Luminous Supernovae

See allHide authors and affiliations

Science  24 Aug 2012:
Vol. 337, Issue 6097, pp. 927-932
DOI: 10.1126/science.1203601


Supernovae, the luminous explosions of stars, have been observed since antiquity. However, various examples of superluminous supernovae (SLSNe; luminosities >7 × 1043 ergs per second) have only recently been documented. From the accumulated evidence, SLSNe can be classified as radioactively powered (SLSN-R), hydrogen-rich (SLSN-II), and hydrogen-poor (SLSN-I, the most luminous class). The SLSN-II and SLSN-I classes are more common, whereas the SLSN-R class is better understood. The physical origins of the extreme luminosity emitted by SLSNe are a focus of current research.

Supernova explosions play important roles in many aspects of astrophysics. They are sources of heavy elements, ionizing radiation, and energetic particles; they drive gas outflows and shock waves that shape star and galaxy formation; and they leave behind compact neutron star and black hole remnants. The study of supernovae has thus been actively pursued for many decades.

The past decade has seen the discovery of numerous superluminous supernova events (SLSNe; Fig. 1). Their study is motivated by their likely association with the deaths of the most massive stars, their potential contribution to the chemical evolution of the universe and (at early times) to its reionization, and the possibility that they are manifestations of physical explosion mechanisms that differ from those of their more common and less luminous cousins.

Fig. 1

The luminosity evolution (light curve) of supernovae. Common SN explosions reach peak luminosities of ~1043 ergs s−1 (absolute magnitude > −19.5). Superluminous SNe (SLSNe) reach luminosities that are greater by a factor of ~10. The prototypical events of the three SLSN classes—SLSN-I [PTF09cnd (4)], SLSN-II [SN 2006gy (12, 13, 77)], and SLSN-R [SN 2007bi (7)]—are compared with a normal type Ia SN (Nugent template), the type IIn SN 2005cl (56), the average type Ib/c light curve from (65), the type IIb SN 2011dh (78), and the prototypical type II-P SN 1999em (79). All data are in the observed R band (80).

With extreme luminosities extending over tens of days (Fig. 1) and, in some cases, copious ultraviolet (UV) flux, SLSN events may become useful cosmic beacons enabling studies of distant star-forming galaxies and their gaseous environments. Unlike other probes of the distant universe, such as short-lived gamma-ray burst afterglows and luminous high-redshift quasars, SLSNe display long durations coupled with a lack of long-lasting environmental effects; moreover, they eventually disappear and allow their hosts to be studied without interference.

Supernovae traditionally have been classified mainly according to their spectroscopic properties [see (1) for a review]; their luminosity does not play a role in the currently used scheme. In principle, almost all SLSNe belong to one of two spectroscopic classes: type IIn (hydrogen-rich events with narrow emission lines, which are usually interpreted as signs of interaction with material lost by the star before the explosion) or type Ic (events lacking hydrogen, helium, and strong silicon and sulfur lines around maximum, presumably associated with massive stellar explosions). However, the physical properties implied by the huge luminosities of SLSNe suggest that they arise, in many cases, from progenitor stars that are very different from those of their much more common and less luminous analogs. In this review, I propose an extension of the classification scheme that can be applied to superluminous events.

I consider SNe with reported peak magnitudes less than −21 mag in any band as being superluminous (Fig. 1) (see text S1 for considerations related to determining this threshold) (2).

Recent Surveys and the Discovery of SLSNe

Modern studies based on large SN samples and homogeneous, charge-coupled device–based luminosity measurements show that SLSNe are very rare in nearby luminous and metal-rich host galaxies (3, 4). Their detection therefore requires surveys that monitor numerous galaxies of all sizes in a large cosmic volume. The first generation of surveys covering large volumes was designed to find numerous distant type Ia SNe for cosmological use. These observed relatively small fields of view to a great depth, placing most of the effective survey volume at high redshift (5).

