Astronomical Science

Seeking for the Progenitors of Type Ia Supernovae

Ferdinando Patat1 Type Ia supernovae are thought to be will allow us to put constraints on the the- Poonam Chandra 2,3 thermonuclear explosions of accreting ory of binary- evolution. Roger Chevalier 2 white dwarfs that reach a critical mass Stephen Justham 4 limit. Despite their importance as cos- Having in mind why we want to do this, Philipp Podsiadlowski 4 mological distance indicators, the na- the next question is, as usual, how. A Christian Wolf 4 ture of their progenitors has remained discriminant between some of the pro- Avishay Gal-Yam 5 controversial. Observations carried posed scenarios would be the detection Luca Pasquini1 out by our team with VLT-UVES led us of circumstellar material (CSM). However, Ian Crawford 6 to the detection of circumstellar mate- notwithstanding the importance of the Paolo Mazzali 7, 8 rial in a normal Type Ia . quest, all attempts at detecting direct sig- Adalbert Pauldrach 9 The expansion velocities, densities and natures of the material being transferred Ken’ichi Nomoto10 dimensions of the circumstellar en - to the accreting white dwarf in normal Stefano Benetti11 velope indicate that this material was SNe Ia were so far frustrated, and only Enrico Cappellaro11 ejected from the system prior to the upper limits to the mass-loss rate could Nancy Elias-Rosa 7, 12 explosion. The relatively low expansion be placed from optical, radio and UV/X- Wolfgang Hillebrandt 7 velocities appear to favour a progenitor ray emission. Claims of possible ejecta- Douglas Leonard13 system where a white dwarf accretes CSM interaction have been made for a Andrea Pastorello14 material from a companion star which is few normal objects, in which the pres- Alvio Renzini11 in the red-giant phase at the time of the ence of CSM is inferred by the detection Franco Sabbadin11 explosion. of high-velocity components in the SN Josh Simon 5 spectra. However, it must be noted that Massimo Turatto11 these features can be explained by a 3D The quest structure of the explosion and, therefore, circumstellar interaction is not necessarily 1 ESO Due to their enormous luminosities and a unique interpretation. Furthermore, no 2 Department of Astronomy, University of their homogeneity, Type Ia Supernovae velocity or density estimate is possible for Virginia, Charlottesville, USA (hereafter SN Ia) have been used in cos- the CSM material, even in the case that 3 Jansky Fellow, National Radio Astron- mology as reference beacons, with the the high-velocity components in the SN omy Observatory ambitious aim of tracing the evolution spectra are indeed the effects of ejecta- 4 Department of Astrophysics, University of the Universe (Riess et al. 1998; Perl- CSM interaction. of Oxford, United Kingdom mutter et al. 1999). Despite the progress 5 Astronomy Department, California Insti- made in this field, the nature of the pro- Two remarkable exceptions are repre- tute of Technology, Pasadena, USA genitor and the physics which gov- sented by the peculiar SNe 2002ic and 6 School of Earth Sciences, Birkbeck erns these powerful explosions are still SN 2005gj, which have shown extreme- College London, United Kingdom uncertain. In general, they are thought ly pronounced hydrogen emission lines, 7 Max-Planck-Institut für Astrophysik, to originate from a close binary system that have been interpreted as a sign of Garching, Germany (Whelan and Iben 1973), where a white strong ejecta-CSM interaction. However, 8 INAF – Osservatorio Astronomico, dwarf accretes material from a compan- the classification of these supernovae as Trieste, Italy ion until it approaches the Chandrasekhar SNe Ia has been questioned, and even if 9 Institut für Astronomie und Astrophysik limit and finally undergoes a thermonu- they were SN Ia, they must be rare and der Ludwig-Maximilians-Universität, clear explosion. This scenario is widely hence unlikely to account for normal Type Munich, Germany accepted, but the nature of both the ac- Ia explosions. As a matter of fact, the only 10 Department of Astronomy, University of creting and the donor star is not yet genuine detection may be represented Tokyo, Japan known, even though favourite configura- by the underluminous SN 2005ke, which 11 INAF – Osservatorio Astronomico, tions do exist (see Parthasarathy et al. has shown an unprecedented X-ray emis- Padova, Italy 2007 for a recent review). But why is it so sion, at a 3.6 S-level, accompanied by 12 Universidad de La Laguna, Tenerife, important to investigate the nature of the a large UV excess (Immler et al. 2006). Spain progenitor system? Besides the funda- These facts have been interpreted as the 13 Department of Astronomy, San Diego mental implications on the cosmological signature of a possible weak interaction State University, USA usage of SNe Ia, there are actually sev- between the SN ejecta and material lost 14 Astrophysics Research Centre, eral other reasons to bother (Livio 2000). by a companion star. Queen’s University Belfast, United First of all, evolution depends on Kingdom the radiation, kinetic energy and nucleo- All the channels explored so far to detect synthesis yields of these powerful events. CSM around Type Ia SN progenitors Secondly, the knowledge of the initial are based on the fact that sooner or later conditions of the exploding system is cru- the fast SN ejecta will crash into the cial for understanding the physics of slow-moving material lost by the system the explosion itself. Finally, identifying the in the pre-explosion phases in the form of progenitors and determining the SN rates stellar wind. This implicitly requires two

