-P1 flux-density sensitivity is 7.8 per cent. We calcu-lated the probability of detecting a radio afterglow in our search tobe 7.2 per cent after factoring in the sky coverage of RACS andVAST-P1 as well as the cadence of our observations. These calcula-tions should be considered as lower limits since less than 20 per centof known GRBs have been observed at radio frequencies and early-time radio detections from the reverse shock mechanism were notconsidered in CF12. Considering the 𝛼 = MNRAS000
2. Predictingthe temporal evolution of the flux density at 𝜈 m as it moves towardsour observing frequency requires further modelling for each burst;we consider this beyond the scope of our work and argue that thesmall difference between the frequency we scaled from (typically1.43 GHz or 8.46 GHz) and our observing frequency would not re-sult in substantial changes to the flux density at 𝜈 m as it movestowards our observing frequency. We therefore only used the twospectral indices to provide a range on the expected maximum fluxdensity for each burst when scaled to 887.5 MHz and noted that thisrange should be shifted upwards slightly if the temporal evolutionof the synchrotron peak was also factored in.Considering the 𝛼 = / 𝜎 -level during the 14 year span covered by CF12. From archival data, the duration, or length of time the radio af-terglow would be detectable, for GRB 980425 and GRB 030329,would be approximately 250 days (Li & Chevalier 1999) and 150days (Frail et al. 2005), respectively. The implied fractional time aradio afterglow from a known GRB is detectable in the sky at theRACS/VAST-P1 flux-density sensitivity is 7.8 per cent. We calcu-lated the probability of detecting a radio afterglow in our search tobe 7.2 per cent after factoring in the sky coverage of RACS andVAST-P1 as well as the cadence of our observations. These calcula-tions should be considered as lower limits since less than 20 per centof known GRBs have been observed at radio frequencies and early-time radio detections from the reverse shock mechanism were notconsidered in CF12. Considering the 𝛼 = MNRAS000 , 1–17 (2021) search for GRB afterglows with ASKAP possible. These two GRBs, associated with SN 1998bw (Galamaet al. 1998; Kulkarni et al. 1998) and SN 2003dh (Hjorth et al.2003; Stanek et al. 2003) respectively, suggest this subset of GRBsat redshifts 𝑧 < . 𝐷 𝐿 < ∼ ,
000 km s − (Modjaz et al. 2016) and withkinetic energies reaching ∼ erg when accompanied by lGRBs(Mazzali et al. 2014). At more cosmological distances probed bydeeper searches, the identification of accompanying supernovae be-comes more difficult, but the dependence of the afterglow max-imum flux density on redshift becomes weaker due to the nega-tive 𝑘 -correction effect (Ciardi & Loeb 2000; Frail et al. 2006).This is represented in Figure 2: the conventional inverse squareluminosity distance decrease (orange dashed line) follows the rela-tion 𝐿 𝜈 = 𝜋𝑆 𝜈 𝐷 𝐿 /( + 𝑧 ) , where 𝐿 𝜈 is the spectral luminosityat the observing frequency, 𝑆 𝜈 is the flux density, 𝐷 𝐿 is the lu-minosity distance and 𝑧 is the redshift; the negative 𝑘 -correctionfactor ( + 𝑧 ) 𝛽 − 𝛼 multiplied to the aforementioned relation re-sults in a flatter flux-density dependence on redshift (purple dash-dotted line); these lines assume a typical spectral luminosity of 𝐿 𝜈 = × erg s − Hz − (CF12) and a post jet break light curvethat is optically thin and flat (i.e. spectral index 𝛼 = /
3, temporalindex 𝛽 =
0, where 𝑆 ∝ 𝜈 𝛼 𝑡 𝛽 ; Frail et al. 2006) . Therefore, theredshift limitations impacting our search would be less severe at theflux-density sensitivities achieved by the deeper EMU survey, and toa larger extent, the Square Kilometre Array Shallow (SKA-Shallow;e.g. Fender et al. 2015) survey in the future. The detectability of afterglows in our search depended on the flux-density sensitivity of our surveys and the degree of host galaxycontamination present. Only radio afterglows that reached our 5 𝜎 -sensitivity threshold of ∼ .
50 mJy beam − during our observationscould be detectable in our search. We expected our detections tobe close to this 5 𝜎 -threshold since the majority of previous radioafterglow detections peaked below this threshold level (see §3).A radio afterglow emitting above the threshold still needed tobe distinguishable from its host galaxy. The radio emission of localgalaxies at redshifts 𝑧 (cid:28) 𝑆 𝜈, gal = . (cid:18) SFR 𝑀 (cid:12) yr − (cid:19) 𝜈 − . 𝐷 − 𝐿, , (1)where SFR is the star formation rate in units of 𝑀 (cid:12) yr − , 𝜈 GHz isthe observing frequency in GHz, 𝐷 𝐿 = 𝐷 𝐿, cm is the lu-minosity distance to the host galaxy and the applied spectral indexof − . The assumption of 𝛼 = / 𝛽 = 𝛼 = 𝑘 -correction factor (and associated lines). afterglow host galaxy (i)(iii) (iv)(ii) z < . ( D L < Mpc) offset extended host z < . ( D L < Mpc)0 . < z < . Mpc < D L < Mpc) coincident host galaxy 0 . < z < . Mpc < D L < Gpc) no host contaminationoffset compact host Figure 3.
Four different scenarios showing how the detectability of radioafterglows may be contaminated by the radio emission of the host galaxies.The typical redshift (and luminosity distance) range for each scenario isgiven assuming: (a) the typical physical offset of an lGRB afterglow fromits host to be ∼ . 𝑀 (cid:12) yr − , and (c) the observations are made using a radio telescopewith ∼ (cid:48)(cid:48) angular resolution. A detailed discussion of these assumptionscan be found in §4. (i) The afterglow is offset and resolvable from its host galaxy.
Host galaxies in this scenario are compact and have angular sepa-rations from their afterglows that are at least as large as the angularresolution of our observations ( ≥ (cid:48)(cid:48) for RACS/VAST-P1). After-glows in this scenario would be detectable if their peak flux densitiesare above our 5 𝜎 -sensitivity threshold of ∼ .
50 mJy beam − .(ii) The afterglow is offset but not clearly resolved from its hostgalaxy.
Host galaxies that are very local (e.g. 𝑧 < .
03 for ra-dio galaxy samples in van Velzen et al. 2012) may have extendedemission at gigahertz frequencies. An afterglow offset from its hostgalaxy centre, even by more than 15 (cid:48)(cid:48) , may be contaminated bythis emission and our 5 𝜎 -sensitivity threshold criteria would notbe enough to determine the detectability of the afterglow. With aRMS scatter in the flux-density scale of ∼ . 𝜎 -detection requires the afterglow to reach fluxdensities ≥ . 𝜎 -sensitivity threshold criteria.(iii) The afterglow is spatially coincident with its host galaxy.
Host galaxies in this scenario are compact and have angular sepa-rations from their afterglows that are smaller than half the angularresolution of our observations ( ≤ (cid:48)(cid:48) for RACS/VAST-P1). Similarto scenario (ii), any afterglow that is spatially coincident with its hostgalaxy would only be detectable in our search with 5 𝜎 -confidenceif it reaches flux densities ≥ . MNRAS , 1–17 (2021)
J. K. Leung et al. (iv)
The afterglow is found without any detectable host galaxyemission associated with it.
Host galaxies of afterglows in this sce-nario are too faint to be detected in any epochs of our observations.These afterglows will not be contaminated by host galaxy emissionand would be detectable if their peak flux densities are above our5 𝜎 -sensitivity threshold of ∼ .
50 mJy beam − .Physical offsets for lGRBs from their host galaxies range from0.075 to 14 kpc with a median offset of ∼ . (cid:48)(cid:48) angular offset from its host galaxy if it is at a redshift of 0.004(0.046), or equivalently, a luminosity distance of 20 Mpc (200 Mpc).This implies that finding afterglows in scenarios (i) and (ii) withappreciable offsets from their hosts using ASKAP surveys is un-likely since all GRBs found in our sample in §3 were located at 𝑧 > .
004 ( 𝐷 𝐿 >
20 Mpc). For instance, our lowest redshift sam-ple, GRB 980425, at 𝑧 = .
008 ( 𝐷 𝐿 =
40 Mpc) had a measuredangular offset of 12 . (cid:48)(cid:48) ± . (cid:48)(cid:48)
052 (Bloom et al. 2002), which is be-low the angular resolution of our observations. This suggests radioafterglows found appreciably offset from their hosts would likely beassociated with an extremely local burst 𝑧 < .
004 ( 𝐷 𝐿 <
20 Mpc).The star formation rate of lGRB host galaxies range from 0.2to 50 𝑀 (cid:12) yr − (Berger 2009) with a median of about 2 𝑀 (cid:12) yr − (Christensen et al. 2004). Assuming a host galaxy star formationrate of 1 𝑀 (cid:12) yr − , we argue using Equation 1 that host galaxycontamination only becomes significant, i.e. 𝑆 𝜈, gal ≥ ∼ 𝜎 ),at 𝑧 < .
04 ( 𝐷 𝐿 <
200 Mpc). For galaxies with a more active starformation rate of 10 𝑀 (cid:12) yr − , the host galaxy contamination wouldbe significant for bursts detected at 𝑧 < .
12 ( 𝐷 𝐿 <
580 Mpc).Since we expected the brightest radio afterglows to be detectable atflux densities ≥ .
50 mJy up to 𝑧 = . 𝐷 𝐿 = 𝐷 𝐿 <
20 Mpc due totheir lower typical luminosities (Wang et al. 1997). Likewise, ra-dio telescopes with higher angular resolution are also more likelyto result in detections corresponding to scenarios (i) or (ii). Radioafterglows detectable in our search needed to be distinguishablefrom other transient phenomena or disentangled from any coinci-dent background sources. This required further multi-wavelengthanalysis and telescope time as demonstrated later in §6.
