Multiple giant eruptions and X-ray emission in the recoiling AGN/LBV candidate SDSS1133
MMNRAS , 1–29 (2021) Preprint 21 January 2021 Compiled using MNRAS L A TEX style file v3.0
Multiple giant eruptions and X-ray emission in the recoilingAGN/LBV candidate SDSS1133
Mitsuru Kokubo , (cid:63) † Astronomical Institute, Tohoku University, 6-3 Aramaki-Aza-Aoba, Aoba-ku, Sendai, Miyagi 980-8578, Japan Department of Astrophysical Sciences, Princeton University, Princeton, New Jersey 08544,USA
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
We present a comprehensive analysis of 20 years worth of multi-color photometriclight curves, multi-epoch optical spectra, and X-ray data of an off-nuclear variableobject SDSS1133 in Mrk 177 at z = 0 . ∼ erg s − with small-scaleflux variations, and peak luminosities during the outbursts reach ∼ erg s − .The optical spectra exhibit enduring broad hydrogen Balmer P-Cygni profiles withthe absorption minimum at ∼ − ,
000 km s − , indicating the presence of fast mov-ing ejecta. Chandra detected weak X-ray emission at a 0 . −
10 keV luminosity of L X = 4 × erg s − after the 2019 outburst. These lines of evidence stronglysuggests that SDSS1133 is an extremely luminous blue variable (LBV) star experi-encing multiple giant eruptions with interactions of the ejected shell with differentshells and/or circumstellar medium (CSM), and strongly disfavors the recoiling Ac-tive Galactic Nuclei (AGN) scenario suggested in the literature. We suggest that pul-sational pair-instability may provide a viable explanation for the multiple energeticeruptions in SDSS1133. If the current activity of SDSS1133 is a precursor of a super-nova explosion, we may be able to observe a few additional giant eruptions and thenthe terminal supernova explosion in future observations. Key words: galaxies: active – stars: mass-loss – stars: variables: general – stars: indi-vidual: SDSS J113323.97+550415.8
SDSS J113323.97+550415.8 (hereafter SDSS1133) has longbeen recognised as an unusually persistent extragalacticvariable object (see also L´opez-Corredoira & Guti´errez 2006;Zhou et al. 2006; Keel et al. 2012; Reines et al. 2013; Kosset al. 2014; Burke et al. 2020b; Ward et al. 2020). SDSS1133is a point source with a quasar-like optical color, located atthe outskirt of a blue compact dwarf galaxy Mrk 177 (UGCA239) at the galactocentric distance of 5 (cid:48)(cid:48) . .
81 kpc (Fig-ure 1). By analysing historic photometry data of SDSS1133,Koss et al. (2014) show that SDSS1133 is optically detectedover 63 yr since 1950. The absolute magnitude of SDSS1133is in between M g ∼ −
13 mag and −
14 mag in its non-outbursting phase (before 1999 and after 2003), and it gottwo orders of magnitude brighter ( M g ∼ −
16 mag) in the (cid:63)
E-mail: [email protected] † JSPS Fellow outbursting phase in 2001 − ∼ ,
000 km s − , and theSDSS pipeline classifies SDSS1133 as a quasi-stellar object(QSO). The narrow line redshift is consistent with that of theMrk 177’s nucleus ( z = 0 . ∼ © a r X i v : . [ a s t r o - ph . H E ] J a n M. Kokubo tion of extremely variable AGNs (e.g., MacLeod et al. 2010,2012; Graham et al. 2017). The broad emission line profilecan be interpreted as complex virial motions of broad lineregion clouds surrounding the AGN. The optical and near-infrared colors remain roughly constant over the observedperiods and are redder than the stellar locus, being con-sistent with the recoiling AGN interpretation (Koss et al.2014).However, other observational properties strongly sug-gest that SDSS1133 is a kind of stellar activity-related vari-ables. The absolute magnitude and temporal behavior ofSDSS1133 are more consistent with gap transients or SNimpostors (namely giant eruptions of extragalactic lumi-nous blue variable stars; hereafter LBV), or faint supernovae(SN) (e.g., Van Dyk et al. 2000; Smith et al. 2011; Smith2017b; Pastorello & Fraser 2019; Perley et al. 2020). Also,the broad Balmer line profile of SDSS1133 exhibits blue-shifted P-Cygni-like absorption features up to ∼ − , − , which is reminiscent of those observed in SNe IIand some giant eruption LBVs (e.g., Smith 2008; Pastorelloet al. 2013; Mauerhan et al. 2013; Koss et al. 2014; Wardet al. 2020). SDSS1133 exhibits narrow/intermediate widthFe II and Ca II emission lines (Koss et al. 2014), which arerarely seen in AGNs while sometimes observed in SNe andLBVs (e.g., Smith et al. 2010; Foley et al. 2011; Taddia et al.2013; Koss et al. 2014; Smith et al. 2018; Ward et al. 2020).Recently, Ward et al. (2020) report another outburstevent of SDSS1133 detected by the Zwicly Transient Facil-ity (reported as a transient ZTF19aafmjfw, also known asGaia19bwn) beginning on 2019 April 7, peaked on 2019 June5, and faded again by 2019 December 1 (see also Stanek et al.2019; Pursimo et al. 2019a,b). Although Koss et al. (2014)and Burke et al. (2020b) suggested a scenario that SDSS1133is a LBV erupting for decades since 1950 and then followedby a terminal explosion observed as a SN IIn after 2001, thedetection of this 2019 outburst event rules out this possi-bility of terminal explosion and any other one-off event sce-narios for SDSS1133 such as SNe, tidal disruption events,or binary mergers. Based on the detection of the recurrentoutbursts and its peculiar enduring spectroscopic properties,Ward et al. (2020) suggest that SDSS1133 is likely to be anerupting LBV with the progenitor still alive, but the AGNscenario cannot be completely ruled out.Although a lot of observational data are available anddiscussed in the literature, there is still no consensus on thenature of SDSS1133. The purpose of this paper is to presenta comprehensive analysis of the multi-wavelength data ofSDSS1133 to conclude the true nature of this enigmatic ob-ject. In Section 2, we present analysis for multi-epoch imag-ing and spectroscopic data of SDSS1133, including alreadypublished and unpublished X-ray, ultraviolet (UV), and op-tical data. In Section 3, we compare the observational prop-erties of SDSS1133 with known LBVs as well as AGNs andSNe, and we conclude that SDSS1133 is one of the bright-est extragalactic giant eruption LBV, akin to η Carinae’sGreat Eruption and known SN progenitor LBV outbursts.We suggest that the X-ray-to-radio emission of SDSS1133is mediated by interactions of an ejected shell with differentshells and/or circumstellar medium (CSM), where the CSMis formed by enduring stellar wind and the ejecta are pro-duced via multiple episodes of non-terminal explosions (e.g.,
Figure 1.
The SDSS DR7 g , r , and i band color image centeredon SDSS1133. The original images are obtained on 2002 April 1.The host galaxy is Mrk 177 at z = 0 . pulsational pair-instability). Conclusions are summarized inSection 4. In this Section, we describe details of the analysis of themulti-wavelength data of SDSS1133 retrieved from variouspublicly-available data archives. The X-ray, UV, and opticalimaging data from the SDSS, PanSTARRS1 (PS1), ZTF,Palomar Transient Factory (PTF), Mayall and Bok LegacySurveys,
Swift , XMM-Newton , and
Chandra data archivesare described in Sections 2.1 − Swift and optical imaging data (SDSS and PS1) have alreadybeen published by Koss et al. (2014), but here we (re)analyseall the data in a consistent manner using
GALFIT version3.0.5 (Peng et al. 2011) to derive host-subtracted PointSpread Function (PSF) magnitude light curves of SDSS1133.Other historic measurements for SDSS1133 reported in theliterature are summarised in Sections 2.9 and 2.10. The opti-cal and X-ray broad-band photometries of SDSS1133 derivedin this work are tabulated respectively in Tables 1 and 2.Figures 2 and 3 show the broad-band light curves. The dataanalysis of the SDSS spectrum and unpublished Keck/LRISspectra retrieved from the Keck Observatory Archive (KOA)is described in Section 2.11.Following Koss et al. (2014), we use d L = 28 . µ = 32 .
30 mag) as a luminosity distance toMrk 177 (taken from NASA/IPAC Extragalactic Database(NED); Tully 1988, 1994) to calculate luminosities and ab-solute magnitudes of SDSS1133 . Since SDSS1133 is a low-redshift object ( z = 0 . K -correction is negligible inmost cases. The magnitudes are reported in the AB mag-nitude unit. The Galactic extinction toward the directionof SDSS1133 is very low, E ( B − V ) = 0 .
010 mag (Fitz-patrick 1999; Schlafly & Finkbeiner 2011). The Galactic ex-tinction coefficients for the SDSS and PS1 systems are taken NED reports an uncertainty on the distance as d L = 28 . ± . µ = 32 . ± .
80 mag). Throughout this paper we donot include the uncertainty on the distance, but it should be keptin mind that the derived luminosity is uncertain by 0 .
27 dex dueto this distance uncertainty. MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133
16 17 18 19 20 21 -16-15-14-13-12-11 - -
12 0 0 2 2 0 0 3 2 0 0 4 2 0 0 5 2 0 0 6 2 0 0 7 2 0 0 8 2 0 0 9 2 0 1 0 2 0 1 1 2 0 1 2 2 0 1 3 2 0 1 4 2 0 1 5 2 0 1 6 2 0 1 7 2 0 1 8 2 0 1 9 2 0 2 0 2 0 2 1 2 0 2 2 M agn i t ude m AB A n s o l u t e m agn i t ude M AB = m AB - µ Year2001-2002 outburst 2014 outburst 2019 outburstSDSS DEIMOSMMT LRIS LRISLowell/NOTSDSS-uSDSS-gSDSS-rSDSS-iSDSS-zPS1-gPS1-rPS1-iPS1-zPS1-yPTF-gPTF-RBok-gBok-rMayall-gMayall-zZTF-gZTF-rZTF-iGaia-G 17 18 19 20 21 22 -15-14-13-12-11 M agn i t ude m AB M AB = m AB - µ UVOT-UVW2UVOT-UVM2UVOT-UVW1OM-UVW1NIRC2-JNIRC2-K 0 5 10 15 20 25 30 52000 53000 54000 55000 56000 57000 58000 59000 60000 0 5 10 15 20 25 30 F l u x ( - e r g / s / c m ) Lu m i no s i t y ( e r g / s ) Modified Julian Date (MJD) SwiftXMMChandra
Figure 2.
Top: the optical light curves of SDSS1133. The PSF magnitudes obtained by the GALFIT modelling are shown, except forthe synthesized SDSS magnitudes from the SDSS spectrum in 2003 and
Gaia G band light curve. Galactic extinction is corrected. Threeoutbursts observed in 2001 − Swift /UVOT and
XMM-Newton /OM PSF measurements are analysed in this work. Bottom: the unabsorbed 0 . −
10 keV X-rayflux light curve. The arrows indicate the 90% upper limits on the X-ray flux.
Table 1.
UV-optical PSF photometry for SDSS1133 obtained by the
GALFIT modelling.MJD Date mag. error in mag. Telescope Filter56701.96 2014-02-13 20.26 0.03 Swift UVW258631.43 2019-05-28 18.61 0.03 Swift UVW258631.44 2019-05-28 18.26 0.03 Swift UVM256531.44 2013-08-27 20.14 0.02 Swift UVW157358.55 2015-12-02 20.52 0.05 XMM UVW157360.53 2015-12-04 20.44 0.04 XMM UVW158631.44 2019-05-28 18.10 0.02 Swift UVW152261.37 2001-12-18 16.35 0.01 SDSS u u g . . . . . . . . . . . . . . . . . .MJD indicates the mean epoch of each observation run. Galactic extinction is uncorrected. All magnitudes are listed as AB magnitude.The statistical photometric errors are reported (systematic errors are not included). The full table is only available in electronic form.MNRAS000
GALFIT modelling.MJD Date mag. error in mag. Telescope Filter56701.96 2014-02-13 20.26 0.03 Swift UVW258631.43 2019-05-28 18.61 0.03 Swift UVW258631.44 2019-05-28 18.26 0.03 Swift UVM256531.44 2013-08-27 20.14 0.02 Swift UVW157358.55 2015-12-02 20.52 0.05 XMM UVW157360.53 2015-12-04 20.44 0.04 XMM UVW158631.44 2019-05-28 18.10 0.02 Swift UVW152261.37 2001-12-18 16.35 0.01 SDSS u u g . . . . . . . . . . . . . . . . . .MJD indicates the mean epoch of each observation run. Galactic extinction is uncorrected. All magnitudes are listed as AB magnitude.The statistical photometric errors are reported (systematic errors are not included). The full table is only available in electronic form.MNRAS000 , 1–29 (2021) M. Kokubo
16 17 18 19 20 21 58200 58300 58400 58500 58600 58700 58800 58900 59000 -16-15-14-13-12-11 M agn i t ude m AB A n s o l u t e m agn i t ude M AB = m AB - µ MJDYear2019 outburstLRIS Lowell/NOT ChandraSwift ZTF-gZTF-rZTF-iGaia-GUVOT-UVW2UVOT-UVM2UVOT-UVW1
Figure 3.
The same as Figure 2, but the plot range is restricted to 58200 < MJD < g band lightcurve. Epochs of optical spectroscopy and X-ray observations ( Swift and
Chandra ) are indicated by vertical bars.
Table 2.
Swift /XRT,
XMM-Newton /EPIC, and
Chandra /ACIS unabsorbed 0 . −
10 keV X-ray flux and luminosity ( L . −
10 keV ) ofSDSS1133. MJD Date Flux Luminosity L . −
10 keV
Telescope(YYYY/MM/DD) (10 − erg s − cm − ) (10 erg s − )56531.4 2013/08/26-28 < . < .
018 Swift56702.0 2014/02/06-25 < . < .
339 Swift57359.6 2015/12/02-04 < . < .
337 XMM58631.4 2019/05/27-29 < . < .
705 Swift58708.9 2019/08/13-14 0 .
393 ( ± . .
393 ( ± . N H = 2 × cm − isassumed. Swift /XRT and
XMM-Newton /EPIC data are non-detection, thus the 90% upper limits are shown. The reported uncertaintyof the
Chandra /ACIS luminosity is ± σ . from NED, and those for other photometric bands are de-rived by assuming a T = 10 ,
000 K black body spectrumat z = 0 . SDSS u , g , r , i , and z band images were obtained onMJD=52261.4 (2001 December 18) and 52365.1 (2002 April1; Figure 1). We downloaded the SDSS corrected framesand calibration files from the SDSS Data Release (DR) 7Data Archive Server. Photometric zero points were calcu-lated from global zero points and airmass values , and PSFmodel images were reconstructed using the calibration files .The background variance maps were calculated by using SExtractor version 2.25.0 (Bertin & Arnouts 1996; Bertin2011), and variance maps were created by adding the back-ground variance and objects’ Poisson variance. http://classic.sdss.org/dr7/algorithms/fluxcal.html https://classic.sdss.org/dr7/products/images/read atlas.html Then, for each image, we performed structural decom-position of SDSS1133 and Mrk177 using
GALFIT to ob-tain host-subtracted PSF magnitudes of SDSS1133. A 2-component model was adopted, where SDSS1133 was mod-elled by a PSF and Mrk 177 was modelled by a S´ersic pro-file (see Figure 4 for an example of the
GALFIT modellingfor a ZTF image). Specifically,
GALFIT χ -minimization wasperformed leaving 3 ( X PSF , Y PSF , mag
PSF ) and 6 ( X Sersic , Y Sersic , mag
Sersic , R e, Sersic , b/a Sersic , PA
Sersic ) parameterswere fitted as free parameters for the point-source and S´er-sic component, respectively. The fitting was performed on ∼ (cid:48)(cid:48) × (cid:48)(cid:48) cutout images centered on SDSS1133.The results of the PSF photometry for SDSS1133 aresummarized in Table 1. The PSF magnitudes derived hereare consistent with the values reported in Koss et al. (2014),who derived host-subtracted PSF magnitudes of SDSS1133by fitting a PSF+linear background model for 5 (cid:48)(cid:48) × (cid:48)(cid:48) imagescentered on SDSS1133. MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 PanSTARRS1 (PS1) 3 π survey started observing the en-tire northern sky since 2010 (Chambers et al. 2016). ThePS1 data obtained before 2015 are now publicly availableas PS1 DR2. We downloaded PS1 g , r , i , z , and y band6000 × (cid:48) . × (cid:48) .
8) single-epoch warp andvariance images at the position around SDSS1133 ( sky-cell .2375.070 of the PS1 sky tessellation) from the PS1DR2 Image Cutout Service. The publicly available single-epoch images as of DR2 are spanning from 2010 February27 to 2015 February 12. Several g and i band images werevisually identified to be affected by chip and readout cellsgaps or telescope tracking errors, and were excluded fromthe analysis. PSF models of the single-epoch images at theposition of SDSS1133 were created using SExtractor and
PSFEx version 3.21.1 (Bertin & Arnouts 1996; Bertin 2011).Photometric zero points calculated by the PS1 data analysispipeline (written in the image headers) were directly used(Waters et al. 2020; Magliocchetti et al. 2020).As in Section 2.1, we performed structural decomposi-tion of SDSS1133 and Mrk177 using
GALFIT for ∼ (cid:48)(cid:48) × (cid:48)(cid:48) cutout images centered on SDSS1133. For each filter dataset,first GALFIT was performed for each image leaving all the9 parameters as free parameters, and then
GALFIT wasperformed again with the 3 S´ersic parameters ( R e, Sersic , b/a Sersic , PA
Sersic ) being fixed to mean values of the fittedparameters. We find that this two-step fitting procedure re-duces internal scatter in the PSF light curves. After the fit-ting, multiple measurements obtained on the same day werecombined into a single measurement by taking a weightedmean. The results of the PSF photometry for SDSS1133 aresummarized in Table 1.
Palomer Transient Factory (PTF) was a wide-field opti-cal time-domain survey using the Samuel Oschin 48-inchSchmidt telescope at Palomar Observatory (Law et al. 2009).PTF observed SDSS1133 in g and Mould- R bands in 2012,2014, and, 2015. Calibrated PTF images were downloadedfrom the IPAC database (Laher et al. 2014). As in Sec-tions 2.1 and 2.2, image background variance maps werecalculated by using SExtractor , and PSF models were cre-ated by using
PSFEx . Photometric zero points defined bya 8 pixel diameter aperture taken from the image header(Laher et al. 2014) were rescaled by using the PSF modelsto represent infinite-aperture zero point magnitudes. Then,
GALFIT analysis was performed to measure the PSF magni-tudes of SDSS1133 on the PTF images. Multiple measure-ments obtained on the same day were combined into a singlemeasurement by taking a weighted mean. The results aresummarized in Table 1.
