High-resolution spectroscopy of SN 2017hcc and its blueshifted line profiles from post-shock dust formation
aa r X i v : . [ a s t r o - ph . H E ] S e p MNRAS , 1– ?? (2012) Preprint 1 October 2020 Compiled using MNRAS L A TEX style file v3.0
High-resolution spectroscopy of SN 2017hcc and itsblueshifted line profiles from post-shock dust formation
Nathan Smith ⋆ and Jennifer E. Andrews Steward Observatory, University of Arizona, 933 N. Cherry Ave., Tucson, AZ 85721, USA
ABSTRACT
SN 2017hcc was remarkable for being a nearby and strongly polarized superlumi-nous Type IIn supernova (SN). We obtained high-resolution echelle spectra that wecombine with other spectra to investigate its line profile evolution. All epochs revealnarrow P Cygni components from pre-shock circumstellar material (CSM), indicatingan axisymmetric outflow from the progenitor of 40-50 km s − . Intermediate-width andbroad components exhibit the classic evolution seen in luminous SNe IIn: symmetricLorentzian profiles from pre-shock CSM lines broadened by electron scattering at earlytimes, transitioning at late times to multi-component, irregular profiles coming fromthe SN ejecta and post-shock shell. As in many SNe IIn, profiles show a progressivelyincreasing blueshift, with a clear flux deficit in red wings of the intermediate andbroad velocity components after day 200. This blueshift develops after the continuumluminosity fades, and in the intermediate-width component, persists at late times evenafter the SN ejecta fade. In SN 2017hcc, the blueshift cannot be explained as occul-tation by the SN photosphere, pre-shock acceleration of CSM, or a lopsided explosionor CSM. Instead, the blueshift arises from dust formation in the post-shock shell andin the SN ejecta. The effect has a wavelength dependence characteristic of dust, ex-hibiting an extinction law consistent with large grains. Thus, SN 2017hcc experiencedpost-shock dust formation and had a mildly bipolar CSM shell, similar to SN 2010jl.Like other superluminous SNe IIn, the progenitor lost around 10 M ⊙ due to extremeeruptive mass loss in the decade before exploding. Key words: binaries: general — stars: evolution — stars: massive — stars: winds,outflows — supernovae (general)
Throughout their lives, massive stars shed mass throughsteady winds, but these winds can be punctuated by episodicmass loss events (Smith 2014). Observationally-derived windmass-loss rates have been revised downward compared torates still used in most stellar evolution codes, undermin-ing some basic predictions of single-star models (Smith2014). Mass-loss rates have been reduced for both line-drivenwinds of hot stars (Bouret et al. 2005; Fullerton et al. 2006;Puls et al. 2008; Sundqvist et al. 2019), and more recentlyalso for dusty red supergiant (RSG) winds (Beasor et al.2020).On the other hand, there has been a shift to strongerinfluence of eruptive mass loss (Smith & Owocki 2006)and binary mass exchange and stripping (Sana et al. 2012; ⋆ E-mail: [email protected]
Moe & Di Stefano 2017; G¨otberg et al. 2018). Eruptivemass loss like luminous blue variables (LBVs) or violent bi-nary interaction are less sensitive to metallicty than line-driven winds (Smith & Owocki 2006; G¨otberg et al. 2017),making them more relevant to supernovae (SNe) that arisein lower-metallicity regions. Eruptive LBV-like mass lossand violent binary interaction may be related, since merg-ers and mass gainers are needed to explain various prop-erties of LBVs (Justham et al. 2014; Smith & Tombleson2015; Aghakhanloo et al. 2017; Smith et al. 2018a).Eruptive mass loss has become especially prominent inthe case of strongly interacting SNe of Types IIn (SNe IIn)and Ibn (SNe Ibn), where observed signatures of interac-tion with dense circumstellar material (CSM) indicate ex-treme and short-lived mass-loss phases immediately preced-ing core collapse; see Smith (2017) and Smith (2014) forreviews. The specific mechanism(s) of the eruptive pre-SNmass loss remains uncertain, but it requires a trigger syn-chronized with core-collapse. Some proposed mechanismsinclude energy deposited in the envelope by wave driv- c (cid:13) Smith et al. ing in advanced nuclear burning phases (Quataert & Shiode2012; Shiode & Quataert 2014; Fuller 2017; Fuller & Ro2018), the pulsational pair instability or other late-phaseburning instabilities (Woosley et al. 2007; Woosley 2017;Smith & Arnett 2014; Renzo et al. 2020), or an inflationof the progenitor’s radius (perhaps caused by the previ-ous mechanisms) that triggers violent binary interaction likecollisions or mergers before core collapse (Smith & Arnett2014). Of these, only the last predicts highly asymmet-ric distributions of CSM (disk-like or bipolar), relevant toasymmetric line profile shapes and high polarization seen inSNe IIn.An extreme case of eruptive pre-SN mass loss is foundin super-luminous SNe (SLSNe) with exceptionally strongCSM interaction, where the very high luminosity is thoughtto result from an explosion with normal to high explo-sion energy (1-5 × erg) running to a very high mass ofCSM, usually of order a few to 20 M ⊙ . Some well stud-ied examples include SN 2006gy, SN 2006tf, SN 2008am,SN 2003ma, SN 2015da, and SN 2010jl (Smith et al. 2007,2010a; Smith & McCray 2007; Chatzopoulous et al. 2011;Rest et al. 2011; Tartaglia et al. 2020; Gall et al. 2014;Fransson et al. 2014; Jencson et al. 2016; Woosley et al.2007) These events have high efficiency in converting ejectakinetic energy into luminosity of order 10% −
50% or more(van Marle et al. 2010), with total radiated energy budgets & erg. An important feature of strongly interacting SNe is theformation of a cold dense shell (CDS) between the for-ward and reverse shocks (Chugai 2001; Chugai et al. 2004;Smith et al. 2008b). Efficient radiative cooling causes thedense post-shock gas to collapse into a thin, dense, clumpy,and probably well-mixed layer (van Marle et al. 2010). Thisrapid cooling that forms a dense shell is a unique featureof interacting SNe, causing their defining narrow-line spec-tra and their high luminosity. This rapid cooling and highdensity may also trigger early dust formation.There are 3 common observational signatures of newdust formation in SNe: (1) a strengthening IR excess con-sistent with dust emission, (2) an increased rate of fading inthe optical continuum that may be attributed to increasedextinction from new dust, and (3) a progressive and system-atic blueshift in emission-line profiles caused by dust thatpreferentially obscures the redshifted portions of the explo-sion. Each of these alone is somewhat ambiguous — an IRexcess could be due to an IR echo from CSM dust (Fox et al.2011; Andrews et al. 2011), the rate of fading can depend onother factors unrelated to dust, and blueshifted line profilescan arise from asymmetry in the ejecta or CSM, or opticaldepth effects at early times — but seeing all three together,as in the case of SN 1987A (Danziger et al. 1989; Lucy et al.1989; Gehrz & Ney 1989; Wooden et al. 1993; Colgan et al.1994), gives a strong indication that new dust has formed.Of these three, the third is unique in helping to diagnose Here “dust formation” may also mean pre-existing grains in theCSM that were incompletely destroyed by the forward shock andthat then regrow. the location of the dust because of the different charactis-tic expansion speeds in the unshocked CSM, SN ejecta, andthe post-shock CDS. The characteristic blueshift of emissionlines is fairly common in interacting SNe, and has been tiedto dust formation in the SN ejecta and post-shock CDS.The first clear case of an interacting SN that showedall three signatures of dust formation was SN 2006jc,where the line-profile evolution of He i lines (this wasa Type Ibn) indicated that new dust formed in thepost-shock CDS (Smith et al. 2008a). This blueshift coin-cided in time with an IR excess and fading in the con-tinuum, but also a burst of X-ray emission and He ii λ η Car(Smith 2010). Similar blueshifted profiles have beenseen in a number of H-rich interacting SNe, includingSN 2006tf (Smith et al. 2008b), SN 2005ip (Smith et al.2009a; Fox et al. 2009; Smith et al. 2017), SN 2007rt(Trundle et al. 2009), SN 2007od (Andrews et al. 2010),SN 2010bt (Elias-Rosa et al. 2018), and SN 2010jl (discussedbelow).The line profile blueshift seen in interacting SNe hasusually been interpreted as post-shock or ejecta dust forma-tion. Line profile blueshift gives a more direct probe of thedust location, as noted above, but it is harder to infer thedust mass from extinction’s effect on line profiles withoutdetailed models and assumptions (Bevan et al. 2018, 2020).One should realize that these options for the location of thedust are not mutually exclusive. Dust may form in both theCDS and the inner SN ejecta at various times, and theremay also be pre-existing CSM dust. In fact, we expect pre-existing dust in the CSM for SNe IIn because of extremeeruptive mass loss (like LBVs or extreme RSGs), and thoseare the same progenitors most likely to have the high CSMdensity to trigger efficient radiative cooling and collapse ofthe CDS, which in turn permits efficient dust formation inthe post-shock region. There has, however, been discussionin the literature about alternatives to dust formation to ex-plain the blueshifts, as illustrated by the case of SN 2010jl.
SN 2010jl was among the nearest SLSNe IIn, leading toconsiderable observational attention with high-quality opti-cal/IR photometry and spectroscopy (Andrews et al. 2011;Stoll et al. 2011; Smith et al. 2011a, 2012a; Zhang et al.2012; Fox et al. 2013; Maeda et al. 2013; Gall et al. 2014;Fransson et al. 2014; Borish et al. 2015; Williams & Fox2015; Jencson et al. 2016; Sarangi et al. 2018; Chugai 2018;Bevan et al. 2020). While pre-existing CSM dust probablycontributed to the observed IR excess (Andrews et al. 2011),even early spectra revealed a prominent blueshift in lineprofiles that strengthened with time (Smith et al. 2012a;Gall et al. 2014).The systematic blueshift of line profiles and their wave-length dependence (more pronouned at shorter wavelengths)
MNRAS , 1– ?? (2012) N 2017hcc led to the suggestion that, like SN 2006jc, SN 2010jl ex-perienced new dust formation in the post-shock region ofthe CDS (Smith et al. 2012a). Several additional studiesconfirmed these signs of dust formation in the post-shockCDS and investigated the dust properties, including addi-tional late-time spectra, IR data, and models (Gall et al.2014; Sarangi et al. 2018; Chugai 2018; Bevan et al. 2020).Dust may have been pre-existing in the CSM, and dust mayhave formed in the ejecta (Andrews et al. 2011; Bevan et al.2020).The conclusion that the blueshift in line profiles wasinfluenced by dust formation in the CDS was not, however,adopted by all authors. Namely, Fransson et al. (2014) pro-posed a different picture where the blueshifted profiles werecaused by acceleration of asymmetric pre-shock CSM alongthe line of sight, and where the broad components werecaused by electron scattering of the narrow CSM emissionfrom that accelerated CSM. This explanation did not ac-count for why the narrow components from the unshockedCSM were not blueshifted even though the broader compo-nents had a blueshifted centroid, or why there was a wave-length dependence to the asymmetry. Subsequent radia-tive transfer models of SN 2010jl’s spectrum (Dessart et al.2015) found that accelerated pre-shock CSM could not ex-plain the observed blueshift in line profile shape, or their be-havior with time. Dessart et al. (2015) showed that while theearly symmetric line profiles were caused by electron scatter-ing of narrow CSM emission, the broad blueshifted profilesat later times arises in the post-shock CDS. These modelsrevealed a blueshifted emission bump that could arise evenwithout dust, at least initially, because the photosphere inthe CSM interaction region could block the redshifted sideof the CDS, as noted earlier by Smith et al. (2012a). Again,however, this mechanism would not account for the observedwavelength dependence of the blueshift (Smith et al. 2012a;Gall et al. 2014).If the systematic blueshift of the broader componentswere due mostly to an optical depth effect in the CDS, asseen in these models for SN 2010jl spectra up until aboutday 200 (Dessart et al. 2015), then there is a clear predictionfor the late-time evolution. Namely, at late times these linesshould become more symmetric, as the optical depth dropsand the continuum luminosity fades, revealing emission fromthe receding side (Smith et al. 2012a; Dessart et al. 2015).This did not happen in SN 2010jl. The continuum lumi-nosity dropped around day 300, but the blueshift persistedeven in spectra beyond day 1000 (Fransson et al. 2014). Thiswould seem to clearly rule out high continuum optical depthsand electron scattering of accelerated CSM as the explana-tion for the persistent blueshift, instead favoring dust forma-tion in the post-shock shell of SN 2010jl (Smith et al. 2012a;Gall et al. 2014; Sarangi et al. 2018; Bevan et al. 2020). SN 2017hcc was discovered on 2017 October 2 by theAsteroid Terrestrial-impact Last Alert System (ATLAS;Tonry 2011), and was classified as a young Type IIn event(Prieto et al. 2017). It was reportedly located a few arcsecsoutheast of the center of an anonymous host galaxy. Fig-ure 1 shows an MMT/Binospec image of the SN (see detailsbelow), indicating that it is located 0.5 arcsec south and 4.5
Figure 1.
