Endurance of SN 2005ip after a decade: X-rays, radio, and H-alpha like SN 1988Z require long-lived pre-supernova mass loss
Nathan Smith, Charles D. Kilpatrick, Jon C. Mauerhan, Jennifer E. Andrews, Raffaella Margutti, Wen-Fai Fong, Melissa L. Graham, WeiKang Zheng, Patrick L. Kelly, Alexei V. Filippenko, Ori D. Fox
aa r X i v : . [ a s t r o - ph . H E ] D ec MNRAS , 1– ?? (2015) Preprint 8 December 2016 Compiled using MNRAS L A TEX style file v3.0
Endurance of SN 2005ip after a decade: X-rays, radio, and H α likeSN 1988Z require long-lived pre-supernova mass loss Nathan Smith ⋆ , Charles D. Kilpatrick , Jon C. Mauerhan , Jennifer E. Andrews ,Raffaella Margutti , , Wen-Fai Fong , , Melissa L. Graham , , WeiKang Zheng ,Patrick L. Kelly , Alexei V. Filippenko , Ori D. Fox Steward Observatory, University of Arizona, Tucson, AZ 85721, USA Department of Astronomy, University of California, Berkeley, CA 94720-3411, USA Center for Interdisciplinary Exploration and Research in Astrophysics (CIERA), and Department of Physics and Astrophysics, NorthwesternUniversity, Evanston, IL 60208, USA Center for Cosmology and Particle Physics, New York University, 4 Washington Place, New York, NY 10003, USA Einstein Fellow Department of Astronomy, University of Washington, Box 351580, Seattle, WA 98195-1580, USA Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA
Accepted 0000, Received 0000, in original form 0000
ABSTRACT
Supernova (SN) 2005ip was a Type IIn event notable for its sustained strong interaction withcircumstellar material (CSM), coronal emission lines, and infrared (IR) excess, interpretedas shock interaction with the very dense and clumpy wind of an extreme red supergiant. Wepresent a series of late-time spectra of SN 2005ip and a first radio detection of this SN, pluslate-time X-rays, all of which indicate that its CSM interaction is still strong a decade post-explosion. We also present and discuss new spectra of geriatric SNe with continued CSMinteraction: SN 1988Z, SN 1993J, and SN 1998S. From 3–10 yr post-explosion, SN 2005ip’sH α luminosity and other observed characteristics were nearly identical to those of the radio-luminous SN 1988Z, and much more luminous than SNe 1993J and 1998S. At 10 yr afterexplosion, SN 2005ip showed a drop in H α luminosity, followed by a quick resurgence overseveral months. We interpret this H α variability as ejecta crashing into a dense shell located . . pc from the star, which may be the same shell that caused the IR echo at earlierepochs. The extreme H α luminosities in SN 2005ip and SN 1988Z are still dominated by theforward shock at 10 yr post-explosion, whereas SN 1993J and SN 1998S are dominated by thereverse shock at a similar age. Continuous strong CSM interaction in SNe 2005ip and 1988Zis indicative of enhanced mass loss for ∼ yr before core collapse, longer than Ne, O, orSi burning phases. Instead, the episodic mass loss must extend back through C burning andperhaps even part of He burning. Key words: circumstellar matter — stars: evolution — stars: winds, outflows — supernovae:general — supernovae: individual (SN 2005ip)
Type IIn supernovae (SNe IIn) have raised important questionsabout the latest phases of evolution in massive stars, because theyrequire enhanced or episodic mass loss shortly before core collapsethat far exceeds known examples of steady winds (see Smith 2014for a general review of mass loss and its connection to interact-ing SNe). SNe IIn have relatively narrow lines of hydrogen in theirspectra (see Filippenko 1997 for a review of SN spectral classifi-cation) caused by slow-moving circumstellar material (CSM) that ⋆ Email: [email protected] is hit by the blast wave or illuminated by ultraviolet (UV) radia-tion from the SN. In some extreme cases, as much as 10–25 M ⊙ seems to have been ejected in just the decade before core collapse(Smith et al. 2007; Woosley et al. 2007; Smith et al. 2008a, 2010;Ofek et al. 2014). In other cases, the luminosity enhancement ismore modest, but the CSM is still dense enough to correspond tothe strongest known winds from extreme and unstable red super-giants (RSGs; e.g., Smith et al. 2009b). SNe IIn are about 8–9% ofall observed core-collapse SNe (Smith et al. 2011; Li et al. 2011),so the physical trigger of the most violent precursors only operatesin a subset of core-collapse events.A key issue is the timing of this mass loss. Are these erup- c (cid:13) Smith et al. tions synchronized to go off just before the SN owing to in-stabilities in the final Ne, O, and Si burning stages of theirprogenitors (Quataert & Shiode 2012; Shiode & Quataert 2014;Smith & Arnett 2014)? We suspect that pre-SN instability maybe more widespread than indicated by the observed fractionof traditional SNe IIn noted above, since these may be onlythe most extreme manifestation of a more generic phenomenon(Smith & Arnett 2014). At a less extreme level, some SNeshow Type IIn signatures for just a day or so after explosion(Gal-Yam et al. 2014; Smith et al. 2015; Khasov et al. 2016; ? ;Quimby et al. 2007; Shivvers et al. 2015), which can only be seenif they are discovered early. At the other extreme, major mass-lossepisodes also seem to precede core collapse by centuries or millen-nia. In the latter case, it could take months or years for the blastwave to reach the CSM. This may often go undetected as follow-upSN observations rarely extend past the first few months because ofsensitivity limits or observers’ compulsion to turn their attention tonewer SNe.There are some observed cases of significantly delayed onsetof CSM interaction. Most famously, SN 1987A began to collidewith its CSM ring nebula after roughly a decade (Sonneborn et al.1998; Michael et al. 1998), and in this case velocities of the ringnebula indicate that it was ejected 10–20 thousand years prior(Meaburn et al. 1995; Crotts & Heathcote 2000). This interactionwas relatively modest compared to that of SNe IIn, however, anddid not lead to a huge increase in luminosity that would have beenobservable in a distant galaxy. An interesting case is SN 2008iy,which did show a large increase in luminosity after a delay. Ini-tially it had an absolute magnitude of about − . , similar to a nor-mal SN II-P with no CSM interaction, but then brightened to about − mag after 400 days as the SN caught up to a massive shellejected a century earlier (Miller et al. 2010a). SN 2001em and morerecently SN 2014C represent cases in which a stripped-envelopeexplosion crashed into a H-rich shell after a year or more and devel-oped strong, relatively narrow H α emission (Chugai & Chevalier2006; Milisavljevic et al. 2015; Margutti et al. 2016). Infrared (IR)echoes from CSM dust illuminated by the SN itself (Wright 1980)or by ongoing interaction (see, e.g., Fox et al. 2011) also provideevidence for massive distant shells. An extreme case is SN 2006gy,which — in addition to its ∼ M ⊙ shell ejected only 8 yr be-fore core collapse (Smith et al. 2010) — also had another moredistant shell of 10–20 M ⊙ ejected ∼ yr earlier (Smith et al.2008b; Miller et al. 2010b; Fox et al. 2015). Yet another case of avery massive dust shell seen as an echo is SN 2002hh (Barlow et al.2005), although interestingly, this aging SN does not yet show signsof shock interaction with this shell even though it probably shouldhave reached it by now (Andrews et al. 2015). For SN 2005ip, theSN considered in this paper, Fox et al. (2013) interpreted the IRecho properties as indicating a past major mass-loss outburst thatwas followed by a less intense wind.We therefore have mounting evidence of major mass-lossepisodes that precede core collapse by centuries or millennia,and not just in the few years beforehand. If the precursor out-bursts of SNe IIn in the decade prior to core collapse can belinked to final nuclear burning phases (Quataert & Shiode 2012;Shiode & Quataert 2014; Smith & Arnett 2014), then what cangive rise to older mass ejections? Wave-driven mass loss dur-ing Ne and O burning can only drive significant mass loss fora few years before core collapse, according to current models(Quataert & Shiode 2012; Shiode & Quataert 2014). Additionalobservations are needed to constrain the physical parameters andtime dependence, but there is mounting evidence of enhanced mass loss on much longer timescales than Ne and O burning. Woosley(2016) notes that the pulsational pair mechanism can produce mass-loss eruptions that precede core collapse on a variety of timescales,but these come from very massive stars and should be relativelyrare (and the eruptions might not be punctuated by an energeticSN, as material falls back to a black hole). This poses an interestingobservational question: What fraction of SNe (including otherwisenormal SNe) have much more extended dense shells? For this rea-son, we have been monitoring some old, nearby SNe to look forchanges in their late-time interaction.In this paper we revisit the aging event SN 2005ip, an un-usual SN IIn located in the host galaxy NGC 2906 at a distanceof ∼
30 Mpc. Smith et al. (2009a) presented the optical photomet-ric and spectroscopic evolution over the first ∼ d, whileFox et al. (2009) discussed the IR photometric evolution over thesame time period. SN 2005ip was notable among SNe IIn for itssustained strong CSM interaction at late times, its rather unusualhigh-ionization spectrum with strong coronal emission lines anda strong blue pseudocontinuum, and its strong IR emission fromdust. SN 2005ip showed some evidence for the formation of newdust at these early times (Smith et al. 2009a; Fox et al. 2009), al-though Fox et al. (2010) concluded that the IR emission was likelyproduced by a combination of new dust formation and an IR echofrom pre-existing CSM dust. Fox et al. (2010) considered distantdust shells of various radii to explain the IR echo, and Fox et al.(2011, 2013) favoured a dust shell located ∼ . pc from the starto explain the IR echo evolution. This is similar to the radius wherewe infer late-time CSM interaction in SN 2005ip, so it is possiblethat the variability of H α emission that we report here is causedby the ejecta hitting shells that were seen as an IR echo at earliertimes. This is discussed below.Overall, the early-time data suggested a scenario wherein theSN blast wave was interacting with a very dense and clumpy RSGwind having a mass-loss rate of more than × − M ⊙ yr − andlikely around − M ⊙ yr − (Smith et al. 2009a). This points toan extreme RSG akin to VY CMa, rather than a normal RSG like α Ori (Smith et al. 2009b).Subsequent papers followed the continuing evolution ofSN 2005ip as it faded during the decade after explosion. By1800–2400 d after discovery, the H α flux had dropped by abouta factor of 10 (Stritzinger et al. 2012), although this was stillmuch stronger than typical interacting SNe at a comparableepoch. Stritzinger et al. (2012) also drew similar conclusions asSmith et al. (2009a) concerning the evidence for dust formation andthe nature of the progenitor’s mass loss, and Stritzinger et al. (2012)mentioned that the X-ray and H α fluxes seem to provide consistentresults. Fox et al. (2015) also presented a late-time (day 3024) op-tical spectrum of SN 2005ip, showing evidence for continued inter-action. Katsuda et al. (2014) studied the evolution of X-ray emis-sion from SN 2005ip during the ∼ d after discovery. Theyfound that the X-ray luminosity was roughly constant until 2009and dropped by a factor of two by 2012. They also showed that theabsorbing column density ( N H ) appeared to be steadily decliningduring the same time period. This prompted Katsuda et al. (2014)to suggest that the progenitor had ejected a massive CSM shell inthe centuries before explosion, and that after ∼ yr, the blast wavewas finally emerging from this shell, which they estimated to havea total mass of ∼ M ⊙ swept up by that time. This implied thatthe epoch of CSM interaction was coming to an end in SN 2005ip.Here we show that CSM interaction clearly has not yet fin-ished. The CSM interaction intensity, traced by H α luminosity,continued to decline very slowly until about mid 2015. After that MNRAS , 1– ?? (2015)(2015)
30 Mpc. Smith et al. (2009a) presented the optical photomet-ric and spectroscopic evolution over the first ∼ d, whileFox et al. (2009) discussed the IR photometric evolution over thesame time period. SN 2005ip was notable among SNe IIn for itssustained strong CSM interaction at late times, its rather unusualhigh-ionization spectrum with strong coronal emission lines anda strong blue pseudocontinuum, and its strong IR emission fromdust. SN 2005ip showed some evidence for the formation of newdust at these early times (Smith et al. 2009a; Fox et al. 2009), al-though Fox et al. (2010) concluded that the IR emission was likelyproduced by a combination of new dust formation and an IR echofrom pre-existing CSM dust. Fox et al. (2010) considered distantdust shells of various radii to explain the IR echo, and Fox et al.(2011, 2013) favoured a dust shell located ∼ . pc from the starto explain the IR echo evolution. This is similar to the radius wherewe infer late-time CSM interaction in SN 2005ip, so it is possiblethat the variability of H α emission that we report here is causedby the ejecta hitting shells that were seen as an IR echo at earliertimes. This is discussed below.Overall, the early-time data suggested a scenario wherein theSN blast wave was interacting with a very dense and clumpy RSGwind having a mass-loss rate of more than × − M ⊙ yr − andlikely around − M ⊙ yr − (Smith et al. 2009a). This points toan extreme RSG akin to VY CMa, rather than a normal RSG like α Ori (Smith et al. 2009b).Subsequent papers followed the continuing evolution ofSN 2005ip as it faded during the decade after explosion. By1800–2400 d after discovery, the H α flux had dropped by abouta factor of 10 (Stritzinger et al. 2012), although this was stillmuch stronger than typical interacting SNe at a comparableepoch. Stritzinger et al. (2012) also drew similar conclusions asSmith et al. (2009a) concerning the evidence for dust formation andthe nature of the progenitor’s mass loss, and Stritzinger et al. (2012)mentioned that the X-ray and H α fluxes seem to provide consistentresults. Fox et al. (2015) also presented a late-time (day 3024) op-tical spectrum of SN 2005ip, showing evidence for continued inter-action. Katsuda et al. (2014) studied the evolution of X-ray emis-sion from SN 2005ip during the ∼ d after discovery. Theyfound that the X-ray luminosity was roughly constant until 2009and dropped by a factor of two by 2012. They also showed that theabsorbing column density ( N H ) appeared to be steadily decliningduring the same time period. This prompted Katsuda et al. (2014)to suggest that the progenitor had ejected a massive CSM shell inthe centuries before explosion, and that after ∼ yr, the blast wavewas finally emerging from this shell, which they estimated to havea total mass of ∼ M ⊙ swept up by that time. This implied thatthe epoch of CSM interaction was coming to an end in SN 2005ip.Here we show that CSM interaction clearly has not yet fin-ished. The CSM interaction intensity, traced by H α luminosity,continued to decline very slowly until about mid 2015. After that MNRAS , 1– ?? (2015)(2015) ndurance of SN 2005ip Table 1.