30 The Messenger 131 – March 2008 conditions to be fulfilled: a) there has to radiation field produced during the explo- ionisation, without the need for direct in- be interaction; and b) the amount of CSM sion. Therefore, is is reasonable to expect teraction. and its density must reach some thresh- variations in the physical conditions of old values in order to produce a detecta- the CSM, like dust evaporation and/or gas With this idea in mind, the experimental ble interaction. Therefore, methods based photoionisation. path was rather clearly traced: obtain on ejecta-CSM interaction will not be multi-epoch, high-resolution spectros- able to reveal this material if its amount is It was while investigating these effects copy of the next bright SN Ia and look for small and/or if it is placed rather far from that we saw a possible way of revealing absorption-line variability. the explosion site. But there is another low amounts of CSM material without possibility of revealing CSM, basically the need of having matter interaction. In because of the transient nature of the SN fact, since in SNe Ia the UV flux blue- SN 2006X in M100 event and its high luminosity. wards of 350 nm undergoes severe line blocking by heavy elements like Fe, Co, The first chance to test our idea came In fact, if the SN is surrounded by a dusty Ti and Cr, they are capable of ionising when SN 2006X was discovered in environment, the scattered light will add possible CSM only within a rather small the Virgo Cluster M100 on with some delay to the SN signal (this is radius. Once the UV flux has significant- 4 February 2006 (Figure 1). A few days why this phenomenon is also known as ly decreased past the post-maximum later, the object was classified as a nor- a light echo), leaving certain signatures in phase, then, if the material has a suffi- mal Type Ia event occurring 1–2 weeks the observed light curves and spectra ciently high density, it can recombine, before maximum light and suffering sub- (see Patat 2005 for a review on this sub- producing time-variable absorption fea- stantial extinction. Prompt Very Large ject). These effects are expected to be tures. Of course, if the material where Array (VLA) observations have shown no dependent on the distance of the scatter- these features arise is reached by the radio source at the SN position, estab- ing material from the SN itself. More pre- fast-moving ejecta, it will be shocked and lishing one of the deepest and earliest cisely, if the dust is contained in a distant ionised, causing the disappearance of limits for radio emission from a Type Ia, cloud (like for example in an interven- such absorptions. and implying a mass-loss rate of less ing spiral arm of the host galaxy), the late than a few 10–8 solar masses per year (for time epochs of the observed SN evolu- Among all possible inter/circumstellar ab- a low wind velocity of 10 km s–1). The SN tion will be completely dominated by the sorption lines, the ubiquitous sodium was not visible in the 0.2–10 keV X-rays light echo, as in the well-known cases D lines (589.0 and 589.6 nm) are the best band down to the SWIFT satellite detec- of SNe 1991T and 1998bu. On the con- candidates for this kind of study. In fact, tion limit. All of this made of SN 2006X a trary, if the scattering material is confined besides falling in an almost telluric ab- perfect candidate to verify our idea. within a small region surrounding the SN, sorption-free spectral region, they are the effects although present at all epochs, produced by a very strong transition, and An ESO Director General Discretionary are subtle and can be confused with in- hence detectable for rather small gas Time proposal was submitted on 15 Feb- trinsic SN properties (Patat et al. 2006). column densities. In addition, the ionisa- ruary and approved immediately after- tion potential of Na I is low (5.1 eV), and wards. The observations started on Of course, if the dust is close enough to this ensures that even a weak UV field is 18 February and were carried out with the SN, this is expected to feel the strong able to have a measurable effect on its the Ultraviolet and Visual Echelle Spec-

Figure 1: The host galaxy M100 before (left) and after (right) the explosion of SN 2006X. The images were taken with VLT-FORS1 in the R passband.

