Multi-epoch spectropolarimetry of SN 2009ip: direct evidence for aspherical circumstellar material
Jon C. Mauerhan, G. Grant Williams, Nathan Smith, Paul S. Smith, Alexei V. Filippenko, Jennifer L. Hoffman, Douglas C. Leonard, Peter Milne, Kelsey I. Clubb, Ori D. Fox, Patrick L. Kelly
aa r X i v : . [ a s t r o - ph . S R ] D ec Mon. Not. R. Astron. Soc. , 1– ?? (2013) Printed 7 October 2018 (MN L A TEX style file v2.2)
Multi-epoch spectropolarimetry of SN 2009ip: directevidence for aspherical circumstellar material
Jon Mauerhan , ⋆ , G. Grant Williams , , Nathan Smith , Paul S. Smith ,Alexei V. Filippenko , Jennifer L. Hoffman , Peter Milne , Douglas C. Leonard ,Kelsey I. Clubb , Ori D. Fox , and Patrick L. Kelly Department of Astronomy, University of California, Berkeley, CA 94720-3411, USA Steward Observatory, University of Arizona, 933 N. Cherry Ave., Tucson, AZ 85721, USA MMT Observatory, Tucson, AZ 85721-0065, USA Department of Physics & Astronomy, University of Denver, 2112 East Wesley Avenue, Denver, CO 80208 Department of Astronomy, San Diego State University, PA-210, 5500 Campanile Drive, San Diego, CA 92182-1221
ABSTRACT
We present spectropolarimetry of SN 2009ip throughout the evolution of its 2012 ex-plosion. During the 2012a phase, when the spectrum exhibits broad P-Cygni lines, wemeasure a V -band polarization of P ≈ .
9% at a position angle of θ ≈ ◦ , indi-cating substantial asphericity for the 2012a outflow. Near the subsequent peak of the2012b phase, when the spectrum shows signs of intense interaction with circumstellarmaterial (CSM), we measure P ≈ .
7% and θ ≈ ◦ , indicating a separate componentof polarization during 2012b, which exhibits a higher degree of asphericity than 2012aand an orthogonal axis of symmetry on the sky. Around 30 days past peak, coincidentwith a substantial bump in the declining light curve, we measure P ≈ .
7% and an-other significant shift in θ . At this point, broad photospheric lines have again becomeprominent and exhibit significant variations in P relative to the continuum, partic-ularly He i /Na I D. By 60 days past peak the continuum polarization has droppedbelow 0.2%, probably declining toward a low value of interstellar polarization. Theresults are consistent with a scenario in which a prolate (possibly bipolar) explosionlaunched during the 2012a phase impacts an oblate (toroidal) distribution of CSMin 2012b. Previous calculations that assumed spherical symmetry for the CSM havesubstantially underestimated the required explosion energy, since only a small fractionof the SN ejecta appears to have participated in strong CSM interaction. An ejectakinetic energy of at least ∼ ergs is difficult to avoid, supporting the interpretationthat the 2012 outburst of SN 2009ip was the result of a core-collapse explosion. Key words: circumstellar matter — stars: evolution — stars: winds, outflows —supernovae: general — supernovae: individual (SN 2009ip)
Interacting supernovae (SNe) are stellar explosions that col-lide with dense circumstellar material (CSM) produced bythe progenitor star. These events are raising critical newquestions about the final evolutionary phases of massivestars and the mass-loss episodes that ensue before core col-lapse (Smith & Arnett 2014). Type-IIn and Ibn SNe, in par-ticular, are interacting SNe that are characterized spectro-scopically by the presence of relatively narrow emission linesof H and He in their spectra (Schlegel 1990; Filippenko 1997; ⋆ E-mail: [email protected]
Pastorello et al. 2008), which arise from dense CSM thatbecomes illuminated by the shock between the fast movingSN ejecta and slower moving CSM (Chevalier & Fransson1994). As such, observations of interacting SNe probe thestellar progenitor’s pre-SN mass-loss history, providing valu-able information on its final evolutionary episodes.Various lines of evidence show that interacting SNe re-quire eruptive pre-SN mass loss that is reminiscent of lumi-nous blue variable (LBV) stars, like η Car, although obser-vations indicate a wide range of mass-loss properties (e.g.,see the review by Smith 2014). The eruptions are often de-tectable as extragalactic transients, commonly referred to as‘SN impostors’ (Van Dyk 2000; Smith et al. 2011a; Kochanek c (cid:13) Mauerhan et al. et al. 2012), which can rival the luminosity of a SN andspectroscopically mimic SNe IIn. Although the observationaldistinction between interacting SNe and SN impostors isnot always clear, a direct link connecting these phenomenahas been established for several objects. The progenitors ofSN 2006jc and SN 2011ht were detected undergoing lumi-nous outbursts 1–2 yr prior to their core-collapse explosions(Foley et al. 2007; Pastorello et al. 2008; Fraser et al. 2013a).SN 2010mc (Ofek et al. 2013a) was also detected 1–2 monthsbefore dramatically brightening, although for this case itappears that the precursor event may actually have beenthe initial phases of the SN caught unusually early (Smith,Mauerhan, & Prieto 2014).Perhaps the most interesting transient observed tohave undergone multiple phases of eruptive mass loss isSN 2009ip. Originally classified as a SN (Maza et al. 2009),this object was actually discovered during an LBV outburst— that is, as a SN impostor (Smith et al. 2010; Foley et al.2011). Several years of continued activity were followed pho-tometrically and spectroscopically, leading up to SN 2009ip’smost extreme outburst in 2012 (Mauerhan et al. 2013; Prietoet al. 2013; Levesque et al. 2014; Smith et al. 2013b; Smith,Mauerhan, & Prieto 2014; Pastorello et al. 2013; Fraser et al.2013b; Margutti et al. 2014; Graham et al. 2014). The 2012event was comprised of two main components: the “2012a”phase, marked by an initial peak at M ≈ − . M = − . M ≈ −
13 mag.Based on spectral similarities with known core-collapseSNe, Mauerhan et al. (2013a) suggested that the relativelyfaint 2012a phase of SN 2009ip marked the initial stages ofa core-collapse SN, while the subsequent 2012b brighteningwas the result of interaction between this SN and dense CSMejected during the earlier LBV outbursts. Levesque et al.(2014) also shared the view that the 2012b brightening wasthe result of interaction between the 2012a outflow and exist-ing CSM. However, several authors have suggested potentialalternatives to a core collapse explosion – that SN 2009ip’s2012 evolution was possibly the result of one or more nonter-minal outbursts (Pastorello et al. 2013; Fraser et al. 2013b;Margutti et al. 2014), perhaps caused by the pulsationalpair-instability mechanism. The motivation for nonterminalscenarios has been based largely on the fact that the totalradiated energy of the light curve (1 . × erg) can beexplained by an explosion energy of < ergs (if spher-ical symmetry is assumed for the CSM), and also becausethe late-time data were interpreted as looking different fromwhat is expected for radioactive decay phases. More recently,Smith, Mauerhan & Prieto (2014; hereafter SMP14) showedthat the 2012 light curve and spectral evolution are consis-tent with published models for core-collapse SNe, which canbe initially faint (like SN 1987A) owing to a relatively com-pact progenitor radius (i.e., a blue supergiant, as expectedfor an LBV, instead of a red supergiant), while the late-timecharacteristics could be explained if the mass of synthesized Ni was half that of SN 1987A. SMP14 further argued thatthe radiated energy did not provide an argument againstcore collapse, since the CSM is probably aspherical, leadingto an inefficient conversion of kinetic energy into radiation.In line with the spectral modeling results and interpreta- tion of Levesque et al. (2014), which are consistent with atoroidal distribution of CSM, SMP14 also proposed a disk-like distribution of CSM, but further argued that significantasphericity is required by the fact that broad photosphericlines are still seen at late times, even after strong CSM inter-action ends. This requires that a large fraction of the totalsolid angle of ejecta was able to expand unimpeded by CSM.SMP14 also pointed out that the ∼ ⊙ , and is incompatible with a 0.1 M ⊙ shell,which would become optically thin much more quickly. Amass of a few M ⊙ moving at high speeds ( ∼ − )directly implies & ergs of kinetic energy.In the case of SNe IIn, the system geometry is criticalif the total radiated energy from CSM interaction is to beused to infer the kinetic energy of the SN explosion. Fortu-nately, spectropolarimetry