Iptf14hls As a Variable Hyper-Wind from a Very Massive Star

arXiv:1911.01740v1 [astro-ph.HE] 5 Nov 2019 ( MNRAS n eosreteeet ste xad hrfr,we Therefore, ⋆ expand. they as ejecta the observe we and n iia pcr yial atfrol 0 asand days 100 only for (e.g., show- luminosity last SNe lower much while typically a years, reach spectra 2 similar almost for ing bright very however, oiyadpoopei ai eesbl,ulk hs ob- those unlike ( subtle, SNe were in radii served photospheric and SNe locity IIP Type in spec- as Its lines year. hydrogen one ( by than dominated more and were months for tra three luminous about very over remained luminosity in rose iPTF14hls PF4l savral ye-idfo eymassive very a from hyper-wind star variable a as iPTF14hls © cetd21 coe 8 eevd21 coe 6 norigi in 16; October 2019 Received 18. October 2019 Accepted clo N rgntn rmteclas fmsiestars massive of collapse the from originating ( typ- are SNe spectra Universe. showed of (SNe) and our SNe ical as in Supernovae observe bright as phenomena regularly stars. became variable iPTF14hls We brightest of the time. fading among in and dynamic brightening is Universe Our INTRODUCTION 1 3 2 1 5 4 raie l 2017 al. et Arcavi raie l 2017 al. et Arcavi 1997 Filippenko aah .Moriya, J. Takashi -al aah.oianoa.p(TJM) [email protected] E-mail: srpyisRsac nttt,LvrolJh orsUn Moores John Liverpool Institute, Research Astrophysics colo hsc n srnm,Fclyo cec,Monash Science, of Faculty Astronomy, and Physics of School ainlAtooia bevtr fJpn ainlInst National Japan, of Observatory Astronomical National NF srpyisadSaeSineOsraoy i .Go P. via Observatory, Science Space and Astrophysics INAF, a-lnkIsiuefrAtohsc,Karl-Schwarzschi Astrophysics, for Institute Max-Planck 09TeAuthors The 2019 nSe aseeto cusoe ml timescale, small a over occurs ejection mass SNe, In 000 , 1 – raie l 2017 al. et Arcavi 7 n eandsmlrframs years 2 almost for similar remained and ) 21)Pern oebr21 oplduigMRSL MNRAS using Compiled 2019 November 6 Preprint (2019) .Drn hspro,cagsi ieve- line in changes period, this During ). ; olra ta.2019 al. et Sollerman 1 , 2 PF4l uigte2ya rgtpaewsmr hnafew a than mas more instantaneous was the phase that the bright find estimate 2-year We we phase. the hypothesis, bright during ste wind the iPTF14hls a the in as on iPTF14hls interpreted Based of are rath when rate. which but natural iPTF14hls, mass-loss supernovae of become Weable like properties scenario, unclear. The outburst supernova still winds. a sudden is stellar a to evolution, similar not spectral b outflow was little spectra supernova-like with iPTF14hls IIP years that Type two had almost which for iPTF14hls, of origin The ABSTRACT id)adi ece smc s10 as much as reached it and wind”) iia hp ota of that to shape similar a bu 10 about was otnu-rvneteewn rmvr asv stars. words: massive Key very from wind extreme continuum-driven a iia