Zwicky Transient Facility (ZTF) has been conducting atime-domain survey using a wide-field camera with 47 squaredegree field of view mounted on the Samuel Oschin 48-inch Schmidt telescope at Palomar Observatory since 2017.Northern Sky Public Survey is a three-day cadence survey ofthe Northern Sky with ZTF- g and ZTF- r bands, and other h m s s s D e c l . ZTF g-band 2019-06-05 h m s s s D e c l . GALFIT model h m s s s D e c l . Residual
Figure 4.
ZTF g band image on 2019 June 5 (SDSS1133’s brightphase; top), GALFIT best-fit model (middle), and residual map(bottom). The images have a size of ∼ (cid:48)(cid:48) × (cid:48)(cid:48) and are centeredon SDSS1133. The images are shown in asinh scale for visualiza-tion purposes. The SDSS coordinates of SDSS1133 and Mrk 177’snucleus are indicated by + and × symbols, respectively. private programs are also obtaining g , r , and i band im-ages over a fraction of the sky (Graham et al. 2019). Asdescribed by Ward et al. (2020), ZTF detected the 2019outburst event of SDSS1133 and reported a transient detec-tion alert on MJD=58596 (2019 April 23), and recorded thetransient as ZTF19aafmjfw (see Patterson et al. 2019, fordetails of the ZTF alert system) .The pipeline-processed data both from public and pri-vate surveys are made public via the ZTF Public Data Re- https://antares.noao.edu/alerts/data/14589650;https://lasair.roe.ac.uk/object/ZTF19aafmjfw/MNRAS000
ZTF g band image on 2019 June 5 (SDSS1133’s brightphase; top), GALFIT best-fit model (middle), and residual map(bottom). The images have a size of ∼ (cid:48)(cid:48) × (cid:48)(cid:48) and are centeredon SDSS1133. The images are shown in asinh scale for visualiza-tion purposes. The SDSS coordinates of SDSS1133 and Mrk 177’snucleus are indicated by + and × symbols, respectively. private programs are also obtaining g , r , and i band im-ages over a fraction of the sky (Graham et al. 2019). Asdescribed by Ward et al. (2020), ZTF detected the 2019outburst event of SDSS1133 and reported a transient detec-tion alert on MJD=58596 (2019 April 23), and recorded thetransient as ZTF19aafmjfw (see Patterson et al. 2019, fordetails of the ZTF alert system) .The pipeline-processed data both from public and pri-vate surveys are made public via the ZTF Public Data Re- https://antares.noao.edu/alerts/data/14589650;https://lasair.roe.ac.uk/object/ZTF19aafmjfw/MNRAS000 , 1–29 (2021) M. Kokubo lease. ZTF DR4 pipeline-processed g , r , and i band single-epoch images (sciimg.fits; primary science image of databasefield = 788) and associated PSF model images (sciimgdaops-fcent.fits; PSF estimate at science image center) were down-loaded from the Science Data System at IPAC (Masci et al.2019; Bellm et al. 2019). The ZTF DR4 includes imagestaken during the period of 2018 March 25 - 2020 June 29,2018 April 6 - 2020 June 29, and 2018 April 6 - 2019 June 18in the g , r , and i band, respectively. INFOBITS in the imageheader was used to remove bad-quality flagged images fromthe analysis below. At the position of SDSS1133, 286, 281,and 29 unflagged images in the g , r , and i band are availablein the ZTF DR4.We performed two-step GALFIT procedure as in Sec-tion 2.2; the PSF magnitudes were obtained by using
GAL-FIT twice where the second fitting was performed with the3 S´ersic parameters ( R e, Sersic , b/a Sersic , PA
Sersic ) being fixedto the mean values of the first fitting results. An example ofthe
GALFIT model and residual map are shown in Figure 4.The magnitude zero-points estimated by the ZTF pipeline(written in the image header) were used to calibrate theinstrument magnitudes to AB magnitudes (specifically, thePS1 system; Masci et al. 2019). Color terms (Masci et al.2020) were neglected, which do not affect the final resultssince SDSS1133’s colors, g − r or r − i , are close to 0. Multi-ple measurements obtained on the same day were combinedinto a single measurement by taking a weighted mean. Theresults are summarized in Table 1. The sky region of SDSS1133 was imaged by the DESI LegacyImaging Surveys (Zou et al. 2019; Dey et al. 2019), specifi-cally by Mayall z band Legacy Survey and by Bok 90Prime g and r band Survey projects, during the period between2016 and 2018. The Mayall Survey is using a mosaic camera(Mosaic-3) mounted on the 4-m Mayall telescope, and Bok90Prime survey is using the 90Prime camera at the primefocus of the Bok 2.3-m telescope. Also, since the sky re-gion of SDSS1133 is located inside the spring Hobby-EberlyTelescope Dark Energy Experiment (HETDEX) field, twoMayall g band images obtained on 2015 March 15-16 duringthe course of an imaging survey for the HETDEX (NOAOprogram ID = 2015A-0075; PI: Robin B. Ciardullo) are avail-able.Mayall and Bok g , r , and z band imaging data (commu-nity pipeline (CP) ooi_v1 products before being processedby Tractor ; Dey et al. 2019) containing SDSS1133 andMrk 177 were downloaded from the NOAO Science Archive .We used SExtractor and
PSFEx to estimate PSF modelsof the images. Then, as in Section 2.2, we used
GALFIT toevaluate the PSF magnitudes of SDSS1133. The zero-pointAB magnitudes were evaluated by comparing instrumentalPSF magnitudes of three field stars around SDSS1133 to thePS1 PSF magnitudes taken from the PS1 DR2
MeanObject
View table (Chambers et al. 2016), with color transforma-tions given in Dey et al. (2019). Table 1 summaries the PSFmagnitudes obtained by the
GALFIT analysis. http://archive1.dm.noao.edu/ Swift (2013-2019)
Swift (Gehrels et al. 2004) X-ray Telescope (XRT) and UV-Optical Telescope (UVOT) have targeted SDSS1133 severaltimes. Koss et al. (2014) carried out Target-of-OpportunityXRT and UVOT UVW1 observations for SDSS1133 on 2013August 26, 27, and 28. Also, unpublished Swift UVW2 dataobtained on 2014 February 6-25 are available (target ID =00032905). Then, several follow-up observations were trig-gered after the 2019 outburst;
Swift observed SDSS1133on 2019 May 27 (MJD=58630.1) and 29 (MJD=58632.8)for 2.4 and 2.0 ks, respectively, with the XRT and UVOTin the UVW2, UVM2, and UVW1 filters (target ID =00011418). We performed the Swift data analysis with HEA-soft software packages (Version 6.25; HEASARC 2014), us-ing calibration datasets in the High Energy AstrophysicsScience Archive Research Center (HEASARC)’s calibrationdatabase (CALDB; accessed 2020 December 17). The Swiftdata were binned into three epochs, where the mean epochsare MJD=56531.4 (2013 August 26-28; UVW1 + XRT),56702.0 (2014 February 6-25; UVW2 + XRT), and 58631.4(2019 May 27-29; UVW2, UVM2, UVW1 + XRT).
We downloaded pipeline-processed
Swift
XRT Level 1 prod-ucts from the HEASARC Archive. Calibrated and cleanedevent files in the XRT photon counting mode were producedusing xrtpipeline with standard screening criteria.We defined three XRT visits as 2013 August26-28 (MJD=56531.4; 18.6 ks), 2014 February 6-25 (MJD=56702.0; 13.8 ks), and 2019 May 27-29(MJD=58631.4; 4.3 ks), and each of the XRT datasets wascoadded separately. Spectral extraction and coadding ofthe event files were performed by using xselect . A circularaperture of 47 (cid:48)(cid:48) in radius centered on the SDSS1133’s opticalcoordinate was used for the source count extraction, and anannular aperture with the inner and outer radii of 100 (cid:48)(cid:48) and200 (cid:48)(cid:48) centered on the SDSS1133’s optical coordinates wasused for the background count extraction. xrtmkarf wasused to create aperture-corrected (uncorrected) auxiliaryresponse files (ARF) for the source (background) spectra,and an exposure time-weighted mean response file for eachvisit was created by using addarf . A common responsematrix file (RMF; swxpc0to12s6 20130101v014.rmf) takenfrom CALDB was used for the subsequent analysis.We found that the X-ray counts of SDSS1133 is quitelow, thus the spectral bins were grouped into a single fullband (0 . −
10 keV) by using
XSPEC/grppha . The back-ground-subtracted full band net count rates were calculatedto be − . ± . × − , − . ± . × − , and4 . ± . × − count s − at MJD= 56531.4, 56702.0,and 58631.4, respectively. We conclude that all of thesemeasurements were non-detection. Under the assumptionof Gaussian statistics with zero mean, 90% upper limitson the count rates are 2 . × − , 2 . × − , and6 . × − count s − , respectively. The corresponding 90%upper limits on the 0 . −
10 keV band flux and luminositywere evaluated by assuming an absorbed power-law model( phabs*powerlaw in XSPEC ) with a fixed power-law photonindex of Γ = 2 (in the form of f ν ∝ ν − Γ ) and a GalacticHI column density of N H = 2 × cm − , and the re- MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 sults are summarized in Table 2. The Swift non-detectionsuggests that the X-ray luminosity of SDSS1133 is less than ∼ × erg s − even at around the peak epoch of the2019 outburst (Figure 3).Our re-analysis of the Swift /XRT data shows that the“marginal”
Swift /XRT X-ray detection on 2013 August 26-28 reported by Koss et al. (2014) was actually non-detection.As described in Section 2.8, this conclusion is further con-firmed by the
Chandra /ACIS data which indicate that theX-ray luminosity of SDSS1133 is just below the
Swift /XRT’sdetection limit.As shown in Section 2.8, not only SDSS1133 but alsothe Mrk 177’s nuclear X-ray emission contribute to the X-ray counts in the r = 47 (cid:48)(cid:48) circular aperture. Thus, pre-cisely speaking, the luminosity upper limits obtained aboveshould be interpreted as the limits for the total luminosity ofSDSS1133 and Mrk 177, giving a conservative upper limitson the SDSS1133’s luminosity. /UVOT observed SDSS1133 through UVW1 (2600˚A),UVM2 (2246˚A), and UVW2 (1928˚A) bands. We downloadedpipeline-processed Swift UVOT Level 2 products from theHEASARC Archive. The UVM2 band data obtained on 2019May 27 and 29 (1.4 ks) are combined into a single imageusing uvotimsum . The UVW2 band data obtained on 2014February 6-25 (total exposure is 13.6 ks) and 2019 May 27-29(1.4 ks) are coadded separately, and the UVW1 band dataobtained on 2013 August 26-28 (19.8 ks) and 2019 May 27-29(1.4 ks) are coadded separately.Time dependent sensitivity corrections (several tenspercent) and small deadtime and coincidence loss correc-tions (at most 2%) were calculated by using uvotsource and applied for the measured fluxes (Poole et al. 2008;Breeveld et al. 2011). The zero point AB magnitudes forthe source count rates for the standard aperture are 18.95,18.54, and 19.11 mag for UVW1, UVM2, and UVW2 band,respectively . Using tables of encircled energy curves for theUVOT PSFs in the CALDB, we estimated zero point ABmagnitudes for the infinite aperture (defined to be 30 (cid:48)(cid:48) inradius) to be used for the PSF photometry as 19.11, 18.72,and 19.24 mag for UVW1, UVM2, and UVW2 band, respec-tively. The PSF models of the UVOT images were createdwith PSFEx , and the PSF models were used in
GALFIT to per-form structural decomposition. The
Swift /UVOT PSF mag-nitudes of SDSS1133 derived by the
GALFIT fitting are sum-marized in Table 1, where the errors on the magnitudes onlyinclude statistical errors reported by
GALFIT . We should notethat, in addition to the statistical errors, the reported mag-nitudes can be affected by various systematic errors; e.g.,the
Swift /UVOT photometric zero-point estimates induce ∼ .
03 mag errors on the magnitudes.The UVW1 magnitude of SDSS1133 on 2013 August 27 The most up-to-date UVOT sensitivity calibration file (in-cluded in CALDB ver.20201215) was used. These photometric zero-points convert count rates into ABmagnitudes, where the count rates are corrected for deadtime,coincidence-loss, aperture corrections, large-scale sensitivity vari-ations, and detector sensitivity variations. reported by Koss et al. (2014) was 21 . ± .
30 mag, whichdiffers from our result (20 .
14 mag) by ∼ .
27 mag. We spec-ulate that the difference may be due to overestimation of thehost galaxy flux contribution in Koss et al. (2014)’s analysis,where the a 5” region around SDSS1133 was fitted with aPSF model and a linear background model. For reference,we measure the 3 (cid:48)(cid:48) radius aperture flux of SDSS1133 on 2013August 27 (with aperture correction applied by using uvot-source and without host galaxy subtraction) as 19 . ± . ± .
03 (sys.) mag, which is close to our PSF mag-nitude estimate (20 .
14 mag). This indicates that the hostgalaxy contamination is small in the UV bands.
XMM-Newton (2015 December)
XMM-Newton (Jansen et al. 2001) X-ray and UV imag-ing data were obtained on 2015 December 2 and 4 (Obs.ID=0762960201 and 0762960301; PI: Michael Koss), whichis unpublished in the literature. We downloaded the two
XMM-Newton datasets (Observation Data File; ODF) fromthe
XMM-Newton
Science Archive. The Current CalibrationFiles (CCF) were downloaded from the CCF ftp server (ac-cessed 2020 October 28). The data analysis was performedby using the
XMM-Newton
Science Analysis System (
SAS version 18.0.0; Gabriel et al. 2004).
XMM - Newton /EPIC
Filtered photon counting event files of European PhotonImaging Camera (EPIC) MOS and PN cameras were gener-ated by using standard
SAS tasks emproc , epproc , and evs-elect . The exposure times of the MOS1, MOS2, and PNdata are 8.2, 8.4, and 4.1 ks on 2015 December 2, and 11.3,11.5, 7.2 ks on 2015 December 4, respectively. MOS and PNspectra of the two datasets (on 2015 December 2 and 4)were extracted and then combined into a single source spec-trum, background spectrum, and response matrix (RSP) byusing SAS/multiespecget . Here the combined RSP file wasdefined by summing up RSP files (=RMF × ARF) of all ofthe MOS and PN data, and the effective exposure time ofthe combined dataset was calculated to be 8.5 ks. A circularaperture of 32 (cid:48)(cid:48) in radius centered on the optical position ofSDSS1133 was used for the source spectral extraction, andan offset circular aperture of 72 (cid:48)(cid:48) in radius was used for thebackground spectral extraction.The same analysis as Section 2.6.1 was applied forthe
XMM-Newton combined spectral data. The effectivebackground-subtracted 0 . −
10 keV net count rate was cal-culated to be 2 .
981 ( ± . × − count s − , thus weconclude that SDSS1133 is undetected in the XMM-Newton data. The 90% upper limit on the effective count rate is2 . × − count s − . Assuming the same absorbed power-law spectral model ( phabs*powerlaw in XSPEC ), we obtainedan 90% upper limit on the unabsorbed X-ray flux in the0 . −
10 keV band as 3 . × − erg cm − s − , corre-sponding to an 90% upper limit on the 0 . −
10 keV lumi-nosity of L . −
10 keV = 3 . × erg s − , as summarizedin Table 2.The source extraction aperture contains not onlySDSS1133 but also the Mrk 177’s nuclear X-ray emission (seeSection 2.8), thus the luminosity upper limit obtained above MNRAS000
10 keV = 3 . × erg s − , as summarizedin Table 2.The source extraction aperture contains not onlySDSS1133 but also the Mrk 177’s nuclear X-ray emission (seeSection 2.8), thus the luminosity upper limit obtained above MNRAS000 , 1–29 (2021)
M. Kokubo only provide a conservative upper limit on the SDSS1133’sluminosity.
XMM - Newton /OM
Both of the two
XMM-Newton /Optical Monitor (OM; Ma-son et al. 2001) imaging mode datasets on 2015 December 2and 4 were obtained through the UVW1 band (2910 ˚A) withthe pre-defined EPIC Imaging mode configuration, wherethe total on-source exposure times were 10.9 and 11.8 ks,respectively. Astrometrically-aligned mosaiced sky images( ∼ (cid:48) × (cid:48) , 2 × .
953 pixel − ) of the twodatasets were separately generated from the ODFs by us-ing SAS/omichain . Since SDSS1133 is faint, the coincidence-loss is negligible. The dead time fraction (0 . . .
566 mag) were directly takenfrom the CCF. We applied the same analysis as Section 2.6.2to the XMM-Newton /OM data;
SExtractor and
PSFEx wererespectively used to subtract global background from theimages and create PSF models, and then
GALFIT was usedto perform the structural decomposition to obtain the PSFmagnitude of SDSS1133. The results of the PSF photometryis listed in Table 1.
Chandra (2019 August)
Chandra (Weisskopf et al. 2002) observed SDSS1133 on 2019August 13 (MJD=58708.4, 16 ks) and 14 (MJD=58709.3,58 ks) with the Advanced CCD Imaging Spectrometer, ACIS(Obs. ID = 21434 and 22684; PI: David Pooley), which isunpublished in the literature. Since the optical light curvesshow that SDSS1133 returned to the optically normal phaseby the end of 2019 June (Figure 3; Pursimo et al. 2019b), the
Chandra observations put a constraint on the X-ray emissionof the SDSS1133 in its post-outburst phase.We used
CIAO v4.12.1 (Fruscione et al. 2006) to down-load and analyse the archival ACIS data. Figure 5 showsa 0 . − ∼ (cid:48)(cid:48) × (cid:48)(cid:48) ) created by using CIAO/merge_obs .SDSS1133 is clearly detected in this image, and also the nu-cleus of the host galaxy Mrk 177 is detected at a similarcount rate.The event files, ARFs, and RMFs, reprocessed with
CIAO/chandra_repro using CALDB 4.9.2.1, were input to
CIAO/specextract to extract count rate spectra, where acircular aperture of 4 (cid:48)(cid:48) in radius and annular aperture with10 (cid:48)(cid:48) inner and 40 (cid:48)(cid:48) outer radii were used for the source andbackground extraction, respectively. The optical position ofSDSS1133 was used to extract the spectra. Aperture cor-rections were applied for the ARFs for the source. Then,the spectra from the two exposures were summed by using
CIAO/combine_spectra to create a 74.2 ks spectrum.The background-subtracted full-band (0 . − .