can yield important clues aboutthe geometry of SNe, allowing us to stringently test the hy-pothesis of aspherical CSM. Here, we report multi-epochspectropolarimetry of the 2012 evolution of SN 2009ip, fromthe end of the initial 2012a phase, through the peak of 2012b,and later into its decline. The results unambiguously demon-strate that the source of bright continuum emission duringthe 2012b phase (i.e., the CSM interaction zone) was as-pherical, consistent with a toroidal or disk-like distributionof CSM proposed by SMP14 and Levesque et al. (2014). Fur-thermore, the results indicate that the initial 2012a phasehas a separate polarization component having an axis ofsymmetry that is distinct from 2012b, which implies thatthe 2012a event did not create the CSM responsible for the2012b brightening. Diminishing polarization at late times,in addition to similarities in wavelength-dependent polar-ization across lines for SN 2009ip and other SNe, providesadditional evidence for a persistent underlying SN photo-sphere. Altogether, the available evidence supports the hy-pothesis of a SN explosion for SN 2009ip in 2012, and arguesstrongly against nonterminal explosion models for this ob-ject (at least those proposed thus far). Spectropolarimetric observations were obtained at the fol-lowing facilities utilised by the University of Arizona: the6.5 m Multiple Mirror Telescope (MMT) on Mt. Hopkinsduring 2012 Sep. 21 and 24 (UT dates are used throughoutthis paper); the Kuiper 61-inch telescope on Mt. Lemmonduring Nov. 11 and 14; and the Bok 2.3 m telescope on KittPeak during Dec. 5, 6, and 7. Observations at these facilitiesall made use of the same CCD Imaging/Spectropolarimeter(SPOL; Schmidt et al. 1992a). The instrument contains a ro-tatable semiachromatic half-waveplate used to modulate in-cident polarization and a Wollaston prism in the collimatedbeam to separate the two orthogonally polarized spectraonto a thinned, anti-reflection-coated 800 × − gratingin first order. The slit selection was based on the seeingconditions and ranged from 1 . ′′ to 5 . ′′ × ′′ . This setupresulted in spectral resolutions of ∼ λλ c (cid:13) , 1– ?? pectropolarimetry of SN 2009ip first-order spectrum was not contaminated by second-orderlight for λ & ◦ , aligned alongnorth-south. A series of four separate exposures that sam-ple 16 orientations of the waveplate yields two independent,background-subtracted measures of each of the normalisedlinear Stokes parameters, Q and U . On some nights sev-eral such polarization sequences of SN 2009ip were obtainedand combined, with the weighting of the individual measure-ments based on photon statistics.The instrumental polarization of SPOL and the Kuiper,Bok and MMT telescopes is extremely small ( < . ◦ Q and U spectra. The linearpolarization position angle on the sky ( θ ) was determined byobserving one or more of the interstellar polarization stan-dards Hiltner 960, VI Cyg ◦
389 (Schmidt etal. 1992b) during all epochs. The adopted correction fromthe instrumental to the standard equatorial frame for θ forall epochs was determined from the average position angleoffset of the polarized standards. Differences between themeasured and expected polarization position angles were < . ◦ The Shane 3 m reflector at Lick Observatory equipped withthe Kast spectrograph (Miller & Stone 1993) was used onOct. 5 and 14, and on Nov. 6 and 14. Like SPOL, Kastis a dual-beam spectropolarimeter that utilises a rotatablehalf-waveplate and Wollaston prism. Only the red channelof Kast was used for spectropolarimetry. A GG455 order-blocking filter blocked all second-order light at wavelengthsshortward of 9000 ˚A. Observations were made with the300 line mm − grating and the 3 ′′ slit, yielding a spectralresolution of ∼
20 ˚A. The orientation of the slit on the skywas always set to a position angle of 0 ◦ , aligned along north-south. Exposures of 900 s were obtained at each of four wave-plate positions (0 . ◦
0, 45 . ◦
0, 22 . ◦
5, and 67 . ◦ ◦ Q and U val-ues to < . ◦ ◦
389 andHD 204827 to obtain the zero-point of the polarization po-sition angle on the sky ( θ ) and to determine the accuracyof polarimetric measurements, which were consistent withpublished values within 0.05%.Since the Oct. 5 observations were hampered by inter-mittent cloud cover, no polarization standard stars could beobserved. Thus, for this epoch we assumed negligible instru-mental polarization, relying on the fact that the instrumen-tal polarization was below 0.05% for the Oct. 14, Nov. 6,and Nov. 14 observations, and we used the position-anglecorrection curve from Oct. 14, which we have verified to bevery stable over time. We also adopted an offset angle of5.9 ◦ to further calibrate the Oct. 5 observations; this is thesame offset value measured in the Oct. 14, Nov. 6, and Nov.14 observations, to within 0.5 ◦ .All data obtained from the Arizona and Lick observa-tories were extracted and calibrated using generic iraf rou-tines and our own idl functions. Spectropolarimetric analy-sis was also performed in iraf and idl following the meth-ods described by Miller et al. (1988) and implemented byLeonard et al. (2001). We also obtained spectropolarimatric data on SN 2009ipfrom the European Southern Observatory (ESO) publicarchive. We analysed data from the Very Large Telescope(VLT) and Focal Reducer and low-dispersion Spectrograph(FORS2) instrument (Appenzeller et al. 1998) obtained on2012 Nov. 7, Dec. 6, and Dec. 10, which were part of thetarget-of-opportunity programme ID 290.D-5006. We usedthe 300V grism data for Nov. 7 and Dec. 6; these data havespectral resolution of ∼
13 ˚A and wavelength coverage of ∼ ∼ ◦ by the instrument rotator. Standard stars from Fos- c (cid:13) , 1– ?? Mauerhan et al. sati et al. (2007) were observed for calibrations. The high-polarization standard NGC 2024-1 was observed on Nov. 6and Dec. 7 to calibrate the zero-point of the polarizationposition angle of the polarimeter on the sky, and the low-polarization standards WD 2039 −
202 and WD 2149+021were observed on Nov. 1 and Dec. 7, respectively, to gaugeany instrumental polarization. The instrumental positionangle spectrum was obtained directly from the ESO publicarchive. On Nov. 1 the observation of WD 2039 −
202 yieldedlinear Stokes parameters of (
Q, U ) < . V band, while on Dec. 6, WD 2149+021 also yielded( Q, U ) < . fors pmos calib and fors pmos science reduction recipeswere used, and executed using the Gasgano graphical userinterface software. After all reduction and calibrations wereperformed, the data were corrected for the redshift of thehost galaxy NGC 7259 ( z = 0 . Linear polarization is expressed as the quadratic sum of the Q and U Stokes parameters, P = p Q + U , and the po-sition angle on the sky is given by θ = 1 / − ( U/Q ),while carefully taking into account the quadrants in the Q – U plane where the inverse tangent angle is located. Since P is a positive-definite quantity, it is significantly overes-timated in situations where the signal-to-noise is low. Itis typical to express the “de-biased” (or “bias-corrected”)form of P as P db = q ( Q + U ) − ( σ Q + σ U ), where σ Q and σ U are the uncertainties in the Q and U Stokes param-eters. If ( σ Q + σ U ) > ( Q + U ), then we set a 1 σ upperlimit on P of q σ Q + σ U . In cases where P/σ P < . θ is essentially undetermined. All polarized spectra presentedherein are displayed in this manner. Note, however, thatat low signal-to-noise ratio P db is also not a reliable func-tion, as it has a peculiar probability distribution (Miller etal. 1988). Thus, for extracting statistically reliable values ofpolarization within a particular waveband, we have binnedthe calibrated Q and U Stokes spectra separately over thewavelength range of interest before calculating P and θ . Allquoted and tabulated values in this paper were determinedin this manner, although spectra are displayed as P db , sothey may exhibit slight offsets from our quoted values. Interstellar polarization (ISP) is commonly a very problem-atic element of polarimetric analysis, as an accurate probeof the ISP from both the Milky Way and the host galaxy canbe very difficult to obtain. Fortunately, several factors indi-cate that the ISP in the direction of SN 2009ip must be verysmall, and should not significantly impact our polarimetryresults and the physical interpretation that follows.First, the high-resolution VLT/FORS2 flux spectrum
Figure 1.
