New Observations of the Very Luminous Supernova 2006gy: Evidence for Echoes

0.38 MB·English

by  A. A. Miller

Draftversion June12,2009
PreprinttypesetusingLATEXstyleemulateapjv.08/13/06
NEW OBSERVATIONS OF THE VERY LUMINOUS SUPERNOVA 2006GY: EVIDENCE FOR ECHOES
A. A. Miller1, N. Smith1, W. Li1, J. S. Bloom1,2, R. Chornock1, A. V. Filippenko1, and J. X. Prochaska3,4
Draft version June 12, 2009
ABSTRACT
Supernova(SN)2006gywasahydrogen-richcore-collapseSNthatremainsoneofthemostluminous
optical supernovae ever observed. The total energy budget (> 2×1051 erg radiated in the optical
9
alone) poses many challenges for standard SN theory. We present new ground-based near-infrared
0
(NIR) observations of SN 2006gy,as well as a single epoch of Hubble Space Telescope (HST) imaging
0
obtained more than two years after the explosion. Our NIR data taken around peak optical emission
2
show an evolution that is largely consistent with a cooling blackbody, with tentative evidence for a
n
growingNIRexcessstartingatday∼100. Ourlate-timeKeckadaptiveoptics(AO)NIRimage,taken
u
on day 736, shows little change from previous NIR observations taken around day 400. Furthermore,
J
theopticalHSTobservationsshowareduceddeclinerateafterday400,andtheSNisblueronday825

2 thanitwasatpeak. Thislate-timedeclineisinconsistentwith56Codecay,andthusisproblematicfor
1
the various pair-instability SN models used to explain the nature of SN 2006gy. The slow decline of

the NIR emissioncanbe explained with a light echo,andwe confirmthat the late-time NIR excess is
]
O theresultofamassive(&10M⊙)dustyshellheatedbytheSNpeakluminosity. Thelate-timeoptical
observationsrequirethe existence ofa scatteredlightecho,whichmaybe generatedby the same dust
C
that contributes to the NIR echo. Both the NIR and optical echoes originate in the proximity of
h. the progenitor, ∼1018 cm for the NIR echo and .20 pc for the optical echo, which provides further
p evidence that the progenitor of SN 2006gy was a very massive star.
- Subject headings: supernovae: general — supernovae: individual(SN 2006gy)
o
r
t
s 1. INTRODUCTION Smith et al. 2009b).
a
Pair-instability SNe (Barkat, Rakavy, & Sack 1967;
[ At the time of discovery, supernova (SN) 2006gy was
Rakavy & Shaviv 1967; Bond, Arnett, & Carr 1984)
themostluminousSNeverfound(Ofeketal.2007;Smith
1 are expected to occur in very massive, low-metallicity
et al. 2007). SN 2006gy generated a great deal of inter-
v stars, such as those that may have been present in the
est;inadditiontobeing∼100timesmoreluminousthan
1 metal-free environment of the very early universe (e.g.,
a typical Type II (hydrogen-rich, core-collapse) SN at
0 Abel, Bryan, & Norman 2000). The detection of a pair-
peak, it exhibited a long rise time (∼70 day) and slow
2 instability SN in the comparatively local universe, then,
2 decline,leadingtospeculationthatitmayhavebeenthe
could potentially reveala great deal about the first gen-
. firstobservedexampleofapair-instabilitySN(Ofeketal.
6 eration of stars. The light curves of pair-instability SNe
2007; Smith et al. 2007) or a pulsational pair-instability
0 areexpectedtoexhibitarelativelyslowrise,followedby
SN (Woosley, Blinnikov, & Heger 2007).
9 a broad turnover after the peak, and a peak luminos-
SN2006gywasclassifiedasaTypeIInSN(seeSchlegel
0 ity that is considerably larger than that of typical SNe
1990 for a definition of the Type IIn subclass and Filip-
: (Scannapieco et al. 2005). Qualitatively each of these
v penko 1997 for a review of its spectral properties) based
i on the relatively narrowemission features present in the characteristics matches those observed for SN 2006gy.
X Other more luminous SNe have been announced since
early-time SN spectrum (Ofek et al. 2007; Smith et al.
r 2007). Some Type IIn supernovae (SNe IIn) are known the discovery of SN 2006gy: SNe 2005ap (Quimby et
a
al. 2007) and 2008es (Miller et al. 2009; Gezari et al.
to be overluminous relative to their typical SN II coun-
2009). This suggests that while these events are rare,
terparts, probably due to the collision of fast-moving
theremaybeanewsubclassofveryluminoussupernovae
SN ejecta and a dense, possibly clumpy, circumstellar
(VLSNe). The peak luminosity and photometric evolu-
medium (CSM; Chugai & Danziger 1994). In a com-
tion of both SNe 2005ap and 2008es are difficult to ex-
panion paper (Smith et al. 2009b), a detailed spectro-
plain via the pair-instability model (Quimby et al. 2007;
scopiccomparisonofSN2006gyismadetootherSNeIIn.
Miller et al. 2009), implying that peak luminosities &
SN 2006gy is unique within the SN IIn subclass, how-
few ×1044 erg s−1 are possible without a pair-instability
ever, because typical interaction models cannot explain
explosion. There is a wide diversity of alternative mod-
its early-time behavior, suggesting the need for alterna-
els that have been developed to explain the early-time
tive models for this particular object (Smith et al. 2007;
observations of SN 2006gy. Smith & McCray (2007) ar-
Woosley, Blinnikov, & Heger 2007; Nomoto et al. 2007;
gue that the peak luminosity and light-curve evolution
1Department of Astronomy, University of California, Berkeley, can be explained via the thermalization of shock energy
CA94720-3411. depositedintoamassive(∼10M⊙),opticallythickshell.
2SloanResearchFellow. Similar explanations were offered for SNe 2005ap and
3Universityof CaliforniaObservatories/Lick Observatory, Uni-
2008es (Quimby et al. 2007; Miller et al. 2009). Based
versityofCalifornia,SantaCruz,CA95064.
4DepartmentofAstronomyandAstrophysics,UniversityofCal- ona model ofthe lightcurve nearpeak, Agnoletto et al.
ifornia,SantaCruz,CA95064. (2009) suggest that the combination of ejecta colliding
2 Miller et al.
withdenseclumpsinthe CSMand∼1–3M⊙ of56Niare tionmustbeappliedwhendeterminingthesourceofany
responsible for the early-time luminosity of SN 2006gy. late-time emission.
Nomotoetal.(2007)areunabletomatchthelightcurve In this paper we present new NIR observations of
withastandardpair-instabilitySNmodel;however,they SN 2006gy, taken around the peak of optical emission,
can reasonably reproduce the (.400 day) light curve of as well as optical and NIR observations obtained more
SN2006gyusingapair-instabilitymodelwheretheyarti- than two years after SN 2006gy exploded. In § 2 we
ficially reduce the ejecta mass, such that the radioactive describe the observations and data reduction. We dis-
heating is less than 100% efficient. Woosley, Blinnikov, cuss the results in § 3, and in § 4 we offer some con-
& Heger (2007) use a model of a pulsational pair insta- clusions. Throughoutthis paperweassumethatthedis-
bility within a massive star to explain the light curve of tancetoNGC1260(thehostgalaxyofSN2006gy)is73.1
SN 2006gy. Mpc, and that the total color excess toward SN 2006gy
Some of the models make predictions for the late-time is E(B −V) = 0.72 mag (Smith et al. 2007), of which
behavior of SN 2006gy. Those with a large yield of 56Ni E(B−V) = 0.54 mag is in NGC 1260 and Galactic ex-
(Nomoto et al. 2007; Smith et al. 2007) would expect a tinctionaccountsforE(B−V)=0.18mag. Allspectral
declineinbolometricluminosityattherateof56Codeay, energydistributions (SEDs) have been correctedfor this
0.98 mag (100 day)−1, or faster if the radioactive-decay color excess assuming R = A /E(B−V) = 3.1 using
V V
energyisnotconvertedtoopticalemissionwith100%ef- thereddeninglawofCardelli,Clayton,&Mathis(1989).
