Draftversion January31,2010 PreprinttypesetusingLATEXstyleemulateapjv.11/10/09 DISCOVERY OF PRECURSOR LBV OUTBURSTS IN TWO RECENT OPTICAL TRANSIENTS: THE FITFULLY VARIABLE MISSING LINKS UGC 2773-OTAND SN 2009ip Nathan Smith1,2, Adam Miller1, Weidong Li1, Alexei V. Filippenko1, Jeffrey M. Silverman1, Andrew W. Howard1, Peter Nugent3, Geoffrey W. Marcy1, Joshua S. Bloom1, Andrea M. Ghez4, Jessica Lu4, Sylvana Yelda4, Rebecca A. Bernstein5, & Janet E. Colucci5 Draft version January 31, 2010 0 ABSTRACT 1 Wepresentprogenitor-stardetections,lightcurves,andopticalspectraofsupernova(SN)2009ipand 0 the 2009 optical transient in UGC 2773 (U2773-OT), which were not genuine supernovae. Precursor 2 variabilityinthedecadebeforeoutburstindicatesthatbothoftheprogenitorstarswereluminousblue n variables(LBVs). Their pre-outburstlightcurvesresemble the S Doradusphases that precededgiant a eruptions of the prototypical LBVs η Carinae and SN 1954J (V12 in NGC 2403), with intermediate J progenitor luminosities. Hubble Space Telescope detections a decade before discovery indicate that 1 the SN 2009ip and U2773-OT progenitors were supergiants with likely initial masses of 50–80 M⊙ 3 and &20 M⊙, respectively. Both outbursts had spectra befitting known LBVs, although in different physical states. SN 2009ip exhibited a hot LBV spectrum with characteristic speeds of 550 km s−1, ] plus evidence for faster material up to 5000 km s−1, resembling the slow Homunculus and fast blast R wave of η Carinae. In contrast, U2773-OT shows a forest of narrow absorption and emission lines S comparable to that of S Dor in its cool state, plus [Ca ii] emission and an infrared excess indicative . of dust, similar to SN 2008S and the 2008 optical transient in NGC 300 (N300-OT). The [Ca ii] h emission is probably tied to a dusty pre-outburst environment, and is not a distinguishing property p of the outburst mechanism. The LBV nature of SN 2009ip and U2773-OT may provide a critical - o link between historical LBV eruptions, while U2773-OT may provide a link between LBVs and the r unusualdust-obscuredtransientsSN2008SandN300-OT.Futuresearcheswilluncovermoreexamples t s of precursor LBV variability of this kind, providing key clues that may help unravel the instability a driving LBV eruptions in massive stars. [ Subjectheadings: circumstellarmatter—stars: evolution—stars: massloss—stars: variables: other 2 — stars: winds, outflows — supernovae: general v 2 9 7 4 1Department of Astronomy, University of California, Berkeley, CA94720-3411. 9. 2email: [email protected]. 0 3Lawrence Berkeley National Laboratory, 1 Cyclotron Rd., Berkeley,CA94720. 9 4DivisionofAstronomyandAstrophysics,UniversityofCalifor- 0 nia,LosAngeles,CA90095-1547. : 5DepartmentofAstronomyandAstrophysics,1156HighStreet, v UCO/LickObservatory, Universityof California,Santa Cruz, CA Xi 95064. r a 2 Smith et al. Fig.1.—FinderchartsshowingthepositionsofSN2009ipandU2773-OTwithrespecttotheirhostgalaxies. Thesmallsquaresindicate the fields of view for the HST/WFPC2 F606W images in Figure 2. These are unfiltered images obtained with the Katzman Automatic ImagingTelescope(KAIT)atLickObservatory. SN 2009ip UGC 2773 OT N E N E Fig.2.— The sites of SN 2009ip and U2773-OT in the HST/WFPC2 F606W images. Each stamp is 10′′×10′′. The location of each transientasderivedfromhigh-resolutionground-basedimagesismarkedwithacirclehavingaradiusthatis15timesthe1σ uncertainty radiusoftheastrometricsolution. Agoodcandidate progenitorisidentifiedforbothtransients. Two LBV Eruptions and their Progenitors 3 1. INTRODUCTION Berger et al. 2009a; Bond et al. 2009). In that case, the initial mass implied by the progenitor’sinfrared (IR) lu- Among the objects discovered in the course of hunt- ing for supernovae (SNe) are transient sources that are minositywouldbecloserto15–20M⊙ forN300-OT,and would therefore be in better agreement with the find- fainter and have slower expansion speeds than most core-collapse SNe. Following the historical examples ings of Gogarten et al. (2009) than an 8–10 M⊙ EAGB star. The somewhat lower luminosity of the progenitor (see Humphreys et al. 1999, and references therein) of the 19th century eruption of η Carinae, the 17th cen- of SN 2008S would imply 12–15 M⊙ under the same as- sumption (Smith et al. 2009). tury eruption of P Cygni, SN 1954J (Variable 12 in In this paper we discuss another pair of newly discov- NGC 2403), and SN 1961V, these transients have usu- ered transients with identified progenitors, the 2009 op- ally been associated with nonterminal eruptions of lu- tical transient in UGC 2773 (hereafter U2773-OT) and minous blue variables (LBVs), rather than final explo- SN2009ip(see Fig.1). U2773-OToccurredinthe dwarf sions that mark the deaths of massive stars. LBVs rep- irregular galaxy UGC 2773, and was discovered (Boles resent a highly unstable, rapid mass-loss phase in the 2009)on2009Aug.18.08(UTdatesareusedthroughout late evolution of massive stars (see Smith & Owocki thispaper). SN2009ipwasdiscovered(Mazaetal.2009) 2006; Smith et al. 2004; Humphreys & Davidson 1994), on 2009 Aug. 26.11 in the Sb galaxy NGC 7259. Both which is still poorly understood. P Cygni and η Car re- objects had discovery absolute magnitudes fainter than main as luminous well-studied blue supergiants,and the −14 mag, and they had spectra with narrow H emission survivor of SN 1954J has been identified as a luminous lines (Berger et al. 2009b; Berger & Foley 2009). These