Late-Time Optical Spectroscopy of SN 2005ip
UT Date Day a Tel./Inst. Res. F (H α ) b (y m d) λ ∆ λ − ergs − cm − c a Here and elsewhere, we note “Day” as the day from discovery. ForSN 2005ip and SN 1988Z this is an unknown amount after explosion,whereas for SN 1993J and SN 1998S the discovery is likely to have beenwithin a day or so of explosion. b Assumed uncertainty is ± %, owing mainly to the systematic uncer-tainty of the placement of the standard star within the slit aperture (whichis, however, difficult to quantify). Measurement uncertanty due to noise inthe spectra is much lower. c There were some thin clouds that came in during subsequent exposures(not included in this estimate), so we increased the uncertainty for this ob-servation when plotted in Figure 1.
Table 2.
Late-Time Optical Spectroscopy of SN 1988Z
UT Date Day Tel./Inst. Res. F (H α ) a (y m d) λ ∆ λ − ergs − cm − b a Assumed uncertainty is ± %, owing mainly to the systematic uncer-tainty of the placement of the standard star within the slit aperture (whichis, however, difficult to quantify). Measurement uncertanty due to noise inthe spectra is much lower. b We increase the uncertainty on this point because of possible calibrationissues. We normalized to the flux of [S II ] emission lines from underlyingH II regions in the same aperture at other epochs. time, however, the H α showed abrupt variations, including a bright-ening by late 2015 and early 2016. The radio and X-ray emissionwas also bright in early 2016, as described below. This indicatesthat the blast wave has continued to run into a distant CSM shell ortorus. We explore possible explanations for the origin of this CSMand implications for the pre-SN evolution of massive stars. Forcomparison, we also discuss similar continued interaction via late-time H α emission seen in newly obtained spectra of SNe 1988Z,1993J, and 1998S. We obtained late-time optical spectroscopy of SN 2005ip overseveral epochs since our previous study (Smith et al. 2009a).These observations included the Bluechannel (BC) spectrographon the 6.5 m Multiple Mirror Telescope (MMT), the Multi-ObjectDouble Spectrograph (MODS; Byard & O’Brien 2000) on theLarge Binocular Telescope (LBT), the Low-Resolution Imag-
Table 3.
Late-Time Optical Spectroscopy of SN 1993J
UT Date Day Tel./Inst. Res. F (H α ) a (y m d) λ ∆ λ − ergs − cm − a Assumed uncertainty is ± %, owing mainly to the systematic uncer-tainty of the placement of the standard star within the slit aperture (whichis, however, difficult to quantify), plus some uncertainty produced byblending with an adjacent oxygen line. Measurement uncertanty due tonoise in the spectra is lower. Table 4.
SN 2005ip on 2016 Feb. 4; VLA
Frequency Flux Density Uncertainty(GHz) (mJy) (mJy)5.0 1.18 0.047.4 0.92 0.048.5 0.86 0.0311.0 0.77 0.03 ing Spectrometer (LRIS; Oke et al. 1995) mounted on the 10 mKeck I telescope, and the Deep Imaging Multi-Object Spectrograph(DEIMOS; Faber et al. 2003) on Keck II. Details of the spectralobservations including measurements of the total H α line flux aresummarized in Table 1, and the spectra are shown in Figure 1.Standard reductions were carried out using IRAF including biassubtraction, flat-fielding, and optimal spectral extraction (see, e.g.,Kelson 2003). Flux calibration was achieved using spectrophoto-metric standards observed at similar airmass to that of each sci-ence frame, and the resulting spectra were median combined into asingle one-dimensional spectrum. For one observation noted in Ta-ble 1, the flux calibration was suspect owing to clouds that came induring subsequent exposures. For this date (day 3425) we increasethe uncertainty in Figure 6 to ± %. SN 1988Z:
In our study, we find it useful to compare the evo-lution of SN 2005ip to that of a few other well-studied, geriatricSNe with late-time interaction. SN 1988Z is a prototypical, long-lasting, and radio-luminous SN IIn that has been studied in de-tail by several authors (Filippenko 1991; Stathakis & Sadler 1991;Turatto et al. 1993; Van Dyk et al. 1993; Chugai & Danziger 1994;Fabian & Terlevich 1996; Aretxaga et al. 1999; Williams et al.2002; Schlegel & Petre 2006). In particular, Aretxaga et al. (1999)have compiled multiwavelength data, including H α fluxes, up today ∼ α line fluxes, are summarized in Table 2 and almost triple the timebaseline for this object. This series of new late-time spectra ofSN 1988Z is shown in Figure 2. These spectra have been corrected IRAF, the Image Reduction and Analysis Facility, is distributed by theNational Optical Astronomy Observatory, which is operated by the Associ-ation of Universities for Research in Astronomy (AURA), Inc., under coop-erative agreement with the National Science Foundation (NSF).MNRAS , 1– ?? (2015) Smith et al.
Figure 1.
Late-time spectra of SN 2005ip (see Table 1) normalized to the red continuum level. The spectra plotted in thin red-orange-green colours representthe appearance of the spectrum in its late-time decline. The black and light-blue spectra are when the H α flux reached its minimum in 2015, while thepurple/blue/grey colours document the resurgence of CSM interaction. See Figure 2 for line identifications. for a redshift of z = 0 . , but no reddening correction has beenapplied as in previous studies (Aretxaga et al. 1999). When we plotthe H α line luminosity in Figure 6, we first subtract a constant fluxof × − erg s − cm − , which is the mean value of the narrowH α + [N II ] line fluxes in late-time spectra (this is only significantfor the last epochs). We attribute this constant narrow emission tocontamination by coincident or nearby H II region emission, or per-haps unshocked distant CSM. Since we compare the H α luminos-ity of SN 2005ip to SN 1988Z and also to other well-studied lateinteractors, we obtained late-time spectra to extend the temporalcoverage of two additional prototypical SNe. SN 1993J:
We procured several late-time spectra of SN 1993J,most recently on 2016 Feb. 10 using LRIS on Keck I. SN 1993Jwas a Type Ib event rather than a Type IIn, but it was very nearbyin M81, was discovered within about 1 day of explosion, andhas shown strong, long-lived CSM interaction that has given riseto a high radio luminosity and prominent optical/UV emissionlines for more than a decade after explosion (Fransson et al. 1996; Matheson et al. 2000; Bietenholz et al. 2002). Table 3 summarizesour late-time spectra reported here, including a measurement of thetotal H α line flux in the calibrated spectra. Here the uncertainty isdominated by the fact that the blue side of H α is blended with astrong oxygen line, as well as uncertainty from possible slit losses.Qualitatively, the spectral appearance of SN 1993J changes verylittle at these late epochs after day 4000, and even resembles theday 2454 spectrum presented by Matheson et al. (2000). The mostrecent epoch is illustrated in the bottom panel of Figure 2 in darkgreen; we also show two earlier epochs on days 4249 and 6159 in alighter shade of green for comparison. Although some of the emis-sion lines have weakened with time, the line profiles and overallappearance of the spectrum have not changed much. SN 1998S:
This has long been considered a prototypicalSN IIn, which was discovered within about a day of explosion, butits early CSM interaction was on the weak side compared to mostSNe IIn (Leonard et al. 2000; Fassia et al. 2001). SN 1998S did,however, exhibit strong continued CSM interaction at late times
MNRAS , 1– ?? (2015)(2015)
MNRAS , 1– ?? (2015)(2015) ndurance of SN 2005ip Figure 2.
The top panel shows late-time spectra of SN 1988Z (see Table 2). Vertical dashed orange lines identify specific emission lines. We include the day3099 spectrum of SN 2005ip (blue) for comparison. The bottom panel shows some of our additional late-time spectra of SN 1993J and SN 1998S. (Pozzo et al. 2004; Mauerhan & Smith 2012). We obtained a lateepoch of SN 1998S on 2016 Mar. 1 using DEIMOS on Keck II,extending the spectral evolution by ∼ d compared to pub-lished data. The last previously published spectrum of SN 1998S(Mauerhan & Smith 2012) was obtained on day 5079, and ournew spectrum corresponds to day 6574. Qualitatively, the appear-ance of this spectrum shows little change compared to the lastone published by Mauerhan & Smith (2012). For this spectrum, however, we measure a strong H α line flux of (2 . ± × − erg s − cm − , which is several times higher than our previ-ous flux measurement (Mauerhan & Smith 2012). The uncertaintyis likely to be dominated by absolute flux calibration and slit place-ment as noted above for SN 2005ip and SN 1988Z, but this increaseappears to be significant. The new spectrum of SN 1998S is shownin the bottom panel of Figure 2. MNRAS , 1– ?? (2015) Smith et al.
Figure 3.
X-ray luminosity evolution of SN 2005ip in the 0.3–10 keV en-ergy band. Circles refer to
Swift data, and the diamonds are fluxes from
Chandra data reported by (Katsuda et al. 2014) as well as from our new
Chandra observations (see text).
We observed the position of SN 2005ip with the Karl G. Jan-sky Very Large Array (VLA; Program 16A-101, PI Kilpatrick)starting on 2016 Feb. 4 (day 3743 after discovery) at 09:47:29(UT dates are used throughout this paper), at mean frequencies of6.2 GHz and . GHz (with side-bands centred at 5.0, 7.4, 8.5,and 11.0 GHz). We obtained 14 min of on-source time for eachfrequency, and used 3C286 and J0925+0019 for bandpass/flux andgain calibration, respectively. Following standard procedures in theAstronomical Image Processing System (AIPS; Greisen 2003) fordata calibration and analysis, we detect a bright source located at α = 09 h m s , δ = +08 ◦ ′ . ′′ (J2000; δα = 0 . ′′ , δδ = 0 . ′′ ), consistent with the optical position. We measure fluxdensities and σ uncertainties for the upper and lower side-bandsusing AIPS/ JMFIT . The radio detections are listed in Table 4. Theradio spectral index of SN 2005ip is α = 0 . ± . , whichis consistent with optically thin emission observed from SNe II(Weiler et al. 2002). We postpone a more detailed analysis of theradio data to a later paper that will examine its temporal variability. Swift -XRT
The X-Ray Telescope (XRT; Burrows et al. 2005) onboard the
Swift satellite (Gehrels et al. 2004) started observing SN 2005ip on2007 Feb. 14 (475 days since discovery). Late-time observationsof SN 2005ip have been carried out until 3630.5 days, as part of a
Swift fill-in program (PI Margutti).XRT data have been analyzed using HEASOFT (v6.18) andcorresponding calibration files, following standard procedures (seeMargutti et al. 2013 for details). In particular, our rebinning schemerequires a minimum of 20 photons in the source region and an in-ferred source count rate at least σ above the background levelto yield a detection. Compared to the previous compilation byKatsuda et al. (2014), our campaign extends the X-ray monitoringof SN 2005ip from ∼ days to ∼ days since explosion,and reveals a continuation of the fading X-ray emission with time.The resulting luminosity light curve is shown in Fig. 3.For the flux calibration of Swift -XRT data acquired at t < days we employ the spectral parameters of the one-
Figure 4.