We compiled a comprehensive catalogue of lGRBs detected in thepost-
Swift era, specifically between 2004 December and 2020 Jan-uary (inclusive). This consisted of 3 ,
005 bursts primarily detectedby the International Gamma-Ray Astrophysics Laboratory (INTE-GRAL; Mereghetti et al. 2003), the
Swift
Burst Alert Telescope(BAT; Lien et al. 2016) and the
Fermi
Large Area Telescope (LAT;Ajello et al. 2019) and Gamma-ray Burst Monitor (GBM; von Kien-lin et al. 2020). These missions typically localised the bursts to afew arcminutes, with the exception of the
Fermi -GBM, which hada cruder typical localisation radius of a few degrees. A better lo-calisation to a few arcseconds is often achieved with the automaticsearch for X–ray and optical afterglows following a trigger fromthe aforementioned missions with the
Swift
X–ray Telescope (XRT; Evans et al. 2009) and
Swift
Ultraviolet/Optical Telescope (UVOT;Roming et al. 2017), respectively. For our search method and anal-ysis, we required localisation to the level of arcseconds so we onlyconsidered lGRBs detected by
Swift ; the wider catalogue we com-piled consisting of bursts from other missions was then used as areference catalogue to assess our search strategy and completenessagainst in §7.2.We selected from the 1 ,
225 lGRBs detected by
Swift up until2020 January (inclusive) 779 lGRBs, representing 63.6 per cent ofthe full
Swift sample, which were localised in the RACS footprintby
Swift -XRT/UVOT to < (cid:48)(cid:48) . The primary goal of our searchwas to find and study bright late-time radio emission from thissubset of well-localised GRBs. We crossmatched our catalogue ofwell-localised GRBs against sources extracted from each epoch ofRACS and VAST-P1 using an association radius of 15 (cid:48)(cid:48) , which is theangular resolution for the observations. For all source associations,we used information on the positions of host galaxies and opticaldata (a demonstration is given in §6) as well as the detectabilitycriteria discussed in §4 to determine whether the emission was likelyhost or afterglow related. For any afterglow candidate(s) found in oursearch, we searched archival data to determine whether any furtherfollow-up observations were required to confirm and/or characterisethe physics of the emission.There is a possibility that some of these candidates were due tochance spatial coincidence with a background source. We estimatedthe number of chance coincidence sources N cc in our search with: N cc ≈ N GRB × 𝐴 ar × 𝜌 vp , (2)where N GRB was the number of well-localised GRBs that weresearched for in the RACS and VAST-P1 footprints, 𝐴 ar was thearea on a celestial sphere corresponding to the association radius ofthe targeted search, and 𝜌 vp was the source density in the VAST-P1 data. Since the measured flux density for most candidates areclose to the detection threshold, we used the source density at thedetection threshold 𝜌 vp ( 𝑆 ≥ .
50 mJy ) ≈
80 sources per squaredegree along with our search parameters as inputs for Equation 2and estimated N cc to be 3.45. We also estimated N cc numerically byrepeating the crossmatch 100 times using simulated GRB positions,with right ascension and cosine of the declination (up to declination + ◦ ) generated from a random uniform distribution. The numericalmethod estimates N cc = . ± .
85, which is in agreement withour analytical estimate.
We found four radio source associations in our search. Using thenumerical estimate of the expected number of chance coincidencesources, we calculated the probability of all four candidates, at leastone candidate and no candidates being afterglow related (ratherthan chance coincidence sources) to be 2.4 per cent, 65.9 per centand 34.1 per cent, respectively. Three (GRBs 080905B, 110312A,160216A) were ruled out as imaging artefacts, host galaxy or back-ground emission as expected from our chance coincidence calcula-tions. The radio source associated with GRB 171205A was the re-maining afterglow candidate. In this section, we discuss its late-timedetection, our follow-up observations with the Australia TelescopeCompact Array (ATCA) radio telescope, our reduction of archivalearly-time radio data and our compilation of multi-wavelength dataon this afterglow. More comprehensive explanations on the vettingof other candidates are provided in Appendix A.
MNRAS000
MNRAS000 , 1–17 (2021) search for GRB afterglows with ASKAP Table 2.
Radio observations of the GRB 171205A afterglow. Column 1 shows the start date of the observations or the mean epoch of the observations forthose spanning more than two days; Column 2 shows the time of the observations in days post-burst; Column 3 shows the radio telescope or survey used forthe observations; Column 4 shows the central frequency of the observations; Column 5 shows the flux-density measurements for detections or the 3 𝜎 -limitsfor non-detections; Column 6 shows references to the initial reports of the observations. Observations above the dividing line are early time observations takenwithin 3 months of the event and observations below the dividing line are late time observations conducted during or after the epoch of our radio afterglowsearch. Date (UT) Δ T(days) Telescope(or Survey) 𝜈 (GHz) 𝑆 𝜈 † (mJy) Reference2017 Dec 09.58 4.31 VLA 5.0 2 . ± .
12 Urata et al. (2019); Laskar et al. (2017)7.1 4 . ± .
05 ""8.5 5 . ± .
05 ""11.0 8 . ± .
06 ""13.5 11 . ± .
09 ""16.0 14 . ± .
11 ""2017 Dec 10.07 4.76 GMRT 1.4 < .
180 Chandra et al. (2017a)2017 Dec 12.00 6.70 ATCA 5.5 4 . ± .
18 This work9.0 5 . ± .
45 ""44.0 18 . ± . < ± . ± .
057 Chandra et al. (2017b)2017 Dec 20.06 14.75 eMERLIN 5.1 7 . ± . . ± .
47 This work5.2 7 . ± .
12 ""5.5 8 . ± .
13 ""5.8 8 . ± .
14 ""6.3 8 . ± .
14 ""8.2 8 . ± .
15 ""8.7 8 . ± .
16 ""9.0 8 . ± .
16 ""9.3 8 . ± .
17 ""9.8 8 . ± .
16 ""44.0 15 . ± . . ± .
46 Lacy et al. (2020)2018 Feb 19.61 76.00 ATCA 4.7 7 . ± .
27 This work5.2 7 . ± .
20 ""5.5 7 . ± .
21 ""5.8 8 . ± .
41 ""6.3 8 . ± .
32 ""8.2 8 . ± .
25 ""8.7 8 . ± .
28 ""9.0 8 . ± .
27 ""9.3 8 . ± .
29 ""9.8 9 . ± .
33 ""2019 Apr 30.39 ‡ . ± .
76 This work2020 Mar 26.63 § . ± .
70 This work2020 Apr 12.46 859.15 ATCA 2.1 0 . ± .
25 This work5.5 0 . ± .
05 ""9.0 0 . ± .
03 ""16.7 0 . ± .
04 ""21.2 < .
09 ""2020 May 07.26 883.95 ATCA 2.1 0 . ± .
29 This work5.5 0 . ± .
05 ""9.0 0 . ± .
03 ""16.7 0 . ± .
05 ""21.2 < .
12 "" † Uncertainties for GCN Circulars are as reported. Uncertainties for VLASS and RACS measurements consist of astatistical and systematic component ( ∼ ∼
30 per cent for RACS) added together in quadrature. ‡ This is the observation of the afterglow initially detected in our search. § Some tiles of RACS were reobserved. RACSe1 refers to the initial observation and RACSe2 refers to the subsequentobservation. All mentions of RACS observations in the text refer to RACSe1 unless specified otherwise.MNRAS , 1–17 (2021)
J. K. Leung et al.
GRB 171205A was detected by the Burst Alert Telescope aboard
Swift on 2019 December 5 at 07:20:43 UT (D’Elia et al. 2017)and an associated Type Ic-BL supernova SN 2017iuk was detectedtwo days later (de Ugarte Postigo et al. 2017b; Wang et al. 2018).Localised to the galaxy 2MASX J11093966 − 𝑧 = . in prep. ), for accurate flux-density measurements, instead ofthe pre-release version used for our search. The RACS tile con-taining the radio source was reobserved 11 months after the initialobservation. While we also provide the flux-density measurementsfrom the reobservation in our paper, we note that at the time of oursearch, the tile had not been reobserved. The flux-density measure-ments from both the initial and subsequent observations were lowsignal-to-noise detections at 6 to 7 𝜎 ; the integrated flux-densitymeasurements were fitting into the noise and could be unreliableas a result, so we used the peak flux-density measurements insteadassuming the radio source was unresolved.With only one data point at the time of the search, the sourcevariability could not be characterised and the host-afterglow degen-eracy could not be broken. Using Equation 1, the range of possibleSFR estimates for the host galaxy from ∼ 𝑀 (cid:12) yr − (Wang et al.2018) to 3 ± 𝑀 (cid:12) yr − (Perley & Taggart 2017) predicted hostemission ranging from 1.2 to 4.9 mJy; while this is consistent withthe measured flux density of the source, there is considerable un-certainty on the expected flux density of the host galaxy. To furtherverify whether the origin of the radio emission was related to after-glow and not host activity, we searched for evidence of positionaloffset between the fitted radio source position and the host galaxy.Since this offset was unlikely to be resolvable in the RACS data,we instead searched for this in a Panoramic Survey Telescope andRapid Response System (Pan-STARRS; Chambers et al. 2016) 𝑔 -band image with sub-arcsecond resolution as shown in Figure 4.This figure shows the positional offset of the radio source fromthe observed optical afterglow position (Izzo et al. 2017) is 1 . (cid:48)(cid:48) − . (cid:48)(cid:48)
7. The RACS tile for this source has anastrometric accuracy of 2 . (cid:48)(cid:48) ± . (cid:48)(cid:48) . (cid:48)(cid:48) ± . (cid:48)(cid:48) We observed the radio source associated with GRB 171205A twicemore, on 2020 April 12 UT and on 2020 May 7 UT, with ATCAunder project C3363 (PI: T. Murphy). For both epochs, the arraywas in the 6A configuration with a maximum baseline of 6 km andthe source was observed at multiple bands – centred on 2.1, 5.5,9.0, 16.7 and 21.2 GHz – each with a bandwidth of 2 ,
048 MHz.We reduced the visibility data using standard Miriad proce-dures (Sault et al. 1995). We used a combination of manual andautomatic radio-frequency interference flagging before calibration,conducted with Miriad tasks uvflag and pgflag, respectively. Inboth epochs for all frequency bands, we used PKS B1934 − −
145 to calibrate thetime-variable complex gains. We also used PKS B1934 −
638 todetermine the bandpass response in both epochs for all frequencybands, except in the 16.7/21.2 GHz bands. For these two exceptions,we used 3C 279 (B1253 − −
115 for the second epoch. We determinedthe flux-density model for these bandpass calibrators using Miriadtask uvfmeas. For the 2.1 GHz band, we flagged the data from shortbaselines ( < ,
800 m) to reduce the amount of confusion from anextended, known NVSS source, J110939.0 − . (cid:48)(cid:48) × . (cid:48)(cid:48)
83 overlaidin red on top of the Pan-STARRS image in Figure 4. We used a leastsquares fit to constrain the spectral index 𝛼 of the radio source inboth epochs; in the first epoch, 𝛼 = − . ± .