An alternative method for surveying a large volume of sky is to use wide-field instruments to cover a large sky area with relatively shallow imaging. With most of the survey volume at low redshift, one can conduct an efficient untargeted survey for nearby SNe. Such surveys provided the first well-observed examples of SLSNe, such as SN 1999as (6), which turned out to be the first example of the extremely 56Ni-rich SLSN-R class (7), and SN 1999bd (8) (Fig. 2), which is probably the first well-documented example of the SLSN-II class (9).

Fig. 2

Spectra of SLSN-II events. A spectrum of SN 1999bd obtained on 22 March 1999 with the 2.5-m Dupont telescope at Las Campanas (blue) is compared with spectra of SN 2006gy (12) (magenta), SN 2008am (57) (red), and a luminous SLSN-II from PTF (PTF10qaf, cyan). The Balmer lines show narrow and intermediate-width components [compare with the narrow host oxygen (O ii) emission lines]. A prominent emission bump around 4600 Å (short black vertical lines) is also a common feature. At a redshift of z = 0.1512, the absolute magnitude at discovery of SN 1999bd was –21.6. Telluric bands are marked and sections of the spectrum of SN 1999bd affected have been excised; the telluric A band strongly absorbs the red wing of the Hα line in this spectrum.

Further important detections resulted from the Texas Supernova Survey (TSS) (10) (text S2). On 3 March 2005, TSS detected SN 2005ap, a hostless transient at 18.13 mag. Its redshift was z = 0.2832, which indicated an absolute magnitude at peak around −22.7 mag, marking it as the most luminous SN detected until then (11). SN 2005ap is the first example of the class defined below as SLSN-I. On 18 November 2006, TSS detected a bright transient located at the nuclear region of the nearby galaxy NGC 1260 [SN 2006gy (12)]. Its measured peak magnitude was ~ −22 mag (12, 13). Spectroscopy of SN 2006gy clearly showed hydrogen emission lines with both narrow and intermediate-width components, leading to a spectroscopic classification of SN IIn; this is the prototype and best-studied example of the SLSN-II class.

During the past few years, several untargeted surveys have been operating in parallel (14). The large volume probed by these surveys and their coverage of a multitude of low-luminosity dwarf galaxies have led, as expected (15), to the detection of numerous unusual SNe not seen before in targeted surveys of luminous hosts; indeed, the SN population in dwarf galaxies has been shown to differ from that observed in giant hosts (16). More details about these surveys [including the Catalina Real-Time Transient Survey (CRTS), the Palomar Transient Factory (PTF), and the Panoramic Survey Telescope and Rapid Response System 1 (PS1)] and their results are provided in the supplementary materials.

Emerging Classes of Superluminous SNe

A total of 18 SLSNe have been discussed in the literature (Table 1). These objects can be grouped into three classes that share observational and physical attributes.

Table 1

SLSN properties by class. We list the reported redshifts, homogenized absolute peak magnitudes (see text S1), and total radiated energies as taken from the literature. The published post-peak magnitude of SN 2006tf (44) (M < −20.7 mag) is below our fiducial cutoff.

View this table:

SLSN-R. Of all classes of SLSNe, this seems to be the best understood. SLSN-R events are powered by large amounts (several solar masses, M) of radioactive 56Ni produced during the explosion of a very massive star. The radioactive decay chain 56Ni → 56Co → 56Fe deposits energy via γ-ray and positron emission, which is thermalized and converted to optical radiation by the expanding massive ejecta. The luminosity of the peak is broadly proportional to the amount of radioactive 56Ni, whereas the late-time decay (which in the most luminous cases begins immediately after the optical peak) follows the theoretical 56Co decay rate (0.0098 mag day−1). The luminosity of this cobalt radioactive tail can also be used to estimate the initial 56Ni mass.