ficiency. The shell-shock model (Smith & McCray 2007)
predicts a rapid decline after ∼200 day, though the au- 2. OBSERVATIONS
thorsnotethatthis decline maybe offsetby the produc- NIRobservationsofSN2006gywereobtainedsimulta-
tion of a large amount of 56Ni or continued CSM inter- neously in J, H, and K with the Peters Automated In-
s
action. The pulsationalpair-instability model (Woosley, fraredImagingTelescope(PAIRITEL;Bloometal.2006)
Blinnikov, & Heger 2007)predicts another SN explosion starting on 2006 October 135 (54 day after explosion6).
at the location of SN 2006gyabout 9 years after the ini- PAIRITEL is a 1.3-m robotic telescope, located on Mt.
tial outburst from SN 2006gy. Hopkins, AZ, which obtained observations for 30 min to
More than a year after the explosion, Smith et al. 1 hr at each epoch through normal queue-scheduled op-
(2008b) detected SN 2006gy in the near-infrared (NIR) erations for the next 24 months. In total there were 124
at a luminosity comparable to that of the peak luminos- epochsobtained. Allimageswereprocessedviaanauto-
ityofmostSNeII.Thispropertyhadnotbeenpredicted mated pipeline (Bloom et al. 2006).
by any of the models. When coupled with the lack of Analysis of SN 2006gyhas provedchallengingbecause
a detection in the radio and X-rays, this led Smith et the SN is located very close to the nucleus of NGC 1260
al. (2008b)to conclude that the luminosity could not be (separation < 1′′; Ofek et al. 2007; Smith et al. 2007).
powered by the continued interaction of the SN ejecta PAIRITEL has a large native scale of 2′′ pixel−1, pre-
with CSM. Kawabata et al. (2009) draw similar conclu- cluding spatial resolution of the SN from the nucleus.
sions based on their detection of weak Hα emission at This necessitates image differencing to obtain the light
a comparable epoch. Smith et al. (2008b) conclude that curve of the SN. Image subtraction was performed with
onlytwopossibilitiesareabletoexplainthelate-timeob- HOTPANTS7, and the flux in each difference pair was
servationsofSN2006gy: (i)theexplosionproduced&2.5 determined via aperture photometry at the location of
M⊙of 56Ni, possibly as a pair-productionSN, which was the SN. An example subtraction, which clearly shows
heating dust and consequently generating the large NIR flux from SN 2006gy after the reference image has been
excess,or(ii)amassive(∼5–10M⊙),dustyshell,located subtracted, is shown in Figure 1.
∼1 light year from the site of the SN, was being heated Despite observations extending more than two years
by the radiation produced at peak, and re-radiated that past the date of discovery, SN 2006gy has not faded be-
energy as a NIR echo (Dwek 1983). For the first case, if yond the point of detectability with PAIRITEL. Con-
the luminosity were powered by radioactive decay, then sequently, all J, H, and K images of SN 2006gy con-
s
futureobservationsshouldindicateacontinueddeclineat tain some flux from the SN. We thus adopted the “NN2
the rate of 56Co decay. A dust echo, on the other hand, method” ofBarrisetal.(2005)to determine the relative
wouldresultinaNIRlightcurvethatstaysroughlycon- NIR flux changes of the SN. The NN2 method treats
stant for ∼1–2 years before exhibiting a rapid decline. all images equally and does not require a template im-
Evolved massive stars, such as red super giants and agewith no light fromthe sourceof interest,SN 2006gy.
luminous blue variables(LBVs), are oftenobservedwith It uses the subtraction of all N(N − 1) possible pairs
massivedust shells, so if these stars explode as Type IIn of images to mitigate against possible errors associated
SNeonemightexpectalate-timeIRecho. ManySNeIIn with the use of a single reference template image. The
have been observed to exhibit a late-time NIR excess downsidetotheNN2methodisthatitonlyproducesthe
(Gerardyetal.2002)whichinsomecaseslasted>1year. differentialfluxbetweeneachoftheN epochsofobserva-
This excess has been attributed to NIR echoes (Gerardy tions. To convert these flux differences into magnitudes
et al. 2002), though we note that the formation of new requiresanabsolutecalibration,whichmustbeobtained
dust has been argued specifically for SN 1998S (Pozzo independently of the results from the NN2 method. Un-
et al. 2004) and SN 1995N (Fox et al. 2009). Dusty re- certainties in the individual subtractions were estimated
gions near the SN should also lead to UV/optical scat-
teredlightechoes(Chevalier1986);thus,dustis capable 5 Alldates inthispaperareUTunlessotherwisenoted.
of providing significant optical and NIR emission at late 6 FollowingSmithetal.(2007),weadopt2006August20asthe
explosiondateforSN2006gy.
times. The optical decline of Type IIn SNe at late times
7http://www.astro.washington.edu/users/becker/hotpants.html
is veryheterogeneous(Li etal.2002),andthereforecau- .
Echoes from SN 2006gy 3
bymeasuringthescatterinfakeSNeinsertedatlocations
TABLE 1
havingasurfacebrightnesssimilartothatatSN2006gy. PAIRITEL Observationsof SN2006gy
Early-time NIR observations of SN 2006gy (defined
here as those made before SN 2006gy passed behind the tmida J magb H magb Ks magb
(MJD) (Vega) (Vega) (Vega)
Sun during 2007) show a remarkably flat light curve, as 54022.24 13.33±0.04 13.05±0.06 12.81±0.05
seen in Figure 2. To transform the relative flux differ- 54023.24 13.33±0.04 13.08±0.09 12.83±0.05
encesfromtheNN2methodtoanabsolutescale,wesub- 54025.26 13.29±0.04 13.03±0.06 12.84±0.06
tractedthearchivalTwoMicronAllSkySurvey(2MASS; 54028.17 13.32±0.04 13.03±0.06 12.78±0.04
54029.19 13.26±0.04 13.03±0.06 12.81±0.05
Skrutskie et al.2006)image from eachof the images ob- 54030.41 13.27±0.03 13.02±0.06 12.81±0.06
tainedon2006Nov. 13.24,14.25,15.38,and16.37tode- 54031.43 13.27±0.04 13.01±0.06 12.79±0.04
termine the J,H, andK magnitudes ofthe SN onthese 54032.41 13.27±0.04 13.01±0.06 12.81±0.05
dates. ThemeanSNfluxswasthenusedtotransformthe 54036.21 13.25±0.04 13.02±0.06 12.78±0.04
54037.20 13.21±0.05 13.00±0.06 12.77±0.04
relativefluxfromtheNN2methodtotheabsolutefluxof 54038.24 13.25±0.03 13.02±0.06 12.76±0.04
the2MASSsystem(Cohen,Wheaton,&Megeath2003). 54040.24 13.23±0.04 13.01±0.06 12.77±0.04
PAIRITEL uses the old 2MASS camera and telescope; 54040.69 13.25±0.04 13.01±0.06 12.77±0.05
54041.71 13.29±0.04 13.01±0.06 12.78±0.04
hence, we do not expect any large systematic effects in
54042.71 13.22±0.05 13.03±0.08 12.79±0.09
the 2MASS subtractions. The final calibrated J, H, and 54044.66 13.26±0.04 13.00±0.06 12.74±0.04