09, and in the secondepoch, 𝛼 = − . ± .
15. With both the flux densities and spectralindices found to be consistent within their uncertainties betweenthe two epochs, we used the data points across both epochs for acombined least squares fit, constraining the spectral index furtherto 𝛼 combined = − . ± . We supplemented our late-time ASKAP and ATCA data witharchival early-time ATCA data to provide a more complete physicalinterpretation for GRB 171205A in §7.1. This subsection describesthese early-time ATCA observations and our data reduction process.
MNRAS000
MNRAS000 , 1–17 (2021) search for GRB afterglows with ASKAP h m s s s s − ◦ Right Ascension (J2000) D ec li n a t i o n ( J ) RACSe1RACSe2
Figure 4.
The Pan-STARRS 𝑔 -band optical image of host galaxy 2MASX J11093966 − 𝜎 -level and increase by a factor of √ (cid:48)(cid:48) , indicative of the astrometric uncertainty of the radio source in RACSe1. The image is 1 (cid:48) × (cid:48) ,centred on the observed position of the GRB optical afterglow from Izzo et al. (2017), with North up and East to the left. GRB 171205A was observed three times with ATCA between2017 December 12 and 2018 February 19 (7 to 76 days post-burst)under target-of-opportunity program CX401 (PI: M. Michałowski).Observations were carried out using four 2048 MHz bands centredon 5.5, 9, 43 and 45 GHz for the first two epochs, while only the5.5 and 9 GHz bands were used for the last epoch. We combinedthe 43/45 GHz data to form a single 4 ,
096 MHz band centred on44 GHz and split the 5.5/9 GHz data into 512 MHz sub-bands aftercalibration to image separately.We reduced the data using Miriad, with automated flaggingapplied and manual flagging where necessary, similar to our late-time ATCA data reduction in §6.2. We used PKS B1934 −
638 asthe primary calibrator to set the flux-density scale for all observa-tions. We also used PKS B1934 −
638 to calibrate the bandpass at5.5/9 GHz, but not at 44 GHz. For the 44 GHz band, we used 3C 279(B1253-055) and PKS B1124-186 to calibrate the bandpass in thefirst and second epochs, respectively. We used PKS B1127-145 andPKS B1124-186 to calibrate the time-variable complex gains at5.5/9 GHz and 44 GHz, respectively. To correct for phase errors, we performed standard self-calibration using Miriad task gpscal, with a single iteration used at5.5 GHz and two at 44 GHz. We detected a point source coincidentwith the position of GRB 171205A in each image at all three epochs.The source was marginally extended at 44 GHz due to scatter in themodel image induced by phase errors, which could not be entirelymitigated with self-calibration. Full details of the flux-density mea-surements are reported in Table 2 (above the divider).
In Figure 5 and Figure 6, we show the multi-wavelength light curveand broadband spectral energy distribution (SED) evolution of theafterglow, respectively. These plots bring together all the multi-wavelength data in the literature on the burst, including radio (Ta-ble 2), millimetre/sub-millimetre (de Ugarte Postigo et al. 2017a;Perley et al. 2017; Smith & Tanvir 2017; Urata et al. 2019), near-
MNRAS , 1–17 (2021) J. K. Leung et al. infrared/optical/ultraviolet (D’Elia et al. 2018; Izzo et al. 2019), andX–ray (flux density at 10 keV; Evans et al. 2010) data points.For the broadband SED in Figure 6, we analysed the X–rayspectra using data from the Swift -XRT, which started observing theburst 145 seconds after the trigger (D’Elia et al. 2017). We extractedthe XRT spectra from the public
Swift repository and analysed themwith Xspec (version 12.10.1f; Arnaud 1996). We fitted the extractedspectra with an absorbed power-law model, adopting the Tubingen-Boulder ISM absorption model (tbabs; Wilms et al. 2000) . TheX–ray spectrum obtained by accumulating all XRT observationsstarting from 4 days post-trigger (225.7 ksec exposure time) was bestfitted by an absorbed power law with equivalent hydrogen columndensity 𝑁 𝐻 = . + . − . × cm − (90 per cent confidence),photon index Γ = . + . − . and de-absorbed flux integrated overthe 0.3–10 keV energy range 𝐹 = . ± . × − erg cm − s − .This photon index, corresponding to a spectral index of 0 . + . − . ,is consistent with the spectral index derived from late-time radioobservations in §6.2. In order to show the evolution of the broadbandSED in Figure 6, we extracted two XRT spectra at 4.3–7.3 days(SP1 – 9 ksec exposure time) and 7.3–14.75 days (SP2 – 38.2 ksecexposure time). We fitted an absorbed power-law model with 𝑁 𝐻 fixed to the value reported above, which is consistent with thatfound from the analysis of the early-time XRT data by D’Elia et al.(2017). The photon spectral indices were Γ SP1 = . + . − . and Γ SP2 = . + . − . , i.e. consistent within their uncertainties. SP1 andSP2 are represented respectively in Figure 6 with blue and redmarkers, corresponding to SED1 and SED3, as described below.The data points in Figure 5 (light curve) are grouped by theirelectromagnetic wavebands, represented by different marker shapesand colours. The radio data points are split into further sub-bandsand are distinguished from other wavelengths by open markers.To highlight the spectral evolution of the afterglow, we grouped theobservations into six gross epochs and show the SED at each of theseepochs in Figure 6 (broadband SED). These epochs correspondto observations taken approximately 4 days (SED1: blue), 7 days(SED2: green), 14 days (SED3: red), 36 days (SED4: pink), 76days (SED5: cyan), and 850 days (SED6: orange) post-burst. Datapoints within these epochs are not perfectly contemporaneous, butwe argue this is acceptable for the subsequent analysis since theevolution of the afterglow on the timescale of days to tens of daysis quite slow. The windows for each epoch are shown by the grayvertical stripes in Figure 5.For each epoch in the radio inset of Figure 6, a spectral fit line(solid) or a comparison line corresponding to a standard afterglowspectral segment (dashed) is displayed on top of the data points. Inthe wider broadband SED plot, data from the first epoch is fit to asmoothly broken power law (blue line, with 1 𝜎 -uncertainty shadedin light blue): 𝑆 𝜈 = 𝑆 peak (cid:20)(cid:0) 𝜈𝜈 peak (cid:1) − 𝑠 𝛿 + (cid:0) 𝜈𝜈 peak (cid:1) − 𝑠 𝛿 (cid:21) − / 𝑠 , (3)where 𝑆 peak is the flux density at the peak, 𝜈 peak is the frequencyat which the turnover occurs, 𝛿 is the slope of the spectral rise, 𝛿 is the slope of the spectral decay, and 𝑠 is the smoothingfactor. The Markov chain Monte Carlo fit was performed using https://heasarc.gsfc.nasa.gov/xanadu/xspec/manual/node265.html the emcee (Foreman-Mackey et al. 2013) Python implementa-tion with uniform priors on all parameters, yielding a fit with 𝑆 peak = . + . − . mJy, 𝜈 peak = . + . − . GHz, 𝛿 = . + . − . , 𝛿 = − . + . − . , and 𝑠 = . + . − . . The value of the fitted peakparameters differ slightly to the true functional peak (of 43.18 mJylocated at 89.00 GHz) due to the smoothing factor. The multi-wavelength light curve in Figure 5 highlights many fea-tures of the GRB 171205A/SN 2017iuk system discussed in theliterature. The X–ray observations probed the prompt emission andrevealed a possible thermal component in the very early phasesof the X–ray light curve; the origins of which is still under de-bate (D’Elia et al. 2018). The near-infrared/optical/ultraviolet wasdominated at early times (up to ∼ ∼ ∼
13 days. We ascribed this radio peak as due to the evolution ofthe characteristic frequencies of the afterglow synchrotron spectrumacross the low-frequency bands.The radio inset of Figure 6 details this spectral evolutionthrough the six epochs described in §6.4. In the first SED (4 dayspost-burst), the radio spectrum featured a steep slope (with somecurvature) consistent with a spectral index 𝛼 = . + . − . by the smoothly broken power law), typical for the self-absorbed segment of the afterglow spectrum. By the second epoch(7 days post-burst), the spectrum exhibited a flatter slope fitted as0 . ± .
