Accepted for publication in the Astrophysical Journal PreprinttypesetusingLATEXstyleemulateapjv.12/16/11 SN 1986J VLBI. III. THE CENTRAL COMPONENT BECOMES DOMINANT Michael F. Bietenholz1,2 and Norbert Bartel2 Accepted for publication in the Astrophysical Journal ABSTRACT We present a new 5-GHz global-VLBI image of supernova 1986J, observed in 2014 at t = 31.6 yr after the explosion, and compare it to previous images to show the evolution of the supernova. Our new image has a dynamic range of ∼100 and a background rms noise level of 5.9 µJy beam−1. There 7 is no significant linear polarization, with the image peak being < 3% polarized. The latest image 1 is dominated by the compact central component, whose flux density is now comparable to that of 0 the extended supernova shell. This central component is marginally resolved with a FWHM width 2 of 900+100 µas, corresponding to a radius of r = 6.7+0.7 ×1016 cm for a distance of 10 Mpc. −500 comp −3.7 r UsingVLBIobservationsbetween2002and2014,wemeasuredthepropermotionsofboththecentral a componentandahot-spottotheNEintheshellrelativetothequasar3C66A. Thecentralcomponent M isstationarytowithintheuncertaintyof12µas yr−1,correspondingto570km s−1. Ourobservations argueinfavorofthecentralcomponentbeinglocatednearthephysicalcenterofSN1986J. Theshell 0 hot-spothadameanvelocityof2810±750km s−1 totheNE,whichisconsistentwithittakingpartin 2 the homologous expansion of the shell seen earlier. The shell emission is evolving in a non-selfsimilar fashion, with the brightest emission shifting inwards within the structure, and with only relatively ] E faint emission being seen near the outer edge and presumed forward shock. An animation is available H in the electronic edition. Keywords: supernovae: individual (SN 1986J) — radio continuum: supernovae . h p - 1. INTRODUCTION Type IIn supernova (Rupen et al. 1987). Due to its rel- o atively high radio flux density, it was one of the first r SN 1986J was one of the most radio luminous su- t pernovae ever observed. Its unusually long-lasting and SNe to be observed with VLBI (Bartel et al. 1987). A s VLBI image, the first of any optically identified super- a strong radio emission and its relative nearness makes it nova, was obtained by Bartel et al. (1991). VLBI obser- [ one of the few supernovae for which it is possible to pro- vations at subsequent epochs up to 2008 led to a series duce detailed images with very-long-baseline interferom- 2 of images. The source morphology was complex, show- etry (VLBI). It is also one of the few supernovae still v inganexpanding,albeitsomewhatdistortedshellwitha detectable more than t = 30 years after the explosion, 7 prominenthot-spotintheshellvisibleoncetheresolution thus we have been able to follow its evolution for longer 4 was sufficient. thanforanyotherSNforwhichthereareresolvedVLBI 4 Other than the relatively slow turn-on, and its high images except for SN 1979C (Bartel & Bietenholz 2008; 8 radio luminosity, the evolution of SN 1986J’s spectral 0 Marcaide et al. 2009). It is one of the few SNe which we energy distribution (SED) was unremarkable till 1998. . have been able to follow observationally as it evolves to- 1 In Paper I, we showed that at that time, an inversion wards a supernova remnant, and it thus helps to build a 0 appearedinthespectrum,withthebrightnessincreasing connection between supernovae and their remnants (see 7 with increasing frequency above ∼10 GHz, up to a high- Milisavljevic & Fesen 2017). We continue here our series 1 frequency turnover at ∼20 GHz. of papers on VLBI observations of SN 1986J: Bietenholz : In Bietenholz et al. (2004), we showed by means of v et al. (2002, 2010b), which we will refer to as Papers I phase-referencedmulti-frequencyVLBIimagingthatthis i and II respectively, as well as Bietenholz et al. (2004). X spectral inversion was associated with a bright, compact SN 1986J was first discovered in the radio, some time r after the explosion (van Gorkom et al. 1986; Rupen componentintheprojectedcenteroftheexpandingshell. a Such a central component has so far not been seen in et al. 1987). The best estimate of the explosion epoch is anyothersupernova3 (seee.g.,Bietenholz2014;Bartel& 1983.2±1.1 (Paper I, see also Rupen et al. 1987; Cheva- Bietenholz 2014). At that time, in late 2002, the central lier 1987; Weiler et al. 1990), which we take as t=0. It componentwasclearlypresentinthe15GHzimage,but occurred in the nearby galaxy NGC 891, for whose dis- not discernible in the 5 GHz one. tance the NASA/IPAC Extragalactic Database (NED) In Paper II we showed that from 2005 on, the central lists 19 measurements with a mean of 10.0±1.4 Mpc, component had become bright also at 5 GHz. We also which value we adopt throughout this paper. showed that, though the morphology was complex, the Opticalspectra,takensoonafterthediscovery,showed radio emission region was still expanding but also decel- a somewhat unusual spectrum with narrow linewidths, erating, with the average outer radius ∝t0.69±0.03. but the prominent Hα lines led to a classification as a 1Hartebeesthoek Radio Observatory, PO Box 443, Krugers- 3Wenotethatcentralemissionatmmwavelengthshasbeenseen dorp,1740,SouthAfrica inSN1987A,whichisattributedtodust. Nocentralsynchrotron 2Department of Physics and Astronomy, York University, emission has been seen in SN 1987A at cm wavelengths (Zanardo Toronto,M3J1P3,Ontario,Canada etal.2014). 