A late-time r -band image of SN 2017hcc and itshost galaxy taken on 2019 July 9 using Binospec on the MMT.SN 2017hcc is found about 0.5 arcsec south and 4.5 arcsec eastof the center of its spiral host galaxy. arcsec east of the center of its host galaxy, and appears co-incident with the leading edge of a spiral arm. The SN wasdiscovered within a few days of explosion, judging by a non-detection a few days earlier. Following Prieto et al. (2017),who presented early photometry, we adopt an explosion dateof 2017 Oct 1.4, which we set as t = 0 in this study. As inPrieto et al. (2017), we adopt a redshift of z =0.0168 and adistance of 73 Mpc. We also adopt E ( B − V )=0.0285 mag forour line-of-sight through the Milky Way interstellar medium(Schlafly & Finkbeiner 2011) to deredden our spectra, al-though this has little impact on our analysis of line profiles.At this distance, its peak V magnitude of 13.7 mag impliesan absolute magnitude of − V -band luminosity about 40-45 daysafter explosion (Prieto et al. 2017). This is a relatively slowrise, although not as slow as the unusual case of SN 2006gy,which took ∼
70 days (Smith et al. 2007). SN 2017hcc wasnot detected at early times in either X-rays with
Chandra or in the radio (Chandra et al. 2017; Nayana & Chandra2017), although such indicators are often faint at early timeswhen X-ray and radio emission are absorbed.Perhaps the most notable property of SN 2017hcc sofar is that spectropolarimetry obtained soon after discoveryby Mauerhan et al. (2017a) indicated very high continuumpolarization of 4.8% or more. This may be the highest levelof polarization seen in any SN to date, and points to signifi-cant asymmetry in the interaction region of this SN IIn. Thisasymmetry indicated by polarization is relevant to the inter-pretation of the observed evolution of emission-line profileshapes for SN 2017hcc that we discuss below. In section 2 wepresent our new spectroscopic observations, in section 3 wepresent results from the analysis of line profiles in our spec-tra, and in section 4 we discuss the interpretation of theseline profiles and corresponding implications for the natureof the CSM and for dust formation in SN 2017hcc.
MNRAS , 1– ?? (2012)(2012)
MNRAS , 1– ?? (2012)(2012) Smith et al.
Table 1.
Low resolution spectroscopy of SN 2017hcc
UT Date day a Tel./Instr. grating ∆ λ ( µ m)2017 10 26 25 MMT/Blue 1200 0.57-0.702017 10 27 26 Bok/BC 300 0.44-0.872017 11 19 49 MMT/Blue 1200 0.57-0.702017 12 09 69 Bok/BC 300 0.44-0.872017 12 15 75 MMT/Blue 1200 0.57-0.702017 12 28 88 MMT/Red 1200 0.62-0.702018 01 01 92 Bok/BC 300 0.44-0.872018 06 30 271 MMT/Blue 1200 0.57-0.702018 08 31 336 MMT/MMIRS zJ/HK 1.0-2.32018 09 05 340 MMT/Blue 300 0.57-0.702018 10 18 382 MMT/Bino 600 0.51-0.752018 11 17 412 MMT/MMIRS zJ/HK 1.0-2.32019 07 08 645 MMT/Bino 600 0.51-0.752019 11 02 762 MMT/Bino 600 0.51-0.752020 01 27 848 Mag/IMACS 1200 0.63-0.67 a We adopt 2017 Oct 1 as day zero, close to the explosion dateinferred by Prieto et al. (2017).
We obtained a deep r -band image of SN 2017hcc on 2019July 9 (day 646) using the imaging mode of Binospec(Fabricant et al. 2019) on the Multiple Mirror Telescope(MMT). The image was reduced using a custom pythonpipeline and is shown in Figure 1. SN 2017hcc is locatedabout 0.5 arcsec south and 4.5 arcsec east of the centerof its spiral host galaxy (a foreground star or cluster pro-jected near the galaxy nucleus would shift the centroid ofthe galaxy light slightly to the east in lower-resolution im-ages). SN 2017hcc is therefore projected a total of 4.53 arcsecfrom center, or ∼ , it is coincident with the UV sourceGALEX 2674128878581058535. The coordinates listed forthis GALEX source are offset a few arcsec from the appar-ent center of the galaxy. SN 2017hcc appears to be locatedon the leading edge of an inner spiral arm or bar, althoughthe ground-based angular resolution is insufficient to deter-mine its proximity to any dust lanes or star forming regions. We obtained spectra of SN 2017hcc using the 6.5-m MultipleMirror Telescope (MMT) with three different instruments,including the Bluechannel (BC) spectrograph, the Redchan-nel (RC) spectrograph, and the newly commissioned Bi-noSpec spectrograph (Fabricant et al. 2019). Each MMTBluechannel or Redchannel observation was taken with a1.0 arcsec slit and either the 1200 lines mm − grating cov-ering a range of approximately 58007000 ˚A, or the 300 linesmm − grating covering a range of about 3600-8000 ˚A. Stan-dard reductions were carried out using IRAF including bias https://github.com/KerryPaterson/ http://simbad.u-strasbg.fr/simbad/ IRAF, the Image Reduction and Analysis Facility, is distributedby the National Optical Astronomy Observatory, which is oper-ated by the Association of Universities for Research in Astronomy subtraction, flat-fielding, and optimal extraction of the spec-tra. Flux calibration was achieved using spectrophotometricstandards observed at an airmass similar to that of each sci-ence frame, and the resulting spectra were median combinedinto a single 1D spectrum for each epoch.We obtained several late epochs of visual-wavelengthspectra using Binospec on the MMT. These data were alltaken using the 600 lines mm − grating centered on 6300 ˚A(covering a range of roughly 5100–7500 ˚A) and with a 1.0arcsec slit. All data were reduced using the Binospec pipeline(Kansky et al. 2019), which includes an internal flux calibra-tion into relative flux units from throughput measurementsof spectrophotometric standard stars.We obtained a few epochs of low-resolution optical spec-tra with the Boller & Chivens (B&C) spectrograph on the2.3 m Bok telescope. We also obtained one late-time spec-trum using the Inamori-Magellan Areal Camera and Spec-trograph (IMACS; Dressler et al. 2011) mounted on the 6.5m Baade telescope of the Magellan Observatory. Data reduc-tion for these followed standard reduction for point sourcesin long-slit optical spectra, as above. Our low/moderate-resolution spectroscopic observations are summarized in Ta-ble 1, and all are plotted in Figure 2. We observed SN 2017hcc on three separate occasions us-ing the Magellan Inamori Kyocera Echelle (MIKE), whichis a double echelle spectrograph designed for use at theMagellan Telescopes at Las Campanas Observatory inChile (Bernstein et al. 2003). We obtained observations ofSN 2017hcc during its main luminosity peak on 2017 Oct25 UT (a total exposure time of 2700 sec), and at two laterepochs after it faded by several magnitudes on 2018 Jul 11and Sep 17 UT (total exposure times of 5400 sec and 3600sec, respectively). All three observations had good trans-parency and seeing (roughly 0.8 arcsec or better), and allhad the 1 arcsec × R = λ /∆ λ of roughly 30,000, or a resolution of typically 10 km s − (al-though we note that the achieved resolution measured fromsky lines was somewhat narrower than this, about 8 km s − ,because the seeing was better than the slit width on all threeepochs).The spectra were reduced using the latest version of theMIKE pipeline (written by D. Kelson). The reduced spec-tra were corrected for SN 2017hcc’s redshift of z = 0.01686,and each epoch had velocities converted to the heliocen-tric reference frame (this correction matters for the narrowcomponents from the pre-shock CSM). Figure 3 shows theregion of the spectrum including the Na i D doublet use-ful for evaluating interstellar extinction and reddening, andalso includes He i λ α and H β , respectively.For H α , the line center was located near the edge of twoadjacent echelle orders, so to display the full profile in Fig-ure 4, we spliced together two adjacent echelle orders. Be- (AURA) under cooperative agreement with the National ScienceFoundation (NSF). http://code.obs.carnegiescience.edu/mike/MNRAS , 1– ?? (2012) N 2017hcc Figure 2.
Sequence of low-resolution optical spectra from the MMT, Bok, and Magellan (see Table 1). Flux is on a log scale. cause of the drop in sensitivity and the large throughputcorrections needed at the ends of echelle orders, we checkedthe resulting overall shape of the line profile by comparingit to lower-resolution single-order spectra taken very closein time (see above). These are plotted in orange in Figure 4,showing very good agreement in overall line shape. Only asingle echelle order is plotted for H β , because it was centeredin the middle of an ehelle order. For H β , we caution that thered wing of the line is partly blended with Fe ii lines, so thisshould not be interpreted as excess redshifted H β flux. Forboth H α and H β , the first epoch line profile is comparedto a Lorentzian line profile shape (thick light-blue curve),meant to match the line wings with FWHM = 2000 km s − for H α and 1600 km s − for H β . Figure 6 shows examplesof Gaussian components that may be fit to the day 351 H α line profile shape, discussed in more detail later.Figure 7 zooms-in on the narrow components of H α , H β ,and H γ seen in MIKE spectra, showing the P Cygni profilesarising from the unshocked CSM. For the first epoch, the “continuum” level is set at the continuum after subtractionof the Lorentzian profiles shown in Figures 4 and 5. For thetwo later epochs, the “continuum” for normalization refersto the flux level of the broader emission component within ±
400 km s − . Similarly, Figures 8 and 9 zoom-in on the nar-row CSM components of He i λ λ β .(We do not show He i λ α (black) and H β (blue) superposed on one another.We have chosen to align these by scaling the line flux tomatch the shape of the blue wing of the emission line, inorder to determine if there are any differences in the shapeof the red wing. The H β line strengths have therefore beenscaled arbitrarily in flux above the continuum level. For thefirst early epoch, the profile shapes of H α and H β agree MNRAS , 1– ?? (2012)(2012)
400 km s − . Similarly, Figures 8 and 9 zoom-in on the nar-row CSM components of He i λ λ β .(We do not show He i λ α (black) and H β (blue) superposed on one another.We have chosen to align these by scaling the line flux tomatch the shape of the blue wing of the emission line, inorder to determine if there are any differences in the shapeof the red wing. The H β line strengths have therefore beenscaled arbitrarily in flux above the continuum level. For thefirst early epoch, the profile shapes of H α and H β agree MNRAS , 1– ?? (2012)(2012) Smith et al.
Figure 3.
MIKE spectrum on day 24 in the region around Na i D and He i λ − grating is shown in orange for comparison. Expectedlocations of the D and D lines of the doublet are indicated byvertical blue bars for the redshift-corrected host galaxy and MilkyWay components. Narrow P Cygni absorption of He i λ i Dabsorption is barely detected in the MMT spectrum. remarkably well. For the later two epochs, however, the redwings of H β are slightly depressed compared to H α . Thisindicates that there is a wavelength dependence to the lineshape that we discuss later. We obtained two epochs of near-IR spectra covering the J , H , and K , bands (roughly 1 − µ m) using the MMT andMagellan Infrared Spectrograph (MMIRS; McLeod et al.2012) mounted on the MMT, with observations listed inTable 1. The standard long-slit zJ and HK single-order spec-tra were reduced using the MMIRS data reduction pipeline(Chilingarian et al. 2013). In this paper, we are most inter-ested in the Pa β µ m line profile shape, plotted alongwith H α line profiles in Figure 11. The high-resolution MIKE echelle spectra allow a sensitiveprobe of the interstellar absorption components from theNa i D doublet. The individual D and D lines are easily re-solved in the echelle spectrum, and the high resolution allowsgreater sensitivity to narrow absorption features, comparedto moderate-resolution spectra. For examining interstellarabsorption, we consider early spectra when the continuumis strongest.Figure 3 shows the region around Na i D, with theday 24 echelle spectrum (black) compared to the moderate-resolution spectrum taken with MMT one day later (or-ange). While both spectra have enough resolution to sep-arate the doublet, the absorption features are more clearlydefined in the MIKE spectrum. One can see the relativelystrong Na i D absorption from the Milky Way interstellar
Figure 4.
Full line profiles of H α as seen in the three epochsof MIKE spectra (black). The orange line profiles are from low-or moderate-resolution spectra taken around the same time, forcomparison. Because the H α line wings are so broad and theline center was located near the end of an echelle order, we usedone order for blueshifted velocities and the adjacent order forthe red wing of the line. The blue solid curve shows a symmetricLorentzian profile for comparison with the early epoch line profileshape. The flux is on a linear scale. Figure 5.
Same as Figure 4, but for H β . Zero velocity of H β was located near the center of a MIKE echelle order, so we onlyshow one full echelle order (there were slight flux mismatchesand increased noise at the edges of adjacent orders. Aside fromthe edges of the orders, the line shapes in echelle spectra andlower resolution spectra are consistent. The bump at high positivevelocity is emission from another line or lines (most likely Fe ii ),rather than a strong excess of high velocity H β emission. The fluxis on a linear scale. medium (ISM), which here is shifted to shorter wavelengthsbecause we are showing wavelengths corrected to the restframe of the host galaxy. One can see the weaker Na i D ab-sorption from the ISM in the host galaxy. In the lower resolu-tion spectrum, the Milky Way absorption is barely detected,and the host galaxy absorption is undetected. Figure 3 alsoshows the strong narrow P Cygni feature of He i λ i D,we measure equivalent widths of EW(D ) = 0.129 ( ± ) = 0.085 ( ± +D ) = MNRAS , 1– ?? (2012) N 2017hcc Figure 6.
