Luminous blue variables and the fates of very massive stars

Abstract

Luminous blue variables (LBVs) had long been considered massive stars in transition to the Wolf–Rayet (WR) phase, so their identification as progenitors of some peculiar supernovae (SNe) was surprising. More recently, environment statistics of LBVs show that most of them cannot be in transition to the WR phase after all, because LBVs are more isolated than allowed in this scenario. Additionally, the high-mass H shells around luminous SNe IIn require that some very massive stars above 40 M_⊙ die without shedding their H envelopes, and the precursor outbursts are a challenge for understanding the final burning sequences leading to core collapse. Recent evidence suggests a clear continuum in pre-SN mass loss from super-luminous SNe IIn, to regular SNe IIn, to SNe II-L and II-P, whereas most stripped-envelope SNe seem to arise from a separate channel of lower-mass binary stars rather than massive WR stars.

This article is part of the themed issue ‘Bridging the gap: from massive stars to supernovae’.

1. Introduction: the traditional view of luminous blue variables

The luminous blue variable (LBV) phase of evolution, punctuated by eruptive mass loss, is thought to be important in the mass-loss evolution of massive stars [1,2]. However, the physics of their instability and its influence on stellar evolution have been challenging to understand quantitatively. In particular, open questions concern the total amount of mass lost in a typical eruption and how it scales with initial mass, the duty cycle and number of repeated LBV giant eruptions, and whether all stars—or only a special subset—suffer the LBV giant eruptions.

The nearby class of LBVs defined in the Milky Way and the Large and Small Magellanic Clouds (LMC/SMC) is thought to be related to some extragalactic non-supernova (SN) transients that may be examples of LBV giant eruptions [3,4]. The episodic mass loss of LBVs has also been a reference point for interpreting the dense circumstellar material (CSM) around Type IIn supernovae (SNe IIn), which is thought to be indicative of extreme pre-SN eruptions (see review by Smith [1], and references therein).

(a). Traditional hallmarks of luminous blue variables

The diverse collection of objects known collectively as LBVs was first proposed by Conti [5], and the standard interpretation of these stars and their role in evolution was established through the 1980s and 1990s (e.g. [6]). The central idea that emerged is that LBVs are massive single stars in a transitional phase between the main sequence and He burning. The review by Humphreys & Davidson [7] (HD94 hereafter) provides a summary of the traditional view of LBVs, which has a few defining characteristics:

• (1) LBVs are presumed to represent a brief transitional phase in the evolution of the most massive stars, between the main-sequence O-type stars and the H-deficient Wolf–Rayet (WR) stars. A typical monotonic evolutionary scheme for a single star of 60–100 M_⊙ initially is

  

  In this scenario, LBVs are a very brief (few ×10⁴ years) transition when the remaining H envelope is removed by winds and eruptive mass loss, propelling the star to become a WR star and to later explode as a hydrogen-free stripped-envelope SN. Because LBVs have already lost some mass in their main-sequence phase, and because their core luminosity has increased, the L/M ratio is high, and they approach the classical Eddington limit. This proximity to the Eddington limit, combined with Fe opacity, leads to a violent instability in the envelope that somehow triggers runaway mass loss. The heavy mass loss of LBVs, in the single-star view, is essential to remove the H envelope to form WR stars. LBVs cannot be immediate SN progenitors in this view, because they are just about to begin core He burning.

• (2) Observed S Dor-type variations, as exemplified most clearly by AG Car in the Milky Way and R127 in the LMC [8–10], are temperature variations at constant L_Bol. The cool (visibly bright) state is caused by a pseudo-photosphere that develops in an optically thick wind, which occurs at the same temperature regardless of luminosity [7,11]. In their quiescent (hot) phases, all LBVs reside on the diagonal ‘S Doradus instability strip’ [12] on the Hertzsprung–Russell (HR) diagram (see figure 1, from Smith et al. [13]).