u eseteebiheigacmaidb extraordinary by 10 accompanied than brightening more extreme ding less but similar a fi tl xss hc ssmlrto similar is which exists, still it if ⋆ ). al .Mazzali, A. Paolo nesne l 2014 al. et Anderson .iT1hswas, iPTF14hls ). M tr:msie–sas asls tr:wns outflows winds, stars: – mass-loss stars: – massive stars: ⊙ neetnl,w n httelgtcreo PF4l a very a has iPTF14hls of curve light the that find we Interestingly, . M dSrß ,878Grhn,Germany Garching, 85748 1, ld-Straße ⊙ vriy C,LvrolSinePr,16Bono il L Hill, Brownlow 146 Park, Science Liverpool IC2, iversity, a om21 etme 04 September 2019 form nal ttso aua cecs -11Oaa iaa oy 181 Tokyo Mitaka, Osawa, 2-21-1 Sciences, Natural of itutes ). n1 er.Tepoeio fiT1hsi esta 150 than less is iPTF14hls of progenitor The years. 10 in η et 0,419Blga Italy Bologna, 40129 101, betti aia uigteGetEuto,wihas experienced also which Eruption, Great the during Carinae nvriy lyo,VC30,Australia 3800, VIC Clayton, University, 3 , 4 e epradsoe at fteeet stm osby goes time as ejecta the of parts slower ( and deeper see 2011 age l 2018 al. et Wang 2018 ohv nsa hsclcniin o Nadi was it and (e.g., SN studies needs previous a in SN iPTF14hls for peculiar Thus, a conditions be to iPTF14hls. physical evolu- claimed unusual in velocity have seen This to not time. was with tion decreases which velocity, emms jcinlk N ecl uha xrm mass short- extreme an violent, such a call We than SN. a rather like star an ejection mass massive to term related a is from iPTF14hls m outflow time-resolved outflow. that long-term continuous variable, propose continuous, long-term we extreme, a paper, was this phenomenon it In explosive if short-time natural makes a becomes What mass as continuous. where is peculiar surface winds, iPTF14hls stellar stellar the velocitie from in line ejection constant properties this observe photospheric do of We example and wind. obvious stellar An a short timescale. is a long kind suc a in situation, over place steady-state loss takes a mass for that natural ejection is mass but timescale, a in expected not azl ta.2004 al. et Mazzali n ln Pian Elena and ako vlto ftepoopei rprisis properties photospheric the of evolution of lack A ; photospheric the of evolution the by tracked is This ). η ose 2018 Woosley aia.Tetopeoeamyb eae to related be may phenomena two The Carinae. M ⊙ yr − ; 1 i ta.2019 al. et Liu h oa asls vrtoyears two over lost mass total The . ; ; ae ose 2009 Woosley & Kasen hgi2018 Chugai 5 ; omn&Skr2019 Soker & Gofman ; lrwn ihvari- with wind llar oe iks2018 Gilkis & Soker M A asls,shed- loss, mass T ⊙ asls rates mass-loss ralong-term a er E tkp bright kept ut ; tl l v3.0 file style X yr -osrt of rate s-loss eepropose here tod with odds at ese tal. et Bersten − vrolL R,UK 5RF, L3 iverpool 1 88,Japan -8588, (“hyper- Dessart as h M ). ass ⊙ s ; 2 T. J. Moriya et al.