dust-enshrouded star (Smith et al. 2001; Van Dyk et al. transient outbursts, still currently ongoing, share prop- 2005). SN1961Visamorecontroversialcase(Chuetal. erties in common with giant LBV eruptions as well as 2004),butthereareindicationsofasurvivingstaraswell SN 2008S and N300-OT. In these new cases, however, (Goodrichetal.1989;Filippenkoetal.1995;VanDyket theprogenitorsarenotasheavilyobscuredbydust. The al. 2002). Because the massive stars are thought to sur- progenitorsaredetectedatopticalwavelengths,andthey vive the events, recent examples of these η Car analogs exhibitpre-eruptionvariabilitythatismatchedbyclassic such as SN 1997bs (Van Dyk et al. 2002) have earned LBVs such as V12 and η Car. We reported the prelim- the label “SN impostors.” For most recent extragalactic inary pre-outburst detections of SN 2009ip in Miller et examples, however, available evidence that the star has al. (2009), while our photometry of U2773-OT is doc- survived remains inconclusive. umented here for the first time. Based on their LBV- Interest in and interpretation of these transients was like pre-outburstvariability,the two new transients help stirredbythesurprisingdiscoverythattworecentevents, bridge the gap between classical LBV eruptions and ob- SN 2008S and the 2008 optical transient in NGC 300 jects similar to SN 2008Sand N300-OT.Other than the (hereafterN300-OT),bothhadrelativelylow-luminosity, historical cases of SN 1954J and SN 1961V, this is the dust-enshroudedprogenitorstars(Prietoetal.2008;Pri- firstdiscoveryof extended 5–10yr pre-eruptionvariabil- eto 2008). Although the outburst properties resembled ity before an extragalactic LBV-like transient. those of known LBV eruptions, interpretation of their dusty progenitors (Prieto et al. 2008; Thompson et al. 2009)impliedinitialmassesbelowtheusuallyrecognized 2. OBSERVATIONS initial-mass range for LBVs extending down to ∼20 M⊙ 2.1. Pinpointing the Progenitors (Smith et al. 2004). This fueled a range of speculation BothSN2009ipandU2773-OThadobservationstaken that these transients might be similar eruptive phenom- ∼10 yr prior to discovery with the Hubble Space Tele- ena extending to somewhat lower mass and cooler stars scope/Wide Field Planetary Camera 2 (HST/WFPC2), (Smithetal.2009;Bergeretal.2009a;Bondetal.2009), which we retrieved from the archive and analyzed. the eruptive birth of a white dwarf and planetary neb- NGC 7259 (SN 2009ip) was observed in the F606W fil- ula in stars with initial masses below 8 M⊙ (Thompson ter on 1999 Jun. 29, and UGC 2773 was observedin the et al. 2009), electron-capture SNe (ecSNe) in extreme F606W and F814W filters on 1999 Aug. 14. asymptotic giantbranch(EAGB) starsaround8–10M⊙ To pinpoint the precise location of the two tran- (Thompsonetal.2009;Botticellaetal.2009),faintcore- sients’progenitorsintheHSTimages,weobtainedhigh- collapse SNe (see Pastorelloet al. 2007 in regardto pre- resolutionground-basedimagesforcomparison. On2009 viousevents),ormergers/mass-transfereventsrelatedto Sep. 9, we observed SN 2009ip in the K′ band with the some other recent transients (e.g., Kulkarni et al. 2007; Near-Infrared Camera 2 (NIRC2) using the laser guide Kashi et al. 2009). Interpretation of these objects re- star (LGS) adaptive optics (AO) system (Wizinowich et mains controversialand puzzling. al. 2006) on the 10-m Keck II telescope. Three mosaic Gogarten et al. (2009) found a likely initial mass of pointings, each with three exposures of 4×15 s, were 12–25M⊙ fortheN300-OTprogenitorbasedonthestar- combined to yield a final stacked image with 9 min to- formation history of its local neighborhood, challenging tal exposure time, a pixel scale of 0′.′04 pixel−1, and a the ecSN or white-dwarf birth hypotheses. Deriving a field of view of 41′′ × 41′′. For U2773-OT, we took a likelyinitialmass around10–12M⊙ fromthe luminosity 20 s guider image with the high-resolution echelle spec- (e.g., Prieto et al. 2008) depends upon the assumption trometer (HIRES; Vogt et al. 1994) on the 10-m Keck I that the progenitor was a cool EAGB star at the up- telescope on 2009 Sep. 9. The image has a pixel scale of permost tip of its AGB. It remains possible, however, 0′.′3 pixel−1 and a field of view of 43′′×58′′. that the progenitors of SN 2008S and N300-OT were To perform astrometric solutions between the ground- relatively blue stars that were heavily obscured by cir- based and HST images, we adopted the technique de- cumstellar dust (Prieto et al. 2008; Smith et al. 2009; tailed by Li et al. (2007) using stars present in both the 4 Smith et al. TABLE 1 TABLE 2 Photometry of SN2009ip Photometry ofU2773-OT MJD filter mag 1σ datasource MJD filter mag 1σ datasource 51358.50 F606W 21.8 0.2 HST 51404.1 F606W 22.83 0.03 HST 53195.47 R >21.17 ... DS 51404.1 F814W 22.22 0.05 HST 53226.33 R >21.43 ... DS 51853.5 unfiltered >20.80 ... KAIT(stack) 53233.25 R >20.97 ... DS 52225.5 unfiltered >21.10 ... KAIT(stack) 53251.23 R >21.01 ... DS 52589.5 unfiltered >21.00 ... KAIT(stack) 53259.21 R >20.65 ... DS 52939.5 unfiltered >20.60 ... KAIT(stack) 53541.44 R 20.61 0.14 DS 53683.0 unfiltered 19.91 0.42 KAIT(stack) 53554.46 R 20.37 0.16 DS 54042.5 unfiltered 19.16 0.07 KAIT(stack) 54305.39 R 20.97 0.22 DS 54394.9 unfiltered 18.93 0.10 KAIT(stack) 54323.35 R 20.92 0.25 DS 54772.1 unfiltered 18.73 0.06 KAIT(stack) 54701.33 R >21.51 ... DS 55051.5 unfiltered 17.96 0.10 KAIT 54702.33 R >21.33 ... DS 55061.5 unfiltered 17.70 0.19 