Full line profile of H α as seen in the day 352 MIKEspectrum, compared to to different multicomponent Gaussians.The top panel (a) shows two Gaussian profiles and their sum(plus continuum) where the centroids are allowed to be offset fromzero, and the bottom (b) shows two similar Gaussians and theirsum (plus continuum), but with the center fixed at 0 km s − , andwhere the sum is allowed to extend above the flux at low velocitiesand the red wing of the line (the excess flux may represent lineflux lost from the intrinsic profile due to absorption). ± E ( B − V ) values of 0.023mag (D ), 0.028 mag (D ), and 0.025 mag (D +D ). Asnoted by Poznanski et al. (2012), there is a 30-40% sys-tematic uncertainty in the relation for E ( B − V ). Theseare in reasonable agreement with the Milky Way line ofsight reddening of E ( B − V )=0.0285 mag adopted earlier(Schlafly & Finkbeiner 2011).For the SN 2017hcc host galaxy ISM absorption com-ponent of Na i D, we measure smaller equivalent widths ofEW(D ) = 0.026 ( ± ) = 0.014 ( ± +D ) = 0.039 ( ± E ( B − V ) values of 0.014 mag (D ), 0.019 mag (D ), and0.016 mag (D +D ). A note of caution is warranted here, be-cause these host galaxy EW values we measure are below therange of EW over which the relations from Poznanski et al.(2012) were calibrated, which extend down to 0.05 ˚A, andthere is some indication that data deviate from the fit at the Figure 7.
Detail showing the narrow components of the brightestBalmer lines H α , H β , and H γ in the three epochs of echelle spec-tra. For the first epoch of each, the observed profile was dividedby an underlying broad Lorentzian (blue in Figure 4). For the twolater epochs, the “continuum” for normalization was the intensitylevel of the plateau in the underlying broad/intermediate-widthline profile. The intensity is on a linear scale. lowest EW values. We therefore regard E ( B − V )=0.016mag as an upper limit to the host galaxy reddening forSN 2017hcc, corresponding to A V . The line profiles of H α and H β are characterized by narrowemission with P Cygni absorption atop a broader emissioncomponent. At early times, the broader component has aLorentzian shape, with a FWHM value of about 2000 kms − or 1600 km s − for H α and H β , respectively (see Figs. 4and 5). At later times, however, the broader line shape be-comes less symmetric. The line shape at later epochs af-ter day ∼
100 is clearly not a single Gaussian or Lorentzian When extrapolated to an EW of 0 ˚A, the slope of the fit givesnon-zero E ( B − V ) values of 0.12-0.17 mag, so E ( B − V ) valuesfor the smallest EWs are probably overestimates in this relation.MNRAS , 1– ?? (2012)(2012)
100 is clearly not a single Gaussian or Lorentzian When extrapolated to an EW of 0 ˚A, the slope of the fit givesnon-zero E ( B − V ) values of 0.12-0.17 mag, so E ( B − V ) valuesfor the smallest EWs are probably overestimates in this relation.MNRAS , 1– ?? (2012)(2012) Smith et al.
Figure 8.
Detail of MIKE echelle spectra on the three epochsshowing the narrow component of He i λ β (orange). Approximate velocities of the early-time narrowabsorption trough of H β (at −
51 km s − ; blue) and He i (at − − ; magenta) are marked. shape, but seems to have at least two subcomponents, withan intermediate-width component at velocities below 2000km s − , and a broader component extending out to around5,000 km s − on the blue and red wings of the line. The lineasymmetry permits multiple ways to fit the line shape.Figure 6 shows two examples of how the same line mightbe approximated with two Gaussian components. Figure 6a(top) shows an example where the centroids of the Gaussiansare permitted to shift, and Figure 6b (bottom) shows Gaus-sian components that have a center fixed at zero velocity,but allowing red wings to fall below the model.In Figure 6a, even two broad components with FWHMvalues of 4000 and 1300 km s − , with line centers shifted by −
200 and −
40 km s − , respectively, are insufficient to givea satisfying fit the detailed line shape. The observed cen-tral intermediate-width component is more boxy than theGaussian, and the broad wings do not match the Gaussian Figure 9.
Same as Figure 8, but for He i λ shape well. An additional broad Gaussian would be neededat roughly +3000 km s − to account for a red emission bumpin excess of the fit in Figure 6a.In Figure 6b, the two broad components with FWHMvalues of 4000 and 1100 km s − and centers at 0 km s − do not fit the line profile either. However, these particularGaussians are chosen because the blue wing and the far redwing are matched very well. The Gaussian model clearlyexceeds the observed flux from −
500 to +3000 km s − , butthis is by design — the motivation for allowing this missingflux is that some of the intrinsic line profile may be absorbedat some velocities. The utility of this seemingly poor fit andthe “missing” flux will be apparent later.Overall, spectra at later times (after the Lorentzian pro-files transition to broader lines) seem to consistently requireat least two separate components in the broader line profileshape. (1) A broad component with FWHM widths of 4000-6000 km s − . This is broader and stronger at first (days100-300), and then narrower and fading at later times (afterday 300). The evolution of the relative strength and width of MNRAS , 1– ?? (2012) N 2017hcc Figure 10.
Comparison between the full line profile shapes ofH α (black) and H β (blue) in the three epochs of echelle spectra.The H β strengths have been scaled to match H α , to correct fortheir different line/continuum flux ratio, and so that their shapecan be compared directly. For the two later epochs, there appearsto be a slight deficit of H β flux on the red wing at intermediatevelocities. The flux is on a linear scale. the broad component with time can be seen in Figure 12. (2)An intermediate-width component with a FWHM of 1000-1500 km s − appears after the Lorentzian components fade,and persists until the latest epochs. As described later, weattribute the broad component to emission from the un-shocked SN ejecta, and the intermediate-width componentto shocked gas in the CDS. This is the typical interpretationof these features in many SNe IIn (Smith 2017). Whetherone prefers the Gaussians offset from zero or the ones that fitthe red side of the line poorly depend on the interpretationof the line profile asymmetry.In any case, the line profile asymmetry is rather mild atday 200-400, but the asymmetry gets more severe at latertimes, like in the day 762 spectrum seen in Figure 11. Inter-estingly, there appears to be a significant change from day645 to day 762. The blue wing of the line is virtually identi-cal at these two epochs, but the red wing changes drastically(examine Figures 11 and 12), with the line becoming muchmore asymmetric and suppressed on the red side. The likelyinterpretation of this change is discussed later.The broader components of He i lines behave dif-ferently from Balmer lines. At early times, He i λ − , whickis faster/broader than the Balmer lines at the same epoch(this broader width may indicate electron scattering in hot-ter gas). The broad He i emission is weak and fades dur-ing the first 100 days. Interestingly, in our spectra, He i λ − − in the day 75 spectrum (Figure 13),and probably indicates that we are beginning to directly seethe fast SN ejecta at this epoch. (A weaker absorption fea-ture can also be seen at − − , but this mightarise from different line transition.) It is not so unusual toobserve fairly strong broad blueshifted absorption from He i λ Figure 11.
Time series of the H α line profile shape in ourmoderate-resolution spectra (MMT Bluechannel and Binospec).For each date, the black line is the observed line profile, whereasthe blue line shows the blue wing of the emission line reflectedacross v=0 km s − . This is meant to show how the red wing of theline would appear if the line were symmetric. The orange spec-tra are the infrared Pa β line from our MMT/MMIRS spectra ondays 336 and 412, for comparison with the H α profile at similarepochs. The Pa β profile is expected to suffer less extinction, andits red wing roughly matches the blue reflected wing of H α . Pa β indicates that significant line flux may be absorbed around zerovelocity in optical lines. The flux is on a log scale. peak (e.g., Mauerhan et al. 2013; Smith et al. 2014). How-ever, since Balmer lines still show strong Lorentzian profilesthat indicate high electron scattering optical depths in theCSM at this same epoch, this suggests that we are able tosee the fast SN ejecta because of asymmetric geometry. Forinstance, we may be looking down on polar regions of the SNejecta, despite high continuum optical depths in the equa-torial CSM. MNRAS , 1– ?? (2012)(2012)
Time series of the H α line profile shape in ourmoderate-resolution spectra (MMT Bluechannel and Binospec).For each date, the black line is the observed line profile, whereasthe blue line shows the blue wing of the emission line reflectedacross v=0 km s − . This is meant to show how the red wing of theline would appear if the line were symmetric. The orange spec-tra are the infrared Pa β line from our MMT/MMIRS spectra ondays 336 and 412, for comparison with the H α profile at similarepochs. The Pa β profile is expected to suffer less extinction, andits red wing roughly matches the blue reflected wing of H α . Pa β indicates that significant line flux may be absorbed around zerovelocity in optical lines. The flux is on a log scale. peak (e.g., Mauerhan et al. 2013; Smith et al. 2014). How-ever, since Balmer lines still show strong Lorentzian profilesthat indicate high electron scattering optical depths in theCSM at this same epoch, this suggests that we are able tosee the fast SN ejecta because of asymmetric geometry. Forinstance, we may be looking down on polar regions of the SNejecta, despite high continuum optical depths in the equa-torial CSM. MNRAS , 1– ?? (2012)(2012) Smith et al.
Figure 12.
Same as Figure 11, but with the late-time H α pro-files plotted on top of one another, scaled arbitrarily to directlycompare line shape. Here we only show the late-time profiles afterthe symmetric Lorentzian scattering profiles no longer dominatethe line shape, instead showing the broad component from theSN ejecta and the intermediate-width component from the CDS.The flux is on a log scale. Our high-resolution MIKE echelle spectra are particularlyuseful for investigating the narrow emission and absorptionarising from the pre-shock CSM. The narrow CSM lines arewell resolved, appearing significantly broader ( ∼
50 km s − )than the instrumental resolution of .
10 km s − .The narrow components of the three Balmer lines H α ,H β , and H γ are shown in Figure 7. All three lines showqualitatively similar evolution, with a strong narrow emis-sion and weak P Cygni absorption at the first epoch on day24, transitioning to much deeper P Cygni absorption andweaker emission at the later two epochs (days 282 and 351).On day 24, the narrow P Cygni absorption from H α isweak and poorly defined, but the narrow H β and H γ ab-sorption is more clear. The centroid of the H β absorption isfound at −
51 km s − ( ± − ), indicated by the verticallight blue bar in Figure 7. Interestingly, the speed of the H γ absorption is a little slower, at roughly −
47 km s − ( ± − ), while the H α absorption (admittedly weak and harderto measure) seems to be at a faster speed of −
55 km s − ( ± − ), so there seems to be a march to the red for absorp-tion components going from H α to H γ . There is a shift inthe opposite sense for emission components, with the peakmarching slightly to the blue as we go from H α to H γ . TheFWHM of the emission components also gets narrower as weproceed up the Balmer series, with FWHM values of 50 ( ± − ), 49 ( ± − ), and 46 ( ±
1) km s − for H α , H β ,and H γ , respectively. Thus, all three trends (shifts in ab-sorption minimum, peak emission, and FWHM) trace loweroutflow velocities for higher order Balmer lines . (We note,however, that the decrease in FWHM may be partly causedby the increase in strength of the P Cygni absorption from Figure 13.