• (3) The strong mass loss of LBVs halts their redward evolution, preventing them from becoming red supergiants (RSG). This explains the observed absence of high-luminosity RSGs above as due to instability and mass loss in evolved massive single stars.

• (4) Some LBVs suffer dramatic ‘giant eruptions’ where huge amounts of mass can be lost. Although there is a formalism to explain the mechanism of this mass loss with super-Eddington continuum-driven winds [14], the underlying reason why single stars suddenly exceed the classical Eddington limit by factors of 5–10 has remained unexplained. Nevertheless, this tremendous mass loss is observed and is likely to be important in the evolution of massive stars.

Figure 1.

LBVs on the HR diagram, from Smith et al. [13], showing the standard locations of the quiescent S Doradus instability strip and the constant temperature strip of LBVs at their maximum light phase in S Dor variations.

(b). Problems with the traditional view

There is mounting evidence that this fairly well-defined picture of LBVs may be unravelling, since several of these hallmarks have not stood the test of time. Continued study has shown that rigid imposition of definitions of LBVs [7] has not been supported by detailed analysis on several fronts, and these definitions would in fact exclude most LBVs from the class. In particular, the important conjecture that S Dor variations are caused by optically thick winds at constant L_Bol may be wrong, and steady super-Eddington winds may not be the correct or dominant driving mechanism in many giant LBV eruptions (see review by Smith [1]).

Studies with quantitative spectroscopy [15–18] have disproven the conjecture that S Doradus brightening events are caused by developing pseudo-photospheres in optically thick winds, as suggested initially by Davidson [11] and reviewed by HD94 [7]. The mass-loss rates of S Dor maxima are not high enough to make such large pseudo-photospheres, and so they are more likely to be caused by envelope inflation or pulsation [19,20]. This inflation may arise due to the subsurface Fe opacity peak. Moreover, bolometric luminosities during S Doradus eruptions are not really constant [17], probably because energy from the radiation field is put into work that must be done to inflate the envelope. Similarly, the idea that giant-eruption maxima are caused by pseudo-photospheres in opaque super-Eddington winds is challenged by light-echo spectra of η Carinae [21,22] (although see [23]), by detailed analysis of the ejecta around η Car that are better matched by explosive models [24–26], and by the fact that many extragalactic giant LBV eruptions are relatively hot at peak luminosity rather than cool [3,27,57].

Two more recent results create severe problems for the traditional view of LBVs: (i) some core-collapse SNe (especially Type IIn) appear to have progenitors that have been identified as LBVs. This is prohibited in the standard view of LBVs as massive stars in transition to core He burning. (ii) The environments of LBVs (and some SNe IIn) are remarkably isolated. As discussed recently by Smith & Tombleson [28], the central paradigm of the role that LBVs play in stellar evolution and their connection to stars with the highest initial masses is probably incorrect, because their isolation from massive O-type stars dictates that they are largely products of binary evolution and not a transitional state in the lives of the most massive stars. There are many ways in which violent binary interaction may be important for understanding LBVs and related transients [1,28–31], but these are not part of the traditional view. These two aspects are discussed in more detail below.

2. The isolation of luminous blue variables

(a). Separations from O-type stars

In a recent study, Smith & Tombleson [28] (ST15 hereafter) analysed the projected spatial distribution of LBVs on the sky and found them to be surprisingly isolated from O-type stars. One of the figures from ST15 is reproduced here in figure 2, which is a plot of the cumulative distribution of projected separation on the sky to any nearby O-type star for various classes of stars—early/mid/late O, LBVs, WR (WNH, WN, WC), RSGs and B[e] supergiants. This is for targets in the LMC/SMC, which avoid the large distance uncertainty and potentially large line-of-sight reddening for Milky Way targets (although it is harder to quantify the statistics, LBVs in the Milky Way seem similarly isolated and tend to avoid young clusters [28]).

Figure 2.

