The accelerating expansion of the Universe, and the need for dark energy, were inferred from observations1,2 of type Ia supernovae. There is a consensus that type Ia supernovae are thermonuclear explosions that destroy carbon–oxygen white dwarf stars that have accreted matter from a companion star3, although the nature of this companion remains uncertain. These supernovae are thought to be reliable distance indicators because they have a standard amount of fuel and a uniform trigger: they are predicted to explode when the mass of the white dwarf nears the Chandrasekhar mass4 of 1.4 solar masses (M⊙). Here we show that the high-redshift supernova SNLS-03D3bb has an exceptionally high luminosity and low kinetic energy that both imply a super-Chandrasekhar-mass progenitor. Super-Chandrasekhar-mass supernovae should occur preferentially in a young stellar population, so this may provide an explanation for the observed trend that overluminous type Ia supernovae occur only in ‘young’ environments5,6. As this supernova does not obey the relations that allow type Ia supernovae to be calibrated as standard candles, and as no counterparts have been found at low redshift, future cosmology studies will have to consider possible contamination from such events.
The supernova SNLS-03D3bb (SN 2003fg) was discovered on 24 April 2003 ut as part of the Supernova Legacy Survey (SNLS). Its redshift is z = 0.2440 ± 0.0003, determined from host galaxy [O ii], [O iii], Hα, and Hβ emission lines. A finding chart and observational details can be found in the Supplementary Information. From the lightcurve (Fig. 1) we measure a peak magnitude in the rest frame V band, V = 20.50 ± 0.06 mag. This corresponds to an absolute magnitude of MV = -19.94 ± 0.06 (H0 = 70 km s-1 Mpc-l, ΩM = 0.3, flat Universe). SNLS-03D3bb falls completely outside the MV distribution of low-z type Ia supernovae (ref. 7), and is 0.87 mag (a factor of 2.2) brighter than the median. We note that neither changes in the Hubble constant nor ΩM significantly affect this brightness difference. Asphericity may account for variations in luminosity in a type Ia supernova at the 25% level, but not by a factor of two8,9. SNLS-03D3bb also does not follow the lightcurve width–luminosity relationship10 for type Ia supernovae that allows them to be calibrated as standard candles—it is too bright for its lightcurve width (‘stretch’, s = 1.13) by 0.61 ± 0.14 mag (4.4σ).
Type Ia supernovae are powered exclusively by the decay of 56Ni and its decay product 56Co (ref. 11), requiring ∼0.6M⊙ of 56Ni to reproduce a normal type Ia supernova12,13,14,15. Because SNLS-03D3bb is 2.2 times overluminous, we infer that it has ∼1.3M⊙ of 56Ni. Such a large 56Ni mass is not possible if the progenitor is limited to the Chandrasekhar mass. Even models that burn the entire 1.4M⊙ to nuclear statistical equilibrium via a pure detonation produce only 0.92M⊙ of 56Ni, with the remainder comprising other iron-peak elements16. At least 40% of the type Ia supernova must be elements other than 56Ni to reproduce observed spectra14,17; this implies a white-dwarf mass of ∼2.1M⊙. Some authors find that rapid rotation may support such a massive white dwarf18. The merger of two massive white dwarfs could also produce a super-Chandrasekhar-mass product19,20.
This simple estimation of the nickel mass is supported by a more detailed calculation using the principle that the luminosity at maximum light is proportional to the instantaneous rate of radioactive decay21,22. The implied Ni mass is23,24: , where Lbol is the bolometric luminosity at maximum light (the luminosity integrated from the ultraviolet to the infrared), and α is the ratio of bolometric to radioactivity luminosities, near unity. Ṡ is the radioactivity luminosity per solar mass of 56Ni from its decay to 56Co and subsequent decay to 56Fe: where tR is the time in days for the supernova to rise from explosion to maximum light. Using tR = s × 19.5 days (ref. 25), for SNLS-03D3bb, we obtain tR = 22 days (see Supplementary Information for the effect of a shorter rise). We use α = 1.2 as a conservative value, although for high 56Ni masses, α may be lower, because nickel above the photosphere will not contribute to the luminosity23.
To convert our V magnitude into a bolometric equivalent, we use a synthetic spectrum calculated to match the observed ultraviolet + optical spectrum (Fig. 3), but extended into the infrared (13% of the bolometric luminosity is from the infrared extrapolation). The bolometric correction (mbc) is 0.07 ± 0.03 mag, such that Mbol = MV + mbc = -19.87 ± 0.06 mag. Using these numbers, we calculate MNi = 1.29 ± 0.07M⊙ for SNLS-03D3bb, in agreement with the simple scaling argument used earlier. The quoted error is from the statistical, k-correction, and bolometric correction errors added in quadrature. SNLS-03D3bb has a significantly larger bolometric luminosity and implied 56Ni mass compared to low-redshift supernovae (Fig. 2).