Chandra
ACIS-S energy spectrum of SN 2005ip from 2016 Apr.3. The red curve is a thermal plasma model (see text), with model residualsshown in the lower panel. component model derived by Katsuda et al. (2014) (their Table2) with a shock temperature evolution T ∝ t − . . For Swift -XRT data acquired at t > days we adopt the spectral pa-rameters constrained by our latest
Chandra spectrum (see below).Figure 3 shows the X-ray evolution of SN 2005ip in the 0.3–10 keV energy range as captured by the
Swift -XRT and
Chan-dra in the first ∼ days since explosion. The steep decline( L x ∝ t − . ) at late times suggests that the shock is sampling asteeply decaying environment density profile ρ ∝ r − s with s > (Fransson & Bj¨ornsson 1998). Chandra
ACIS
SN 2005ip was observed with the
Chandra X-ray Observatory (CXO) and Advanced CCD Imaging Spectrometer (ACIS) on 2016Apr. 3 (3812 days) under a single-epoch DDT program (PI Mauer-han, ID 18802). The ACIS-S array was used and the total on-sourceexposure time was 35.59 ks.Photometry and energy spectra were extracted using the specextract package within the HEASOFT
Ciao softwaresuite. The point source associated with SN 2005ip yielded 820counts within a circular aperture 2 ′′ in radius. A 5 ′′ radiusbackground annulus surrounding the source aperture yielded 121counts. The energy spectrum of the source is shown in Figure 4.Note the apparent detection of Fe K α emission near 6.8 keV.The source and background spectra were modeled simultane-ously using the Sherpa package. An absorbed single-temperaturethermal plasma model ( apec ) having solar abundances as definedby Asplund et al. (2009) was fit to the source for photon energiesin the range 0.5–8.0 keV, and a simple power law was used tofit to the background. The He, C, N, O, and Fe elemental frac-tions, by number relative to H, are . × − , . × − , . × − , . × − , and . × − , respectively. A C -statistic was used in the fitting. We fixed an equivalent interstellarneutral hydrogen column density of N H ( ISM ) = 3 . × cm − ,adopting this value from Katsuda et al. (2014), and allowed foran additional intrinsic source of absorption for the SN. Our bestfit yielded N H ( SN ) = 1 . +0 . − . × cm − and a plasma tem-perature of kT = 5 . +1 . − . keV. The associated uncertainties are90% confidence envelopes ( ∼ . σ ). The absorbed energy flux is MNRAS , 1– ?? (2015)(2015)
Ciao softwaresuite. The point source associated with SN 2005ip yielded 820counts within a circular aperture 2 ′′ in radius. A 5 ′′ radiusbackground annulus surrounding the source aperture yielded 121counts. The energy spectrum of the source is shown in Figure 4.Note the apparent detection of Fe K α emission near 6.8 keV.The source and background spectra were modeled simultane-ously using the Sherpa package. An absorbed single-temperaturethermal plasma model ( apec ) having solar abundances as definedby Asplund et al. (2009) was fit to the source for photon energiesin the range 0.5–8.0 keV, and a simple power law was used tofit to the background. The He, C, N, O, and Fe elemental frac-tions, by number relative to H, are . × − , . × − , . × − , . × − , and . × − , respectively. A C -statistic was used in the fitting. We fixed an equivalent interstellarneutral hydrogen column density of N H ( ISM ) = 3 . × cm − ,adopting this value from Katsuda et al. (2014), and allowed foran additional intrinsic source of absorption for the SN. Our bestfit yielded N H ( SN ) = 1 . +0 . − . × cm − and a plasma tem-perature of kT = 5 . +1 . − . keV. The associated uncertainties are90% confidence envelopes ( ∼ . σ ). The absorbed energy flux is MNRAS , 1– ?? (2015)(2015) ndurance of SN 2005ip Figure 5.
A detail of the observed H α line profile, with velocity rel-ative to the centroid of the narrow component. This shows two of thehigher-resolution spectra obtained with MMT/Bluechannel (day 2249) andKeck/DEIMOS (day 3770), before and after the dip in 2015, respectively. F abs = 2 . +0 . − . × − erg s − cm − (0.5–8.0 keV) and the un-absorbed model flux is F unabs = 2 . +0 . − . × − erg s − cm − .For our adopted distance the flux implies an X-ray luminosity of L X = 3 . × erg s − for the SN at epoch 3812 days. Wealso experimented with fitting the data using a χ statistic with aGehrels variance function, and obtained a reduced χ value of 43.6for 65 degrees of freedom. The resulting physical parameters areconsistent with those from the C statistic, quoted above. The modelfit is also shown in Figure 4. α Evolution of SN 2005ip
Figure 1 shows our newly obtained series of late-time optical spec-tra of SN 2005ip, and Figure 5 displays the H α line profile in afew epochs with the best signal-to-noise ratio. In addition to H α ,the spectrum shows consistent emission from other lines commonlyseen in late-time spectra of strongly interacting SNe IIn. Here wefocus mainly on the H α line, since it is the dominant line in thespectrum of SNe IIn observed at late times, and is a good proxy forthe strength of interaction.Our higher-resolution spectra (i.e., all but the 2014 May and2015 Nov. spectra) are able to cleanly resolve the intermediate-width emission of the post-shock gas from the narrow pre-shockCSM or underlying H II region emission (Fig. 5). In these spec-tra, the intermediate-width component of H α (within about ± km s − ) has a consistent profile shape, which is asymmetric andshifted to the blue. The peak of this emission is located roughlyat − to − km s − , shifting toward more positive veloc-ities with time. This asymmetric blueshifted profile was seen atearly times in the broad and intermediate-width lines (Smith et al.2009a), and was attributed to new dust formation in the SN ejectaor in the post-shock gas. Blueshifted lines with this type of shapecan arise simply from occultation of the far side by the SN pho-tosphere at early times (Smith et al. 2012; Fransson et al. 2014;Dessart et al. 2015), but the fact that the blueshifted asymmetry per-sists for a decade in SN 2005ip, long after the continuum opacityof the photosphere has vanished, means that the blocking of thered side of the lines must be caused by dust mixed within the SN (or to intrinsically asymmetric CSM). Much of the dust is likelyto be mixed in the post-shock region where the intermediate-widthH α originates, since even early-time spectra showed a blueshift inlines formed within the post-shock shell, as well as in the broad SNejecta lines (Smith et al. 2009a). The narrowest components fromthe pre-shock CSM are unresolved or only marginally resolvedin our spectra, indicating expansion speeds of 80 km s − or less(Fig. 5).The temporal evolution of the H α line luminosity is shownin Figure 6. This plot uses H α line fluxes in the first 1000 d fromour earlier paper (Smith et al. 2009a), and supplements these withour new spectra of SN 2005ip reported here. The line fluxes wereconverted to H α luminosity with the same assumptions as in ourearlier study. The H α luminosity remains roughly constant for thefirst 1000 d, as we noted earlier, and then declines slowly over thenext several years as found previously by Stritzinger et al. (2012).Our new spectra show that this slow decline continued to aroundday ∼ (mid 2015), when the H α luminosity was in decline,and then suddenly reversed to be on a quick rise that has continuedto the time of writing. Additional monitoring will be needed to de-termine how high it will rise, if it levels off, or fades again. Possiblecauses of this are discussed further in Section 4.Figure 6 also compares the H α luminosity of SN 2005ipto that of several other SNe with strong late-time interaction.We see here that SN 2005ip really is quite unusual, with a sus-tained H α luminosity that was essentially the same as the cur-rent record holder for such interaction: SN 1988Z (see below). Atdays 1000–2000, the H α luminosity was 2 orders of magnitudestronger than in some other well-studied objects like SN 1998S(Mauerhan & Smith 2012), SN 1980K (Milisavljevic et al. 2012),and SN 1993J (Chandra et al. 2009), and an order of magnitudestronger than in the SN IIn PTF11iqb (Smith et al. 2015). The late-time CSM interaction in these other objects can be explained bya very strong RSG wind with − M ⊙ yr − (Mauerhan & Smith2012), but SN 2005ip and SN 1988Z are clearly more extreme situ-ations. Thus, the H α luminosity implies that around 4000 days afterexplosion, the shock is interacting with a wind that had a mass-loss rate of at least − M ⊙ yr − . With a CSM speed around40 km s − , this implies that the high mass-loss rate was occurringwithin a few thousand years before core collapse. This, in turn, re-quires several M ⊙ being lost from the star in the last few millenniaof its life. Our estimate is conservative compared with some oth-ers, which favour higher rates of & . M ⊙ yr − and total CSMmasses greater than 10 M ⊙ (Katsuda et al. 2014; Stritzinger et al.2012).In our earlier paper (Smith et al. 2009a), we drew comparisonsbetween SN 2005ip and SN 1988Z, including a direct compari-son of the H α luminosity. During the first several hundred days,SN 2005ip was significantly less luminous in both H α and con-tinuum emission. However, during this time SN 2005ip remainedroughly constant in H α luminosity, while SN 1988Z steadily de-clined. Figure 6 extends this same comparison, and shows that byday 1000 the two objects had about the same luminosity, and forthe rest of its first decade, SN 2005ip was an almost exact twin ofSN 1988Z in terms of its CSM interaction strength traced by H α .As in our previous paper, Figure 6 incorporates H α luminosities forSN 1988Z published by Turatto et al. (1993), extended to almost This is a guess, but is lower than the 80 km s − resolution limit ofour data, and similar to CSM speeds for SN 1998S seen in echelle spec-tra (Shivvers et al. 2015).MNRAS , 1– ?? (2015) Smith et al.
Figure 6.