Cumulative distribution plot illustrating the differing degrees of isolation among various classes of massive stars in the LMC, from Smith & Tombleson [28]. Classes of objects that are more clustered with young O-type stars appear farther to the left. The relative isolation is represented here by the distribution of distances to the nearest O-type star of any spectral type or luminosity class. The sample includes all O-type stars, WR stars, LBVs, sgB[e] stars and RSGs within a 10^° projected radius of 30 Dor (except for the SMC stars). The O-type stars are further subdivided into early (O5 and earlier; green), mid (O6+O7, orange) and late (O8+O9, cyan) subtypes. For WR stars, we show WC stars (magenta), a collection of all WN stars including WNH stars (solid blue), as well as WN stars without WNH (dashed blue). The mustard dot-dashed line is for all H-poor WR stars (WN+WC). For LBVs, we include both LMC and SMC targets (the separation of the three SMC targets has been multiplied by 1.2 to adjust for the difference in distance), and we include both confirmed and candidate LBVs. RSGs (red) are stars with spectral types later than K3 and luminosity classes of I, Ia or Iab, and the supergiant B[e] sample (cyan dashed) is from the literature [32,33]. See Smith & Tombleson [28] for further details. (Online version in colour.)

ST15 [28] concluded that the results were inconsistent with expectations for the traditional picture of LBVs in single-star evolution [7], wherein LBVs are descended from very massive main-sequence O-type stars, and where LBVs are the key agent that provides the required mass loss to drive them into the WR phase. In particular, ST15 [28] found that LBVs were far from early O-type stars, and were in fact more dispersed from O-type stars on the sky than WR stars, making it impossible for the observed population of LBVs to turn into the observed population of WR stars. They concluded that, instead of the traditional view wherein LBVs descend from the most massive single stars, many LBVs are likely to be the product of binary evolution, where stars are spun-up, chemically enriched, and rejuvenated by mass transfer after a delay, and possibly kicked by their companion's SN explosion [28]. In this view, LBVs are essentially evolved massive blue stragglers.

What about selection effects? Most selection effects go the wrong way, if one is trying to resurrect the traditional view. There is no conceivable bias that would prevent one from recognizing bright LBVs in known star clusters; for a young cluster, LBVs in the traditional view have finished core H burning and are somewhat more luminous than O-type stars that still remain on the main sequence, but LBVs have moved to the right on the HR diagram. At these cooler temperatures, they vastly outshine their O-type siblings at visible wavelengths. It therefore seems impossible that we have missed a large number of LBVs in massive young clusters. While there are a couple of LBVs in clusters, these clusters are older than they should be for the LBV's effective single-star mass [28], so the avoidance of young clusters by a majority of LBVs is a robust result. It seems highly unlikely that any unobscured massive young clusters surrounding these LBVs have escaped detection, since these LBVs have been studied with deep high-contrast imaging to search for nebulosity. Moreover, we do indeed see late O-type stars around some LBVs, but it is the early O-type stars (i.e. the more easily detected ones) that are missing. The lack of brighter early O-type stars cannot be a detection bias when fainter late O-type stars are clearly detected. The population of late O-type dwarfs in the field may indeed be incomplete, but we can ask what happens to figure 2 if there are many O-type stars scattered throughout the field in the LMC. In that case, everything will shift slightly to the left in figure 2; this will not rectify the problem of LBVs being more isolated than WR stars.