327 ( ± . × − count s − with the signal-to-noise ratio of S/N = 3 . The UVW1-filter zero point is defined by an aperture of 35unbinned pixels (16 (cid:48)(cid:48) .
7) radius, and for count rates corrected forthe coincidence-loss, dead time, and sensitivity degradation. h m s s s D e c l . Figure 5.
Chandra /ACIS 0 . − ∼ (cid:48)(cid:48) × (cid:48)(cid:48) , 0.492 arcsec pixel − ) obtainedon 2019 August 13 and 14 (74.2 ks; Obs. ID = 21434 and 22684).The colorbar indicates the X-ray counts. North is up and Eastis left. The source extraction aperture (4 (cid:48)(cid:48) in radius) centeredon SDSS1133 is shown as a dotted line. The X-ray source located ∼ (cid:48)(cid:48) north west from SDSS1133 corresponds to the nucleus of thehost galaxy Mrk 177. The SDSS optical coordinates of SDSS1133and Mrk 177’s nucleus are indicated by + and × symbols, respec-tively. the observed X-ray counts is too small for the spectral anal-ysis (see Figure 5), thus here we report integrated quantitiesin a 0 . − CIAO/modelflux with theARFs and RMFs, and assuming a power-law spectral modelwith a fixed power-law photon index of Γ = 2 ( xspowerlaw )and a Galactic HI column density of N H = 2 × cm − ( xsphabs ), we obtained an unabsorbed X-ray flux in the0 . − .
109 ( ± . × − erg cm − s − ,or 0 . − L . − = 3 .
107 ( ± . × erg s − . The corresponding 0 . −
10 keV luminosityis 3 . × erg s − (Table 2). As already mentioned inSection 2.6.1, this faint X-ray luminosity suggests that the Swift /XRT observations were not sensitive enough to detectX-ray emission from SDSS1133.Under the assumption that SDSS1133 is a LBV, anoptically-thin thermal plasma model may be a more ade-quate reference model (e.g., Naz´e et al. 2012). By using athin thermal plasma emission model xsapec (Smith et al.2001) in
CIAO/modelflux , and fixing the temperature to k B T = 0 . N H = 2 × cm − , and metal abun-dance to Z (cid:12) , we obtained an unabsorbed X-ray flux in the0 . − .
425 ( ± . × − erg cm − s − , or0 . − L . − = 3 .
423 ( ± . × erg s − . The difference of the estimated luminositybetween the power-law model and thermal plasma modelis small, suggesting that the luminosity estimates are notvery sensitive to the choice of the spectrum models. In thesubsequent sections, we only refer to the X-ray luminosities MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 obtained under the assumption of the absorbed power-lawmodel summarised in Table 2.By applying the same analysis on the nucleus ofMrk 177, the count rate was evaluated as 1 .
770 ( ± . × − count s − . We obtained an unabsorbed X-ray flux inthe 0 . − .
339 ( ± . × − erg cm − s − ,or 0 . − L . − = 2 .
337 ( ± . × erg s − . The corresponding 0 . −
10 keV luminosity is2 . × erg s − . Accidentally, this luminosity is verysimilar to that of SDSS1133. Since the separation betweenSDSS1133 and Mrk 177’s nucleus is 5 (cid:48)(cid:48) .
8, the low spatial res-olution X-ray telescopes (
Swift and
XMM-Newton ) cannotresolve the two X-ray components, as already pointed out inSections 2.6.1 and Sections 2.7.1.
Gaia (2015-2021)
The
Gaia mission (Gaia Collaboration et al. 2016) has beenobserving SDSS1133 in
Gaia ’s G band since 2015 January18, including the period of the second flare event in 2019.The Gaia G band is a white-light filter covering entire op-tical wavelengths (Weiler 2018). We downloaded
Gaia lightcurve data of SDSS1133 (recorded as Gaia19bwn; Pursimoet al. 2019a) from the
Gaia
Photometric Science AlertsDatabase webpage (accessed 2021 January 12), excluding‘null’ and ‘untrusted’ data points . The Gaia light curveis plotted in Figure 2, but we should note that the
Gaia photometry is host-unsubtracted.
Historic photometric measurements for SDSS1133 based onPhotographic Digital Sky Survey (DSS) plates from thePOSS I and II surveys are reported by Koss et al. (2014).Although the detections of SDSS1133 are not significant (seeFigure 1 of Koss et al. 2014), the visual magnitudes on 1950March 20, 1994 April 14, and 1999 April 25 are measured as18 . ± .
7, 18 . ± .
4, and 18 . ± . u - and g band), GG395( g band), and IIIaF emulsion ( i band), respectively.In addition, a g band upper limit is reported by Zhouet al. (2006) on January 2005; the observation using the 2.16-m telescope of the Beijing Observatory failed to detect thesource within 3 mag of the 2002 SDSS observation, providinga rough limit on the g band brightness as g > . J and K p band Adap-tive Optics assisted imaging observations for SDSS1133 withKeck/the Near Infrared Camera 2 (NIRC2) on 2013 June 16(MJD=56459.3) These observations put constraints on thesize of the SDSS1133’s emission region to be <
12 pc, and J -and K p band AB magnitudes are estimated as 19 . ± . . ± .
16 mag, respectively.
GALEX
All-sky Imaging survey provides a NUV band(1750 − λ eff = 2267 ˚A) magnitude on 2004 March 6as 21 . ± .
40 mag (Koss et al. 2014), though the detectionsignificance is not high and it can be a false detection.At a radio band, Perez-Torres et al. (2015) report a non-detection by 5.0 GHz electronic European VLBI Network http://gsaweb.ast.cam.ac.uk/alerts/tableinfo. (eEVN) radio observations for SDSS1133 on 2014 December12 provides a 3 σ upper limit of 150 microJy beam − , corre-sponding to a 5.0 GHz luminosity of 5 . × erg s − Hz − .Archival FIRST radio observations at 1.4 GHz also providesa 3 σ upper limit of about 450 microJy beam − , correspond-ing to a luminosity of 1 . × erg s − Hz − . Koss et al. (2014) present multi-epoch spectra ofSDSS1133; the SDSS spectrum on 2003 March 9(MJD=52707), Keck/DEIMOS spectrum on 2013 Decem-ber 13 (MJD=56639), and the Multiple Mirror Telescope(MMT) spectrum on 2014 January 3 (MJD=56660). Pur-simo et al. (2019a,b) conducted optical spectroscopic follow-up observations for the 2019 outburst event of SDSS1133(refereed to as Gaia19bwn in their Fig. 2) with ALFOSCon Nordic Optical Telescope (NOT) on 2019 May 29(MJD=58632), near the light curve maximum epoch (Fig-ure 3). Ward et al. (2020) present an optical spectrum takenon 2019 May 29 with the Lowell Discovery Telescope, thesame date as the Pursimo et al. (2019a)’s observation.Other than these already published data, there are un-published Keck/LRIS spectroscopic data of SDSS1133 ob-tained on 2015 July 17 (MJD=57220.2; 560 s ×
2) and2019 March 10 (MJD=58552.5; 560 s) in the KOA. Thesecond Keck/LRIS was obtained just before the 2019 out-burst event, probing the pre-outburst gradually rising phase(Figure 3). Below we describe the data reduction of theKeck/LRIS spectra (Section 2.11.1), and spectral decom-position analysis of the Keck/LRIS and SDSS spectra (Sec-tion 2.11.2). The comparisons of the multi-epoch spectra arepresented in Section 2.11.3.
We downloaded the raw Keck/LRIS spectral data ofSDSS1133, spectrophotometric star (HZ44), and associatedcalibration data from KOA (PI: Fiona Harrison, programID: C280LA, C300). All the LRIS data were obtained witha 1 (cid:48)(cid:48) . (cid:48)(cid:48) .
27 pixel − ). Theobservations were carried out under non-photometric (vari-able seeing and atmospheric extinction) conditions. The PSFFWHM were evaluated as ∼ (cid:48)(cid:48) .
2) on 2015 July 17and ∼
11 pixel (3 (cid:48)(cid:48) .
0) on 2019 March 10 from the standardstar’s spatial profiles on the dispersed images.Standard
IRAF data reduction was performed on theLRIS data, where overscan, split-illumination flat, wave-length calibration, and image distortion corrections wereapplied. Pixels affected by cosmic-rays were identified andmasked by using lacos_spec (van Dokkum 2001). The ob-ject spectra were extracted by using IRAF/apall ’s apertureextraction task with a 8 pixel (2 (cid:48)(cid:48) .
16) aperture. The host IRAF is distributed by the National Optical Astronomy Ob-servatory, which is operated by the Association of Universities forResearch in Astronomy (AURA) under a cooperative agreementwith the National Science Foundation.MNRAS000
16) aperture. The host IRAF is distributed by the National Optical Astronomy Ob-servatory, which is operated by the Association of Universities forResearch in Astronomy (AURA) under a cooperative agreementwith the National Science Foundation.MNRAS000 , 1–29 (2021) M. Kokubo F l u x ( - e r g / s / c m / A ng s t r o m ) Rest-frame wavelength (Angstrom)2003 Mar 92019 Mar 102015 July 17 HeIFeII[OIII]H β H γ H δ H ε [OII] FeII H α + [NII] [CaII] CaII triplet[SII] ⊕⊕ SDSS 2003 Mar 9LRIS 2015 July 17LRIS 2019 Mar 10
Figure 6.
Keck/LRIS and SDSS spectra of SDSS1133. Several emission/absorption features are labelled. Hydrogen Balmer series showsbroad P-Cygni-like features (see Section 2.11). Galactic extinction is corrected. The Keck/LRIS spectra are not corrected for the telluricabsorption (indicated by ⊕ ). The Galactic-extinction corrected g band magnitudes at the epochs of the SDSS and LRIS observationswere estimated as 18 .
56, 19 .
95, and 19 .
38 mag, respectively. F l u x ( - e r g / s / c m / A ng s t r o m ) Rest-frame wavelength (Angstrom)H β [OIII] [OIII]abs.FeII FeIISDSS 2003 Mar 9Bestfit model F l u x ( - e r g / s / c m / A ng s t r o m ) Rest-frame wavelength (Angstrom)H α [SII]abs.[OI] SDSS 2003 Mar 9Bestfit model Figure 7.
Spectral decomposition of the SDSS spectrum of SDSS1133 at the wavelength ranges of H β (top) and H α (bottom). The redline represents the summed model spectra. A power-law continuum, iron (mostly Fe II) pseudo-continuum template, broad, very broad,and narrow Balmer lines, and several narrow emission lines are included in the model (black lines). The power-law continuum and broadand very broad Balmer emission lines are assumed to be affected by a broad P-Cygni-like absorption; solid and dotted lines are absorbedand unabsorbed model components, respectively (see Section 2.11.2 for details). MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 F l u x ( - e r g / s / c m / A ng s t r o m ) Velocity (km/s)SDSS 2003 Mar 9Bestfit model F l u x ( - e r g / s / c m / A ng s t r o m ) Velocity (km/s)SDSS 2003 Mar 9Bestfit model F l u x ( - e r g / s / c m / A ng s t r o m ) Velocity (km/s)LRIS 2015 July 17Bestfit model F l u x ( - e r g / s / c m / A ng s t r o m ) Velocity (km/s)LRIS 2015 July 17Bestfit model F l u x ( - e r g / s / c m / A ng s t r o m ) Velocity (km/s)LRIS 2019 Mar 10Bestfit model F l u x ( - e r g / s / c m / A ng s t r o m ) Velocity (km/s)LRIS 2019 Mar 10Bestfit model
Figure 8.
The same as Figure 7, but the SDSS and Keck/LRIS spectra are shown as a function of velocity relative to the narrow H β linecenter. The Galactic extinction-corrected g band magnitudes are g = 18 .
66, 19.95, and 19.38 mag at the epochs of the SDSS and the twoLRIS observations, respectively. galaxy component was not subtracted, and telluric absorp-tion correction was not applied. The Galactic extinction wascorrected by using Fitzpatrick (1999)’s extinction curve. Therelative flux offset between the blue arm and red arm spectraof each epoch was corrected by requiring the fluxes at around λ obs = 5500 ˚A to be consistent with each other. Then, theabsolute fluxes of the spectra were roughly calibrated byscaling the spectra to match the g band photometric magni-tudes at the epochs of the spectroscopic observations, wherethe g band photometric magnitudes were evaluated by lin-early interpolating the Galactic extinction-corrected g bandlight curve in Figure 2. The Galactic-extinction corrected g band magnitudes were estimated as 19 .
95 and 19 .
38 mag on2015 July 17 and 2019 March 10, respectively, which are re-spectively 1 .
39 and 0 .
82 mag fainter than the g = 18 .
56 magestimated from the SDSS spectrum on 2003 March 9. The Keck/LRIS spectra and publicly available SDSSspectrum of SDSS1133 are shown in Figure 6. We can seethat overall spectral properties of SDSS1133 are unchangedduring the periods of these spectroscopic observations. TheP-Cygni profile is detected in all of the hydrogen Balmerseries and possibly in He I λ − (see also Koss et al. 2014; Wardet al. 2020). Many narrow-to-intermediate width metal emis-sion lines (e.g., Fe II, Ca II, and O I) are clearly detectedespecially in the Keck/LRIS spectra. The most remarkable feature of the optical spectra ofSDSS1133 is the broad Balmer P-Cygni profile. The spec-tral decomposition including the broad P-Cygni absorption
MNRAS000
MNRAS000 , 1–29 (2021) M. Kokubo
Table 3.
Best-fit model parameters and 68% percentile range derived from the spectral fitting.Parameters SDSS 2003 March 9 LRIS 2015 July 17 LRIS 2019 March 10H β region ( λ obs = 4400 − λ H β, n,obs (˚A) 4901 . +0 . − . . +0 . − . . +0 . − . σ H β, n,rest (km s − ) 91 . +1 . − . . +1 . − . . +2 . − . f H β, n (10 − erg s − cm − ) 102 . +7 . − . . +1 . − . . +3 . − . λ H β, b,obs (˚A) 4901 . +0 . − . . +0 . − . . +0 . − . σ H β, b,rest (km s − ) 609 . +58 . − . . +102 . − . . +181 . − . f H β, b (10 − erg s − cm − ) 311 . +127 . − . . +24 . − . . +54 . − . λ H β, vb,obs (˚A) 4912 . +7 . − . . +3 . − . . +37 . − . σ H β, vb,rest (km s − ) 2197 . +867 . − . . +594 . − . . +7030 . − . f H β, vb (10 − erg s − cm − ) 822 . +1461 . − . . +239 . − . . +157 . − . f [OIII]4959 (10 − erg s − cm − ) 84 . +4 . − . . +0 . − . . +1 . − . f [OIII]5007 (10 − erg s − cm − ) 242 . +6 . − . . +1 . − . . +2 . − . α λ − . +0 . − . − . +0 . − . − . +0 . − . f λ, z ) (10 − erg s − cm − ˚A − ) 12 . +0 . − . . +0 . − . . +0 . − . λ H β, abs,obs (˚A) 4865 . +2 . − . . +2 . − . . +1 . − . σ H β, abs,rest (km s − ) 2671 . +247 . − . . +234 . − . . +663 . − . τ H β, . +66 . − . . +30 . − . . +15 . − . scale Fe . +0 . − . . +0 . − . . +0 . − . z Fe /z H β, n . +0 . − . . +0 . − . . +0 . − . H α region ( λ obs = 6200 − λ H α, n,obs (˚A) 6617 . +0 . − . . +0 . − . . +0 . − . σ H α, n,rest (km s − ) 120 . +4 . − . . +2 . − . . +2 . − . f H α, n (10 − erg s − cm − ) 531 . +35 . − . . +3 . − . . +4 . − . λ H α, b,obs (˚A) 6624 . +1 . − . . +0 . − . . +0 . − . σ H α, b,rest (km s − ) 961 . +45 . − . . +10 . − . . +18 . − . f H α, b (10 − erg s − cm − ) 4556 . +161 . − . . +6 . − . . +15 . − . λ H α, vb,obs (˚A) 6619 . +17 . − . . +1 . − . . +2 . − . σ H α, vb,rest (km s − ) 3509 . +46 . − . . +85 . − . . +172 . − . f H α, vb (10 − erg s − cm − ) 8289 . +315 . − . . +20 . − . . +35 . − . f [NII]6548 (10 − erg s − cm − ) 13 . +12 . − . . +0 . − . . +3 . − . f [NII]6584 (10 − erg s − cm − ) 0 . +26 . − . . +0 . − . . +2 . − . f [SII]6717 (10 − erg s − cm − ) 97 . +3 . − . . +0 . − . . +1 . − . f [SII]6731 (10 − erg s − cm − ) 68 . +2 . − . . +0 . − . . +1 . − . f [OI]6300 (10 − erg s − cm − ) 31 . +2 . − . . +0 . − . . +1 . − . α λ − . +0 . − . − . +0 . − . − . +0 . − . f λ, z ) (10 − erg s − cm − ˚A − ) 12 . +0 . − . . +0 . − . . +0 . − . λ H α, abs,obs (˚A) 6559 . +11 . − . . +0 . − . . +1 . − . σ H α, abs,rest (km s − ) 2274 . +100 . − . . +78 . − . . +103 . − . τ H α, . +9 . − . . +2 . − . . +3 . − . scale Fe . +0 . − . . +0 . − . . +0 . − . z Fe /z H α, n . +0 . − . . +0 . − . . +0 . − . The spectral fitting was performed at the H β and H α spectral ranges independently. The Galactic extinction is corrected, and thehost galaxy extinction is uncorrected. λ obs , σ rest , and f indicate the observed-frame central wavelength, rest-frame velocity standarddeviation, and integrated flux of the narrow (n), broad (b), and very broad (vb) Gaussian line components. The narrow lines are onlymarginally spectrally-resolved. The continuum is assumed to have a power-law form of f λ = f λ, z ) [ λ/ ((1 + z )5100˚A)] α λ , andthe P-Cygni absorption strength is modelled as e − τ λ where τ λ = τ (2 πσ , obs ) − / e − (( λ − λ abs , obs ) /σ abs , obs ) . scale Fe and z Fe /z H indicate the scaling factor to the Boroson & Green (1992) Fe template spectrum and relative redshift of the Fe template to thehydrogen Balmer narrow line. The details of the fitted spectral model is described in Section 2.11.2. feature is not performed by Koss et al. (2014), thus we con-ducted spectral decomposition analysis of the H β and H α P-Cygni profile for the SDSS and Keck/LRIS spectra.Each spectrum at the wavelength range of λ obs =4400 − β lines, and narrow Gaussian H β and OIII emission lines, where the power-law continuumand broad and very broad H β were absorbed by a factorof e − τ λ where τ λ was modelled as a Gaussian. Since theoptical spectra of SDSS1133 clearly show Fe II line forest(Koss et al. 2014), we also included a Fe II pseudo-continuumtemplate spectrum in the model. Since no template spec- MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 F l u x ( a r b i t r a r il y sc a l ed and s h i ft ed ) Velocity (km/s)SDSS 2003 Mar 9DEIMOS 2013 Dec 13MMT 2014 Jan 3LRIS 2015 July 17LRIS 2019 Mar 10Lowell 2019 May 29
Figure 9.