SNLS-03D3bb also has an unusually low ejecta velocity, as shown in the Keck spectrum taken two days after maximum light (Fig. 3). With a Si ii velocity of 8,000 ± 500 km s-1, it falls well outside the range of velocities seen for this feature at maximum light (Fig. 4). This is hard to understand in the Chandrasekhar-mass model, which predicts higher velocities for more luminous type Ia supernovae, in contrast to the unusually low velocities in SNLS-03D3bb.
The kinetic energy (EK) of a type Ia supernova arises from the difference between the nuclear energy (En) obtained from the synthesis of elements via fusion in the explosion and the binding energy (Eb) of the white dwarf13. Thus the kinetic energy velocity is: carbon–oxygen white dwarf is 0.5 × 1051 ergs (ref. 14). For a 2M⊙ white dwarf and a central density of 4 × 109 g cm-3, the binding energy is 1.3 × 1051 ergs (ref. 18).
There are only three classes of elements in a type Ia supernova (iron-peak elements, intermediate mass elements, and unburned carbon and oxygen), so a simple model can be developed for the nuclear-energy generation, En. Burning a mixture of equal parts carbon and oxygen to the iron peak produces EFe = 1.55 × 1051 erg M⊙-1, while the synthesis of 28Si produces 76% as much energy13. Thus: En = EFeMwd(fFe + 0.76fIME), where Mwd is in solar masses, and fFe and fIME are the fractional compositions of iron peak and intermediate-mass elements. If fFe and fIME do not sum to one, the remainder is the fraction of unburned carbon and oxygen (fC), which does not contribute to the nuclear energy. The 56Ni makes up approximately 70% of iron-peak elements14,16, so we adopt MNi = 0.7MwdfFe, where MNi is the mass of 56Ni.
In the Chandrasekhar-mass model, more-luminous supernovae, with more 56Ni, have a higher vke (Fig. 4). Increasing the fraction of unburned carbon and oxygen, fC, can lower the kinetic energy, perhaps accounting for some of the dispersion in type Ia supernova velocities. However, this also lowers the available 56Ni, so it cannot account for the low velocity seen in SNLS-03D3bb.
The kinetic energy gives the velocity of the supernova averaged over the entire mass, approximately equivalent to the velocity a few weeks after maximum light13. The most appropriate observational signature of this velocity is unclear, because type Ia supernova line velocities change with time, and different ions can have different relative velocities. We find good agreement between the Si ii velocity at 40 days after maximum light26 and the theoretical kinetic-energy velocity, but we emphasize that this is an imperfect comparison.
A super-Chandrasekhar mass reproduces the low velocities seen in SNLS-03D3bb (Fig. 4). Because Chandrasekhar models with more Ni produce higher velocities, the low velocities of SNLS-03D3bb imply an increased progenitor binding energy and thus a larger total mass. As a caveat, we note that this simple calculation is only intended to illustrate general trends. Future theoretical studies will have to assess such complications using different ions, different white-dwarf density structures, and a wider range of binding energies.
Super-Chandrasekhar-mass type Ia supernovae should be more likely to occur in a young stellar population, where the most massive stars exist19,20. The low-mass, star-forming host of SNLS-03D3bb is consistent with this scenario (see Supplementary Information). Thus, the apparent existence of super-Chandrasekhar-mass type Ia supernovae may explain why the most luminous type Ia supernovae only occur in young stellar environments5,6. The standard Chandrasekhar-mass model offers no explanation for this behaviour, because the total amount of fuel and triggering mechanism should be independent of the mass of the progenitor stars.
Supernovae such as SNLS-03D3bb will have to be screened out in cosmological studies. Younger stellar environments produce more-luminous supernovae, so as the mean stellar age decreases with redshift the mean properties of type Ia supernovae will change5. This can be calibrated if all supernovae obey the same stretch–luminosity relationship, but SNLS-03D3bb does not. Its peculiarity was so obvious that it was excluded from the SNLS cosmological result7, but less extreme objects could lurk in supernova samples. Future cosmology studies will need to carefully scrutinise type Ia supernovae from young populations to see whether they obey the same lightcurve width–luminosity relationship as other type Ia supernovae.