K photometry is summarized in Table 1. 54045.66 13.28±0.05 13.00±0.06 12.79±0.05
s
On2006Nov. 01,Ofeketal.(2007)obtainedPalomar 54046.65 13.33±0.04 13.02±0.07 12.80±0.05
54047.64 13.26±0.03 12.99±0.06 12.76±0.05
Hale (5 m) AO images in the J and Ks bands, clearly 54048.62 13.29±0.04 13.02±0.06 12.74±0.05
resolving SN 2006gy from the host-galaxy nucleus, and 54049.62 13.27±0.04 13.01±0.06 12.78±0.05
measured its flux. In the K band our calibration and 54050.66 13.25±0.03 13.00±0.06 12.78±0.04
s
the Ofek et al. (2007) measurement agree to within 1σ, 54051.70 13.25±0.04 13.00±0.06 12.76±0.04
54052.64 13.24±0.04 13.00±0.06 12.77±0.04
while the agreement in the J band is somewhat worse 54053.66 13.25±0.03 13.01±0.06 12.76±0.04
(∼2σ). Ofek et al. (2007) had only a single 2MASS star 54054.78 13.27±0.04 13.01±0.06 12.76±0.04
withinthefieldofviewoftheirAOimages,whereas>100 54055.78 13.28±0.04 13.01±0.06 12.77±0.04
2MASS stars were used to calibrate the PAIRITEL im- 54058.57 13.27±0.03 13.00±0.06 12.75±0.04
54059.59 13.24±0.04 13.01±0.06 12.74±0.05
ages; when coupled with the difficulty associated with 54060.61 13.27±0.03 13.01±0.06 12.79±0.04
photometry of AO images, this may explain the differ- 54061.60 13.24±0.04 13.01±0.06 12.75±0.05
encesbetweenthetwomeasurements. Wenotethatwere 54063.63 13.28±0.04 13.00±0.06 12.74±0.05
we to adopt the Ofek et al. (2007) measurements as our 54064.63 13.30±0.04 13.01±0.06 12.76±0.04
54065.63 13.24±0.07 13.08±0.11 12.83±0.19
calibration,there would be an overallsystematic shift of 54066.65 13.31±0.04 13.01±0.06 12.76±0.05
our J and Ks light curves to brighter values, which in 54069.55 13.27±0.04 13.04±0.06 12.82±0.05
turn would lead to worse agreement between the NIR 54070.96 13.31±0.04 13.02±0.06 12.77±0.04
54071.98 13.31±0.05 13.02±0.06 12.83±0.05
data and early-time optical spectra (see Figure 4). This
54072.97 13.28±0.04 13.01±0.06 12.77±0.04
suggests that our calibration method is sufficient. We 54074.92 13.31±0.04 13.03±0.06 12.82±0.05
note that the uncertainties in our photometry are dom- 54075.96 13.31±0.04 13.01±0.06 12.77±0.05
inated by the uncertainty in the calibration, which is 54076.95 13.37±0.04 13.02±0.06 12.79±0.04
54077.95 13.37±0.04 13.02±0.06 12.80±0.04
∼0.03maginJ, ∼0.06maginH,and∼0.04maginKs. 54080.01 13.34±0.05 13.03±0.06 12.80±0.05
This uncertainty is the same for all epochs, so a change 54081.01 13.41±0.05 13.06±0.06 12.80±0.04
in the calibrationwould lead to a systematic shift of the 54082.01 13.40±0.04 12.98±0.06 12.82±0.05
entire light curve. 54082.99 13.32±0.07 13.07±0.07 12.79±0.05
54083.99 13.36±0.09 13.00±0.06 12.75±0.05
As shown by Smith et al. (2008b), and subsequently 54084.97 13.39±0.05 13.05±0.06 12.84±0.08
confirmed by Agnoletto et al. (2009), the NIR evolution 54085.96 13.40±0.05 13.08±0.07 12.84±0.05
of SN 2006gy is very slow at late times. Given the rel- 54086.96 13.48±0.06 13.06±0.07 12.88±0.06
atively small change in flux, and the reduced signal-to- 54090.91 13.42±0.04 13.11±0.07 12.88±0.05
54091.91 13.59±0.09 13.11±0.07 12.93±0.05
noiseratiofollowingthefadingoftheSN,wewereunable 54093.12 13.51±0.11 13.06±0.06 12.90±0.07
to recover reliable flux measurements from PAIRITEL 54094.95 13.68±0.11 13.12±0.07 12.90±0.05
data taken after 2007 Sep. Furthermore, unlike the case 54095.95 13.66±0.11 13.13±0.07 13.00±0.16
atearlytimes,theSNhadfadedbelowthe2MASSdetec- 54100.97 13.79±0.08 13.15±0.07 13.03±0.13
54101.42 13.81±0.07 13.16±0.07 13.16±0.14
tion limit, which means that the late-time subtractions
relativetothe2MASStemplateimagedonotyieldmean-
aMidpointbetweenthefirstandlastexposuresinasinglestacked
ingful results despite the fact that at K ≈15 the SN is
s image.
well above the PAIRITEL detection limit. In principle, bObservedvalue;notcorrectedforGalacticextinction.
if deep PAIRITEL images are obtained after SN 2006gy
fades well beyond the detection limit, it should be pos- SN 2006gy, as summarized in Table 2.8 We also present
sible to recoverthe late-time NIR light curve using tem- the first measurement of the H-band flux from the AO
plate images that contain no light from the SN. images taken on 2007 Dec. 2. To obtain this H-band
We also observed SN 2006gy with the Near-Infrared measurement,despite a lackof2MASSstarsinthe field,
Camera 2 (NIRC2) using the laser guide star (LGS) AO we measured the H −K′ color of SN 2006gy relative to
system(Wizinowichetal.2006)onthe10-mKeckIItele-
scope in Hawaii on 2008 Aug. 25. We have re-reduced 8 Late-time AO Keck observations were all made in K′-band.
The uncertainty associated with the transformation between K′-
the LGS AO images presented by Smith et al. (2008b)
and derive revisedvalues for the K′-bandmagnitudes of ubnancedrtaanindtyKasn-bdatnhderiesfosmreaflolrcothmeplaarteed-titmoethAeOabimsoalguetseKca′li≈brKatsio.n
4 Miller et al.
Fig. 1.— Example subtraction of PAIRITEL NIR images of SN 2006gy, which have all been registered to the same coordinates. Each
imageis∼2′ ×3′ insize. Left: PAIRITELKs-bandimagetakenon2006Nov. 15. Middle: PAIRITELKs-bandimagetakenon2007Jan.
01. Right: Image(leftpanel)minusreference(middlepanel)subtractionimage. TheimagesubtractionwasperformedwithHOTPANTS,
andSN2006gyisclearlyvisibleinthedifferenceimageasabrightpointsourcenearthegalaxynucleus.
TABLE 3
ga] 1143 KAIT/opticalPAIRITEL/NIRP200/NIRR + 0.5 FFFi45lt55e05rWWHSTλ45c(d53e˚Aa19n)t97ya8251∆(92˚A5Oλ27)b6bser22v12a..51(tmV16ioea±±gngacs00)..o00f32SN592..930F71(0µl6±±uJgxyy00d)..1189
Ve J + 1.0 F675W 6697 866 21.07±0.03 11.29±0.34
ag [ 15 KH + 0.5 F814W 7924 1500 20.91±0.04 10.80±0.43
M
16 aCentral wavelength of the filter, based on WFPC2 calibrations
presentedinTable8ofHoltzmanetal.(1995b).
bFilterwidth,basedonWFPC2calibrationspresentedinTable8
ofHoltzmanetal.(1995b).
17
cObservedvalue;notcorrectedforGalacticextinction.
dObserved flux using the photometric calibrations presented in
0 50 100 150 200 Table9ofHoltzmanetal.(1995b).
Days since explosion [observed frame]
Dolphin(2000)todochargetransferefficiencycorrection
and photometric reduction of the WFPC2 images. Our
Fig. 2.—Earlyphotometric evolutionofSN 2006gy. Unfiltered
(approximately R band) observations are taken from Smith et al. measurementsofthe SN2006gymagnitudesaresumma-
(2007). PAIRITELJ,H,andKs observations arefromthiswork. rized in Table 3. A false-color image of this detection is
WealsoshowthePalomarAOphotometryfromOfeketal.(2007). shown in Figure 3.