(b). Some back and forth

Following the paper by ST15 [28], an alternative view of LBV locations was presented by Humphreys et al. [34] (H16 hereafter). H16 argued that if, from the original 19 LBVs in ST15, one chooses the three with the smallest separations to O-type stars and remakes the same cumulative distribution plot as in figure 2, one finds that LBVs are not as isolated as ST15 claimed. H16 claimed that this revalidates the traditional view of LBVs. However, Smith [35] noted several problems with the H16 argument. First, the evidence based on environments is essentially a statistical case. One cannot select the three objects at the tail end of a cumulative distribution and remake the same comparison. Second, even if one does this, there is no statistical significance. Moreover, the argument made by H16 [34] that including candidate LBVs corrupts the sample is false, since there is no statistical difference in their environments [35]. Third, even if one were to accept the selection criteria of H16, they interpreted the result incorrectly. H16 suggested that the three most luminous LBVs (which have implied initial masses of 60–100 M_⊙) have distributions on the sky similar to O-type stars; they took this as consistent with the traditional view. However, Smith [35] pointed out that these are late O-type dwarfs (O8 and O9 V), which have initial masses that are far lower than for those LBVs. The implication is actually the opposite of that claimed by H16. There are more points to the debate, and we refer the reader to H16 [34] and Smith [35] for more details.

More recently, Smith & Stassun [36] have examined the implications of revised distances to some Galactic LBVs as listed in the Gaia DR1. A surprising result was that the canonical LBV star AG Carinae is actually at a much closer distance than previously thought, which lowers its true luminosity by about a factor of 10. This would move it way off the S Doradus instability strip on the HR diagram (figure 1). This is potentially a big problem for the traditional view of LBVs as massive single stars in transition to the WR phase. While the Gaia DR1 distances to these LBVs are still considered tentative [36], one may anticipate that the second data release from Gaia might shake things up even more.

(c). An alternative evolutionary view

The fact that LBVs are more isolated than WR stars, plus the fact that some stars appear to remain in an LBV phase until death as SNe IIn, suggests an entirely different picture. Instead of the monotonic evolutionary scheme for single stars, the spatial distributions of LBVs suggest that massive star populations and especially LBVs are dominated by bifurcated evolutionary trajectories:

In this scenario, O-type stars evolve off the main sequence, and through binary interaction the majority of massive stars either (i) lose their H envelope through mass transfer to a companion, become WR stars and die as stripped-envelope SNe, or (ii) gain mass from a companion, become blue supergiants (BSGs), sgB[e]s and LBVs, and then retain their H envelopes until they die as SNe IIn or other varieties of SNe II. These mass gainers would be overluminous compared to surrounding stars, they would get spun-up, and probably enriched in products of CNO burning. These are all observed traits of LBVs. Similar to being mass gainers, many of the same observed results would occur if LBVs are the products of binary mergers [37]. This sort of mass-gainer or merger scenario for LBVs was mentioned early on by Kenyon & Gallagher [38], but seemed to be disfavoured or largely ignored until recently.

Figure 3 shows an HR diagram of LBVs with single-star evolutionary tracks, as well as an example of a binary evolutionary track that exemplifies the bifurcated scheme mentioned above.

Figure 3.

HR diagram (similar to figure 1 but with an expanded L scale) comparing LBVs and some model evolutionary tracks. The LBVs plotted here are the same as in Smith et al. [13], with some updates as noted by Smith & Tombleson [28]. The evolutionary tracks are from fig. 4a in Langer & Kudritzki [39], although as they note, the single-star tracks up to 60 M_⊙ are originally from Brott et al. [40]. The blue dotted tracks are single-star evolution tracks for Z_⊙ and an initial rotation speed of 100 km s⁻¹. The solid blue and red tracks are for a binary system that undergoes Roche lobe overflow (RLOF)on the main sequence, with an initially 16 M_⊙ mass donor and an initially 14 M_⊙ mass gainer. This illustrates just one example of how a star that initially had a relatively low mass can end up as a much more luminous star that could resemble the low-luminosity LBVs; binary systems with higher initial masses might obviously populate the more extreme LBVs in a similar manner [37]. For reference, the approximate locations of supergiant B[e] stars and BSGs are shown, as are the progenitor of SN 1987A and the putative companion of SN 1993J's progenitor. (Online version in colour.)