The first well-observed example of this group was SN 2007bi, detected by the PTF “dry run” experiment (7). Its large 56Ni mass was measured using both the peak luminosity (−21.35 mag absolute) and the cobalt decay tail, followed for >500 days. Estimates derived from the observations as well as via comparison to other well-studied events (SN 1987A and SN 1998bw) converge on a value of M56Ni ≈ 5 M. The large amount of radioactive material powers a long-lasting phase of nebular emission, during which the optically thin ejecta are energized by the decaying radionuclides. Analysis of late-time spectra obtained during this phase (7) provides independent confirmation of the large initial 56Ni mass via detection of strong nebular emission from the large mass of resulting 56Fe, as well as the integrated emission from all elements, powered by the remaining 56Co.

Estimation of other physical parameters of the SN 2007bi event—in particular, the total ejected mass (which provides a lower limit on the progenitor star mass), its composition, and the kinetic energy it carries—is more complicated. There are no observed signatures of hydrogen in this event [either in the ejecta or as traces of circumstellar material (CSM) interaction]; hence, the ejecta mass directly constrains the mass of the exploding helium core, which is likely dominated by oxygen and heavier elements. Scaling relations based on the work of Arnett (17), as well as comparison of the data to custom light-curve models (7), indicate an ejecta mass of M ≈ 100 M. Analysis of the nebular spectra provides an independent lower limit on the mass, M > 50 M, with a composition similar to that expected from theoretical models of massive cores exploding via the pair-instability process. A lower ejecta mass (M = 43 M) has been proposed (18). In any case, there is no doubt that these explosions are produced by extremely massive stars, with the most massive exploding heavy-element cores known to date. The scaling relations used in (7) also indicate extreme values of ejecta kinetic energy (approaching Ek = 1053 ergs). Finally, the integrated radiated energy of this event over its very long lifetime is high (>1051 ergs).

SN 1999as, one of the first genuine SLSNe detected (6), was shown to be similar to SN 2007bi during its photospheric phase, reaching −21.4 mag absolute at peak (7) (Fig. 3); another analysis (19) suggested physical attributes (56Ni mass, kinetic energy, and ejected mass) that are close to but somewhat lower than those of SN 2007bi. Unfortunately, no late-time data have been published for this object, so it is impossible to conduct the same analysis carried out for SN 2007bi, but the similarities suggest that this was likely another member of the SLSN-R class.

Fig. 3

Photospheric spectra of SLSN-R events SN 2007bi [blue (7)], SN 1999as [magenta (7)], PTF10nmn [black (23)], and SN 2010hy (cyan); all spectra were obtained close to peak. Identification of prominent spectral features as well as a synthetic SYNOW fit [red, from (7)] are also shown.

Recently, the Lick Observatory Supernova Search [LOSS (20)] detected the luminous type Ic SN 2010hy (21, 22). After this detection, the event was also recovered in PTF data (and designated PTF10vwg). Although final photometry is not yet available for this event, preliminary Katzman Automatic Imaging Telescope and PTF data indicate a peak magnitude of −21 mag or brighter. The event is spectroscopically similar to other SLSNe-R (Fig. 3), which suggests that it is also likely a member of this class.

Objects of this subclass are exceedingly rare, and additional examples are scarce. During the past 2 years, the PTF survey has detected another likely member, PTF10nmn (23) (Fig. 3), with properties similar to those of SN 2007bi, while PS1 may have discovered another similar object at a higher redshift (24). Assembling a reasonable sample of such events may thus be a time-consuming process.

Some photospheric spectra of SLSNe-R (7, 23, 25) (Fig. 3) show forbidden line emission, notably Ca ii and probably also Mg ii. Such lines are usually only observed during the nebular phase of SNe, when the ejecta are optically thin, which is clearly not the case here. The superposition of this nebular-like emission on an underlying photospheric spectrum may hint at a complex geometry of the emitting region (motivating spectropolarimetric studies), but no explanations for this phenomenon have been put forth so far.