Thedatahavenotbeencorrectedforextinctioninthehostorthe
Gcoalolarxiyn.creTahseingNIsRteaedviolyluatfiotenridsayre∼m7a0r.kaTbhlyisflbaeth,avwiiotrhisthceonRsis−tenKt 3. DISCUSSION
withacoolingblackbody (seetext). 3.1. Early-Time NIR Observations
While the early evolution of SN 2006gy is remarkably
TABLE 2 flat in the NIR (see Figure 2), we find that these mea-
Keck AOObservationsof SN2006gy surements are consistent with radiation from a cooling
blackbody. Fromblackbodyfitstoopticalspectra,Smith
date epocha Filter magb
et al. (2009b) find that the temperature of SN 2006gy
(UT) (day) (Vega)
2007Sep. 29 401 K′ 14.91±0.17 monotonically cools after day ∼50, roughly two weeks
2007Dec. 02 469 H 16.8±0.3 beforeopticalmaximum,untilday∼165,wherethetem-
2007Dec. 02 469 K′ 15.02±0.17 perature levels off at ∼6300 K. In the top panel of Fig-
2008Aug. 25 736 K′ 15.59±0.21
ure4weshowtheevolutionoftheSEDofSN2006gyfor
threeepochs(day59,109,and135)duringourearly-time
aObserved-framedayfromtheadoptedexplosiondate,2006Aug.
NIR observations. For each epoch we show the single-
20(Smithetal.2007).
bObservedvalue;notcorrectedforGalacticextinction. componentblackbodyspectrum,afteradoptingthetem-
perature from Smith et al. (2009b) and normalizing the
the H −K′ color of the host galaxy, and calibrated this spectra to the photometric measurements from Smith et
against the H −K color of the galaxy in the archival al. (2007). Their photometric observations come from a
s
2MASS images. The large uncertainty for this measure- seriesofunfilteredobservationstakenwiththe Katzman
ment reflects the accuracy with which we can determine Automatic Imaging Telescope (KAIT; Filippenko et al.
the color of the galaxy from the 2MASS images. 2001), which are best matched by the R band (Riess et
As part of a Hubble Space Telescope (HST) snapshot al. 1999; Li et al. 2003). However, the scatter in the
survey(GO-10877;PILi),SN2006gywasobservedwith transformation between KAIT unfiltered and R can be
theWideFieldPlanetaryCamera2(WFPC2;Holtzman quitelarge(Ganeshalingam 2009),soweadopta0.2mag
et al. 1995a) using the F450W, F555W, F675W, and uncertainty for the calibrationof the blackbody spectra.
F814W filters on 2008 Nov. 22. The data were reduced We also show the NIR flux during these epochs, deter-
in the standardfashion using multidrizzle (Koekemoer mined from the absolute calibration of the 2MASS sys-
et al. 2002). NGC 1260 and SN 2006gy were located at tem (Cohen, Wheaton, & Megeath 2003). This shows
the center of the PC chip of WFPC2, which has a na- excellent agreement with an extrapolation of the early-
tivepixelscaleof0.0455′′ pixel−1.Wefollowtherecipeof timespectraofSN2006gy(Smithetal.2009b). Between
Echoes from SN 2006gy 5
day 59 day 59, TBB = 12000
day 109, TBB = 7660
day 133, TBB = 6645
mJy) 10 day 109
F (ν
day 133
BB 0.4 10000
)/fB 0.2
B
− f 000...000
(f obs
4000 6000 8000 10000 20000
λ (Å)
eff
Fig. 4.— SED evolution of SN 2006gy at early times show-
ing a possible trend towards a growing IR excess above a cooling
blackbody. Top: thesolid,dotted,anddashedlinesshowasingle-
componentblackbodyspectrum,normalizedtounfilteredobserva-
tionsfromSmithetal.(2007),onday59,109,and135,respectively.
Bottom: thefractionalexcessemissioninJ,H,andKs,relativeto
theblackbodymodel. Forclaritywedonotincludetheuncertain-
tiesinthesemeasurements,whicharelargeanddominatedbythe
uncertaintyinthetransformationbetweenunfiltereddataandthe
Rband(seetext). Eachpointiswithin∼1σofshowingnoexcess;
however,theNIRexcess doesappeartogrowwithtime.
is growingin the H and K bands as a function of time.
s
Thetrendmaybeindicativeofradiationfromwarmdust
(see below), though we note that this effect would be
small. If this trend toward a NIR excess is created by
the same source as the late-time NIR excess, we would
expect a K -band excess of ∼0.17 mag, comparable to
s
the uncertainty in our blackbody calibration.
3.2. Late-Time NIR Observations
With data obtained more than a year after explosion,
Smith et al. (2007) discovered a significant NIR excess
fromSN2006gy. OurNIRobservationstaken736dayaf-
terexplosionshowthattheK′-bandfluxfromSN2006gy
has only faded by a factor of ∼2 over the course of the
previous year. We cannot determine the total IR lu-
minosity at late times, because our NIR data do not
cover the peak of the SED. Furthermore, as first noted
bySmithetal.(2008b),theveryredH−K′ coloratlate
times indicates that the IR emission likely peaks in the
Fig. 3.—Late-timeoptical andNIRimagesofSN2006gy. Top: mid-IR. We can, however,place lower limits on the NIR
Keck AO H and K′ false-color image from Smith et al. (2008b). luminosity by assuming that the SED peaks in the K′
TheSNisclearlyredintheNIR.Notethatthefieldofviewforthis
band. Followingthis assumption, the NIR excess consti-
imageisslightlysmallerthanthatoftheothers. Middle: KeckAO
K′ image of SN 2006gy taken on day 736. The SN is still clearly tutes a slowly varying luminosity of & 2×108 L⊙ for >
visibleinthe NIR,andhas shownlittlechange sinceobservations 1 yr.
takenaroundday400. Bottom: HSTWFPC2false-colorimagein- The most likely explanation for this large luminosity
cludingthefourfiltersinwhichwedetectSN2006gy: F450W (cor- in the NIR is warm dust. The observed H −K′ color
respondingtoblue),F555W,F675W,andF814W (corresponding
tored). TheSNisclearlyblueintheopticalcomparedtothelight from day 436 corresponds to a temperature of ∼1000K,
fromthesurroundingstars. assuming a single-temperature blackbody. If the dust is
radiatingasasingleblackbodywithT =1000K,this
day ∼55–135roughly2–4%of the bolometricluminosity peak
of SN 2006gy was emitted in the NIR. would increase the above luminosity to & 3×108 L⊙.
We note that our NIR observations are only sensitive to
In the lower panel of Figure 4 we show the fractional
the warmest dust; it is possible that cooler dust with
excess of the photometric observations relative to the
T ≈ 600 K9 could dominate the IR emission, in which
single-component blackbodies (note that by definition
casetheluminositycouldbesignificantlyhigherthanthe
the R-band excess is set to zero). For clarity we do
values quoted above.
not show the uncertainties associated with the excess,
but after accounting for the large uncertainty in the R-
9 Smith et al. (2008b) show that the combination of the peak
bandcalibration,eachpointiswithin∼1σofzeroexcess.
luminosity and distance to the dust suggest an equilibrium tem-
Nevertheless, there is an apparent trend that the excess peraturearound600K.