The Messenger 131 – March 2008 31 Astronomical Science Patat F. et al., Seeking for the Progenitors of Type Ia Supernovae

Figure 2: Time evolution of the sodium trograph (UVES) mounted at the Very 1.0 D component region as a function of Large Telescope on four different epochs, 2 0.8 elapsed time since B-band maximum which correspond to days –2, +14, +61 Day –2 light. The heliocentric velocities have and +121 with respect to the B-band 0.6 Day +14 been corrected to the rest-frame using maximum light. Additionally, a fifth epoch the host galaxy recession velocity. All (day +105) was covered with the High 1.0 spectra have been normalised to their continuum. In each panel, the dotted Resolution Echelle Spectrometer mount- x 0.8 curve traces the atmospheric absorp- u l Day +14 ed at the 10-m Keck telescope. The data F tion spectrum. 0.6

d Day +61

show a wealth of interstellar features, but e s i the most remarkable finding is the clear l a 1.0 m evolution seen in the profile of the Na I D r

o 0.8 lines (Patat et al. 2007a). In fact, besides N Day +61 a strongly saturated and constant com- 0.6 Day +105 ponent, arising in the host galaxy disc, a number of features spanning a velocity 1.0 –1 range of about 100 km s appear to vary 0.8 significantly with time (Figures 2 and 3). Day +105 0.6 SN 2006X is situated on the receding Day +121 side of the galaxy, and the component of –100 –50 0 50 100 150 the rotation velocity along the line of sight Restframe Heliocentric Velocity v (km s–1) at the apparent SN location is about h +75 km s–1, which coincides with the Figure 3: Evolution of the Na I D and strongly saturated Na I D component, the 1.0 2 Ca II K line profiles between day –2 saturated Ca II H and K lines, and a weak- 0.8 (black), day +14 (red) and day +61 ly saturated CN vibrational band (see Fig- (blue, Na I D only). The vertical dotted 0.6 2 ure 2). This feature, and its lack of time CN R(1) R(0) P(1) lines mark the four main variable com- evolution, proves that the deep absorp- ponents at –3 (A), +20 (B), +38 (C) and A B C D +45 (D) km s–1. For comparison, the tion arises within the disc of M100 in an 1.0 upper panel shows R(0), R(1) and P(1) interstellar molecular cloud (or system

x line profiles of the CN (0–0) vibrational u of clouds) that is responsible for the bulk l 0.8 band. Colour coding is as for the other F

of the reddening suffered by SN 2006X. d two panels. e s i l 0.6 a I Na D2

In contrast, the relatively blue-shifted m r I o structures of the Na D lines show a rath- N er complex evolution. For the sake of 1.0 discussion, four main components, which we will indicate as A, B, C and D, can be 0.8 tentatively identified in the first two epochs (Figure 3). Components B, C and 0.6 D strengthen between day –2 and day Ca II K +14 while component A remains constant during this time interval. The situation –50 0 50 100 150 Restframe Heliocentric Velocity v (km s–1) becomes more complicated on day +61: h components C and D clearly start to de- crease in strength; component B remains it has been attributed by some authors For this reason we conclude that the Na I almost constant; component A becomes to line-of-sight geometrical effects, due to features seen in SN 2006X, arising in a definitely deeper and is accompanied by the fast GRB expansion coupled to the number of expanding shells (or clumps), a wide absorption that extends down to patchy nature of the intervening absorb- evolve because of changes in the CSM a rest-frame heliocentric velocity of about ing clouds. Our data clearly show that ionisation conditions induced by the vari- –50 km s–1 (Figure 3). After this epoch despite the marked evolution in the Na I D able SN radiation field. In this context, there is no evidence of evolution, and lines, Ca II H and K components do not the different behaviour seen in the Na I component A remains the most intense change with time (see Figure 2). There- and Ca II lines is explained in terms of the feature up to the last phase covered by fore, in the case of SN 2006X, trans- lower ionisation potential of Na I (5.1 eV) our observations, more than four months verse motions in the absorbing material with respect to Ca II (11.9 eV), their differ- after the explosion. and line-of-sight effects due to the fast ent recombination coefficients and pho- SN photosphere expansion (typically toionisation cross sections, coupled to a Variable interstellar absorption on compa- 10 4 km s –1) can be definitely excluded, UV-deficient radiation field. Regrettably, rably short timescales has been claimed since they would cause variations in all not much is known about the UV emis- for some Gamma Ray Bursts (GRB), and absorption features. sion of SNe Ia shortwards of 110 nm. As