VLT/FORS2 total-flux spectrum (normalised to thecontinuum) near the Na D region, obtained on 2012 Nov. 7. Thespectrum has not been corrected for cosmological redshift. Galac-tic interstellar Na D absorption exhibits a very small equivalentwidth of ∼ § on Nov. 7, shown in Figure 1, reveals the resolved com-ponents of Galactic interstellar Na I D absorption lines at5890.1 ˚A and 5896.3 ˚A. The equivalent width (EW) of eachcomponent is a very small ∼
40 m˚A, which implies a Galactic E ( B − V ) < .
01 mag (Munari & Zwitter 1997). Accordingto Serkowski et al. (1975), the value of E ( B − V ) can beused to derive an upper limit on Galactic ISP. The so-calledSerkowski relation is given by ISP < E ( B − V ), which im-plies that ISP < .
1% in the direction of SN 2009ip.Second, earlier photometric estimates of the total red-dening toward the host galaxy NGC 7259 and SN 2009ipindicate a low value of E ( B − V ) = 0 .
019 mag (Schlegel etal. 1998; Smith et al. 2010; Foley et al. 2011), which impliesthat the host galaxy also does not produce substantial ISPalong the line of sight. This is also consistent with the lack ofNa D absorption at the redshift of NGC 7259; in fact, Na Dappears in weak emission, perhaps from a local H ii regionor from gas excited by SN radiation, with no indication ofadditional absorption components.The low host extinction is not particularly surprising,given SN 2009ip’s large radial distance from the center ofthis nearly face-on spiral galaxy. If we were to naively applythe Serkowski relation to the total measured reddening, wewould obtain ISP < . Figure 2 shows the total flux spectra, P , and θ of SN 2009ipfor epochs between Sep. 21 and Dec. 6. To gauge the total V -band polarization P and θ , we first binned Q and U sepa-rately in the wavelength range 5050–5950 ˚A. The extractedvalues are listed in Table 1. Figure 3 shows the resulting po- c (cid:13) , 1– ?? pectropolarimetry of SN 2009ip Figure 2.
Total flux, fractional debiased polarization P db , andposition angle θ for SN 2009ip during its 2012 outburst, at epochsbetween Sep. 21 and Dec. 6. The three Kuiper/SPOL epochs ( red )have been averaged for clarity. The Sep. 21 and 24 MMT/SPOLdata ( light blue ) have been averaged and binned to 300 ˚A; allother epochs are binned to 50 ˚A. Note the ∼ ◦ offset in θ between the 2012a phase and the peak of 2012b. larization light curve and the temporal evolution in the Q – U plane. In addition to V -band, we also sampled “continuum”regions of 5200–5500 ˚A and 6100–6200 ˚A, because they ap-pear mostly devoid of strong line features; these ranges areclose to the respective “green” and “red” wavelength regionsused by Hoffman et al. (2008). We caution, however, thatthere is no location within the spectrum that is completelyfree of line emission/absorption during all epochs, so the re-gions we have selected should be regarded as pseudocontinuathat avoid the strong lines.The Sep. 21 and 24 epochs took place near the end ofSN 2009ip’s 2012a phase, before the sharp rise in bright-ness that marks the onset of the 2012b phase. The to-tal flux spectrum during 2012a exhibits the characteristicsof SNe IIn, displaying intermediate-width ( ∼ − )emission lines superimposed on a spectrum with broad P-Cygni lines ( ∼ − ) that is reminiscent of more com-mon SNe II-P (Mauerhan et al. 2013). During this early Figure 3.
Upper panel : Temporal evolution of the total V -band(5050–5950 ˚A) polarization for SN 2009ip during the 2012a and2012b phases. For reference, an arbitrarily scaled version of the V -band light curve is plotted in the background. Lower panel : Evo-lution in the Q – U plane. Interstellar polarization is constrainedto < θ V = 166 ◦ and 72 ◦ measuredat the first epoch on Sep. 21 and peak polarization on Oct. 14,respectively. Our latest continuum measurement (5200–5500 ˚A)for Dec. 6 is marked with a red cross near the origin, consistentwith our ISP limit. The approximate point-reflection symmetrybetween 2012a and the peak of 2012b suggests two separate androughly orthogonal components of polarization on the sky. Notethe temporary shift in θ associated with the bump in the lightcurve at ∼
30 days past peak. epoch, SN 2009ip is polarized at ∼ θ ≈ ◦ , and it occupies quadrant IV in the Q – U plane.By Oct. 5, SN 2009ip has entered the 2012b phase andthe flux spectrum changes substantially; the broad emissioncomponents have become strongly diluted by the strengthen-ing continuum, which is quite blue, while the intermediate-width emission features have developed broad Lorentzianwings, a commonly observed characteristic of SNe IIn andan indication of high optical depth (Chugai 2001; Dessart etal. 2009; Smith et al. 2008), probably resulting from CSMinteraction. At this time, P has changed only marginally to c (cid:13) , 1– ?? Mauerhan et al.
Table 1.
Spectropolarimetry of SN 2009ipDate (2012 UT) day a Tel./Instr. P green (%) P red (%) P V (%) θ V (deg)Sep. 21 −
13 MMT/SPOL 0.96 (0.10) ::: 0.89 (0.08) 166 (5)Sep. 24 −
10 MMT/SPOL 1.15 (0.09) 0.99 (0.11) 0.90 (0.08) 178 (5)Oct. 5 b − < θ red = 132 (3) a Day is with respect to the day of peak in the V band (JD 2456207.72). b Data for Oct. 5 interrupted by poor weather; required calibration using Oct. 14 standard-star observations; to be used with caution. ∼ θ ≈ ◦ , moving into quadrant II of the Q – U plane. In-terestingly, this dramatic change in position angle indicatesseparate polarization components for the 2012a and 2012bphases, with distinct axes of symmetry. The lack of a cor-respondingly significant change in fractional P associatedwith the large rise in brightness could possibly be explainedby the partial cancellation of the polarization vectors fromthe 2012a and 2012b components, as the latter rises in lu-minosity.By Oct. 11, the source has risen to peak flux. The spec-trum has retained the overall appearance of Oct. 5, whilethe continuum becomes less blue and broad-line P-Cygni ab-sorption becomes apparent for He i /Na D. Our Lick/Kastand Kuiper/SPOL measurements on Oct. 14 show that P has risen to a maximum of ∼ θ has reached ∼ ◦ .At this time, strong deficits in polarization are seen for theintermediate-width emission components of H α and, to alesser extent, H β . This is likely to be the result of dilutionfrom the strong, intrinsically unpolarized emission lines inthe outer CSM. Meanwhile, a slight increase in P is seen at5600–5800 ˚A, suggesting a potential association with broadHe i λ Q – U plane for the Oct. 14 Lickepoch, binned to 100 ˚A. For reference, we also show scaledlogarithmic flux spectra in the same figure, which has beennormalised with a low-order spline (avoiding the strongestemission features for the spline nodes). On Oct. 14, nearpeak brightness, the entire optical spectrum of SN 2009ipoccupies a compact portion of the Q – U plane, in quad-rant II . The points form a slightly elongated cluster thatshows a continuous trend in wavelength, with the bluestpoints exhibiting the lowest position angle in the Q – U plane.The cluster as a whole exhibits a small but non-negligiblespread in position angle, which corresponds to a range of ∼ ◦ on the sky. The dilution in polarization associatedwith intermediate-width H α emission is apparent as a pro-trusion from the cluster of data points which roughly pointstoward the origin. Note that our binning causes the line to be unresolved, and this affects the strength and position an-gle of the protrusion; the unbinned data show the protrusionextending very close to the origin. In total polarization P , wesee broad, shallow variations that appear to be associatedwith broad absorption components of H α and He i /Na Dblueward of 5876 ˚A. It is interesting to note that the scaledflux spectrum included in Figure 4 shows that the broad P-Cygni emission features from fast ejecta, first seen during2012a, are still present throughout the peak of 2012b; thishas important physical consequences for the energetics ofSN 2009ip (see § i /Na Dabsorption has deepened, and the intermediate-width com-ponent of H α has developed an absorption component ontop of the broad underlying feature, perhaps part of its ownP-Cygni profile. Interestingly, this epoch coincides with theappearance of a temporary albeit substantial bump in thedeclining light curve. The Nov. 7 VLT/FORS2 spectropo-larimetry data are shown in detail in the lower panels ofFigure 4. At this time, the V -band polarization has droppedto ∼ ∼ ◦ . The enhancement of polarization near He i /Na i absorption feature has become more pronounced and sharp;this feature is also apparent in Lick data from Nov. 6, shownin Figure 2. Meanwhile, the sharp dip in polarization pro-duced by the dilution from unpolarized intermediate-widthBalmer lines is no longer apparent.We examine the polarization enhancements of two spe-cific line features in Figure 5, which show the Stokes pa-rameters plotted as a function of velocity, with respect tothe He i λ α rest wavelengths. On Nov. 7, the en-hanced polarization of He i /Na D is substantially blueshiftedto a velocity of 13,000 km s − (if it is associated with He)and exhibits a velocity width of ∼ − . The peak isalso blueshifted by ∼ − with respect to the fluxminimum of the absorption line. The enhanced He i /Na Dfeature forms a linear-shaped loop that roughly traces theaverage polarization axis of the system for that date. c (cid:13) , 1– ?? pectropolarimetry of SN 2009ip Figure 4. Q and U Stokes parameters and polarization for SN 2009ip on 2012 Oct. 14 (from Lick/Kast, upper panels), Nov. 7 andDec. 6 (VLT/FORS2, 300V grism, middle and lower panels). The colors have been chosen to correspond with wavelength, but note thedifferent wavelength scales for the Lick and VLT data. The data have been binned to 100 ˚A. Scaled versions of the total flux spectra areplotted in the background for reference (solid grey curves).c (cid:13) , 1– ?? Mauerhan et al.