10 12

11 ) -1

) 10 -1

erg s 9

1 42 8

6 0.1 5 wind velocity (1000 km s β = 2 3 4

5 bolometric luminosity (10 10 3 0.01 0.1 1 10 100 2 0 100 200 300 400 500 600 700 radius (1015 cm) observed days after discovery

Figure 1. Possible wind velocity structure in iPTF14hls. Figure 2. Bolometric LC of iPTF14hls from Sollerman et al. (2019). The epochs of the 18 spectra we selected for our analysis are marked by vertical purple lines. All spectra (reported in Fig. −1 outflow from massive stars, which exceeds 1 M⊙ yr in the A1) were fitted with a blackbody and the fitting parameters were case of iPTF14hls as we shall see, a “hyper-wind.” used to estimate the photospheric properties.

the terminal wind velocity is achieved well above the photo- 2 IPTF14HLS AS A STELLAR WIND sphere. If this velocity structure does not change much, we We first discuss the velocity structure of the wind in can observe P-Cygni profiles that do not change much over iPTF14hls leading to its extremely slow spectral evolution. time because this velocity structure is fixed in the radius. Arcavi et al. (2017) estimated that the velocity measured This explains the spectra without apparent velocity change from the P-Cygni profile of Fe ii λ5169 is ≃ 4000 km s−1, in iPTF14hls which is not expected in expanding ejecta. R which can be considered as the velocity at the photosphere. The wind launching radius 0 strongly depends on the The velocities measured from the hydrogen lines are faster, assumed β if we adopt the β velocity law (Fig. 1). They are 14 i.e., 8000 − 6000 km s−1 for Hα and 7000 − 5000 km s−1 for (5 − 10)× 10 cm with β ≃ 2 − 5. The progenitor radius of 14 Hβ. The absorption in the hydrogen lines extends to ≃ iPTF14hls is, therefore, likely around (5 − 10)× 10 cm. 10000 km s−1 and, therefore, the maximum velocity reached by the wind is likely around 10000 km s−1. We interpret that 10000 km s−1 is the terminal velocity of the wind forming iPTF14hls and the lower velocities are found as the line for- 3 MASS-LOSS RATES OF THE PROGENITOR mation occurs in the wind accelerated towards the terminal If iPTF14hls is a stellar wind, the next important question velocity. is the mass-loss rates of the progenitor causing iPTF14hls. The acceleration of the wind is often approximated by To estimate the mass-loss rates of iPTF14hls, we first eval- using the so-called β velocity law, i.e., uate photospheric radii and temperatures of iPTF14hls by R β fitting blackbody functions to selected spectra. The epochs V r = V 0 wind( ) ∞ 1 − r , (1) of the spectra when we performed the blackbody fitting are shown in Fig. 2. We took the spectra at these epochs from −1 1 where V∞ = 10000 km s is the terminal wind velocity and WISeREP (Yaron & Gal-Yam 2012). These spectra are reported by Arcavi et al. (2017). The spectra in the database R0 is the radius where the wind is launched. β represents how fast the wind is accelerated. As β increases, the wind is more were flux calibrated based on the photometry (Arcavi et al. slowly accelerated. Extended stars such as red supergiants 2017). All the flux calibrated spectra we used are shown in (RSGs) are known to have a slow wind acceleration with Appendix A with photometry. β & 2 (e.g., Bennett 2010). Because iPTF14hls has a large Our blackbody fitting was performed by taking line 15 blanketing in the blue and ultraviolet into account. Even photospheric radius (Rph ≃ 2.5 × 10 cm) as shown in the following Section 3, the wind acceleration in iPTF14hls is if photons are emitted with the blackbody energy distribu- likely slow and β could be at least as large as those found in tion from the photosphere, photons can be absorbed and re- RSGs. emitted above the photosphere and the spectra do not keep 15 −1 the original blackbody shape. This line blanketing mostly Taking Rph ≃ 2.5 × 10 cm and vph ≃ 4000 km s , we can calibrate Eq. (1) for given β and see the wind velocity absorbs blue and ultraviolet photons and they are re-emitted structure in iPTF14hls. Although the wind does not neces- in redder wavelengths. Thus, this effect changes the origi- sarily follow the β law, we can have a general idea of the nal blackbody spectra especially in the blue and ultravio- wind structure in this way. let. Thus, we fit the spectra by matching the redder part The wind velocity and density structure estimated with the above method is shown in Fig. 1. Although the wind velocity at the photosphere (≃ 2.5 × 1015 cm) is 4000 km s−1, 1 https://wiserep.weizmann.ac.il

MNRAS 000, 1–7 (2019) iPTF14hls as a variable hyper-wind 3 of the continuum spectra to the blackbody function and al- 4 9 radius temperature lowing flux excess in the blue and ultraviolet. The previous studies obtaining the blackbody properties of iPTF14hls 3.5 8 (Arcavi et al. 2017; Sollerman et al. 2019) simply fitted pho- cm) tometry with the blackbody function and did not take the 15 3 line blanketing effect into account. 7

The results of the blackbody ﬁtting with errors are pre- 2.5 sented in Appendix A. Fig. 3 shows the estimated photospheric radii and temperatures. There is some diﬀerence be- 6 2 tween our estimates and the results mentioned in previous

works by Arcavi et al. (2017); Sollerman et al. (2019) caused photospheric radius (10 5 1.5 by the line blanketing. In particular, we found almost con- phoospheric temperature (1000 K) stant photospheric radii ((2−3)×1015 cm) and slightly higher 1 4 photospheric temperatures (6250 − 7500 K) than the previ- 50 100 150 200 250 300 350 400 450 500 ous estimates (5000 − 6000 K). The temperatures we derive days after discovery in the rest frame are consistent with the recombination temperature of ionised hydrogen and support the idea that the spectra are formed Figure 3. Photospheric radii and temperatures of iPTF14hls. at the electron scattering photosphere of the hydrogen-rich They are derived by ﬁtting a blackbody function to the observed wind, as also suggested by the increasingly strong Hα emis- spectra with line blanketing in mind. sion component. This causes the spectra of iPTF14hls to resemble those of Type IIP SNe. Based on our measurements of the photospheric radii 14 11 mass-loss rate and temperatures, we estimate the mass-loss rate of bolometric light curve