6 Miller et al.
The location of this dust, and whether it is newly 1011
formed or pre-existing at the time of the SN explosion,
rmpfoelreamaihnaneiintanhtgteiontegbhxeatttehdneeadtdeeprdurmsoetlix:oncne(edgis)e.sd.rTaWhhdeeieaodtacuocsstnoitsvu-icedrocehoerelifiansotguinnrtegipemdforeesodsimisbtsioh5li6oteCirxetos-, osity (L)O • 1010 NTOoIpRttai clL aLul umLmuinimnooisnistoiytsyity
n
decay, (ii) collisional excitation of pre-existing dust, (iii) umi 109
heatingvia radiationfromcircumstellarinteraction,and L
c
(iv) a late-time IR dust echo, where pre-existing dust is etri
heatedby the radiationproducedwhile the SN wasnear m
o 108
its optical peak. ol
B
3.2.1. Radioactive Heating from 56Co 107 2.5 MO • 56Ni
Smith et al. (2008b) noted that the observed K′-band 200 400 600 800
Days since explosion
decaybetween2007Septemberand2007Decembercould
be explained with radioactive heating from a minimum
Fig. 5.— Evolution of the bolometric luminosity of SN 2006gy
of2.5M⊙ of56Ni,ifasufficientamountofdustformedin duringthefirst800dayafterexplosion. Opticaldataareshownas
order to move the luminosity into the NIR (though they black circles, and NIR data are red squares. Early-time data (<
favoredanother interpretation; see below). We show the 250day)arefromSmithetal.(2009b). Theopticaldetectionnear
day 400 is from Kawabata et al. (2009). All other data are from
bolometricevolutionofSN2006gythroughthefirst∼800
thiswork. NIRmeasurementsareonlylowerlimitstothetotalIR
day post-explosion, including both optical and NIR de- luminosity,becauseourobservationsdonotsampleredwardof2.2
tections,aswellasthe expectedlightcurvefrom2.5M⊙ µm (see text). The dashed line shows the expected decline if the
of 56Ni, in Figure 5. The early-time measurements (< luminositywerepoweredby2.5M⊙of56Ni. Theflatnatureofthe
light curve indicates that radioactivity is not the primary source
250 day) come from Smith et al. (2009b). The optical
ofthelate-timeemission.
measurementnearday400comesfromphotometricmea-
surements by Kawabata et al. (2009), while the optical materialinthe expandingblastwave. Theintense ultra-
luminosity on day 825 comes from a direct integration violet/opticaloutputfromaSNatitspeakvaporizesany
of our HST observations (see § 3.3.1). The three late- dust in the vicinity of the SN (Dwek 1983). This radia-
time NIR luminosities represent lower limits based on tion near peak creates a dust-free cavity into which the
themeasuredK′-bandflux(seeabove,§3.2). Thedecay SNejectamayexpandatearlytimes;however,theejecta
rate of 56Co, 0.98 mag (100 day)−1, is much faster than blastwavewilleventuallyreachtheedgeofthe dust-free
the observed decay of the K′ flux from SN 2006gy,∼0.2 cavity, at which point collisional excitations may gener-
mag (100 day)−1. The late-time NIR excess declines at ate NIR emission.
a rate that is too slow to be explained by 56Co heating The large peak luminosity of SN 2006gy, 8×1010 L⊙
alone. (Smith et al. 2009b), provides significant challenges to
The pair-instability model of Nomoto et al. (2007), thiscollisionalheatingscenario. Dwek(1983)showsthat
whichprovidedgoodagreementwiththeearlylightcurve theradiusofthedust-freecavitycanbedeterminedfrom
of SN 2006gy after an artificial reduction of the total the peak luminosity and the dust vaporization temper-
ejecta mass from the evolutionary calculation, was able ature, assuming the grain emissivity Q is proportional
v
to reproduce the late-time NIR luminosity observed by to (λ/λ )−1:
0
Smith et al. (2008b). This model required less than
100% efficiency in the conversion of gamma-rays (from Q¯νL0(rmL⊙) 0.5
radioactivedecay)tooptical/NIRemission,whichmeans R1(pc)=23(cid:20) λ (µm)T5 (cid:21) , (1)
that the light curve should decay faster than 0.98 mag 0 v
(100 day)−1. The possibility of a pair-instability SN where R is the radius of the dust-free cavity, Q¯ is the
1 ν
was first invoked to explain the large peak luminosity meangrainemissivity, L is the peak luminosity, andT
0 v
of SN 2006gy(Ofek et al. 2007;Smith et al. 2007). This is the dust vaporization temperature. Following Dwek
scenariowouldhaverequiredtheproductionof&10M⊙ (1983), if we assume Q¯ = 1 and that λ = 0.2 µm, we
ν 0
56Ni, which in turn would produce a large late-time lu- find that R ≈ 1.4×1018 cm, for a vaporization tem-
1
minosity that decays at the rate of 56Co. The late-time perature T ≈1000K. If the absorptioncoefficient is in-
v
NIRluminosityisnotaccountedforineitherthe general steadproportionalto λ−2, then the cavitybecomes even
pair-instability models or the artificial model of Nomoto larger. Basedonthe observedwidth of the Hα line from
et al. (2007); it therefore provides a serious challenge to SN2006gy,Smith et al.(2007)estimate the speedof the
the pair-instability SN hypothesis.10 blast wave to be 4000 km s−1, which would mean that
this material would take ∼70 yr to reach the edge of
3.2.2. Collisional Heating of the Dust the dust-free cavity. Even if there are ejecta traveling
at more typical SN velocities of 10,000 km s−1, it would
Another possibility is that the dust existed prior to
take over 28 yr for this material to reachthe edge of the
the SN explosion, at large distances from the explosion
dust-freecavity. GiventhattheNIRexcessispresent∼1
site, and was heated via collisions with the expanding
yr following explosion, and that the SN ejecta could not
reach the dust in such a short time, collisional heating
10 Thepulsationalpair-instabilitymodelofWoosley,Blinnikov,
of dust by SN ejecta cannot explain the observed NIR
&Heger(2007)hasnot,however,beenexcluded;seealsoSmithet
al.(2009b). excess.
Echoes from SN 2006gy 7
3.2.3. Newly Formed Dust emission comes from dust near the explosion site which
isheatedbytheearly-timeUV/opticalemissionfromthe
The emission features that arise in SNe IIn, as in
SN.The modelpredicts a fastrisein the IR,followedby
SN 2006gy, originate from the interaction of the SN
anextendedplateau(seebelow)beforetheIRluminosity
ejecta with a dense CSM. Shocked gas, located between
follows a rapid decline. SN 2006gy exhibits a very slow
the forward shock which is plowing into the CSM and
decline in the NIR, only a factor of ∼2 in flux over the
a reverse shock of the SN ejecta, can cool, and if the
course of ∼1 yr, consistent with being on the extended
density is large enough, it may form new dust grains.
plateau phase as predicted by the model.
The cooling time for suchdust is short(∼10–100s), but
The plateau occurs because the hottest dust, which
a persistent heat source can generate an extended pe-
exists right at the edge of the dust-free cavity described
riod of NIR emission. The physical conditions for post-
above, dominates the emission. According to the view
shockdustformationarenotnecessarilyeasytoproduce:
of a distant observer, the emitting volume is a series of
strongly interacting SNe are often X-ray sources, which
paraboloidlightfrontsthatexpandthroughoutthedust-
could prevent the formation of new dust grains in the
free cavity (see Figure 1 of Dwek 1983 or Figure 4 of
post-shockgas. Inafew cases,however,evidence forthe
Smith et al. 2008b). As a paraboloid expands it contin-
formationofdustinthepost-shockregionhasbeenfound
ually heats dust at R , which is what gives rise to the
for SNe 1998S (Pozzo et al. 2004), 2006jc (Smith, Foley, 1
plateauinthelightcurve. Afteratime∼2R /c,thever-
& Filippenko 2008),and2005ip(Smith etal. 2009a;Fox 1
tex of a given paraboloidwill reach the back edge of the
et al. 2009), and possibly also SN 2006tf (Smith et al.