The day 24 MIKE spectrum around He i λ − , and the day 75 MMT spectrum showing the broadP Cygni absorption seen in He i λ − − on day 75 could be a fast component ofHe i , or may be from a different line. H α to H γ , where the stronger P Cygni features decrease theflux on the blue side of the emission component.)There is also a change in the centroid of the P Cygniabsorption to lower velocities at later times in Balmer lines(Figure 7), but it is not a simple shift of the absorption toslower speeds. Rather, on days 282 and 351, the blue edge ofthe P Cyg absorption stays roughly the same as in the firstepoch, but the absorption gets wider for all three Balmerlines. It therefore appears that the absorption at later timesis tracing a larger range in CSM expansion speeds alongthe line(s) of sight, including additional slowly expandinggas that was not seen in the first epoch, rather than a netshift to lower speeds in the CSM. The widening absorptionat later epochs also eats into the emission peak, pushing itfurther to the red and making it weaker for all three lines.The narrow CSM components of He i lines show someinteresting differences compared to Balmer lines. Figure 8shows narrow components of He i λ β ,while Figure 9 shows the same for He i λ i lines, but it is at a slower outflow speed than Balmer lines.Both He i lines have an absorption minimum at −
38 ( ±
1) kms − , more than 10 km s − slower than H β . The He i linestherefore seem to extend the trend of marching to slowerspeeds at higher excitation and ionization. Unlike the shiftat late times in Balmer lines, this is not a widening of theabsorption, but a clear shift of the narrow component toslower outflow speeds. Also unlike Balmer lines, He i linesdo not develop much deeper and broader absorption at latetimes, but instead, the absorption gets weaker (there seemsto be weak He i λ −
38 km s − on day 282) or the lines are not detected.There is some lingering emission in He i λ i MNRAS , 1– ?? (2012) N 2017hcc λ Above in Section 3.2 we noted a mild asymmetry in the H α profile at later epochs, requiring either a slightly blueshiftedcentroid for Gaussian fits, or a deficit of flux on the red wingcompared to symmetric profiles (Figure 6). There is also asubtle wavelength dependence and a time dependence to thisasymmetry that we discuss in more detail here.Figure 10 shows the same MIKE spectra of H α and H β that appeared earlier, but here they are plotted together.The day 24 profiles of H α and H β are nearly identical, andboth are consistent with symmetric Lorentzian profiles. Thelater spectra on days 282 and 351 show clear differences inthe line profile shapes. The blue wings of the lines matchquite well at the later two epochs; while we have admittedlyscaled the line strengths to overlap on the blue side, it isalso true that the shape of the blue wings match for H α andH β . In contrast, the red sides of the H α and H β lines donot match on days 282 and 351, with H β showing a cleardeficit of emission on the red wing of the intermediate-widthcomponent on both dates. The blueshifted asymmetry in theH α profile becomes more pronounced at H β .A consistent extension of this trend in the wavelengthdependence is seen as we move to the IR. Figure 11 includesa time series of H α line profiles in low-resolution spectra,but it also includes the 1.28 µ m Pa β line from our two lateepochs of IR spectra taken with MMIRS on the MMT. Com-paring H α and Pa β , it is evident that these profiles have verydifferent shapes. If we match the flux on the blue side of theline profile, then Pa β shows an excess in the peak of theintermediate-width component (within ±
800 km s − ) andexcess flux over most of the red wing of the line. Interest-ingly, the excess of Pa β over H α in Figure 11 is quite similarto the excess of a symmetric Gaussian model above the ob-served H α profile shown in Figure 6b. Overall, we concludethat there is a subtle but clear wavelength dependence inthe H emission lines, such that lines at shorter wavelengthshave a more pronounced blueshift.The asymmetry in emission profiles is also time depen-dent. Figure 11 also shows the blue wing of the line reflectedacross to the red side of the profile (in light blue) in orderto illustrate deviations from a symmetric profile shape. Wecan characterize the evolution in three main phases:1. Early times (up to day 100) show little asymmetry,with a narrow component atop broader symmetric Lorenz-tian profiles. Small deviations from symmetry are that days75 and 88 seem to show some excess flux on the red side(i.e. in the opposite sense of the blueshifted asymmetry atlater times). This might indicate some broad P Cyg absorp-tion that suppresses the blue wing of H α (recall that broadblueshifted He i is clearly seen at this epoch) or a slightredward shift in the centroid of the Lorentzian profile.2. Intermediate phases (roughly days 200-400 sampledby our spectra) show both a broad component and anintermediate-width component. These epochs exhibit a sig-nificant deficit of flux on the red side of the line in both theintermediate-width and broad components, although the de-viation from symmetry changes from one epoch to the next. This asymmetry is wavelength dependent, with more pro-nounced blueshift at shorter wavelengths.3. Late phases (after day 500-600 or so) are dominat-edf by an intermediate-width component, with the broadcomponent having faded substantially or become narrowerso that it is blended with the intermediate-width compo-nent. During this late phase, the intermediate-width com-ponent shows a clear blueshifted asymmetry that becomesmore pronounced with time. Note the discrepancy betweenthe observed red wing compared to a reflected blue wing,moving from day 645 to 762 (Figure 11).The profile in the day 762 spectrum has a striking asym-metry, with a peak shifted to about −
350 km s − . This pro-file cannot be fit by a symmetric Gaussian or Lorentzian thathas a shifted centroid. Rather, the peak is skewed to theblue, missing emission at low redshifted velocities, thoughthe broader wings are almost symmetric.The change from phase 2 to 3 is even more evident inFigure 12, where we overplot the H α profiles at various times(excluding the early Lorentzian phase). Here we see that thebroad component decreases in strength and/or width fromday 271 to days 340 and 382, and then the broad componentis gone by days 645, 762, and 848, while the width of theintermediate component changes very little except for theincrease in blueshifted asymmetry of the peak. In the discussion below, numbered days refer to days sincethe inferred explosion date of 2017 Oct 1 (see above). Onthis scale, for reference, the time of peak visual light wasaround day 40-45, and the peak bolometric luminosity wasaround day 30 (Prieto et al. 2017).
SN 2017hcc exhibits the classic evolution of line profileshapes that is common in strongly interacting SNe IIn,which transition from symmetric Lorentzian profiles at earlytimes (before and during peak), to irregular, broader, andasymmetric shapes at late times well after peak. This is un-derstood as a shift from narrow CSM lines broadened byelectron scattering to emission lines formed in the post-shockCDS (Smith 2017; Smith et al. 2008b; Dessart et al. 2015).The early profiles are characterized by a very narrowemission-line core (about ∼
50 km s − ) that also has anarrow P Cygni absorption component. These early pro-files have broad wings that follow a symmetric Lorentzianshape, due to incoherent electron scattering of narrow emis-sion from pre-shock gas (Chugai 2001, 2018; Smith et al.2008b, 2010a, 2012a; Smith 2017). At these times, the nar-row line width traces the pre-SN mass-loss speed, whereasthe line wings are caused by thermal broadening and are notdue to expansion speeds. This indicates that at these earlytimes (usually up to and including the time of peak luminos-ity), the continuum photosphere is in the CSM ahead of theshock, hiding emission from the CDS and SN ejecta. By day65, however, we may begin to see the fast SN ejecta directlyvia broad P Cygni absorption in He i λ MNRAS , 1– ?? (2012)(2012)
50 km s − ) that also has anarrow P Cygni absorption component. These early pro-files have broad wings that follow a symmetric Lorentzianshape, due to incoherent electron scattering of narrow emis-sion from pre-shock gas (Chugai 2001, 2018; Smith et al.2008b, 2010a, 2012a; Smith 2017). At these times, the nar-row line width traces the pre-SN mass-loss speed, whereasthe line wings are caused by thermal broadening and are notdue to expansion speeds. This indicates that at these earlytimes (usually up to and including the time of peak luminos-ity), the continuum photosphere is in the CSM ahead of theshock, hiding emission from the CDS and SN ejecta. By day65, however, we may begin to see the fast SN ejecta directlyvia broad P Cygni absorption in He i λ MNRAS , 1– ?? (2012)(2012) Smith et al.
At later times, however, emission from the fast SNejecta and CDS become visible after optical depths dropand the photosphere recedes (in mass), and as the shockovertakes the photosphere. For SN 2017hcc, this transitiontook place sometime during a gap in our spectral cover-age between days 92 and 271, missed because SN 2017hccwas behind the Sun. From day 271 onward, spectra reveala more complex line profile shape in H α with at least twobroader components (Figure 6). These two include a broademission component with FWHM = 4000 km s − that weinterpret as tracing the fast, unshocked SN ejecta, as well asan intermediate-width component with FWHM = 1100 kms − , which we interpret as emission from the post-shock gasin the CDS. Narrow emission and P Cyg absorption fromthe pre-shock CSM persist to late times as well.This transition is typically seen in SNe IIn (Smith et al.2008b; Smith 2017), and essential properties of the tran-sition are reproduced in radiative transfer simulations ofSNe IIn (Dessart et al. 2015). These simulations affirm theinterpretation of electron scattering in the CSM at earlytimes and emission from post-shock gas at later times.An interesting aspect of the H α line-profile evolutionin SN 2017hcc is the clear identification of a broad emis-sion component reaching ± − that we attributeto the fast SN ejecta (Figures 6b, 11, and 12). After day ∼ α line. As timeprogresses, the broad component declines in strength andwidth and disappears by day 752, whereas the intermediate-width component persists at all late epochs with a roughlyconstant width (ignoring affects associated with asymmetricabsorption; see below). This different time evolution con-firms that the two emission components have a differentorigin from one another. Seeing strong emission from thefreely expanding SN ejecta is rare in SLSNe IIn, wherecontinuum optical depths often hide the emission fromunderlying ejecta, or where stronger emission from theintermediate-width component dominates the lines. For ex-ample, the broader component from SN ejecta was absent ormuch weaker in day >
200 spectra of SN 2010jl (Smith et al.2012a; Gall et al. 2014; Fransson et al. 2014), and SN 2006tfshowed only weak H α emission and faint O i and He i ab-sorption at fast blueshifts (Smith et al. 2008b). Since we seethe SN ejecta more clearly in SN 2017hcc, we may be view-ing from a different orientation (looking from the poles, forexample).Identifying the broad emission with the SN ejecta alsohas important implications for interpreting the asymmetryin SN 2017hcc’s line profiles discussed below, and for itshigh observed continuum polarization. The broad emissioncomponent of H α shows only mild asymmetry (Figure 12),with a small portion of its red wing depressed compared toa symmetric profile (Figure 6). Importantly, though, the red One might infer that the “broad” width of 4,000 km s − is notso fast when compared to typical speeds of & − inthe SN ejecta of non-interacting SNe. However, recall that we areseeing these broad lines at late times after day 200, when broadlines are long gone in normal SNe II-P, and only narrow nebularlines from the inner ejecta remain. In this context, the longevityof lines with widths of 4,000 km s − in SN 2017hcc is remarkable. and blue wings of the broad component are symmetric at ve-locities faster than ± − . Moreover, the line profileof the infrared line Pa β , where any dust absorption shouldbe less influential than in the optical, is quite symmetricas well (Figure 11). This means that the intrinsic emission-line profile from the fast SN ejecta is symmetric, which hastwo critical implications: (1) High continuum optical depthsassociated with the SN photosphere are not blocking thereceding side of the explosion at late times, because thebroad lines are symmetric at the highest velocities. There-fore, something else is causing the blueshifted line asymme-try in late-time spectra. (2) The symmetric emission sug-gests that the underlying SN explosion itself was not highlyaspherical. This, in turn, would mean that faster or denserSN ejecta in a particular direction (i.e. a lopsided explosion)are not causing stronger CSM interaction in a particulardirection in SN 2017hcc, and asymmetry in the underlyingSN explosion is probably not responsible for SN 2017hcc’shigh polarization. Instead, the polarization is likely relatedto aspherical CSM, and the blueshifted line profiles are likelycaused by selective absorption, discussed later. A novel aspect of this study is that we obtained three epochsof high-resolution echelle spectra, which provides a uniqueview of the narrow-line emission from slowly expandingCSM. Only a few examples of high-resolution echelle spectrafor SNe IIn have been published, including SN 1998S on day1 (Shivvers et al. 2015), SN 1997ab a few months after ex-plosion (Salamanca et al. 1998), SN 1997eg on roughly day200 (Salamanca et al. 2002), and SN 2005gj (a Type Ia/IInhybrid) on days 86 and 374 (Trundle et al. 2008). These hadinferred progenitor wind speeds deduced from narrow linesof 40 km s − (SN 1998S), 90 km s − (SN 1997ab), 160 kms − (SN 1997eg) and 60-130 km s − (SN 2005gj).It is clear that with CSM expansion velocities of only40-50 km s − in the case of SN 2017hcc, the CSM emis-sion and absorption are completely washed out in most low-resolution optical spectra that are usually used for observingSNe (typically R = λ /∆ λ of 300-1000, or up to 300 km s − resolution for full broad wavelength range optical spectra).The narrow line profiles are underresolved even in moderate-resolution spectra like those we typically obtain using a 1200lpm grating with Bluechannel on the MMT (Andrews et al.,in prep.), or spectra obtained with X-shooter on the VLT(typically R of 4000-1000 or 30-80 km/s). Examples of thenarrow components in echelle spectra as compared to mod-erate resolution MMT ( R ∼ − . The blue edge of the absorption is around −
60 or −
70 km s − (different values for different lines).2. At later times, the blue edge velocity remains about MNRAS , 1– ?? (2012) N 2017hcc Figure 14.