Figure 5.
VLT/FORS2 Q and U Stokes parameters and polarization for SN 2009ip on 2012 Nov. 7 for spectral regions near 5876 ˚A(upper panels; 300V grism; binned to 40 ˚A) and H α (lower panels; 1200R grism; binned to 30 ˚A). The right-hand plots are with respectto velocity. The colors have been chosen to correspond with velocity, but note the different velocity scales for the upper and lowerpanels. Scaled versions of the total-flux spectra are plotted in the background for reference (black curves in right-hand panels). Enhancedpolarization is associated with He i λ − − . The enhancementforms a linear loop in the Q – U plane that roughly follows along the V -band (5050–5950 ˚A) symmetry axis for that epoch (2 θ V , thickblack line). Note that the weak emission line redward of 5876 ˚A is Na D emission (see Figure 1). For H α , enhanced polarization appearsto be associated with the intermediate-width absorption feature of H α , while the broad underlying feature exhibits a loop in the Q – U plane, within the range of position angles bounded by 2 θ r and 2 θ V . There also appears to be an enhancement in fractionalpolarization for the absorption component blueward of theintermediate-width H α emission feature (at − − ).The origin of this absorption feature is unclear. It couldbe an absorption component of a P-Cygni profile for theintermediate-width H α component, or it could be the resultof blueshifted absorption for He i λ α P-Cygni, then this enhanced polarization feature does notappear blueshifted as in the case of the broad He i /Na D ab-sorption. In the Q – U plane, the track across this potentialH α /He i feature appears to be roughly consistent with theoverall axis of symmetry. There also appears to be some in-dication of polarization enhancement for blend of absorptioncomponents seen at the higher velocities of ∼ km s − and larger, which are probably from H α . Overall, the broademission component of H α appears to make a loop in the Q – U plane.Averaging all of our measurements from Kuiper/SPOL on Nov. 12 and 14, we detect ∼ V band, similar to the value of the preceding VLT/FORS2epoch on Nov. 7. The Lick/Kast data solely from Nov. 14show a value of polarization similar to the Nov.12–14 com-bined average of the Kuiper/SPOL data. The polarized spec-tra are not shown in Figure 2, to avoid cluttering the figure,although their broadband measurements are presented inFigure 3. Interestingly, although the overall degree of polar-ization has not changed substantially between Nov. 6 and 14,the position angle has shifted again back toward the axis ofsymmetry of the 2012b peak. It thus appears that the bumpin the light curve on Nov. 6–7 is associated with anothercomponent of polarization, perhaps related the 2012a lumi-nosity source.A few weeks later, on Dec. 5–7, we detect weak polar-ization in an average of three separate measurements withBok/SPOL, obtaining P V ≈ .
3% and only an upper limitfor the continuum of P green < . c (cid:13) , 1– ?? pectropolarimetry of SN 2009ip Figure 6.
Same a Figure 5 but for later epochs. On Dec. 6–10 (lower two panels), enhanced polarization is now associated with thelower velocity component of the double-dipped absorption feature, produced by either He i or Na D. Enhanced polarization is also seenfor the blend of absorption components blueward of the H α line center. The He i /Na D and H α features both appear to make loops inthe Q – U plane, roughly along the average symmetry axes of the given epochs. measurement takes place on Dec. 6, for which we obtain P green ≈ .
16% and P V ≈ . III with θ ≈ ◦ .The flux spectrum during this late epoch shows that thecontinuum has become redder, while the broad P-Cygni fea-tures are still present. He i /Na D absorption has becomea double-dipped profile. Figure 6 shows that by Dec. 6 theenhanced polarization of this feature has shifted redward,and is probably associated with the bluer component of thedouble-dipped profile. A complicated blend of multiple ab-sorption and perhaps emission features appears blueward ofH α at this time, and also exhibits evidence for enhancedpolarization, roughly along the average axis of symmetry. Examining more closely the temporal evolution of SN 2009ipin the Q – U plane, shown in Figure 3 (lower panel), the rangeof possible ISP values constrained by the value of E ( B − V )(see § Q and U during Dec. is the result of a migration back to the point of ISP, somewhere near the origin. In-deed, our latest measurement of the continuum polarizationyielded P green < .
22% from Bok/SPOL for Dec. 5–7, and P green = 0 .
16% from VLT/FORS2 on Dec. 6 ( q = − . u = − . Q and U are destined to drift back to the location inquadrant IV occupied by our earliest MMT/SPOL measure-ments in September (i.e., that the 2012a phase is intrinsicallyunpolarized, and instead reflects the location of ISP). Butsuch a large value of ISP ( & E ( B − V )and EW(Na D); the structure in P for the September epochsalso does not follow a Serkowski law, which is inconsistentwith the 2012a polarization being the result of ISP. Further-more, it would be a rather remarkable coincidence for theposition angle of ISP to be orthogonal to the polarizationaxis of SN 2009ip at peak.Therefore, the evidence is strong that our earliest po-larization detections from MMT/SPOL in September revealtrue polarization intrinsic to the source during the 2012aphase, and that the ISP is within < .
2% of the origin,and likely . . c (cid:13) , 1– ?? Mauerhan et al. measurement. As one can see, any value of ISP within thisallowable range will affect our results only slightly, and willnot change the fact that the 2012a and 2012b phases appearto have two separate components of polarization that willremain roughly orthogonal between 2012a and the peak of2012b, no matter where the ISP lies within the range illus-trated by Figure 3 (lower panel).
The most natural explanation for linear polarization of a SNis Thomson scattering by free electrons, either from the SNphotosphere or within the zone of intense CSM interaction,which includes the forward shock, the reverse shock, andthe cold dense shell (CDS) in between. For most SNe IIn,including SN 2009ip, evidence of electron scattering is alsoindicated by the broad Lorentzian wings typically seen atthe bases of the intermediate-width emission-line profiles(Chugai et al. 2001; Smith et al. 2008; Dessart et al. 2009).But in the case of scattered photons emerging from a spheri-cally symmetric distribution of material, the sum of all elec-tric vectors will cancel out and yield zero net polarization.Thus, the detection of net polarization from any SN directlyimplies aspherical structure, either intrinsic or imposed frompartial obscuration of the polarized surface by interveningmaterial (see Wang & Wheeler 2008 for a review on SN po-larization).Core-collapse SNe of Type II-P have been observed tobe weak sources of polarization while in their plateau phases,which implies that the outer layer of ejecta is roughly spher-ical in some cases. However, near the end of their plateaus,SNe II-P can exhibit brief but significant increases in polar-ization (Leonard et al. 2000, 2006; Chornock et al. 2010),because as the ejecta cool and recombine the photosphererecedes from the spherical outer layers of H-rich materialinto more aspherical He-rich and metal-rich layers deeperinside the ejecta. Thus, to first order, SNe II-P can be char-acterized by a single component of photospheric polarizationthat scales inversely with the decreasing luminosity. SNe IIn,on the other hand, appear to exhibit a high degree of po-larization while in their brightest phases, at least for thefew objects that have been studied in sufficient detail (e.g.,Leonard et al. 2001; Hoffman et al. 2008). This implies thatCSM interaction luminosity must be a major source of thenet polarization from SNe IIn. In this case, since both theSN photosphere and the CSM interaction zone are potentialsources of polarized photons, changes in the relative lumi-nosities of these components can lead to variations in netpolarization, even if the fractional polarization and positionangles of each component remain constant with time. Below,we propose this to be the case for SN 2009ip.