Evolution of the total H α luminosity for SN 2005ip (black diamonds) compared to observations of other representative SNe with ongoing CSMinteraction, as well as expected trends for a RSG wind profile (Chevalier & Fransson 1994) and radioactive decay (dashed line). This figure is adapted fromSmith et al. (2015), which was adapted from Mauerhan & Smith (2012). H α luminosities for SN 1998S are from Mauerhan & Smith (2012), SN 1980K fromMilisavljevic et al. (2012), SN 1993J from Chandra et al. (2009), and PTF11iqb from Smith et al. (2015). For SN 1988Z, we use H α fluxes from Turatto et al.(1993) as described in our previous paper (Smith et al. 2009a), extended to almost day 3000 by Aretxaga et al. (1999), plus our own measurements fromunpublished spectra taken at later times of SN 1988Z (see Table 2), SN 1993J, and SN 1998S. The dashed black line shows the expected rate of fading forradioactive decay luminosity from Co, for comparison. day 3000 by Aretxaga et al. (1999). We also supplement these pub-lished H α luminosities with measurements from our own spectraof SN 1988Z (see Table 2).With its recent resurgence in H α luminosity, SN 2005ip equalsSN 1988Z again at a comparable epoch, making it tied as the recordholder for the highest late-time H α luminosity at +10 yr. Integratingthrough time in Figure 6, the total energy radiated by SN 2005ip inthe H α line alone is × erg. For comparison, the total in-tegrated 0.3–10 keV X-ray energy is about a factor of 10 largerat ∼ × erg, and the integrated optical luminosity yieldsa similar total radiated energy of ∼ × erg. This is cer-tainly an underestimate of the total radiated energy, since we ap-pled no correction for flux at other wavelengths. The asymmetricblueshifted profiles of H α , in particular, suggest some dust extinc-tion in SN 2005ip (with a lack of such evidence for dust extinctionin SN 1988Z); correcting for this could raise the H α and contin-uum luminosities further, and thus make SN 2005ip significantlymore luminous than SN 1988Z. The radiated energy so far is there-fore likely to be at least erg for SN 2005ip. The kinetic energyimparted to the swept-up CSM shell (about 10 M ⊙ accelerated to ∼ km s − ) is roughly × erg. Thus, CSM interactionmay have already tapped a substantial fraction of the explosionenergy in SN 2005ip; remaining kinetic energy of the freely ex-panding ejecta that are still inside the reverse shock may powerSN 2005ip for the next decade. It will be interesting to see where itgoes from here. A significant detail is that after it rebrightened, thespectrum showed stronger emission from the intermediate-widthcomponents of [O I ] λλ Despite its redshift of 0.022, SN 1988Z has remained detectablefor nearly three decades. It was a very luminous radio SN thathas been studied extensively as it faded slowly after discovery,showing remarkable longevity in the radio, X-rays, and opticalemission lines like H α (Filippenko 1991; Stathakis & Sadler 1991;Turatto et al. 1993; Van Dyk et al. 1993; Chugai & Danziger 1994;Fabian & Terlevich 1996; Aretxaga et al. 1999; Williams et al.2002; Schlegel & Petre 2006). It has been compared to SN 1986J,and most studies of its first decade and a half favour the interpre-tation that it is powered by an energetic explosion (as much as ∼ erg; Aretxaga et al. 1999) from a massive progenitor (aninitial mass of something like 20–30 M ⊙ or more), that was inter-acting with roughly 10 M ⊙ produced by a star that had a wind witha mass-loss rate of ∼ − M ⊙ yr − for ∼ yr before core col-lapse. The rate of decline in H α and radio suggested that the mass-loss rate ramped up in the final millennium before core collapse(Van Dyk et al. 1993; Aretxaga et al. 1999; Williams et al. 2002).Even the somewhat lower mass-loss rate traced by later CSM inter-action was extreme, however, allowing SN 1988Z to remain as oneof the most radio luminous SNe even after a decade, matched onlyby SN 2005ip.Here we present a series of optical spectra, listed in Table 2 MNRAS , 1– ?? (2015)(2015)
Evolution of the total H α luminosity for SN 2005ip (black diamonds) compared to observations of other representative SNe with ongoing CSMinteraction, as well as expected trends for a RSG wind profile (Chevalier & Fransson 1994) and radioactive decay (dashed line). This figure is adapted fromSmith et al. (2015), which was adapted from Mauerhan & Smith (2012). H α luminosities for SN 1998S are from Mauerhan & Smith (2012), SN 1980K fromMilisavljevic et al. (2012), SN 1993J from Chandra et al. (2009), and PTF11iqb from Smith et al. (2015). For SN 1988Z, we use H α fluxes from Turatto et al.(1993) as described in our previous paper (Smith et al. 2009a), extended to almost day 3000 by Aretxaga et al. (1999), plus our own measurements fromunpublished spectra taken at later times of SN 1988Z (see Table 2), SN 1993J, and SN 1998S. The dashed black line shows the expected rate of fading forradioactive decay luminosity from Co, for comparison. day 3000 by Aretxaga et al. (1999). We also supplement these pub-lished H α luminosities with measurements from our own spectraof SN 1988Z (see Table 2).With its recent resurgence in H α luminosity, SN 2005ip equalsSN 1988Z again at a comparable epoch, making it tied as the recordholder for the highest late-time H α luminosity at +10 yr. Integratingthrough time in Figure 6, the total energy radiated by SN 2005ip inthe H α line alone is × erg. For comparison, the total in-tegrated 0.3–10 keV X-ray energy is about a factor of 10 largerat ∼ × erg, and the integrated optical luminosity yieldsa similar total radiated energy of ∼ × erg. This is cer-tainly an underestimate of the total radiated energy, since we ap-pled no correction for flux at other wavelengths. The asymmetricblueshifted profiles of H α , in particular, suggest some dust extinc-tion in SN 2005ip (with a lack of such evidence for dust extinctionin SN 1988Z); correcting for this could raise the H α and contin-uum luminosities further, and thus make SN 2005ip significantlymore luminous than SN 1988Z. The radiated energy so far is there-fore likely to be at least erg for SN 2005ip. The kinetic energyimparted to the swept-up CSM shell (about 10 M ⊙ accelerated to ∼ km s − ) is roughly × erg. Thus, CSM interactionmay have already tapped a substantial fraction of the explosionenergy in SN 2005ip; remaining kinetic energy of the freely ex-panding ejecta that are still inside the reverse shock may powerSN 2005ip for the next decade. It will be interesting to see where itgoes from here. A significant detail is that after it rebrightened, thespectrum showed stronger emission from the intermediate-widthcomponents of [O I ] λλ Despite its redshift of 0.022, SN 1988Z has remained detectablefor nearly three decades. It was a very luminous radio SN thathas been studied extensively as it faded slowly after discovery,showing remarkable longevity in the radio, X-rays, and opticalemission lines like H α (Filippenko 1991; Stathakis & Sadler 1991;Turatto et al. 1993; Van Dyk et al. 1993; Chugai & Danziger 1994;Fabian & Terlevich 1996; Aretxaga et al. 1999; Williams et al.2002; Schlegel & Petre 2006). It has been compared to SN 1986J,and most studies of its first decade and a half favour the interpre-tation that it is powered by an energetic explosion (as much as ∼ erg; Aretxaga et al. 1999) from a massive progenitor (aninitial mass of something like 20–30 M ⊙ or more), that was inter-acting with roughly 10 M ⊙ produced by a star that had a wind witha mass-loss rate of ∼ − M ⊙ yr − for ∼ yr before core col-lapse. The rate of decline in H α and radio suggested that the mass-loss rate ramped up in the final millennium before core collapse(Van Dyk et al. 1993; Aretxaga et al. 1999; Williams et al. 2002).Even the somewhat lower mass-loss rate traced by later CSM inter-action was extreme, however, allowing SN 1988Z to remain as oneof the most radio luminous SNe even after a decade, matched onlyby SN 2005ip.Here we present a series of optical spectra, listed in Table 2 MNRAS , 1– ?? (2015)(2015) ndurance of SN 2005ip and shown in Figure 2, which extend this late-time evolution fromthe mid-1990s to the present epoch. We find that SN 1988Z’s strongCSM interaction has continued to fade very slowly, indicating a re-markably extended and dense wind into which the blast wave con-tinues to crash. This traces mass loss over ∼ yr preceedingcore collapse. Based on the luminosity of H α and the character ofthe optical spectra, we group SN 1988Z’s late evolution into a fewdifferent epochs:(1) The first 500–3000 days . This is the phase that has al-ready been studied extensively as noted above. During this time,SN 1988Z showed a steady and very slow decline in radio andH α luminosity, although remaining far more luminous than mostSNe having signs of CSM interaction. Williams et al. (2002) noteda possible break in the decline rate of the H α luminosity around day1000, although this break is not clear in Figure 6 where the declinelooks rather continuous within the uncertainties. Of interest in thepresent paper is that throughout the time period of roughly 1000–3000 days, SN 2005ip was a nearly identical twin of SN 1988Z interms of its decline rate, its luminosity in the radio (admittedly onlya single epoch so far for SN 2005ip), H α , and X-rays, and in the ap-pearance of its spectrum. In our previous paper (Smith et al. 2009a)we noted similarities in the spectra of SN 2005ip and SN 1988Z atearly times, although SN 2005ip exhibited stronger coronal lines.From day 1000 to 3000, however, their spectra are almost indistin-guishable. Figure 2 includes a recent spectrum of SN 2005ip on day3098 for comparison, which appears very similar to the days 2006and 2542 spectra of SN 1988Z in the same plot. This phase is char-acterized by very strong intermediate-width H α , a steep H α /H β decrement of > , several intermediate-width and narrow coronallines such as [Fe VII ], and intermediate-width lines that indicatevery high electron densities in the post-shock gas, such as [N II ] λ VII ] line).(2)
A “plateau” over days 3000–7000 . During this time pe-riod, the H α luminosity appears to level-off for a decade. While H α remains constant, the rest of the spectrum shows some interestingchanges. The coronal lines and other indicators of high density fadeaway, leaving mostly narrow H II region lines and broad, weak oxy-gen lines (possibly from the reverse shock; see below) that are alsoseen in young SN remnants. The intermediate-width H β line alsofades, yielding an H α /H β ratio that is higher than before ( > ).(3) After day 7000.
Eventually, other lines fade away too, leav-ing only intermediate-width H α , which also resumed a decline influx at these very late times. Even the broad/intermediate-widthoxygen lines mostly disappear. Note that in Figure 2, we have sub-tracted a constant value for the contribution of narrow H α +[N II ]emission from underlying H II regions in the spectrum, which typ-ically contributes about × − erg s − cm − in our Keck slitaperture after subtraction of adjacent background. This suggeststhat SN 1988Z resides in or near a complex of H II regions, pro-viding another possible indication of a high initial mass for theprogenitor. At SN 1988Z’s distance of ∼ Mpc, however, ′′ corresponds to almost 500 pc, which could include a large complexof H II regions like the Tarantula nebula, and is not a very reliableor precise indicator of initial mass. It is, however, a good indicatorof a relatively strong ambient UV radiation field.This last point is interesting in the context of something wediscuss later. Namely, Mackey et al. (2014) have suggested that ex-ternal ionization of an otherwise normal RSG wind can produce astalled dense shell, which might offer an alternative to very strong or eruptive mass loss as an origin for SNe IIn. Despite its prob-able location in an H II region, this idea does not seem to applywell to SN 1988Z, which from its earliest phases shows an uninter-rupted decline in CSM interaction strength, consistent with a strongshock moving through a freely expanding wind. The total massof 15–20 M ⊙ or more of CSM that has been swept up so far bySN 1988Z rules out a normal RSG wind that has been confined to athin shell, because such shells only contain a fraction (about a third)of the total RSG mass loss, and are therefore expected to have muchlower total mass (Mackey et al. 2014). With such a strong wind,the expected location of a stalled shell would probably be differentfrom the models presented by Mackey et al. (2014), however. Wereturn to this issue later, because despite its similarity to SN 1988Z,SN 2005ip does show a rapid interruption of the smooth decline inits CSM interaction strength, which might be indicative of externalinfluence of this sort. This may hint that two otherwise very similarevents may have their CSM modified in different ways. Then again,the relatively coarse sampling of SN 1988Z in its past decade doesnot rule out the possibility of some brief drops or spikes in luminos-ity indicative of density fluctuations in its CSM. Indeed, one mayspeculate that the late plateau phase in SN 1988Z’s H α luminositymight represent the shock traversing a relatively constant density,ionized portion of the wind outside such a neutral dense shell. Forthis reason as well, it will be interesting to see how SN 2005ip be-haves in the coming decade as compared to SN 1988Z. All the SNe discussed in this paper show signs of long-lived, strongCSM interaction, but there are interesting differences. ExaminingFigures 1 and 2, it is clear that both SN 2005ip and SN 1988Z havelate-time spectra dominated by their intermediate-width compo-nents of H α and other lines, with line widths of about 2000 km s − .Both SN 1993J and SN 1998S have significantly broader line pro-files of 5000–10,000 km s − . This difference is probably closelylinked to their total H α luminosity. Both SN 2005ip and SN 1988Zare ∼ orders of magnitude more luminous than the other two at acomparable late epoch. The combination of higher H α luminosity(as well as higher radio and X-ray luminosities), high-density trac-ers, and narrower lines likely stems from the same cause: denserCSM and a radiative forward shock . If a blast wave is expandinginto extremely dense and extended CSM, the forward shock can re-main radiative, giving higher CSM interaction luminosity and caus-ing the intermediate-width components to continue to dominate thespectrum as the blast wave continues to expand into the CSM andto decelerate slowly. (Strong global asymmetry or a high degree ofclumping can provide very high CSM densities without necessarilyhaving a huge total CSM mass, so an intermediate-width compo-nent might also be present at lower luminosity in some cases.)At lower average CSM densities and luminosities, as inSN 1993J and SN 1998S, radiation from the reverse shock dom-inates the late-time emission, and so the spectrum exhibits muchbroader lines from oxygen-rich SN ejecta crossing the reverse Note that this scenario cannot explain the CSM interaction observed atearly times in normal SNe IIn, because their narrow emission line com-ponents have resolved widths and P-Cygni absorption indicating that theCSM near the star is outflowing — i.e., the CSM shells are not stalled andare not ambient material (see Smith 2014). Stalled shells of this sort, how-ever, could potentially influence the shock interaction observed at very latetimes.MNRAS , 1– ?? (2015) Smith et al.