01, which we interpreted to be in mid-transition towardsthe 𝛼 = / 𝜈 sa < 𝜈 < 𝜈 m . The slope in the third SED (14 days post-burst)remained consistent with 𝛼 = 𝛼 = 𝛼 = / 𝜈 = 𝛼 = / 𝑟 from the progenitor, is described as 𝑛 ( 𝑟 ) = 𝐴𝑟 − 𝑘 .In the wind (homogeneous ISM) environment, the density drops(remains constant) with distance from the progenitor, i.e. 𝑘 = MNRAS000
01, which we interpreted to be in mid-transition towardsthe 𝛼 = / 𝜈 sa < 𝜈 < 𝜈 m . The slope in the third SED (14 days post-burst)remained consistent with 𝛼 = 𝛼 = 𝛼 = / 𝜈 = 𝛼 = / 𝑟 from the progenitor, is described as 𝑛 ( 𝑟 ) = 𝐴𝑟 − 𝑘 .In the wind (homogeneous ISM) environment, the density drops(remains constant) with distance from the progenitor, i.e. 𝑘 = MNRAS000 , 1–17 (2021) search for GRB afterglows with ASKAP Time post-burst (seconds) − − − − F l u x D e n s i t y ( m J y ) RACS (887.5 MHz)L/S (1 − − −
12 GHz)K u /K/K a (12 −
40 GHz)V (40 −
75 GHz)mm/sub-mm × × × Time post-burst (days)
Figure 5.
Multi-wavelength light curve for GRB 171205A/SN 2017iuk showing both the prompt, thermal and afterglow emission. The light curve showsdata points for all observations of GRB 171205A/SN 2017iuk presented in this work and in the literature to-date. Observations at radio frequencies probe theafterglow emission of the burst. They are represented by open circular markers (3 𝜎 -limits represented by open triangular markers) and are distinguished incolour by their radio band. The details of each radio data point is summarised in Table 2. Observations at other wavelengths may have additional contributionsfrom prompt and thermal supernova emission (see §7.1 text for details). They are represented by solid markers and are distinguished in colour by theirelectromagnetic spectrum classification. For clarity, limits are omitted and data points in the millimetre/sub-millimetre, infra-red, and X–ray (10 keV) bandshave been scaled by a factor of 20, 5, and 50, respectively. The gray-shaded vertical stripes represent the gross epochs from which the SEDs shown in Figure 6were constructed. ( 𝑘 = 𝐴 is just the particle densityof the ISM in the homogeneous scenario, 𝑛 , it is related to themass-loss rate (cid:164) 𝑀 𝑤 and stellar wind velocity 𝑣 𝑤 of the progenitorstar by 𝐴 = (cid:164) 𝑀 𝑤 / 𝜋𝑣 𝑤 = × 𝐴 ∗ g cm − in the wind scenario.Here, 𝐴 ∗ is a normalisation of 𝐴 , such that 𝐴 ∗ = (cid:164) 𝑀 𝑤 = − 𝑀 (cid:12) yr − and stellar wind velocity 𝑣 𝑤 = ,
000 km s − .We argue the GRB 171205A/SN 2017iuk event occurred ina stellar wind environment induced by the progenitor prior to itscore-collapse (see also Suzuki et al. 2019). The radio spectral evo-lution in the first five epochs, in particular, the emergence of the 𝛼 = / 𝜈 sa , across the gigahertz frequencyrange. According to standard afterglow theory (e.g. GS02), 𝜈 sa onlyevolves with time, shifting towards lower frequencies, in the windscenario. The flux-density levels in the fourth and fifth SED, sam-pled after the passage of 𝜈 sa through the band, were also very similar;this is consistent with the wind scenario, where the flux density ispredicted to remain constant in the 𝜈 sa < 𝜈 < 𝜈 m spectral segment(GS02). Our conclusion that the afterglow is produced by the rela-tivistic outflow decelerating in a circumburst medium with a winddensity profile is in agreement with the conclusion obtained by the independent analysis of Maity & Chandra (2020) using observationsfrom the upgraded Giant Metrewave Radio Telescope.The very late-time SED (850 days post-burst) enabled us toconstrain the slope of the electron spectrum 𝑝 , where the distributionof electron energies is 𝑁 ( 𝐸 ) ∝ 𝐸 − 𝑝 . According to Maity & Chandra(2020), even at ∼ ,
000 days post-burst, the blast wave had nottransitioned to the non-relativistic (Newtonian) regime. A spectralinversion occurred between the fifth and sixth SED (76 and 850 dayspost-burst), where the spectral index transitioned from 𝛼 = / 𝛼 = − .
92. This indicates that 𝜈 m had moved from between themillimetre and radio frequencies at 4 days post-burst and passedthrough the entire radio frequency band shown in the SED at sometime between 76 and 850 days post-burst; by 850 days post-burst, 𝜈 m was located below the 887 . 𝜈 m < 𝜈 < 𝜈 c , where 𝜈 c is the frequency ofthe cooling break, with characteristic spectral index 𝛼 = ( − 𝑝 )/ 𝛼 = − . ± .
07, we constrained the slope of the electronspectrum to 𝑝 = .
84. While this value deviates from the typical 𝑝 ∼ . 𝑝 -valuesobtained from afterglow modelling (Ryan et al. 2015; Wang et al. MNRAS , 1–17 (2021) J. K. Leung et al. a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a ν (Hz) − − − − − − F l u x D e n s i t y ( m J y ) SED1: 4dSED2: 7dSED3: 14d SED4: 36dSED5: 76dSED6: 850d1 5 10 50 ν (GHz) F l u x D e n s i t y ( m J y ) ν ν . ν ν / ν − . Figure 6.
The broadband SED of the GRB 171205A afterglow taken at six epochs (approximately 4, 7, 14, 36, 76, 850 days post-burst) from X–ray downto radio frequencies, with the inset zooming into the radio frequencies. Observations in the near-infrared to ultraviolet regime are shown in gray and are notincluded in any fits because they are contaminated by an additional thermal supernova component. Non-detections are given by their 3 𝜎 -limits (downwardtriangular markers). Spectral fits to data points are displayed as solid lines, while comparison lines representing typical spectral indices from segments of thestandard afterglow spectrum are displayed as dashed lines. For SED1 (4 days post-burst), a smoothly broken power-law fit is applied (blue, with 1 𝜎 -uncertaintyshaded in light blue) to show the injection frequency 𝜈 m located between the radio and millimetre/sub-millimetre frequencies. By analysing the evolution ofthe radio and broadband SED across the six epochs, we inferred the circumburst medium of the burst to be consistent with a stellar wind environment and thelate-time radio emission (from RACS and ATCA observations) to be afterglow dominated (see §7.1 text for details). 𝑝 .The fit of SED1 (blue points in Figure 6) with a single spectralcomponent between the peak in the radio and the X–ray data pointssuggests that already at these early epochs the cooling frequency 𝜈 c lies above the X–ray energy range. This is consistent with thesimilar spectral slopes found in the two X–ray spectra we analysedsince 𝜈 c should increase with time in the wind scenario.The broadband SED corresponding to SED1–3 in Figure 6shows that the radio/millimetre (above 89 GHz) through X–ray spec-trum is consistent with a single non-thermal component, while thenear-infrared/optical data points (gray markers) contaminated bythermal supernova emission are located above the correspondingempirical fit (blue line with shaded region). Moreover, the extrap-olation of the very late-time 𝛼 = − .