10 ) iPTF14hls. The mass-loss rate at the photosphere is ex- 12 -1

pressed as ) 9 -1 erg s 42 MÛ = R2 V yr 10 4π ph ρph ph, (2) ⊙ 8 R where ph is the photospheric radius, ρph is the den- 8 7 sity at photosphere, and vph is the photospheric velocity. In order to form the P-Cygni profiles of hydrogen lines 6 6 as observed in iPTF14hls, the electron number density mass-loss rate (M 5 is required to be ≃ (8 ± 1) × 109 cm−3 when the photo- 4 spheric temperature is 6000 − 7000 K as found in iPTF14hls 4 bolometric luminosity (10 (Eastman et al. 1996; Owocki & Shaviv 2016). Assuming a 2 3 solar composition, the corresponding photospheric density 50 100 150 200 250 300 350 400 450 500 −14 −3 is ρph ≃ (1.8 ± 0.3)× 10 g cm . The velocity at the pho- days after discovery in the rest frame −1 tosphere is Vph ≃ 4000 km s throughout the peak phase based on Fe ii λ5169 line measurements (Arcavi et al. 2017; Figure 4. Estimated mass-loss rate history of iPTF14hls (left Sollerman et al. 2019). We use the photospheric radii pre- axis). The bolometric LC of iPTF14hls (Sollerman et al. 2019) is overplotted for comparison (right axis). sented in Fig. 3. The estimated mass-loss rate history of iPTF14hls is presented in Fig. 4. The estimated mass-loss rates exceed −1 1 M⊙ yr and they are extremely large (“hyper-wind”). The the temperature remains roughly constant, indicating an op- −1 mass-loss rate is estimated to be ≃ 9 M⊙ yr on average tically thick regime. In the second observing season after from 100 days to 260 days and about 4 M⊙ are lost in this ∼ 350 days, the mass-loss rate becomes smaller as iPTF14hls period. Initially, up to ∼ 150 days, the rise in luminosity pre- gets fainter. The terminal velocity of the outflow probably cedes that of the mass-loss rate. Simultaneously, the tem- decreases somewhat, as shown by the slow evolution of the perature decreases. This indicates an initial expansion, as observed velocity of both the Hα and Hβ P-Cygni absorption traced by the increase in radius. This phase likely repre- components. The photospheric radius also starts to decrease, sents the onset of the high mass-loss episode. Although the suggesting that the hydrogen-dominated gas is progressively mass-loss rate starts to increase when the luminosity also recombining. This is similar to the situation of a Type IIP does, the first peak in the mass-loss rate at the photospheric SN at the end of the plateau phase, and is confirmed by the radius is reached some 2 months after the first luminosity increasing importance of Hα emission with respect to overall peak. This is the time it takes gas moving at 4000 km s−1 luminosity. to reach the photospheric radius (2.5 × 1015 cm). Thereafter, Between 350 and 460 days the average mass-loss rate is −1 the mass-loss rate follows the behaviour of the light curve about 6 M⊙ yr , and the total mass lost during this period (LC) – when the LC gets brighter, the mass-loss rates be- is about 2 M⊙. The total mass lost in 2 years may be close to come larger, and vice-versa. Several peaks in the mass-loss 10 M⊙. Because the mass that is ejected is accelerated to a rates can be seen, some sharper and some smoother. In par- terminal velocity of 10000 km s−1, the total kinetic energy of ticular, a sharp peak is seen at 220 − 250 days, when instan- the outflow may be around 1052 erg – much higher than the −1 51 taneous mass-loss rates in excess of 10 M⊙ yr are reached. standard SN explosion energy of 10 erg (Pejcha & Prieto The radius responds to these changes in mass-loss rates, but 2015).

MNRAS 000, 1–7 (2019) 4 T. J. Moriya et al.