Long-term time-series of the H β velocity spectra ofSDSS1133. Keck/DEIMOS and MMT spectra are taken fromFig. 9 of Koss et al. (2014), and Lowell Discovery Tele-scope/DeVeny spectrum is taken from Fig. 14 of Ward et al.(2020). The spectra are horizontally shifted so that the narrowH β peaks correcpond to 0 km s − , and the flux is arbitrarilyshifted and scaled for the purpose of clarity. Dotted lines in-dicate continuum levels estimated by linear regression for rela-tively iron-free spectral windows at − , ±
200 km s − and+7 , ±
200 km s − (see Figure 8). Persistent blue-shifted broadabsorption is clearly seen, with the maximum absorption velocityat ∼ − ,
000 km s − . trum for SNe/stellar objects is available in the literature,we adopted the Fe emission template spectrum of Boroson &Green (1992) which is created from a spectrum of the narrowline Seyfert 1 galaxy PG0050+124 (I Zw 1). The Fe emis-sion template spectrum has the intrinsic FWHM of I Zw 1 of900 km s − , and we directly used the Fe II spectrum withoutfurther Gaussian convolution. The Fe template spectrum hasa flux density of 1 . × − erg s − ˚A − at λ rest = 5000 ˚A,and the flux scaling factor to the Fe template spectrum andits redshift are fitting parameters.The same model fitting was also performed at the wave-length range of λ obs = 6200 − χ minimization model fitting was performed by usingthe Levenberg-Marquardt algorithm implemented as opti-mize.least_squares in SciPy . The fitting was performed inthe observed-frame. Uncertainties on the best-fitting param- B a l m e r de c r e m en t H α / H β NarrowBroadVery Broad A λ ( m ag ) Rest-frame wavelength (Angstrom)H β H α SMC (R V =2.93, E(B-V) host =0.628)LMC (R V =3.16, E(B-V) host =0.608)MW (R V =3.08, E(B-V) host =0.650) Figure 10.
Top: The Balmer decrement for the narrow, broad, andvery broad emission line components of SDSS1133. The medianBalmer decrement for the narrow component, H α /H β = 5 .
49, isindicated by the dot-dashed line. Theoretically expected value ofH α /H β = 2 .
86 for Case B recombination H II region with T =10 K and n e = 10 cm − is indicated by the dotted line. Bottom:the Pei (1992)’s SMC, LMC, and MW dust extinction curvesscaled to account for the observed narrow line Balmer decrement. eters were evaluated by 10,000 trials of Monte Carlo resam-pling, where 10,000 mock spectra were generated by addingGaussian flux noise to the original spectrum using the cal-culated standard deviations of the flux density.The fitting parameters are tabulated in Table 3, andthe best-fitting model for the SDSS spectrum is shown inFigure 7. The narrow lines are only marginally spectrally-resolved, thus the narrow velocity widths reflect the instru-mental broadening. The narrow emission line redshift is mea-sured to be z = 0 . β emission line, whilethe Fe II emission at the H α wavelength range is very weakand hard to be seen. The very broad component is not wellconstrained especially at the H β wavelength range, but isrequired to explain the extended wing of the H α profile.The asymmetric P-Cygni-like Balmer line profile canlargely be explained by the combination of the variable MNRAS000
86 for Case B recombination H II region with T =10 K and n e = 10 cm − is indicated by the dotted line. Bottom:the Pei (1992)’s SMC, LMC, and MW dust extinction curvesscaled to account for the observed narrow line Balmer decrement. eters were evaluated by 10,000 trials of Monte Carlo resam-pling, where 10,000 mock spectra were generated by addingGaussian flux noise to the original spectrum using the cal-culated standard deviations of the flux density.The fitting parameters are tabulated in Table 3, andthe best-fitting model for the SDSS spectrum is shown inFigure 7. The narrow lines are only marginally spectrally-resolved, thus the narrow velocity widths reflect the instru-mental broadening. The narrow emission line redshift is mea-sured to be z = 0 . β emission line, whilethe Fe II emission at the H α wavelength range is very weakand hard to be seen. The very broad component is not wellconstrained especially at the H β wavelength range, but isrequired to explain the extended wing of the H α profile.The asymmetric P-Cygni-like Balmer line profile canlargely be explained by the combination of the variable MNRAS000 , 1–29 (2021) M. Kokubo broad blue-shifted absorption on the very broad symmet-ric Balmer line, and the broad blue-shifted absorption alsoaccount for the continuum absorption observed at the blue-ward spectral region of the Balmer lines. The best-fittingmodels for the SDSS and LRIS spectra are shown in Figure 8as a function of rest-frame velocity from − ,
000 km s − to10,000 km s − relative to the narrow Balmer emission lines.The best-fitting absorption model suggests that the absorp-tion peaks at ∼ − ,
000 km s − and the highest velocitycomponent is ∼ − ,
000 km s − . β line profile Figure 9 summarises the multi-epoch H β spectra ofSDSS1133 described at the beginning of Section 2.11. The g band magnitudes at the epochs of the SDSS, DEIMOS,MMT, LRIS 2015, LRIS 2019, and Lowell observations areevaluated by linearly interpolating the g band light curve inFigure 2; g = 18 .
66 (SDSS 2003), 19.65 (DEIMOS 2013),19.65 (MMT 2014), 19.95 (LRIS 2015), 19.38 (LRIS 2019),and 17.36 mag (Lowell 2019), respectively.Figure 9 reveals that the P-Cygni absorption featuresof SDSS1133 have been persistent at least for 16 years since2003, even during the outburst phase in 2019 at the epochof the Lowell observation when SDSS1133 was ∼ − − − ,while the maximum velocity component is roughly kept con-stant at ∼ − ,
000 km s − . The persistent P-Cygni absorp-tion feature suggests that the high velocity ejecta producingthe P-Cygni absorption is not due to an one-off transientevent but is due to continuous or multiple ejections. The dust extinction toward SDSS1133 may be inferred byusing the Balmer decrement, specifically the flux ratio of H α to H β . Figure 10 presents the H α /H β Balmer decrementfor the narrow, broad, and very broad emission line com-ponents of SDSS1133. While the Balmer decrements of thebroad and very broad components are not well constraineddue to the line profile modelling uncertainties, the Balmerdecrements of the narrow component constrain H α /H β tobe ∼ −
6. By combining the three measurements of narrowBalmer decrements, the median and 68% percentile range isH α /H β = 5 . +0 . − . .The theoretically expected value of the Balmer decre-ment is 2.86 for Case B recombination H II region with T = 10 K and n e = 10 cm − (e.g., Osterbrock 1989;Dom´ınguez et al. 2013). This means that the median valueof the relative extinction at the H β and H α wavelengthsis A H β − A H α = 2 . (5 . / .
86) = 0 .
702 mag. Assum-ing Pei (1992)’s Small Magellanic Cloud (SMC)-like, LargeMagellanic Cloud (LMC)-like, and Milky Way (MW)-likedust extinction curves ( R V = A V /E ( B − V ) = 2 .
93, 3 . .
08, respectively), the Balmer decrement of the narrowcomponent in SDSS1133 corresponds to E ( B − V ) host = 0 . +0 . − . (SMC) , . +0 . − . (LMC) , . +0 . − . (MW) , (1) Table 4.
Source properties and possible scenariosSource Property LBV AGNLarge-amplitude UV/optical outbursts YES UnusualDecades-long UV/optical emission YES YESUV/optical luminosity Unusual UnusualX-ray luminosity Unusual NOBroad P-Cygni feature Unusual NONarrow line BPT diagnostics YES NOCa II emission lines YES UnusualNarrow line variability? YES NOYES indicates that the observational feature can be readily ex-plained by the scenario. NO indicates it cannot be explained.Unusual indicates it can be explained but it is unusual of ob-served systems. in the unit of magnitude, respectively (Figure 10). Here E ( B − V ) host ≡ A B, host − A V, host indicates the color ex-cess due to the dust extinction in the SDSS1133’s host en-vironment. The hydrogen column density corresponding tothe derived value of E ( B − V ) host is N H, host = (2 . +0 . − . ,1 . +0 . − . , 0 . +0 . − . ) × cm − for the SMC-like, LMC-like, and MW-like dust-to-gas ratios, respectively (Pei 1992).The inferred large dust extinction ( A V = R V E ( B − V ) =1 . − η Carinae’s dusty Homunculus nebula (e.g., Smith2012; Morris et al. 2017), or interstellar materials locatedalong our line of sight towards SDSS1133. E ( B − V ) host derived above assumes that the hydro-gen Balmer line-emitting region is in the Case B recombina-tion condition. The intrinsic Balmer decrements can be muchlarger than 2.86 when the collisional excitation becomes im-portant and/or the line-emitting region is optically-thick tothe hydrogen Balmer lines (e.g., Osterbrock 1989; Kokuboet al. 2019). In SDSS1133, the narrow Balmer lines maypossibly have multiple origins, and the Balmder decrementmeasurements can be affected by the flux contaminationsfrom the dense circumstellar photo-ionizing region where theintrinsic balmder decrements are larger than 2.86.Also, the assumptions of the dust extinction curves and R V directly affect the estimates of the dust extinction. Sincethe SDSS1133’s host galaxy Mrk 177 is a low-mass, low-metallicity galaxy (Section 3.1), the SMC-like dust extinc-tion with R V = 2 .
93 may provide a good approximationto the interstellar dust extinction in Mrk 177. However, weknow little about the size distribution, dust composition,and dust-to-gas ratio of the circumstellar dust grains possi-bly surrounding SDSS1133. For example, the intense radi-ation from SDSS1133 may preferentially remove small dustgrains and result in a flat extinction curve (e.g., Tazaki et al.2020). Detailed discussion about the properties of the cir-cumstellar dust grain of the SDSS1133 is beyond the scopeof this paper, and will be discussed elsewhere.
In this Section we discuss the nature of SDSS1133 inferredfrom the multi-epoch, multi-band photometric and spectro-scopic measurements described in the previous Section. Asmentioned in Section 1, SDSS1133 is suggested to be ei-
MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 ther an recoiling (offset) AGN or extragalactic outburstingLBV star (=SN impostor). The multiple outbursts and per-sistent emission rule out the possibility of an one-off tran-sient event. Following Koss et al. (2014) and Burke et al.(2020b), we summarize the major observational propertiesof SDSS1133 and evaluate each scenario in Table 4. Thoughsome of the source properties are quite unusual and rarelyobserved among LBVs, the LBV scenario is favored overthe AGN scenario. We conclude that SDSS1133 belongs toan extreme population of LBVs experiencing multiple non-terminal explosions and subsequent strong interactions ofthe ejected shell with different shells and/or CSM. Detailsof these source properties and comparisons with other gianteruption LBVs and SN impostors are discussed below. The SDSS1133’s host galaxy Mrk 177 is a dwarf galaxy withthe g band absolute magnitude of M g = − . M ∗ ∼ . M (cid:12) (Koss et al. 2014). Thedisturbed morphology of Mrk 177 indicates that Mrk 177is a post-merger galaxy (see Koss et al. 2014, for details).The gas-phase oxygen abundance is estimated to be 12 +log(O/H) = 8 .
58 from the SDSS optical spectroscopy of theMrk 177’s nucleus. The low metallicity galaxies are preferredsites for massive star formation, thus the low metallicity inMrk 177 may be consistent with the idea that the progenitorof SDSS1133 is a massive star like other LBVs.Under the LBV scenario, SDSS1133 is a massive starand the local environment of SDSS1133 should be an ac-tive star-forming region. The local environment of SDSS1133cannot be examined by current observations, and furtherhigh spatial resolution observations for the emission linesand dust emission around SDSS1133 are needed to judgethe validity of the LBV scenario based on the host galaxy’slocal environment.Koss et al. (2014) estimate a star-formation rate ofMrk 177 as 0 . M (cid:12) yr − using GALEX
UV photome-try.
W ISE magnitudes of Mrk 177 are 13.25, 13.20, 10.03,and 7.68 Vega mag at W1, W2, W3, and W4 band, respec-tively, and the WISE colors W1 − W2=0.05 Vega mag andW2 − W3=3.17 Vega mag are outside of empirical AGN se-lection windows (Hainline et al. 2016). The SDSS opticalspectrum of the nucleus of Mrk 177 exhibits star-forminggalaxy-like emission line ratio of [NII]/H α and [OIII]/H β ,indicating that Mrk 177 does not harbor a strong AGN atthe galactic center (Koss et al. 2014; Toba et al. 2014). Also,the weak X-ray luminosity of L . −
10 keV ∼ × erg s − at the nucleus of Mrk 177 observed by Chandra (Section 2.8)is well below the empirical threshold luminosity to identifyAGNs ( ∼ × erg s − ), and can be explained by in-tegrated luminosities from X-ray binary populations in thegalaxy without the need for AGN (e.g., Lehmer et al. 2016;Birchall et al. 2020; Schirra et al. 2020). The recoiling AGNscenario requires that SDSS1133 is a merged SMBH thatwas ejected from the center of Mrk 177 at the time of theSMBHB coalescence, and it requires that the Mrk 177’s nu-cleus currently possesses no SMBH. The absence of an AGNin the nucleus of Mrk 177 is not inconsistent with the re-coiling AGN scenario, though the observations do not ex- clude the possibility of the presence of an inactive SMBH inMrk 177.SDSS1133 is located at a galactocentric distance of0.81 kpc. The narrow line redshift of SDSS1133 is consistentwith that of Mrk 177’s nucleus, indicating that the line-of-sight velocity of SDSS1133 relative to the systemic velocityis currently much less than 100 km s − (Section 2.11.2).Under the recoiling AGN scenario, the required life-time ofthe AGN activity of SDSS1133 to explain the large galacto-centric distance is roughly at least 0 .
81 kpc /
100 km s − ∼ yrs, which is somewhat longer than the life-time ex-pected for recoiling AGNs (e.g., Blecha & Loeb 2008; Wardet al. 2020). Thus, the small relative velocity of SDSS1133disfavors the recoiling AGN scenario. Figure 11 shows the 70 years worth of absolute magnitudelight curve of SDSS1133, which include the POSS-I andPOSS-II photographic plate measurements in 1950, 1994,and 1999 (see Section 2.10). Except for the POSS data,the absolute magnitudes from the g band measurementsare shown. Two light curves are shown; one is uncorrectedand the other us corrected for the possible extinction inthe host galaxy estimated from the Balmer decrement (Sec-tion 2.11.4). We should note that while the brief 2014 out-burst indicated in Figure 2 is not sampled with the g bandand does not appear in Figure 11.SDSS1133 has been a bright variable object at leastsince 1950, as first noticed by Koss et al. (2014). In ad-dition to the outbursts, SDSS1133 also exhibits small am-plitude flux variations even in the non-outbursting phase.After 1994, SDSS1133 light curve probably shows a long-term declining trend since 1994. We calculate the long-termlinear trend between 1994 and 2020 by fitting a straightline for the data points at MJD > < MJD < < MJD < − and 0.059 mag yr − for host extinction uncorrected and cor-rected light curves, respectively.As reference models, consider simple SN light curvemodels. Late-phase SN luminosity can be powered eitherby radioactive decays in cobalt or SN ejecta-CSM shock in-teractions. The former light curve is approximately an ex-ponential form of L ( t ) ∝ e − t/τ Co where τ Co = 111 . Co, and the latter can be approximatelya power-law form of L ( t ) ∝ t − / ( n − (assuming a steadystellar wind CSM) where n = 7 −
12 depends on the radialdensity structure of the ejecta ( ρ ej ∝ r − n ; e.g., Moriya et al.2013b). In either case, a single SN exploded somewhere inthe 20th century cannot explain the very slow decline trendof the long-term light curve of SDSS1133. For the same rea-soning, we can conclude that the non-outbursting phase lightcurve of SDSS1133 cannot be attributed to a tidal disrup-tion of a star by a SMBH ( L ( t ) ∝ t − / ; Phinney 1989), orin general to any other known one-off explosive events.AGNs are known to show decades-long optical variabil-ity at amplitudes of a few 0.1 mag (e.g., MacLeod et al.2012), thus the long-term light curve in the non-outburstingphase of SDSS1133 alone cannot exclude the possibility of MNRAS , 1–29 (2021) M. Kokubo -18-16-14-12-10-8 -20000 -15000-65 -60 -55 -50 -45 -40 A n s o l u t e m agn i t ude M AB = m AB - µ DaysYearsPOSS-I -4000 -3000 -2000 -1000 0 1000 2000 3000 4000 5000 6000 7000 8000 9000-10 -5 0 5 10 15 20 25DaysYearsPOSS-II Host extinction correction (SMC)No host extinction correction η Car(Great Eruption)SN 2009ip(precursor) SN 2009ip(explosion)iPTF14hls
Figure 11.
The historical light curve of SDSS1133. The horizontal axis is the observed-frame days since the peak epoch (2001 December18). The upper light curve (open symbols) is corrected for the host galaxy extinction assuming SMC-like extinction (Section 2.11.4).Square symbols indicate the POSS-I/II photographic plate photometries, and circle symbols indicate the g band PSF light curve M g inFigure 2. Note that the brief 2014 outburst indicated in Figure 2 is not sampled with the g band. The long-term linear decline ratesbetween 1994 and 2020 are 0.068 mag yr − and 0.059 mag yr − for host extinction uncorrected and corrected light curves, respectively(Section 3.2). The V band light curve for the η Carinae’s Great Eruption (peaks in 1845), pseudo R band light curve of pre-SN eruptions(peaks on 2009 August 28) and subsequent terminal SN explosion of SN 2009ip in September 2012, and g band light curve of iPTF14hls(peaks on 2015 May 22) for comparison. the AGN origin. However, the co-existence of the decades-long trend and short time-scale large-amplitude outburstsis rarely seen in AGNs. As shown below, the long-lastinglight curves and short time-scale outbursts in SDSS1133 arereminiscent of the LBV variability. As shown in Figures 2 and 11, we find that SDSS1133 ex-perienced at least three optical outbursts since 2001; thesethree outbursts are referred to as 2001-2002 outburst, 2014outburst, and 2019 outburst, respectively.SDSS1133 was in a very bright outbursting phase dur-ing the SDSS imaging observations in 2001 − g = 16 . g = 18 . ± . r = 18 . ± . i = 18 . ± .