92 spectral slope from theradio frequencies through to the X–rays is consistent with the verylate-time upper limits derived from XRT observations. Along withthe consistency of the X–ray spectral indices presented in §6.4 withthe very late-time radio spectral index, these suggest the emissionwe sampled in the radio band starting from 4 days post-burst wasdominated by the afterglow component. As the radio SEDs are dominated by the afterglow componentof the burst event, we used these observables to provide reasonableestimates for the microphysical shock parameters of the blast wave,including the normalised wind-density parameter 𝐴 ∗ , the fractionalenergy in electrons 𝜖 𝑒 , and the fractional energy in the magneticfield 𝜖 𝐵 . These observables are related to the microphysical shockparameters by a set of analytical relations detailed in Panaitescu &Kumar (2000). For a stratified wind medium, the self-absorptionfrequency is: 𝜈 sa = . × 𝐸 − / 𝐴 / ∗ 𝜖 − 𝑒, − 𝜖 / 𝐵, − 𝑇 − / 𝑑 Hz , (4)where 𝐸 = 𝐸 erg is the kinetic energy of the blast wave and 𝑇 𝑑 is the time (in days) since the burst explosion . The flux densityat the self-absorption frequency, 𝑆 𝜈, sa , is then related to 𝜈 sa at time 𝑇 𝑑 by the relation: 𝑆 𝜈, sa = . 𝐷 − 𝐿, 𝐸 𝐴 − ∗ 𝜖 𝑒, − 𝜈 , . 𝑇 𝑑, − mJy , (5)where 𝐷 𝐿 = 𝐷 𝐿, cm is the luminosity distance of the source. We note that times and frequencies should be converted from the sourceto the observer frame accounting for the source redshift. However, in thiscase, the 𝑧 = .000
92 spectral slope from theradio frequencies through to the X–rays is consistent with the verylate-time upper limits derived from XRT observations. Along withthe consistency of the X–ray spectral indices presented in §6.4 withthe very late-time radio spectral index, these suggest the emissionwe sampled in the radio band starting from 4 days post-burst wasdominated by the afterglow component. As the radio SEDs are dominated by the afterglow componentof the burst event, we used these observables to provide reasonableestimates for the microphysical shock parameters of the blast wave,including the normalised wind-density parameter 𝐴 ∗ , the fractionalenergy in electrons 𝜖 𝑒 , and the fractional energy in the magneticfield 𝜖 𝐵 . These observables are related to the microphysical shockparameters by a set of analytical relations detailed in Panaitescu &Kumar (2000). For a stratified wind medium, the self-absorptionfrequency is: 𝜈 sa = . × 𝐸 − / 𝐴 / ∗ 𝜖 − 𝑒, − 𝜖 / 𝐵, − 𝑇 − / 𝑑 Hz , (4)where 𝐸 = 𝐸 erg is the kinetic energy of the blast wave and 𝑇 𝑑 is the time (in days) since the burst explosion . The flux densityat the self-absorption frequency, 𝑆 𝜈, sa , is then related to 𝜈 sa at time 𝑇 𝑑 by the relation: 𝑆 𝜈, sa = . 𝐷 − 𝐿, 𝐸 𝐴 − ∗ 𝜖 𝑒, − 𝜈 , . 𝑇 𝑑, − mJy , (5)where 𝐷 𝐿 = 𝐷 𝐿, cm is the luminosity distance of the source. We note that times and frequencies should be converted from the sourceto the observer frame accounting for the source redshift. However, in thiscase, the 𝑧 = .000 , 1–17 (2021) search for GRB afterglows with ASKAP A similar relation to Equation 4 providing the location of the injec-tion frequency, 𝜈 m , at time 𝑇 𝑑 is given by: 𝜈 m = . × 𝐸 / 𝜖 𝑒, − 𝜖 / 𝐵, − 𝑇 − / 𝑑 Hz . (6)To estimate the microphysical shock parameters using our ob-servables and these relations, we fixed the luminosity distance pa-rameter to 𝐷 𝐿 = . × cm corresponding to the afterglowredshift of 𝑧 = . 𝐸 = . 𝐸 prompt = . × erg (D’Elia et al. 2018).In Equations 4 and 5, we estimated the time 𝑇 𝑑 and flux density 𝑆 𝜈, sa at which the self-absorption frequency 𝜈 sa passed 5 GHz usingSED3 and SED4 (14 and 36 days post-burst) in Figure 6. In the windscenario, the flux density is expected to rise linearly with 𝑡 in the self-absorbed spectral segment and stay constant in the spectral segmentbetween 𝜈 sa and 𝜈 m . Given this, we argue, with the 5.1 GHz datapoint shown on SED3 lying on the 𝛼 = / 𝜈 sa was likely already at 5 GHz around 14 days post-burst. We thus used the self-absorption break at 𝜈 sa = 𝑆 𝜈, sa = . 𝑇 𝑑 =
14 days as our observables for Equations4 and 5. These numbers are only estimates since we do not see aclear self-absorption break in any SED; 𝜈 sa may have crossed 5 GHzat any time between 14 days and 36 days. For Equation 6, while noneof the radio SEDs in Figure 6 show a spectral break correspondingto the injection frequency 𝜈 m , SED6 (850 days post-burst) providesan observable boundary condition; it suggests 𝜈 m at 𝑇 𝑑 =
850 dayspost-burst lies below 887 . 𝐴 ∗ , 𝜖 𝑒 and 𝜖 𝐵 , we solved the set of non-linearEquations 4–6 using a non-linear least squares optimisation rou-tine in the SciPy (Virtanen et al. 2020) Python library. We im-posed the following constraints on the free parameters: 𝐴 ∗ > < 𝜖 𝑒 , 𝜖 𝐵 < < 𝜈 m < . 𝐴 ∗ = 𝜖 𝑒 = . 𝜖 𝐵 = . 𝜈 m =
46 MHz for thissolution. The low value of 𝜖 𝐵 we derived is consistent with thewide distributions of values estimated from the modelling of GRBafterglows (e.g. 10 − to 10 − in Barniol Duran 2014, 10 − to 10 − in Santana et al. 2014). These low values can explain the proper-ties of the afterglow emission at very high energies as detected by Fermi -LAT (Beniamini et al. 2015) and have been interpreted to bedue to the possible decay of the magnetic field in the downstreamregion behind the external shock (Lemoine et al. 2013). While wehave shown plausible estimates of the microphysical parameters canbe derived from our data, these estimates are very sensitive to ourassumptions; more accurate estimates would require full afterglowmodelling across relativistic and sub-relativistic regimes, which isbeyond the scope of this paper.
In this work, we limited our search to well-localised bursts to reducethe number of false, or coincident, associations. Since this locali-sation is often achieved by
Swift -XRT/UVOT, we introduced a biastowards bursts with detectable afterglows at X–ray, UV and opti-cal wavelengths. For example, bursts with detectable afterglows aremore likely to be associated with certain physical properties, suchas a higher circumburst density (e.g. Piran 2004). If we removedthis localisation requirement and searched for radio afterglow can-didates within the GRB error radius, we would be able to provide follow-up observations for approximately three times as many, or ∼ ,
000 more, GRBs.In this scenario, the problem of false associations becomessignificant. Using a typical error radius of 5 (cid:48) in Equation 2, wefind that the number of expected false associations would be 1 , ∼ ∼
7. When a deep, singleepoch sky map from EMU (see §7.3) is completed, it can be usedas a reference epoch for this type of time-domain analysis; thiswill reduce the false association rate for searches in future surveys,regardless of whether these surveys are multi-epoch or not.Due to these considerations, there were significant challengesin using previous generations of wide coverage radio surveys, inparticular, NVSS and SUMSS, to conduct a similar search. Thesesurveys conducted in the pre-
Swift era would have significantlyfewer well-localised bursts to follow-up and the lack of multi-epochdata hindered the ability to conduct further variability analysis forthese bursts with larger error radius.Given our choice to limit our search to well-localised bursts,we also did not consider any bursts occurring prior to 2004, thelaunch year of
Swift . Due to this constraint, our search was unableto test late-time rebrightening theories for GRB/SN systems sincethis phenomenon is predicted to occur on the timescales of a fewdecades and usually at sub-mJy flux-density levels (e.g. Peters et al.2019). Still, our search was sensitive to local, radio-bright GRB/SNafterglows as demonstrated by the detection of late-time emissionfrom the GRB 171205A/SN 2017iuk system. This suggests orphanafterglow searches using the current generation of wide coverageradio surveys, such as RACS and VAST-P1, may also be sensitiveenough to detect off-axis emission from these systems, providingan alternative to orphan afterglow searches guided by supernovaassociation (e.g. off-axis GRB detection via association with TypeIc-BL SN 2020bvc; Izzo et al. 2020).
Our ASKAP search for radio afterglows probed for radio emissionfrom 587 GRBs that were not previously observed by the ArcminuteMicrokelvin Imager (AMI; Anderson et al. 2018), ATCA, or VLA,and resulted in one detection. With the sensitivity improvementsexpected in future radio surveys, for example, the SKA-Shallowsurvey targeting a 5 𝜎 -detection threshold of 1 µ Jy beam − , simi-lar searches will be able to study greater classes of GRB events,including sGRBs, useful for identifying electromagnetic counter-parts to gravitational wave events, and high-redshift GRBs, usefulfor probing Population-III stars. These expected improvements tosensitivity in such searches will enable sufficient detections forpopulation analyses of radio afterglows to be conducted, providinguseful experiments to better understand the nature of radio-quiet and-loud GRBs, for example. In this subsection we compare the GRByields of our search using RACS and VAST-P1 to various surveys, MNRAS , 1–17 (2021) J. K. Leung et al. including the EMU Pilot Survey, commensal ASKAP surveys andthe SKA-Shallow survey.Our search covered ∼ ,
000 square degrees and the detec-tion of one low-luminosity GRB afterglow above our 5 𝜎 -sensitivitythreshold of ∼ .
50 mJy beam − implies a transient surface density 𝜌 (cid:38) × − deg − . This is consistent with the gigahertz frequencytransient surface density prediction for low-luminosity GRBs in Fig-ure 3 of Metzger et al. (2015) calculated to be 𝜌 ≈ × − deg − .Following our detection of the GRB 171205A radio afterglow upto 842 days post-burst in RACS, we extended the time baselinefor our rates analysis in §3 from 2011 to 2020. We searched on-line databases for radio afterglow detections up to 2020 Januaryand manually searched the literature for radio flux-density measure-ments of each detection. We found that aside from GRB 171205A,no lGRB radio afterglows reached a peak flux density above the 5 𝜎 -sensitivity threshold of ∼ .
50 mJy beam − at 887.5 MHz. However,the radio afterglow from Very High Energy ( 𝐸 >
100 GeV) lGRB190829A, occurring in the VAST-P1 footprint two days after ourfirst epoch of observations, may have been marginally detectable,i.e. at 4 𝜎 -significance, with more favourable timing, having peakedat 1 . ± .