SN 2007bi is hosted by a dwarf galaxy [with luminosity similar to that of the Small Magellanic Cloud (SMC)], with relatively low metallicity (ZZ/3) (25)—somewhere between those of the Large Magellanic Cloud (LMC) and the SMC. Thus, although the progenitor star of this explosion probably had subsolar metal content, there is no evidence that it had very low metallicity. The host galaxy of SN 1999as is more luminous (and thus likely more metal-rich) than that of SN 2007bi, but still fainter than typical giant galaxies such as the Milky Way (3), whereas the host galaxy of PTF10nmn seems to be as faint as or fainter than that of SN 2007bi. This class of objects may thus typically explode in dwarf galaxies.

The observations of SLSNe-R strongly indicate that these events are powered by massive-star explosions that synthesize several solar masses of radioactive 56Ni, but the physical nature of the explosion is a matter of some controversy. Theoretical work suggests two options. The first is an extreme version of the iron core collapse model that is generally assumed to take place in explosions of massive stars that manifest themselves as common type II SNe (18, 26). The second is the pair-instability mechanism [e.g., (2732)]. The pair instability occurs during the evolution of very massive stars that develop oxygen cores above a critical mass threshold (~50 M). These cores achieve high temperatures at relatively low densities; substantial amounts of electron-positron pairs are created prior to oxygen ignition; loss of pressure support, rapid contraction, and explosive oxygen ignition follow, leading to a powerful explosion that disrupts the star. Extensive theoretical work indicates that such a result is unavoidable for massive oxygen cores; when the core mass in question is large enough (~100 M, as inferred for SN 2007bi), many solar masses of radioactive nickel are naturally produced. It has been shown (18, 26) that a carbon-oxygen core with a mass of ~43 M (just below the pair-instability threshold) that explodes with an ad hoc large explosion energy (>1052 ergs) can produce the required large amounts of nickel (26) as well as the light curve shape of the SLSN-R prototype, SN 2007bi (18). Both the pair-instability model and the massive core collapse model fit the light curve shape of SN 2007bi equally well. However, because the progenitors of pair-instability explosions have larger cores and thus larger initial stellar masses—which are, assuming a declining initial mass function, intrinsically more scarce—the core collapse model has been claimed to be favored for SN 2007bi (33, 34).

The two models agree about the nickel mass but strongly differ in their predictions about the total ejected mass. Total heavy-element masses above the 50 M threshold would indicate a core that is bound to become pair-unstable, and would thereby rule out the core collapse model. The core collapse model of (18), which assumes a similar amount of radioactive 56Ni and lower total ejected mass (to avoid the pair instability), predicts very strong nebular emission lines that are not consistent with the data for SN 2007bi. Thus, this model is not viable for this prototypical SLSN-R object, supporting instead a pair-instability explosion as originally claimed. It remains to be demonstrated whether the massive core collapse model applies to real supernovae; if it does, the resulting SLSNe should show large amounts of radioactive nickel but relatively small amounts of total ejecta.

Assuming, for the sake of the current discussion, that these explosions do arise from the pair instability, a clear prediction of the relevant theoretical models [e.g., (27, 28)] is that for each luminous, 56Ni-rich explosion (from a core around 100 M) there would be numerous less luminous events with smaller 56Ni masses but large ejecta masses (M > 50 M; fig. S2). These should manifest as events with very slow light curves (long rise and decay times) and moderate or even low peak luminosities.

SLSN-II. This is probably the most commonly observed class of SLSN. Whereas some examples were identified relatively early (e.g., SN 1999bd; Fig. 2), these objects became a focus of attention only after the discovery of SN 2006gy (12, 13). Since then, several additional examples have been studied in some detail. SLSNe-II show strong hydrogen features in their spectra; these explosions therefore typically occur within thick hydrogen envelopes. This makes investigations of their nature more complicated because all information carried by electromagnetic radiation from the exploding core is reprocessed by the outer envelope. For this reason, our knowledge about their energy source (or sources) is still mostly speculative. By contrast, the physics responsible for converting the explosion energy into the observed radiation is better understood.