32 The Messenger 131 – March 2008 we have already anticipated, from a theo- The H mass turns out to be a few 10–4 not ruled out by the lack of radio emis- retical point of view, one expects a severe solar masses (this value is reduced by a sion from SN 2006X. In fact, in the light of UV line blocking, which reduces drama- factor 100 for material at about 1016 cm, our current understanding of the ejecta- tically the flux of ionising radiation. Never- the most likely distance for components CSM interaction mechanism, the pres- theless, in the case of a SN Ia, the mod- C and D; see below). Even in the case ence of similar shells with masses smaller els show that this flux is sufficient to fully of complete ionisation, such an H mass than a few 10–4 solar masses, cannot ionise Na I up to a distance of one light would produce an HA luminosity which be excluded by radio non-detections of year (~ 1018 cm). However, since the re- is two orders of magnitude below the SNe Ia in general. Our findings are also combination timescale must be of the upper limits set by our observations at all consistent with upper limits on the radio order of 10 days to account for the ob- epochs and by any other SN Ia observed flux set by our VLA observations, ob- served variations, this requires a large so far. Therefore, the absence of narrow tained about ten months after the explo- electron density (10 5 cm –3). Given the low emission lines above the detection limit is sion, which also confirm that SN 2006X abundance of any other element, such not in contradiction with the presence is not more radio luminous than any other a high electron density can be produced of partially ionised H up to masses of the normal SNe Ia. only by partial hydrogen ionisation. On order of 0.01 solar masses. account of the severe line blocking, the flux of photons capable of ionising H is However, photo-ionisation alone cannot A new beginning? expected to be very small and this implies account for the fact that not all features that the gas where the Na I time-depend- increase in depth with time (Figure 2). In If we adopt the velocity of the CN lines as ent absorptions arise must be confined fact, on day +61, components C and D indicative of the host galaxy rotation within a few 1016 cm from the SN. turn back to the same low intensity they component along the line of sight at the had on day –2. One possible explanation SN location, then our observations pro- In a SN Ia, the UV flux decreases by a is that the gas is re-ionised by some other vide solid evidence of CSM expanding factor of ten in the first two weeks after mechanism, like the ejecta-CSM interac- at velocities that span a range of about maximum light. Since at a distance of tion. In this case, the absorbing material 100 km s–1 (Figure 2). 10 16 cm from the SN, the ionisation time- generating components C and D must scale for Na I is much shorter than the be close enough to the SN so that the The most important implication of these recombination timescale, the ionisation ejecta can reach it in about one month observations is that they show that this fraction grows with time following the after the explosion (1016 cm for maximum circumstellar material was ejected from increase of the UV flux during the pre- ejecta velocities of 4 × 10 4 km s–1). Simi- the progenitor system in the recent past. maximum phase, while after maximum it larly, in order not to be reached by the This almost certainly rules out a double- decreases following the recombination ejecta more than four months after, com- degenerate scenario for SN 2006X, where timescale. This would explain the overall ponents A, B and the broad high-velocity the supernova would have been triggered growth of the blue component’s depth, components must arise at larger dis- by the merger of two CO white dwarfs. In as shown by our data, in terms of an in- tances (> 5 × 1016 cm). This scenario is this case, no significant mass loss would creasing fraction of neutral Na, while the different evolution of individual compo- nents would be dictated by differences in the densities and distances from the SN. Moreover, once all the sodium has re- combined (which should happen within a month), there should be no further evolu- tion, in qualitative agreement with the observations. Additionally, since the flux of photons that can ionise Ca II is more than four orders of magnitude less than in the case of Na I, the corresponding ioni- sation fraction is expected to be of a few Figure 4: Artist’s impression of the per cent only. Therefore, the recombina- favoured configuration for the progeni- tion does not produce measurable effects tor system of SN 2006X before the explosion. The White Dwarf (on the on the depth of Ca II H and K lines, as is right) accretes material from the Red indeed observed. Giant star, which is losing gas in the form of stellar wind (the diffuse An upper limit to the H mass contained in material surrounding the giant). Only the clumps generating the observed part of the gas is accreted by the White Dwarf, through an accretion disc absorptions can be estimated from our which surrounds the compact star. observations, after making some con- The remaining gas escapes the sys- servative assumptions and using the Na I tem and eventually dissipates into column densities deduced from the data. the interstellar medium (see ESO Press Release 31/07).