22 mJy beam − at an observing frequency of 1.3 GHz(Rhodes et al. 2020). The new information improves the probabilityof detecting a radio afterglow calculated in §3 to 13.3 per cent (11.3per cent), assuming the 𝛼 = / ( ) spectral scaling scenario. Whilethis probability suggests the detection of GRB 171205A afterglowin our search was unlikely, its detection could be attributed to theextreme nature of the event, being the second brightest afterglowdetected to date at radio frequencies.Surveys reaching lower flux-density thresholds are requiredto probe for classes of GRBs beyond the subset of local, radio-bright GRB/SN systems. We consider the expected GRB yield ofsome of these surveys here, assuming a more conventional 1.4 GHzobserving frequency, spectral scaling with 𝛼 = / 𝜎 -sensitivity threshold of 100 µ Jy beam − ,but covers ∼
10 per cent ( ∼ 𝑧 = . . (cid:48)(cid:48) ∼ . ∼ . https://swift.gsfc.nasa.gov/archive/grb_table.html instead, the angular offset would be < (cid:48)(cid:48) . Variability analysis overmultiple epochs would be required and this is not possible with thesingle epoch EMU Pilot Survey.The full ASKAP surveys (e.g. EMU and VAST) will be con-ducted with a commensal approach; depending on the cadence,multi-epoch deep surveys with large sky coverage may be available.For example, the full EMU survey aims to cover the entire RACSfootprint at the EMU Pilot Survey sensitivity, requiring ∼
10 hr ofintegration per tile. In the extreme cases, the 10 hours could be ob-served through 10 min pointings with 60 visits per tile over 5 years(i.e. 1 visit per month) or through two 5 hr pointings per tile sepa-rated by over a year; a hybrid between these two extremes is likelymore realistic. The former will reach a 5 𝜎 -sensitivity thresholdsimilar to our RACS/VAST-P1 search and has a 47.8 per cent prob-ability (based of a similar analysis to that in §3) of detecting a radioafterglow from a known local radio-bright GRB/SN system, whilethe latter will reach a 5 𝜎 -sensitivity threshold of ∼ µ Jy beam − and has a 62.4 per cent (16.2 per cent if we do not account for GRBsthat do not receive radio follow-up) probability of detecting a radioafterglow from a broader class of known on-axis GRBs.These ASKAP surveys, which are expected to yield a num-ber of radio afterglow detections of order unity, are precursors tomore potent SKA-era surveys. The proposed SKA-Shallow sur-vey, for example, covering half the sky to sub- µ Jy sensitivity withdaily cadence would provide radio follow-up to all bursts in thehalf-sky and deeper non-detection limits than any other radio tele-scope to date. While the ASKAP surveys used in our work still donot reach the flux-density sensitivity required for placing stringenttests on radio rebrightening theories and conjectures regarding dis-tinct radio-quiet/loud GRB populations, applying a similar searchmethodology and analysis to SKA-era surveys will enable these ex-periments as well as further insights into the general radio afterglowpopulation to be realised.
We have conducted a search for radio afterglows from 779 well-localised lGRBs occurring after 2004 by crossmatching their posi-tions with radio sources in the RACS and VAST-P1 data. Analysingthe positional offset of the radio source and host galaxy in opticalimages was the primary method we used to distinguish afterglowfrom background or host related emission for matched sources. Oursearch produced four candidates, three of which were ruled out asbackground source or host-galaxy related, while one was confirmedas a radio afterglow from GRB 171205A, a low-luminosity lGRBassociated with SN 2017iuk.We conducted further late-time observations of theGRB 171205A radio afterglow with ATCA at 859 and 884 dayspost-burst. Both the flux densities and spectral indices were mea-sured to be consistent within their uncertainties between the twoepochs. Our late-time radio observations provided a robust measure-ment of the electron spectral index and also provided the boundedconstraint on the injection frequency (see Equation 6) required tosolve for the other microphysical parameters. Using this late-timedata alongside archival data from early-time radio observations, weshowed the progenitor of this GRB-SN system originated in a stellarwind environment with microphysical shock parameters of the burstestimated as 𝑝 = . 𝐴 ∗ = 𝜖 𝑒 = . 𝜖 𝐵 = . <
10 per cent for detectingany radio afterglows in our search, we were able to probe for asubset of radio-bright lGRBs in the local Universe at 𝑧 < . MNRAS000
10 per cent for detectingany radio afterglows in our search, we were able to probe for asubset of radio-bright lGRBs in the local Universe at 𝑧 < . MNRAS000 , 1–17 (2021) search for GRB afterglows with ASKAP 𝐷 𝐿 < ACKNOWLEDGEMENTS
We thank the anonymous referee and Phil Edwards for useful com-ments that improved the quality of this manuscript. JL, DD and JPare supported by Australian Government Research Training Pro-gram Scholarships. TM acknowledges the support of the AustralianResearch Council through grants FT150100099 and DP190100561.DK is supported by NSF grant AST-1816492. Parts of this researchwere conducted by the Australian Research Council Centre of Excel-lence for Gravitational Wave Discovery (OzGrav), project numberCE170100004.The Australian Square Kilometre Array Pathfinder is part of theAustralia Telescope National Facility which is managed by CSIRO.Operation of ASKAP is funded by the Australian Government withsupport from the National Collaborative Research InfrastructureStrategy. ASKAP uses the resources of the Pawsey SupercomputingCentre. Establishment of ASKAP, the Murchison Radio-astronomyObservatory and the Pawsey Supercomputing Centre are initiativesof the Australian Government, with support from the Government ofWestern Australia and the Science and Industry Endowment Fund.We acknowledge the Wajarri Yamatji as the traditional owners ofthe Murchison Radio-astronomy Observatory site. The AustraliaTelescope Compact Array is part of the Australia Telescope Na-tional Facility which is funded by the Australian Government foroperation as a National Facility managed by CSIRO. We acknowl-edge the Gomeroi people as the traditional owners of the Paul WildObservatory site.The Pan-STARRS1 Surveys (PS1) and the PS1 public sciencearchive have been made possible through contributions by the In-stitute for Astronomy, the University of Hawaii, the Pan-STARRSProject Office, the Max-Planck Society and its participating in-stitutes, the Max Planck Institute for Astronomy, Heidelberg andthe Max Planck Institute for Extraterrestrial Physics, Garching,The Johns Hopkins University, Durham University, the Univer-sity of Edinburgh, the Queen’s University Belfast, the Harvard-Smithsonian Center for Astrophysics, the Las Cumbres Observa-tory Global Telescope Network Incorporated, the National CentralUniversity of Taiwan, the Space Telescope Science Institute, theNational Aeronautics and Space Administration under Grant No.NNX08AR22G issued through the Planetary Science Division ofthe NASA Science Mission Directorate, the National Science Foun-dation Grant No. AST-1238877, the University of Maryland, EotvosLorand University (ELTE), the Los Alamos National Laboratory,and the Gordon and Betty Moore Foundation. The national facilitycapability for SkyMapper has been funded through ARC LIEF grantLE130100104 from the Australian Research Council, awarded to theUniversity of Sydney, the Australian National University, SwinburneUniversity of Technology, the University of Queensland, the Uni- versity of Western Australia, the University of Melbourne, CurtinUniversity of Technology, Monash University and the AustralianAstronomical Observatory. SkyMapper is owned and operated byThe Australian National University’s Research School of Astron-omy and Astrophysics. This work made use of data supplied by theUK Swift Science Data Centre at the University of Leicester.
Software:
Astropy (Astropy Collaboration et al. 2013, 2018),emcee (Foreman-Mackey et al. 2013), matplotlib (Hunter 2007),NumPy (Harris et al. 2020), pandas (McKinney 2010), SciPy (Vir-tanen et al. 2020), ASKAPsoft (Cornwell et al. 2011; Guzman et al.2019), Selavy (Whiting & Humphreys 2012), Miriad (Sault et al.1995), and Xspec (Arnaud 1996).
DATA AVAILABILITY
The ASKAP data used in this paper (RACS and VAST-P1) canbe accessed through the CSIRO ASKAP Science Data Archive(CASDA ) under project codes AS110 and AS107. The ATCA dataused in this paper can be accessed through the Australia TelescopeOnline Archive (ATOA ) under project codes CX401 and C3363.Other auxiliary datasets can be made available upon request viaemail to the corresponding author. REFERENCES
Ajello M., et al., 2019, ApJ, 878, 52Akerlof C., et al., 1999, Nature, 398, 400Anderson G. E., et al., 2018, MNRAS, 473, 1512Arnaud K. A., 1996, in Jacoby G. H., Barnes J., eds, Astronomical Societyof the Pacific Conference Series Vol. 101, Astronomical Data AnalysisSoftware and Systems V. p. 17Astropy Collaboration et al., 2013, A&A, 558, A33Astropy Collaboration et al., 2018, AJ, 156, 123Barniol Duran R., 2014, MNRAS, 442, 3147Barniol Duran R., Giannios D., 2015, MNRAS, 454, 1711Beniamini P., Nava L., Duran R. B., Piran T., 2015, MNRAS, 454, 1073Berger E., 2009, ApJ, 690, 231Blanchard P. K., Berger E., Fong W.-f., 2016, ApJ, 817, 144Bloom J. S., Kulkarni S. R., Djorgovski S. G., 2002, AJ, 123, 1111Burlon D., Murphy T., Ghirlanda G., Hancock P. J., Parry R., Salvaterra R.,2016, MNRAS, 459, 3356Carilli C. L., Yun M. S., 1999, ApJ, 513, L13Chambers K. C., et al., 2016, arXiv e-prints, p. arXiv:1612.05560Chandra P., Frail D. A., 2012, ApJ, 746, 156Chandra P., Nayana A. J., Bhattacharya D., Cenko S. B., Corsi A., 2017a,GRB Coordinates Network, 22222, 1Chandra P., Nayana A. J., Bhattacharya D., Cenko S. B., Corsi A., 2017b,GRB Coordinates Network, 22264, 1Chevalier R. A., Li Z.-Y., 2000, ApJ, 536, 195Christensen L., Hjorth J., Gorosabel J., 2004, A&A, 425, 913Ciardi B., Loeb A., 2000, ApJ, 540, 687Clark B. G., 1980, A&A, 89, 377Condon J. J., 1992, ARA&A, 30, 575Condon J. J., Cotton W. D., Greisen E. W., Yin Q. F., Perley R. A., TaylorG. B., Broderick J. J., 1998, AJ, 115, 1693Cornwell T., Humphreys B., Lenc E., Voronkov M., Whiting M., 2011,ASKAP-SW-0020: ASKAP Science Processing, ASKAP ScienceCase Memo Series 027, https://data.csiro.au/dap/public/casda/casdaSearch.zul https://atoa.atnf.csiro.au/query.jsp MNRAS , 1–17 (2021) J. K. Leung et al.