Sketch illustrating how a hypothetical CSM geom-etry may explain the observed behavior of narrow emission andabsorption lines in SN 2017hcc. Panel (a) represents early timeswithin the first ∼
100 days, and panel (b) shows later times afterday 200. The same bipolar CSM configuration (blue) is shown inboth panels, where the structure is meant to symbolize latitude-dependent outflow velocity, rather than physical size and radius(although these are related if the CSM was made by an episodicmass ejection). The main difference in the two panels is the changein the emitting source of continuum and broad/intermediate-width line emission: (a) at early times, the emitting source (or-ange) is the SN photosphere, which still has a relatively small ra-dius. This light passes through a small section of the polar CSM,resulting in a narrow range of absorbed velocities (∆ v abs ). (b)at later times, the ejecta and shock have expanded, and most ofthe luminosity is generated in a more extended CSM interactionshock (dark red) and in the expanded SN ejecta (orange gradi-ent). This light passes through a much larger section of the CSM,yielding a broader range of ∆ v abs . A hypothetical Earth-basedobserver is at left. This observer sees narrow emission from allparts of the CSM shell, but only sees absorption generated alongthe line of sight to the continuum or broad-line emitting regions. the same, but the range of absorbed speeds is wider, extend-ing to slower speeds, and the absorption is deeper.3. By day 75, we begin to see the fast SN ejecta, eventhough the optical depths in the CSM remain high. Thisrequires non-spherical geometry.4. Deeper Balmer line absorption at late times reachesdown to ∼
20% of the continuum, but this “continuum” ismostly the underlying broad components of the same emis-sion line. This means that the absorption source is along ourline of sight to both the photosphere and the broad com-ponents. The continuum luminosity and intermediate-widthcomponent may arise in the same CSM interaction region. 5. The narrow emission (especially its red wing) staysroughly the same at all epochs, and has a velocity widthsimilar to the early absorption speed around 40-50 km s − .This indicates that the asymmetry in CSM velocity is fairlymild, and that the CSM speed ahead of the shock remainssimilar over a range of radii as the shock moves out throughit, regardless of the changing luminosity of the SN. This, inturn, suggests that any non-sphericity in the CSM geometryis probably axisymmetric rather than one-sided.6. The higher excitation and higher ionization lines,which are normally formed deeper in a stellar wind or closerto the shock in a SN, have slightly lower speeds (i.e. 38km s − in He i vs. 55 km s − in H α , and slightly de-creasing speeds in higher-order Balmer lines). This sametrend was also observed in narrow components of SN 2005gj(Trundle et al. 2008). We are not aware of any other SN IInwhere this has been seen in high-resolution data. Interest-ingly, the opposite was seen in spectra of SN 1998S, wherethe narrow cores of higher-ionization lines were system-atically broader than lower-ionization lines (Shivvers et al.2015). This difference could be partly due to the fact thatthe high-resolution spectrum of SN 1998S was taken within1 day of explosion, which still included emission from theinner acceleration zone of the progenitor wind.One conjecture that is immediately apparent from thislist is that a single-speed, constant velocity spherical windcannot reproduce these properties. Instead, some more com-plicated geometry or time dependence is needed. Figure 14 shows a sketch that illustrates how a hypotheticalCSM geometry might account for the observed behavior ofnarrow lines seen in SN 2017hcc. While this is admittedlynot a unique explanation, it does account for several traitsthat a spherical wind cannot. The basic idea representedin this figure is that the SN explodes inside a CSM shellthat has a bipolar configuration, borrowing from nebulae of-ten seen around LBVs like η Car (Smith 2006; Smith et al.2018b), except with much slower expansion speeds than η Car. This slow CSM nebula (tinted blue in Figure 14),viewed by an Earth-based observer at some intermediatelatitude, has higher velocities, larger radii, and a larger in-ner cavity along the poles, and it has higher densities andlower velocities in the equator. As in the case of η Car,such a structure might arise from an LBV-like eruption ina massive binary interaction or merger event (Smith et al.2018a). The emission component of the narrow lines arisesin the portions of the CSM that are not along the line-of-sight (again, the blue region in Figure 14, and this emissionretains approximately the same velocity at all times becauseit is the integrated emission from most of the CSM.The key point conveyed in Figure 14 is that the under-lying source of luminosity (tinted orange/red) is expandingwith time, causing the emitted radiation to traverse differ-ent paths (and hence different velocity ranges) through theasymmetric CSM. At early times during the SN peak (Fig-ure 14a), the emitting radius is small compared to the sizeof the nebula, and so the continuum radiation traces a pen-cil beam through only a small section of the CSM. In thiscase, it passes through only a small section of the thin polarcap, yielding weak blueshifted absorption near the maximum
MNRAS , 1– ?? (2012)(2012)
MNRAS , 1– ?? (2012)(2012) Smith et al. speed with only a narrow range of velocity (small ∆ v abs ). Atlate times (Figure 14b; past day 200 when the continuum lu-minosity has dropped), the emitting source that is absorbedbecomes more complicated. The continuum luminosity hasmostly faded, and the photons being absorbed by the CSMare now the broad component from the freely expanding SNejecta (orange gradient) as well as the intermediate-widthcomponent from the strong CSM interaction, which maybe dominated by emission from the post-shock CDS in thedense equatorial CSM (red arcs). This broad line emissionpasses through a much larger sample of the CSM and tracesa wider range of speeds (hence, larger observed ∆ v abs ). Inthis case, the blue absorption edge is still about the same,corresponding to the polar speed – but absorption now oc-curs at speeds all the way to zero, because some of the ab-sorbing CSM is moving transversely in the plane of the skyand has no Doppler shift. The absorption is also deeper atthese later times, because most of the source photons passthrough denser material and longer path lengths throughthe low-latitude portions of the CSM. Thus, this CSM con-figuration meets the first 4 requirements listed above.At some epoch intermediate between Figures 14a and14b, a transition occurs. The photosphere is outside theshock and obscures the CDS and SN ejecta at early times,but eventually the CSM interaction at the equator becomesdominant, and the optical depths through polar regions thinfirst. At this point, we may begin to see SN ejecta directlydown the poles. This may explain the broad blueshifted He i λ i , or higher orderBalmer lines like H γ , may be achieved in this configurationas follows. These lines will prefer regions of higher tempera-ture and ionization. In the configuration shown in Figure 14,these are most likely to be found closer to the source of SNluminosity and closer to the strongest shock interaction - i.e.,in the regions of the pinched waist near the equator wherethe expansion velocities are slower and directed out of ourline of sight. Lower velocities at smaller radii arise naturallyin CSM created in episodic mass loss, as opposed to a steadywind. Thus, a bipolar configuration may naturally satisfy therequirement of lower velocities for higher excitation. This isharder to imagine in a spherical CSM configuration. At latetimes, the He i lines may weaken significantly, as observed(Figures 8 and 9), for a few reasons: this may happen be-cause the inner equatorial CSM has been swept up, or sim-ply because the SN’s luminosity has dropped and the shockspeed has slowed, allowing the remaining pre-shock gas tocool and He to be mostly neutral. The sixth requirement above (slower speeds for higher exci-tation) also has important consequences regarding any accel-eration of pre-shock CSM by the radiation from the SN itself(Chugai 2019). Radiative acceleration of pre-shock CSM hassometimes been invoked to account for blueshifted line pro-files seen in some SNe IIn, like SN 2010jl (Fransson et al.2014; Zhang et al. 2012). However, this effect should bemost important for the peak-luminosity phase of the SN, whereas the blueshifted line profiles persist to late times inmost SNe IIn. Moreover, radiative transfer simulations forSN 2010jl (Dessart et al. 2015) indicate that radiative accel-eration cannot reproduce the observed blueshift.For SN 2017hcc, the observed velocity patterns sug-gest that any effect of pre-shock acceleration of the CSMis unimportant. Radiation from the shock could, in princi-ple, propagate upstream so that photon momentum fromthe tremendous SN luminosity could radiatively acceleratethe unshocked CSM, just as stellar luminosity radiativelyaccelerates winds in massive stars. If this happens in CSMwith high optical depth, as required in SNe IIn, then weexpect the strongest pre-shock acceleration in upstream re-gions that are close to the shock, tracing regions of highertemperature and ionization, and milder or minimal accel-eration at larger and cooler radii in the CSM. Radiativetransfer simulations confirm that H α arises at larger radiithan lines like H γ or He i , which come from a similar deeperregion (e.g., Shivvers et al. 2015; Dessart et al. 2015). So inthe case of radiative acceleration by a SN IIn, one expectshigher outflow speeds in He i lines or in H γ as comparedto H α , because these lines are closer to the shock. Thiseffect was indeed seen in SN 1998S (Shivvers et al. 2015).SN 2017hcc shows the opposite trend, however, with lower outflow speeds in He i and H γ . This means that the ob-served differences in speed are not tracing acceleration ofthe unshocked CSM by radiation from the SN, and that anyradiative acceleration of pre-shock CSM is smaller than the10 km s − difference in these lines. Radiative acceleration ofthe pre-shock CSM must also, therefore, play no role in theblueshifted asymmetry seen in emission lines at late times.Instead, the likely explanation for the slower outflow speedsin He i and H γ may be geometric, as noted earlier.The bipolar CSM geometry invoked in Figure 14 ishardly unusual; in fact, axisymmetric CSM tends to be thenorm rather than the exception among massive star nebu-lae (Nota et al. 1995; Smith 2014, 2017), probably due tothe pervasiveness of binary interaction in massive star evo-lution (Sana et al. 2012; Moe & Di Stefano 2017). Overall,the CSM expansion around SN 2017hcc is fairly slow, withthe bulk speed of around 40-50 km s − and a blue edge to theP Cygni absorption at only −
70 km s − . Because the absorp-tion speed seen at early times is comparable to the emissionwidth, and because the blue edge stays the same throughoutits evolution, it is likely that we view SN 2017hcc along asightline corresponding to a mid-latiutude or high latitude(say within 45 ◦ or so of the symmetry axis). This is differ-ent from SN 2010jl, where the P Cyg absorption speed isslower than the narrow emission width (Smith et al. 2011a),suggesting a view from low latitudes.Although bipolar nebulae are common around LBVs(Nota et al. 1995), most notably around η Carinae (Smith2006), the expansion speed of the CSM around SN 2017hccis relatively slow compared to typical LBV winds and neb-ular expansion speeds of &
100 km s − (Smith 2014). How-ever, because of their bipolar geometry, some LBVs exhibita wide range of outflow speeds in a single object; for ex-ample, although η Car has a wind speed of 500 km s − (Smith et al. 2003; Hillier et al. 2006) and the poles of itsnebula are expanding at 650 km s − (Smith 2006), it hasmuch slower speeds of only 40 km s − in its equatorial re-gions (Zethson et al. 1999; Smith 2006; Smith et al. 2018b). MNRAS , 1– ?? (2012) N 2017hcc There are also several well-studied blue supergiants thathave very slow (10-40 km s − ; much slower than their stel-lar winds) expansion in their resolved equatorial ring neb-ulae. These include Sher 25 (Brandner et al. 1997), SBW1(Smith et al. 2013), NaSt1 (Mauerhan et al. 2015), the mas-sive eclipsing binary RY Scuti (Smith et al. 2002, 2011b),HD 168625 (Smith 2007), and of course the progenitor ofSN 1987A (Meaburn et al. 1995; Crotts & Heathcote 2000).These slow disks are thought to arise from binary inter-action episodes, allowing outflows much slower than theirraditively driven winds. The class of B[e] supergiants arealso blue supergiants that are inferred to have slow, dense,equatorial outflows (Zickgraf 2006; Kraus 2019). While theCSM around SN 2017hcc is expanding faster than the windsof normal RSGs (typically 10-20 km s − ), there is a sub-set of extreme, high-luminosity RSGs with faster winds andhigh mass-loss rates, some of which also have asymmetricor axisymmetric structures in their CSM. One pertinentexample is VY CMa, which has mildly bipolar geometry,with the bulk outflow at 35-40 km/s, but with some fasterfeatures up to 70 km/s (Smith 2004; Smith et al. 2009a;Decin et al. 2016). Another extreme RSG with bipolar ge-ometry seen in water masers is VX Sgr (Berulis et al. 1999;Pashchenko et al. 2006).Similar bipolar/disk geometries, viewed from dif-ferent orintation angles, have been invoked for sev-eral other SNe with strong interaction, includingSN 2009ip (Mauerhan et al. 2014; Smith 2014), SN 2010jl(Andrews et al. 2011; Smith et al. 2011a; Gall et al. 2014;Fransson et al. 2014; Dessart et al. 2015), SN 2007rt(Trundle et al. 2009), PTF11iqb (Smith et al. 2015),SN 2014ab (Bilinski et al. 2020), SN 2010jp (Smith et al.2012b), iPTF14hls (Andrews & Smith 2018), SN 2013L(Andrews et al. 2017), SN 2013ej (Mauerhan et al.2017b), SN 1998S (Leonard et al. 2000), and SN 1997eg(Hoffman et al. 2008), among others. Common indicationsof bipolar geometry and slow disks around SNe IIn mighthint at a common mechansim related to pre-SN binaryinteraction (Smith & Arnett 2014). Armed with reliable estimates of the speeds of the CSMand CDS, and the observed SN luminosity, we can makeseveral rough estimates of the pre-SN mass loss. The ob-served narrow CSM lines indicate an outflow speed of 40-50 km s − , so we take an average velocity for the pre-SNmass loss of V CSM =45 km s − . The measured FWHM of theintermediate-width component of H α , emitted by the CDS,is 1100 km s − . We multiply this radial velocity by √ V CDS =1600 km s − .The luminosity generated by CSM interaction (seeSmith 2017 for a review) is given by L = 12 wV (1)where w = 4 πR ρ = ˙ M/V
CSM is the wind density parame-ter. Here, L is the total bolometric luminosity generated byCSM interaction, which includes line emission and contin-uum across all wavelengths. The observed UV/optical/IR continuum luminosity is a lower limit for this, but duringthe main peak of the SN when the CSM interaction shockis below the photosphere that resides in the optically thickCSM, the optical luminosity is a good proxy for L . At latertimes when the material becomes more optically thin andthe visual-wavelength luminosity drops, an increasing frac-tion of the luminosity may escape as line emission and X-rays. (The value of L may also change as the SN evolves andthe shock decelerates or runs into varying density CSM.)The progenitor’s mass-loss rate can be expressed as˙ M CSM = 2
L V
CSM V (2)where this is likely a conservative value if L is derivedfrom the visual-wavelength continuum luminosity, which isa lower limit to the true bolometric L . Again, the pre-SNmass-loss rate was eruptive and episodic, so ˙ M CSM may nothave been sustained for very long, and only represents thedensity of the CSM into which the SN shock expands duringthe main light curve peak.As noted in the introduction, SN 2017hcc had a peak V absolute magnitude of about − × L ⊙ or roughly 6.5 × erg s − with no bolometric correction. With such a high luminosity,it is unlikey that the underlying SN ejecta radiation con-tributes significantly to the total luminosity, so we assumethat CSM interaction dominates the emergent luminosity.This is the peak luminosity, so we adopt L ≃ × ergs − as a rough average for L over the first ∼
100 days. Thus,with the values V CSM = 45 km s − and V CDS =1600 km s − adopted above, the CSM required to power the main peakof SN 2017hcc through CSM interaction would have a winddensity parameter of W =2 × g cm − , corresponding toan average mass-loss rate of 8.8 × g s − or ∼ M ⊙ yr − . Including a bolometric correction or some efficiencyfactor for converting kinetic energy to radiation would raisethe CSM mass, so our estimate of the mass-loss rate canbe considered conservative. Of course, if the CSM is aspher-ical, then the true wind density parameter is higher thanquoted above, but occupies only a portion of the solid angleencountered by the SN ejecta.Since we do not detect significant changes in the CSMspeed as the SN evolves, this may have been a constantvelocity but very short duration wind. The time period pre-ceding explosion over which this wind was active can be es-timated as t = t obs × ( V CDS /V CSM ). The episodic wind musthave operated for about 6-12 years pre-explosion in order tocreate the CSM that powered SN 2017hcc for the first 100-200 days. Thus, the progenitor of SN 2017hcc shed at least8-16 M ⊙ in the decade before it died. This is comparableto estimated values of mass ejected in the decade or so be-fore explosion for other well-studied SLSNe like SN 2006gy,SN 2006tf, and SN 2010jl (Smith et al. 2007, 2008b, 2010b; We note that in a recent paper, Kumar et al. (2019) use a sim-ilar method to estimate a mass-loss rate for SN 2017hcc’s progen-itor, but they derive a lower value of 0.12 M ⊙ yr − . However, wenote that their quoted luminosity of 6 × erg s − is too lowby about a factor of 10 for the peak absolute magnitude of − M would be 10 times larger, in good agreement with our estimate.MNRAS , 1– ?? (2012)(2012)
100 days. Thus,with the values V CSM = 45 km s − and V CDS =1600 km s − adopted above, the CSM required to power the main peakof SN 2017hcc through CSM interaction would have a winddensity parameter of W =2 × g cm − , corresponding toan average mass-loss rate of 8.8 × g s − or ∼ M ⊙ yr − . Including a bolometric correction or some efficiencyfactor for converting kinetic energy to radiation would raisethe CSM mass, so our estimate of the mass-loss rate canbe considered conservative. Of course, if the CSM is aspher-ical, then the true wind density parameter is higher thanquoted above, but occupies only a portion of the solid angleencountered by the SN ejecta.Since we do not detect significant changes in the CSMspeed as the SN evolves, this may have been a constantvelocity but very short duration wind. The time period pre-ceding explosion over which this wind was active can be es-timated as t = t obs × ( V CDS /V CSM ). The episodic wind musthave operated for about 6-12 years pre-explosion in order tocreate the CSM that powered SN 2017hcc for the first 100-200 days. Thus, the progenitor of SN 2017hcc shed at least8-16 M ⊙ in the decade before it died. This is comparableto estimated values of mass ejected in the decade or so be-fore explosion for other well-studied SLSNe like SN 2006gy,SN 2006tf, and SN 2010jl (Smith et al. 2007, 2008b, 2010b; We note that in a recent paper, Kumar et al. (2019) use a sim-ilar method to estimate a mass-loss rate for SN 2017hcc’s progen-itor, but they derive a lower value of 0.12 M ⊙ yr − . However, wenote that their quoted luminosity of 6 × erg s − is too lowby about a factor of 10 for the peak absolute magnitude of − M would be 10 times larger, in good agreement with our estimate.MNRAS , 1– ?? (2012)(2012) Smith et al.