The key recognition here is that the stars we call LBVs—which may currently sit along an evolutionary track of a massive single star with a certain apparent initial mass—might instead arise from stars that had roughly half their implied initial mass, but have gained mass or merged through binary interaction. If this mass gaining or merging occurs late in life, an LBV could have had a much longer lifetime than we would naively expect from comparing its location on the HR diagram to single-star models. This could explain the apparent discrepancy between the high-luminosity LBVs and the lack of O-type stars of the expected turn-off mass in their vicinity. Mass gainers in binary evolution may also get a kick when their companion explodes, which may further their isolation. The kick may be quite mild (perhaps only approx. 10 km s⁻¹ or so), but that motion can last for an additional 1–2 Myr to send the star into the middle of nowhere. Whether such SN kicks are needed to explain the observed statistics remains an open question.

(d). Winds or explosions?

In addition to their characteristic S Doradus-type variability, LBVs are also thought to suffer giant eruptions. These giant eruptions of LBVs may shed large amounts of mass in a single event that lasts only a few years, unlike S Dor episodes, which are not major mass-loss events. An extreme case is the nineteenth-century Great Eruption of η Car, which shed more than 10 M_⊙ in only a decade or so [24,41], although there are also a number of less extreme events. LBV giant eruptions are also associated with some extragalactic events that go by various names, including ‘SN impostors’ [3,4,57]. Depending on how many of these eruptions occur for a given star, these giant eruptions could be the dominant mode of mass loss for some very massive stars that cannot become RSGs [2].

The traditional physical explanation for LBV giant eruptions is that they are driven by super-Eddington winds [11,14]. Even in quiescence, LBVs are thought to be close to the classical Eddington limit. During giant eruption episodes, their luminosity is observed to increase dramatically, which is thought to push them beyond the Eddington limit and initiate strong mass loss in a continuum-driven wind [14]. The trigger of this luminosity increase and the source of the extra energy has not been explained. Qualitatively, the narrow-lined spectra with strong H Balmer lines seen in SN impostors seem consistent with winds, although the spectra of SN impostors have not been modelled quantitatively in the way that quiescent LBV spectra have been [16–18,42].

There is growing observational evidence, however, that some LBV giant eruptions may be explosive in nature—i.e. driven by a shock wave, instead of a relatively steady radiation-driven wind. Studies of η Car show a very high ratio of kinetic energy to radiated energy, very thin walls in its bipolar nebula that seem to be consistent with shock compression, and very high speeds in the outer ejecta that suggest the presence of a strong blast wave [24–26]. Radiation-driven winds have a harder time accounting for the combination of very fast speeds while also having most of the mass in a thin shell moving relatively slowly. Although we normally think of an explosion as occurring on a dynamical time, Smith [43] showed that the decade-long plateau of η Car's Great Eruption could be explained by a shock wave ploughing through dense CSM, as in the standard picture for SNe IIn. More recently, studies of the SN impostor UGC 2773-OT showed clear signatures of shock excitation in its spectrum during its decade-long eruption that closely resembled η Car [44]. While the continuum luminosity and colour temperature stayed relatively constant (and cool) over many years, strong changes in ionization were seen in the spectrum that indicated shock excitation [44]. Figure 4 shows a comparison of the early and late-time spectra of UGC 2773-OT, where a cooler absorption-dominated spectrum with slow outflow speeds was seen to transition into an emission-line-dominated spectrum with faster outflow speeds and much stronger Hα. The difference between these two (i.e. the residual of a subtraction) closely resembles the spectrum of a core-collapse SN IIn at late times, where the emission is dominated by a strong shock running through distant CSM. The Hα line profiles of UGC 2773-OT show a peculiar blue bump, which may be indicative of a bipolar structure (figure 5).

Figure 4.