Figure 7.
Late-time (1500–6000 days) 5 GHz luminosity observed towardSNe 1986J, 1988Z, 1995N, and 2005ip ( d = 10 , 100, 24, 33 Mpc, respec-tively; Williams et al. 2002; Bietenholz et al. 2002; Chandra et al. 2009;Smith et al. 2009a). We have assumed that the radio emission from eachevent is emitted isotropically. The dotted curve shows the best-fit model tolate-time ( > day) emission from SN 1988Z. shock. In our late spectra of SN 1993J, the flat-topped and double-peaked profiles described by Matheson et al. (2000) still persist.These may be the result of interaction with a disk-like CSM, whichpinches the waist of the CSM interaction region and makes the re-verse shock brighter at the equator. One may envision a reverse-shock geometry similar to that of SN 1987A (France et al. 2015).Interestingly, we may see a transitional phase in SN 1988Z,where weak and broad oxygen lines remain after the intermediate-width coronal lines have faded (days 5138 and 6640 in Figure 2). Ingeneral, though, the lack of strong oxygen lines in most SNe IIn ata few hundred days after explosion is most likely a consequence ofthe fact that the forward shock still dominates the emission spec-trum. Since this is tracing shock-heated CSM material, it reflectsthe composition of the CSM and not the composition of the innerejecta. Thus, a lack of strong oxygen features should not be takenas any indication of the abundances in the SN ejecta as long asintermediate-width lines are still seen.Another relevant detail concerns the line profiles. WhileSN 2005ip and SN 1988Z are otherwise very similar, they differin that SN 2005ip shows asymmetric blueshifted line profiles, es-pecially in its intermediate-width H α line at late times (Figure 1).The intermediate-width H α components of SN 1988Z remain muchmore symmetric at all epochs (Figure 2). This could signify a dif-ference in dust-formation efficiency in the two SNe, or perhapsviewing-angle effects if the CSM interaction is not spherically sym-metric (for example, SN 1988Z could be a ring or disk-like in-teraction seen pole-on, in which case dust cannot block emissionfrom receding material). Interestingly, the late-time line profiles ofSN 1998S, and to a somewhat lesser extent SN 1993J, do showblueshifted asymmetry that may be indicative of dust formationin these SNe. This was already discussed in detail for the caseof SN 1998S by Mauerhan & Smith (2012). There have been nopublished studies of early-time spectropolarimetry of SN 1988Z orSN 2005ip, however. Very few radio SNe have been observed beyond 10 yr. Previ-ous examples such as SNe 1986J, 1988Z, and 1995N (see, e.g.,Van Dyk et al. 1993; Williams et al. 2002; Bietenholz et al. 2002;Chandra et al. 2009) were noted for having unusually luminous ra-dio emission that peaked around 900 to 1100 days after the inferredexplosion date. In Figure 7, we compare the 5 GHz radio luminos-ity observed toward SN 2005ip to these radio SNe observed at latetimes (1500–6000 days). SN 2005ip appears very similar in its ra-dio luminosity to SN 1988Z at a comparable epoch, which is per-haps not surprising in light of the H α results above. The best fitto the observed radio light curve of SN 1988Z is based on a five-parameter model derived from Weiler et al. (1986): S ν = K (cid:16) ν GHz (cid:17) α (cid:18) t − t day (cid:19) β (cid:18) − e τ τ (cid:19) , (1) τ = K (cid:16) ν GHz (cid:17) − . (cid:18) t − t day (cid:19) δ , (2)with α, β, δ, K , and K as free parameters. The explosion date t can be fixed from early-time optical photometry. In Williams et al.(2002), the fit parameters for the late-time ( > day) light curveof SN 1988Z were α = − . , β = − . , δ = − . , K =9 . × , and K = 3 . × . These parameters imply that theSN progenitor underwent mass loss of at least . × − M ⊙ yr − for the last 10,000 yr before exploding. SN 2005ip is about 20%less luminous than SN 1988Z at 5 GHz and t ≈ days, butit is a factor of a few more luminous in X-rays at the same epoch.Analysis of X-rays suggests significantly larger mass-loss rates of − to − M ⊙ yr − (Katsuda et al. 2014).It is difficult to infer the physical parameters of the CSMsurrounding SN 2005ip based on a single epoch of radio obser-vations. However, given the similarity between SNe 2005ip and1988Z at both optical and radio wavelengths, we can hypothesizethat SN 2005ip has an extremely dense and clumpy CSM profileout to large radii. We plan to continue to monitor the radio and X-ray evolution of SN 2005ip in light of its recent H α brightening,and the results will be presented in a future paper. As noted in Section 2.3, the rate of fading of the X-ray luminos-ity of SN 2005ip implies a CSM density that falls off more steeplythan for a steady wind with ρ ∝ r − s and s = 2 , which shouldyield a t − profile in the case of free-free emission from a wind.This would imply that the star’s mass-loss rate was ramping up inthe last few thousand years of its life as it approached core col-lapse. This is qualitatively consistent with implications from theH α evolution noted above. The wind density being steeper than asteady wind is also consistent with the earlier analysis of X-rays byKatsuda et al. (2014). Looking at X-rays alone, the rate of fadingappears to be consistent with a continuation of the trend inferredby Katsuda et al. (2014). This is interesting, since the H α luminos-ity shows a more sudden drop followed by a resurgence in recentdata. It is possible that the coarse time sampling of the X-ray datahas missed this peculiar variability, so future additional X-ray ob-servations are required to resolve this.Although the X-rays from SN 2005ip imply a CSM that fallsoff more steeply than a steady wind, it is interesting to note thatthe X-ray luminosity of SN 1988Z falls off even faster with timeat a similar epoch after explosion than SN 2005ip (e.g., Figure 8 MNRAS , 1– ?? (2015)(2015)
Late-time (1500–6000 days) 5 GHz luminosity observed towardSNe 1986J, 1988Z, 1995N, and 2005ip ( d = 10 , 100, 24, 33 Mpc, respec-tively; Williams et al. 2002; Bietenholz et al. 2002; Chandra et al. 2009;Smith et al. 2009a). We have assumed that the radio emission from eachevent is emitted isotropically. The dotted curve shows the best-fit model tolate-time ( > day) emission from SN 1988Z. shock. In our late spectra of SN 1993J, the flat-topped and double-peaked profiles described by Matheson et al. (2000) still persist.These may be the result of interaction with a disk-like CSM, whichpinches the waist of the CSM interaction region and makes the re-verse shock brighter at the equator. One may envision a reverse-shock geometry similar to that of SN 1987A (France et al. 2015).Interestingly, we may see a transitional phase in SN 1988Z,where weak and broad oxygen lines remain after the intermediate-width coronal lines have faded (days 5138 and 6640 in Figure 2). Ingeneral, though, the lack of strong oxygen lines in most SNe IIn ata few hundred days after explosion is most likely a consequence ofthe fact that the forward shock still dominates the emission spec-trum. Since this is tracing shock-heated CSM material, it reflectsthe composition of the CSM and not the composition of the innerejecta. Thus, a lack of strong oxygen features should not be takenas any indication of the abundances in the SN ejecta as long asintermediate-width lines are still seen.Another relevant detail concerns the line profiles. WhileSN 2005ip and SN 1988Z are otherwise very similar, they differin that SN 2005ip shows asymmetric blueshifted line profiles, es-pecially in its intermediate-width H α line at late times (Figure 1).The intermediate-width H α components of SN 1988Z remain muchmore symmetric at all epochs (Figure 2). This could signify a dif-ference in dust-formation efficiency in the two SNe, or perhapsviewing-angle effects if the CSM interaction is not spherically sym-metric (for example, SN 1988Z could be a ring or disk-like in-teraction seen pole-on, in which case dust cannot block emissionfrom receding material). Interestingly, the late-time line profiles ofSN 1998S, and to a somewhat lesser extent SN 1993J, do showblueshifted asymmetry that may be indicative of dust formationin these SNe. This was already discussed in detail for the caseof SN 1998S by Mauerhan & Smith (2012). There have been nopublished studies of early-time spectropolarimetry of SN 1988Z orSN 2005ip, however. Very few radio SNe have been observed beyond 10 yr. Previ-ous examples such as SNe 1986J, 1988Z, and 1995N (see, e.g.,Van Dyk et al. 1993; Williams et al. 2002; Bietenholz et al. 2002;Chandra et al. 2009) were noted for having unusually luminous ra-dio emission that peaked around 900 to 1100 days after the inferredexplosion date. In Figure 7, we compare the 5 GHz radio luminos-ity observed toward SN 2005ip to these radio SNe observed at latetimes (1500–6000 days). SN 2005ip appears very similar in its ra-dio luminosity to SN 1988Z at a comparable epoch, which is per-haps not surprising in light of the H α results above. The best fitto the observed radio light curve of SN 1988Z is based on a five-parameter model derived from Weiler et al. (1986): S ν = K (cid:16) ν GHz (cid:17) α (cid:18) t − t day (cid:19) β (cid:18) − e τ τ (cid:19) , (1) τ = K (cid:16) ν GHz (cid:17) − . (cid:18) t − t day (cid:19) δ , (2)with α, β, δ, K , and K as free parameters. The explosion date t can be fixed from early-time optical photometry. In Williams et al.(2002), the fit parameters for the late-time ( > day) light curveof SN 1988Z were α = − . , β = − . , δ = − . , K =9 . × , and K = 3 . × . These parameters imply that theSN progenitor underwent mass loss of at least . × − M ⊙ yr − for the last 10,000 yr before exploding. SN 2005ip is about 20%less luminous than SN 1988Z at 5 GHz and t ≈ days, butit is a factor of a few more luminous in X-rays at the same epoch.Analysis of X-rays suggests significantly larger mass-loss rates of − to − M ⊙ yr − (Katsuda et al. 2014).It is difficult to infer the physical parameters of the CSMsurrounding SN 2005ip based on a single epoch of radio obser-vations. However, given the similarity between SNe 2005ip and1988Z at both optical and radio wavelengths, we can hypothesizethat SN 2005ip has an extremely dense and clumpy CSM profileout to large radii. We plan to continue to monitor the radio and X-ray evolution of SN 2005ip in light of its recent H α brightening,and the results will be presented in a future paper. As noted in Section 2.3, the rate of fading of the X-ray luminos-ity of SN 2005ip implies a CSM density that falls off more steeplythan for a steady wind with ρ ∝ r − s and s = 2 , which shouldyield a t − profile in the case of free-free emission from a wind.This would imply that the star’s mass-loss rate was ramping up inthe last few thousand years of its life as it approached core col-lapse. This is qualitatively consistent with implications from theH α evolution noted above. The wind density being steeper than asteady wind is also consistent with the earlier analysis of X-rays byKatsuda et al. (2014). Looking at X-rays alone, the rate of fadingappears to be consistent with a continuation of the trend inferredby Katsuda et al. (2014). This is interesting, since the H α luminos-ity shows a more sudden drop followed by a resurgence in recentdata. It is possible that the coarse time sampling of the X-ray datahas missed this peculiar variability, so future additional X-ray ob-servations are required to resolve this.Although the X-rays from SN 2005ip imply a CSM that fallsoff more steeply than a steady wind, it is interesting to note thatthe X-ray luminosity of SN 1988Z falls off even faster with timeat a similar epoch after explosion than SN 2005ip (e.g., Figure 8 MNRAS , 1– ?? (2015)(2015) ndurance of SN 2005ip Figure 8.