Fransson et al. 2014; Woosley et al. 2007). Based on thelate-time interaction that continued well after the main lu-minosity peak, the progenitor of SN 2017hcc probably shedmass at a somewhat lower rate for many decades before that.
SN 2017hcc shows the progressively increasing blueshift inits emission-line profiles that is common in SNe IIn. Therehave been four different suggestions for the potential ori-gin of the systematic blueshift seen in SNe IIn, discussedin the next four subsections. The only one that is a viableexplanation in the case of SN 2017hcc is the hypothesis ofpost-shock and/or ejecta dust formation.
Slow pre-shock CSM could be accelerated by the tremendousluminosity in a SLSN IIn, and with high optical depths in theCSM, the narrow CSM emission lines could be broadened byelectron scattering to have intermediate-width Lorentzianline profiles with a blueshifted centroid. This was proposedto explain the strongly blueshifted line profiles in SN 2010jl(Fransson et al. 2014). While the resulting line shape is asymmetric Lorentzian profile, the blueshifted centroid re-quires a large acceleration, and reqires that we cannot seethe redshifted CSM because it is blocked by the SN pho-tosphere, or that the highly asymmetric CSM is mostly onour side of the SN. There are several problems in reconcilingthis idea with observed properties of SNe IIn:(1) Since the Lorentzian wings originate as narrow-linephotons that are scattered and broadened by thermal elec-trons, the Lorentzian profile should have the same centroidas the narrow emission, but the observed narrow componentis usually not blueshifted and remains narrow even thoughthe center of the Lorentzian is blueshifted. This is the casefor both SN 2017hcc and SN 2010jl.(2) The blueshift is small or absent at early times andbecomes progressively more pronounced at later times, butthe opposite is expected from this mechanism. The strongestpre-shock acceleration should occur when the SN luminosityis highest (Dessart et al. 2015). In SN 2017hcc, the narrowlines with Lorentzian wings are symmetric and centered atzero velocity for the first ∼
200 days as the luminosity risesto peak and then falls.(3) The expected amount of radiative acceleration ismuch smaller than the observed shift (Dessart et al. 2015).Moroever, in SN 2017hcc, the constant narrow emission com-ponents and constant blue edges of the P Cyg absorptionconfirm very minimal (if any) pre-shock acceleration of theCSM (less than 10 km s − ), even though the peak of theintermediate-width component becomes blueshifted by asmuch as 300-500 km s − , similar to SN 2010jl.(4) The net blueshift of the intermediate-width com-ponent persists to very late times, but as continuum opticaldepth drops, the ability of electron scattering to create broadLorentzian wings also drops. Lines should become narrowerand symmetric as the continuum optical depth goes away,and we should see the far side of the CSM. The oppositeis observed: the blueshift becomes progressively stronger asthe continuum fades. (5) Electron scattering predicts no wavelength depen-dence, so we should observe the same blueshift in all lines(except perhaps for lines of different excitation levels, whereas noted above, we might expect higher speeds for higherexcitation closer to the shock). Observations indicate, how-ever, that the blueshift is wavelength dependent. As notedabove, the profiles of H β , H α , and Pa β in SN 2017hcc in-dicate that the lines become progressively more asymmet-ric at shorter wavelengths, inconsistent with a cause of theblueshift being the result of wavelength-indepenedent elec-tron scattering. A similar wavelength-dependence was seenin SN 2010jl (Smith et al. 2012a; Gall et al. 2014).So in summary, although radiative forces may producesome acceleration of CSM, it cannot dominate the expansionof the CSM for reasons noted here and in Section 4.2.3. Occultation by the continuum photosphere of the SN couldblock emission from ejecta or CSM interaction regions aris-ing on the redshifted side of the SN (Smith et al. 2012a).Dessart et al. (2015) presented radiative transfer simula-tions that showed this effect could produce a blueshiftedemission bump in line profiles of SLSNe IIn, once the pho-tosphere recedes from the pre-shock CSM and direct emis-sion from the post-shock CDS is revealed. As with theprevious mechanism, however, this requires high contin-uum optical depths, so this effect should be strongest atrelatively early times, and the lines should become sym-metric when the photosphere recedes and the continuumluminosity fades (Smith et al. 2012a; Dessart et al. 2015).However, in most SNe IIn exhibiting blueshifted lines, in-cluding SN 2017hcc and SN 2010jl, the opposite is seen.The blueshifted lines persist and even become increasinglyblueshifted at late times well after the continuum luminosityhas dropped by many magnitudes, as noted in the introduc-tion. In SN 2017hcc, we see the most pronounced blueshiftin the intermediate-width component after day ∼ In principle, a Type IIn event might show a blueshiftedprofile shape if there is stronger CSM interaction occur-ing on the near side of the SN, either because of a non-axisymmetric density distribution in the CSM (with higherdensities or smaller radii on our side), or because the ex-plosion was asymmetric with faster or denser SN ejectaaimed preferentially at us. Although one expects axisym-metric CSM from rotating stars and binaries, one-sided CSMmight not necessarily be so unusual. Events like mergers orgrazing collisions at periastron in eccentric binaries mightsend a spray of CSM in one preferred direction, as in theouter ejecta around η Car (Kiminki et al. 2016; Smith et al.2018a). Some SNe IIn do show signs of significant non-axisymmetric CSM and SN ejecta (Bilinski et al. 2018). Itis, of course, statistically unlikely that one-sided CSM couldlead SNe IIn to show a preference for blueshifted lines, but
MNRAS , 1– ?? (2012) N 2017hcc non-axisymmetric CSM may nevertheless be important forindividual objects. However, if the CSM is one-sided or theexplosion is lopsided, observational clues may indicate this.For the specific case of SN 2017hcc, one-sided CSMor a lopsided explosion is ruled out because the narrowand broader components of the line profiles are symmet-ric at early times (excluding blushifted P Cyg absorption ofcourse), and their continued evolution is consistent with anaxisymmetric explosion and CSM. The intermediate-widthcomponents of H β and H α are nearly symmetric during days200-400, displaying only subtle blueshifted asymmetry. Im-portantly, there is no sign of asymmetry in the Pa β profileduring this time, indicating that the intrinsic line profile issymmetric. But then at later times, the blueshifted asymme-try grows even though the blue wing of the line profile staysthe same. This behavior cannot be due to significantly one-sided CSM. Also, even though the core of the line becomesasymmetric and blueshifted, Figures 6 and 11 show that thehigh-velocity wings of H α (beyond ± − ) are sym-metric. This symmetry at high velocity indicates that thereis no significant front/back asymmetry in the fast SN ejecta,and that we can see direct emission from the far redshiftedside of the SN ejecta (this latter point is important con-sidering the location of the dust, see below). Finally, thismechanism (which depends on true asymmetry in the gasdensity) predicts no systematic wavelength dependence forthe blueshift, contrary to observations.There are certain patterns in line profile evolution thatone would expect with one-sided CSM or lopsided explo-sions. For example, if there were high-mass CSM concen-trated mostly on the near side of a SN (producing a blueshiftin emission from the post-shock CDS), then we would expectto see some corresponding and opposite asymmetry in thebroader component from the ejecta; namely, in this case thefast SN ejecta on the blue side should hit the shock soonerand the remaining unshocked ejecta should have slower ve-locities, whereas the red side of the SN ejecta could expandless impeded, allowing us to see a broader red wing in theSN ejecta. This is not seen in SN 2017hcc, although preciselythis behavior was seen in SN 2012ab (Bilinski et al. 2018).Thus, while we see good evidence for axisymmetry inthe CSM of SN 2017hcc as noted earlier, there is no evidencefor a one-sided CSM density distribution or a significantlylopsided explosion that might yield a net blueshift. The formation or regrowth of dust grains is the clearly fa-vored explanation for the blueshift of line profiles observedin SN 2017hcc because it self-consistently explains at leastfour properties that cannot be reconciled with the previ-ous three mechanisms: (1) extinction by dust within theline-emitting gas is the only mechanism consistent with theobserved wavelength dependence, where H lines at shorterwavelengths (i.e. H β even more so than H α ) show a strongerdeficit of flux on their redshifted portions than those atlonger wavelengths (Pa β ), (2) increased extinction from dustwould explain why the blue wings of the line profiles re-main constant even as their profile shapes become moreblueshifted, because dust can only absorb redshifted emis-sion in the line but does not influence the blue side of theline), (3) the gradual formation and buildup of dust as the Figure 15.
The same day 351 H α and H β profiles from Figure 10,but including the 2-Gaussian model (orange) from Figure 6b andthe day 336 Pa β profile (red histogram) from Figure 11. The plot-ted velocity range zooms-in on the regions where dust influencesthe line profile. The Gaussian model is meant to approximate thetrue intrinsic line profile shape before absorption by dust. Figure 16.