05, and z = 18 . ± .
08 mag(Koss et al. 2014), which are ∼ − − − . . g -band magnitude peaksat M g = − .
30 mag ( M g = − .
26 mag after correctingfor the possible host galaxy extinction), thus the 2001-2002outburst event can be regarded as a SN impostor (Pastorelloet al. 2013; Pastorello & Fraser 2019; Perley et al. 2020).The PTF and PS1 photometry reveals that SDSS1133experienced a brief optical outburst between 2014 March8 (MJD=56738.5; PTF R = 18 .
99 mag) and 2014 July 8(MJD=56846.3; PS1 y = 18 .
65 mag). The observed peakmagnitude is PS1 i = 18 . ± .
01 mag on 2014 May
MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133
10 (MJD=56787.3). The duration of the 2014 outburst isroughly constrained to be 56846 . − . . − g , r , and i band observations provide thedensely-sampled multi-band light curves for the 2019 out-burst (Figure 3). The g and r band light curves peak on 2019June 5 (MJD=58639.2) at g = 17 .
03 and r = 16 .
81 mag,about 3 mag brighter than the non-outbursting phase. Theabsolute g -band magnitude at the peak is M g = − .
27 mag( M g = − .
46 mag after correcting for the possible hostgalaxy extinction), thus the 2019 outburst should be re-garded as another SN impostor event in SDSS1133 after the2001-2002 outburst. We can see from Figure 3 that the dura-tion of the entire 2019 outburst is about ∼
100 days (see alsoWard et al. 2020), thus the 2019 outburst is ∼ . − . g -band light curve, we obtain t rise = 23 . t fade = 3 . t rise + t fade = 27 . η Carinae analog giant eruptions. The optical outbursts inthe S Dor-type LBV eruptions are associated with the de-crease of the temperature so that the bolometric luminosityis unchanged while the photospheric radius increases (Fig-ure 15; Wolf 1989; Vink 2011; Smith et al. 2011; Pastorelloet al. 2013; Smith 2014, 2017b; Kilpatrick et al. 2018, andreferences therein), which is not the case for SDSS1133’soutbursts. Instead, we suggest that the flux variations inSDSS1133 are due to the intrinsic increase in the bolometric luminosity as observed in some of η Carinae analog gianteruption LBVs = SN impostor, e.g., pre-SN outbursts ofSN 2009ip (e.g., Humphreys et al. 1999; Smith et al. 2011;Pastorello et al. 2013; Davidson 2020; Weis & Bomans 2020,and references therein). We will return to this point in Sec-tion 3.4.
In Figure 11, the absolute g -band magnitude light curve ofSDSS1133 is compared with the historical V band light curveof the 19th century Great Eruption of η Carinae’s compiledby Frew (2004) and Smith & Frew (2011). The original lightcurve is shifted by 0.02 mag to match the AB magnitudesystem, and the absolute magnitude is calculated by assum-ing a distance to η Carinae as 2.3 kpc ( µ = 11 .
81 mag), and A V = 1 . V -band magnitude of η Carinae had been in a range between − −
11 mag be-fore 1822 and reached ∼ −
14 mag during the Great Eruptionduring 1822 − − . η Carinae’s brief precursor eruptions in 1838 and 1843 justbefore the final brightening in 1844 December (see Smith &Frew 2011). Figure 11 suggests that SDSS1133 remains tobe brighter than the η Carinae’s Great Eruption over theperiod since 1950. Since the total mass-loss from η Carinaeduring the Great Eruption is estimated to be about ∼ M (cid:12) (Smith & Frew 2011), the mass-loss from SDSS1133 shouldbe much larger than 10 M (cid:12) if we assume the mass-loss rate( ∝ kinetic energy) is roughly in proportion to the radiationenergy. Also this may possibly mean that the SDSS1133’scurrent mass is more massive than η Carinae, i.e., > M (cid:12) (e.g., Smith 2008).We also show a R band absolute magnitude light curveof SN 2009ip (Smith et al. 2010; Mauerhan et al. 2013; Pa-storello et al. 2013) in Figure 11. The R band light curvein Figure 11 is constructed from R band, F606W band, andunfiltered data between 1999 June 29 and 2009 August 28in Smith et al. (2010), combined with the R band data be-tween 2009 August 30 and 2012 October 16 in Pastorelloet al. (2013). The absolute magnitude is calculated by as-suming a distance modulus to SN 2009ip as µ = 31 .
55 mag,and A R = 0 .
05 mag (Mauerhan et al. 2013; Pastorello et al.2013), and the light curve is shifted by 0.21 mag to matchthe AB magnitude system. SN 2009ip is believed to have ex-ploded as a terminal SN (namely SN IIn) in 2012 September(but see Pastorello et al. 2013), and before that there weremultiple precursor eruptions. The first identified eruptionwas in 2009 August-September whose peak was on 2009 Au-gust 28 (MJD=55071.75; m R = 17 mag; Smith et al. 2010).The second eruption was observed in 2011 May-October,and then in 2012 there were two large eruptions (2012aevent from August 8 to September 24, and 2012b eventfrom September 24 to December; Smith et al. 2010; Pas-torello et al. 2013; Margutti et al. 2014). The 2012b eventis considered to be the terminal SN explosion of SN 2009ip,and the pre-SN eruptions (observed as SN impostor) aresuggested to be originated from the LBV giant eruptionsof the progenitor star (Mauerhan et al. 2013; Smith et al.2014; Graham et al. 2014). The peak absolute magnitude ofSN 2009ip during the eruptions (2009 and 2012a events) is MNRAS , 1–29 (2021) M. Kokubo similar to that of η Carinae’s giant eruptions, and thus veryclose to the brightness of SDSS1133 (Figure 11).Another interesting object to be compared withSDSS1133 is iPTF14hls. iPTF14hls is a hydrogen-rich opti-cal transient at z = 0 . d L = 156 Mpc), exhibiting mul-tiple peaks before and after the maximum epoch, and lastingfor 2 years (Arcavi et al. 2017; Yalinewich & Matzner 2019;Sollerman et al. 2019). The spectrum shows broad hydro-gen Balmer P-Cygni absorption features at − − but no signs of narrow P-Cygni features (Arcavi et al.2017), similar to SDSS1133. Although the absolute magni-tude reaches over −
18 mag, the peculiar light curve shapeof iPTF14hls is inconsistent with normal SNe. There is asuggestion that iPTF14hls is not a terminal SN explosion;specifically, Moriya et al. (2020) suggest that iPTF14hls isa non-terminal eruption event like η Carinaes’s Great Erup-tion. Arcavi et al. (2017) report that iPTF14hls was detectedon the POSS-I image on 1954 February 23 at 20 . ± . g band magnitudelight curve of iPTF14hls, where the Galactic extinction co-efficient is assumed to be A g = 0 .
05 mag. The brightnessof iPTF14hls is consistent with SDSS1133 if the large hostgalaxy extinction of E ( B − V ) host = 0 .
628 (SMC-like; Equa-tion 1) of SDSS1133 is taken at its face value. The time scaleof the light curve evolution of iPTF14hls is similar to the2001-2002 outburst of SDSS1133.Figure 11 shows that there exists a vast diversity in thelight curve properties of the LBV giant eruptions (and gianteruption candidates), and at least we can say that the long-term and outburst properties of SDSS1133 are within thediversity in the LBV giant eruptions.
The similar brightness and variability time scales betweenSDSS1133 and other giant eruption LBVs suggest a commonphysical processes behind these objects. The best studiedobject among the giant eruption LBVs is η Carinae, whichis suggested to be a binary system composed of a ∼ M (cid:12) LBV star and O-type main sequence star, with a orbitalperiod of 5.5 yrs (Damineli 1996; Parkin et al. 2009; Smith &Frew 2011). The giant eruptions of η Carinae are suggestedto have occurred at the times of pericenter passage of thecompanion star (Smith & Frew 2011).The brightness and durations of each of the outburstsin SDSS1133 are different with each other (Section 3.2.2),which may imply that the outbursts in SDSS1133 cannotbe explained by the simple binary model. Nevertheless, ifSDSS1133 is a binary system, the binary orbital period maybe inferred from the recurrence timescale of the observedoutbursts.We observed three large outbursts from SDSS1133 in2001-2002, 2014, and 2019; specifically, the observed peakepochs are MJD=52261.4 (SDSS on 2001 December 18),56787.3 (PS1 i band on 2014 May 10), and 58639.2 (ZTFon 2019 June 5), respectively. The observations before 2014were scarce, and we may have missed other outburstsbetween 2003 and 2014. The time interval between the2014 and 2019 outbursts is ∆ t − , obs = 1851 . P in SDSS1133 is, if it exists, P = (1851 . /n ) days where n is integer. Assuming thatthe period remains unchanged during the observations, thetime interval between the 2001-2002 and 2014 outbursts(∆ t / − , obs = 4525 . P . In a range of n <
10, we find n = 2( P = 926 . . . The results of the spectral decomposition analysis in Ta-ble 3 indicate that the low-ionization nebular narrow lines([OIII],[NII], and [SII]) seem to display flux variations overthe spectroscopic periods (see also Pursimo et al. 2019a).The flux variations of the narrow lines are rarely observed inAGNs since the narrow line regions in AGNs are extendedover ∼ kpc scales. The narrow line variability may implythat these lines in SDSS1133 are originated from circumstel-lar materials in close vicinity of the central hot star. But weshould note that careful spectroscopic analyses (especiallyhost galaxy subtraction and slit-loss corrections) are neededto examine the significance of the narrow line flux variabilityof SDSS1133, which is beyond the scope of this paper.The narrow [SII] doublet ratio [SII]6717/[SII]6731 isconsistent at ∼ .
4, indicating that the electron density ofthe narrow emission line region is consistent with typicalelectron densities for local galaxies, n e = 10 −
100 cm − (Osterbrock 1989).Figure 12 presents the Baldwin-Phillips-Terlevich(BPT) narrow line diagnostic diagrams for SDSS1133 (Bald-win et al. 1981; Kewley et al. 2001, 2006; Kauffmann et al.2003). The [NII] narrow emission lines of SDSS1133 aremuch weaker compared to the H α narrow line in SDSS1133,while the narrow line ratio of log[OIII]5007/H β is relativelyhigh compared to typical star-forming galaxies. Also, the[OI]6300/H α and [SII]6717,6731/H α narrow line ratios aremuch less than 1. According to the BPT diagrams, the nar-row line ratios of SDSS1133 are consistent with low-massstar-forming galaxies, and suggest that the narrow emissionline region of SDSS1133 has a high ionization parameter andlow-metallicity (e.g., Kewley et al. 2013; Simmonds et al.2016). The BPT diagnostics for SDSS1133 disfavor the stan-dard AGN ionization, as already pointed out by Koss et al.(2014) and Simmonds et al. (2016).However, we should note that, since the narrow emis-sion lines are not spectrally-resolved, the narrow lines inSDSS1133 are probably originated from multiple regions,such as the host galaxy H II region and photo-ionized cir-cumstellar region. In that case, the interpretation of theBPT diagnostics is ambiguous. MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 -1-0.5 0 0.5 1 1.5 -2 -1.5 -1 -0.5 0 0.5 l og ( [ O III] / H β ) log([NII]6584/H α )HII AGNComposite -1 -0.5 0 0.5log([SII]6717,6731/H α )HII AGN LINER -2 -1.5 -1 -0.5 0log([OI]6300/H α )HII AGN LINERSDSS 2003 Mar 9LRIS 2015 July 17LRIS 2019 Mar 10 Figure 12.
BPT classification diagrams for the narrow emission line ratios of SDSS1133. The SDSS and two Keck/LRIS measurements areindicated by filled circles, open circles, and filled squares, respectively. The extreme starburst lines (solid lines; Kewley et al. 2001), purestar-formation line (dotted line; Kauffmann et al. 2003), and Seyfert-LINER (low-ionization nuclear emission-line region) classificationlines (dot-dashed lines; Kewley et al. 2006) are shown. F l u x ( - e r g / s / c m / A ng s t r o m ) Rest-frame wavelength (Angstrom)OI CaII CaII CaIISDSS 2003 Mar 9Bestfit model F l u x ( - e r g / s / c m / A ng s t r o m ) Rest-frame wavelength (Angstrom)OI CaII CaII CaIILRIS 2015 July 17Bestfit model F l u x ( - e r g / s / c m / A ng s t r o m ) Rest-frame wavelength (Angstrom)OI CaII CaII CaIILRIS 2019 Mar 10Bestfit model
Figure 13.
The SDSS and Keck spectra at the Ca II IR λλ , , As shown in Figure 7, the optical spectra of SDSS1133are heavily contaminated by iron emission line forests.In addition to the iron emission lines, another interest-ing spectroscopic feature of SDSS1133 is intermediate-width or narrow calcium emission lines clearly detectedin the optical spectra, as already reported by Koss et al.(2014) and Ward et al. (2020). In the LRIS spectra,[Fe II] λ λλ , λ λλ , , λ λλ , n e ∼ cm − (e.g., Humphreys et al. 2013).Figure 13 shows a power-law + Gaussian emission linemodel fitting to the SDSS and LRIS spectra at the Ca tripletspectral region. The rest-frame velocity widths are 386.45,172.67, and 170.43 km s − for the SDSS and two LRIS spec-tra (the later two are spectrally-unresolved), respectively,suggesting that the line widths of these lines are variable.The emission strength ratio of Ca II triplets is roughly 1:1:1rather than the theoretically predicted value of 1:9:5, sug-gesting that the line-forming region is optically-thick (e.g.,Apparao & Tarafdar 1988; Banerjee et al. 2021). The re-semblance between the Ca II and Fe II line widths in theSDSS spectrum implies that Ca II and Fe II emission linesare originated from the same, high density photo-ionizationregion. The broader line widths of Ca II and Fe II comparedto the other H, N, S, and O nebular narrow lines in the SDSSspectrum (Table 3) suggest that the Ca II and Fe II emis-sion region is separated from the low density, low velocitydispersion nebular responsible for the nebular narrow lines,and is located in closer vicinity to SDSS1133. The narrowCa II lines are occasionally seen in stellar explosion eventsand LBVs (e.g., Smith et al. 2010; Foley et al. 2011; Wardet al. 2020) while are rarely seen in AGNs, favoring the sce-nario that SDSS1133 is LBV-origin (e.g., Koss et al. 2014;Ward et al. 2020). As shown in Section 2.11.2, the hydrogen Balmer linesin SDSS1133 exhibit multi component P-Cygni profiles,
MNRAS000
MNRAS000 , 1–29 (2021) M. Kokubo composed of an unresolved narrow component, FWHM= 1 , − ,
000 km s − broad component, FWHM ∼ ,
000 km s − very broad wing component, and blue-shiftedbroad absorption component. The spectral changes in thebroad line profile can be reproduced by the model wherelarge variations in the absorption component are allowed.As shown in Figure 9, the spectroscopic properties of the H β and H α profiles of SDSS1133 remain essentially unchangedduring the epochs of the spectroscopic observations since2003. The P-Cygni absorption feature is clearly detectedwith the absorption peak at ∼ − ,
000 km s − , and theabsorption feature continuously extends from ∼ − up to ∼ − ,
000 km s − (Figure 8).The optical spectra of SDSS1133 resemble those of typeII SNe (Pursimo et al. 2019a), but the long-duration lightcurve and multiple outbursts observed in SDSS1133 rule outthe possibility that SDSS1133 is an one-off event, and arein favor of the LBV scenario (see also Ward et al. 2020).Under the assumption that SDSS1133 is a LBV star, thepersistent broad absorption feature implies that the ejectedmaterials are moving at a high velocity of ∼ − ,
000 km s − (up to − ,
000 km s − ). Observations for giant eruptionLBVs (namely η Carinae and SN 2009ip’s pre-SN eruptions)have revealed that at least a fraction of outflowing materialsejected during the LBV giant eruptions can have a veloc-ity higher than 5 ,
000 km s − (e.g., Smith 2008; Foley et al.2011; Pastorello et al. 2013; Mauerhan et al. 2013). The per-sistent broad P-Cygni absorption seen in SDSS1133 suggeststhat the mass-loss in SDSS1133 is one of the most energeticamong the known examples of the LBV giant eruptions (Pa-storello et al. 2013). Such high velocity mass ejections (mov-ing faster than the star’s escape velocity) are hard to be ex-plained solely by the radiation-driven winds of the star, andthey imply the existence of non-terminal explosions in thestellar core inducing outward blast waves into the envelope,which are observed like scaled-down SNe (e.g., Smith 2008;Pastorello et al. 2013; Smith 2013, 2014, 2017b; Smith et al.2014, 2018; Strotjohann et al. 2020); see Section 3.3.4.About 10% of broad line AGNs are known to showblue-shifted broad absorption line (BAL) features in theUV wavelengths due to line absorption by high velocity diskwinds (e.g., Hall et al. 2002). However, only a few examplesof hydrogen Balmer BAL AGNs are known to date, becausethe disk winds are generally too low-density and too highly-ionized to keep neutral hydrogen atoms populated in the n = 2 shell (e.g., Hall 2007; Zhang et al. 2015; Kokubo 2017;Burke et al. 2020a). Considering the intrinsic rarity of theBalmer BAL AGNs, it is unreasonable to assume that the P-Cygni features of SDSS1133 (selected as a rare offset AGN)are due to the Balmer BAL nature. We should note that thediscovery of SDSS1133 suggests that a fraction of BalmerBAL AGN candidates known to date (e.g., SDSS J1025 inBurke et al. 2020a) can actually be SDSS1133-like LBVs. Most LBVs generally exhibit narrow hydrogen Balmer emis-sion lines from the stellar winds with the velocity of a few100 km s − , reflecting the low escape speed of blue super-giants stars (Smith 2014; Davidson 2020). In contrast to thenormal LBVs, the line width of the broad emission lines inSDSS1133 is much broader than 1,000 km s − ; the FWHM of the broad component is FWHM b = 2 √ σ b ∼ , − ,
000 km s − , and very broad component is FWHM vb =5 , − ,
000 km s − , which are comparable to the linewidths observed in SNe.As with the broad P-Cygni absorption, the broadBalmer emission lines of SDSS1133 are in common withthe pre-SN outbursts of SN 2009ip (e.g., Smith 2008; Fo-ley et al. 2011; Pastorello et al. 2013; Mauerhan et al. 2013;Ward et al. 2020). The broad P-Cygni absorption feature im-plies the presence of the high velocity outflowing materials(Section 3.3.3), and the broad emission lines are presumablyemitted from the same high velocity materials. As alreadymentioned in Section 3.3.3, although the relatively narrowlines ( (cid:46) ,
000 km s − ) observed in typical LBV eruptionscan be attributed to stellar winds (e.g., Smith 2014; David-son 2020), the broad lines of > ,
000 km s − observed inSDSS1133 and pre-SN outbursts of SN 2009ip are too broadto be explained by the same wind mechanism (e.g., Smith2008; Pastorello et al. 2013; Smith 2014), suggesting thatthe broad line-emitting fast materials are originated fromthe non-terminal explosions (Smith 2008, 2013; Smith et al.2018).It is known that SNe IIn exhibit broad/very broad emis-sion lines and absorption features originated from ejecta-CSM interaction regions, where the ejecta kinematics andelectron scattering broaden the emission lines up to ∼ − (e.g., Chugai 2001; Kokubo et al. 2019). Theblue optical continuum and broad and narrow emission linesproduced by the ejecta-CSM interactions in SNe IIn areknown to closely resemble those of type 1 AGNs (Filippenko1989; Baldassare et al. 2016), as observed in SDSS1133. Wecan expect that a similar mechanism can produce broademission and absorption features in the case of mass-lossejecta-CSM interactions or collisions between shells of mat-ter ejected from a LBV (e.g., Woosley et al. 2007; Dessartet al. 2009; Yoshida et al. 2016; Woosley 2017). Smith (2013)suggests a model for the 19th century eruption of η Carinaewhere a strong wind blowing for 30 yrs and subsequent non-terminal ∼ erg explosion occurring in 1844 result inejecta-CSM interactions like a scaled-down SN IIn (see alsoSmith 2008; Smith et al. 2018). If SDSS1133 is an eventsimilar to these η Carinae analog giant eruptions, the per-sistent blue UV-optical continuum and broad emission linesfrom SDSS1133 can naturally be explained by these non-terminal explosions and subsequent shock interactions. Theshock interactions also produce X-ray photons (Chevalier1982; Chevalier et al. 2006; Dessart et al. 2009; Dwarkadas& Gruszko 2012; Smith 2013), and X-ray emission escapedfrom the shock region can naturally explain the observedX-ray luminosity of SDSS1133 (see Section 3.4.4). We willdiscuss the energetics of this non-terminal explosion scenarioin detail in Sections 3.5.SDSS1133 shows no signs of narrow P-Cygni absorp-tion features. SNe IIn with strong ejecta-CSM interactionstypically show the narrow P-Cygni absorption, which is at-tributed to unshocked pre-SN stellar wind components (e.g.,Taddia et al. 2013). But observationally, some SNe IIn lacknarrow P-Cygni absorption features (e.g., Kokubo et al.2019), which is probably due to viewing angle effects causedby asphericity in the CSM matter distribution. Therefore,under the scaled-down SN IIn scenario, the absence of nar-
MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 λ L λ ( e r g / s ) Rest-frame wavelength (Angstrom)log(T eff [K])=4.04log(L[erg/s])=41.04log(T eff [K])=4.03log(L[erg/s])=41.67 UVOT+ZTF (2019-5-27/29)SMC corr.LMC corr.MW corr.UVOT+PS1 (2014-2-6/25)SMC corr.LMC corr.MW corr.