Two main physical processes have been invoked to explain the conversion of explosion energy to emitted radiation in SLSNe-II. The first process assumes that the explosion launches a powerful shock wave expanding outward from the center of the star. This shock heats the material it traverses until it eventually escapes from the effective outer edge of the star, where this effective edge is the radius around which the material is no longer optically thick to radiation (35). The energy deposited by the shock is then slowly reemitted by the hydrogen-rich material as photons diffuse out, in analogy to the more common and much less luminous type II-P SNe, where this process occurs within the envelope of a red supergiant star. To account for the much higher observed SLSN-II luminosities, the radius of the effective edge of the star must be substantially larger than the radii of even the largest red supergiants. Deposition of shock energy into more compact stars is radiatively inefficient because the deposited energy is quickly drained by adiabatic expansion. The observed luminosities probably also require an energetic explosion shock. The shape of the light curve is determined mostly by the density structure and composition of the material into which the energy was deposited. Several options have been suggested to explain the large effective radii (>1015 cm) required for this mechanism to work. These include very large (bloated) stars [e.g., (36)], energy deposited into massive (unbound) optically thick shells ejected by previous eruptions of the exploding star that have expanded to the required radius [e.g., (37, 38)], or energy deposited into an optically thick massive stellar wind [e.g., (3942)] extending out to the required radius.

The second mechanism invoked in converting large explosion energies into optical emission is strong interaction between the expanding ejecta and massive CSM previously lost from the progenitor star. This mechanism converts the kinetic energy carried by the expanding ejecta into radiation via strong shocks, and is commonly invoked for type IIn SNe (43). Because CSM envelopes can be extremely extended, this process can in principle remain active for many years, and is thus more useful to explain very long-lived events [e.g., SN 2006tf (44), SN 2003ma (45)]. On the other hand, conversion of kinetic energy into radiation should manifest itself as an observed decline in expansion velocities; this process is therefore disfavored for events showing high expansion velocities that do not decrease substantially with time [e.g., SN 2008es (36)]. Possible mechanisms invoked to eject large quantities of mass from the star prior to explosion include luminous blue variable (LBV)–like activity [e.g., (12, 44)] and pulsational pair instability [e.g., (46)].

Regardless of the conversion mechanism, the total emitted energy in several recently observed objects (>1051 ergs) is difficult to reproduce in models of regular iron core collapse explosions (where >99% of the initial explosion energy, ~3 × 1053 ergs, is carried away by neutrinos). This led several authors to speculate about additional energy sources contributing to these powerful explosions. Pair-instability explosions can provide large kinetic energies and synthesize large amounts of radioactive 56Ni; however, SLSNe-II studied at late times did not follow the expected 56Co radioactive decay rate, in contrast to SLSNe-R. The derived limits on the amount of initial radioactive nickel generally argue against SLSNe-II resulting from energetic pair-instability explosions. Spin-down of nascent magnetars (rapidly spinning neutron stars with strong magnetic fields) has been proposed as an alternative energy source (47, 48). This process may be relevant at least for some SLSNe-II (e.g., SN 2008es). One can also consider the collapsar scenario, in which energy is extracted from material rapidly accreting onto a newly formed black hole—a process that may be driving cosmological γ-ray bursts (49). When occurring within a massive star with a thick hydrogen envelope, this process may deposit the energy in the expanding envelope, where it may be thermalized and reemitted as optical photons (11, 50). Unfortunately, because any energy injected by such processes deep inside exploding stars is then reprocessed by the optically thick outer hydrogen layers, investigations of such exotic processes in SLSNe-II are difficult and have remained mostly speculative.

In any case, it is clear that SLSNe-II are explosions of massive stars that retained their hydrogen envelopes until they exploded. For some objects, spectroscopy indicates that these stars have lost substantial amounts of mass prior to explosion [e.g., SN 2006gy (12), SN 2006tf (44)], which suggests that perhaps the progenitor stars are similar to massive LBVs, which are known to undergo episodic eruptions involving extreme mass loss [e.g., (51, 52)].