2 Bietenholz & Bartel 2. OBSERVATIONSANDDATAREDUCTION “native” or highest resolution possible in the data, MS- CLEAN simultaneously deconvolves images at that na- 2.1. VLBI Observations tiveresolutionandseverallargerones,andthencombines We obtained VLBI observations of SN 1986J on 2014 the different resolutions into a final image. MS-CLEAN Oct.22,usingaglobalVLBIarraywhichconsistedofthe has generally been shown to produce superior results for following21antennas. FromtheNationalRadioAstron- the deconvolution of extended sources (see, e.g., Hunter omyObservatory(NRAO)4:theVLBA(9×25mdiame- et al. 2012; Rich et al. 2008; Bietenholz et al. 2010a). ter,MaunaKeadidnotobserve),theJanskyVeryLarge Greisen et al. (2009) and Cornwell (2008) discuss Array(inphasedmode,equivalentdiameter130m);from the properties of MS-CLEAN as implemented in AIPS. the European VLBI Network: Badary (32 m), Effels- Aside from the selection of resolutions, the main param- berg (100 m diameter), Hartebeesthoek (26 m), Jodrell eters for tuning the MS-CLEAN algorithm are the or- BankLovell(70m),Noto(32m),Onsala(25m),Svetloe der in which the different resolutions are treated, and (32m), Torun(32m), Westerbork(phasedmode, equiv- the depth to which they are CLEANed (Greisen et al. alent diameter 94m), Yebes (40 m) and Zelenchukskaya 2009). We made two different simulated visibility data (32 m). sets for testing, both designed to be relatively similar to The observations of SN 1986J were interleaved with the brightness distribution of SN 1986J. The first one ones of the quasar 3C 66A, only 40(cid:48) away on the sky, was a geometrical model consisting of an almost circu- which we used as a phase-reference source. SN 1986J’s lar disk of diameter ∼8 mas and a slightly off-center declination of +42◦ enabled us to obtain dense and only marginally-resolved Gaussian source with full-width at moderately elliptical u-v coverage. As usual, a hydro- half-maximum (FWHM) 0.4 mas. The second model gen maser was used as a time and frequency standard consisted of the CLEAN components derived from a de- at each telescope. We recorded both senses of circular convolution of the real data. We replaced the observed polarization with the RDBE/Mark5C wide-band system visibility values with the relevant Fourier transforms of atasample-rateof1Gbps, andcorrelatedthedatawith thesemodelsandthenaddednoiseatalevelcomparable NRAO’sVLBADiFXcorrelator(Delleretal.2011). We to that in the observed visibilities. We deconvolved the used a bandwidth of 128 MHz centered on 4.996 GHz. resulting simulated visibilities, and compared the result- The data reduction was carried out with NRAO’s As- ingCLEANimagestothemodelimagesafterconvolving tronomicalImageProcessingSystem(AIPS).Theinitial the latter with the CLEAN beam. flux density calibration was done through measurements We found that the most accurate reconstruction was of the system temperature at each telescope, and im- achieved for both models when the CLEAN iterations proved through self-calibration of the 3C 66A data. were stopped when the flux density of the CLEAN com- We used a cycle time of ∼3.2 minutes to phase- ponentshadreachedthermsimagebackgroundlevel. We reference to 3C 66A, with about 1.8 minutes spent on also found that a multi-scale CLEAN with three resolu- SN 1986J. Our positions in this paper are given relative tions,withFWHMbeamareasof2.2,10.5and36.1mas2 toanassumedpositionofthebrightnesspeakof3C66A gave good results. Using this imaging strategy, the rms of RA = 02h 22m 39s.611500, decl. = 43◦ 02(cid:48) 07(cid:48).(cid:48)79884 error within the CLEAN window, i.e., the difference be- taken from the International Celestial Reference Frame, tween the CLEANed image and the original model, was ICRF15. only 3% and 16% larger than the rms image background We determined the instrumental polarization leakage for the two models respectively, with the worst-case er- from our observations of 3C 66A, which is almost un- rors being (cid:46)5× the image background rms. polarized, using a linear approximation (AIPS task LP- CAL). We did not calibrate absolute position angle, so 3. VLBIIMAGES our polarization measurements resulted in correct mag- We show the 5-GHz VLBI image of SN 1986J in Fig- nitudes but unknown position angles. ure 1. The image was made in AIPS, using complex weighting with the robustness parameter chosen so as to 2.2. VLBI Imaging Considerations attainclosetonaturalweighting(AIPSROBUST=2.2). Furthermore,weusedmulti-scaleCLEANdeconvolution Imaging and deconvolving interferometer data can be with the strategy discussed in § 2.2 above. The total difficult,particularlyinthecaseofanon-uniformandrel- CLEANed flux density was 1622 µJy, the peak bright- atively sparse array like that used for VLBI. We use the ness 617 µJy beam−1, and the background rms bright- multi-scale extension of the original CLEAN algorithm, nesswas5.9µJy beam−1,foranimagedynamicrangeof MS-CLEAN (Wakker & Schwarz 1988), for the decon- ∼100. We estimate that the on-source brightness uncer- volution. MS-CLEAN works similarly to the traditional tainty is ∼7 µJy beam−1, only slightly larger than the CLEAN algorithm (see, e.g., Cornwell et al. 1999), but off-source rms. instead of just using a single basis function based on the The image is dominated by the central component. In central lobe of the dirty beam, that is one based on the contrast, the highest brightness in the more extended emission is only ∼ 100 µJy beam−1, or (cid:46)15% of the 4TheNationalRadioAstronomyObservatory,NRAO,isafacil- peak brightness of the central component (we estimate ityoftheNationalScienceFoundationoperatedundercooperative agreementbyAssociatedUniversities,Inc. the brightness ratio between the central