D’Elia V., D’Ai A., Lien A. Y., Sbarufatti B., 2017, GRB CoordinatesNetwork, 22177, 1D’Elia V., et al., 2018, A&A, 619, A66Djorgovski S. G., Frail D. A., Kulkarni S. R., Bloom J. S., Odewahn S. C.,Diercks A., 2001, ApJ, 562, 654Evans P. A., et al., 2009, MNRAS, 397, 1177Evans P. A., et al., 2010, A&A, 519, A102Fender R., Stewart A., Macquart J. P., Donnarumma I., Murphy T., DellerA., Paragi Z., Chatterjee S., 2015, Advancing Astrophysics with theSquare Kilometre Array (AASKA14), p. 51Fong W., Berger E., Margutti R., Zauderer B. A., 2015, ApJ, 815, 102Foreman-Mackey D., Hogg D. W., Lang D., Goodman J., 2013, PASP, 125,306Frail D. A., Waxman E., Kulkarni S. R., 2000, ApJ, 537, 191Frail D. A., Soderberg A. M., Kulkarni S. R., Berger E., Yost S., Fox D. W.,Harrison F. A., 2005, ApJ, 619, 994Frail D. A., et al., 2006, ApJ, 646, L99Galama T. J., et al., 1998, Nature, 395, 670Ghirlanda G., et al., 2013, MNRAS, 435, 2543Ghirlanda G., et al., 2014, Publ. Astron. Soc. Australia, 31, e022Ghisellini G., Nardini M., Ghirlanda G., Celotti A., 2009, MNRAS, 393,253Goodman J., 1997, New Astron., 2, 449Granot J., Sari R., 2002, ApJ, 568, 820Granot J., van der Horst A. J., 2014, Publ. Astron. Soc. Australia, 31, e008Granot J., Piran T., Sari R., 1999, ApJ, 513, 679Guzman J., et al., 2019, ASKAPsoft: ASKAP science data processor soft-ware, Astrophysics Source Code Library, record ascl:1912.003
Hallinan G., et al., 2017, Science, 358, 1579Hancock P. J., Gaensler B. M., Murphy T., 2013, ApJ, 776, 106Harris C. R., et al., 2020, Nature, 585, 357Hjorth J., et al., 2003, Nature, 423, 847Högbom J. A., 1974, A&AS, 15, 417Horesh A., Cenko S. B., Perley D. A., Kulkarni S. R., Hallinan G., BellmE., 2015, ApJ, 812, 86Hunter J. D., 2007, Computing In Science & Engineering, 9, 90Izzo L., et al., 2017, GRB Coordinates Network, 22180, 1Izzo L., et al., 2019, Nature, 565, 324Izzo L., Auchettl K., Hjorth J., De Colle F., Gall C., Angus C. R., RaimundoS. I., Ramirez-Ruiz E., 2020, A&A, 639, L11Kulkarni S. R., et al., 1998, Nature, 395, 663Kulkarni S. R., et al., 1999, ApJ, 522, L97Lacy M., et al., 2020, PASP, 132, 035001Laskar T., Coppejans D. L., Margutti R., Alexand er K. D., 2017, GRBCoordinates Network, 22216, 1Laskar T., Hull C. L. H., Cortes P., 2020, ApJ, 895, 64Lemoine M., Li Z., Wang X.-Y., 2013, MNRAS, 435, 3009Levan A. J., et al., 2014, ApJ, 781, 13Levinson A., Ofek E. O., Waxman E., Gal-Yam A., 2002, ApJ, 576, 923Li Z.-Y., Chevalier R. A., 1999, ApJ, 526, 716Lien A., et al., 2016, ApJ, 829, 7Lloyd-Ronning N. M., Fryer C. L., 2017, MNRAS, 467, 3413Lloyd-Ronning N. M., Gompertz B., Pe’er A., Dainotti M., Fruchter A.,2019, ApJ, 871, 118Ma C., et al., 2009, IERS Technical Note, 35, 1MacFadyen A. I., Woosley S. E., 1999, ApJ, 524, 262Maity B., Chandra P., 2020, arXiv e-prints, p. arXiv:2012.05166Mauch T., Murphy T., Buttery H. J., Curran J., Hunstead R. W., PiestrzynskiB., Robertson J. G., Sadler E. M., 2003, MNRAS, 342, 1117Mazzali P. A., McFadyen A. I., Woosley S. E., Pian E., Tanaka M., 2014,MNRAS, 443, 67McConnell D., et al., 2020, Publ. Astron. Soc. Australia, 37, e048McKinney W., 2010, in Stéfan van der Walt Jarrod Millman eds,Proceedings of the 9th Python in Science Conference. pp 56–61,doi:10.25080/Majora-92bf1922-00aMereghetti S., Götz D., Borkowski J., Walter R., Pedersen H., 2003, A&A,411, L291Metzger B. D., Williams P. K. G., Berger E., 2015, ApJ, 806, 224 Modjaz M., Liu Y. Q., Bianco F. B., Graur O., 2016, ApJ, 832, 108Murphy T., et al., 2013, Publ. Astron. Soc. Australia, 30, e006Nakar E., Piran T., 2011, Nature, 478, 82Norris R. P., et al., 2011, Publ. Astron. Soc. Australia, 28, 215Onken C. A., et al., 2019, Publ. Astron. Soc. Australia, 36, e033Panaitescu A., Kumar P., 2000, ApJ, 543, 66Perez-Torres M., Cenko S. B., Horesh A., Alberdi A., 2018a, GRB Coordi-nates Network, 22302, 1Perez-Torres M., Moldon J., Varenius E., Beswick R., Alberdi A., CenkoS. B., Horesh A., 2018b, GRB Coordinates Network, 22350, 1Perley D. A., Taggart K., 2017, GRB Coordinates Network, 22194, 1Perley D. A., Schulze S., de Ugarte Postigo A., 2017, GRB CoordinatesNetwork, 22252, 1Peters C., et al., 2019, ApJ, 872, 28Piran T., 2004, Reviews of Modern Physics, 76, 1143Planck Collaboration et al., 2016, A&A, 594, A13Rhodes L., et al., 2020, MNRAS, 496, 3326Roming P. W. A., et al., 2017, ApJS, 228, 13Ryan G., van Eerten H., MacFadyen A., Zhang B.-B., 2015, ApJ, 799, 3Santana R., Barniol Duran R., Kumar P., 2014, ApJ, 785, 29Sari R., Piran T., 1999, ApJ, 520, 641Sari R., Piran T., Narayan R., 1998, ApJ, 497, L17Sault R. J., Wieringa M. H., 1994, A&AS, 108, 585Sault R. J., Teuben P. J., Wright M. C. H., 1995, in Shaw R. A., Payne H. E.,Hayes J. J. E., eds, Astronomical Society of the Pacific ConferenceSeries Vol. 77, Astronomical Data Analysis Software and Systems IV.p. 433Sironi L., Spitkovsky A., 2011, ApJ, 726, 75Smith I. A., Tanvir N. R., 2017, GRB Coordinates Network, 22242, 1Stanek K. Z., et al., 2003, ApJ, 591, L17Suzuki A., Maeda K., Shigeyama T., 2019, ApJ, 870, 38Taylor G. B., Frail D. A., Berger E., Kulkarni S. R., 2004, ApJ, 609, L1Toma K., Sakamoto T., Mészáros P., 2011, ApJ, 731, 127Trushkin S. A., Erkenov A. K., Tsybulev P. G., Nizhelskij N. A., 2017, GRBCoordinates Network, 22258, 1Urata Y., et al., 2019, ApJ, 884, L58Virtanen P., et al., 2020, Nature Methods, 17, 261Wang L., Höflich P., Wheeler J. C., 1997, ApJ, 483, L29Wang X.-G., et al., 2015, ApJS, 219, 9Wang J., et al., 2018, ApJ, 867, 147Waxman E., Kulkarni S. R., Frail D. A., 1998, ApJ, 497, 288Whiting M., Humphreys B., 2012, Publ. Astron. Soc. Australia, 29, 371Wilms J., Allen A., McCray R., 2000, ApJ, 542, 914Woosley S. E., 1993, ApJ, 405, 273Zheng H., et al., 2017, MNRAS, 464, 3486de Ugarte Postigo A., et al., 2017a, GRB Coordinates Network, 22187, 1de Ugarte Postigo A., Izzo L., Kann D. A., Thoene C. C., Pesev P., ScarpaR., Perez D., 2017b, GRB Coordinates Network, 22204, 1van Velzen S., Falcke H., Schellart P., Nierstenhöfer N., Kampert K.-H.,2012, A&A, 544, A18van der Horst A. J., et al., 2008, A&A, 480, 35von Kienlin A., et al., 2020, ApJ, 893, 46 MNRAS000