KAIT 55043.50 unfiltered 18.5 0.4 CBET1928 55072.5 unfiltered 17.91 0.06 KAIT 55069.61 unfiltered 17.9 0.3 CBET1928 55077.5 unfiltered 17.76 0.05 KAIT 55071.75 unfiltered 17.0 0.3 CBET1928 55079.5 unfiltered 17.76 0.03 KAIT 55073.50 R 18.2 0.1 KAIT 55081.5 unfiltered 17.82 0.04 KAIT 55085.10 unfiltered 20.2 0.2 guider 55082.5 unfiltered 17.76 0.03 KAIT 55096.50 unfiltered 18.3 0.2 guider 55083.5 unfiltered 17.79 0.03 KAIT 55097.50 R 18.3 0.2 KAIT 55084.5 unfiltered 17.61 0.12 KAIT 55090.5 unfiltered 17.74 0.03 KAIT 55091.5 unfiltered 17.73 0.03 KAIT ground-based and HST images. Due to the small field 55092.5 unfiltered 17.75 0.03 KAIT of view of the ground-based images, we are only able to 55094.5 unfiltered 17.75 0.03 KAIT measurethe positionsforfive starsineachfield. The as- 55095.5 unfiltered 17.72 0.03 KAIT trometricsolutionsusingIRAF/GEOMAPyieldapreci- 55096.5 unfiltered 17.68 0.03 KAIT 55097.5 unfiltered 17.71 0.03 KAIT sionof36masand21masforSN2009ipandU2773-OT, 55098.5 unfiltered 17.67 0.03 KAIT respectively6. The positions of the transients are then 55099.5 unfiltered 17.67 0.03 KAIT mappedontotheHSTimages. Duetothesmallavailable 55100.5 unfiltered 17.74 0.03 KAIT 55102.5 unfiltered 17.71 0.03 KAIT number of stars, we only use second-order polynominals 55105.5 unfiltered 17.67 0.04 KAIT in the astrometic transformations. We also conducted 55110.5 unfiltered 17.69 0.05 KAIT an additional error analysis by taking out one star and 55113.5 unfiltered 17.71 0.03 KAIT leaving four with which to perform the solution, and re- 55120.5 unfiltered 17.65 0.03 KAIT 55123.5 unfiltered 17.59 0.03 KAIT peatingthisforallstars. Forbothtransients,theaverage 55126.5 unfiltered 17.62 0.03 KAIT positionfromthefiveseparatemeasurementsandtheas- 55129.5 unfiltered 17.57 0.03 KAIT sociated uncertainty is fully consistent with the position 55132.5 unfiltered 17.55 0.03 KAIT measured from the solution using all five stars together, 55133.5 unfiltered 17.57 0.03 KAIT 55136.5 unfiltered 17.67 0.04 KAIT so our solutions appear to be stable despite the small 55149.5 unfiltered 17.52 0.03 KAIT number of stars involved. 55152.5 unfiltered 17.57 0.03 KAIT Figure 2 shows a 10′′×10′′ region of the sites for the 55155.5 unfiltered 17.65 0.03 KAIT transientsintheHST/WFPC2images. Acandidatepro- 55158.5 unfiltered 17.60 0.03 KAIT 55161.5 unfiltered 17.72 0.03 KAIT genitorisdetectedforbothtransientswithin1σprecision 55164.5 unfiltered 17.56 0.03 KAIT of the astrometric solution. We therefore confirm the 55169.5 unfiltered 17.66 0.03 KAIT candidateprogenitorsmarkedinFigure2, firstproposed 55173.5 unfiltered 17.60 0.05 KAIT for SN 2009ip by our group (Miller et al. 2009) and for U2773-OTbyBerger&Foley(2009). TheHSTphotom- etryfortheprogenitorsismeasuredwithHSTphot(Dol- phin 2000a, 2000b) and listed in Tables 1 and 2. HST- phot also reports that both objects have a stellar point- TABLE 3 PAIRITEL Photometry of U2773-OT spread function (PSF): they belong to type “1” (good star), with a sharpness well within the range of ± 0.3 as MJD J (mag) H (mag) K (mag) suggested by Leonard et al. (2008). The coordinates of the two transients, as measured directly from the World 55082.50 15.57±0.09 14.64±0.15 14.97±0.25 55090.36 15.99±0.13 15.07±0.12 15.19±0.16 CoordinateSystem(WCS) in the WFPC2FITS images, 55091.40 15.89±0.15 14.95±0.16 14.87±0.19 areα =3h32m07s.22,δ =47◦47′39′.′4forU2773- 55092.36 15.99±0.16 15.13±0.15 15.1±0.2 OTaJn2d00α0 =22h23m08Js.22000,0δ =−28◦56′52′.′6for 55095.35 15.82±0.14 14.84±0.09 14.88±0.24 J2000 J2000 55096.41 15.97±0.13 15.06±0.11 14.94±0.23 SN 2009ip. 55099.42 15.85±0.17 14.93±0.12 14.98±0.14 55102.43 15.87±0.13 14.89±0.14 14.99±0.09 2.2. Pre-Eruption Photometry 55113.41 15.96±0.21 15.04±0.18 14.91±0.17 55119.41 15.69±0.14 14.84±0.15 14.78±0.18 In Figure 3 we plot absolute magnitudes derived from 55122.43 15.8±0.14 14.84±0.11 14.8±0.19 the photometry listed in Tables 1 and 2. For SN 2009ip 55125.43 15.75±0.12 15.18±0.15 15.11±0.18 55180.13 15.55±0.1 14.71±0.13 14.68±0.14 6 Althoughtheground-basedAOimageofSN2009iphasmuch 55182.07 15.47±0.12 14.79±0.11 14.61±0.13 higherresolutionthantheU2773-OTimage,theastrometricsolu- tionislessprecisebecauseofthefaintnessofstarsintheimage. Two LBV Eruptions and their Progenitors 5 TABLE 4 Spectroscopic Observations Date Tel./Inst. Target day Exp. (s) λ(˚A) comment 2009Sep.10 Keck/HIRES U2773-OT 22 1800 6543-7990 onlyredlinesdetected 2009Sep.11 Keck/HIRES SN2009ip 16 1200 6543-7990 onlyHαdetected 2009Sep.21 Keck/LRIS U2773-OT 34 300 3800-5100 med. resolution 2009Sep.21 Keck/LRIS U2773-OT 34 300 6250-7850 med. resolution 2009Sep.21 Keck/LRIS U2773-OT 34 300 3600-9200 lowresolution 2009Sep.21 Keck/LRIS SN2009ip 25 780 3800-5100 med. resolution 2009Sep.21 Keck/LRIS SN2009ip 25 640 6250-7850 med. resolution 2009Sep.21 Keck/LRIS SN2009ip 25 300 3600-9200 lowresolution Fig.3.— Our absolute-magnitude light curves of the LBV-like transients SN 2009ip and U2773-OT (Tables 1 and 2) compared to the historicalLBVsη Carinaeinits19thcentury “GreatEruption”(compiledbyFrew2004) andSN1954J (V12inNGC2403; Tammann& Sandage 1968). The orange curveis the LBVeruption of V1inNGC 2363 (Drissenetal. 