4 IPTF14HLS AND THE “GREAT ERUPTION” -19

−1 -18 The huge mass-loss rates exceeding 1 M⊙ yr , accompanied by significant stellar brightening in iPTF14hls are reminis- -17 cent of η Carinae (Davidson & Humphreys 1997). η Cari- -16 nae suddenly became bright in 1838 and it maintained a -15 high luminosity for about 10 years (Fernández-Lajús et al. 2009). During this period known as the “Great Erup- -14 tion”, η Carinae ejected a large amount of mass, which we -13 currently observe as the nebula called the “Homunculus” absolute magnitude -12 (Davidson & Humphreys 1997). The mass-loss rate during the Great Eruption is estimated to have been more than -11 −1 -10 iPTF14hls (g band) 1 M⊙ yr (Morris et al. 1999; Smith 2013, 2006). Interest- η Carinae (visual), scaled ingly, the spectra of η Carinae during the Great Eruption -9 obtained by observing its light echoes have similar hydro- -500 0 500 1000 1500 2000 gen P-Cygni profiles and an indication of a high velocity days ≃ −1 component in hydrogen ( 10000 km s ) as in iPTF14hls Figure 5. Comparison of the LCs of iPTF14hls (Rest et al. 2012; Prieto et al. 2014; Smith et al. 2018). (Arcavi et al. 2017) and η Carinae at the Great Eruption If iPTF14hls is a similar mass-loss event to the Great (Fernández-Lajús et al. 2009). The LC of η Carinae is scaled by Eruption of η Carinae, the progenitor may not have dis- a factor of 0.15 in time and 100 (5 magnitudes) in brightness. appeared. Its last observed luminosity, some 3 years after The triangles are the unfiltered upper limits from Catalina Sky 6 discovery, was ≈ 5 × 10 L⊙ (Sollerman et al. 2019). The Survey (Arcavi et al. 2017). progenitor mass of iPTF14hls, therefore, must be smaller than 150 M⊙ if it still exists (e.g., Ekström et al. 2012; Georgy et al. 2013; Szécsi et al. 2015; Yoon et al. 2012). of a very massive star. However, its spectrum looked like that This mass is similar to (or slightly larger than) the mass of a Type IIP SN. This is not a problem, as long as the conditions in the outflow (density, temperature) are similar to of η Carinae (∼ 100 M⊙) (Davidson & Humphreys 1997). The mass-loss and brightening mechanisms of η Carinae are those of the line-forming region of a Type IIP SN, where hy- not understood well, but we found that there is a striking drogen recombination causes a sudden jump in opacity and similarity between the LC of iPTF14hls and that of η Cari- electron scattering is the main source of opacity below that nae during the Great Eruption, when the latter is scaled region. One of the diagnostics of hydrogen recombination is in time and luminosity (Fig. 5). iPTF14hls was 100 times the dominance of the emission component over the absorp- brighter and 6.5 times shorter-lived than the Great Erup- tion one in the Hα P-Cygni profile. This is seen in Type IIP tion. The similarity may indicate that iPTF14hls and η Cari- SNe as time advances, both in the plateau and in the later nae share a common wind driving mechanism but with a dif- radioactive tail phase, but the same is seen in iPTF14hls. ferent timescale, presumably because of the difference in the We therefore use the example of Type IIP SNe to estimate luminosity. The Great Eruption has indeed been suggested whether recombination radiation can be the main method to be a continuum-driven wind rather than a short-term of reprocessing the stellar radiation that makes iPTF14hls mass ejection (e.g., Owocki & Shaviv 2016; Davidson 1987). bright, just as in Type IIP SNe. Both iPTF14hls and η Carinae during the Great Eruption We selected the nearby (7.7 Mpc) and well-observed were well above their Eddington luminosity in the bright SN 1999em as a prototype Type IIP SN. First we estimated phases and an optically-thick continuum-driven wind can the pseudo-bolometric luminosities – or integrated optical be triggered in both cases, resulting in the very high mass- luminosities – of iPTF14hls and SN 1999em in the range − ˚ loss rates and the ionization of the material ejected (e.g., 3500 8500 A (reported in Fig. 6 as blue and red filled cir- B i van Marle et al. 2008). cles, respectively). For iPTF14hls we used g photometry If the iPTF14hls progenitor survived and became as from Arcavi et al. (2017) and Sollerman et al. (2019), after E(B − V) = luminous as η Carinae, iPTF14hls should have been set- applying a correction for Galactic reddening of . tled at around 27.5 magnitudes in optical by now given 0 014 and using a distance of 145 Mpc; for SN 1999em we UBVI the rapid photometric decline rate observed in early 2018 used the photometry reported in Hamuy et al. 2001; (Sollerman et al. 2019). Future observations by Hubble Space Leonard et al. 2002; Elmhamdi et al. 2003; Anderson et al. R Telescope or James Webb Space Telescope will be able to 2014; Faran et al. 2014; Galbany et al. 2016 (the filter was α judge whether there is a surviving massive star at the loca- avoided as it includes the strong H emission), dereddened E(B − V) = tion of iPTF14hls. with 0.0346. From the dereddened spectra of iPTF14hls and SN1999em we evaluated the Hα emission line net luminosities for both sources, i.e. we subtracted the absorbed from the emitted flux. These are reported in Fig. 6 5 DISCUSSION as light blue and orange crosses, respectively. In order to compare the light curves of SN1999em and iPTF14hls we 5.1 Luminosity source of iPTF14hls set t = 0 in iPTF14hls to coincide with the beginning of the first bright phase, which we argue is driven by recombina- 5.1.1 Early phase luminosity tion radiation (as is also required by the fact that the peak We support the interpretation that iPTF14hls was a contin- in luminosity precedes that in the mass-loss rate). uous, long-lasting outflow caused by a sudden re-brightening As it is known that some of the luminosity of a Type IIP