The observational characteristics of SLSNe-II are quite diverse. The peak luminosities of the brightest events reach well above −22 mag absolute [e.g., SN 2008fz (53)]. However, these seem to be the top of a broad distribution, with examples of peak magnitudes smoothly extending from these extreme values down to luminosities typical of the general SN population [e.g., SN 2010jl (54, 55); see (56) for a review of older events]. The light curve shapes are quite diverse, with some SLSNe-II showing a rapid rise and decline [e.g., SN 2008es (36)], some showing light curves with a slow rise (>50 days) to a broad peak [e.g., SN 2006gy (12, 13), SN 2008fz (53)], and some with rapid rise and very slow decline [e.g., SN 2003ma (45), SN 2008am (57), and probably also SN 2006tf (44)]. The spectra (Fig. 2) also show diversity, with most objects showing narrow hydrogen Balmer lines, and SN 2008es uniquely not showing such lines. Narrow Balmer lines arise from a slow wind blown by the progenitor star before its explosion. This wind is assumed to have been photoionized by the explosion and to then recombine. The lack of such narrow lines in the spectra of SN 2008es suggests that its progenitor star was not blowing massive winds for an extended period prior to its explosion; any substantial mass loss must have been episodic [e.g., (38)] or otherwise time-variable (41).

The environments of SLSNe-II are also quite diverse. This is the only SLSN subclass that has been detected in luminous, Milky Way–like galaxies [e.g., SN 2006gy (12, 13), CSS100217:102913+404220 (58)]. Still, like other SLSNe, most SLSN-II events reside in dwarf star-forming hosts (3). Published SLSN-II events in giant host galaxies seem to have been detected very close to their host nuclei, which suggests that perhaps specific conditions that are unique to this environment (e.g., circumnuclear star-forming rings) somehow mimic the conditions in star-forming dwarf galaxies.

SLSN-I. This class of SLSNe was initially the most difficult to understand, with the first two reported events—SN 2005ap (11) and SCP 06F6 (59)—sparsely observed. It was only after the discovery of additional members of this class by the PTF survey, bridging the redshift gap between the relatively nearby SN 2005ap (z = 0.2832) and the high-redshift SCP 06F6 (z = 1.189), that a comprehensive view of this class of objects could be formed [(4), including spectroscopic redshifts based on Mg ii absorption lines, and the first correct identification of the redshift of SCP 06F6; Fig. 4]. A major reason for the initial difficulties was that the early spectra of these objects are quite featureless and the absorption lines that do appear are mostly of high-excitation low-mass elements (Fig. 4); the elements commonly observed in most SN classes (neutral oxygen, magnesium, iron, and the ubiquitous ionized calcium) appear only much later (60) (fig. S1).

Fig. 4

Early spectra of all published SLSN-I events: SN2005ap (11); SN 2010gx, PTF09cnd, SN 2009jh, and PTF09atu (4); PS1-10awh and PS1-10ky (68); and SCP 06F6 [(59); combined version from (4)]. The optical O ii blends and near-UV C ii, Si iii, and Mg ii lines identified by (4) are marked. The spectroscopic similarity among these objects is quite striking.

Observationally, these events are characterized by extreme peak luminosities (often brighter than −22 mag absolute), very blue spectra with copious UV flux persisting for many weeks, and (relative to other classes of SLSNe) fast-evolving light curves with rise times below 50 days and post-peak slopes that decline substantially faster than radioactive cobalt decay rates (4, 60). Contrary to early reports [e.g., (11, 36, 38)], these events do not show hydrogen in their spectra (4, 60) (fig. S5) and thus do not belong to the spectroscopic class of type II SNe. Technically, these events should be classified as type Ic SNe because they also do not show strong He features in spectra taken around peak. However, because the class of type Ic SNe is not positively defined and a physical connection has not been firmly established between the events considered here and more common SN Ic events (see below), this class is denoted SLSN-I.