2001; Petit et al.2006), although shiftedto an arbitrarydate. WealsoincludeR-bandlightcurvesoftherecenttransients N300-OT(fromBondetal.2009)andSN2008S(fromSmith etal.2009), along withtheirpre-outburst bolometricmagnitudes derived fromSpitzer/IRAC data(Prieto etal.2008; Prieto2008; Bond et al. 2009). For U2773-OT, we show both upper limits or detections fromindividual exposures (arrows and error bars, respectively), as well as upper limitsand detections instacked seasonal data (blue stars). The earliestdetections of both objects werefrom archival HST F606Wdata,correctedappropriatelyforGalacticextinction. in NGC 7259,we adopta distance modulus of m−M = filter that has a blue cutoff at λ≈ 6100 ˚A, and are oth- 31.55 mag, and a Galactic reddening and extinction of erwisesimilartounfilteredphotometry,whichwetaketo E(B−V)=0.019magandAR =0.05mag,respectively. be comparable to R-band (see below). DS photometry Similarly, for U2773-OT we adopt m−M =28.82 mag, of SN 2009ip was calibrated relative to the USNO-B1.0 E(B −V) = 0.56 mag, and AR = 1.51 mag. The ob- redmagnitudes,withtypicaluncertaintiesof0.1–0.2mag servationsaredescribedbelow. The lightcurvesarealso when several USNO stars are used. shown in Figure 4, which focuses on the time around ThefullhistoricalDSlightcurveofSN2009ipisshown peak luminosity. inFigure3,while thephotometryisreportedinTable1. ThefieldofSN2009ipwasimagedmultiple times dur- Figures 3 and 4 also include data around the time of ing the operations of the Palomar-Quest survey, and discovery (Maza et al. 2009) and our own R-band pho- those observations have been reprocessed as part of the tometry obtained with the 0.76-m Katzman Automatic DeepSky project7 (DS; Nugent 2009). As first noted by Imaging Telescope (KAIT; Filippenko et al. 2001; Filip- Milleretal.(2009),asourceatthelocationofSN2009ip penko2003)atLickObservatory(seebelow). TheKAIT was observed in 2005,4 yr prior to discovery,and it was R-band photometry is calibrated relative to the USNO- atcomparableluminosityin2007. DStypicallyhasmore B1.0 red magnitudes. It is unclear how the unfiltered than one image on any given night when a field was ob- discoverymagnitudes reportedbyMazaetal.(2009)are served,sowestackallDSimagestakenonthesamenight calibrated,sowegivelargeuncertaintiesforthesepartic- to improvethe limiting magnitude for eachepoch. Deep ular measurements. Moreover, as discussed by Li et al. Sky images were obtained through a nonstandard red (2003), unfiltered data are often clearly matched to the broad R band. A source at the position of SN 2009ip is 7 http://supernova.lbl.gov/˜nugent/deepsky.html. seenintheHST/WFPC2imagetaken∼10yrbeforedis- 6 Smith et al. covery,withmF606W =21.8±0.2mag. Atthedistanceof locationsinthehostgalaxythathadasurfacebrightness NGC 7259this implies MV ≈−9.8mag. The earlyHST similartothatofthetransient’slocation. Wesubtracted detection, if this is the quiescent progenitor star, thus the 2MASS image from the images with fake stars, and requires a high luminosity of at least log(L/L⊙) ≈ 5.9 measured the scatter in the fake-star flux to determine (higherforanonzerobolometriccorrection),andimplies the uncertainty in the flux of the transient. The final ahighinitialmassof50–80M⊙ (e.g.,Lejeune&Schaerer JHKs photometry is listed in Table 3 and is plotted in 2001). Figure 5. During the time of our observations,the near- UGC 2773, the dwarf irregular host galaxy of U2773- IR flux from U2773-OThas shown a slight increase over OT,ismonitoredregularlywithKAIT.Weanalyzedpre- ∼100 days, commensurate with the slow brightening in discoveryunfiltered(approximatelyR band)imagesand optical photometry, and with relatively constant color. detected a source at the position of U2773-OT during UsingthestackedK′ imageofSN2009ip(§2.1),weat- the ∼5 yr before discovery, as well as upper limits be- temptedtomeasureits K′ magnitude. InadeepPAIRI- fore that. There were multiple observations each year, TELimage, inwhich we do notdetect SN 2009ip,we do soweproducedstackedseasonalaveragestoimprovethe detect a star located at α = 22h23m08s.503, δ J2000 J2000 sensitivity. A stacked image from the year 2000, when =−28◦56′47′.′75, which is common to both the PAIRI- U2773-OT was not detected, is used as a template im- TEL and AO images. Calibrating relative to 2MASS we age in animage-subtractiontechnique to cleanly remove measureKs = 15.93± 0.12 mag for this star. Assuming thegalaxycontaminationatthepositionofU2773-OTin Ks ≈ K′ (the K′ and Ks filters are approximately the later images. The resulting upper limits and detections same), we therefore measure K′ = 19.68±0.12 mag for are listed in Table 2, while these averaged data as well SN 2009ip on 2009 Sep. 9. as individual epochs are shown in Figure 3. As noted above, U2773-OT was also detected in 2.4. Spectroscopy archival HST/WFPC2 images about 10 yr before dis- A summary of our spectroscopic observations is listed covery, and these HST magnitudes are listed in Table 2 in Table 4. We obtained high-resolution (R = λ/∆λ ≈ as well. F606W is not a standard V-band filter, so we 60,000)opticalechellespectraofU2773-OTon2009Sep. used SYNPHOT to convert it in order to interpret the 10 using HIRES (Vogt et al. 2004) on the 10-m Keck I F606W−F814W color. Using SYNPHOT and adopting telescope. The HIRES spectra were reduced using stan- the Galactic reddening of E(B−V) = 0.564 mag along dard procedures. These observations correspond to day the line of sight to U2773, the object has MV ≈ −7.8 22after discoveryforU2773-OT.We usedthe B5decker mag, and an intrinsic V − I color of ∼0.09 mag, or (0′.