Hallinan G., et al., 2017, Science, 358, 1579Hancock P. J., Gaensler B. M., Murphy T., 2013, ApJ, 776, 106Harris C. R., et al., 2020, Nature, 585, 357Hjorth J., et al., 2003, Nature, 423, 847Högbom J. A., 1974, A&AS, 15, 417Horesh A., Cenko S. B., Perley D. A., Kulkarni S. R., Hallinan G., BellmE., 2015, ApJ, 812, 86Hunter J. D., 2007, Computing In Science & Engineering, 9, 90Izzo L., et al., 2017, GRB Coordinates Network, 22180, 1Izzo L., et al., 2019, Nature, 565, 324Izzo L., Auchettl K., Hjorth J., De Colle F., Gall C., Angus C. R., RaimundoS. I., Ramirez-Ruiz E., 2020, A&A, 639, L11Kulkarni S. R., et al., 1998, Nature, 395, 663Kulkarni S. R., et al., 1999, ApJ, 522, L97Lacy M., et al., 2020, PASP, 132, 035001Laskar T., Coppejans D. L., Margutti R., Alexand er K. D., 2017, GRBCoordinates Network, 22216, 1Laskar T., Hull C. L. H., Cortes P., 2020, ApJ, 895, 64Lemoine M., Li Z., Wang X.-Y., 2013, MNRAS, 435, 3009Levan A. J., et al., 2014, ApJ, 781, 13Levinson A., Ofek E. O., Waxman E., Gal-Yam A., 2002, ApJ, 576, 923Li Z.-Y., Chevalier R. A., 1999, ApJ, 526, 716Lien A., et al., 2016, ApJ, 829, 7Lloyd-Ronning N. M., Fryer C. L., 2017, MNRAS, 467, 3413Lloyd-Ronning N. M., Gompertz B., Pe’er A., Dainotti M., Fruchter A.,2019, ApJ, 871, 118Ma C., et al., 2009, IERS Technical Note, 35, 1MacFadyen A. I., Woosley S. E., 1999, ApJ, 524, 262Maity B., Chandra P., 2020, arXiv e-prints, p. arXiv:2012.05166Mauch T., Murphy T., Buttery H. J., Curran J., Hunstead R. W., PiestrzynskiB., Robertson J. G., Sadler E. M., 2003, MNRAS, 342, 1117Mazzali P. A., McFadyen A. I., Woosley S. E., Pian E., Tanaka M., 2014,MNRAS, 443, 67McConnell D., et al., 2020, Publ. Astron. Soc. Australia, 37, e048McKinney W., 2010, in Stéfan van der Walt Jarrod Millman eds,Proceedings of the 9th Python in Science Conference. pp 56–61,doi:10.25080/Majora-92bf1922-00aMereghetti S., Götz D., Borkowski J., Walter R., Pedersen H., 2003, A&A,411, L291Metzger B. D., Williams P. K. G., Berger E., 2015, ApJ, 806, 224 Modjaz M., Liu Y. Q., Bianco F. B., Graur O., 2016, ApJ, 832, 108Murphy T., et al., 2013, Publ. Astron. Soc. Australia, 30, e006Nakar E., Piran T., 2011, Nature, 478, 82Norris R. P., et al., 2011, Publ. Astron. Soc. Australia, 28, 215Onken C. A., et al., 2019, Publ. Astron. Soc. Australia, 36, e033Panaitescu A., Kumar P., 2000, ApJ, 543, 66Perez-Torres M., Cenko S. B., Horesh A., Alberdi A., 2018a, GRB Coordi-nates Network, 22302, 1Perez-Torres M., Moldon J., Varenius E., Beswick R., Alberdi A., CenkoS. B., Horesh A., 2018b, GRB Coordinates Network, 22350, 1Perley D. A., Taggart K., 2017, GRB Coordinates Network, 22194, 1Perley D. A., Schulze S., de Ugarte Postigo A., 2017, GRB CoordinatesNetwork, 22252, 1Peters C., et al., 2019, ApJ, 872, 28Piran T., 2004, Reviews of Modern Physics, 76, 1143Planck Collaboration et al., 2016, A&A, 594, A13Rhodes L., et al., 2020, MNRAS, 496, 3326Roming P. W. A., et al., 2017, ApJS, 228, 13Ryan G., van Eerten H., MacFadyen A., Zhang B.-B., 2015, ApJ, 799, 3Santana R., Barniol Duran R., Kumar P., 2014, ApJ, 785, 29Sari R., Piran T., 1999, ApJ, 520, 641Sari R., Piran T., Narayan R., 1998, ApJ, 497, L17Sault R. J., Wieringa M. H., 1994, A&AS, 108, 585Sault R. J., Teuben P. J., Wright M. C. H., 1995, in Shaw R. A., Payne H. E.,Hayes J. J. E., eds, Astronomical Society of the Pacific ConferenceSeries Vol. 77, Astronomical Data Analysis Software and Systems IV.p. 433Sironi L., Spitkovsky A., 2011, ApJ, 726, 75Smith I. A., Tanvir N. R., 2017, GRB Coordinates Network, 22242, 1Stanek K. Z., et al., 2003, ApJ, 591, L17Suzuki A., Maeda K., Shigeyama T., 2019, ApJ, 870, 38Taylor G. B., Frail D. A., Berger E., Kulkarni S. R., 2004, ApJ, 609, L1Toma K., Sakamoto T., Mészáros P., 2011, ApJ, 731, 127Trushkin S. A., Erkenov A. K., Tsybulev P. G., Nizhelskij N. A., 2017, GRBCoordinates Network, 22258, 1Urata Y., et al., 2019, ApJ, 884, L58Virtanen P., et al., 2020, Nature Methods, 17, 261Wang L., Höflich P., Wheeler J. C., 1997, ApJ, 483, L29Wang X.-G., et al., 2015, ApJS, 219, 9Wang J., et al., 2018, ApJ, 867, 147Waxman E., Kulkarni S. R., Frail D. A., 1998, ApJ, 497, 288Whiting M., Humphreys B., 2012, Publ. Astron. Soc. Australia, 29, 371Wilms J., Allen A., McCray R., 2000, ApJ, 542, 914Woosley S. E., 1993, ApJ, 405, 273Zheng H., et al., 2017, MNRAS, 464, 3486de Ugarte Postigo A., et al., 2017a, GRB Coordinates Network, 22187, 1de Ugarte Postigo A., Izzo L., Kann D. A., Thoene C. C., Pesev P., ScarpaR., Perez D., 2017b, GRB Coordinates Network, 22204, 1van Velzen S., Falcke H., Schellart P., Nierstenhöfer N., Kampert K.-H.,2012, A&A, 544, A18van der Horst A. J., et al., 2008, A&A, 480, 35von Kienlin A., et al., 2020, ApJ, 893, 46 MNRAS000 , 1–17 (2021) search for GRB afterglows with ASKAP APPENDIX A: RULED-OUT AFTERGLOW CANDIDATES
In §6, we ruled out three of the four afterglow candidates in thevetting process. For each candidate, we provide an explanation andshow in Figure A1 their corresponding radio and optical images.
Candidate 1 (GRB 080905B) – This matched radio sourcefound in RACS is likely to be related to the host galaxy, 2MASXJ20065732 − . (cid:48)(cid:48) . (cid:48)(cid:48) Candidate 2 (GRB 110312A) – This matched radio sourceis the only candidate located within the VAST-P1 footprint and itwas detected in all epochs of our search. The emission is either hostgalaxy related or from a coincident background source since theradio source is not spatially consistent with the GRB position (14 . (cid:48)(cid:48) Candidate 3 (GRB 160216A) – This matched radio sourcefound in RACS is not spatially consistent with the GRB position(5 . (cid:48)(cid:48) Candidate 4 (GRB 171205A) – This matched radio sourcefound in RACS is a confirmed afterglow candidate. The details arediscussed in §6.
This paper has been typeset from a TEX/L A TEX file prepared by the author.MNRAS , 1–17 (2021) J. K. Leung et al. h m s s m s s − ◦ D ec li n a t i o n ( J ) ASKAP: Candidate 1 (GRB 080905B)0 . . . . F l u x D e n s i t y ( m J y b e a m − ) h m s s m s s − ◦ SkyMapper: Candidate 1 (GRB 080905B)10 h m s s s − ◦ D ec li n a t i o n ( J ) ASKAP: Candidate 2 (GRB 110312A)0123456 F l u x D e n s i t y ( m J y b e a m − ) h m s s s − ◦ Pan-STARRS: Candidate 2 (GRB 110312A)20 h m s s s − ◦ D ec li n a t i o n ( J ) ASKAP: Candidate 3 (GRB 160216A) − . − . . . . . . . . F l u x D e n s i t y ( m J y b e a m − ) h m s s s s − ◦ SkyMapper: Candidate 3 (GRB 160216A)11 h m s s s − ◦ Right Ascension (J2000) D ec li n a t i o n ( J ) ASKAP: Candidate 4 (GRB 171205A) − . . . . . . F l u x D e n s i t y ( m J y b e a m − ) h m s s s − ◦ Right Ascension (J2000)Pan-STARRS: Candidate 4 (GRB 171205A)
Figure A1.
The ASKAP image (left) and optical image (right) for each afterglow candidate found in our search. All ASKAP images are from RACS, withthe exception of the Candidate 2 (GRB 110312A) ASKAP image taken from epoch 8 of VAST-P1. The optical images shown are 𝑔 -band images from eitherSkyMapper (Onken et al. 2019, Candidates 1 and 3) or Pan-STARRS (Candidates 2 and 4) with radio contours from their corresponding ASKAP imagesoverlaid (orange). The lowest contours start at the 3 𝜎 -level and increase by a factor of √ . (cid:48) × . (cid:48) , centred on their corresponding Swift -XRT/UVOT localised afterglow positions, with typicallocalisation errors of a few arcseconds. North is up and East is to the left. MNRAS000