The similarity between the late-time spectra of SN 2010gx (SLSN-I) and the spectra of broad-line type Ic SNe (4, 52) (fig. S5) prompts discussion of a possible connection between these two classes. However, there are several physical differences between SLSNe-I and SNe Ic.

Detailed modeling of SNe Ic [e.g., (6163)] indicates that their luminosity is dominated by radioactive 56Ni decay. Indeed, these objects show a correlation between the peak luminosity and the synthesized 56Ni mass [e.g., (64, 65)]. The same is true for SLSNe-R (7) but not for SLSNe-I (4), in the sense that the nickel mass required to power the observed luminous peaks is in conflict with the later evolution of the light curve [e.g., (4, 60)]. The luminosity of SLSNe-I must therefore come from a different source.

Two very luminous broad-line SNe Ic [SN 2007D (65), SN 2010ay (66)] have been observed with peak luminosities approaching those of SLSNe-R (−20.6 mag and −20.23 mag absolute for SN 2007D and SN2010ay, respectively), but with less 56Ni (~1 M). Other processes may be contributing to the large observed peak luminosity of these events [e.g., an internal engine (66)]. Perhaps they are intermediate events between the class powered purely by radioactivity (normal SNe Ic and SLSNe-R) and SLSNe-I for which the contribution from 56Ni is negligible, and which must be powered by some other process, as discussed below.

Another important physical distinction between SLSNe-I and SNe Ic is the size of the emitting region. It has been shown that the energy radiated by SLSNe-I must have been deposited at large initial radii, ~1015 cm (4). However, early observations of SNe Ic indicate that the progenitor stars had an initial radius orders of magnitude smaller [<1011 cm (61, 67)], implying that both the explosion shock energy and radioactivity must be contained within a small initial radius.

It thus seems that the observed spectroscopic similarity between SLSNe-I (at late times) and broad-line SNe Ic suggests similar ejecta composition and a large kinetic energy (evident as a substantial amount of mass at high velocities), but other physical properties (energy source, physical size) are different and suggest a different physical mechanism powering these two classes of objects.

It has been shown (4) that the observed luminosity of these objects requires the deposition of a large amount of internal energy, taking place at large radii (1015 cm, about 10 times the size of the largest red supergiants), into material expanding at high velocities (104 km s−1). The data can rule out the traditional energy conversion mechanisms discussed above (radioactivity, photon diffusion, interaction with massive hydrogen-rich CSM). Viable options include interaction with expanding shells of hydrogen-free material (40), perhaps ejected by the pulsational pair instability [e.g., (46)], or reemission of energy injected by an internal engine, such as magnetar spin-down (47, 48) or a “collapsar”-like accreting black hole [e.g., (11, 60, 68, 69)]. Even though these events are not enshrouded by massive, opaque hydrogen shells, the physical nature of the energy source remains speculative; the energy conversion mechanism is also not clearly understood.

The host galaxies of these events are again typically dwarf galaxies, although at higher redshifts, luminosity upper limits on undetected hosts are less constraining (3, 4). The most natural explanation for these objects not occurring in more luminous galaxies is that a lower metallicity is required to form the progenitor stars of these events, but other explanations are also possible (e.g., different star formation modes or a top-heavy initial mass function in dwarf galaxies). The extreme intrinsic luminosity and plentiful UV flux of these sources make them ideal probes of dwarf galaxies at high redshifts.