SN 2009ip’s photospheric component is made evident by itsspectropolarimetric similarities with the Type IIb SN 1993J,shown in Figure 7. Near peak brightness, both SNe exhibita comparable degree of polarization across the optical band, with SN 2009ip being stronger in the continuum, and theyboth display remarkably similar wavelength-dependent vari-ations across their polarization spectra, particularly evidentat wavelengths of 5200–7000 ˚A. In the case of SN 1993J,these variations are associated with the broad P-Cygni linesof H α and He i +Na D, and the same is probably true forSN 2009ip. The variations are produced by the depolariz-ing effect of line scattering (Trammell et al. 1993; H¨oflich1995; H¨oflich et al. 1996; Tran et al. 1997). This is not tobe confused with the intrinsically unpolarized recombina-tion emission from the outer optically thin CSM that we seein SN 2009ip, which dilutes the polarized continuum andresults in relatively narrow “absorption” features of polar-ization for the H α and H β lines. Rather, true depolarizationis important for the broad emission lines that develop in theionised ejecta above the electron-scattering photosphere ofthe SN. The persistence of these lines in SN 2009ip for ∼ ⊙ ), consistent with a SN outflow (see SMP14).CSM is also a potential contributor to the polarizationproperties of SN 1993J (H¨oflich et al. 1996; Tran et al. 1997),via electron scattering within a SN/CSM interaction zoneand/or the scattering of photons by dust. After all, the con-tinuum polarization of SN 1993J being entirely due to anaspherical ejecta photosphere is somewhat difficult to recon-cile with the evolution of a highly spherical radio remnantrevealed by VLBI (Bietenholz, Bartel, & Rupen 2003), al-though this structure might only trace the morphology of theoutermost H-rich layers, not the more asymmetric He-richlayers potentially responsible for the polarized continuum.Still, the source did exhibit evidence for pre-existing CSMat early times, in the form of narrow emission-line featuresexhibiting small velocity widths of ∼
170 km s − , indicativeof an ambient stellar wind (Benetti et al. 1994). However,these features were less pronounced than those of SN 2009ipand lasted for a shorter amount of time, indicating a muchless dense CSM. SN 1993J’s X-ray properties were also in-dicative of CSM interaction (Leising et al 1994), while its ra-dio evolution indicated a changing pre-SN mass-loss rate inthe range ˙ M = 10 − –10 − M ⊙ yr − (Van Dyk et al. 1994),which, although roughly 100 times weaker than the pre-SNmass-loss rate inferred for SN 2009ip (Ofek et al. 2013b),could generate enough CSM to influence the object’s polar-ization properties.Therefore, like SN 2009ip, SN 1993J could have twocomponents of polarization — photospheric and CSM in-teraction. In the case of SN 2009ip, however, it appearsrather obvious that CSM interaction is the dominant sourceof polarization near photometric peak. The opposite couldbe true for SN 1993J: its polarization could be dominatedby an aspherical explosion and only moderately influencedby the CSM component. In this regard, a noteworthy dif-ference between the two SNe is that near peak brightnessSN 1993J exhibits more variation than SN 2009ip in θ asa function of wavelength, suggesting that SN 2009ip has amore highly ordered axis of symmetry in 2012b resultingfrom its luminous and highly aspherical CSM interactionzone. By 50–60 days after peak, however, both SNe exhibitremarkable similarity in P and θ as a function of wavelength.It appears that as CSM interaction fades after the peak of2012b and the broad P-Cygni lines return by Nov. 6, thepolarization characteristics of SN 2009ip begin to look more c (cid:13) , 1– ?? pectropolarimetry of SN 2009ip Figure 7.
Comparison between the flux and polarization properties of SN 2009ip ( black ) and SN 1993J ( red ). Both unbinned andsmoothed versions of the data are shown for each SN. No scaling or adjustments have been applied. The SN 1993J April 30 data weretraced from H¨oflich et al. 1996) and the May 11 data are from Tran et al. (1997). For the earlier epochs, which are near the mainphotometric peaks of both SNe, they appear very similar in spectral polarization. SN 2009ip exhibits little deviation in θ while thevariations for SN 1993J (Tran et al. 1997) are more pronounced. At 50–60 days past peak polarization, both SNe exhibit very similarvariations across their spectra, and their position angle variations become more comparable. like a SN photosphere again. It is thus interesting that thesource makes a pronounced shift in position angle on Nov. 6,associated with the obvious bump in the light curve (see Fig-ure 3), which could represent the temporary brightening ofthe photospheric component of polarization that dominatedthe 2012a phase. Taken at face value, SN 2009ip’s degree of peak polarizationfor the 2012a and 2012b phases (from P ≈ .
9% to 1 . < < >
15% and >
30% asphericity), if we assumethat polarization translates directly to the apparent geom-etry via comparison with the oblate electron-scattering at-mosphere models of H¨olfich (1995). However, the true phys-ical axial ratios could be substantially more aspherical thanthese limits, for two key reasons: (1) inclination angle, and(2) partial cancellation between the 2012a and 2012b polar-ization vector components. If both the 2012a peak and thelate decline of the 2012b phase (from the Nov. 6–7 bumponward) are dominated by the SN photosphere component,as appears to be the case, then partial cancellation could beparticularly important for the 2012b peak. Indeed, the ∼ ◦ shift in position angle between the earliest 2012a observa-tion and the peak of the 2012b phase implies two separatecomponents of polarization that are roughly orthogonal toone another. Since orthogonal polarization vectors cancel, the degree of intrinsic polarization for the 2012b componentmust be higher than it appears (i.e., even more aspherical),because the orthogonality of the separate 2012a vector com-ponent, if still present during 2012b, will partially cancel the2012b component.We can estimate the true intrinsic polarization of the2012b peak component as follows. If we rotate the Q – U val-ues from Figure 3 such that they are aligned with the axisof symmetry at maximum polarization ( θ = 72 ◦ ), whichroughly traces the average axis of symmetry for the en-tirety of 2012b, then the majority of polarization duringthe 2012b phase is shifted onto the positive Q axis, whilethe orthogonal 2012a polarization gets shifted onto the neg-ative Q axis. Figure 8 shows the evolution of the resultingrotated Stokes parameters. The Q rot curve represents vari-ations along the main axis of symmetry, while U rot tracesdeviations from the average axis of symmetry. The resultimplies that if the 2012a component maintains its orthog-onal contribution to the total electric vector during 2012b,then the separation between the 2012a and 2012b compo-nents will approximately define the true degree of intrinsicpolarization for the 2012b peak, which, therefore, could beas high as ∼ . . > ◦ inclination angle, hypothetically, it is possible that the in-trinsic geometry of the 2012b component is a highly flat-tened one. Below we discuss a plausible geometric scenario c (cid:13) , 1–, 1–
30% asphericity), if we assumethat polarization translates directly to the apparent geom-etry via comparison with the oblate electron-scattering at-mosphere models of H¨olfich (1995). However, the true phys-ical axial ratios could be substantially more aspherical thanthese limits, for two key reasons: (1) inclination angle, and(2) partial cancellation between the 2012a and 2012b polar-ization vector components. If both the 2012a peak and thelate decline of the 2012b phase (from the Nov. 6–7 bumponward) are dominated by the SN photosphere component,as appears to be the case, then partial cancellation could beparticularly important for the 2012b peak. Indeed, the ∼ ◦ shift in position angle between the earliest 2012a observa-tion and the peak of the 2012b phase implies two separatecomponents of polarization that are roughly orthogonal toone another. Since orthogonal polarization vectors cancel, the degree of intrinsic polarization for the 2012b componentmust be higher than it appears (i.e., even more aspherical),because the orthogonality of the separate 2012a vector com-ponent, if still present during 2012b, will partially cancel the2012b component.We can estimate the true intrinsic polarization of the2012b peak component as follows. If we rotate the Q – U val-ues from Figure 3 such that they are aligned with the axisof symmetry at maximum polarization ( θ = 72 ◦ ), whichroughly traces the average axis of symmetry for the en-tirety of 2012b, then the majority of polarization duringthe 2012b phase is shifted onto the positive Q axis, whilethe orthogonal 2012a polarization gets shifted onto the neg-ative Q axis. Figure 8 shows the evolution of the resultingrotated Stokes parameters. The Q rot curve represents vari-ations along the main axis of symmetry, while U rot tracesdeviations from the average axis of symmetry. The resultimplies that if the 2012a component maintains its orthog-onal contribution to the total electric vector during 2012b,then the separation between the 2012a and 2012b compo-nents will approximately define the true degree of intrinsicpolarization for the 2012b peak, which, therefore, could beas high as ∼ . . > ◦ inclination angle, hypothetically, it is possible that the in-trinsic geometry of the 2012b component is a highly flat-tened one. Below we discuss a plausible geometric scenario c (cid:13) , 1–, 1– ?? Mauerhan et al.