cavity and will no longer be heating dust at R . Emis-
2008a). This newly formed dust is then heated by the 1
sion from the dust will be dominated by the UV/optical
energy from the shock as it continues to propagate into
radiationproduced by the SN at peak, and once this ra-
the CSM, which gives rise to the prolonged NIR excess.
diation sweeps past the back edge of the cavity the IR
This scenario is difficult to reconcile with the case of
luminosity begins to rapidly decline. Therefore, if we
SN2006gy,however. Spectratakenaroundday200show
know the duration of the plateau phase we can deter-
adeclineintheluminosityoftheHαemissionline,which
mine the size of the dust-free cavity. Constraints on the
indicatesareductionintheCSMinteractionatthistime
cavitysizecanbe determinedby assumingthatthe echo
(Smith et al. 2009b). Furthermore,as detailed by Smith
starts at our first K′ observation on day 405 and ends
et al. (2008b), the lack of X-ray,Hα, and radio emission
on our last observation on day 736. In this case we find
a little more than a year after the SN explosion implies
thatR ≈4×1017 cm, however,it wouldbe contrivedif
that the observed NIR luminosity cannot be explained 1
our observations were to perfectly bookend the plateau
by shock-heated dust. In fact, the radio nondetection
phaseoftheecho: theechoalmostcertainlystartedprior
of SN 2006gy continues through day 648 (Bietenholz &
to day 405 and continued beyond day 736. If we instead
Bartel 2008). Kawabata et al. (2009) do claim the de-
assumethat the slightNIR excess seenat day∼100(see
tection of very weak Hα emission from SN 2006gy on
day 394, but the inferred luminosity of ∼1039 erg s−1 Figure 2) is the rise of an IR echo, then we find that
R ≈ 8×1017 cm, which shows reasonable agreement
is nearly three orders of magnitude lower than the ob- 1
with the value we calculated for the dust-free cavity in
served NIR luminosity at this time. Kawabata et al.
§ 3.2.2, R ≈ 1.4×1018 cm, considering the uncertain-
(2009) also claim an R-band detection of 19.4 mag on 1
ties in the dust temperature and the likelihood that the
day 394 (though we note that Smith et al. 2008b re-
plateau extends beyond day 736.
port an upper limit of R > 20 mag, and Agnoletto et
OurNIRobservationsareonlysensitivetothewarmest
al. 2009 give an upper limit of R> 20.3 mag, at a simi-
dust, but if we assume that the NIR luminosity peaks
larepoch), which wouldconstitute an opticalluminosity
in the K′ band, then a lower limit to the total energy
similar to that observed in the NIR. The source of this
optical luminosity could potentially be the heat source emitted in the IR is EIR ≈ 2×108 L⊙ × 600 d, or &
4×1049 erg. This is comparable to the canonical optical
fornewlyformeddustatday∼400,thoughitcannotex-
output of a normal SN II. Typically, this would pose
plainthe similarNIRluminosityseenonday736. While
a significant problem for the IR echo hypothesis as it
SN 2006gy was detected in the optical at > 800 day to
wouldrequireorderunityoftheoriginalopticalradiation
have a luminosity similar to that seen in the NIR, the
from the SN to be absorbedand re-radiatedin the NIR.
veryblue nature ofopticaldata suggestthat it is a scat-
Indeed, this is precisely the argument used by Fox et al.
teredlightecho(see below)ofUV/opticalemissionfrom
(2009) to argue against the light-echo hypothesis for SN
the SN peak. The light echo must originate outside the
2005ip. In the case of SN 2006gy this is not a problem,
dust-forming region and therefore cannot heat the dust.
however, because the NIR emission falls well within the
While we cannot strongly rule out dust formationin the
availableenergybudgetto heatthe dust. After applying
post-shock region of SN 2006gy it is implausible to ex-
a bolometric correction, Smith et al. (2009b) find that
plainthelate-timeNIRluminosityvianewdust,because
∼ 2.5×1051 erg were emitted by SN 2006gy during the
there are no signs of a heating source that lasts for > 1
first ∼220 day of the light-curve evolution, which means
yr.
that less than 2% of this energy has been absorbed and
re-radiated by the dust. This value is comparable to
3.2.4. NIR Dust Echo
those found by Dwek (1983) for SNe 1979C and 1980K.
Perhapsthemostnaturalexplanationforthelate-time If we assume that the dust radiates as a blackbody,
NIR excess is a dust-heated echo, as first discussed by we can estimate the optical depth of the dust shell as
Smith et al. (2008b). The qualitative behavior of IR τ ≈EIR/(E+EIR)≈0.016. WenotethatlikeEIRabove,
echoes predicted by Dwek (1983) provides an excellent thisvalueconstitutesalowerlimitforτ. EveniftheNIR
match to the observed behavior of SN 2006gy. The IR luminosity is a factor of 10 larger than our lower limits,
8 Miller et al.
thiswouldimplyτ ≈0.1,whichcorrespondstoanoptical
extinction of AV = 1.086τ ≈ 0.1. This is far less than Keck LGS AO − day 465
the total observed extinction from the host, A ≈ Keck LGS AO − day 736
V,host
1.67 (Smith et al. 2007), suggesting that the dusty shell −1) 10 HST − day 825
s
responsiblefortheIRechocannotproduceallofthenon- −2
m
Galactic reddening observed toward SN 2006gy. That
c
there is a considerable amount of dust along the line of g
er
sight, which is not in the CSM of SN 2006gy,should not −13
come as a surprise given that there is a prominent dust 0
1
lane in NGC 1260 (see Figure 3). F (ν
With this information in hand, it is now possible to ν
estimate the totaldustmassofthe shellthatcreatesthe
NIRecho;accordingtoDwek (1983),it is simply related 1
tothedustproperties,theopticaldepthofthedust,and
4000 6000 800010000 20000
theinnerandouterradiiofthe shell. Aboveweestimate
λ (Å)
the inner radius of the dust shell to be ∼8×1017 cm. eff
The value of the outer radius can be deciphered from Fig. 6.— Late-time SED of SN 2006gyshowing evidence for
the decay of the NIR light curve following the plateau. twodistinctemissioncomponents. Weshowthefluxintheoptical
As we have not yet observed the NIR decay, we adopt (HST,day825)andNIR(KeckAO,day469and736),correctedfor
host-galaxyandMilkyWayreddening. Thesolidbluelineshowsa
an outer radius of 2R , which minimizes the dust mass
1 12,000Kblackbody,normalizedtotheF814W detection. Thered
in the shell. Finally, we follow the same assumptions as dotted line shows a single-component blackbody fit to the H and
Dwek (1983) about the dust properties: a grain density K′-banddetectionsfromday469(notethatthisfithaszerodegrees
ρ = 3 g cm−3, grain radius a = 0.1µm, and mean ab- offreedom). TheIRevolutionisslow,andthedetectionfromday
gr 736 indicates the degree of fading in the NIR. The SED shows a
sorption efficiency Q¯ν = 1. With these asumptions we rapid decline toward the red portion of the optical, as well as a
find a total dust mass of Md ≈ 0.1 M⊙. Assuming a strong rise toward the red within the NIR. These characteristics
typical value for the dust-to-gas ratio, 1:100, the total are precisely those predicted by the IR and scattered light echo
models.