The Messenger 131 – March 2008 33 Astronomical Science

be expected in the phase immediately culations have difficulty in matching the tions of the line of sight with respect to preceding the supernova. Thus, a single- velocities in our observations if the nova the orbital plane may exist. degenerate model is the favoured model shells are decelerated in a spherically for SN 2006X, where the progenitor ac- symmetric wind. However, if the wind is What we have seen in 2006X is far from creted from a non-degenerate compan- concentrated towards the orbital plane being completely understood and we ion star. this discrepancy could be removed, since are certainly left with more questions the nova shell would be more strongly than answers. Even though the results Mean velocities for the circumstellar ma- decelerated in the equatorial plane; in that obtained with multi-epoch, high-resolu- terial of 50 km s–1 are comparable to case we would be observing the super- tion observations of this event have al- those reported for the winds of early red nova close to the orbital plane. Not only ready triggered a couple of similar studies giant (RG) stars; velocities matching our might this be expected a priori, but (Patat et al. 2007b, Simon et al. 2007), observations are also expected for late observations of the 2006 outburst of RS the sample is simply too small to allow for subgiants. The observed material is mov- Ophiuchi show that the nova ejecta are any conclusion. Most likely, there is more ing more slowly than would be expected bipolar and that there is an equatorial than one channel leading to the same for winds from main-sequence donor density enhancement which strongly re- explosive theme, on top of which nature stars or from compact helium stars. Of strains the expansion of the nova shell, adds some variations, as the non perfect the two major formation channels pro- thus providing some support for such a homogeneity of SNe Ia seems to tell us. posed for SN Ia with a non-degenerate scenario. donor star, these wind velocities seem Rather than the end of an old story, we more consistent with the shorter-period One crucial issue is whether what we consider these findings as the beginning end of the ‘symbiotic’ formation chan- have seen in SN 2006X represents the of a new one. We hope that the tele- nel. Symbiotic systems are interacting bi - rule or is rather an exceptional case. scope time that has been allocated to our naries consisting of a late-type mass- Other cases of SNe Ia showing negative project will bring more insights into this losing giant in orbit with a hot companion, velocity components are known, like field, answering at least a few of the which accretes material from wind or SNe 1991T and 1998es. Unfortunately, questions that SN 2006X has posed us. Roche lobe overflow; they have been pro- multi-epoch high-resolution spectroscopy posed as a viable channel for Type Ia SN is not available for these objects (to our explosions (Munari and Renzini 1992). knowledge, the SN 2006X data set is References The observed structure of the circumstel- unique in this respect), and therefore time Immler S. I. et al. 2006, ApJ 648, L119 lar material could be due to variability variability cannot be demonstrated. Livio M. 2000, in “Type Ia Supernovae: Theory and in the wind from the companion RG; con- Nevertheless, the data clearly show com- Cosmology”, eds. J. C. Niemeyer and siderable variability of RG mass loss is ponents approaching the observer at J. W. Truran, (Cambridge: CUP), 33 generally expected. velocities which reach at least 50 km s–1 Munari U. and Renzini A. 1992, ApJ 397, L87 Parthasarathy M. et al. 2007, New Astronomy with respect to the deep absorption that Reviews 51, 524 A potentially more interesting interpreta- we infer to be produced within the discs Patat F. 2005, MNRAS 357, 1161 tion of these distinct features is that they of the respective host . This, Patat F. et al. 2006, MNRAS 369, 1949 arise in the remnant shells (or shell frag- and the fact that SN 2006X has shown Patat F. et al. 2007a, Science 317, 924 Patat F. et al. 2007b, A&A 474, 931 ments) of successive nova outbursts, no optical, UV and radio peculiarity, sup- Perlmutter S. et al. 1999, ApJ 517, 565 which can create dense shells in the slow- ports the conclusion that what we have Riess A. G. et al. 1998, AJ 116, 1009 moving material released by the com- witnessed for this object might be com- Simon J. D. et al. 2007, ApJ 671, L25 panion, also evacuating significant vol- mon at least for some normal SN Ia, even Whelan J. and Iben I. 1973, ApJ 186, 1007 umes around the progenitor star. Our cal - though variations due to different inclina-

2

Colour coded K-band adaptive optics image obtained with VLT NACO and the Laser Guide Star Facility of the gal- axy NGC 4945 which contains an ob-

N scured active nucleus. Close to the nucleus several luminous star clusters E can be resolved. See ESO Press Photo 27e/07 for more details.

34 The Messenger 131 – March 2008