Figure 8.
Temporal evolution of the rotated Stokes parameters,aligned to the axis of symmetry exhibited during maximum polar-ization in 2012b phase ( θ = 72 ◦ ). The plot is meant to illustratethe full possible degree of intrinsic polarization for the 2012b com-ponent ( ∼ § for SN 2009ip’s 2012 evolution, facilitated by Figures 9 and10. For the 2012b component of polarization the situationappears relatively clear. Since CSM interaction dominatesthe luminosity during this phase, and since the associatedpolarization source exhibits a rather steady and well-definedaxis of symmetry, the net polarization must reflect the phys-ical structure of the dense CSM, which in this case couldrepresent a toroidal geometry (a disk or ring). Toroidal ge-ometry for the interacting CSM has already been suggestedfor SN 2009ip by Levesque et al. (2014), based on spec-tral modeling and analysis of the Balmer emission lines.Those authors found that the H α emission during the peakof 2012b requires a total radiating area of ∼ ,while the Balmer decrement requires a high CSM densityof n e > cm − , for Case-B recombination (note thatsuch a high density would naturally explain the absence offorbidden emission lines in SN 2009ip’s spectrum). The com-bination of high density and large area implies a flattenedgeometry (i.e., a disk/ring), possibly as thin as 10 AU. Al-ternatively, a lower density CSM of n e > cm − in alimb-brightened spherical shell configuration was also pro-posed by Levesque et al. (2014) as an additional possibility,although less favourably because it requires that the opticaldepth of H β be lower than for H α , which would be unusual.Independently, SMP14 also favoured disk-like geometry forthe CSM based on energetic arguments and on the persis-tence of broad P-Cygni lines even after the phase of intenseCSM interaction had declined, which implies that a largefraction of the ejecta did not decelerate and, thus, must nothave participated in strong CSM interaction (toroidal ge-ometry would allow a large fraction of the ejected mass to bypass strong CSM interaction in directions orthogonal tothe CSM plane). Our spectropolarimetric results are consis-tent with toroidal geometry for the CSM around SN 2009ip,and contradict models that invoke spherical distributions ofCSM. In this case, the axial ratio of the polarized emissionsource at the peak of 2012b (constrained to < .
7, and prob-ably < § inconsistent withsome physical scenarios previously proposed as possibilitiesfor SN 2009ip’s 2012 evolution. Since it is the CSM geome-try that determines the polarization properties at the peakof the interaction phase, then the earlier eruption that pro-duced such CSM should reflect the same geometry. Thus, theorthogonal position angle shift we observe between 2012aand the peak of 2012b is not compatible with the interpre-tation that these respective phases represent the launchingand collision of successive shells of CSM (e.g., from pair-instability eruptions), a possibility suggested by other au-thors (Pastorello et al. 2013; Margutti et al. 2014), sincesuch a scenario should result in similar axes of symmetryfor the two phases, not orthogonal geometries. Rather, theorthogonality between the 2012a and 2012b events implies aphysically distinct origin for each of the associated outflows.In addition, the large shift in polarization and position an-gle between the 2012b peak and the subsequent bump inthe declining light curve on Nov. 6–7 also suggests that thebump is the result of a polarized luminosity component thatis separate from that of the 2012b peak — for example, theSN photospheric component brightening again as the CSMinteraction component fades away. This is inconsistent withthe interpretation that the bump in the light curve is the re-sult of a fluctuating central source that variably illuminatesthe CSM, such as activity from a central surviving star, asproposed by Martin et al. (2013).Toroidal CSM geometry would not be unique toSN 2009ip; it has been proposed before for other SNe IIn,specifically in the cases of SN 1988Z (Chugai & Danziger1994, their Model B), SN 1998S (Leonard et al. 2000), andSN 1997eg (Hoffman et al. 2008). This type of geometry isalso unsurprising in the context of SN 2009ip’s stellar pro-genitor, since rings and tori are common phenomena aroundLBVs, B[e] supergiants, and other evolved massive stars(Smith, Bally, & Walawender 2007; Smith et al. 2011a). Notat all clear, however, is the connection of such geometryto the erratic behaviour of SN 2009ip over the last decade.It is possible that dense equatorial CSM is the result ofbinary-influenced mass loss, as has been suggested to explainthe toroidal CSM of B[e] supergiants. A connection betweensuch objects and SN 2009ip was speculated upon by Clarket al. (2013) in the case of LHA 115-S 18, a B[e] supergiantexhibiting erratic long-term variability that is reminiscent ofLBVs. It is thus interesting to note that the radial dimen-sions of B[e] disks (10 –10 AU; Zickgraf et al. 1986) arecomparable to the size estimates for SN 2009ip’s toroidalCSM (Smith et al. 2013; SMP14).The origin of the polarization for the 2012a phase isless clear. Since the spectrum of SN 2009ip on Sep. 21–24resembles that of more common core-collapse SNe, the sim-plest explanation might be electron scattering within thephotosphere of the early SN, which has yet to impact the c (cid:13) , 1– ?? pectropolarimetry of SN 2009ip SN SN2009ip - 2012a phase SN ejectaphotospherebroad P-Cygniforward shock dense CSM dense CSM SN P NET = Obs. view rarified CSM φ Figure 9.
Illustration of a potential SN/CSM configuration dur-ing the 2012a event (not to scale), before the onset of intenseCSM interaction. The presence of broad P-Cygni lines implies arapidly expanding photosphere. The detection of significant con-tinuum polarization at this phase suggest an aspherical geome-try for the SN photosphere and/or partial absorption by toroidalCSM (inset). The approximate orthogonality of the polarizationposition angle with respect to the 2012b CSM-interaction phasesuggests that the SN photosphere might have a bipolar geometry.The angle φ represents the angle subtended by the CSM from thepoint of view of the explosion. dense CSM lying farther out. In this case, the axial ratioof < .
85 and the orthogonal geometric axis with respectto the peak of the CSM interaction phase would suggest aprolate/bipolar shape for the 2012a outflow. Interestingly,bipolar geometry has been suggested previously for othercases of highly polarized core-collapse SNe (e.g., see Wanget al. 2001). Alternatively, net polarization from the 2012aphase could also arise if the near side of the soon-to-be-shocked toroidal CSM is partially blocking the waist of theSN photosphere (Figure 9, inset), which could result in apolarization axis that is orthogonal to 2012b if the polar-ized poles of the photosphere remain relatively unobscured.Finally, scattering of the SN photons by dust in the outerCSM could provide another source of polarized photons. Af-ter all, infrared data obtained during the 2012a phase showstrong evidence for circumstellar dust at a radius of ∼ SN SN2009ip - 2012b phase forward shock
SN ejectabroad P-CygniH α / He Ι / Na Ι P P dense CSMdense CSM
CDSe - scatteringpolarized continuumCDSe - scatteringpolarized continuum P NET = Obs. view
Figure 10.
Illustration of a possible configuration for the 2012bphase, after the onset of intense CSM interaction. A toroidaldistribution of dense CSM gives rise to strong shocks, electronscattering, and a luminous polarized continuum (marked by abold-faced “P”). At the same time, diluted yet persistent broadP-Cygni lines from 2012a indicate an underlying SN photospherefrom fast ejecta that have not participated in strong CSM inter-action. With respect to the observer, fast He/Na-rich ejecta ina quasi-Hubble flow could partially obscure the inner polarizedcontinuum, resulting in the sharp polarization “enhancements”we see at specific velocities (see Figures 5 and 6). for the photosphere or partial blocking of its waist by thetoroidal CSM.