mass of the circumstellar shell is ∼10 M⊙. This value
shows good agreement with the initial estimate of 5–10 rateofdecline,the SNunderwentsignificantcolorevolu-
M⊙ from Smith et al. (2008b), which was based on only tion. Assuming that F555W and F814W approximate
observationstakenbetween400and465day afterexplo- theV andI bands,respectively,wefindthatV−I =0.60
sion. We note that ∼10 M⊙ constitutes a lower limit to magonday825. By contrast,witha12000K colortem-
the total mass of the shell, because the actual value of perature at peak (Smith et al. 2009b), the observe color
the durationofthe plateauandE couldpotentially be of SN 2006gywould have been V −I ≈1.01mag. Thus,
IR
much larger than the values we adopted above. the spectrum on day 825 is bluer than the spectrum at
At a distance of ∼1018 cm from the SN explosionsite, peak. Indeed, the ∼0.4 mag color change shows excel-
it is notimmediately evidentthat the dust giving riseto lentagreementwithaλ−1 scatteringlaw. Aλ−2 scatter-
theNIRechoispartoftheprogenitor’sCSMorthelocal ing law, on the other hand, would predict a −0.83 mag
interstellar medium (ISM). Smith et al. (2008b) argue change to the V −I color, or roughly V −I ≈0.20 mag
that the large dust mass needed at ∼1018 cm could be during the late-time epochs.
explained if the progenitor passed through a phase of Thisbehaviorissimilartowhatonewouldexpectfrom
eruptive mass loss ∼1000–1500 yr prior to explosion, in ascatteredlightecho: asinthecaseofanIRecho,when
an event similar to observed outbursts from η Car. Pre- successive paraboloids sweep out to progressively larger
existing dust in the walls of a giant HII region, where radii, dust in the circumstellar (or in some cases the in-
a massive star like SN2006gy’s progenitor might have terstellar)environmentcanscatter that light towardthe
lived, cannot account for the IR echo because at typical observer, creating a plateau in the optical light curve of
distancesof∼10pcormorefromtheSN,thedustwould the SN. At the same time, scattering preferentially se-
be far too cold to reproduce the NIR excess. lects shorter wavelengths, which results in a spectrum
that is bluer than that of the SN at peak. This is the
3.3. Observed Color Evolution from HST Observations
precise behavior observed for the Type Ia SNe 1991T
Chevalier(1986)predictsthatanySNwithanIRecho (Schmidt et al. 1994), 1998bu (Cappellaro et al. 2001),
due to pre-existing CSM dust should also show a faint and2006X(Wang etal.2008)whichbothshowedsignif-
scatteredlightechointhe opticalaswell. He showsthat icant departures from the expected decline rate of Type
this dust can lead to optical emission characterizedby a aSNe. Spectra takenroughlytwo yearsafter maximum,
λ−α scatteringlaw,whereαissomevaluebetween1and for SNe 1991T and 1998bu, and ∼10 months after max-
2. MorethanayearafterexplosionSN2006gyhadfaded imum for SN 2006X, show emission features similar to
rapidly in the optical (Smith et al. 2008b; Agnoletto et thoseseeninthe spectranearpeaksuperposedontopof
al. 2009; Kawabata et al. 2009). Therefore, we obtained a blue continuum for both these SNe. Light echoes have
HST images to search for a possible late-time scattered alsobeenobservedaroundanumberofcore-collapseSNe,
light echo. e.g. SN 2003gd (Sugerman 2005; Van Dyk, Li, & Filip-
Our optical detection on day 825 shows a significant penko 2006) and SN 1993J (Sugerman & Crotts 2002),
changeintheopticaldeclinerateofSN2006gy. Between thoughwenotethatlate-timespectraofthe lightechoes
day 200–400 the SN faded by & 3 mag in the optical do not exist in these cases.
(Smith et al. 2008b), whereas the decline between day In Figure 6 we show the late-time SED of SN 2006gy,
400–825 was only ∼1 mag. In addition to this reduced including our NIR detections on day 469 and 736 and
Echoes from SN 2006gy 9
locations.
With only a single epoch of the unresolved scattered
light echo, a precise determination of the location of the
scatteringdustcannotbeobtained. Nevertheless,wecan
gain some insight into the dust location based on the
total optical and NIR emission. For the case where the
NIR-echo dust also gives rise to the scattered light echo,
and assuming the SN can be modeled as a single short
pulse of UV/optical radiation, the ratio of the scattered
light luminosity (L ) to the IR luminosity at late-times
s
can be shown to be (Chevalier 1986):
L ω (ct)2 (1−g)2
s
= 1− (2)
L 1−ω (cid:18) 4R2 (cid:19) 2g2
IR 2
(1−g)2+2g (1−g)2+gct/R
2
× − ,
(cid:26)[(1−g)2+4g]1/2 [(1−g)2+2gct/R ]1/2(cid:27)
2
whereω isthedustalbedo,cisthespeedoflight,tisthe
timesincetheSNexplosion,R isouterradiusofthedust
2
shell,andg isameasureofthedegreeofforwardscatter-
Fig. 7.—Evidenceforascatteredlightechofromtheblueoptical ing. The case g = 0 corresponds to isotropic scattering
emissionfromtheSN.TheHST/WFPC2photometryofSN2006gy
and g = 1 is completely forward scattering. Note that
(blackdots)arecomparedtoobservedandreflected spectra, both
normalized to the HST F814W flux density. The gray lineis the in Equation 2 we have assumed that ct/2 < R1, where
observed spectrum of SN 2006gy on day 71 during its peak lumi- R is the inner radius of the dust shell, and that the
1
nosity phase, from Smith et al. (2009b). We plot the observed albedo is independent of frequency. Empirical estimates
wavelength (i.e., not corrected forthe SN redshift), and the spec-
and numerical calculations (White 1979) show g ≈ 0.6.
trum has no correction forreddening. Theblack line(and dotted
linesabove andbelow) show the samespectrum, butadopting an For the dust albedo we adopt ω ≈ 0.6 (Mathis, Rumpl,
assumed wavelength dependence for the dust reflectance propor- &Nordsieck1977). Asmentionedabove,wecannotcon-
tionaltoλ−1.2(±0.15). strainR withourcurrentobservations,butwedoknow
2
the optical detections from day 825. The solid blue line thatR mustbegreaterthanR ,andthevalueofEqua-
2 1
represents a single-component blackbody at T =12,000 tion 2 does not change when R → ∞. For reasonable
2
K, roughly the peak temperature of SN 2006gy (Smith limitstotheouterradius,1.5R <R <∞,wefindthat
1 2
et al. 2009b), normalized to our F814W detection. The L /L ≈0.88–0.60.
s IR
SED clearly shows that the late-time emission is signifi- Direct integration of the HST optical flux yields an
cantly bluer than a blackbody spectrum. We note that optical luminosity of ∼(7.4±0.5)×107 L⊙. This value is
the early-time spectra of SN 2006gy show considerable somewhatuncertainbecauseourdatadonotextendinto
lineblanketingbluewardof∼4000˚A,suchthattheemis- the ultraviolet;however,opticalspectratakennearpeak
sion decreases blueward of our F450W detection. The show that line blanketing severely reduces the flux blue-
NIRemissionrapidlyrisestowardtheIR,suggestingthat wardof∼4000˚A(Smithetal.2009b),sothisuncertainty
thereis considerableIRemissionto whichourNIRmea- shouldnothaveasignificanteffectonthetotalscattered
surements are not sensitive. light luminosity. If we assume that the NIR flux contin-
Figure 7 illustrates the observed HST photometry uesto decline atthe samerate asthatobservedbetween
along with the spectrum of SN 2006gy obtained near day405and736,then wefind a lowerlimit to the IRlu-
peak (Smith et al. 2009b). Note that no correction minosity(againbecauseourNIRobservationsonlyprobe
for reddening has been made. The photometry shows emission from the hottest dust) of LIR > 8.6×107 L⊙
excellent agreement with a λ−α scattering law, where on day 825. Therefore, on day 825, L /L <0.86, con-
s IR
α = 1.2±0.15. The lack of an actual reflection spec- sistentwiththepredictedvaluesaboveinthecasewhere
trum precludes a more detailed model of the scattering the scattering dust is the same dust responsible for the
dust at this time. Given that the other typical scenarios IR echo.