Evidence for shock powering of LBV giant eruptions, from Smith et al. [44]. (a) A comparison of the early (day 34) and late-time (day 1836) spectra for UGC 2773-OT, highlighting the excess blue emission. Both spectra are normalized to the red continuum level, and in fact the red spectra appear very similar except for Hα and Ca⁺. In the blue,there is a clear excess of line emission at late times. (b) Isolates this excess blue emission by subtracting the normalized day 34 spectrum from the normalized day 1836 spectrum. The grey plot is the residual, and the black plot is a smoothed version of the residual emission. This residual is compared to the ‘blue pseudo-continuum’ seen in SN 2005ip (red) on day 905 [45], which is characteristic of the forest of blue emission lines seen in interacting SNe IIn. This indicates that the excess blue line emission in UGC 2773-OT is most likely powered by a shock interacting with CSM. (Online version in colour.)

Figure 5.

Comparison between the asymmetric Hα line profile observed in UGC 2773-OT and the H₂ emission from the Homunculus nebula around η Carinae. (a) The two-dimensional long-slit spectrum of H₂ S(1-0) 2.122 μm emission from the Homunculus [25]. Panel (b) compares the lineprofile of this H₂ emission integrated along the slit (meant to mimic the integrated H₂ line profile observed for the whole Homunculus nebula). This is the thick orange curve, which is compared to the Hα profile observed in UGC 2773-OT (black). From Smith et al. [44]. (Online version in colour.)

Whether explosive models are the unique explanation for LBV giant eruptions remains unclear—something needs to provide the slow CSM into which the shock expands, and that could be a precursor super-Eddington wind. Moreover, a time-dependent wind-driven model, wherein a fast wind overtakes a previous slow wind [46], may also lead to similar signatures of shocks in the spectra and resulting ejecta.

3. Luminous blue variables as SN progenitors

Perhaps, the most severe discrepancy with the traditional picture of LBVs comes from their end fate. If LBVs are massive stars that have just finished their main-sequence evolution and are in transition to their core He burning phase as H-free WR stars, then we should never expect a star to undergo core collapse during the LBV phase. However, observational studies in the past decade have revealed several cases where LBV-like progenitors of SNe are consistent with the data, and in some cases, LBVs are the only known type of star that could be a plausible progenitor. The different lines of evidence are discussed below.

(a). Dense circumstellar material in SNe IIn

SNe IIn are powered largely by CSM interaction, and for a given SN explosion energy, more massive CSM shells surrounding the star will lead to more luminous SNe. The first case (and one of the most extreme examples) to highlight the clear need for an eruptive LBV-like progenitor was the super-luminous SN 2006gy [47,48]. In this case, a huge mass of H-rich CSM of order 20 M_⊙ was ejected at speeds of a few hundred kilometres per second just 8 years prior to the SN [49,50]. These properties match the nineteenth-century eruption of η Car remarkably well. Many other SNe IIn are not as extreme as SN2006gy, but LBV-like mass loss is nevertheless implied for many of them. From the observed luminosity and CSM speed measured in spectra, one can deduce the progenitor mass-loss rate that is needed to account for the observed CSM interaction luminosity. To make CSM interaction luminosity compete with the normal photospheric luminosity of the SN, very dense CSM is required. Figure 6 (from [51]) shows the required pre-SN mass-loss rates derived for a number of interacting SNe, as compared to mass-loss rates of known types of stars (from [1]). While some of the lower-luminosity SNe IIn could be explained by the dense and slow winds of the most extreme RSGs and yellow hypergiants, many interacting SNe seem to be consistent with CSM that has been created by LBV-like eruptive mass loss. This does not necessarily mean that all LBVs are poised to explode as SNe within the next decade—in fact, many LBVs have massive shells that are centuries or millennia old—but it does mean that LBVs may offer a valuable window into the eruptive instability of SN IIn progenitors.

Figure 6.