X-ray evolution of SN 2005ip (this paper) and SN 1988Z(Schlegel & Petre 2006). The X-ray luminosity plotted for SN 1988Z is alower limit to the true luminosity of the transient, as no correction for any in-trinsic neutral hydrogen column density was attempted by Schlegel & Petre(2006). For SN 1988Z Schlegel & Petre (2006) assumed a 5 keV plasmain collisional equilibrium. For comparison with this X-ray light curve ofSN 1988Z, here we show the X-ray fluxes for SN 2005ip reduced in thesame way (not corrected for intrinsic absorption, integrated over the sameenergy range of 0.5–2 keV). This is different from the data shown in Fig-ure 3, but allows for a more direct comparison with SN 1988Z. The twoSNe overlap in X-ray luminosity in this energy band, but have somewhatdifferent deecay rates. and Schlegel & Petre 2006). Note that a direct comparison betweenX-rays from SN 2005ip and SN 1988Z is a bit tricky, becauseSchlegel & Petre (2006) made somewhat different assumptions intheir analysis. Therefore, for the purpose of comparing the twoobjects in Fig 8, we have recalibrated our SN 2005ip X-ray datawith the same assumptions (integrated over 0.5–2 keV, uncorrectedfor intrinsic absorption). Here we find that the X-ray luminositiesare about the same at ∼ d, although the somewhat faster de-cline of SN 2005ip is still apparent. Yet, despite this faster declinein X-rays around a decade after explosion, the H α luminosity ofSN 1988Z then hit a plateau and did not decline for another decadeafter this epoch. This constant H α luminosity seems inconsistentwith the steep drop in wind density implied by the X-rays. Over-all, it seems that a clear lesson from comparing various diagnosticsfrom different late-time interactors is that any single diagnostic byitself provides an incomplete picture, so readers should be awarethat some of the corresponding mass-loss rate estimates are likelyto be lower limits. Note that the value of ˙ M that we derive fromradio emission is an order of magnitude lower than from X-ray andH α observations of SN 2005ip at the same epoch.Consider the comparison of SN 1988Z and SN 2005ip adecade after explosion. The two have roughly equal X-ray andH α luminosities, but SN 1988Z is more luminous in radio emis-sion. A caveat is that the quoted X-ray luminosities are not bolo-metric X-ray luminosities, although we have tried to recalibrateour SN 2005ip X-ray data in the same way as had been donefor SN 1988Z (Schlegel & Petre 2006). One might imagine thatthese inconsistencies could be attributed to some combination ofclumping or aspherical geometry, which can influence the emissiv-ity of H α or the escape of X-rays and radio differently for a given CSM mass. Dense regions that emit strong H α emission mightbe self absorbed in the radio, for example, and a range of den-sities may exist simultaneously if the emitting region is clumpy.Katsuda et al. (2016) discussed how asymmetry might influenceX-ray emission in SNe IIn. We noted earlier that there are someindications of asymmetry and different viewing angles betweenthe two SNe (SN 2005ip shows asymmetric blueshifted line pro-files in H α that imply dust formation blocks the far side, whereasSN 1988Z does not; this could perhaps be ascribed to a pole-onview of SN 1988Z and a nearly edge-on view of SN 2005ip).While such aspects are difficult to constrain with high confidence,the large binary fraction among massive stars (Moe & Di Stefano2016; Sana et al. 2012; Chini et al. 2012; Kiminki & Kobulnicky2012; Kiminki et al. 2012; Kobulnicky et al. 2014) and the high in-cidence of aspherical geometry in resolved CSM around nearbymassive stars suggests that effects such as asymmetry and clump-ing may be the norm rather than an exception. In this paper we have emphasized the remarkable similarity be-tween the late-time evolution of SN 2005ip and that of SN 1988Z,which is the prototypical long-lived SN IIn. So far there are threemain differences in their evolution.First, the CSM interaction in SN 1988Z was stronger at earlytimes and declined for the first 1000 days, whereas SN 2005ip hada more delayed onset of its strongest CSM interaction. SN 2005ip’sinteraction remained roughly constant until day 1000, after whichit tracked that of SN 1988Z. This relatively delayed onset inSN 2005ip was accompanied by stronger narrow coronal emissionlines at early times, probably signifying a more highly clumpedinner-wind region (compared to SN 1988Z), allowing the densestpre-shock regions to be photoionized by X-rays from a very fastblast wave (Smith et al. 2009a).The second major difference is that SN 2005ip has just re-cently shown a drop and then a quick resurgence in its H α lumi-nosity over the course of a few months, indicative of strong den-sity fluctuations in its distant CSM. The recent sudden increase inH α would reflect a density increase by a factor of 2–5. SN 1988Zhas shown no such behaviour, although brief blips cannot be ruledout by the coarse time sampling of our late-time SN 1988Z spec-tra. For a blast wave expanding at 2000–5000 km s − , fluctuationsseen ∼ days after SN 2005ip’s explosion would correspond toradii of (0.6–1.5) × cm, or roughly 0.02–0.05 pc from the star.Interestingly, this is consistent with the radius of a warm dust shellresponsible for the near/mid-IR echo seen at early times (Fox et al.2009, 2010), which was inferred to reside at ∼ . pc from thestar (Fox et al. 2011, 2013). This dust was ascribed to a shell ofenhanced density at that location. It therefere seems likely that thesudden increase in the H α luminosity that we observe around day3500 could in fact be caused by the forward shock reaching andovertaking this same dusty CSM shell. If so, continued study of thisinteraction may provide a novel way to investigate the destructionor survival of CSM dust that is hit by a SN blast wave.Third, despite their similar H α luminosity evolution,SN 2005ip declines more slowly in X-rays than SN 1988Z, but it isalso somewhat less luminous at radio wavelengths than SN 1988Zat a comparable epoch (see Sections 3.4 and 3.5 for details andcaveats). This suggests that any one diagnostic taken alone givesan incomplete picture of the progenitor mass loss.The recent H α variability suggests that SN 2005ip has strong MNRAS , 1– ?? (2015) Smith et al. deviations from a monotonically decreasing wind density at largeradii, but what caused this? A simple interpretation would be thatchanges in CSM density reflect previous changes in wind mass-loss rate. At a radius of 0.03 pc, and with a constant wind speedof ∼ km s − (somewhat faster wind speeds are appropriate formore luminous RSGs; see Smith 2014), this would point to substan-tial changes in mass loss about 1000 yr preceding core collapse. Asnoted in Section 1, sudden bursts of mass loss on these timescalesbefore core collapse do not fall within the realm of currently sug-gested ideas that involve Ne and O burning, which are too brief forobjects like SN 2005ip and SN 1988Z (Quataert & Shiode 2012;Shiode & Quataert 2014; Smith & Arnett 2014). Instead, it appearsas if the mass-loss rate was ramping up in the last few thousandyears before core collapse in SN 2005ip and SN 1988Z. Some ev-idence for such behaviour is observed among nearby populationsof RSGs (Beasor & Davies 2016), and it may be expected theoreti-cally (Heger et al. 1997; Yoon & Cantiello 2010; Smith 2014).Although RSG winds may ramp up near the end of a star’s life,strong fluctuations in CSM density may point to episodic mass lossakin to eruptions of luminous blue variables (LBVs) or unsteadybinary mass-transfer episodes. Indeed, LBV-like eruptive mass losshas been suggested as the possible origin of the CSM shells aroundSN 2005ip (Fox et al. 2013; Katsuda et al. 2014). Smith & Arnett(2014) have discussed how binary interaction triggered by pre-SNevolution might instigate LBV-like mass loss through a rapid on-set of common-envelope evolution. Recalling that LBVs are a phe-nomenological class and binary interaction is a physical mecha-nism, these might be two names for the same phenomenon. Somecurrent ideas for LBVs do favour binary interaction as a necessaryingredient in their evolution (Smith & Tombleson 2015).A drop in pre-shock density followed immediately by an in-crease in density, as indicated by the H α fluctuation of SN 2005ip,might also result from a temporary increase in wind speed thatswept material into a thinner shell. Thus, variations in wind speed,as opposed to just wind mass-loss rate, might also give rise to suchstructures. Wind speeds could vary significantly if the progenitorexperienced a blue loop on the Hertzsprung-Russell diagram, forexample. Perhaps the common assumption of ballistic speeds isnaive, but nevertheless, this material must be far from the star andquite old compared to the CSM overtaken by normal SNe IIn dur-ing their early bright phases.Another puzzling aspect of SN 2005ip’s late-time behaviourconcerns its X-ray emission. In 2013–2014, analysis of X-rays fromSN 2005ip (Katsuda et al. 2014) indicated a drop in both the X-ray flux and the absorbing neutral H column density ( N H ) alongour line of sight (reaching the expected Galactic line-of-sight valueby the last epoch). This prompted Katsuda et al. (2014) to sug-gest that SN 2005ip’s blast wave had reached the outer boundaryof the dense shell. However, the similarity to SN 1988Z (whichcontinued unabated for another decade after this epoch), the recentresurgence in H α emission from SN 2005ip, and strong radio andX-ray emission, indicate that its CSM interaction is not yet goingaway. In that case, how should we reconcile the observed drop in N H ? Katsuda et al. (2014) considered the possibility that the outerRSG wind is ionized by the luminosity from CSM interaction, butfound the ionization to be insufficient for expected shock param-eters. However, they did not discuss the possibility of external in-fluences on the RSG wind, and interpreted the drop in density as ashell resulting from a past LBV-like eruption, as noted above. (Theyalso noted that a change in the degree of clumping might explainthe drop in N H .)An alternative explanation for the origin of a sudden jump in density at a large radius may be an otherwise normal steadywind that becomes confined by external pressure at large radii(Garcia-Segura et al. 1996; Chita et al. 2008; Mackey et al. 2014).If the progenitor was indeed a more massive star (above 18–20 M ⊙ )so that it was an O-type star when on the main sequence, then itshot shocked main-sequence wind will provide an external pressurethat might decelerate the RSG wind and confine it to a thin shell. (Ifthe progenitor has similarly massive O-type neighbours in a youngstellar cluster, the hot interior of the H II region might create the re-quired external pressure and ionization.) Mackey et al. (2014) havediscussed the interesting possibility that some SNe IIn may arisefrom the interaction between the SN shock and this type of exter-nally confined, thin neutral shell. In practice, this may be quite rare:most normal RSG progenitors are at relatively low initial masses(i.e., 10–15 M ⊙ ), which means that they were not O-type stars onthe main-sequence, and they reach the RSG phase long after as-sociated O-type stars have died. On the other hand, more-massiveRSGs that die sooner have much higher wind momentum in theRSG phase than assumed in the models, so it remains unclear ifnormal RSG winds are likely to yield SNe IIn through this process.Although this scenario was proposed as a way to explainSNe IIn without appealing to eruptive pre-SN mass loss, it mayalso provide a way for an interacting SN to be “rejuvenated” inits old age, since the stalled shell is denser than a freely expand-ing wind. Mackey et al. (2014) estimate that such a shell can haveits density enhanced by a factor of 80, and may contain a signifi-cant fraction (perhaps a third) of the total mass lost during the RSGphase. Their models demonstrate that a SN can show a bolomet-ric luminosity spike at very late times from such a collision. In thiscase it is difficult to estimate the age of the shell, since its expansionmay have stalled, and its radius depends on the external pressure asmuch as on the momentum of the RSG wind. Future monitoring ofSN 2005ip may help to test this hypothesis, as the shock contin-ues to make its way through this dense neutral shell and out intothe ionized portion of the RSG wind, or if SN 2005ip resumes avery slow decline as SN 1988Z did. Also, higher-resolution spectramay be able to place tighter constraints on the pre-shock CSM ex-pansion speed, since the narrow components are unresolved in ourmoderate-resolution spectra presented here.Regardless of the true explanation for the density fluctuationin SN 2005ip’s outer CSM, it is useful to realize that mass-lossrates this high (several times − to − M ⊙ yr − or more)are quite extreme, and do not correspond to normal RSGs likeBetelgeuse, where the wind is two orders of magnitude less dense(Smith et al. 2009b). The cool evolved stars with winds this denseare rare (see, e.g., Smith 2014 for an overview of mass-loss rates),corresponding to the most luminous and extreme RSGs that areenshrouded by their own dusty winds, like VY CMa, NML Cyg,S Per, etc., or to self-enshrouded super-asymptotic giant branch(AGB) stars. In the cases of SN 1988Z and SN 2005ip, however, thevery large CSM mass needed for their sustained high-luminosityCSM interaction has been estimated as 10 M ⊙ or more. This con-fidently rules out super-AGB stars (total mass including a neutronstar of only ∼ M ⊙ ) as their possible progenitors. This would beunlikely anyway, as these super-AGB stars that may undergo O-Ne-Mg core collapse in an electron-capture event are expected toproduce a low SN kinetic energy of ∼ ergs (Nomoto et al.1982; Nomoto 1987). Thus, electron-capture SNe that produceSNe IIn through interaction may be expected to fade quickly (see,e.g., Mauerhan et al. 2013). SN 1988Z was hyperenergetic; up today 3000, Aretxaga et al. (1999) estimate a total radiated energyfor SN 1988Z of erg (model-dependent E rad ), and at least MNRAS , 1– ?? (2015)(2015)