′86 slit width) and a total exposure time of 1800 s. less if there was additional circumstellar reddening as The instrument covers the wavelength range 3642–7990 we strongly suspect (see below). This corresponds to an ˚A,butwithgapsinthewavelengthcoveragebecausethe earlyA-typesupergiantorhotter,withlog(L/L⊙)≥5.1, spectrum is dispersed onto three different detectors; be- and an initial mass of at least 20 M⊙. This is much like causeofthelowsignal-to-noiseratio(S/N),weonlyused the LBV star HD 168625 (Smith 2007), and it has the same spectral type but is less luminous than the yellow dataonthe redchipoverthe interval6543–7990˚A.This hypergiant IRC+10420 which, interestingly, has a spec- was a single exposure, so individual cosmic rays in the trum similar to that of the U2773-OT outburst, as we extractedspectrumweremaskedout;thesefeatureswere discuss below. We find it quite likely, however, that the alwaysafewpixelswideanddidnotsignificantlyimpact progenitor star had some additional circumstellar dust the line profiles. based on its IR excess (§3.5), in which case it was even The spectrumhas verylowS/Nin the blue range,but hotter, more luminous, and more massive than our esti- several emission lines are detected at red wavelengths. mates from HST data alone. The spectrumrevealsnarrowemissionlines of Hα, [N ii] λλ6548, 6583, and [S ii] λλ6716, 6731 having full width 2.3. Infrared Photometry During Eruption athalf-maximumintensity(FWHM)≈49kms−1 (much We observed U2773-OT simultaneously in the JHKs wider than the instrument resolution of ∼5 km s−1); bands with the 1.3-m Peters Automated Infrared Imag- these are due to an underlying H ii region that was not ingTelescope(PAIRITEL;Bloometal.2006),beginning subtracted, or perhaps an extended circumstellar neb- on 2009 Sep. 8 (day 21 after discovery) and at several ula. In addition, Hα has a broader base attributable subsequent epochs as listed in Table 3. The PAIRI- to the transient. The HIRES spectrum of U2773-OT in TEL observations were scheduled and obtained auto- the wavelength range around Hα is shown in Figure 6. matically by the robotic telescope, and the images were The underlying broad feature has a clear P Cygni pro- processed and reduced as part of an automatic pipeline. file, which can be fit with a combination of an emission ArchivalTwoMicronAllSkySurvey(2MASS;Skrutskie line havingGaussianFWHM ≈360kms−1 andanother et al. 2006) images of the field taken on 1998 Nov. 11 Gaussianin absorption. The P Cygni absorptiontrough were used as reference data for image differencing. We is at −350 km s−1, in agreement with velocities quoted performed PAIRITEL−2MASS image subtraction using by Berger & Foley (2009) in an earlier spectrum. HOTPANTS8, and measuredthe flux of the transientin We obtained a HIRES spectrum of SN 2009ip on the thedifferenceimagesviaaperturephotometry. Thepho- followingnight,2009Sep.11(day16). Despitegoodcon- tometry was calibrated relative to the 2MASS stars in ditions,the 20-minexposurewasnotdeepenoughto de- the field. To estimate the uncertainties we inserted fake tectthecontinuum,andsubsequentanalysisofunfiltered starsatthe measuredmagnitude ofU2773-OTontopof photometryofthe guiderimage revealedthat SN 2009ip had faded more than we anticipated, to 20.2 mag. Our 8http://www.astro.washington.edu/users/becker/hotpants.html. HIRES observation was conducted at or near the mini- Two LBV Eruptions and their Progenitors 7 Fig.4.— Same as Figure 3, but zooming in on the time period near maximum light. Plotting symbols for SN 2009ip, U2773-OT, SN2008S,andN300-OTarethesameasinFigure3. Forcomparisonatthisscale,wehavealsoaddedtheV-bandlightcurveofSN1997bs (orange; VanDyketal.2000) andtheunfiltered/R-band lightcurveofSN2002kg (blue; VanDyketal.2006). Therightpanel showsan expandedviewoftheleftpanel. spectra of both SN 2009ip and U2773-OT. We observed 14.0 K - 0.7 withthesameLRISsetupforbothSN2009ipandU2773- OT, consisting of medium-resolution (0′.′7 slit width; 14.5 1200linesmm−1grating;3˚Apixel−1)andlow-resolution H (1′.′0 slit width; 400 lines mm−1 grating; ∼6 ˚A pixel−1) 15.0 red-side settings. On the blue side, a 400 lines mm−1 mag15.5 J grism was used in both settings, providing spectra with ∼3 ˚A pixel−1. The medium-resolution setting yielded 16.0 blue and red spectra in the wavelength ranges 3120– 16.5 R + 1 5594 ˚A and 6250–7880 ˚A, respectively, while the low- resolution setting covered the full range 3246–10,260 ˚A 17.0 (the bluest wavelengths were clipped due to noise). For -20 0 20 40 Day6s0 80 100 120 SN 2009ip, we used exposure times of 780 s (blue side, medium resolution), 2×320 s (red side, medium resolu- Fig. 5.—Thenear-IRapparentJHKsphotometryofU2773-OT tion), and300s (both sides, low resolution). For U2773- indaysafterdiscovery(2009Aug.18.08)obtainedwithPAIRITEL, comparedtotheKAITunfiltered(∼R-band) opticalphotometry. OT, all exposures were 300 s. The reduction of the U2773-OT spectrum was com- plicated because the long-slit spectra revealed extended mum of a sudden fading episode, discussed at length in emission along the slit due to a background H ii region §3.2. Atthisepoch,however,wediddetecttheHαemis- or extended circumstellar nebula within a few tenths of sion line, which has a total line flux of 0.7×10−14 erg an arcsecond from the transient. This background H ii s−1 cm−2 (this is one third the value measuredin lower- regionemission was sampled on either side of the source resolution spectra obtained a few days later; see below). and carefully subtracted from the spectrum, although TheHIRESHαlineprofileinSN2009ipisshowninFig- some