component and 5 We note that a slightly revised position for 3C 66A is given the remainder below in § 3.2.1). in the ICRF2 (Fey et al. 2009), which is different from the one To highlight the evolution of SN 1986J, we show in we used by −43,−24 µas in RA and decl. respectively (Ma et al. Figure2ourearlierimagesat8.4GHZ,reproducedfrom 1998). However, for consistency, we use in this paper the same ICRF1referencepositionfor3C66AasweusedinPaperII,which PaperI,andat5.0GHz,reproducedfromPaperII.Inthe hasnoeffectonthedifferentialpositionsoranyofourconclusions. first image, at t=5.5 yr and 8.4 GHz, the source is still 3 5 4 3 200 2 s 4 nd 1 321 150 o c e s 0 c r A 100 i -1 l l i M -2 50 -3 -4 0 -5 4 2 0 -2 -4 MilliArc seconds Figure 1. The5-GHzVLBIimageofSN1986Jmadefromobservationson2014Oct.23,atage31.6yr. Thecontoursaredrawnat−3,3, 5,10,15,20,30,50,70and90%ofthepeakbrightness,withthe50%contourbeingemphasized. Thepeakbrightnesswas617µJybeam−1, thetotalCLEANedfluxdensitywas1622µJy,andthebackgroundrmsbrightnesswas5.9µJybeam−1. Thecolororgrayscaleislabeled in µJybeam−1, and is saturated so as to better show the low-level emission. The (green) overlapping circles indicate the position of the centralcomponentatages20.3,22.6,25.6and31.6yr,whilethe(green)crosseslabeled“1”to“4”showthepositionoftheshellhot-spot at ages 15.9, 19.6, 22.6, and 25.6 yr, respectively, along with their estimated uncertainties of 120 µas in each coordinate. In the present image(t=31.6yr),theshellhot-spotisnolongerclearlyidentifiable. Northisupandeasttotheleft,andtheFWHMoftheconvolving beamof2.21mas×0.89masatp.a.−15◦ isindicatedatupperleft. rather compact but with the protrusions already visible. 4.4µJy beam−1,whilethatinS was4.6µJy beam−1. pol By t = 7.4 yr the protrusions had expanded indicating No significant polarization was detected, with the maxi- a complex brightness distribution. At t = 15.9 yr and mumvalueofS observedbeing14µJy beam−1 or3σ. pol at 5 GHz, the shell hot-spot is visible to the NE of the Inparticularthepolarizedintensityatthelocationofthe center. Byt=22.6yr,thecentralcomponentisstarting centralcomponentwas7.0±5.6µJy beam−1,andweput to appear, and by t = 25.6 yr, the central component a 3σ upper limit on the linear polarization of the central has become brighter than the shell hot-spot, which is component of 3.3%. This limit is comparable to upper fading. By the present epoch, at t=31.6 yr, the central limitswereportedforSN1993J,andconsistentwiththe componentdominatesandtheshellhot-spotisnolonger expectationthatinternalFaradaydepolarizationisquite clearly visible. strong (Bietenholz et al. 2003). 3.1. Polarization 3.2. The Central Component We also made images in Stokes Q, U, and V. As men- We determined the characteristics of the central com- tioned in section 2.1 above, we did not calibrate the ab- ponentbyfittinganellipticalGaussianmodelaswellasa solute polarization p.a., but we did calibrate the leak- zeroleveltothecenteroftheCLEANed(StokesI)image age terms. Therefore, our measurements of the linearly by least squares. Although the properties of the fitted polarized intensity are accurate, but that of the p.a. Gaussianaresomewhatdependentontheexactchoiceof are completely uncertain. The lack of p.a. calibration the fitting window because the central component can- turns out to be immaterial, since no significant polar- notbeseparateduniquelyfromthecomplexbackground ized emission is seen in either linear or circular polariza- emission due to the shell, the results can nonetheless be tion. We combined the Q and U images into an image used to characterize the central component. of the linearly polarized intensity, S , corrected for the pol Ricean noise bias6. The rms in Stokes Q and U was correctionhasbeenimplementedintheAIPStaskCOMB,which 6 S = (cid:112)Q2+U2 where Q and U are the measured Stokes we used to calculate our linearly polarized flux densities. This pnaorisaemipesotltehrse.reBfoyredbefiiansietdio.nT,hSipsobliiassphoassitbiveee,nacnadlciunlattheedparnedsenacbeiaosf pofrooccecdausiroenparlloydupcreosduthciencgourrnepchtymsiecaanlnveagluaetiovfeSvpaolul,esa.ttheexpense 4 Bietenholz & Bartel 6 66 a) 5.5 yr (1988 Sep 29) 8.4 GHz e) 22.6 yr (2005 Oct 25) 5.0 GHz 440000 7 4 44 6 330000 ds 2 5 dsds 22 on m onon MilliArc sec-02 34 Jy/beam MilliArc secMilliArc sec--0022 121200000000 2 -4 1 --44 00 0 -666 4 2 0 -2 -4 -6 --66666 44 22 00 --22 --44 --66 b) 7.4 yr (199M0il lJiAurlc 2s1ec) onds 8.4 GHz 4 f) 25.6 yr (200MM8iill llOiiAAcrrcct 2ssee6cc) oonnddss 5.0 GHz 350 4 4 300 3 m 250 Arc seconds 20 2 Jy/beam Arc seconds 20 125000 Milli-2 Milli-2 100 1 50 -4 -4 0 0 -6 -6 66 4 2 0 -2 -4 -6 66 4 2 0 -2 -4 -6 c) 15.9 yr (19M99ill iFAercb s 2ec2o)n ds 5.0 GHz g) 31.6 yr (20M14ill iOArcct s2e3c)o nds 5.0 GHz 600 1500 4 4 500 ds 2 ds 2 400 n 1000 n o o c c e e c s 0 c s 0 300 Ar Ar Milli-2 500 Milli-2 200 -4 -4 100 0 0 -6 -6 66 4 2 0 -2 -4 -6 6 4 2 0 -2 -4 -6 MilliArc seconds MilliArc seconds d) 19.6 yr (2002 Nov 11) 5.0 GHz 500 4 400 s 2 d on 300 c e c s 0 MilliAr 200 -2 100 -4 0 -6 6 4 2 0 -2 -4 -6 MilliArc seconds Figure 2. A sequence of all our VLBI images of SN 1986J at 5 or 8.4 GHz, showing its evolution from t = 5.5 yr to 31.6 yr after the explosion, or 1988 to 2014. Each panel has the lowest contour at 3× the image rms, and the remainder at 10, 20, 30, 40,50,70,and90%,ofthepeakbrightness,withthe50%onebeingemphasized(andbeingthefirstwhitecontourintheblack andwhiteimages). Ineachpanel,theageanddatearegivenattopleft,thefrequencyatthetopright,andtheFWHMsizeof theconvolvingbeamisindicatedatlowerleft. Northisupandeasttotheleft. Thegray-orcolorscaleisinµJy beam−1 unless indicated otherwise. For details of panels a) and b), see Paper I, c) through f), see Paper II. Panel g) is the same as Figure 1, but not saturated. This figure is also available as an animation in the electronic edition. 