Figure 14.
UV-optical SEDs of SDSS1133 on 2014 February 6-25 (outbursting phase) and 2019 May 27-29 (outbursting phase).The host galaxy extinction-corrected SEDs assuming the extinc-tion curves determined by the Balmer decrement (Figure 10) arealso shown. The best-fitting black body spectra are overplotted;the extinction-corrected SEDs are consistent with Rayleigh-Jeanspower-law ( λL λ ∝ λ − ). row P-Cygni absorption may imply the aspherical CSM in-teractions in SDSS1133. The UVOT data obtained on 2014 February 6-25 (non-outbursting phase) and 2019 May 27-29 (during the 2019outburst) are useful to constrain the UV-to-optical SEDof SDSS1133. While the 2019 UVOT data cover the fullUV wavelength range with the UVW2, UVM2, and UVW1bands, the 2014 UVOT data are obtained only with theUVW2 band. Since simultaneous optical broad band pho-tometries were not available at the epochs of the UVOTobservations, ZTF g , r , and i band and PS1 z and y bandmagnitudes at the epochs of the UVOT observations wereestimated by linearly-interpolating the light curves.Assuming a spherical black body radiator with a radius R , the observed flux is modelled as f ν, obs = (cid:18) rd L (cid:19) (1 + z ) πB ν, rest ( T eff ) . (2)There are two free parameters; T eff and overall scaling factor C ≡ ( r/d L ) . The bolometric luminosity L can be evaluatedfrom T eff and C : L = 4 πd L Cσ SB T , (3)from the Stefan-Boltzmann law, (cid:82) πB ν ( T eff ) dν = σ SB T ,where σ SB is the Stefan-Boltzmann constant and B ν is the l og ( L / L s un ) Temperature (K) η Car (today) η Car (1890) η Car (Great Eruption)UGC2773-OTSN2009ip (precursor)SDSS1133 HD-limit S D o r i n s t ab ili t y s t r i p ZAMS
S Dor-type LBVGiant eruption LBV
Figure 15.
The H-R diagram for SDSS1133 in 2014 (non-outbursting phase) and 2019 (outbursting phase), compared withknown S Dor-type LBVs and giant eruption LBVs taken fromRest et al. (2012). Except for η Carinaes, the same objects areconnected with dotted lines. The S Dor instability strip (LBVs inquiescence) and constant temperature eruptive LBV strip (LBVsin outburst) are indicated by gray bands, and the empiricalHumphreys − Davidson (HD) limit is indicated by a solid line.The dashed-dotted line indicates the Zero-Age Main Sequence(ZAMS) for stars with M = 25 − M (cid:12) (no rotation and Z = Z (cid:12) ; Ekstr¨om et al. 2012). The temperatures and luminositiesfor SDSS1133 and SN 2009ip’s pre-SN eruption in 2009 are lowerlimits because of the possible effects of host galaxy reddening. black body intensity. Therefore, there two fitting parame-ters, T eff and L . The SED fitting was performed with the χ minimization, where the model broad band magnitudescalculated by the filter convolution were compared to theobserved magnitudes.Figure 14 shows the UV-optical SEDs of SDSS1133 on2014 February 6-25 (non-outbursting phase) and 2019 May27-29 (outbursting phase). The SEDs corrected for possiblehost galaxy extinction assuming the extinction curves deter-mined by the Balmer decrement (Figure 10; Section 2.11.4)are also shown. The best-fitting black body spectra areoverplotted. The best-fitting black body parameters for theSEDs with no host galaxy extinction correction are T eff , L =10 . K , . L (cid:12) and 10 . K , . L (cid:12) for the 2014 and2019 data, respectively. The corresponding black body ra-dius is R = 10 . R (cid:12) and 10 . R (cid:12) , respectively.This fitting implies that the effective temperature ofSDSS1133 did not vary significantly between the outburst-ing and non-outbursting phases. As already discussed in Sec-tion 3.2.2, this is in contrast to normal LBV eruptions orS Dor-type variables. The eruptions in the normal LBVscan be explained by the outward motion of the photosphereeither due to a true increase in the star’s hydrostatic ra- MNRAS000
The H-R diagram for SDSS1133 in 2014 (non-outbursting phase) and 2019 (outbursting phase), compared withknown S Dor-type LBVs and giant eruption LBVs taken fromRest et al. (2012). Except for η Carinaes, the same objects areconnected with dotted lines. The S Dor instability strip (LBVs inquiescence) and constant temperature eruptive LBV strip (LBVsin outburst) are indicated by gray bands, and the empiricalHumphreys − Davidson (HD) limit is indicated by a solid line.The dashed-dotted line indicates the Zero-Age Main Sequence(ZAMS) for stars with M = 25 − M (cid:12) (no rotation and Z = Z (cid:12) ; Ekstr¨om et al. 2012). The temperatures and luminositiesfor SDSS1133 and SN 2009ip’s pre-SN eruption in 2009 are lowerlimits because of the possible effects of host galaxy reddening. black body intensity. Therefore, there two fitting parame-ters, T eff and L . The SED fitting was performed with the χ minimization, where the model broad band magnitudescalculated by the filter convolution were compared to theobserved magnitudes.Figure 14 shows the UV-optical SEDs of SDSS1133 on2014 February 6-25 (non-outbursting phase) and 2019 May27-29 (outbursting phase). The SEDs corrected for possiblehost galaxy extinction assuming the extinction curves deter-mined by the Balmer decrement (Figure 10; Section 2.11.4)are also shown. The best-fitting black body spectra areoverplotted. The best-fitting black body parameters for theSEDs with no host galaxy extinction correction are T eff , L =10 . K , . L (cid:12) and 10 . K , . L (cid:12) for the 2014 and2019 data, respectively. The corresponding black body ra-dius is R = 10 . R (cid:12) and 10 . R (cid:12) , respectively.This fitting implies that the effective temperature ofSDSS1133 did not vary significantly between the outburst-ing and non-outbursting phases. As already discussed in Sec-tion 3.2.2, this is in contrast to normal LBV eruptions orS Dor-type variables. The eruptions in the normal LBVscan be explained by the outward motion of the photosphereeither due to a true increase in the star’s hydrostatic ra- MNRAS000 , 1–29 (2021) M. Kokubo dius or the radius of a pseudo-photosphere in the opaquewind; the optical outbursts are accompanied with lower-ing of effective temperature so that the bolometric lumi-nosity remains nearly constant (see Figure 15; Vink 2011;Smith 2011; Smith et al. 2011; Van Dyk & Matheson 2012;Humphreys et al. 2014). The constant effective temperatureduring the eruptions in SDSS1133 is more similar to LBVgiant eruptions, like the 19th century Great Eruption of η Carinae (e.g., Davidson & Humphreys 1997; Smith et al.2011; Smith 2013, 2014, 2017b).Although the effective temperature and bolometric lu-minosity from the fitting for the host galaxy extinction-uncorrected SEDs are physically reasonable, the observedSED shape is much redder than the black body spectra,probably being consistent with the large Balmer decrementsdescribed in Section 2.11.4. Actually, as shown in Figure 14,the host galaxy extinction correction make the SEDs con-sistent with Rayleigh-Jeans power-law ( λL λ ∝ λ − ). If thehost galaxy extinction coefficients are taken at their facevalues, the weak UV turnover (if any) in the SEDs in Fig-ure 14 prevents us from determining the effective tempera-ture. This implies that the true temperature of SDSS1133 ismuch higher than 10 K. In this sense, T eff and L derivedfor the host galaxy extinction-uncorrected SEDs should betaken as the lower limits, as indicated in Figure 15. Also, tak-ing into account the uncertainties in the host galaxy extinc-tion estimates (Section 2.11.4), the host galaxy extinction-corrected SEDs shown in Figure 14 should be regarded asthe upper limits.We should note that the artificial excess emission in theUVM2 band seen in the host galaxy extinction-correctedSEDs in Figure 14 is due to the 2175˚A bump in theextinction curves. Assuming that the intrinsic continuumSEDs should be smooth, the 2175˚A extinction bump towardSDSS1133 should be suppressed than the MW, SMC, LMCextinction curves. Since the 2175˚A extinction bump is con-sidered to be caused by Polycyclic Aromatic Hydrocarbons(PAHs) and/or small graphite dust grains (e.g., Draine &Lee 1984; Tazaki et al. 2020), the weak 2175˚A extinctionbump may indicate that these small grains are eliminatedtoward the line of sight, probably due to the strong emis-sion of SDSS1133. Figure 15 shows the Hertzsprung-Russell (H-R) diagram forSDSS1133 in 2014 (non-outbursting phase) and 2019 (out-bursting phase), compared with known S Dor-type LBVsand giant eruption LBVs taken from Rest et al. (2012). Theluminosity and temperature of SDSS1133 are obtained fromthe black body fitting without any host galaxy extinctioncorrections, described in Section 3.4.1. Normal S Dor-typeLBV eruptions are characterized by transitions from a hotquiescent state in the S Dor instability strip (late O typeor early B type) to cool state as an F-type supergiant with T ∼ − η Carinae analogs dis-tribute over the H-R diagram, typically above the empiricalHumphreys & Davidson (1979) (HD) limit (e.g., Davidson2020). The HD-limit is interpreted as the Eddington limitfor massive stars, thus the giant eruption LBVs are consid- ered to be in an unstable super-Eddington state (e.g., Smith2006, 2014).SDSS1133 locates well above the HD limit in the H-Rdiagram even in its relatively faint phase in 2014, placingit in the parameter space consistent with other giant erup-tion LBVs (e.g., Davidson 2020). This means that SDSS1133remains to be in an eruption phase at least since 1950, , re-sembling the decades-long 19th century Great Eruption of η Carinae (Figure 11; Smith & Frew 2011).It is suggested that the giant eruption LBVs may besubdivided into two classes; ’hot LBV’ subclasses exempli-fied by the SN impostor SN 2009ip, and ’cool LBVs’ ex-emplified by UGC 2773-OT (e.g., Smith et al. 2010; Fo-ley et al. 2011; Smith 2013; Ward et al. 2020). The hotLBVs have high temperature of > ,
000 K, high lumi-nosity, and broad line profiles, while the cool LBVs havecooler 7 , − ,
000 K spectra of F-type supergiants su-perposed with narrower emission/absorption features (Fig-ure 15; e.g., Smith et al. 2010). From Figure 15, we concludeSDSS1133 resembles SN 2009ip’s precursor (pre-SN) erup-tion, and thus belongs to the hot LBV subclass. Combinedwith the other observational evidence (high velocity P-Cygniprofile and X-ray emission; Sections 3.3.3 and 3.3.4), we canspeculate (as pointed out by Smith et al. 2010) that the hotLBVs (including SDSS1133) may be powered by ejecta-CSMor ejecta-ejecta interactions by a large fraction, while thecool LBVs are dominated by the photospheric emission ofopaque radiation-driven outflows (e.g., Davidson 2020, andreferences therein).
The dust extinction in the host galaxy of SDSS1133 inferredby the Balmer decrement also implies the presence of X-rayabsorption. As mentioned in Section 2.11.4, the hydrogencolumn density in the host galaxy inferred from the colorexcess is N H , host = (2 . , . , . × cm − under theassumption of the SMC-like, LMC-like, and MW-like extinc-tion, respectively. The X-ray absorption due to the gas of thecolumns density of N H, host is estimated by adding zvphabs ( xszvphabs ) component in the XSPEC ( CIAO ) X-ray spectralmodelling, where the heavy-element abundances relative toMW are assumed to be 1/8, 1/3 for SMC and LMC, respec-tively (Pei 1992), the absorber’s redshift is fixed to 0 . ∼
3. This mean that the uncertainty on the X-ray absorptionstrength in the host galaxy have only little influence on ourorder-of-magnitude discussion about the X-ray luminosity ofSDSS1133.Figure 16 shows the X-ray-to-NIR SED of SDSS1133in the non-outbursting phase in 2014 and outbursting phasein 2019. The Elvis et al. (1994)’s template SED for radio-quiet AGNs is overplotted in Figure 16 for comparison. The
Swift /XRT X-ray upper limits obtained simultaneously withthe Swift/UVOT UV measurements reveal that the X-rayemission of SDSS1133 is much less than the X-ray emissionexpected from the typical X-ray-to-UV/optical luminosityratio in AGNs (see also Simmonds et al. 2016). Moreover,the
Chandra
X-ray detection in the non-outbursting phase in
MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 Table 5.
The 0 . −
10 keV X-ray luminosity ( L . −
10 keV ) of SDSS1133 corrected for the host galaxy X-ray absorption assuming SMC,LMC, and MW-like extinction (Section 3.4.4).MJD Date Luminosity L . −
10 keV (SMC, LMC, MW) Telescope(YYYY/MM/DD) (10 erg s − )56531.4 2013/08/26-28 < . < . < . < . .
669 ( ± . .
644 ( ± . .
522 ( ± . N H, host = (2 . , . , . × cm − (assuming SMC, LMC, and MW-likeextinction, respectively) in the host galaxy is included in the X-ray spectral modelling. λ L λ ( e r g / s ) Rest-frame wavelength (Angstrom)UVOT+PS1 ( 2014-2-6/25 )Swift/XRT ( 2014-2-6/25 )UVOT+ZTF ( 2019-5-27/29 )Swift/XRT ( 2019-5-27/29 )Chandra ( 2019-8-13/14 )Keck/NIRC2 ( 2013-6-16 )AGN template SED λ L λ ( e r g / s ) Rest-frame wavelength (Angstrom)Host extinction corrected (SMC)
Figure 16.
The same as Figure 14, but including X-ray and NIRdata of SDSS1133. Top: the X-ray-to-NIR SED uncorrected forthe host galaxy extinction. Bottom: the X-ray-to-NIR SED cor-rected for the host galaxy extinction assuming the SMC-like ex-tinction (Section 3.4.4). The X-ray SED is calculated as λL λ = L . −
10 keV / ln(10 keV / . L r ad i o G H z ( e r g / s / H z ) L X (erg/s) 06jd88Z05kd86J95N05ip78K98S09ipSDSS1133 Figure 17.
The radio and X-ray luminosity of SDSS1133 observedin the non-outbursting phase, compared with the radio and X-raydata of ejecta-CSM interacting SNe IIn at the radio peak (blackdots) taken from Margutti et al. (2014). The two data points forSDSS1133 indicate the X-ray luminosity uncorrected or correctedfor the possible host galaxy extinction assuming the SMC-likeextinction (Table 5). The data for SN 2009ip are for the 2012bevent (Section 3.2). Note that the LBVs known to date have theX-ray luminosity much less than 10 erg s − (Naz´e et al. 2012). ∼ erg s − ; Lehmer et al. 2016;Birchall et al. 2020; Schirra et al. 2020), thus SDSS1133 can-not be classified as an AGN based on the X-ray luminosity.The UV-optical black body modelling for the hostextinction-uncorrected SED in the non-outbursting phase in2014 suggests that the UV-optical luminosity of SDSS1133is L ∼ erg s − (Section 3.4.1). By using the Chandra
X-ray luminosity ( L X = 4 × erg s − ), the X-ray-to-optical luminosity ratio is log( L X /L ) ∼ − .