Plot of mass-loss rate as a function of wind velocity, comparing values for interacting SNe to those of known types of stars. The solid coloured regions correspond to values for various types of evolved massive stars taken from Smith [1], corresponding to asymptotic giant branch (AGB) and super-AGB stars, red supergiants (RSGs) and extreme RSGS (eRSG), yellow supergiants (YSG), yellow hypergiants (YHG), LBV winds and LBV giant eruptions, binary Roche-lobe overflow (RLOF), luminous WN stars with hydrogen (WNH) and H-free WN and WC Wolf–Rayet (WR) stars. A few individual stars with well-determinedvery high mass-loss rates are shown with circles (VY CMa, IRC+10420, η Car's eruptions and P Cyg's eruption). Also shown with ‘X's are some representative examples of SNe IIn (and one SN Ibn) that have observational estimates of the pre-shock CSM speed from the narrow emission component as measured in moderately high-resolution spectra as well as estimates of the progenitor mass loss required, taken from the literature. The diagonal lines are wind density parameters () of 5×10¹⁶ and 5×10¹⁵ g cm⁻¹, which are typically the lowest wind densities required to make a SN IIn. Note that in some cases, an observationally derived value for the mass of the CSM has been converted to a mass-loss rate with a rough estimate of the time elapsed since ejection (figure from [51]). (Online version in colour.)

(b). Variable speed and density of circumstellar material

Rather than arguments based on the amount of CSM mass, LBVs have also been identified as possible SN progenitors based on variability in the CSM speed or density structure. Trundle et al. [52] noted that the two component absorption troughs in some SNe IIn resemble features seen in line profiles of the LBV AG Carinae, where the two velocities are thought to arise from the star crossing the bistability jump. The modulation of the radio light curves of some SNe IIb suggests density variations consistent with what may arise from changes in wind properties during the S Doradus cycles of LBVs [53]. This led Kotak & Vink [53] to make one of the early suggestions that LBVs may explode as core-collapse SNe.

(c). Directly detected SN progenitors

The most straightforward way to connect a SN to its progenitor is, in principle, to directly detect the progenitor star before the SN. This has been done for a number of SNe using archival pre-explosion data, reviewed at length elsewhere in this issue. There have also been a few reports for interacting SNe, which seem to point towards LBVs as likely progenitors. One of the least ambiguous cases was SN 2005gl, where a very luminous star akin to P Cygni was suggested [54] and where the progenitor faded after the event [55]. The historical case of SN 1961V is another strong case [3,56], where an extremely luminous and variable progenitor was detected for decades before the SN, and where that source subsequently faded by more than 5 mag at visible wavelengths. It is also very faint in the thermal IR, confirming that the very luminous progenitor is gone rather than surviving as dust-enshrouded [56]. (SN 1961V is often overlooked because it had previously been interpreted as a SN impostor, not a true SN.) Another example is SN 2010jl, where a very luminous and blue source (if a direct detection of the progenitor) or very young environment (if a host cluster) both imply a massive progenitor [57]. Fox et al. [58] recently showed that the candidate blue source is a young host cluster and not the progenitor star. SN 2009ip has a very luminous progenitor detection consistent with a 50–80 M_⊙ star [59,60]. (Note that here the lower value of 50 M_⊙ is derived assuming that the progenitor was in an eruptive state, whereas the higher mass estimate of 80 M_⊙ derives from the necessary bolometric correction if the star was quiescent and hot.) At this meeting we have heard reports from Elias-Rosa et al. of yet another example, SN 2015bh, which appears to be very similar to SN 2009ip. Having multi-filter (colour) information in pre-explosion data is usually hard to come by, but the very high luminosities of these progenitors are more luminous than the most luminous known RSGs, and are far too bright to be WR stars. LBVs seem like the most straightforward interpretation of these progenitors.

Placing the detected source on an HR diagram to deduce its initial mass, however, is not so straightforward because of one caveat—pre-SN eruptions are inferred to account for their dense CSM, so they might have been in an outburst state at the moment when the pre-SN archival data were obtained (which would, of course, still point to an LBV-like progenitor). In fact, some cases of pre-SN eruptions have been detected, as described next.