5 3.2.1. Flux Density 103 The best fitting point source model, i.e., an elliptical GaussianofwidthfixedtothatoftheCLEANbeam,had a flux density of 511±5µJy. We note that the value of µJy) 511µJy is, strictly speaking, closer to being an estimate nsity ( ofthepeakbrightnessinµJy beam−1thananestimateof de the total flux density, since if the component is resolved Flux 102 the latter would be higher. We do in fact find that the centralcomponentisslightlyresolved(see§3.2.3below). However,itisdifficulttoreliablyestimateboththewidth Central component and the flux density simultaneously. In order to be able Shell hot-spot to compare the flux density from this epoch with those 15 20 25 30 from previous ones consistently, we therefore take the Age (yr) valuesobtainedbyfittinganunresolvedsourceandazero level to the image as the estimates of the flux density, Figure 3. The 5-GHz flux densities of the central component (blue circles) and the shell hot-spot (red squares) as determined keeping in mind that we are underestimating the total from the VLBI images (see text), as a function of age. For the flux density to the extent that the central component is epochs at which the component or spot could not be clearly dis- resolved. cerned, we plot 50% of the surface brightness at the relevant lo- cation as the upper limit, on the grounds that if the feature had Thisvalueofthefluxdensityofthecentralcomponent more than half the brightness of the background emission at that at ν = 5 GHz corresponds to a spectral luminosity, L , ν location,itwouldbeclearlydiscernibleintheimage. of 6×1025 erg s−1 Hz−1, or ∼20 times that of the Crab Nebula. If we approximate the luminosity as νL , the ν 3.2.2. Position and Proper Motion central component’s luminosity is ∼80L . (cid:12) If we fit the images from earlier epochs in the same We take the quadratically interpolated peak bright- way, we obtain the values tabulated in Table 1. The ness position from the phase-referenced image as the uncertainty in these flux-density estimates is difficult to best estimate of the central component’s position at determineaccuratelyasitdependsontheseparationbe- t = 31.6 yr, which was RA = 02h 22m 31s.321434 and tween the central component and the shell-emission in decl.=42◦ 19(cid:48) 57(cid:48).(cid:48)25941withsmallstatisticaluncertain- the fit and how resolved the central component is. How- ties of < 10 µas. A fit of an elliptical Gaussian to the ever,thegeneralpictureseemsclear: fromt=22.6yrto image results in a position consistent with the one given t = 31.6 yr (2005 to 2014) the central component’s flux above to within 20 µas. densityincreased by∼300µJy(from177to511µJy),or For a realistic estimate of the uncertainty on the posi- byafactorofalmost3. AsafractionofSN1986J’stotal tion which includes systematic effects, we take the same fluxdensity,ithasincreasedfromabout7%att=22.6yr conservativeestimateof120µasaswedidinPaperII.We to32%att=31.6yr. Wecanthereforesaywithreason- confirmed this estimate with a Monte-Carlo simulation able confidence that, in the last 6 yr, the central com- based on the present data. The largest part of this un- ponent’s fraction of SN 1986J’s total 5-GHz flux density certaintyisduetothedifficultyinseparatingthecentral has roughly doubled, and that the 5-GHz flux density of componentfromtheshellemissionaswellasduetonoise, the central component has increased. We plot the flux with a smaller part being due to the possible instability density of the central component in Figure 3. inthereferencesourceanderrorsinthephase-referencing The peak brightness of the central component, as esti- suchasthetroposphereorerrorsinthestationpositions mated by the point-source fit, is 511±5 µJy beam−1. (for estimates of the phase-referencing errors, see Pradel The average brightness of the shell emission is etal.2006). Wetabulatethepositionofthecentralcom- ∼50µJy beam−1,comparableto,butslightlylowerthan, ponent, expressed as offsets from our nominal explosion the shell-brightness of 70±8 µJy beam−1 estimated by center position, in Table 1. our fit at the central component location. The central Forournominalexplosioncenterposition,wetookthe component is therefore ∼ 10× brighter than the shell is estimate of the explosion center location that we found on average. inPaperII,whichwasRA=02h 22m 31s.321457,decl.= Theshellemissionisopticallythin,sinceithasaspec- 42◦ 19(cid:48) 57(cid:48).(cid:48)25951, with an uncertainty of 200 µas in each tral index of ∼ −0.6 (where S ∝ να). The central coordinate. This estimate is the position of the center ν component, on the other hand, had a positive spectral of the shell, averaged over 5 epochs from 1999 to 2008, index at 5 GHz at t∼25 yr and was therefore still opti- all determined with respect to 3C 66A. If the supernova cally thick, with the transition to an optically thin spec- is expanding symmetrically, this shell center position is trum occurring near 13 GHz (Paper II). Extrapolating identicaltotheexplosioncenter. Onlyintwosupernovae the SED from Paper II to t = 31.6 yr, it seems unlikely do we have observational constraints on the symmetry that the SED would have evolved to the point it was op- of the ejecta in the sky plane. In SN 1993J, in projec- ticallythinat5GHz,sothecentralcomponentisalmost tion, the ejecta are