subtraction residuals remained at low levels (see, ure 7; the spectrum is verynoisy, eventhough the pixels for example, the [S ii] lines in the red-side spectrum). have been binned by a factor of 4. Given the quality of Figures8and9showthefinalmedium-resolutionLRIS the data, the line can be approximated adequately by spectra of both transients in the blue and red, respec- a Lorentzian profile with FWHM ≈ 550 km s−1; this is tively, while Figure 10 zooms in on the Hα profiles of the same profile that is well fit to the lower-resolution each. Figure 11 displays the wavelength range around spectrum(see below). We do notclearly detect anynar- the red [Ca ii] lines seen in the LRIS spectra for both row component from an underlying H ii region around targets, as well as the HIRES data for U2773-OT. Fig- SN 2009ip, although this is not necessarily surprising as ure 12 shows the wavelength range around Na i D and many nearby LBVs are not in bright H ii regions. He i λ5876 in the low-resolution LRIS spectra, which is Later, on 2009 Sep. 21, we observed SN 2009ip again, in a spectral region excluded by the medium-resolution this time with the Low Resolution Imaging Spectrome- LRIS data. Finally, the full low-resolution LRIS spectra ter (LRIS; Oke et al. 1995) on Keck I. To our surprise, ofbothSN2009ipandU2773-OTareshowninFigure13. unfiltered photometry in the guider image revealed that InFigure13wecomparetheselow-resolutionLRISspec- SN 2009iphadrebrightenedto 18.3 mag (listed in Table traofbothtransientstospectraoftheyellowhypergiant 1), returning almost to its peak brightness. The condi- IRC+10420(fromSmithetal.2009),aswellasSN2008S tionswerephotometric,with0′.′85seeing,soweobtained 8 Smith et al. (Smithetal.2009)andtheLBVSN1997bs(VanDyket been subject to detailed non-LTE (localthermodynamic al.2002)duringeruption. Thelow-resolutionLRISspec- equilibrium) modeling of its outburst spectrum, reveal- trum of U2773-OT was qualitatively similar, although ing log(L/L⊙) ≈ 106.4 and R/R⊙ ≈ 300–400 during its superior to, a spectrum obtained at Lick Observatory outburst, and a strong stellar wind with v∞ ≈ 300 km three nights earlier, which is not shown here. s−1 (Drissenetal.2001). Theradiusofthephotosphere, 3. RESULTS however, only increased at a rate of ∼4 km s−1 (i.e., much slower than the steady wind expansion), so this 3.1. Light-Curve Comparison was not an explosion. V1 lives within the well-known Absolute-magnitude light curves for SN 2009ip and “mini-starburst”giantHiiregionNGC 2366(Drissenet U2773-OTaredisplayedinFigure3,wheretheyarecom- al. 1997) in the galaxy NGC 2363, surrounded by many pared with those of several other objects. The gray line young, massive stars. This is a well-established case of showsthehistoricalvisuallightcurveofη Carduringits a super-Eddington wind outburst from a massive star, 19th century “Great Eruption,” compiled from histori- despite the apparently low luminosity of its quiescent- cal sources by Frew (2004). It showed gradual brighten- phase progenitor. The LBV outburst of V1 is similar in ing 10 yr before peak, a pre-outburst event in 1837,and duration and absolute magnitude to the precursor out- then finally the M ≈−14 mag peak of its eruption bursts ofthe twonew transientspresentedin this paper, Visual in 1843, followed by a slow and irregular decline over although the V1 LBV outburst has not (as yet) culmi- the next decade. While η Car is the most famous and nated in a comparably bright giant eruption phase with beststudiedLBV,itiscertainlynotrepresentativeofall MV .−12 mag; instead, it appears to be an example of LBVs. The prolonged eruption, in particular, is highly a fainter and prolonged LBV eruption. unusual. BothSN2009ipandU2773-OTstandinbetweenηCar We also show the B-band light curve of the prototyp- and SN 1954J, with all four objects showing precursor ical LBV eruption SN 1954J (V12 in NGC 2403; shaded variability in the decade leading up to the peak of their grayinFigure3)fromTammann&Sandage(1968),cor- giant eruptions. Again, the cause for this is not known, rected for Galactic extinction. Before the LBV eruption but the phenomenon is a well-established property of began,the quiescentstarwasconsiderablylessluminous someLBVs(e.g.,Humphreysetal.1999). Thisprecursor thanη Car,withanabsoluteB magnitudeofonlyabout variabilityandtherangeofluminosityoftheprogenitors −7to−7.5mag. Yet,Tammann&Sandage(1968)noted makes a strong case that SN 2009ip and U2773-OT are, that it was a key example of the class of bright blue ir- in fact, both bona fide giant LBV eruptions from mas- regular variables like those in M31 and M33 (Hubble & sivestarsthatwereinaprolongedoutburstphasebefore Sandage1953),latertermed“LBVs,”andthestarisnow their discovery. knowntohavesurvivedtheevent(Smithetal.2001;Van Figure 3 also includes some available information for Dyk et al. 2005). The progenitor, V12, is a member of SN 2008S and N300-OT, and their dust-obscured pro- thelessluminousclassofLBVswithinitialmassesof20– genitors. The light curves of their eruptions show peak 40 M⊙ (Smith et al. 2004). V12 showed rapid irregular absolute magnitudes of roughly −14 mag, with a rela- variability in the ∼5 yr leading up to its outburst, with tively fast decline over 100–200 days resembling that of oscillations of 1–2 mag. (This “flickering” is discussed SN1954J.TheirsimilaritytoLBVeruptionswasalready more in §3.2 below.) Such rapid variability is