(d). Directly detected precursor eruptions

Observations have already confirmed the idea that SNe with strong CSM interaction may have suffered violent LBV-like eruptive instability in the few years before core collapse, because there are several cases where the precursor eruptions were directly detected as dramatic photometric variability. The first documented case was the outburst 2 years before SN 2006jc [61,62], which was unusual in that it was an H-poor event (Type Ibn). SN 1961V, mentioned above, also showed slow S Dor-like variability in the years before its SN (see [3]), although for many years it was not considered as a SN. The Type IIn-P event SN 2011ht also showed pre-explosion outburst behaviour [63]. The most dramatic and well-studied case so far has been SN 2009ip. In addition to its progenitor star mentioned above, it showed variability consistent with an S Dor cycle [59] as well as several dramatic pre-SN outbursts [59,64,65], and high-quality spectra of its pre-SN eruptions were obtained [59,60]. (Again, a talk at this meeting by Elias-Rosa and collaborators discussed the recent event SN 2015bh, where a sequence of precursor outbursts similar to those in SN 2009ip appear to be detected.) Two other studies have examined a larger number of events to search for precursor eruptions in PTF data [66] and Lick Observatory Supernova Search data [67]. SN 2010mc had a double-peaked light curve, although it is a debated case of a precursor eruption because the initial fainter peak may have been the SN itself followed by a rebrightening due to CSM interaction [64,68]. Blue progenitors (like LBVs) may be expected to yield relatively faint SNe initially, until CSM interaction kicks in to boost the luminosity. CSM interaction may turn on after a short delay, as in SN 2009ip [68,69].

4. Conclusion

This contribution has discussed several ways in which our standard view of LBVs is changing. Their role in stellar evolution may be very different from the standard view that has been developed by comparison with single-star evolutionary models with strong mass loss. The main ways in which LBVs contradict the traditional view of ‘massive stars in transition’ are summarized briefly as follows:

• (1) Their S Dor eruptions are not caused by pseudo-photospheres resulting from increased mass loss, but rather, these events appear to be more consistent with envelope inflation.

• (2) Their giant eruptions are not necessarily wind-driven events. Many LBV giant eruptions and SN impostors appear to be more consistent with explosive events, although highly variable winds are difficult to rule out.

• (3) LBVs appear too isolated from O stars to be the immediate descendants of the most massive early O-type stars. In particular, they are farther from O stars than WR stars.

• (4) Some LBVs appear to be exploding as core-collapse SNe, which could not be the case if they are in transition to the beginning of He burning.

Instead of the traditional single-star view, a picture is emerging wherein LBVs are mainly the product of binary evolution, as part of a bifurcated evolutionary path of massive stars. Their isolated environments suggest that they are evolved massive blue stragglers that have been rejuvenated as the result of mergers or mass gainers in binary evolution [28]. If this is true, then an interesting question concerns the source of their peculiar eruptive instability. Is the LBV instability a physical result of that binary evolution, perhaps linked to the extra angular momentum associated with a merger or mass transfer episode? Are giant eruptions of LBVs simply the merger events caught in action?

Moreover, removing LBVs from their standard role as the agent that removes the H envelope in massive stars raises further questions. It may suggest that WR stars and stripped-envelope SNe are largely the result of mass transfer in binaries, where these stars are the descendants of the initially more massive of the two stars (the mass donor). This would allow a much wider range of initial masses to give rise to SNe Ibc. On the flip side, the apparent continuum in the degree of CSM interaction from SNe II-P to II-L and up to IIn [70] may be determined by a combination of the initial masses and the total mass accreted by the mass gainers in binaries [71]. Without the participation of LBVs, the fate of the most massive single stars with initial masses above 40 M_⊙ (if there are any single stars at these masses) remains an open question.

Data accessibility

This article has no supporting data.

Competing interests

The authors declare that there are no competing interests.

Funding

Support for NS was provided by NSF grants AST-131221 and AST-151559 and by a Scialog grant from the Research Corporation for Science Advancement.