circularly symmetric to within 5.5% certainly still partly absorbed at 5 GHz. The intrinsic about the explosion position (Bietenholz et al. 2001). In brightness ratio between the central component and the SN 1987A, there is a complicated structure, with mostly shell is therefore likely even higher than 10. bilateral symmetry, but with one-sided asymmetry of at 6 Bietenholz & Bartel Table 1 Fluxdensitiesandpositionsoffsetsofthecentralcomponentandtheshellhot-spot MidpointDate Agea Frequency Centralcomponent Shellhot-spot Fluxdensityb Positionoffsetc Fluxdensityb Positionoffsetc RA dec. RA dec. (yr) (GHz) (µJy) (mas) (mas) (µJy) (mas) (mas) 1999Feb22 15.94 5.0 ··· ··· ··· 1300 0.996 0.821 2002Nov11 19.66 5.0 ··· ··· ··· 380 1.172 0.809 2003Jun22 20.27 15.4 ··· −0.261 −0.207 ··· ··· ··· 2003Jun23 20.28 8.4 ··· −0.289 −0.097 ··· ··· ··· 2005Oct25 22.62 5.0 177 −0.195 −0.221 290 1.376 0.895 2008Oct26 25.62 5.0 282d −0.283 −0.025 120 1.460 1.143 2014Oct23 31.61 5.0 511 −0.263 −0.095 ··· ··· ··· a TheageofSN1986J,takenwithrespecttoanexplosionepochof1983.2(seePaperI). bThefluxdensityofanunresolvedsourcefittedtotheimage. Avariablezero-levelwassimultaneouslyfitted. Thisfitwillunderestimatethetotalfluxdensityofaresolvedsource,butweusethisconsistentestimateto comparethevariousepochs,sincewecannotreliablydetermineboththeextentandthefluxdensity. cThepositionoffsetfromtheournominalexplosioncenterposition. Weusetheinterpolatedpeak-brightness locationfromtheun-selfcalibrated5-GHzimagesasourestimateofthepositions,andtheestimateduncer- taintiesare120µasineachcoordinate. Ournominalexplosioncenterpositionistheaveragepositionofthe center of the shell from 1999 to 2008, RA = 02h22m31s.321457, decl. = 42◦19(cid:48)57(cid:48).(cid:48)25951, and we estimate thatthispositioniswithin∼200µasofthetrueexplosioncenter(seePaperII). d NotethatinPaperIIwegiveanapproximateandsomewhatlargerfluxdensityof390µJyforthecentral component which was derived just from the peak brightness in the image. Fitting the zero-level, as we do here,givesasomewhatlower,butmoreaccurateestimateofthefluxdensityofthecomponent. least 10% at some azimuth angles (Zanardo et al. 2013). For SN 1986J, the presence of two distinct blue-shifted components in the optical spectra also suggest an asym- metric structure (Milisavljevic et al. 2008). 2.0 central comp.: RA In 2014, therefore, SN 1986J’s central component is central comp.: decl −263,−95 µas from the explosion center in RA and dec. shell hot-spot: RA 1.5 respectively, with a combined uncertainty of 230 µas in shell hot-spot: decl each coordinate. We regard this displacement as sugges- tivWe,ebucatnnodtetceornmcliunseivteh.e proper motion of the central et (mas) 1.0 component by comparing the position in the present offs n e2p00o8ch(PwaitphercoIIn)s,iswtehnicthly-adreetetrambuinlaetdedoninesTfraobmle210.0A5 alinnd- Positio 0.5 ear regression gives a proper motion of 0±12 and 8±12 µas yr−1 in RA and decl., respectively, corresponding to 0.0 380±570km s−1atp.a.∼0◦. Weplotthepropermotion of the central component in Figure 4. 0.5 0 5 10 15 20 25 30 3.2.3. Size and Expansion Velocity Age (yr) Wealsoallowedthefittedmodeltobenonpoint-likeby Figure 4. The proper motion of the central component and the allowingthewidthsandp.a.ofthefittedellipticalGaus- shell hot-spot. The “Age” axis, t, is plotted non-linearly, so that sian to be free in the fit. We obtained a fitted FWHM, any offset ∝t0.69 would appear as a straight line in the plot. We choose this scaling because we determined in Paper II that the deconvolvedfromtheCLEANbeam, of900×550µasat shellexpandswithradius∝t0.69. Thepositionsweredetermined p.a. 125◦, a flux density of 707 µJy, and a fitted zero- fromimagesphase-referencedto3C66A,andarelistedinTable1, level of 45 µJy beam−1. The rms residuals to this fit along with the frequency of each image. The position offsets are were 16 µJy beam−1, about 2.7 times larger than the determinedrelativeto3C66A,butareexpressedrelativetoRA= 02h22m31s.321457,decl.=42◦19(cid:48)57(cid:48).(cid:48)25951),whichistheaverage off-source background rms. Unfortunately, the fitted de- position of the center of the shell (see Paper II). The plotted 1σ convolvedsizeisbiasedwhenitisnear0, sinceitcannot uncertaintiesof120µasareestimateswhichincludebothstatistical be less than 0. We performed a Monte-Carlo simula- and systematic contributions added in quadrature. The plotted tion (n = 400 trials), where we used model sources of linesindicatethefitsoftheformx=x0+b∗t0.69,withthedot- dashedlinesbeingRAandthedashedlinesbeingdecl. knownsizeandrmsimage-noiselevelsof16µJy beam−1 (equal to the rms residuals in our fit to the real image). Thissimulationshowedusthatthereisa4%chancethat (cid:46)1000µas, thereforeanapproximatelycircularshapeis we would obtain a fitted major axis of 900 µas when the compatible with our measurements. These FWHM val- sourcewasinfactunresolved,resultinginfinalvaluesfor ues can be compared to our FWHM resolution, which the fitted FWHM and p.a., along with the correspond- was1710×760µasatp.a.