MNRAS 000, 1–7 (2019) iPTF14hls as a variable hyper-wind 5

SN during the plateau is due to nuclear decay energy, we re- 5.2 Energy source moved from the plateau phase (i.e. up to ∼ 110 days) of the integrated optical LC of SN 1999em the integrated optical LC of SN 1987A (referring to a similar wavelength range as We argue that iPTF14hls is a hyper-wind rather than a mass that adopted for SN 1999em, as it is derived from UBVRI eruption. Given the extremely large luminosity far beyond photometry, Hamuy et al. 1988), which did not have an ex- the Eddington luminosity, the wind is probably driven by tended hydrogen envelope and therefore did not display a radiation as suggested for η Carinae and other luminous blue plateau, but rather a delayed (∼ 60 days) luminosity peak variables (e.g., Smith & Owocki 2006). The total estimated 52 caused by the deposition of gamma-rays and positrons and kinetic energy of the outflow is estimated to be 10 erg and the diffusion of the optical radiation created by their ther- it is not clear how the progenitor gained such a huge energy malisation. This subtracted LC is the recombination lumi- to drive the hyper-wind. nosity of SN 1999em at epochs prior to ∼ 110 days (represented by black open circles in Fig. 6), when the plateau Pulsational pair-instability is one possibility dominates the emission. We then computed the recombina- (Woosley et al. 2007; Woosley 2017; Marchant et al. tion luminosity in iPTF14hls assuming that it had the same 2018; Vigna-Gómez et al. 2019). Pulsational pair-instability ratio to Hα as it did in SN 1999em. The result is shown as SNe (PPISNe) have already been related to iPTF14hls open magenta diamonds in Fig. 6. This shows that if our in previous studies (e.g., Woosley 2018). PPISNe from assumption is correct, there is more than enough radiation progenitors having zero-age main-sequence mass of around recombination in iPTF14hls to power the early bright phase 110 M⊙ have pulsation duration of several years (Woosley of the LC. 2017) matching the duration of the continuum-driven wind After the plateau phase, Type IIP SN LCs settle on in iPTF14hls as well as the estimated progenitor mass to a linear phase called the radioactive tail, where the en- (Section 4). A caveat is that the pulsational pair-instability ergy that is deposited from radioactivity is immediately re- may just lead to a violent mass ejection rather than a con- emitted and diffuses out as optical radiation without delay. tinuous mass outflow. Also, the PPISN mass outflows are Therefore, we computed the ratio of the integrated optical predicted to have kinetic energy up to 3 × 1051 erg (Woosley luminosity to the Hα luminosity in SN 1999em after 110 days 2017), which is about a factor 3 smaller than the energy 52 and multiplied it to the Hα luminosity in iPTF14hls, to deduced for iPTF14hls (10 erg). Given the uncertainties check whether similar conditions apply in the two events, in modelling the pulsation (e.g., Takahashi et al. 2016), as the spectral similarity suggests: the increasing prevalence however, it is possible that pulsational instabilities may of Hα emission suggests an increasingly transparent flow, achieve such a high energy. Such an energetic pulsation of where the underlying luminosity (deposited radioactive en- a massive star is predicted to be followed by the formation ergy in the Type IIP SN and stellar luminosity in iPTF14hls) of a black-hole. This may indeed have been the fate of the is reflected immediately in the luminosity of the outflow and progenitor of iPTF14hls. Later observations may reveal diffusion times are negligible. The estimated recombination whether any source is present at the location of iPTF14hls. luminosity of iPTF14hls past 110 days (open magenta triangles in Fig. 6) shows that this approximation is also rea- It has been suggested that the Great Eruption of sonably good. Therefore, iPTF14hls in the early phase can η Carinae is a result of a stellar merger in a close be thought of as obeying radiation laws of a Type IIP SN binary system triggered by a third massive star (e.g., without being a SN: it is a massive outflow which progres- Portegies Zwart & van den Heuvel 2016). In such a case, a sively decreases in mass flow as the driving stellar luminosity strong continuum-driven wind is suggested to occur follow- fades. ing the release of tidal energy dissipation prior to the merger. The variability in the mass-loss rate may be related to the rotational period of the eccentric merging binary system. 5.1.2 Late phase luminosity The total energy available to blow the wind depends on At very late phases, when it faded after the 2-year the binary mass and separation and it can, in principle, 52 bright phase, iPTF14hls showed observational signa- reach 10 erg. The huge difference in the luminosity be- tures of circumstellar interaction (Andrews & Smith 2018; tween iPTF14hls and η Carinae could be due to the much Sollerman et al. 2019). Circumstellar interaction would nat- closer initial binary separation in iPTF14hls. The timescale urally be expected after the extraordinary mass loss from of the wind could also be affected by the initial binary con- the progenitor because the outflow can crash into the pre- figuration and the properties of the third massive star. existing circumstellar matter (e.g., Woosley et al. 2007). The possible brightening of the progenitor of iPTF14hls in Finally, the instability at the surface of massive red su- 1954 (Arcavi et al. 2017) may represent a previous pulsa- pergiants (RSGs) that is triggered when the luminosity-to- tional event and may be responsible for the formation of an mass ratio is high is another possible mechanism to initiate outer dense shell of circumstellar matter. Thus, the source a strong wind (Heger et al. 1997; Yoon & Cantiello 2010). of luminosity of iPTF14hls in the late phase could be inter- Moriya & Langer (2015) investigated the surface instabil- action converting outflow kinetic energy into radiation. The ity of a 150 M⊙ RSG and found that mass-loss enhance- effect of the collision was not significant during the bright ment triggered by the surface instability could last for sev- phase, which was powered by a very high stellar luminos- eral years. However, the estimated mass-loss rates from the −1 ity reprocessed via hydrogen ionization and the ensuing re- pulsation were up to ∼ 0.1 M⊙ yr and did not go beyond −1 combination, as for Type IIP SNe. The interaction became 1 M⊙ yr as estimated for iPTF14hls. However, how the apparent after iPTF14hls faded. pulsation drives the wind is uncertain.