If strong CSM interaction is confined to a toroidal geome-try, then large amounts of high-velocity SN ejecta could ex-pand relatively unimpeded. This is already indicated by thepersistence of the broad P-Cygni lines (SMP14), but couldalso explain some of the higher-order effects we see in thespectropolarimetric data. For example, the sharp blueshiftedenhancement in polarization associated with the He i /Na Dfeature seen on Nov. 6–7 (Figure 5) is an indication of high-velocity material. This feature has a blueshifted velocity of13,000 km s − , while the position angle across the featureapproximately follows along the main axis of symmetry ofthis epoch in the Q – U plane. The fact that the peak ofthe enhanced polarization feature is not coincident with theflux minimum of the absorption line implies that the en-hancement cannot simply be the result of absorption of un-polarized or less polarized flux (which would increase thefractional polarization at that wavelength). Rather, the netblueshift probably results from an intervening layer of fastHe-rich (or Na-rich) material that lies near the outer edge ofthe ejecta and is partially absorbing the inner polarized con-tinuum. From the point of view of the observer, such partialblocking can increase the net polarization over that range ofvelocities by interfering with the cancellation of orthogonalpolarization vectors distributed across the polarized source c (cid:13) , 1–, 1–
If strong CSM interaction is confined to a toroidal geome-try, then large amounts of high-velocity SN ejecta could ex-pand relatively unimpeded. This is already indicated by thepersistence of the broad P-Cygni lines (SMP14), but couldalso explain some of the higher-order effects we see in thespectropolarimetric data. For example, the sharp blueshiftedenhancement in polarization associated with the He i /Na Dfeature seen on Nov. 6–7 (Figure 5) is an indication of high-velocity material. This feature has a blueshifted velocity of13,000 km s − , while the position angle across the featureapproximately follows along the main axis of symmetry ofthis epoch in the Q – U plane. The fact that the peak ofthe enhanced polarization feature is not coincident with theflux minimum of the absorption line implies that the en-hancement cannot simply be the result of absorption of un-polarized or less polarized flux (which would increase thefractional polarization at that wavelength). Rather, the netblueshift probably results from an intervening layer of fastHe-rich (or Na-rich) material that lies near the outer edge ofthe ejecta and is partially absorbing the inner polarized con-tinuum. From the point of view of the observer, such partialblocking can increase the net polarization over that range ofvelocities by interfering with the cancellation of orthogonalpolarization vectors distributed across the polarized source c (cid:13) , 1–, 1– ?? Mauerhan et al. (e.g., also see Kasen et al. 2003). The high-velocity He/Na D-rich material was probably launched during the SN in 2012aphase, while the CSM responsible for the 2012b phase waslikely to have been created by the LBV progenitor eruptionsduring the years prior (e.g., see Graham et al. 2014, theirFigure 7).
A toroidal or disk-like distribution of CSM around SN 2009iphas important consequences for the kinetic energy of the ex-plosion that is inferred from the total radiated energy mea-sured during the 2012b CSM-interaction phase. Taking the2012a explosion time of −
47 days, suggested by SMP14, andthe fastest ejecta velocity of 13,000 km s − indicated by theblue edge of the H α P-Cygni line during 2012a, implies aradial distance of ∼
350 AU from the point of explosion tothe inner edge of the dense CSM; this is consistent with thelimits on the minimum radius of the dust component in-ferred from infrared measurements during 2012a ( >
120 AU;Smith et al. 2013b), and with the 300 AU radial estimatesby Levesque et al. (2014). The latter authors showed thatthe CSM need only be ∼
10 AU thick to explain the H α lu-minosity in the toroidal scenario. For a 10 AU toroidal scaleheight, the 300–350 AU distance from the explosion wouldimply a very small subtended angle of φ ≈ ± ◦ for the CSMfrom the point of view of the explosion. Such a configura-tion inclined at an angle of 35–45 ◦ with respect to our line ofsight (from edge-on) would appear as an ellipse on the skyhaving an axial ratio that is consistent with the geometricconstraints placed by the value of peak polarization during2012b (see § π ) intercepted byCSM will be only 2–3% ( <
10% conservatively). This minoramount of interacting area will result in an inefficient con-version of the total kinetic energy of the SN into radiation.Thus, the ∼ ergs of kinetic energy inferred from theradiated energy of the explosion by Fraser et al. (2013) andMargutti et al. (2014), both of whom assume a sphericalCSM configuration in their calculations, must be a signifi-cant underestimate, by 1 dex or more. A kinetic energy of & ergs, as suggested by SMP14, seems more likely, giventhe demand for aspherical CSM by spectropolarimetry.One must also keep in mind that the shape of the ex-plosion is an influential factor as well. A bipolar explosion,which we have shown is a plausible geometry for the 2012aoutflow, will intersect even less area of the toroidal CSM. Ifthe material is clumpy, then the effective intersecting areacould be even lower than we have estimated, implying aneven lower efficiency conversion of kinetic energy into radi-ation. We have presented multi-epoch spectropolarimetry of the2012 outburst of SN 2009ip, covering the 2012a and 2012bphases. Since the available data imply a very low amountof interstellar absorption and ISP < . < < ≈
40 days apart(Margutti et al. 2014), since such a scenario should producesimilar axes of symmetry for the 2012a and 2012b phases,not orthogonal geometries. Furthermore, the geometric pa-rameters constrained by spectropolarimetry imply that onlya small fraction of the SN ejecta ( < ∼ ergs explosion that are based onthe total radiated energy, which assume spherical symme-try for the CSM (Fraser et al. 2013; Margutti et al. 2014),are likely to be substantially underestimated by an order ofmagnitude.The strong similarities between SN 2009ip andSN 1993J, from the time near their peak through theirdecline, provide another strong line of evidence for theexistence of an underlying SN photosphere from high-velocity ejecta (5000–8000 km s − ) that persists throughoutSN 2009ip’s 2012 evolution ( ∼
100 days). Such a long-lastingoptically thick component indicates a mass of at least a fewM ⊙ for the fast outer component of the ejecta, and the as-sociated speeds imply a kinetic energy of & ergs for theejecta, as suggested by SMP14.After peak polarization, higher-order structure in the Q – U plane from He- i /Na D indicates the presence of fastmetal-rich ejecta that have overrun the dense CSM, consis-tent with a high-velocity flow that bypassed intense CSMinteraction. Meanwhile, the weakening of polarization withdeclining continuum flux at late times suggests that thediminishing intensity of electron scattering is responsiblefor the decline in polarization. The fact that the flux andpolarization drop more rapidly for SN 2009ip relative toother SNe IIn, such as SN 1997eg (Hoffman et al. 2008) orSN 2010jl (Grant Williams, 2014, private communication),also suggests that the CSM around SN 2009ip has a compactconfiguration, compared to these other explosions.The lines of evidence presented here thus favour a sce-nario first proposed by Mauerhan et al. (2013), in which theluminosity of the 2012a phase of SN 2009ip was the resultof a & ergs explosion (i.e., a core-collapse SN) thatsubsequently plowed into an aspherical (probably toroidal)distribution of CSM during 2012b, creating the jump in lu- c (cid:13) , 1– ?? pectropolarimetry of SN 2009ip minosity and the strong polarization of the source. Nonter-minal scenarios are difficult to support in light of the currentbody of evidence, and we would have to invent a processcapable of generating ejecta having & ergs of kineticenergy without destroying the star. Still, the best evidenceof progenitor death is a vanishing of the star at late times.However, CSM interaction can persist for decades or more,and generate enough luminosity to make progenitor disap-pearance a difficult observable to confirm. SN 1961V is aparticularly controversial example (Filippenko et al. 1995;Chu et al. 2004; Smith et al. 2011b; Kochanek et al. 2011;Van Dyk et al. 2012). Moreover, it is very plausible that theprogenitor of SN 2009ip’s 2012 explosion left a massive com-panion at the explosion site, not only because massive starsare rarely solitary (Sana et al. 2012), but because toroidalor disk-like geometry for the CSM is suggestive of binaryinfluence. As discussed by Smith & Arnett (2014), repeatedbrief pre-SN eruptions could result from binary interactionif the primary suddenly increases its radius during late nu-clear burning stages, thereby triggering binary interactionand mass ejection. Finally, we note that strong net polariza-tion consistently observed from SNe IIn, as a class, impliesthat CSM interaction is a dominanting influence on theirpolarization properties, and that an aspherical distributionof CSM is a commonplace feature of the stellar progenitorsof these explosions. ACKNOWLEDGEMENTS
We thank the referee for their review of this manuscript. We thank the staffsat Lick, the MMT, and Steward Observatories for their excellent assistance.We thank S. Bradley Cenko for assisting with the Lick observations. Someobservations reported here were obtained at the MMT Observatory, a jointfacility of the University of Arizona and the Smithsonian Institution. Thisresearch was also based, in part, on observations made with ESO Telescopesat the La Silla Paranal Observatory under programme ID 290D-5006. Wethank Hien Tran for supplying polarimetric data on SN 1993J for our com-parison. J.C.M. thanks Dan Kasen at UC Berkeley for insightful discussion,and we also thank Leah N. Huk at U. Denver for helpful commentary. Thisresearch was supported by NSF grants AST-1210599 (UA), AST-1211916(UC Berkeley), AST-1210311 (SDSU), and AST-1210372 (U. Denver). Thesupernova research of A.V.F.’s group at U.C. Berkeley is also supportedby Gary & Cynthia Bengier, the Richard & Rhoda Goldman Fund, theChristopher R. Redlich Fund, and the TABASGO Foundation.