Relative extinction derived from emission line pro-files of Pa β , H α , and H β , normalized to H β , compared to ex-tinction laws from Cardelli et al. (1989) for various values of R V = A V /E ( B − V ). gas expands and cools is consistent with the fact that theobserved asymmetry increases with time as the SN fades,and (4) once dust forms in the post-shock gas, it contin-ues to cause extinction of the redshifted line emission fromreceding material; this explains why the blueshift persiststo very late times, long after the continuum luminosity hasfaded and the continuum optical depths have dropped. Thefact that line profiles begin symmetric and become progres-sively more blueshifted with time points to dust formationwithin axisymmetric material.SN 2017hcc’s asymmetry in line profiles is admittedlymore subtle than in SN 2010jl, but the wavelength depen-dence is quantifiable. Figure 15 compares the profiles of H β ,H α , and Pa β scaled to match the blue wings. On the red MNRAS , 1– ?? (2012)(2012)
Relative extinction derived from emission line pro-files of Pa β , H α , and H β , normalized to H β , compared to ex-tinction laws from Cardelli et al. (1989) for various values of R V = A V /E ( B − V ). gas expands and cools is consistent with the fact that theobserved asymmetry increases with time as the SN fades,and (4) once dust forms in the post-shock gas, it contin-ues to cause extinction of the redshifted line emission fromreceding material; this explains why the blueshift persiststo very late times, long after the continuum luminosity hasfaded and the continuum optical depths have dropped. Thefact that line profiles begin symmetric and become progres-sively more blueshifted with time points to dust formationwithin axisymmetric material.SN 2017hcc’s asymmetry in line profiles is admittedlymore subtle than in SN 2010jl, but the wavelength depen-dence is quantifiable. Figure 15 compares the profiles of H β ,H α , and Pa β scaled to match the blue wings. On the red MNRAS , 1– ?? (2012)(2012) Smith et al. side of the peak, H β is missing only about 3-4% of the fluxas compared to H α . However, the difference between Pa β and either H α or H β is more striking, with Pa β showingonly a mild blueshift (with the caveat that the Pa β spec-trum was obtained somewhat earlier than H α and H β ). Wetherefore infer that the grains forming in SN 2017hcc (atleast at 200-400 days) are probably not as large and theextinction not as gray as in SN 2010jl, where Pa β and H γ both showed similar asymmetric profiles (Gall et al. 2014).One could calculate the wavelength dependence of the ex-tinction (i.e. R = A V / E ( B − V )) from the missing flux in eachline, provided that the intrinsic line profile is known. UnlikeSN 2010jl, however, SN 2017hcc was behind the Sun whenthe emission from the CDS first appeared with a symmetricprofile uncorrupted by dust. This phase was not traced inour data, but perhaps other observations can reveal it.If instead we take the symmetric Gaussian line profilefit in Figure 6b as a proxy for the intrinsic profile, then wecan evaluate the relative extinction in each line. This sym-metric fit is also shown alongside the observed line profilesin Figure 15. The Pa β asymmetry at low velocities is mild,but the broader red wing of Pa β shows a more significantflux deficit at +1000 to +3000 km s − . This difference mayhint at larger grains that make more gray extinction in thefast SN ejecta, and smaller grains in the post-shock CDS.We integrate the flux of each line between − − , and taking the Gaussian fit as the in-trinsic flux, we calculate extinctions of A Paβ =0.066 mag, A Hα =0.148 mag, and A Hβ =0.179 mag (note again thatmost of the extinction in Pa β occurs at the higher velocities,probably associated with the SN ejecta and not the CDS).These are plotted in Figure 16, where the wavelength depen-dence of this extinction from H emission lines is comparedto a standard extinction law (Cardelli et al. 1989) with var-ious values of R V . It is clear from Figure 16 that the dustin SN 2017hcc is not similar to typical grains in the MilkyWay ISM with R V =3.1. Instead, the newly formed dust inSN 2017hcc seems to prefer larger values of R V around 6 to10 or larger. One could attempt a more complicated fit witha mix of gray dust and various R V values, as demonstratedfor SN 2010jl by Gall et al. (2014), but the quality of our IRdata is insufficient to provide meaningful constraints on thedust properties at a more detailed level. Moreover, we cau-tion that there may be different grain properties ( R V andgrain size) for the dust forming in the post-shock CDS andin the SN ejecta, based on the behavior of Pa β noted above.In any case, the data in Figure 16 and the large implied R V value are sufficient to indicate the formation of ratherlarge grains in SN 2017hcc. Large grains are common in theCSM of massive stars where dust forms rapidly in eruptiveevents, and the formation of relatively large grains has beeninferred for other SNe IIn (Gall et al. 2014; Nielsen et al.2018; Bevan et al. 2020). Deriving a mass of dust from theobserved extinction in lines and the grain size distributionis complicated, as it depends on a number of assumptionsabout grain properties and clumping better addressed in asophisticated model (e.g., Bevan et al. 2018).But where is the new dust located? Is dust located inthe pre-shock CSM, the SN ejecta, or the post-shock CDS?The answer to this question is probably: “yes”. There islikely to be dust in all three zones. Any progenitor star thathas suffered enough mass loss to make a SLSN through early CSM interaction must have experienced very strong episodicmass loss. Engulfed in a ∼ M ⊙ cocoon, any such progen-itor is likely to be surrounded by large amounts of dust.While such pre-existing dust in the CSM could give rise toan IR echo when illuminated by SN luminosity, pre-existingCSM dust cannot produce the blueshifted asymmetry. Thisrequires the formation of new dust or the re-growth behindthe shock of incompletely destroyed CSM dust.Blueshifted line profiles that become more pronouncedwith time could, in principle, arise from the formation orre-growth of grains in either the post-shock zone of the CDSor in the unshocked SN ejecta, and both may be at work inany SN IIn, with a differing relative contribution to the totaldust at different times. Detailed analysis of early to late-timeoptical and IR observations of SN 2010jl demonstrated earlydust formation in the post-shock CDS and continual graingrowth in the SN ejecta (Smith et al. 2012a; Gall et al. 2014;Sarangi et al. 2018; Chugai 2018).SN 2017hcc exhibits clear evidence of blueshift in theintermediate-width component emitted by the CDS at ve-locities of −
500 to +1000 km s − , and it also shows a fluxdeficit at faster redshifted velocites of +1000 to +3000 kms − . It is therefore likely that there is dust forming in boththe CDS and ejecta. The blueshift of the intermediate-widthcomponent is especially apparent at late times, and persistsafter the broad lines from the SN ejecta have faded away.Compare days 645 and 762, where the blueshift becomesmuch more pronounced, even though there is essentially nobroad component left in either line.Consider the hypothesis that dust forms only in thefreely expanding SN ejecta. In this case, we might ex-pect dust formation to occur relatively late, increasing withtime. Ejecta dust is likely to be centrally concentrated, asis seen in the resolved ejecta of SN 1987A (Cigan et al.2019). In SN 2017hcc, this location would correspond tothe orange-gradient colored inner ejecta in Figure 14b. Thisdust could cause a blueshift in the broad component of anyline emitted by the SN ejecta, but it could also influencethe intermediate-width component by blocking some emis-sion from redshifted post-shock gas behind the ejecta. Thjiswould remove flux mostly at the highest redshifted speeds inthe CDS. However, it would be very unlikely for dust in thecentral regions of the SN ejecta to significantlty absorb lineflux around zero velocity in the intermediate-width compo-nent, because there is no SN ejecta along the line-of-sight tothis emission (Figure 14b). In SN 2017hcc and many otherinteracting SNe with these blueshifted profiles, it is evidentthat there is significant missing flux in the intermediate-width component at low velocities, often even absorbingsome blueshifted velocities (shifting the observed peak to −
300 km s − or so). The clearest illustration of this miss-ing flux at low velocities is in the difference between H α and Pa β , where in the IR line we clearly see extra emis-sion at velocities of ±
500 km s − that is missing in H α andH β . This emission arises primarily in portions of the CDSthat are expanding near the plane of the sky, perpendicularto our line of sight, and for which their intermediate-widthemission passes through little or no SN ejecta on its way tous. This blueshift cannot be due primarily to an underly-ing blueshift in the broad component from the SN ejecta,because the blueshift remains at late times (i.e. day 762and after; Figures 11 and 12) even when the broad compo- MNRAS , 1– ?? (2012) N 2017hcc nent has faded and only the intermediate-width H α compo-nent remains. We therefore conclude that while there maybe dust forming in the SN ejecta as well, there must be somenew dust formation in the post-shock gas within the CDSof SN 2017hcc. We present a series of high-resolution echelle spectra andmoderate resolution spectra of SN 2017hcc to investigatethe evolution of emission-line profiles in this highly polarizedSLSN IIn. Here we briefly summarize the main findings:1. Narrow emission and absorption in high-resolutionspectra reveal slow CSM expanding at 40-50 km. This CSMoutflow resembles either the slow equatorial ejection in BSGbinaries (LBVs, B[e] supergiants, interacting/merging bina-ries, etc.) or the dense CSM cocoons in some extreme RSGs.2. The narrow profiles and their variation with bothtime and excitation suggest a mildly bipolar shell geometry.Lines formed in gas at higher temperatures and ionization(He i and upper Balmer series) exhibit lower sppeds, proba-bly tracing regions near the equator, closer to the SN radia-tion and to the CSM-interaction shock in the pinched waist.Faster speeds (the blue edge of the narrow P Cygni absorp-tion) are constant, and may arise from a polar shell. Wesuggest that the stark changes in narrow P Cyg absorptionstrength and width arise because of the changing size andlocation of the underlying radiation souerce, arising fromthe relatively compact SN continuum at early times and themore distended CSM interaction region at late times.3. The narrow line profiles show no evidence for sig-nificant acceleration of the pre-shock CSM with time, andthe lower velocities seen in higher excitation/ionization linesseem to contradict expectations for pre-shock acceleration.4. SN 2017hcc exhibits the typical evolution in broaderemission profiles seen in many SNe IIn, changing fromsymmetric Lorentzian profiles at early times to multi-component, asymmetric blueshifted profiles at late times.This reflects a change in the source of emission from a pho-tosphere in the pre-shock CSM at early times, when narrowCSM lines undergo strong electron scattering that producesymmetric profiles, to late times when the intermediate-width and broad components of the line profile trace thepost-shock CDS and SN ejecta, respectively.5. Despite the high polarization reported at early times(Mauerhan et al. 2017a), the line profiles during the first ∼
100 days are remarkably symmetric, and even at latertimes, the observed asymmetry in line profile shape is mildand limited to low radial velocities. This suggests that whilethe explosion and CSM geometry may be non-spherical, theyare likely to be axisymmetric.6. The time-dependence and wavelength-dependence ofthe blueshifted line profiles is most likely caused by new dustformation in the post-shock CDS and probably also the SNejecta, with different amounts as a function of time and pos-sibly different grain properties in the two regions. For vari-ous reasons discussed in detail, the blueshift cannot be dueto pre-shock acceleration of CSM, the SN continuum block-ing far side of the CDS, or one-sided CSM/ejecta. Instead,the blueshift must be due to dust formation. The degree ofblueshifted asymmetry has a wavelength dependence con- sistent with extinction from dust, with a stronger effect inlines at shorter wavelengths. The wavelength-dependent ex-tinction suggests rather large grains, consistent with a total-to-selective extinction ratio of R V =6-10 or more.7. From the observed expansion speeds of the CSM andCDS, combined with published estimates of the luminosity,we infer a high mass-loss rate of roughly 1 M ⊙ yr − in thedecade or so before explosion, with a CSM shell of around10 M ⊙ or more. This is quite similar to extreme CSM massvalues derived from a number of other SLSNe IIn. There mayhave been strong mass loss for decades or centuries beforethat as well, judging from the ongoing CSM interaction andconstant P Cyg absorption velocity from the CSM.While we have shown that the blueshifted emission-lineprofiles in SN 2017hcc must arise from the formation of rela-tively large dust grains in the CDS and SN ejecta, we expectthat additional observations (especially in the IR) can im-prove our understanding of SN 2017hcc. Namely, better IRspectra over more epochs and multi-band IR photometry,combined with modeling of the line profiles, could signifi-cantly improve detailed estimates for the grain propertiesas well as the relative amounts of dust in the pre-existingCSM, the CDS, and the SN ejecta. Reconciling the nearlysymmetric line profiles of SN 2017hcc (which imply mildasymmetry in the form of axisymmetric CSM) with its veryhigh polarization may be an interesting challenge that leadsto new insights about pre-SN mass loss in SLSNe IIn. ACKNOWLEDGEMENTS
We thank an anonymous referee for helpful suggestions. Sup-port for NS was provided by NSF award AST-1515559,and by the National Aeronautics and Space Administra-tion (NASA) through HST grant AR-14316 from the SpaceTelescope Science Institute, operated by AURA, Inc., underNASA contract NAS5-26555. Some data reported here wereobtained at the MMT Observatory, a joint facility of theUniversity of Arizona and the Smithsonian Institution. Thispaper includes data gathered with the 6.5 meter MagellanTelescopes located at Las Campanas Observatory, Chile.Facilities: MMT (Bluechannel, Redchannel, Binospec,MMIRS), Magellan (MIKE, IMACS), Bok (B&C)
DATA AVAILABILITY
The data underlying this article will be shared on reasonablerequest to the corresponding author.
REFERENCES
Aghakhanloo M, Murphy J, Smith Nm Hlozek R. 2017, MNRAS,472, 591Andrews J. E., Smith N. 2018, MNRAS, 477, 74Andrews, J. E., Gallagher, J. S., Clayton, G. C., et al. 2010, ApJ,715, 541Andrews, J. E., Clayton, G. C., Wesson, R., et al. 2011, AJ, 142,45Andrews J. E., Smith N., McCully C., Fox O.D., Valenti S., How-ell D.A. 2017, MNRAS, 471, 4047Beasor ER, Davies B, Smith N,van Loon JT, Gehrz RD, FigerDF. 2020, MNRAS, 492, 5994MNRAS , 1– ?? (2012)(2012)
Aghakhanloo M, Murphy J, Smith Nm Hlozek R. 2017, MNRAS,472, 591Andrews J. E., Smith N. 2018, MNRAS, 477, 74Andrews, J. E., Gallagher, J. S., Clayton, G. C., et al. 2010, ApJ,715, 541Andrews, J. E., Clayton, G. C., Wesson, R., et al. 2011, AJ, 142,45Andrews J. E., Smith N., McCully C., Fox O.D., Valenti S., How-ell D.A. 2017, MNRAS, 471, 4047Beasor ER, Davies B, Smith N,van Loon JT, Gehrz RD, FigerDF. 2020, MNRAS, 492, 5994MNRAS , 1– ?? (2012)(2012) Smith et al.