MNRAS 000, 1–7 (2019) 6 T. J. Moriya et al.

Figure 6. Recombination luminosity (magenta open symbols) of iPTF14hls, estimated based on the behaviour of Type IIP SNe. The zero time of iPTF14hls is set to coincide with the start of the bright phase. All optical LCs (ﬁlled circles) were computed in the range 3500 − 8500 A˚ (as described in the text, where we call them ”integrated” optical LCs). The Hα luminosities of iPTF14hls and SN 1999em, estimated from the spectra as detailed in the text, are reported as light blue and orange crosses, respectively. The optical LC of SN 1987A (green circles) was rescaled to match the 56Co radioactive tail of SN 1999em; the diﬀerence between optical luminosities of SN 1999em and normalized SN 1987A is reported as black open diamonds; the recombination luminosity of iPTF14hls, computed from its Hα luminosity and from the recombination and Hα luminosities of SN 1999em, is represented by magenta diamonds and triangles before and after 110 days, respectively (see text).

6 CONCLUSIONS ACKNOWLEDGEMENTS PAM and EP are grateful for kind hospitality at NAOJ during completion of this work. This work was supported by NAOJ Research Coordination Commit- tee, NINS, Grant Number 19FS-0506 and 19FS-0507. We have proposed that iPTF14hls is a continuous outflow TJM is supported by the Grants-in-Aid for Scien- like a stellar wind rather than a mass ejection like a SN, and tific Research of the Japan Society for the Promo- unveiled its mass-loss rate history. The slow change in its tion of Science (JP17H02864, JP18K13585). We made spectroscopic properties over 2 years is naturally explained use of WISeREP (https://wiserep.weizmann.ac.il, by such a continuous wind. The mass-loss rates exceed a Yaron & Gal-Yam 2012) and the Open Supernova Catalog −1 few M⊙ yr during the bright phase of iPTF14hls and tem- (https://sne.space/, Guillochon et al. 2017). −1 porarily become as high as 10 M⊙ yr . This hyper-wind is a super-Eddington continuum-driven wind in which the hydrogen recombination likely powers the early bright phase REFERENCES of iPTF14hls. We have shown that η Carinae during the Anderson J. P., et al., 2014, ApJ, 786, 67 Great Eruption share similar properties to iPTF14hls in- Andrews J. E., Smith N., 2018, MNRAS, 477, 74 cluding their mass-loss rates and LC shapes. The two mas- Arcavi I., et al., 2017, Nature, 551, 210 sive stars may have similar mass (around 100 M⊙) as well. Bennett P. D., 2010, in Leitherer C., Bennett P. D., Morris P. W., The exact luminosity source of iPTF14hls is not clear, but Van Loon J. T., eds, Astronomical Society of the Pacific Con- PPISNe or massive close binary mergers may be related as ference Series Vol. 425, Hot and Cool: Bridging Gaps in Mas- suggested for the Great Eruption of η Carinae. sive Star Evolution. p. 181 (arXiv:1004.1853)