X-ray evolution of SN 2005ip (this paper) and SN 1988Z(Schlegel & Petre 2006). The X-ray luminosity plotted for SN 1988Z is alower limit to the true luminosity of the transient, as no correction for any in-trinsic neutral hydrogen column density was attempted by Schlegel & Petre(2006). For SN 1988Z Schlegel & Petre (2006) assumed a 5 keV plasmain collisional equilibrium. For comparison with this X-ray light curve ofSN 1988Z, here we show the X-ray fluxes for SN 2005ip reduced in thesame way (not corrected for intrinsic absorption, integrated over the sameenergy range of 0.5–2 keV). This is different from the data shown in Fig-ure 3, but allows for a more direct comparison with SN 1988Z. The twoSNe overlap in X-ray luminosity in this energy band, but have somewhatdifferent deecay rates. and Schlegel & Petre 2006). Note that a direct comparison betweenX-rays from SN 2005ip and SN 1988Z is a bit tricky, becauseSchlegel & Petre (2006) made somewhat different assumptions intheir analysis. Therefore, for the purpose of comparing the twoobjects in Fig 8, we have recalibrated our SN 2005ip X-ray datawith the same assumptions (integrated over 0.5–2 keV, uncorrectedfor intrinsic absorption). Here we find that the X-ray luminositiesare about the same at ∼ d, although the somewhat faster de-cline of SN 2005ip is still apparent. Yet, despite this faster declinein X-rays around a decade after explosion, the H α luminosity ofSN 1988Z then hit a plateau and did not decline for another decadeafter this epoch. This constant H α luminosity seems inconsistentwith the steep drop in wind density implied by the X-rays. Over-all, it seems that a clear lesson from comparing various diagnosticsfrom different late-time interactors is that any single diagnostic byitself provides an incomplete picture, so readers should be awarethat some of the corresponding mass-loss rate estimates are likelyto be lower limits. Note that the value of ˙ M that we derive fromradio emission is an order of magnitude lower than from X-ray andH α observations of SN 2005ip at the same epoch.Consider the comparison of SN 1988Z and SN 2005ip adecade after explosion. The two have roughly equal X-ray andH α luminosities, but SN 1988Z is more luminous in radio emis-sion. A caveat is that the quoted X-ray luminosities are not bolo-metric X-ray luminosities, although we have tried to recalibrateour SN 2005ip X-ray data in the same way as had been donefor SN 1988Z (Schlegel & Petre 2006). One might imagine thatthese inconsistencies could be attributed to some combination ofclumping or aspherical geometry, which can influence the emissiv-ity of H α or the escape of X-rays and radio differently for a given CSM mass. Dense regions that emit strong H α emission mightbe self absorbed in the radio, for example, and a range of den-sities may exist simultaneously if the emitting region is clumpy.Katsuda et al. (2016) discussed how asymmetry might influenceX-ray emission in SNe IIn. We noted earlier that there are someindications of asymmetry and different viewing angles betweenthe two SNe (SN 2005ip shows asymmetric blueshifted line pro-files in H α that imply dust formation blocks the far side, whereasSN 1988Z does not; this could perhaps be ascribed to a pole-onview of SN 1988Z and a nearly edge-on view of SN 2005ip).While such aspects are difficult to constrain with high confidence,the large binary fraction among massive stars (Moe & Di Stefano2016; Sana et al. 2012; Chini et al. 2012; Kiminki & Kobulnicky2012; Kiminki et al. 2012; Kobulnicky et al. 2014) and the high in-cidence of aspherical geometry in resolved CSM around nearbymassive stars suggests that effects such as asymmetry and clump-ing may be the norm rather than an exception. In this paper we have emphasized the remarkable similarity be-tween the late-time evolution of SN 2005ip and that of SN 1988Z,which is the prototypical long-lived SN IIn. So far there are threemain differences in their evolution.First, the CSM interaction in SN 1988Z was stronger at earlytimes and declined for the first 1000 days, whereas SN 2005ip hada more delayed onset of its strongest CSM interaction. SN 2005ip’sinteraction remained roughly constant until day 1000, after whichit tracked that of SN 1988Z. This relatively delayed onset inSN 2005ip was accompanied by stronger narrow coronal emissionlines at early times, probably signifying a more highly clumpedinner-wind region (compared to SN 1988Z), allowing the densestpre-shock regions to be photoionized by X-rays from a very fastblast wave (Smith et al. 2009a).The second major difference is that SN 2005ip has just re-cently shown a drop and then a quick resurgence in its H α lumi-nosity over the course of a few months, indicative of strong den-sity fluctuations in its distant CSM. The recent sudden increase inH α would reflect a density increase by a factor of 2–5. SN 1988Zhas shown no such behaviour, although brief blips cannot be ruledout by the coarse time sampling of our late-time SN 1988Z spec-tra. For a blast wave expanding at 2000–5000 km s − , fluctuationsseen ∼ days after SN 2005ip’s explosion would correspond toradii of (0.6–1.5) × cm, or roughly 0.02–0.05 pc from the star.Interestingly, this is consistent with the radius of a warm dust shellresponsible for the near/mid-IR echo seen at early times (Fox et al.2009, 2010), which was inferred to reside at ∼ . pc from thestar (Fox et al. 2011, 2013). This dust was ascribed to a shell ofenhanced density at that location. It therefere seems likely that thesudden increase in the H α luminosity that we observe around day3500 could in fact be caused by the forward shock reaching andovertaking this same dusty CSM shell. If so, continued study of thisinteraction may provide a novel way to investigate the destructionor survival of CSM dust that is hit by a SN blast wave.Third, despite their similar H α luminosity evolution,SN 2005ip declines more slowly in X-rays than SN 1988Z, but it isalso somewhat less luminous at radio wavelengths than SN 1988Zat a comparable epoch (see Sections 3.4 and 3.5 for details andcaveats). This suggests that any one diagnostic taken alone givesan incomplete picture of the progenitor mass loss.The recent H α variability suggests that SN 2005ip has strong MNRAS , 1– ?? (2015) Smith et al. deviations from a monotonically decreasing wind density at largeradii, but what caused this? A simple interpretation would be thatchanges in CSM density reflect previous changes in wind mass-loss rate. At a radius of 0.03 pc, and with a constant wind speedof ∼ km s − (somewhat faster wind speeds are appropriate formore luminous RSGs; see Smith 2014), this would point to substan-tial changes in mass loss about 1000 yr preceding core collapse. Asnoted in Section 1, sudden bursts of mass loss on these timescalesbefore core collapse do not fall within the realm of currently sug-gested ideas that involve Ne and O burning, which are too brief forobjects like SN 2005ip and SN 1988Z (Quataert & Shiode 2012;Shiode & Quataert 2014; Smith & Arnett 2014). Instead, it appearsas if the mass-loss rate was ramping up in the last few thousandyears before core collapse in SN 2005ip and SN 1988Z. Some ev-idence for such behaviour is observed among nearby populationsof RSGs (Beasor & Davies 2016), and it may be expected theoreti-cally (Heger et al. 1997; Yoon & Cantiello 2010; Smith 2014).Although RSG winds may ramp up near the end of a star’s life,strong fluctuations in CSM density may point to episodic mass lossakin to eruptions of luminous blue variables (LBVs) or unsteadybinary mass-transfer episodes. Indeed, LBV-like eruptive mass losshas been suggested as the possible origin of the CSM shells aroundSN 2005ip (Fox et al. 2013; Katsuda et al. 2014). Smith & Arnett(2014) have discussed how binary interaction triggered by pre-SNevolution might instigate LBV-like mass loss through a rapid on-set of common-envelope evolution. Recalling that LBVs are a phe-nomenological class and binary interaction is a physical mecha-nism, these might be two names for the same phenomenon. Somecurrent ideas for LBVs do favour binary interaction as a necessaryingredient in their evolution (Smith & Tombleson 2015).A drop in pre-shock density followed immediately by an in-crease in density, as indicated by the H α fluctuation of SN 2005ip,might also result from a temporary increase in wind speed thatswept material into a thinner shell. Thus, variations in wind speed,as opposed to just wind mass-loss rate, might also give rise to suchstructures. Wind speeds could vary significantly if the progenitorexperienced a blue loop on the Hertzsprung-Russell diagram, forexample. Perhaps the common assumption of ballistic speeds isnaive, but nevertheless, this material must be far from the star andquite old compared to the CSM overtaken by normal SNe IIn dur-ing their early bright phases.Another puzzling aspect of SN 2005ip’s late-time behaviourconcerns its X-ray emission. In 2013–2014, analysis of X-rays fromSN 2005ip (Katsuda et al. 2014) indicated a drop in both the X-ray flux and the absorbing neutral H column density ( N H ) alongour line of sight (reaching the expected Galactic line-of-sight valueby the last epoch). This prompted Katsuda et al. (2014) to sug-gest that SN 2005ip’s blast wave had reached the outer boundaryof the dense shell. However, the similarity to SN 1988Z (whichcontinued unabated for another decade after this epoch), the recentresurgence in H α emission from SN 2005ip, and strong radio andX-ray emission, indicate that its CSM interaction is not yet goingaway. In that case, how should we reconcile the observed drop in N H ? Katsuda et al. (2014) considered the possibility that the outerRSG wind is ionized by the luminosity from CSM interaction, butfound the ionization to be insufficient for expected shock param-eters. However, they did not discuss the possibility of external in-fluences on the RSG wind, and interpreted the drop in density as ashell resulting from a past LBV-like eruption, as noted above. (Theyalso noted that a change in the degree of clumping might explainthe drop in N H .)An alternative explanation for the origin of a sudden jump in density at a large radius may be an otherwise normal steadywind that becomes confined by external pressure at large radii(Garcia-Segura et al. 1996; Chita et al. 2008; Mackey et al. 2014).If the progenitor was indeed a more massive star (above 18–20 M ⊙ )so that it was an O-type star when on the main sequence, then itshot shocked main-sequence wind will provide an external pressurethat might decelerate the RSG wind and confine it to a thin shell. (Ifthe progenitor has similarly massive O-type neighbours in a youngstellar cluster, the hot interior of the H II region might create the re-quired external pressure and