Bernstein R, Shectman SA, Gunnels SM, Mochnacki S, AtheyAE. 2003, SPIE, 4841, 1694Berulis II, Pashchenko MI, Rudnitskij GM. 1999, A&AT, 18, 77Bevan A. 2018, MNRAS, 480, 4659Bevan AM, Krafton K, Wesson R, Andrews JE, Montiel E,Niculescu-Duvaz M, Barlow MJ, De Looze I, Clayton GC.2020, ApJ, 894, 111Bilinski C, Smith N, Williams GG, et al. 2018, MNRAS, 475,1120Bilinski C, Smith N, Williams GG, et al. 2020, MNRAS, in press.Borish HJ, Huang C, Chevalier RA, Breslauer BM, Kingery AM,Privon GC. 2015, ApJ, 801, 7Bouret, J. C., Lanz, T., Hillier, D. J. 2005, A&A, 438, 301Brandner W., Chu Y. H., Eisenhauer F., Grebel E. K., Points S.D., 1997, ApJ, 489, L153Cardelli, J. A., Clayton, G. C., Mathis, J. S. 1989, ApJ, 345, 245Chandra P, Fransson C, Chevalier RA. 2017, ATel, 10936, 1Chatzopoulos E, Wheeler JC, Vinko J, Quimby R, et al. 2011,ApJ, 729, 143Chilingarian, I., Brown, W., Fabricant, D., et al. 2013, Astronom-ical Data Analysis Software and Systems XXII, 475, 189Chugai NN. 2001, MNRAS,326, 1448Chugai NN. 2018, MNRAS, 481, 3643Chugai NN. 2019, AStron. Lett., 45, 71Chugai NN, et al. 2004, MNRAS, 352, 1213Cigan P, Matsuura M, Gomez HL, et al. 2019, AjP, 886, 51Colgan, S. W. J., Haas, M. R., Erickson, E. F., Lord. S. D., &Hollenbach, D. J. 1994, ApJ, 427, 874Crotts A. P. S., Heathcote S. R., 2000, ApJ, 528, 426Danziger, I. J., Gouiffes, C., Bouchet, P., & Lucy, L. B. 1989, IAUCirc., 4796, 1Decin L, Richards AMS, Millar TJ, Baudry A, De Beck E, HomanW, Smith N, Van de Sande M, Walsh C. 2016, A&A, 592, A76Dessart L, Audit E, Hillier DJ. 2015, MNRAS, 449, 4304Di Carlo E, Corsi C, Arkharov AA, Massi F, Larionov VM, Efi-mova NV, Dolci M, Napoleone N, Di Paola A. 2008, ApJ, 684,471Dressler A. et al., 2011, PASP, 123, 288Elias-Rosa N, Van Dyk SD, Benetti S, et al. 2018, ApJ, 860, 68Fabricant, D., Fata, R., Epps, H., et al. 2019, PASP, 131, 075004Fox, O., Skrutskie, M. F., Chevalier, R. A., et al. 2009, ApJ, 691,650Fox, O. D., Chevalier, R. A., Skrutskie, M. F., et al. 2011, ApJ,741, 7Fox, O. D., Filippenko, A.V., Skrutskie, M. F., Silvermn, J.M.,Ganeshalingam M., Cenko S.B., Clubb, K.I. 2013, AJ, 146, 2Fransson C, Ergon M, Challis PJ, et al. 2014, ApJ, 797, 118Fuller J., 2017, MNRAS, 470, 1642Fuller J., Ro S. 2018, MNRAS, 476, 1853Fullerton, A. W., Massa, D. L., Prinja, R. K. 2006, ApJ, 637,1025Gall, C., Hjorth, J., Watson, D., et al. 2014, Natur, 511, 326 (G14)Gehrz, R. D., & Ney, E. P. 1989, Proc. Nat. Acad. Sci., 87, 4354G¨otberg Y., de Mink S. E., Groh J. H., 2017, A&A, 608, A11G¨otberg Y., de Mink S. E., Groh J. H., Kupfer, T.; Crowther, P.A.; Zapartas, E.; Renzo, M. 2018, A&A, 615, A78Hackwell J. A., Gehrz R. D., Grasdalen G. L., 1979, ApJ, 234,133Hillier, D. J., Gull, T., Nielsen, K., et al. 2006, ApJ, 642, 1098Hoffman JL, Leonard DC, Chornock R, Filippenko AV, BarthAJ, Matheson T. 2008, ApJ, 688, 1186Immler S, Modjaz M, Landsman W, et al. 2008, ApJ, 674, L85Jencson JE, Prieto JL, Kochanek CS, Shappee BJ, Stanek KZ,Pogge RW. 2016, MNRAS, 456, 2622Justham S, Podsiadlowski P, Vink JS. 2014, ApJ, 796, 121Kansky, J., Chilingarian, I., Fabricant, D., et al. 2019, PASP, 131,075005Kiminki MM, Reiter M, Smith N. 2016, MNRAS 463, 845 Kraus M. 2019, Galaxies, 7, 83Kumar B, Eswaraiah C, Singh A, et al. 2019, MNRAS, 488, 3089Leonard DC. Filippenko AV, Barth AJ, Matheson T. 2000, APJ,536, 239Lucy, L. B., Danziger, I. J., Gouiffes, G., & Bouchet, P. 1989, inStructure and Dynamics of the Interstellar Medium, ed. G.Tenorio-Tagle, et al. (Berlin: Springer-Verlag), 164McLeod, B., Fabricant, D., Nystrom, G., et al. 2012, PASP, 124,1318Maeda K, Nozawa T, Sadu D, et al. 2013, ApJ, 776, 5Mauerhan JC, Smith N, Filippenko AV, et al. 2013, MNRAS, 430,1801Mauerhan JC, Williams GG, Smith N, et al. 2014, MNRAS, 442,1166Mauerhan J. C., Smith N., Van Dyk, S.D., et al. 2015, MNRAS,450, 2551Mauerhan J. C., Filippenko A. V., Brink T. G., Zheng W., 2017a,Astron. Telegram, 10911, 1Mauerhan J. C., Van Dyk S. D., Johansson J., Hu M., Fox O. D.,Wang L., Graham M. L., Filippenko A. V., Shivvers I. 2017b,ApJ, 834, 118Meaburn J., Bryce M., Holloway A. J., 1995, A&A, 229, L1Moe M, Di Stefano R. 2017, ApJS, 230, 15Monnier J. D., Tuthill P. G., Danchi W. C., 2002, ApJ, 567, L137Nayana AJ, Chandra P. 2017, ATel, 11015, 1Nielsen ASB, Hjorth J, Gall C. 2018, A&A, 611, A67Nota A, Livio M, Clampin M, Schulte-Ladbeck R. 1995, ApJ, 448,788Pashchenko MI, Rudnitskij GM, Samodurov VA, Tolmachev AM.2006, AA&T, 25, 399Poznanski D, Prochaska X, Bloom JS. 2012, MNRAS, 426, 1465Prieto J. L. et al., 2017, Res. Notes AAS, 1, 28Puls, J., Vink, J. S., Najarro, F. 2008, A&ARv, 16, 209Renzo M, Farmer R, Justham S, G¨otberg Y, de Mink SE,Zapartas E, Marchant P, Smith N. 2020, A&A, in press,arXiv:2002.05077Rest A, Foley RJ, Gezari S, Narayan G, Draine B, Olsen K et al.2011, ApJ, 729, 88Quataert E., Shiode J., 2012, MNRAS, 423, L92Salamanca I, Cid-Fernandes R, Tenorio-Tagle G, Telles E, Ter-levich RJ,Salamanca I, Terlevich RJ, Tenorio-Tagle G, 2002, MNRAS, 330,844Sana H, de Mink SE, de Koter A, et al. 2012, Science, 337, 444Sarangi A, Dwek E, Arendt RG. 2018, ApJ, 859, 66Schlafly E. F., Finkbeiner D. P., 2011, ApJ, 737, 103Shiode J. H., Quataert E., 2014, ApJ, 780, 96Shivvers I, Groh JH, Mauerhan JC, Fox OD, Leonard DC, Filip-penko AV. 2015, ApJ, 806, 213Smith N. 2004, MNRAS, 349, L31Smith N. 2006, MNRAS, 644, 1151Smith N., 2007, AJ, 133, 1034Smith N. 2010, MNRAS, 402, 145Smith N. 2014, ARAA, 52, 487Smith N., 2017, in Alsabti A.W., Murdin P., eds, Handbook ofSupernovae. Springer International Publishing, Berlin, p. 403Smith N, Arnett D., 2014, ApJ, 785, 82Smith N, McCray R. 2007, ApJ, 671, L17Smith N, Owocki SP. 2006, ApJ, 645, L45Smith N, Tombleson R. 2015, MNRAS, 447, 602Smith N., Gehrz R. D., Stahl O., Balick B., Kaufer A., 2002, ApJ,578, 464Smith N, Davidson, Gull TR, Ishibashi K, Hillier DJ. 2003c, ApJ,586, 432Smith N, Li W, Foley RJ, Wheeler JC, Pooley D, Chornock R,Filippenko AV, Silverman JM, Quimby R, Bloom J, HansenC. 2007, ApJ, 666, 1116Smith N., Foley R. J., Filippenko A. V., 2008a, ApJ, 680, 568MNRAS , 1– ?? (2012) N 2017hcc Smith, N., Chornock, R., Li, W., et al. 2008b, ApJ, 686, 467Smith N. et al., 2009a, ApJ, 695, 1334Smith N, Hinkle KH, Ryde N. 2009b, MNRAS, 137, 3558Smith N. et al., 2010a, AJ, 139, 1451Smith N, Chornock R, Silverman JM, Filippenko AV, Foley RJ.2010b, ApJ, 709, 856Smith N, Li W, Miller A, et al. 2011a, ApJ, 732, 63Smith N., Gehrz R. D., Campbell R., Kassis M., Le Mignant D.,Kuluhiwa K., Filippenko A. V., 2011b, MNRAS, 418, 1959Smith N., Silverman J. M., Filippenko A. V., Cooper M. C.,Matheson T., Bian F., Weiner B. J., Comerford J. M., 2012a,AJ, 143, 17Smith N., Cenko SB, Butler N, et al. 2012b, MNRAS, 420, 1135Smith N., Arnett W. D., Bally J., Ginsburg A., Filippenko A. V.,2013, MNRAS, 429, 1324Smith N, Mauerhan JC, Prieto JL. 2014, MNRAS, 438, 1191Smith N, Mauerhan JC, Cenko SB, et al. 2015, MNRAS, 449,1876Smith N, Kilpatrick CD, Mauerhan JC, Andrews JE, MarguttiR, Fong WF, Graham ML, Zheng W, Kelly PL, FilippenkoAV, Fox OD. 2017, MNRAS, 466, 3021Smith N, Andrews JE, Rest A, et al., 2018, MNRAS, 480, 1466Smith N, Ginsburg A, Bally J. 2018, MNRAS, 474, 4988Stoll R, Prieto JL, Stanek KZ, Pogge RW, Szczygiel DM, Poj-manski G, Antognini J, Yan H. 2011, ApJ, 730, 34Sundqvist JO, Bjo¨orklund R, Puls J, Najarro F. 2019, A&A, 632,A126Tartaglia, L., Pastorello, A., Sollerman, J., et al. 2020, A&A, 635,A39Tonry J. L., 2011, PASP, 123, 58Trundle C, Kotak R, Vink JS, Meikle WPS. 2008, A&A, 483, L47Trundle C, Pastorello A, Benetti S, et al. 2009, A&A, 504, 945van Marle AJ, Smith N, Owocki SP, van Veelen B. 2010, MNRAS,407, 2305Williams P. M. et al., 1990, MNRAS, 243, 662Williams BJ, Fox OD. 2015, ApJ, 808, L22Wooden, D. H., Rank, D. M., Bregman, J. D., Witteborn, F. C.,Tielens, A. G. G. M., Cohen, M., Pinto, P. A., & Axelrod, T.S. 1993, ApJS, 88, 477Woosley SE, Blinnikov S, Heger A. 2007, Nature, 450, 390Woosley SE. 2017, ApJ, 836, 244Zethson, T., Johansson, S., Davidson, K., Humphreys, R. M.,Ishibashi, K., Ebbets, D. 1999, A&A, 344, 211Zhang T, et al. 2012, AJ, 144, 131Zickgraf F. J., 2006, in Kraus M., Miroshnichenko A. S., eds, ASPConf. Ser. Vol. 355, Stars With the B[e] Phenomenon. Astron.Soc. Pac., San Francisco, p. 135MNRAS , 1– ?? (2012)(2012)