MNRAS 000, 1–7 (2019) iPTF14hls as a variable hyper-wind 7

Bersten M. C., Benvenuto O., Hamuy M., 2011, ApJ, 729, 61 This paper has been typeset from a TEX/LATEX file prepared by Chugai N. N., 2018, Astronomy Letters, 44, 370 the author. Davidson K., 1987, ApJ, 317, 760 Davidson K., Humphreys R. M., 1997, ARA&A, 35, 1 Dessart L., 2018, A&A, 610, L10 Eastman R. G., Schmidt B. P., Kirshner R., 1996, ApJ, 466, 911 Ekström S., et al., 2012, A&A, 537, A146 Elmhamdi A., et al., 2003, MNRAS, 338, 939 Faran T., et al., 2014, MNRAS, 442, 844 Fernández-Lajús E., et al., 2009, A&A, 493, 1093 Filippenko A. V., 1997, ARA&A, 35, 309 Galbany L., et al., 2016, AJ, 151, 33 Georgy C., et al., 2013, A&A, 558, A103 Gofman R. A., Soker N., 2019, arXiv e-prints, p. arXiv:1905.05573 Guillochon J., Parrent J., Kelley L. Z., Margutti R., 2017, ApJ, 835, 64 Hamuy M., Suntzeff N. B., Gonzalez R., Martin G., 1988, AJ, 95, 63 Hamuy M., et al., 2001, ApJ, 558, 615 Heger A., Jeannin L., Langer N., Baraffe I., 1997, A&A, 327, 224 Kasen D., Woosley S. E., 2009, ApJ, 703, 2205 Leonard D. C., et al., 2002, PASP, 114, 35 Liu T., Song C.-Y., Yi T., Gu W.-M., Wang X.-F., 2019, Journal of High Energy Astrophysics, 22, 5 Marchant P., Renzo M., Farmer R., Pappas K. M. W., Taam R. E., de Mink S., Kalogera V., 2018, arXiv e-prints, p. arXiv:1810.13412 Mazzali P. A., Deng J., Maeda K., Nomoto K., Filippenko A. V., Matheson T., 2004, ApJ, 614, 858 Moriya T. J., Langer N., 2015, A&A, 573, A18 Morris P. W., et al., 1999, Nature, 402, 502 Owocki S. P., Shaviv N. J., 2016, MNRAS, 462, 345 Pejcha O., Prieto J. L., 2015, ApJ, 806, 225 Portegies Zwart S. F., van den Heuvel E. P. J., 2016, MNRAS, 456, 3401 Prieto J. L., et al., 2014, ApJ, 787, L8 Rest A., et al., 2012, Nature, 482, 375 Smith N., 2006, ApJ, 644, 1151 Smith N., 2013, MNRAS, 429, 2366 Smith N., Owocki S. P., 2006, ApJ, 645, L45 Smith N., et al., 2018, MNRAS, 480, 1457 Soker N., Gilkis A., 2018, MNRAS, 475, 1198 Sollerman J., et al., 2019, A&A, 621, A30 Szécsi D., Langer N., Yoon S.-C., Sanyal D., de Mink S., Evans C. J., Dermine T., 2015, A&A, 581, A15 Takahashi K., Yoshida T., Umeda H., Sumiyoshi K., Yamada S., 2016, MNRAS, 456, 1320 Vigna-Gómez A., Justham S., Mandel I., de Mink S. E., Podsi- adlowski P., 2019, ApJ, 876, L29 Wang L. J., et al., 2018, ApJ, 865, 95 Woosley S. E., 2017, ApJ, 836, 244 Woosley S. E., 2018, ApJ, 863, 105 Woosley S. E., Blinnikov S., Heger A., 2007, Nature, 450, 390 Yaron O., Gal-Yam A., 2012, PASP, 124, 668 Yoon S.-C., Cantiello M., 2010, ApJ, 717, L62 Yoon S. C., Dierks A., Langer N., 2012, A&A, 542, A113 van Marle A. J., Owocki S. P., Shaviv N. J., 2008, MNRAS, 389, 1353

APPENDIX A: RESULTS OF THE BLACKBODY FITTING We present all the results of the blackbody ﬁtting used in the paper in Fig. A1.

MNRAS 000, 1–7 (2019) 8 T. J. Moriya et al.

3 3 3 8000 K, 1.7 x 1015 cm 2015−01−08 (#1) 7500 K, 2.05 x 1015 cm 2015−02−11 (#2) 7250 K, 2.1 x 1015 cm 2015−03−04 (#3) 7500 K, 1.9 x 1015 cm 7000 K, 2.25 x 1015 cm 6750 K, 2.35 x 1015 cm 2.5 7000 K, 2.1 x 1015 cm 2.5 6500 K, 2.45 x 1015 cm 2.5 6250 K, 2.6 x 1015 cm

) 2 ) 2 ) 2 −1 −1 −1 Å Å Å