ionization.) Mackey et al. (2014) havediscussed the interesting possibility that some SNe IIn may arisefrom the interaction between the SN shock and this type of exter-nally confined, thin neutral shell. In practice, this may be quite rare:most normal RSG progenitors are at relatively low initial masses(i.e., 10–15 M ⊙ ), which means that they were not O-type stars onthe main-sequence, and they reach the RSG phase long after as-sociated O-type stars have died. On the other hand, more-massiveRSGs that die sooner have much higher wind momentum in theRSG phase than assumed in the models, so it remains unclear ifnormal RSG winds are likely to yield SNe IIn through this process.Although this scenario was proposed as a way to explainSNe IIn without appealing to eruptive pre-SN mass loss, it mayalso provide a way for an interacting SN to be “rejuvenated” inits old age, since the stalled shell is denser than a freely expand-ing wind. Mackey et al. (2014) estimate that such a shell can haveits density enhanced by a factor of 80, and may contain a signifi-cant fraction (perhaps a third) of the total mass lost during the RSGphase. Their models demonstrate that a SN can show a bolomet-ric luminosity spike at very late times from such a collision. In thiscase it is difficult to estimate the age of the shell, since its expansionmay have stalled, and its radius depends on the external pressure asmuch as on the momentum of the RSG wind. Future monitoring ofSN 2005ip may help to test this hypothesis, as the shock contin-ues to make its way through this dense neutral shell and out intothe ionized portion of the RSG wind, or if SN 2005ip resumes avery slow decline as SN 1988Z did. Also, higher-resolution spectramay be able to place tighter constraints on the pre-shock CSM ex-pansion speed, since the narrow components are unresolved in ourmoderate-resolution spectra presented here.Regardless of the true explanation for the density fluctuationin SN 2005ip’s outer CSM, it is useful to realize that mass-lossrates this high (several times − to − M ⊙ yr − or more)are quite extreme, and do not correspond to normal RSGs likeBetelgeuse, where the wind is two orders of magnitude less dense(Smith et al. 2009b). The cool evolved stars with winds this denseare rare (see, e.g., Smith 2014 for an overview of mass-loss rates),corresponding to the most luminous and extreme RSGs that areenshrouded by their own dusty winds, like VY CMa, NML Cyg,S Per, etc., or to self-enshrouded super-asymptotic giant branch(AGB) stars. In the cases of SN 1988Z and SN 2005ip, however, thevery large CSM mass needed for their sustained high-luminosityCSM interaction has been estimated as 10 M ⊙ or more. This con-fidently rules out super-AGB stars (total mass including a neutronstar of only ∼ M ⊙ ) as their possible progenitors. This would beunlikely anyway, as these super-AGB stars that may undergo O-Ne-Mg core collapse in an electron-capture event are expected toproduce a low SN kinetic energy of ∼ ergs (Nomoto et al.1982; Nomoto 1987). Thus, electron-capture SNe that produceSNe IIn through interaction may be expected to fade quickly (see,e.g., Mauerhan et al. 2013). SN 1988Z was hyperenergetic; up today 3000, Aretxaga et al. (1999) estimate a total radiated energyfor SN 1988Z of erg (model-dependent E rad ), and at least MNRAS , 1– ?? (2015)(2015) ndurance of SN 2005ip × erg (observed E rad ). SN 2005ip appears very similar inmany respects, and is even more luminous and slower to decline inX-rays.The extreme RSGs mentioned have initial masses of order 25–35 M ⊙ , higher than normal unobscured RSGs, with more tightlybound cores that may require more energetic explosions than nor-mal SNe II-P. This reinforces earlier suggestions that these ex-treme RSGs (as opposed to normal RSGs) are the likely progen-itors of some SNe IIn, including long-lived SN 1988Z-like events(Smith et al. 2009a,b; Van Dyk et al. 1993; Williams et al. 2002).More massive LBV stars may also be able to supply the requiredamount of CSM mass (Smith 2014; Smith & Owocki 2006). How-ever, the steady late-time decline over decades in the case ofSN 1988Z argues in favour of a freely expanding, slow, and sus-tained dense RSG wind, rather than the sporadic eruptive massloss characteristic of LBVs. For SN 2005ip, future observationsof its continuing evolution through the extended CSM may helpus choose more confidently between an extreme RSG — perhapsexposed to external ionizing radiation from its environment — oran eruptive progenitor. In any case, the strong variability in H α around day 3500 in SN 2005ip has not been seen so clearly beforeat such late times in SNe IIn, and suggests that similar late-timeobservations of other objects can provide important constraints onthe variability in their pre-SN mass loss.Strong and variable mass loss of the type seen here occur-ring 1000 yr before death presents a challenge for models of pre-SNevolution. Once we have a larger number of known SNe IIn withsuch enduring strong CSM interaction, it would be interesting toinvestigate their statistical proximity to H II regions as compared toother populations of SNe. ACKNOWLEDGMENTS
We thank an anonymous referee for helpful suggestions. Support wasprovided by the National Science Foundation (NSF) through grants AST-1210599 and AST-1312221 to the University of Arizona. C.D.K.’s researchreceives support from NASA through Contract Number 1255094 issued byJPL/Caltech. W.F. was supported by NASA through an Einstein Postdoc-toral Fellowship. The supernova research of A.V.F.’s group at U.C. Berkeleyis supported by Gary & Cynthia Bengier, the Richard & Rhoda GoldmanFund, the Christopher R. Redlich Fund, the TABASGO Foundation, andNSF grant AST-1211916.We thank the staffs at the MMT and Keck Observatories for their as-sistance with the observations. Observations using Steward Observatory fa-cilities were obtained as part of the large observing program AZTEC: Ari-zona Transient Exploration and Characterization. Some of the observationsreported in this paper were obtained at the MMT Observatory, a joint facilityof the University of Arizona and the Smithsonian Institution. This researchwas also based in part on observations made with the LBT. The LBT isan international collaboration among institutions in the United States, Italy,and Germany. The LBT Corporation partners are the University of Arizonaon behalf of the Arizona university system; the Istituto Nazionale di As-trofisica, Italy; the LBT Beteiligungsgesellschaft, Germany, representingthe Max-Planck Society, the Astrophysical Institute Potsdam, and Heidel-berg University; the Ohio State University and the Research Corporation,on behalf of the University of Notre Dame, University of Minnesota, andUniversity of Virginia. Some of the data presented herein were obtained atthe W.M. Keck Observatory, which is operated as a scientific partnershipamong the California Institute of Technology, the University of California,and NASA; the observatory was made possible by the generous financialsupport of the W.M. Keck Foundation. The authors wish to recognize andacknowledge the very significant cultural role and reverence that the summitof Mauna Kea has always had within the indigenous Hawaiian community.We are most fortunate to have the opportunity to conduct observations fromthis mountain.
REFERENCES
Andrews JE, Smith N, Mauerhan JC, 2015, MNRAS, 451, 1413Aretxaga I, et al. 1999, MNRAS, 309, 343Asplund M, Grevesse N, Sauval AJ, Scott P, 2009, ARA&A, 47, 481Barlow MJ, et al., 2005, ApJ, 627, L113Beasor ER, Davies B, 2016, MNRAS, in press (arXiv:1608.03895)Bietenholz MF, Bartel N, Rupen MP, 2002, ApJ, 581, 1132Burrows DN, et al., 2005, SSR, 120, 165Byard PL, OBrien TP, 2000, Proc. SPIE, 4008, 934Chandra P, et al., 2009, ApJ, 690, 1839Chevalier RA, Fransson C, 1994, ApJ, 420, 268Chini R., Hoffmeister V. H., Nasseri A., Stahl O., Zinnecker H., 2012, MN-RAS, 424, 1925Chita SM, Langer N, van Marle AJ, Garca-Segura G, Heger A, 2008, A&A,488, L37Chugai NN, Chevalier RA, 2006, ApJ, 641, 1051Chugai NN, Danziger IJ, 1994, MNRAS, 268, 173Crotts APS, Heathcote SR, 2000, ApJ, 528, 426Dessart L, Audi E, Hillier DJ 2015, MNRAS, 449, 4304Faber SM, et al., 2003, in Iye M, Moorwood AFM, ed., Proc. SPIE Vol.4841, Instrument Design and Performance for Optical/Infrared Ground-based Telescopes. SPIE, Bellingham, p. 1657Fabian AC, Terlevich R, 1996, MNRAS, 280, L5Fassia A, et al., 2001, MNRAS, 325, 907Filippenko AV, 1991, in SN 1987A and Other Supernovae, ed. IJ Danziger& K Kjar (Garching: ESO), 343Filippenko AV, 1997, ARAA, 35, 309Fransson C, et al., 1996, ApJ, 461, 993Fransson C, Bj¨ornsson C-I, 1998, ApJ, 509, 861Fransson C, et al. 2014, ApJ, 797, 118Fox OD, et al., 2009, ApJ, 691, 650Fox OD, Chevalier RA, Dwek E, Skrutskie MF, Sugerman BEK, LeisenringJM, 2010, ApJ, 725, 1768Fox OD, Chevalier RA, Skrutskie MF, Soderberg AM, Filippenko AV,Ganeshalingam M, Silverman JM, Smith N, Steele TN, 2011, ApJ, 741,7Fox OD, et al., 2013, AJ, 146, 2Fox OD, et al., 2015, MNRAS, 454, 4366France K, et al., 2015, ApJ, 801, L16Gal-Yam A, et al., 2014, Nature, 509, 471Garca-Segura G, Langer N, Mac Low MM, 1996, A&A, 316, 133Gehrels N, et al., 2004, ApJ, 611, 1005Greisen EW, 2003, Information Handling in Astronomy — Historical Vis-tas, 285, 109Heger A, Jeannin L, Langer N, Baraffe I, 1997, A&A, 327, 224Katsuda S, Maeda K, Nozawa T, Pooley D, Immler S, 2014, ApJ, 780, 184Katsuda S, et al. 2016, arXiv:1609.09093Kelson D, 2003, PASP, 115, 688Khasov D, et al., 2016, ApJ, 818, 3Kiewe M, et al., 2012, ApJ, 744, 10Kiminki D.C., Kobulnicky H.A., 2012, ApJ, 751, 4Kiminki D.C. et al. 2012, ApJ, 747, 41Kobulnicky H.A. et al., 2014, ApJS, 213, 34Leonard DC, et al., 2000, ApJ, 536, 239Li W, et al., 2011, MNRAS, 412, 1441Mackey J, Mohamed S, Gvaramadze VV, Kotak R, Langer N, Meyer DMA,Moriya TJ, Neilsen HR, 2014, Nature, 512, 282Margutti R, et al., 2013, MNRAS, 428, 729Margutti R, et al., 2016, preprintMatheson T, et al., 2000, AJ, 120, 1499Mauerhan J, Smith N, 2012, MNRAS, 424, 2659Mauerhan JC, Smith N, Silverman JM, et al., 2013, MNRAS, 431, 2599Meaburn J, Bryce M, Holloway AJ, 1995, A&A, 229, L1Michael E, McCray R, Borkowski KJ, Pun CSJ, Sonneborn G, 1998, ApJ,492, L143Miller AA, et al., 2010a, MNRAS, 404, 305MNRAS , 1– ?? (2015) Smith et al.
Miller AA, Smith N, Bloom JS, Chornock R, Filippenko AV, Prochaska JX,2010b, AJ, 139,2218Milisavljevic D, Fesen RA, Chevalier RA, Kirshner RP, Challis P, TurattoM, 2012, ApJ, 751, 25Milisavljevic D, et al., 2015, ApJ, 815, 120Moe M, Di Stefano R, 2016, preprint (arXiv:1606.05347)Niemela VS, Ruiz MT, Phillips MM. 1985, ApJ, 289, 52Nomoto K, 1987, ApJ, 277, 791Nomoto K, Sugimoto D, Sparks WM, Fesen RA, Gull TR, Miyaji S, 1982,Nature, 299, 803Ofek EO, et al., 2014, ApJ, 781, 42Oke JB, Cohen JG, Carr M, et al., 1995, PASP, 107, 375Pozzo M, et al., 2004, MNRAS, 352, 457Quataert E, Shiode J, 2012, MNRAS, 423, L92Quimby R, Wheeler JC, Hoflich P, Akerlof CW, Brown PJ, Rykoff ES.2007, ApJ, 666, 1093Sana H, de Mink SE, de Koter A, et al., 2012, Science, 337, 444Schlegel EM, Petre R, 2006, ApJ, 646, 378Shiode J, Quataert E, 2014, ApJ, 780, 96Shivvers I, Mauerhan JC, Leonard DC, Filippenko AV, Fox OD, 2015, ApJ,806, 213Smith N, 2014, ARA&A, 52, 487Smith N, Arnett WD, 2014, ApJ, 785, 82Smith N, Owocki SP, 2006, ApJ, 645, L45Smith N, Tombleson R, 2015, MNRAS, 447, 602Smith N, Li W, Foley RJ, et al., 2007, ApJ, 666, 1116Smith N, Chornock R, Li W, Ganeshalingam M, Silverman JM, Foley RJ,Filippenko AV, Barth AJ, 2008a, ApJ, 686, 467Smith N, et al., 2008b, ApJ, 686, 485Smith N, et al., 2009a, ApJ, 695, 1334Smith N, Hinkle KH, Ryde N, 2009b, AJ, 137, 3558Smith N, et al., 2010, ApJ, 709, 856Smith N, et al., 2011, MNRAS, 412, 1522Smith N, et al., 2012, AJ, 143, 17Smith N, et al., 2015, MNRAS, 449, 1876Sonneborn G, et al., 1998, ApJ, 492, L139Stathakis RA, Sadler EM, 1991, MNRAS, 250, 786Stritzinger M, et al., 2012, ApJ, 756, 173Taddia F, et al., 2013, A&A, 38, 113Turatto M, et al., 1993, MNRAS, 262, 128Van Dyk SD, Weiler KW, Sramek RA, Panagia N 1986, ApJ, 419, L69Weiler KW, Sramek RA, Panagia N, van der Hulst JM, Salvati M, 1986,ApJ, 301, 790Weiler KW, Panagia N, Montes MJ, Sramek RA. 2002, ARA&A, 40, 387Williams CL, Panagia N, Van Dyk SD, Lacey CK, Weiler KW, Sramek RA,2002, ApJ, 581, 396Woosley SE, 2016, preprint (arXiv:1608.08939)Woosley SE, Blinnikov S, Heger A, 2007, Nature, 450, 390Wright EL, 1980, ApJ, 242, L23Yoon SC, Cantiello M, 2010, ApJ, 717, L62 MNRAS , 1– ?? (2015)(2015)