4. On the otherhand, if the host galaxy extinction correction shown in Fig-ure 16 is taken at its face value, the UV-optical luminosityof SDSS1133 can be as large as L ∼ erg s − , thus MNRAS000
4. On the otherhand, if the host galaxy extinction correction shown in Fig-ure 16 is taken at its face value, the UV-optical luminosityof SDSS1133 can be as large as L ∼ erg s − , thus MNRAS000 , 1–29 (2021) M. Kokubo log( L X /L ) ∼ − .
4. Naz´e et al. (2012) show that the X-ray-to-optical luminosity ratio (log( L X /L )) of known LBVsis at most ∼ − −
6. We shouldnote that the X-ray luminosity of SDSS1133 is at least 4orders of magnitude larger than that observed in GalacticLBVs ( L . − < erg s − ; Naz´e et al. 2012; Kosset al. 2014). Therefore, we can conclude that SDSS1133 isextremely X-ray bright compared to other normal LBVs. The large discrepancy between the X-ray luminosities ofSDSS1133 and other LBVs known to date (Naz´e et al. 2012)is probably because the current LBV sample observed at X-ray is limited to normal S Dor-type LBVs, and essentiallyno detailed X-ray data is available for giant eruption LBVsor SN impostors.As discussed in Sections 3.3.3 and 3.3.4, a populationof LBV giant eruptions can be powered by shock interac-tions between ejecta materials and CSM (e.g., Smith 2008,2013; Smith et al. 2010, 2018). As observationally confirmedin ejecta-CSM interacting SNe IIn, the interaction regioncan produces strong UV-optical, X-ray, and radio emission(Smith 2013, 2017a). It is naturally expected that the ob-servational properties of the interaction-powered LBVs re-semble the SNe IIn, and observed as scaled-down SNe IIn(Smith 2013).In Figure 17, the X-ray and radio luminosities ofSDSS1133 are compared with those of a sample of ejecta-CSM interacting SNe IIn compiled by Margutti et al. (2014).We should note that the
Chandra
X-ray observation wascarried out just after the 2019 outburst (Figure 3), thusthe X-ray luminosity may be an “afterglow” emission of the2019 outburst event, while the radio luminosity reflects theemission in the non-outbursting phase before the 2014 out-burst. In this figure, we can see that SDSS1133 may locateat the faint end of the X-ray vs. radio correlation space ofthe interacting SNe. The X-ray luminosity of SDSS1133 isabout 2 orders of magnitude lower than the typical valuesof the interacting SNe IIn, suggesting that SDSS1133 is ascaled-down version of these objects. In terms of the multi-wavelength energetics, we can say that SDSS1133 is in be-tween LBVs and interacting SNe.Koss et al. (2014) point out that the Keck/NIRC2 K band measurement may indicate the presence of an IR excessemission at λ ∼ ,
000 ˚A(see Figure 16). Since the hot dustemission of various strength is one of the observational fea-tures of interacting SN (e.g., Szalai et al. 2019; Kokubo et al.2019), the IR excess in SDSS1133 can also be naturally ex-plained by the above mentioned scaled-down SN IIn model.Future optical-NIR simultaneous observations are needed toconfirm this weak NIR excess emission.
In this Section we discuss the energetics of the UV-opticalemission in the non-outbursting and outbursting phases ofSDSS1133. Constructing models to reproduce the detailedstructure of the multi-wavelength light curves is beyondthe scope of this paper and will be presented elsewhere.Here we show that the required radiation energy to ex- plain SDSS1133 can be generated within existing observa-tional and theoretical frameworks; namely, some form ofnon-terminal explosions and subsequent efficient kinetic-to-radiation energy conversion by ejecta-CSM or ejecta-ejectainteractions.The UV-optical luminosity of SDSS1133 is at least L outburst = 10 . L (cid:12) = 10 . erg s − during the 2019 out-burst (no host galaxy extinction correction; Section 3.4.1),that continued for more than 27.5 days (see Section 3.2.2).Thus, the total energy required to explain the 2019 outburstis at least E rad , outburst ∼ erg × (cid:18) L outburst . L (cid:12) (cid:19) (cid:18) ∆ t . (cid:19) . (4)SDSS1133 experienced multiple outbursts of similar radia-tion energy at least three times since 2001. On the otherhand, the persistent continuum luminosity in the non-outbursting phase since 1950 is roughly L non-outburst =10 . L (cid:12) = 10 . erg s − , thus the the total radiationenergy released so far is E rad , non-outburst ∼ erg × (cid:18) L non-outburst . L (cid:12) (cid:19) (cid:18) ∆ t
70 yrs (cid:19) . (5)We should note that these estimates for E rad are lower-limit,as the possible host galaxy extinction is uncorrected correc-tion in the above calculations.The Eddington ratio or Eddington parameter Γ giventhe luminosity L and stellar mass M isΓ = κ e L πGMc = 2 . × (cid:18) L L (cid:12) (cid:19) (cid:18) M M (cid:12) (cid:19) − , (6)where κ e = 0 .
34 cm g − is the opacity due to Thom-son scattering, and we assume M = 100 M (cid:12) (the currentmass of η Carinae; e.g., Smith 2008) as a reference value.The luminosity of SDSS1133 (
L > L (cid:12) ) implies thatit exceeds the classical Eddington limit of Γ = 1 even inthe non-outbursting phase at least for 70 yrs. This situa-tion is reminiscent of the η Carinae’s Great Eruption, thathad been luminous at Γ = 5 for over decades. Like the η Carinae’s Great Eruption, SDSS1133 is probably contin-uously launching inhomogeneous, porous super-Eddington(SE) continuum-driven winds (e.g., Smith & Owocki 2006;Owocki & Shaviv 2012; Dotan & Shaviv 2012). A dense CSMaround SDSS1133 can be produced by these radiation-drivenwinds (e.g., Smith 2014). The radiation-driven wind mech-anisms, however, cannot produce high velocity outflows of > ,
000 km s − (well beyond the star’s escape velocity) ob-served in SDSS1133 as the broad absorption, thus we needsome form of non-terminal explosions and subsequent en-hanced mass ejections (like scaled-down SN) to explain thehigh velocity materials observed in SDSS1133 (Sections 3.3.3and 3.3.4).The radiated energy of SDSS1133 is comparable to thatof a single core-collapse SN ( E rad ∼ erg). As suggestedin Sections 3.3.3 and 3.3.4, the large radiation energy ofSDSS1133 can be naturally explained by mass ejections dueto non-terminal explosions ( E kin ∼ − erg, where E kin is the kinetic energy of the explosion) and subsequentCSM interactions. Because the CSM interactions can havehigh efficiency of converting kinetic energy into radiation (afew 10%; e.g., Moriya et al. 2013a; Smith et al. 2014), theshocks induced by the non-terminal explosions and prop- MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 agating into the dense CSM can produce enough radia-tion energy to account for both the outbursting and non-outbursting phases of SDSS1133. In the same way as theCSM interactions, if massive shells ejected by the multipleexplosion events have different velocities, collisions betweenthe shells may produce distinct outbursts (e.g., Heger et al.2003; Woosley et al. 2007; Dessart et al. 2009; Moriya et al.2013a; Yoshida et al. 2016; Arcavi et al. 2017).Consider a simple ejecta-CSM shock interaction model(e.g., Smith 2013; Moriya et al. 2013b). The CSM is as-sumed to be formed via a steady wind with a velocity of v w . The kinetic energy of the ejected massive shell is con-verted into radiation energy at the shocked region mov-ing at a velocity of v s . The shocked shell velocity v s istime-dependent, but here we replace it by the initial ve-locity of the ejecta: v s ∼ (cid:112) E kin /M ej (cid:39) ,
000 km s − × ( E kin / erg) / ( M ej / M (cid:12) ) − / , where M ej is theejected mass. A kinetic-to-radiation conversion efficiency, (cid:15) ,is usually assumed to be in a range of 0 . − . M can be relatedto the bolometric luminosity via (cid:15) as: L = 12 (cid:15) ˙ Mv w v s = 10 . L (cid:12) × (cid:16) (cid:15) . (cid:17) (cid:18) ˙ MM (cid:12) yr − (cid:19) × (cid:16) v w
100 km s − (cid:17) − (cid:18) E kin erg (cid:19) / (cid:18) M ej M (cid:12) (cid:19) − / . (7)Although most of the parameters are uncertain in the caseof SDSS1133, this estimate suggests that the luminositiesin the non-outbursting andoutburst phases can be producedby the CSM interactions if the persistent mass-loss of theorder of ˙ M ∼ M (cid:12) yr − and non-terminal explosions of E kin ∼ erg and M ej ∼ M (cid:12) . Such a large mass-loss rate is suggested to present in giant eruption LBVs(Smith 2014), and actually the mass-loss rate at the timeof the η Carinaes’s Great Eruption is estimated to be a few M (cid:12) yr − from observations for the Homunculus nebula (e.g.,Humphreys 2005; Smith 2013). The large mass-loss rate alsoindicates that SDSS1133 is a massive star of the mass of100 M (cid:12) like η Carinae.While the dense CSM may be naturally formed via theSE winds (e.g., Smith 2014), it is unclear what mechanismscan trigger (multiple) non-terminal explosions in LBVs (e.g.,Smith et al. 2014). From the viewpoint of energy deposition,it is suggested that the non-terminal explosions of E kin ∼ erg may not be explained by the envelope instability,and require to transfer a large amount of extra energy fromthe stellar core into the outer envelope (e.g., Smith 2014;Smith et al. 2014; Strotjohann et al. 2020).Many mechanisms to generate the mass-loss episodesin the giant eruption LBVs or SN impostors have been sug-gested in the literature, including pulsational pair instability(PPI; e.g., Woosley et al. 2007; Woosley 2017; Arcavi et al.2017), wave-driven mass loss (e.g., Quataert & Shiode 2012;Shiode & Quataert 2014), and stellar collisions or mergers ina binary system (e.g., Kashi & Soker 2010; Smith 2011; Hiraiet al. 2020) (see Smith et al. 2011; Smith 2014, for a review).Although η Carinae is known to be a binary (or triple) sys-tem, currently there is no reason to consider that SDSS1133is in a binary system and its eruptions are caused by binary interactions. Among these scenarios, the PPI-driven massejection is of particular interest since this mechanism maypossibly provide a natural explanation for the multiple gi-ant eruptions (=non-terminal explosions) every few yearswith the explosion energy of ∼ erg, ejected mass of ∼ M (cid:12) , and ejecta velocity of ∼ ,
000 km s − , as ob-served in SDSS1133 and other extreme LBV giant eruptions(e.g., Smith 2008; Yoshida et al. 2016; Woosley 2017; Arcaviet al. 2017). An interesting possibility is that, since the PPI(as well as the wave-driven mass loss) is predicted to oc-cur at the very late stage of massive star evolution, we maybe able to observe the terminal SN explosion of SDSS1133after a few additional pre-SN eruptions, like other precur-sor eruptions observed prior to SNe IIn (e.g., Gal-Yam &Leonard 2009; Smith et al. 2011; Langer 2012; Mauerhanet al. 2013; Moriya et al. 2014; Ofek et al. 2014; Arcaviet al. 2017; Smith 2017b; Strotjohann et al. 2020). Furthermulti-wavelength follow-up observations are encouraged totrace the evolution of SDSS1133. Finally we note the rarity of the SDSS1133-like objects inthe local universe. The volumetric rate of the SDSS1133-like LBV objects can roughly be estimated by the fact thatthere are only two objects found to be offset from the hostgalaxy nucleus among 3,579 broad H α emission lines objectsat z < .
31 targeted by the SDSS-I/II spectroscopic survey( > ,
000 km s − ; Stern & Laor 2012; Koss et al. 2014).Among the two objects, one is SDSS1133, and another isMrk 739, but Mrk 739 is confirmed to be a dual AGN (Kosset al. 2011). Thus, here we assume that SDSS1133 is the onlyextreme LBV giant eruption object in the z < .
31 SDSSLegacy spectroscopic survey sample. The survey area of theSDSS Legacy spectroscopic survey is 8,032 deg (Abazajianet al. 2009). The comoving volume within redshift z = 0 . × (8 , / , , thus the vol-umetric rate of SDSS1133-like objects may be estimated as ∼ . − . We should note that this rate is a very con-servative lower limit considering the large luminosity varia-tions of LBVs; if SDSS1133 was not in the outburst phasein 2001 − z ∼ .
01 (correspondingto ∼
20 mag in the non-outbursting phase of SDSS1133) isthe maximum possible redshift where SDSS1133-like objectsin their faint phase can be detected, the same calculationresults in the volumetric rate of ∼ ,
000 Gpc − .The estimated rate of the SDSS1133-like LBV objectsis much less than the core-collapse SN rate in the local Uni-verse ( ∼ ,
000 Gpc − yr − ; Perley et al. 2020). It may in-dicate that at most a few percent of core-collapse SN progen-itors experience the SDSS1133-like long-lasting ( >
60 years)bright LBV eruption phase. This may be consistent with thescenario that LBVs with enhanced mass-loss end up with(super-luminous) SNe IIn, i.e., a rare population of core-collapse SNe (Ofek et al. 2014; Smith 2017b; Strotjohannet al. 2020). ΛCDM cosmology of Ω Λ = 0 .
7, Ω m = 0 .
3, and H =70 km s − Mpc − is assumedMNRAS000
3, and H =70 km s − Mpc − is assumedMNRAS000 , 1–29 (2021) M. Kokubo
The above volumetric rate estimate ignores several ob-jects similar to SDSS1133 individually reported in the litera-ture; for example, PHL 293B, which is found in a local bluecompact dwarf galaxy at z = 0 . z = 0 .
013 with the peak absolute magnitudeof M g = − .
34; Perley et al. 2020), which is classified as aneruption of a LBV. It is worth noticing that the spectral re-semblance between SDSS1133 and broad line AGNs suggeststhat many SDSS1133-like objects are missed by the cur-rent surveys; for example, a fraction of nuclear/off-nuclearAGN-like objects can actually be LBVs (e.g., Izotov et al.2007; Izotov & Thuan 2008; Baldassare et al. 2016; Sim-monds et al. 2016; Burke et al. 2020a; Ward et al. 2020). Inthis sense, the volumetric rate of the SDSS1133-like objectscould be higher than the above estimate, and more compre-hensive survey is needed to quantify the true occurrence rateof the SDSS1133-like objects.
In this paper we have presented the comprehensive analy-sis of the multi-wavelength data of SDSS1133. Our analysissuggests that SDSS1133 is likely to be an extremely lumi-nous giant eruption LBV, and strongly disfavor the recoilingAGN scenario suggested in the literature (Table 4).The observed peak absolute magnitude of the out-bursts reaches M g ∼ −
16 mag ( −
18 mag after correct-ing for the possible host galaxy extinction), which placesSDSS1133 as one of the brightest LBV giant eruptions ofknown Galactic and extragalactic LBVs and SN impostors.The observational properties of SDSS1133 (multiple out-bursts, broad P-Cygni profile, UV-optical luminosity, highblack body temperature of > ,
000 K) is reminiscent ofpre-SN eruptions of SN 2009ip (Sections 3.2.2, 3.3.3, 3.3.4,and 3.4.1), suggesting that SDSS1133 belongs to the hotLBV subclass (Section 3.4.2). The persistent P-Cygni ab-sorption feature seen in the hydrogen Balmer lines extendsup to − ,
000 km s − . The fast ejected materials respon-sible for the P-Cygni absorption in SDSS1133 likely origi-nate from multiple non-terminal explosive outbursts, ratherthan radiation-driven stellar winds (Section 3.3.3). We sug-gest that both of the bright persistent emission and UV-optical outbursts in SDSS1133 are produced by interactionsof the ejected shell with CSM and/or different shells (Sec-tions 3.3.4 and 3.5). The interaction model naturally ex-plains the long-duration SN IIn-like multi-wavelength pho-tometric and spectroscopic properties of SDSS1133 includ-ing X-ray luminosity, persistent broad emission/absorptionlines, and blue UV-optical continuum (Sections 3.3.4, 3.4.4,and 3.5).The origin of the non-terminal explosions is highly de-bated. Among the many theoretical possibilities, we suggestthat pulsational pair-instability may provide a viable expla-nation for the multiple energetic eruptions in SDSS1133, interms of the energetics and occurence rate of the eruptions(Sections 3.2.2 and 3.5). If the current activity of SDSS1133is a bona-fide precursor of a terminal SN explosion, we may be able to observe a few additional giant eruptions and thenthe terminal SN explosion in future observations (in a fewyears to a few thousand years). ACKNOWLEDGEMENTS
We thank Masaomi Tanaka and Hirofumi Noda for a fruitful dis-cussion. A part of this work was initiated during my stay at Cal-tech under the Caltech-Japan visiting program in Astronomy, andM. K. is grateful to Matthew J. Graham for his support and hos-pitality during the visit.This research made use of
Astropy , a community-developedcore Python package for Astronomy (Astropy Collaboration et al.2013). We used numpy (Harris et al. 2020), scipy (Virtanen et al.2020), extinction (Barbary 2016), and speclite in the dataanalysis.This research has made use of the Keck Observa-tory Archive (KOA), which is operated by the W. M.Keck Observatory and the NASA Exoplanet Science Insti-tute(NExScI), under contract with the National Aeronauticsand Space Administration. The Keck archival data (KOAID= LB.20150717.21554, LB.20150717.22197, LB.20190310.44548,LR.20150717.21568, LR.20150717.22257, LR.20190310.44557) ob-tained through the programs C280LA and C300 (PI: F. Harrison)were used.This work has made use of data from the European SpaceAgency (ESA) mission Gaia ( ), processed by the Gaia
Data Processing and Analysis Con-sortium (DPAC, ). Funding for the DPAC has been provided by na-tional institutions, in particular the institutions participating inthe
Gaia
Multilateral Agreement.We acknowledge ESA Gaia, DPAC and the Photometric Sci-ence Alerts Team (http://gsaweb.ast.cam.ac.uk/alerts).This research has made use of the NASA/IPAC Infrared Sci-ence Archive, which is operated by the Jet Propulsion Laboratory,California Institute of Technology, under contract with the Na-tional Aeronautics and Space Administration.This research has made use of data obtained from the Chan-dra Data Archive and the Chandra Source Catalog, and softwareprovided by the Chandra X-ray Center (CXC) in the applicationpackages CIAO, ChIPS, and Sherpa.Based on observations obtained with
XMM-Newton , anESA science mission with instruments and contributions directlyfunded by ESA Member States and NASA.We acknowledge the use of public data from the Swift dataarchive.Based on observations obtained with the Samuel Oschin 48-inch Telescope at the Palomar Observatory as part of the ZwickyTransient Facility project. ZTF is supported by the National Sci-ence Foundation under Grant No. AST-1440341 and a collabora-tion including Caltech, IPAC, the Weizmann Institute for Science,the Oskar Klein Center at Stockholm University, the University ofMaryland, the University of Washington, Deutsches Elektronen-Synchrotron and Humboldt University, Los Alamos National Lab-oratories, the TANGO Consortium of Taiwan, the University ofWisconsin at Milwaukee, and Lawrence Berkeley National Labo-ratories. Operations are conducted by COO, IPAC, and UW.The Pan-STARRS1 Surveys (PS1) and the PS1 public sci-ence archive have been made possible through contributions bythe Institute for Astronomy, the University of Hawaii, the Pan-STARRS Project Office, the Max-Planck Society and its partic-ipating institutes, the Max Planck Institute for Astronomy, Hei-delberg and the Max Planck Institute for Extraterrestrial Physics, https://github.com/desihub/specliteMNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 Garching, The Johns Hopkins University, Durham University,the University of Edinburgh, the Queen’s University Belfast, theHarvard-Smithsonian Center for Astrophysics, the Las CumbresObservatory Global Telescope Network Incorporated, the Na-tional Central University of Taiwan, the Space Telescope ScienceInstitute, the National Aeronautics and Space Administrationunder Grant No. NNX08AR22G issued through the PlanetaryScience Division of the NASA Science Mission Directorate, theNational Science Foundation Grant No. AST-1238877, the Uni-versity of Maryland, Eotvos Lorand University (ELTE), the LosAlamos National Laboratory, and the Gordon and Betty MooreFoundation.The Legacy Surveys consist of three individual and comple-mentary projects: the Dark Energy Camera Legacy Survey (DE-CaLS; NOAO Proposal ID
DATA AVAILABILITY
The data underlying this article will be shared on reasonable re-quest to the corresponding author.