for a large late-time luminosity from a SN, 56Co decay The alternative possibility is that the scattering dust
and extended CSM interaction, have been ruled out, we isnotlocatedintheCSM,insteadexistingatsomeother
conclude that the late-time HST detection is a scattered locationinNGC 1260alongthe lineofsight. Cappellaro
light echo. et al. (2001) inferred that this was the case with SNe
1991Tand1998bu,andlikethosetwoSNeweknowthat
3.4. Location of the Scattering Dust SN2006gyhasasignificantamountofdustlocatedalong
thelineofsight. AsshownbyCappellaroetal.(2001),if
The location of the scattering dust is of interest as it
we approximate the SN light curve as a shortpulse with
canhelpdeterminepropertiesoftheenvironmentlocalto
duration ∆t , then the scattered echo luminosity is
SN2006gy. Aboveweshowthatthe CSMdustresponsi- SN
blefortheNIRechocannotaccountforallthereddening L (t)=L ∆t f(t), (3)
echo SN SN
observedtowardSN2006gy,suggestingthattheSNlight
where L ∆t can be obtained via direct integration
passes through external, non-CSM dusty regions in the SN SN
+∞
host galaxy prior to reaching us. Therefore, the scat- of the observed light curve, LSN∆tSN = 0 LSN(t)dt,
tered light echo could originate in a number of different andf(t)isthefractionoflightscatteredtRotheobserver.
10 Miller et al.
Under the assumption that light is being scattered by SN models. Rather, the NIR observations can best be
dust in a sheet of thickness ∆D ≪ D, where D is the explained with an IR echo, and the optical detections
distance between the dust and the SN, then (Chevalier are most compatible with a scattered light echo. Both
1986; Cappellaro et al. 2001) IR and scattered light echoes should be present if there
is a massive dust shell external to the SN (Dwek 1983;
c ωτ 1−g2 Chevalier 1986).
f(t)= , (4) At the time of discovery SN 2006gy was the most
8πD+ct{1+g2−2g[D/(D+ct)]}3/2 luminous SN known (Ofek et al. 2007; Smith et al.
where τ is the optical depth of the dust. Using the rela- 2007), which when combined with its proximity to the
tionA =1.086τ (Cox2000),andthefactthatA out- host-galaxy nucleus (∼300 pc), suggested that perhaps
V V V
side the dusty CSM is ∼1.6 mag, we find τ ≈1.5. From there was something unique about star formation and
Smithetal.(2009b)weknowthatL ∆t ≈2.5×1051 SNe in that region. Since the discovery of SN 2006gy
SN SN
erg. Therefore, using L ≈2.9×1041 erg s−1, we find several other VLSNe have been found, including SNe
echo
(from Equation 4) that dust located at D ≈ 20 pc can 2005ap(Quimbyetal.2007),2006tf(Smithetal.2008a),
explain the observed scattered light luminosity. 2008am(Yuanetal.2008),and2008es(Milleretal.2009;
Our observations appear to be consistent with one of Gezarietal.2009). IneachofthesecasestheSNhasbeen
two scenarios: either (i) the scattered light echo is due localized well outside the central kpc of the host galaxy,
to CSM dust, which has also been heated and is radiat- with the exception of SN 2008es, for which a definitive
ing in the IR, or (ii) a dusty region ∼20 pc from the SN hosthasnotyetbeenidentified(Milleretal.2009;Gezari
is responsible for the late-time optical luminosity. This etal.2009). ThissuggeststhatthelocationofSN2006gy
latter scenario seems very plausible if the progenitor of was not in fact special. Current and future surveys will
SN 2006gywas a very massive star, & 100 M⊙(Smith et dramatically increase the rate of discovery of VLSNe,
al. 2009b). If this were the case we might expect that whichshould,in turn,allowfor a moresystematic study
the progenitorresides in the center of a giantHII region of their host environments (for a discussion of the rates
with multiple dense, dusty clouds nearby. This scenario of VLSNe see Miller et al. 2009).
mightbeverysimilartotheCarinanebulasurroundingη Given our interpretation for the late-time optical and
Car,whichhasmultiple dustymolecularcloudsonly10– NIR emission arising from a dust shell at ∼8×1017 cm
20 pc from η Car (e.g. Smith & Brooks 2007). Perhaps fromtheSN,whatmightweexpectfromfuture observa-
given a similar progenitor birth environment, evidence tionsofSN2006gy? TheIR-echomodelsofDwek(1983)
for patchy dust at the tens of pc scale around GRB en- showthatafteratime t≈2R1/c,theIRluminosityfalls
vironments also appears to be secure (e.g. Perley et al. offthe plateauandexhibitssignificantdecline. Basedon
2009). If the progenitor of SN 2006gy were surrounded ourpredictedvalueforthesizeofthedust-freecavity(see
by such a nebula then our assumption of a single, thin §3.2.2),wewouldexpectthattheplateauphaseoftheIR
scatteringsurfacewouldnolongerbevalid,andthescat- echohasended,andthattheIRfluxisnowinsteadyde-
tering dust could potentially be located at a number of cline. Futureobservationsofthisdeclinewillplacelimits
different locations. on the outer extent of the dust shell, which will in turn
With only a single epoch of observations, we cannot provide better limits on the total mass of the dust shell.
distinguishbetweenthetwopossibledustlocations. One Our ability to predict the behavior of the scatteredlight
additional epoch of optical imaging should prove suffi- echois limited by the currentdegeneracyconcerningthe
cient to determine which scenario is correct. As long as physicallocationofthescatteringdust. Ifthe samedust
D ≫ct,thenf(t)isroughlyconstant,whereasthemod- is responsibleforboth the IR andscatteredlightechoes,
els of Chevalier (1986) show a continuous decline in the then we would expect that the optical emission is now
scattered light echoes when g = 0.6 and the scattering fading, in a manner similar to what ought to be seen in
dust is in the CSM. A relatively flat optical light curve the NIR. Our observations are also consistent with dust
wouldputthedustat∼20pc,whereassignificantdecline located at ∼20 pc from the SN, in which case we might
fromtheobservedday825fluxwouldsuggestCSMdust. expect the scattered light echo to continue at roughly a
Finally,wenotethattheprospectsforfullyresolvingthe constant flux for several more years. Continued obser-
SN2006gylightechoaredim: atadistanceof73Mpc it vations with HST would allow us to distinguish between
wouldtakeseveraldecadesbeforethelightechocouldbe these two situations.
resolvedwithhighspatialresolutionimages(i.e.,compa- Theshell-shockmodelofSmith&McCray(2007)may
rable to HST in the optical). alsoexplainthepeakluminosityofseveralotherVLSNe:
2005ap, 2006tf, 2008es (Quimby et al. 2007; Smith et
4. CONCLUSIONSANDFUTUREWORK al. 2008a; Miller et al. 2009; Smith et al. 2009b). This
We have presentednew observationsof SN 2006gy,in- modelrequireslargeCSMdensitiesatadistanceofafew
cluding early-time NIR data from PAIRITEL, as well as ×1015 cm, which can be accomplished if the SN progen-
optical and NIR detections more than two years after itor undergoes significant (∼0.1–1 M⊙ yr−1) mass loss
theexplosion. TheearlyNIRdatashowgoodagreement during the few decades prior to explosion (Smith et al.
withasingle-component,coolingblackbodybetweenday 2007). The most luminous LBVs have observed mass
54 and 133. There is tentative evidence for a slight NIR shells with &10 M⊙ (Smith & Owocki 2006 and refer-
excess towardthe end of this period. The late-time NIR encestherein), indicativeof gianteruptions (Humphreys
and optical detections show a rate of decline which is & Davidson 1994) but not LBVs in their more typical
slower than that of 56Co, indicating that the luminosity state with less violent wind variability (see Smith, Vink,
is not generated via heating by radioactive decay. This & de Koter 2004 for a general reference on LBV winds).
slowrateofdeclineisnotexpectedinthepair-instability The possible connection between large CSM densities