unusual, notedinpreviouspapers(Smithetal.2009;Bergeretal. but V12 also has unusually well-sampled pre-eruption 2009a; Bond et al. 2009; Prieto et al. 2008; Thompson photometry. The cause of these wild oscillations is un- et al. 2009), but they were surprising because of their known, but they probably signify a growing instability relativelylow-luminosityprogenitorscomparedtoLBVs. in the star and herald the approaching runaway of the However,Figure 3 shows that their progenitor luminosi- SN 1954J event itself.9 ties are not that low after all. The plotted quantities in Another case to consider is the more recent and well- Figure3arethebolometricluminositiesderivedfromfits studied LBV outburst V1 in the nearby dwarf irregular tothe mid-IRspectralenergydistributions (SEDs)mea- galaxyNGC2363. TheabsoluteV lightcurvefromDris- sured in Spitzer data (Prieto et al. 2008; Prieto 2008; senetal.(1997,2001;seealsoPetitetal.2006)isshown Bond et al. 2009). While their luminosities may over- with the orange curve in Figure 3. (Its dwarf host with lap with the most extreme AGB stars at the very tip of metallicitysimilartothatoftheSmallMagellanicCloud theirevolution(Thompsonetal.2009),theluminosityof is particularly relevant for U2773-OT.) V1 is notewor- N300-OT is the same as the pre-outburst luminosity of thy because, again, its LBV eruption came from a star V12inNGC2403,knowntobeanLBV,andisveryclose whose initial luminosity was low compared to that of η to the pre-outburst luminosity of U2773-OT.The IR lu- Car —in fact, its quiescentpre-outburstluminosity was minosity of the SN 2008S progenitor is only a factor of equivalent to the IR luminosity of SN 2008S,also shown ∼2less,andiscomparabletothequiescentluminosityof in Figure 3 (more details are given below). During its the LBV V1 in NGC 2363. This makes it plausible that decade-long outburst in the 1990s, its MV brightened both of these transients were moderately massive stars, by ∼3.5 mag and remained so for several years. Un- comparableto or somewhatless massivethanV12. This like the historical examples of V12 and η Car, V1 has LBV connection is reinforced by a spectral comparison, discussed later. 9 While the peak of the SN 1954J eruption was fainter than An obvious caveat is that we are comparing inte- othersinFigure3,itisworthnotingthatitwasnotobservedfor∼7 grated IR luminosities of dust-enshrouded progenitors months before peak, and Tammann & Sandage (1968) suspected (for SN 2008S and N300-OT) to estimates of the lu- that a brighter peak magnitude may have occurred during that hiatusfromobserving. Weindicatethistimeintervalwithadashed minosity based on visual-wavelength photometry. The lineinFigs.3and4. Two LBV Eruptions and their Progenitors 9 comparison in Figure 3 assumes bolometric corrections of zero for the optically identified sources, and does not include possible IR excesses or correction for local ex- tinction, so these luminosities are actually lower limits. However,theintegratedmid-IRluminositiesofSN2008S and N300-OT are also lower limits, since they only rep- resent the luminosity absorbed and reradiated by warm dust, whereas radiation may escape in other directions thatwecannotseewithoutheatingdustifthedustshells are nonspherical (in the case of an edge-on dust torus, for example). In any case, further corrections beyond thevaluesshowninFigure3becomeveryuncertain,but could only raise the luminosities shown here. 3.2. Peak Luminosity, Decline Rates, and “Flickering” The photometric behavioraroundthe time ofpeak lu- minosity is unusual for both transients. Figure 4 shows thesamelightcurvesasinFigure3,butonanexpanded scaletoillustratethedetailsofthegianteruptionsthem- selves. Inthe5yrbeforediscovery,U2773-OTwasapparently Fig. 6.—HIRESspectrumofU2773-OTonday22,showingHα and the narrow [N ii] lines. The thick gray curve is a Gaussian in an unstable pre-outburst state. It continually rose in withFWHM=360km s−1,andthethinblackcurveisthesame brightness,culminatinginitspeakabsoluteRmagnitude Gaussian with a blueshifted P Cygni absorption component sub- of about −12.8. The total increase was ∆m≈5 mag. It tracted. Theabsorptionminimumisat−350kms−1. Thenarrow hasremainedroughlyatthatluminosityforafewmonths components of all lines have Gaussian FWHM ≈ 49 km s−1, and afterward,showingminoroscillationswithamplitudesof ariseinanunderlyingHiiregionorextendedcircumstellarnebula thatwasnotsubtractedfromtheHIRESdata. no more than 0.2 mag on timescales of several days. SN 2009ip was different and rather astonishing. After its precursor eruption ∼5 yr before discovery, it settled to a fainter state, with only upper limits of about −10 mag at ∼1 yr before discovery. This may mark a tem- poraryreturnto itsquiescentstateseenbyHST.Itthen brightened to MR = −13 mag, and continued to rise to a peak magnitude even brighter than that of η Car, at MR =−14.5mag. The total increase in brightness from itspre-outburststatewas∼4.7mag. Unexpectedly,how- ever,SN 2009ipsuddenly faded by 3.2 mag in ∼16days, only to recoveragain soon after (Fig. 4). We announced our discovery of this startling dip and recovery shortly before submitting this paper (Li et al. 2009). Inconnectionwiththe transientsSN2008SandN300- OT,onemaywonderiftheirfastdeclineratesareconsis- tent with LBVs. In fact, severalwell-studied LBVs have shown extremely fast decline rates. At ∼100 days after peak, V12/SN 1954J exhibited a very rapid decline rate of 0.05 mag day−1, comparable to those of SN 2008S and N300-OT (Fig. 4). SN 1997bs also showed an ex- tremely rapid 1.5 mag decline from its peak magnitude Fig. 7.