−15◦. Theresidualstothefit ing uncertainties, of 900+100 µas, at p.a. 125◦ ± 25◦. were,asmentioned,∼2.7×largerthantheexpectedim- −500 In the NS direction, our resolution is poorer, and we age noise. This is likely due to the shell emission in our can only say that the component’s perpendicular axis is fit being represented only by a constant offset, which is 7 almost certainly an over-simplification. It could also be 3.4. The Shell Size and Expansion Curve due to the central component being resolved and having Inourearlierpapers,wedeterminedtheouterradiusof a shape different from an elliptical Gaussian. SN1986Jateachepoch,anddeterminedthecorrespond- Our fitted value of 900+−150000 µas for the FWHM major ingexpansioncurve. Wefoundthattheouterradiuswas axis of the central component suggests that the central expanding with time as r ∝t0.69±0.03 (Paper II). Can out component is somewhat resolved in an approximately weagainestimatetheangularradiusreliablyforournew SE-NW direction. Both the fitted FWHM width and observations to track the expansion curve? p.a. are consistent with what can be seen in the higher In our earlier papers, we took as a representative resolution 22-GHz image in Paper II. value for the angular outer radius of SN 1986J the value At 10 Mpc, the major axis FWHM corresponds to (cid:112) θ ,whichisequalto area/πofthecontourwhich r = 6.7+0.7 ×1016 cm. The average expansion ve- 90%flux loccomitpy, over t−h3e.731.6 yr, assuming the central component encompasses90%ofthetotalfluxdensity. Sinceθ90%flux is somewhat dependent on the convolving beam size, we originated in the SN explosion, is then 680+−83080 km s−1. convolved our images with an approximately co-moving beam whose size increases ∝ t0.70 so as to minimize the bias in the measured radius evolution. We convolved our new, t = 31.6 yr, image with the co-moving beam (FWHM: 3.91 mas ×1.96 mas at p.a. −1◦), and found that the formal value of θ is 90%flux 3.3. The Shell Hot-Spot now 3.97 mas. This value of θ90%flux would suggest that SN 1986J has shrunk over the last six years, down from We first noted a bright hot-spot to the NE in the shell θ = 4.23 mas at t = 25.6 yr. The reason for the inPaperI,wherewecalleditC1. Inourcurrentimage,at 90%flux apparent decrease in size is that θ is no longer t=31.6yr(Fig.1),theshellhot-spotisnolongerclearly 90%flux a good estimate of the extent of the shell. As can be discernible,althoughthereisstillanenhancementofthe seen in Figure 1, the extended shell emission is now only brightness near the location where the shell hot-spot is just visible, and the image is strongly dominated by the expected. The shell hot-spot has been fading steadily central component. Since the central component is at sincewefirstnotedit. Wegiveestimatesofits5-GHzflux bestmarginallyresolved,anincreaseinitsbrightnesswill density, or limits thereon in the last epoch, determined cause θ to shrink. We attempted to compensate in the same way (§ 3.2.1) as were those of the central 90%flux for the dominance of the central component by calcu- component (and which will similarly underestimate the latingθ fromanimagewithanartificiallylimited true flux density if the shell hot-spot is in fact resolved). 90%flux brightness. Thisprocedureindeedresultsinlargervalues We plot the flux densities and limits in Figure 3. of θ , but the exact value is dependent on the sur- We tabulate the shell hot-spot’s position, again ex- 90%flux face brightness cutoff used, and, as the cutoff is lowered, pressed as offsets from our nominal explosion center po- the level of the adjusted θ contour becomes com- sition,inTable1. Thepositions,likethoseofthecentral 90%flux parabletothebackgroundrms,andthevalueofθ component (see § 3.2.2), were the quadratically interpo- 90%flux therefore unreliable. lated peak brightness positions on the phase-referenced We therefore can no longer determine the outer radius images). We plot the positions(along with those of the of SN 1986J due to the low signal-to-noise ratio of the central component) in Figure 4. Since we cannot de- shell emission. Our new VLBI image suggests stronger termine a position for the hot-spot in the present im- deceleration since 2008, but a continued powerlaw ex- age, we have only the same set of positions that we pansion with r ∝t0.69 is also compatible with our VLBI already determined a proper motion from in Paper II. image. There does however, seem to be an evolution of However, a new analysis (using a weighted fit) gives a theradialdistributionoftheemission, withtheemission consistent but slightly more accurate proper motion of near the outside edge having faded relative to that near 59±16 µas yr−1, corresponding to a projected velocity the center in the last image. of 2810±750 km s−1, at p.a. (57±15)◦. AswenotedinPaperII,adeceleratedmotionlikethat seen for the shell, with r ∝ t0.69, is also consistent with 4. DISCUSSION the measured positions and we show a fit of this form ThegeneralpictureofradioemissioninSNe(see, e.g., in Figure 4. The shell hot-spot therefore, has a radially Chevalier 1982; Chevalier & Fransson 2016) is that it outward proper motion consistent with the homologous arises from the shocks formed as the cloud of ejecta expansionoftheshell, withaprojectedspeedabouthalf interacts with the surrounding circumstellar material that of the outer edge of the radio emission, consistent (CSM). In particular, a forward shock is driven into the withitsprojectedpositionabouthalfwaybetweentheav- CSM, a reverse shock is driven back into the expand- erage outer radius and the center. Such a proper motion ing ejecta. The radio emission is thought to arise from wouldbeexpectedifitwereduetoadensecondensation shock-accelerated electrons and amplified magnetic field in the CSM. between these two shocks. Both shocks decelerate with We note that the p.a. of the shell hot-spot is approx- time as the ejecta transfer part of their kinetic energy to imately in the same direction as the elongation of the the swept-up CSM. In the case of spherical symmetry central component which is visible in our 2006 image at and power-law