REFERENCES
Abazajian K. N., et al., 2009, ApJS, 182, 543Allan A. P., Groh J. H., Mehner A., Smith N., Boian I., FarrellE. J., Andrews J. E., 2020, MNRAS, 496, 1902Apparao K. M. V., Tarafdar S. P., 1988, A&A, 192, 255Arcavi I., et al., 2017, Nature, 551, 210Astropy Collaboration et al., 2013, A&A, 558, A33Baldassare V. F., et al., 2016, ApJ, 829, 57Baldwin J. A., Phillips M. M., Terlevich R., 1981, PASP, 93, 5Banerjee G., Mathew B., Paul K. T., Subramaniam A., Bhat-tacharyya S., Anusha R., 2021, MNRAS, 500, 3926Barbary K., 2016, extinction v0.3.0, doi:10.5281/zenodo.804967, https://doi.org/10.5281/zenodo.804967
Bellm E. C., et al., 2019, PASP, 131, 018002Bertin E., 2011, in Evans I. N., Accomazzi A., Mink D. J., RotsA. H., eds, Astronomical Society of the Pacific Conference Se-ries Vol. 442, Astronomical Data Analysis Software and Sys-tems XX. p. 435Bertin E., Arnouts S., 1996, A&AS, 117, 393Birchall K. L., Watson M. G., Aird J., 2020, MNRAS, 492, 2268Blecha L., Loeb A., 2008, MNRAS, 390, 1311Boroson T. A., Green R. F., 1992, ApJS, 80, 109Breeveld A. A., Landsman W., Holland S. T., Roming P., KuinN. P. M., Page M. J., 2011, in McEnery J. E., Racusin J. L.,Gehrels N., eds, American Institute of Physics Conference Se-ries Vol. 1358, American Institute of Physics Conference Se-ries. pp 373–376 ( arXiv:1102.4717 ), doi:10.1063/1.3621807Burke C. J., Liu X., Chen Y.-C., Shen Y., Guo H., 2020a, arXive-prints, p. arXiv:2011.10053Burke C. J., et al., 2020b, ApJ, 894, L5Chambers K. C., et al., 2016, arXiv e-prints, p. arXiv:1612.05560Chevalier R. A., 1982, ApJ, 258, 790Chevalier R. A., Fransson C., Nymark T. K., 2006, ApJ, 641, 1029Chugai N. N., 2001, MNRAS, 326, 1448Damineli A., 1996, ApJ, 460, L49Davidson K., 2020, Galaxies, 8, 10Davidson K., Humphreys R. M., 1997, ARA&A, 35, 1Dessart L., Hillier D. J., Gezari S., Basa S., Matheson T., 2009,MNRAS, 394, 21Dey A., et al., 2019, AJ, 157, 168MNRAS000
Bellm E. C., et al., 2019, PASP, 131, 018002Bertin E., 2011, in Evans I. N., Accomazzi A., Mink D. J., RotsA. H., eds, Astronomical Society of the Pacific Conference Se-ries Vol. 442, Astronomical Data Analysis Software and Sys-tems XX. p. 435Bertin E., Arnouts S., 1996, A&AS, 117, 393Birchall K. L., Watson M. G., Aird J., 2020, MNRAS, 492, 2268Blecha L., Loeb A., 2008, MNRAS, 390, 1311Boroson T. A., Green R. F., 1992, ApJS, 80, 109Breeveld A. A., Landsman W., Holland S. T., Roming P., KuinN. P. M., Page M. J., 2011, in McEnery J. E., Racusin J. L.,Gehrels N., eds, American Institute of Physics Conference Se-ries Vol. 1358, American Institute of Physics Conference Se-ries. pp 373–376 ( arXiv:1102.4717 ), doi:10.1063/1.3621807Burke C. J., Liu X., Chen Y.-C., Shen Y., Guo H., 2020a, arXive-prints, p. arXiv:2011.10053Burke C. J., et al., 2020b, ApJ, 894, L5Chambers K. C., et al., 2016, arXiv e-prints, p. arXiv:1612.05560Chevalier R. A., 1982, ApJ, 258, 790Chevalier R. A., Fransson C., Nymark T. K., 2006, ApJ, 641, 1029Chugai N. N., 2001, MNRAS, 326, 1448Damineli A., 1996, ApJ, 460, L49Davidson K., 2020, Galaxies, 8, 10Davidson K., Humphreys R. M., 1997, ARA&A, 35, 1Dessart L., Hillier D. J., Gezari S., Basa S., Matheson T., 2009,MNRAS, 394, 21Dey A., et al., 2019, AJ, 157, 168MNRAS000 , 1–29 (2021) M. Kokubo
Dom´ınguez A., et al., 2013, ApJ, 763, 145Dotan C., Shaviv N. J., 2012, MNRAS, 427, 3071Draine B. T., Lee H. M., 1984, ApJ, 285, 89Dwarkadas V. V., Gruszko J., 2012, MNRAS, 419, 1515Ekstr¨om S., et al., 2012, A&A, 537, A146Elvis M., et al., 1994, ApJS, 95, 1Filippenko A. V., 1989, AJ, 97, 726Fitzpatrick E. L., 1999, PASP, 111, 63Foley R. J., Berger E., Fox O., Levesque E. M., Challis P. J., IvansI. I., Rhoads J. E., Soderberg A. M., 2011, ApJ, 732, 32Frew D. J., 2004, Journal of Astronomical Data, 10, 6Fruscione A., et al., 2006, in Silva D. R., Doxsey R. E., eds,Society of Photo-Optical Instrumentation Engineers (SPIE)Conference Series Vol. 6270, Society of Photo-Optical Instru-mentation Engineers (SPIE) Conference Series. p. 62701V,doi:10.1117/12.671760Gabriel C., et al., 2004, in Ochsenbein F., Allen M. G., EgretD., eds, Astronomical Society of the Pacific Conference SeriesVol. 314, Astronomical Data Analysis Software and Systems(ADASS) XIII. p. 759Gaia Collaboration et al., 2016, A&A, 595, A1Gal-Yam A., Leonard D. C., 2009, Nature, 458, 865Gehrels N., et al., 2004, ApJ, 611, 1005Graham M. L., et al., 2014, ApJ, 787, 163Graham M. J., Djorgovski S. G., Drake A. J., Stern D., MahabalA. A., Glikman E., Larson S., Christensen E., 2017, MNRAS,470, 4112Graham M. J., et al., 2019, PASP, 131, 078001HEASARC 2014, HEAsoft: Unified Release of FTOOLS andXANADU (ascl:1408.004)Hainline K. N., Reines A. E., Greene J. E., Stern D., 2016, ApJ,832, 119Hall P. B., 2007, AJ, 133, 1271Hall P. B., et al., 2002, ApJS, 141, 267Harris C. R., et al., 2020, Nature, 585, 357–362Heger A., Fryer C. L., Woosley S. E., Langer N., Hartmann D. H.,2003, ApJ, 591, 288Hirai R., Podsiadlowski P., Owocki S. P., Schneider F. R. N.,Smith N., 2020, arXiv e-prints, p. arXiv:2011.12434Humphreys R. M., 2005, in Humphreys R., Stanek K., eds, As-tronomical Society of the Pacific Conference Series Vol. 332,The Fate of the Most Massive Stars. p. 14Humphreys R. M., Davidson K., 1979, ApJ, 232, 409Humphreys R. M., Davidson K., 1994, PASP, 106, 1025Humphreys R. M., Davidson K., Smith N., 1999, PASP, 111, 1124Humphreys R. M., Davidson K., Grammer S., Kneeland N., Mar-tin J. C., Weis K., Burggraf B., 2013, ApJ, 773, 46Humphreys R. M., Weis K., Davidson K., Bomans D. J., BurggrafB., 2014, ApJ, 790, 48Izotov Y. I., Thuan T. X., 2008, ApJ, 687, 133Izotov Y. I., Thuan T. X., 2009, ApJ, 690, 1797Izotov Y. I., Thuan T. X., Guseva N. G., 2007, ApJ, 671, 1297Jansen F., et al., 2001, A&A, 365, L1Kashi A., Soker N., 2010, ApJ, 723, 602Kauffmann G., et al., 2003, MNRAS, 346, 1055Keel W. C., et al., 2012, MNRAS, 420, 878Kewley L. J., Dopita M. A., Sutherland R. S., Heisler C. A.,Trevena J., 2001, ApJ, 556, 121Kewley L. J., Groves B., Kauffmann G., Heckman T., 2006, MN-RAS, 372, 961Kewley L. J., Dopita M. A., Leitherer C., Dav´e R., Yuan T., AllenM., Groves B., Sutherland R., 2013, ApJ, 774, 100Kilpatrick C. D., et al., 2018, MNRAS, 473, 4805Kokubo M., 2017, MNRAS, 467, 3723Kokubo M., et al., 2019, ApJ, 872, 135Komossa S., 2012, Advances in Astronomy, 2012, 364973Koss M., et al., 2011, ApJ, 735, L42Koss M., et al., 2014, MNRAS, 445, 515 Laher R. R., et al., 2014, PASP, 126, 674Langer N., 2012, ARA&A, 50, 107Law N. M., et al., 2009, PASP, 121, 1395Lehmer B. D., et al., 2016, ApJ, 825, 7L´opez-Corredoira M., Guti´errez C. M., 2006, in Lerner E. J.,Almeida J. B., eds, American Institute of Physics ConferenceSeries Vol. 822, First Crisis in Cosmology Conference. pp 75–92 ( arXiv:astro-ph/0509630 ), doi:10.1063/1.2189124MacLeod C. L., et al., 2010, ApJ, 721, 1014MacLeod C. L., et al., 2012, ApJ, 753, 106Magliocchetti M., et al., 2020, arXiv e-prints, p. arXiv:2002.02980Margutti R., et al., 2014, ApJ, 780, 21Masci F. J., et al., 2019, PASP, 131, 018003Masci et al. F. J., 2020, The ZTF Science Data System (ZSDS)Explanatory Supplement, Version 5Mason K. O., et al., 2001, A&A, 365, L36Mauerhan J. C., et al., 2013, MNRAS, 430, 1801Moriya T. J., Blinnikov S. I., Tominaga N., Yoshida N., TanakaM., Maeda K., Nomoto K., 2013a, MNRAS, 428, 1020Moriya T. J., Maeda K., Taddia F., Sollerman J., Blinnikov S. I.,Sorokina E. I., 2013b, MNRAS, 435, 1520Moriya T. J., Maeda K., Taddia F., Sollerman J., Blinnikov S. I.,Sorokina E. I., 2014, MNRAS, 439, 2917Moriya T. J., Mazzali P. A., Pian E., 2020, MNRAS, 491, 1384Morris P. W., Gull T. R., Hillier D. J., Barlow M. J., Royer P.,Nielsen K., Black J., Swinyard B., 2017, ApJ, 842, 79Naz´e Y., Rauw G., Hutsem´ekers D., 2012, A&A, 538, A47Ofek E. O., et al., 2014, ApJ, 789, 104Osterbrock D. E., 1989, Astrophysics of gaseous nebulae and ac-tive galactic nucleiOwocki S. P., Shaviv N. J., 2012, Instability & Mass Loss near theEddington Limit. p. 275, doi:10.1007/978-1-4614-2275-4 12Parkin E. R., Pittard J. M., Corcoran M. F., Hamaguchi K.,Stevens I. R., 2009, MNRAS, 394, 1758Pastorello A., Fraser M., 2019, Nature Astronomy, 3, 676Pastorello A., et al., 2013, ApJ, 767, 1Patterson M. T., et al., 2019, PASP, 131, 018001Pei Y. C., 1992, ApJ, 395, 130Peng C. Y., Ho L. C., Impey C. D., Rix H.-W., 2011, GALFIT:Detailed Structural Decomposition of Galaxy Images, Astro-physics Source Code Library (ascl:1104.010)Perez-Torres M., Piconcelli N. R.-O. E., Alberdi A., Komossa S.,Herrero-Illana R., 2015, The Astronomer’s Telegram, 7388, 1Perley D. A., et al., 2020, ApJ, 904, 35Phinney E. S., 1989, in Morris M., ed., Vol. 136, The Center ofthe Galaxy. p. 543Poole T. S., et al., 2008, MNRAS, 383, 627Pursimo T., Ighina L., Ihanec N., Mandarakas N., Skillen K.,Terefe S., 2019a, Contributions of the Astronomical Observa-tory Skalnate Pleso, 49, 539Pursimo T., Galindo-Guil F., Dennefeld M., Ighina L., IhanecN., Mandarakas N., Skillen K., Terefe S., 2019b, The As-tronomer’s Telegram, 12911, 1Quataert E., Shiode J., 2012, MNRAS, 423, L92Reines A. E., Greene J. E., Geha M., 2013, ApJ, 775, 116Rest A., et al., 2012, Nature, 482, 375Schirra A. P., Habouzit M. K.-R. S., Fornasini F., Nelson D.,Pillepich A., 2020, arXiv e-prints, p. arXiv:2011.02501Schlafly E. F., Finkbeiner D. P., 2011, ApJ, 737, 103Shiode J. H., Quataert E., 2014, ApJ, 780, 96Simmonds C., Bauer F. E., Thuan T. X., Izotov Y. I., Stern D.,Harrison F. A., 2016, A&A, 596, A64Smith N., 2006, ApJ, 644, 1151Smith N., 2008, Nature, 455, 201Smith N., 2011, MNRAS, 415, 2020Smith N., 2012, All Things Homunculus. p. 145, doi:10.1007/978-1-4614-2275-4 7Smith N., 2013, MNRAS, 429, 2366 MNRAS , 1–29 (2021) xtragalactic giant eruption LBV SDSS1133 Smith N., 2014, ARA&A, 52, 487Smith N., 2017a, Interacting Supernovae: Types IIn and Ibn.p. 403, doi:10.1007/978-3-319-21846-5 38Smith N., 2017b, Philosophical Transactions of the Royal Societyof London Series A, 375, 20160268Smith N., Frew D. J., 2011, MNRAS, 415, 2009Smith N., Owocki S. P., 2006, ApJ, 645, L45Smith R. K., Brickhouse N. S., Liedahl D. A., Raymond J. C.,2001, ApJ, 556, L91Smith N., et al., 2010, AJ, 139, 1451Smith N., Li W., Silverman J. M., Ganeshalingam M., FilippenkoA. V., 2011, MNRAS, 415, 773Smith N., Mauerhan J. C., Prieto J. L., 2014, MNRAS, 438, 1191Smith N., et al., 2018, MNRAS, 480, 1466Sollerman J., et al., 2019, A&A, 621, A30Stanek K. Z., et al., 2019, The Astronomer’s Telegram, 12794, 1Stern J., Laor A., 2012, MNRAS, 423, 600Strotjohann N. L., et al., 2020, arXiv e-prints, p. arXiv:2010.11196Szalai T., Zs´ıros S., Fox O. D., Pejcha O., M¨uller T., 2019, ApJS,241, 38Taddia F., et al., 2013, A&A, 555, A10Tazaki R., Ichikawa K., Kokubo M., 2020, ApJ, 892, 84Terlevich R., Terlevich E., Bosch G., D´ıaz ´A., H¨agele G., CardaciM., Firpo V., 2014, MNRAS, 445, 1449Toba Y., et al., 2014, ApJ, 788, 45Tully R. B., 1988, Nearby galaxies catalogTully R. B., 1994, VizieR Online Data Catalog, p. VII/145Van Dyk S. D., Matheson T., 2012, The Supernova Impostors.p. 249, doi:10.1007/978-1-4614-2275-4 11Van Dyk S. D., Peng C. Y., King J. Y., Filippenko A. V., TreffersR. R., Li W., Richmond M. W., 2000, PASP, 112, 1532Vink J. S., 2011, Ap&SS, 336, 163Virtanen P., et al., 2020, Nature Methods, 17, 261Ward C., et al., 2020, arXiv e-prints, p. arXiv:2011.11656Waters C. Z., et al., 2020, ApJS, 251, 4Weiler M., 2018, A&A, 617, A138Weis K., Bomans D. J., 2020, Galaxies, 8, 20Weisskopf M. C., Brinkman B., Canizares C., Garmire G., MurrayS., Van Speybroeck L. P., 2002, PASP, 114, 1Wolf B., 1989, A&A, 217, 87Woosley S. E., 2017, ApJ, 836, 244Woosley S. E., Blinnikov S., Heger A., 2007, Nature, 450, 390Yalinewich A., Matzner C. D., 2019, MNRAS, 490, 312Yoshida T., Umeda H., Maeda K., Ishii T., 2016, MNRAS, 457,351Zhang S., et al., 2015, ApJ, 815, 113Zhou H., Wang T., Yuan W., Lu H., Dong X., Wang J., Lu Y.,2006, ApJS, 166, 128Zou H., et al., 2019, ApJS, 245, 4van Dokkum P. G., 2001, PASP, 113, 1420This paper has been typeset from a TEX/L A TEX file prepared bythe author.MNRAS000