REFERENCES
Appenzeller, I., Fricke, K., F¨urtig, W., et al. 1998, The Messenger, 94, 1Arnett W.D. 1996, Supernovae and Nucleosynthesis (Princeton: PrincetonUniv. Press)Benetti, S., Patat, F., Turatto, M., et al. 1994,A&A, 285, L13Bietenholz, M. F., Bartel, N., & Rupen, M. P. 2003, ApJ, 597, 374Chevalier, R. A., & Fransson, C. 1994, ApJ, 420, 268Chornock, R., Filippenko, A. V., Li, W., & Silverman, J. M. 2010, ApJ,713, 1363Chu, Y.-H., Gruendl, R. A., Stockdale, C. J., et al. 2004, AJ, 127, 2850Chugai, N. N. 2001, MNRAS, 326, 1448Chugai N. N., Danziger I. J. 1994, MNRAS, 268, 173Clark, J. S., Bartlett, E. S., Coe, M. J., et al. 2013, A&A, 560, A10Dessart L., Hillier D.J., Gezari S., Basa S., & Matheson T. 2009, MNRAS,394, 21Filippenko, A. V. 1997, ARA&A, 35, 309Filippenko, A. V., Barth, A. J., Bower, G. C., et al. 1995, AJ, 110, 2261Foley, R. J., Smith, N., Ganeshalingam, M., et al. 2007, ApJL, 657, L105Foley R. J., Berger E., Fox O., et al. 2011, ApJ, 732, 32Fossati, L., Bagnulo, S., Mason, E., & Landi Degl’Innocenti, E. 2007, TheFuture of Photometric, Spectrophotometric and Polarimetric Standard-ization, 364, 503Fraser, M., Magee, M., Kotak, R., et al. 2013a, ApJL, 779, L8Fraser, M., et al. 2013b, MNRAS, 433, 1312Graham, M. L., Sand, D. J., Valenti, S., et al. 2014, arXiv:1402.1765 Hoffman, J.L., Leonard, D.C., Chornock, R., Filippenko, A.V., Barth,A.J., & Matheson, T. 2008, 688, 1186H¨oflich, P. 1995, ApJ, 440, 821H¨oflich, P., Wheeler, J. C., Hines, D. C., & Trammell, S. R. 1996, ApJ,459, 307Kasen, D., Nugent, P., Wang, L., et al. 2003, ApJ, 593, 788Kochanek, C. S., Szczygiel, D. M., & Stanek, K. Z. 2012, ApJ, 758, 142Kochanek, C. S., Szczygiel, D. M., & Stanek, K. Z. 2011, ApJ, 737, 76Leising, M. D., Kurfess, J. D., Clayton, D. D., et al. 1994, ApJL, 431, L95Leonard D. C., Filippenko A. V., Barth A. J., Matheson T. 2000, ApJ,536, 239Leonard, D. C., Filippenko, A. V., Ardila, D. R., & Brotherton, M. S.2001, ApJ, 553, 861Leonard, D. C., Filippenko, A. V., Chornock, R., & Li, W. 2002, AJ, 124,2506Leonard, D. C., Filippenko, A. V., Ganeshalingam, M., et al. 2006, Nature,440, 505Levesque, E. M., Stringfellow, G. S., Ginsburg, A. G., Bally, J., & Keeney,B. A. 2014, AJ, 147, 23Margutti, R., Milisavljevic, D., Soderberg, A. M., et al. 2014, ApJ, 780,21Martin, J. C., Hambsch, F.-J., Margutti, R., et al. 2013, arXiv:1308.3682Mauerhan, J. C., Smith, N., Filippenko, A. V., et al. 2013, MNRAS, 430,1801Maza, J., et al. 2009, CBET, 1928, 1Miller, J. S., Robinson, L. B., & Goodrich, R. W. 1988, Instrumentationfor Ground-Based Optical Astronomy, 157iller J. S., Stone R. P. S., 1993, Lick Obs. Tech. Rep. 66. Lick Obs., SantaCruzMunari, U., & Zwitter, T. 1997, A&A, 318, 269Ofek, E. O., et al. 2013a, Nature 494, 65Ofek, E. O., et al. 2013b, ApJ, 768, 47Pastorello, A., et al. 2008, MNRAS, 389, 131Pastorello, A., et al. 2013, ApJ, 767, 1Prieto, J.L., Brimacombe, J., Drake, A.J., & Howerton, S. 2013, ApJ, 763,L27Sana, H., de Mink, S. E., de Koter, A., et al. 2012, Science, 337, 444Schlegel, E. M. 1990, MNRAS, 244, 269Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525Schmidt, G. D., Elston, R., & Lupie, O. L. 1992a, AJ, 104, 1563Schmidt, G. D., Stockman, H. S., & Smith, P. S. 1992b, ApJL, 398, L57Serkowski, K., Mathewson, D. S., & Ford, V. L. 1975, ApJ, 196, 261Smith, N., ARA&A, 52, in press (arXiv:1402.1237)Smith, N., & Arnett, D. 2014, ApJ, in press (arXiv:1307.5035)Smith N. 2006, ApJ, 644, 1151Smith, N., & Owocki, S. P. 2006, ApJL, 645, L45Smith, N., Bally, J., & Walawender, J. 2007, AJ, 134, 846Smith N. 2011, MNRAS, 415, 2020Smith, N., Gehrz, R. D., Campbell, R., et al. 2011a, MNRAS, 418, 1959Smith, N., Li, W., Silverman, J. M., Ganeshalingam, M., & Filippenko,A. V. 2011b, MNRAS, 415, 773Smith N. 2013a, MNRAS, 429, 2366Smith, N., Mauerhan, J. C., & Prieto, J. L. 2014 (SMP14), MNRAS, 438,1191Smith N., & McCray, R. 2007, ApJ, 671, L17Smith N., Vink J., de Koter A., 2004, ApJ, 615, 475Smith, N., Li, W., Foley, R. J., et al. 2007, ApJ, 666, 1116Smith N., Chornock R., Li W., et al. 2008, ApJ, 686, 467Smith N., Miller A., Li W., et al. 2010, AJ, 139, 1451Smith N., et al. 2011a, MNRAS, 415, 773Smith N., Mauerhan, J.C., Kasliwal, M., & Burgasser, A. 2013, MNRAS,434, 2721Trammell, S. R., Hines, D. C., & Wheeler, J. C. 1993, ApJL, 414, L21Tran, H. D., Filippenko, A. V., Schmidt, G. D., et al. 1997, PASP, 109,489Van Dyk S. D., Peng C. Y., King J. Y., et al. 2000, PASP, 112, 1532Van Dyk, S. D., Weiler, K. W., Sramek, R. A., Rupen, M. P., & Panagia,N. 1994, ApJL, 432, L115Van Dyk, S. D., & Matheson, T. 2012, ApJ, 746, 179Wang, L., Howell, D. A., H¨oflich, P., & Wheeler, J. C. 2001, ApJ, 550,1030Wang, L., & Wheeler, J. C. 2008, ARAA, 46, 433Zickgraf, F.-J., Wolf, B., Leitherer, C., Appenzeller, I., & Stahl, O. 1986,A&A, 163, 119 c (cid:13) , 1–, 1–