—HIRESspectrumofSN2009iponday16,showingHα. in the first 20 days, although that decline rate varied The spectrum has been binned by a factor of 4 to reduce noise, later. Evenη Carhadarapidriseanddeclineassociated andthecontinuumwasnotclearlydetected. Thethickgraycurve with its events in 1837 and 1843 (Fig. 3). isaLorentzianprofilewithFWHM=550kms−1. In this context, the astonishingly fast decline of SN 2009ip from its maximum luminosity is quite impor- larity of progression.” tant. Its prediscovery luminosity and variability estab- To continue quoting Herschel: “What origin can we lish that it is a true LBV, yet it fades faster than any of ascribe to these sudden flashes and relapses?” This old the historical LBVs, SN impostors, or the controversial mystery persists, and is made even more extreme by the transientsSN2008SandN300-OT.Thesharpfadingand case of SN 2009ip. A fading of over 3 mag in 16 days rebrightening of SN 2009ip is even more extreme than is too fast for most physical mechanisms one can imag- the fluctuations experienced by η Car in 1837 and 1843, ine, and is faster than most SNe. It cannot be a huge providing some assurances that the rapid 19th century increaseinextinctioncausedbyasimplepuffofdustfor- fluctuations observedby J. F. W. Herschel may have, in mation, since the time for ejected material to reach the fact, been real, leading him (see Herschel 1847) to de- dustsublimationradiusof170–230AU(forthe observed scribe η Car as “astar fitfully variable to anastonishing luminosity of ∼2×107 L⊙) expanding at ∼500 km s−1 extent... apparently with no settled period and no regu- is much longer — roughly 600–800 days depending on 10 Smith et al. Fig.8.—Medium-resolutionLRISspectra of SN 2009ip (day 25) and U2773-OT (day 34) inthe blue, dereddened byE(B−V) values of0.019magand0.564magforSN2009ipandU2773-OT,respectively. TheforestofabsorptionlinesinU2773-OTisreal. grain-condensation temperatures. We can rule out sub- OT may have been an oversimplification, which is why stantialdustanyway,basedonthe lackofIRexcessdur- wealsoplotindividualmeasurementsinFigures3and4. ing the fading, as discussed below in §3.5. Furthermore, afterthedip,theluminosityofSN2009iprecoveredfaster 3.3. Spectral Morphology thancanbeexplainedbythesubsequentthinningofthat Figures 8 and 9 show the medium-resolution hypotheticaldust. Onecanimaginethathydrogeninthe Keck/LRISspectra atblue andred wavelengths,respec- high-density wind could recombine quickly, but this re- tively, of SN 2009ip (day 25) and U2773-OT (day 34). quires that the source of ultraviolet (UV) photons was Theseareusefulfordiscussingthe generalappearanceof suddenly quenched, and it seems inconsistent with our the spectra, and the differences between the two tran- detection of Hα during the dip. Unless it was muchhot- sients. We describe the spectral morphology in detail ter than typical LBV eruptions, the photospheric radius below; in brief, SN 2009ip resembles typical spectra of of SN 2009ipmust have been comparable to the orbit of LBVs in their hotter state, whereas U2773-OT is ex- Saturn, but fluctuating as fast as (or faster than) the emplary of the complex spectra of LBVs in their cooler wind’s expansion speed, so the sudden ejection of an state. optically thick shell is perhaps the most likely culprit. The day 25 spectrum of SN 2009ip in Figures 8 and Davidson& Humphreys (1997)noted that in the case of 9 is dominated by strong Balmer lines with Lorentzian η Car,this rapidfluctuationchallenges eventhe dynam- FWHM widths of ∼550 km s−1 and a smooth contin- ical timescale of the star itself. A fading by more than a uumwithanapparentblackbodytemperatureof∼104K. factor of 10 over such a short time in SN 2009ip is truly It also shows broad emission profiles of He i λλ5876, spectacular. 6678,and7065withsimilarwidths(notethatHeiλ5876 Other LBVs show qualitatively similar fading and re- was in the gap between the blue and red LRIS medium- brightening episodes, although less extreme, which we resolutionspectra in Figs. 8 and 9, but it can be seen in refer to as “flickering.” V12 in NGC 2403, for exam- the low-resolution spectrum in Fig. 12). SN 2009ip also ple, oscillated wildly in the decade before its eruption exhibitsseveralnarroweremissionlineswhicharemostly (Fig. 3), as noted above, with several changes of more Fe ii, probably produced in the outer wind. Overall,the than 1 mag on equally short timescales (Tammann & spectrumistypicalofclassicalLBVsintheirhotterstates Sandage1968). Inthe same galaxy,V37/SN2002kghad (e.g., Hillier et al. 2001; Szeifert et al. 1996; Stahl et al. a rapid fading and rebrightening episode about 350–380 1993;Stahl 1986). In the dereddened low-resolutionday days after peak, as plotted in Figure 4 (see Van Dyk et 25spectrumofSN2009ip(Fig.13),wemeasureaBalmer al.2006). Thissortofrapidvariabilityarguesthatshort decrement of Hα : Hβ : Hγ = 2.74 : 1.0 : 0.48. This is cadences are valuable when obtaining photometry and veryclosetoCaseBrecombinationvalues,andissimilar spectroscopy of these objects, lest one miss a significant tothedecrementHα: Hβ : Hγ =2.6: 1.0: 0.5observed mass-loss event. In that case, our method of stacking in SN 1997bs(Van Dyk et al. 2000)in a spectrum taken seasonal data for the prediscovery variability of U2773- shortly after discovery. Note, however, that the spec-