density distributions, with the CSM den- 22 GHz (Paper II). Furthermore there is a suggestion in sity, ρ ∝rs andthatoftheejecta, ρ ∝rn, aself- CSM eject someoftheimagesinFig.2ofapossibleenhancementon similar solution exists, and the shock radii are expected the SW side, opposite the shell hot-spot. We elaborate to evolve ∝ rm where m = (n−3)/(n−s) (Fransson on this coincidence in the Discussion section (§ 4). et al. 1996). In this picture, one expects radio emission 8 Bietenholz & Bartel from a spherical shell region. In such a region, the outer center of SN 1986J. boundoftheradioemissioncorrespondstotheprojected In this case the central component may be due to outer edge of the shell which is the forward shock. If the thesupernovashockinteractingwithahighlystructured volumeemissivityisuniform, thebrightnessdistribution CSM produced by a binary companion (Chevalier 2012; is highest at the projected inner radius of the shell, and Chevalier 2014; see also Barkov & Komissarov 2011; Pa- has a minimum in the center. In the case of SN 1993J, pish et al. 2015), where the shock travels much more indeed,atextbookshellstructureisseenintheVLBIim- slowly in the denser parts of the CSM near the binary ages (e.g., Bietenholz et al. 2003; Marcaide et al. 1995; orbit plane, thus producing a bright but compact radio Mart´ı-Vidal et al. 2011) emission region. InSN1986J,thestructureisnotasclear,withtheshell The central component could also be emission from a being somewhat distorted. If we equate the outer edge compact remnant of the supernova. If that remnant is a of the radio emission region with the forward shock7, we neutron star, the emission could be from a pulsar-wind could determine that m = 0.69±0.03 up to t = 25.6 yr nebulaorperhapsfromtheneutronstar’saccretiondisc. (Paper II). The outer edge of the radio emission moved Iftheremnantisablackhole,whichquitepossiblegiven outwards between t = 0 and 25.6 yr with an average the probably massive progenitor (Weiler et al. 1990), speed of 7800 km s−1, and a speed at t = 25.6 yr of then the central component could be emission from the 5400 km s−1. In the present, t = 31.6 yr, image, the black hole’s accretion disc. In this respect, the coinci- outer edge of the radio emission is poorly determined dence(mentionedin§3.3)oftheapproximatealignment due to the low signal-to-noise ratio. Although continued of the elongation of the central component with the NE expansion with r ∝ t0.69 is compatible with our mea- hot spot and the possible SW enhancement becomes in- surements,itrequiresthattheemissionneartheforward triguing. Although inconclusive, it is suggestive of the shock is fading. In other words, the locus of the bright- action of a jet and counterjet, which has been suggested est radio emission seems to have stopped expanding and asanalternatetothedelayedneutrinoshockmechanism may even be moving inward. The radio emission region, forcorecollapseSNe(e.g.,Soker2010;Gilkisetal.2016). even aside from the emergence of the central component We will discuss the nature of the central component in seems to be evolving in a non self-similar fashion. It more detail in a forthcoming Paper IV on the evolution is interesting to note that Dwarkadas & Gruszko (2012) of the spectral energy distribution. foundthatalsotheobservedX-rayemissionofSN1986J was hard to reconcile with a self-similar evolution since 5. SUMMARYANDCONCLUSIONS early times. 1. We obtained a new phase-referenced global-VLBI im- The central component is stationary to within our un- age of SN 1986J at 5 GHz, showing the continued evo- certainties of 570 km s−1, corresponding to 8% of the lution of this supernova in the radio. An animation is average expansion speed of the shell. It is marginally re- available in the electronic edition. solved, and has a radius (HWHM) of r = 6.7+0.7 × comp −3.7 2. The 5-GHz VLBI image is now dominated by a 1016 cm. If it had radius = 0 at the time of the explo- marginally resolved central component. The ejecta shell sionin1983.2,itsaverageexpansionspeedsincethenwas is only barely visible. The peak brightness of the central 680+80 km s−1, or 9+1% of that of the shell. component is ∼10 times higher than that of the shell. −380 −5 The motions of both the central component and the 3. The flux density of the central component is still in- shellhot-spot,andtheexpansionofthemorediffuseshell creasing, both in absolute terms and as a fraction of the emission,areallthereforecompatiblewithahomologous total. Since its first detection in 2003 (t = 20.3 yr), its expansion with r ∝ t0.69, originating from the geomet- 5-GHz flux density has increased by a factor of ∼4. Its rical center of the shell, which is coincident within the currentluminosity(νL at5GHz)is20∼30timesthat uncertaintieswiththepositionofthecentralcomponent. ν of the Crab Nebula (3.5∼4.5×1035 erg s−1). However, the expansion does not in fact seem to be en- tirelyhomologous,sincetheemissionneartheouteredge 4. The 3σ upper limit on the linearly polarized fraction seems to be fading relative to that nearer the center. of the image peak, which is the central component, was What is the nature of the central component? There 3.3% are several hypotheses. We suggested in Paper II that 5. The emission from the shell is decreasing, and the it could be emission from the supernova shock running brightness of the outer edge seems to have faded more into a dense condensation in the CSM fortuitously near than that nearer the center. The angular outer radius the center of the shell in projection. As mentioned in of the radio emission, identified