Rates of SLSNe

The only measurement of the rate of SLSNe is a rough estimate based on TSS statistics (4), which, normalizing the rate of SLSNe-I at z ≈ 0.3 relative to that of SNe Ia, yields ~10−8 Mpc−3 year−1. This rate is substantially lower than the rates of core-collapse SNe (~10−4 Mpc−3 year−1) and is also well below those of rare subclasses such as broad-line SNe Ic (hypernovae; ~10−5 Mpc−3 year−1) or long gamma-ray bursts [>10−7 Mpc−3 year−1 (70, 71)]. The reported discovery statistics suggest that the rate of SLSNe-II is comparable or larger than that of SLSNe-I, whereas SLSNe-R are rarer by a factor of ~5, correcting for their slightly lower peak luminosities. SLSNe-R are the rarest type of explosions studied so far, and quite possibly they arise from stars that are at the very top of the initial mass function.


During the past dozen years, numerous superluminous SN events have been discovered and studied. The accumulated data suggest that these can be grouped into three distinct subclasses according to their observational and physical attributes. Radioactively powered SLSNe-R seem to be the best understood (and rarest) class; hydrogen-rich SLSNe-II and the most luminous hydrogen-poor SLSNe-I are more common, but the physical origins of the extreme luminosity they emit is not clear at this time. With several ongoing surveys efficiently detecting additional examples, the amount of information about these objects is likely to increase substantially in the next few years.

Supplementary Materials

Texts S1 to S3

Figs. S1 and S2

References (82103)

References and Notes

  1. Even the brightest SNe associated with cosmological gamma-ray bursts (e.g., the nickel-rich SN 1998bw) fall well below our SLSN threshold.
  2. Possible early detections of SLSNe by these surveys and their modern counterparts are discussed in text S2.
  3. The study of the peculiar SN 1997cy (72) was perhaps the first modern study of a supernova that was substantially more luminous than the norm. The absolute peak magnitude of this event (–20.1 mag), as well as that of similar objects detected since [e.g., SN 2002ic (73)], falls below our fiducial limit defined above, and thus we do not discuss them here.
  4. The models considered by (33) require stars with exceedingly large initial masses (>310 M) to form pair-unstable cores at the moderate metallicity indicated for SN 2007bi (25). However, alternative models (74) predict that stars with much lower initial masses (150 M to 250 M) explode as pair-instability SNe at SMC- or LMC-like metallicities, although they may have to be tweaked to explain the lack of hydrogen in observed SLSN-R spectra.
  5. This effective edge does not necessarily coincide with the physical edge of the star, which we define as the radius inside which material is gravitationally bound to the star.
  6. There is no conflict between this process (converting kinetic energy to radiation) and the previous one (converting shock energy stored as internal heat energy into radiation) and both can contribute in any given object, although there is debate about which one is dominant for particular cases.
  7. A similar relation has been established in the case of a lower-luminosity SN IIn [SN 2005 gl (75, 76)].
  8. See text S3 for additional details.
  9. Acknowledgments: I thank O. Yaron; Caltech Core-Collapse Project members E. Enriquez, A. Soderberg, S. B. Cenko, D. Leonard, D. Fox, D. Moon, and D. Sand; M. Phillips and P. Nugent; E. Chatzopoulos; L. Chomiuk and R. Quimby for use of data presented here; E. Nakar, P. Mazzali, D. Xu, R. Waldman, A. Pastorello, I. Arcavi, A. Howell, E. O. Ofek, A. Drake, S. Smartt, C. Wheeler, and A. Miller for useful advice; members of the PTF collaboration, and in particular J. S. Bloom, M. M. Kasliwal, S. R. Kulkarni, N. M. Law, and D. Poznanski, for use of unpublished PTF material; and the anonymous reviewers for useful and constructive suggestions and comments. Supported by grants from the Israeli Science Foundation, the U.S.-Israel Binational Science Foundation, the German-Israeli Foundation, the Minerva Foundation, ARCHES (Award for Research Cooperation and High Excellence in Science), and the Lord Sieff of Brimpton Fund. This research has made use of the NASA/IPAC (Infrared Processing and Analysis Center) Extragalactic Database, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. All spectra shown are available in digital form from the Weizmann Interactive Supernova Data Repository (WISeREP) (81) at
View Abstract

Stay Connected to Science

Navigate This Article