with the location of the § 3.2.1, the brightness of the central component is at forward shock, is no longer well determined. Continued least 10× brighter than the shell. This high and still expansion with r ∝ t0.69 seen earlier is consistent with increasing brightness, coupled with the central compo- our new VLBI image, but it is also possible that the nent’s stationarity, and its long lifetime argue against deceleration has increased. that hypothesis. It seems likely therefore that the cen- 6. In earlier observations, there was a prominent hot- tralcomponentreallyisinornearthethree-dimensional spot to the NE in the shell. The shell hot-spot, which has almost faded from view in our present, t = 31.6 yr, 7 As we noted in Paper II, the angular radius of the forward image, was moving outward with a projected speed of shock is probably not identical to θ90%flux, our estimate of the 2810±750 km s−1 at p.a. 57◦ ±15◦ between t = 15.9 outer edge of the radio emission, but we expect the two to be and 25.6 yr. The shell hot spot’s motion is consistent close, and moreover, since we used a convolving beam that scales with the expansion, for the ratio between θ and the shock with it taking part in a homologous expansion together 90%flux radiustoremainrelativelyconstantastheSNexpands. with the shell, in other words, having an origin at the 9 explosion center in 1983.2 and moving radially outward Bietenholz,M.F.,Bartel,N.,Milisavljevic,D.,etal.2010a, with r ∝t0.69. MNRAS,409,1594 Bietenholz,M.F.,Bartel,N.,&Rupen,M.P.2001,ApJ,557,770 7. The central component seems to be marginally re- —.2002,ApJ,581,1132,(PaperI) solvedinourobservations. WefounditsFWHMangular —.2003,ApJ,597,374 —.2004,Science,304,1947 diameteratt=31.6yrtobe900+100 µas,corresponding −500 —.2010b,ApJ,712,1057,(PaperII) to r = 6.7+0.7 ×1016 cm. If it has expanded since Chevalier,R.A.1982,ApJ,259,302 comp −3.7 —.1987,Nature,329,611 theexplosionin1983.2,thentheaverageprojectedspeed —.2012,ApJ,752,L2 ofexpansionwas680+−83080 km s−1,or9%thespeedofthe Chevalier,R.A.2014,inIAUSymposium,Vol.296,Supernova outer edge of the shell. EnvironmentalImpacts,ed.A.Ray&R.A.McCray,95–102 Chevalier,R.A.,&Fransson,C.2016,ArXive-prints, 8. The central component has a proper motion corre- arXiv:1612.07459 sponding to a projected speed of 380±570 km s−1 be- Cornwell,T.,Braun,R.,&Briggs,D.S.1999,inAstronomical SocietyofthePacificConferenceSeries,Vol.180,Synthesis tweent=20.3yrand31.6yr(2003and2014),consistent ImaginginRadioAstronomyII,ed.G.B.Taylor,C.L.Carilli, with being stationary. Its position is consistent within &R.A.Perley,151 the uncertainties of the explosion position that we esti- Cornwell,T.J.2008,IEEEJournalofSelectedTopicsinSignal mated earlier. Processing,2,793 Deller,A.T.,Brisken,W.F.,Phillips,C.J.,etal.2011,PASP, 9. The latest observations argue in favor of the cen- 123,275 tral component being located at or near the three- Dwarkadas,V.V.,&Gruszko,J.2012,MNRAS,419,1515 dimensional center of SN 1986J, rather then being as- Fey,A.L.,Gordon,D.,&Jacobs,C.S.,eds.2009,IERS TechnicalNote,Vol.35,TheSecondRealizationofthe sociated with the expanding shell and being central only InternationalCelestialReferenceFramebyVeryLongBaseline in projection. Interferometry(Frankfurt:VerlagdesBundesamtsfu¨r KartographieundGeoda¨sie),1 Fransson,C.,Lundqvist,P.,&Chevalier,R.A.1996,ApJ,461, ACKNOWLEDGMENTS 993 We thank N. Soker for comments on the manuscript. Gilkis,A.,Soker,N.,&Papish,O.2016,ApJ,826,178 The European VLBI Network is a joint facility of Eu- Greisen,E.W.,Spekkens,K.,&vanMoorsel,G.A.2009,AJ, 137,4718 ropean and Chinese radio astronomy institutes funded Hunter,D.A.,Ficut-Vicas,D.,Ashley,T.,etal.2012,AJ,144, by their national research councils. This research was 134 supported by both the National Sciences and Engineer- Ma,C.,Arias,E.F.,Eubanks,T.M.,etal.1998,AJ,116,516 ing Research Council of Canada and the National Re- Marcaide,J.M.,Mart´ı-Vidal,I.,Perez-Torres,M.A.,etal.2009, searchFoundationofSouthAfrica. Wehavemadeuseof A&A,503,869 Marcaide,J.M.,Alberdi,A.,Ros,E.,etal.1995,Nature,373,44 the NASA/IPAC Extragalactic Database (NED) which Mart´ı-Vidal,I.,Marcaide,J.M.,Alberdi,A.,etal.2011,A&A, is operated by the Jet Propulsion Laboratory, Califor- 526,A142 nia Institute of Technology, under contract with the Na- Milisavljevic,D.,&Fesen,R.A.2017,ArXive-prints, tional Aeronautics and Space Administration (NASA), arXiv:1701.00891 as well as NASA’s Astrophysics Data System Abstract Milisavljevic,D.,Fesen,R.A.,Leibundgut,B.,&Kirshner,R.P. 2008,ApJ,684,1170 Service. Papish,O.,Soker,N.,&Bukay,I.2015,MNRAS,449,288 Pradel,N.,Charlot,P.,&Lestrade,J.-F.2006,A&A,452,1099 Rich,J.W.,deBlok,W.J.G.,Cornwell,T.J.,etal.2008,AJ, REFERENCES 136,2897 Rupen,M.P.,vanGorkom,J.H.,Knapp,G.R.,Gunn,J.E.,& Barkov,M.V.,&Komissarov,S.S.2011,MNRAS,415,944 Schneider,D.P.1987,AJ,94,61 Bartel,N.,&Bietenholz,M.F.2008,ApJ,682,1065 Soker,N.2010,MNRAS,401,2793 Bartel,N.,&Bietenholz,M.F.2014,inIAUSymposium,Vol. vanGorkom,J.,Rupen,M.,Knapp,G.,etal.1986,IAUCirc., 296,IAUSymposium,ed.A.Ray&R.A.McCray,53–57 4248,1 Bartel,N.,Rupen,M.P.,Shapiro,I.I.,Preston,R.A.,&Rius, Wakker,B.P.,&Schwarz,U.J.1988,A&A,200,312 A.1991,Nature,350,212 Weiler,K.W.,Panagia,N.,&Sramek,R.A.1990,ApJ,364,611 Bartel,N.,Ratner,M.I.,Rogers,A.E.E.,etal.1987,ApJ,323, Zanardo,G.,Staveley-Smith,L.,Ng,C.-Y.,etal.2013,ApJ,767, 505 98 Bietenholz,M.2014,in12thEuropeanVLBINetwork Zanardo,G.,Staveley-Smith,L.,Indebetouw,R.,etal.2014, SymposiumandUsersMeeting(2014),publishedbySISSA, ApJ,796,82 Trieste,ed.A.Tarchi,M.Giroletti,&L.Feretti,51