Perlmutter, S. et al. Measurements of Omega and Lambda from 42 high-redshift supernovae. Astrophys. J. 517, 565–586 (1999)
Riess, A. G. et al. Observational evidence from supernovae for an accelerating Universe and a cosmological constant. Astron. J. 116, 1009–1038 (1998)
Hoyle, F. & Fowler, W. A. Nucleosynthesis in supernovae. Astrophys. J. 132, 565–590 (1960)
Chandrasekhar, S. The maximum mass of ideal white dwarfs. Astrophys. J. 74, 81–82 (1931)
Sullivan, M. et al. Rates and properties of type Ia supernovae as a function of mass and star-formation in their host galaxies. Astrophys. J. (in the press); preprint at http://www.arXiv.org/astro-ph/0605455 (2006)
Hamuy, M. et al. The absolute luminosities of the Calan/Tololo type IA supernovae. Astron. J. 112, 2391–2397 (1996)
Astier, P. et al. The Supernova Legacy Survey: measurement of ΩM, ΩΛ and w from the first year data set. Astron. Astrophys. 447, 31–48 (2006)
Howell, D. A., Höflich, P., Wang, L. & Wheeler, J. C. Evidence for asphericity in a subluminous type Ia supernova: spectropolarimetry of SN 1999by. Astrophys. J. 556, 302–321 (2001)
Kasen, D., Nugent, P., Thomas, R. C. & Wang, L. Could there be a hole in type Ia supernovae? Astrophys. J. 610, 876–887 (2004)
Phillips, M. M. The absolute magnitudes of type IA supernovae. Astrophys. J. Lett. 413, L105–L108 (1993)
Colgate, S. A. & McKee, C. Early supernova luminosity. Astrophys. J. 157, 623–644 (1969)
Woosley, S. E. & Weaver, T. A. The physics of supernova explosions. Annu. Rev. Astron. Astrophys. 24, 205–253 (1986)
Branch, D. The Hubble constant from nickel radioactivity in type IA supernovae. Astrophys. J. 392, 35–40 (1992)
Nomoto, K., Thielemann, F.-K. & Yokoi, K. Accreting white dwarf models of Type I supernovae. III—Carbon deflagration supernovae. Astrophys. J. 286, 644–658 (1984)
Hoflich, P. Analysis of the type IA supernova SN 1994D. Astrophys. J. 443, 89–108 (1995)
Khokhlov, A., Mueller, E. & Hoeflich, P. Light curves of type IA supernova models with different explosion mechanisms. Astron. Astrophys. 270, 223–248 (1993)
Fisher, A., Branch, D., Hatano, K. & Baron, E. On the spectrum and nature of the peculiar type IA supernova 1991T. Mon. Not. R. Astron. Soc. 304, 67–74 (1999)
Yoon, S.-C. & Langer, N. On the evolution of rapidly rotating massive white dwarfs towards supernovae or collapses. Astron. Astrophys. 435, 967–985 (2005)
Tutukov, A. V. & Yungelson, L. R. Merging of binary white dwarfs neutron stars and black-holes under the influence of gravitational wave radiation. Mon. Not. R. Astron. Soc. 268, 871–879 (1994)
Howell, D. A. The progenitors of subluminous type Ia supernovae. Astrophys. J. Lett. 554, L193–L196 (2001)
Arnett, W. D. Type I supernovae. I—Analytic solutions for the early part of the light curve. Astrophys. J. 253, 785–797 (1982)
Arnett, W. D., Branch, D. & Wheeler, J. C. Hubble's constant and exploding carbon-oxygen white dwarf models for type I supernovae. Nature 314, 337–338 (1985)
Nugent, P. et al. Low Hubble constant from the physics of type Ia supernovae. Phys. Rev. Lett. 75, 394–397 (1995)
Nugent, P. E. Non-local thermodynamic equilibrium spectrum synthesis of type IA supernovae. PhD thesis, Univ. Oklahoma (1997)
Conley, A. et al. The rise time of type Ia supernovae from the Supernova Legacy Survey. Astron. J. (in the press); preprint at http://www.arXiv.org/astro-ph/0607363 (2006)
Benetti, S. et al. The diversity of type Ia supernovae: evidence for systematics? Astrophys. J. 623, 1011–1016 (2005)
Nugent, P., Kim, A. & Perlmutter, S. K-corrections and extinction corrections for type Ia supernovae. Publ. Astron. Soc. Pacif. 114, 803–819 (2002)
Bessell, M. S. UBVRI passbands. Publ. Astron. Soc. Pacif. 102, 1181–1199 (1990)
Fisher, A., Branch, D., Nugent, P. & Baron, E. Evidence for a high-velocity carbon-rich layer in the type IA SN 1990N. Astrophys. J. Lett. 481, L89–L92 (1997)
Marion, G. H. et al. Low carbon abundance in type Ia supernovae. Astrophys. J. 645, 1392–1401 (2006)
SNLS relies on observations with MegaCam, a joint project of CFHT and CEA/DAPNIA, at the Canada–France–Hawaii Telescope (CFHT). We used data products from the Canadian Astronomy Data Centre as part of the CFHT Legacy Survey. Some data were obtained at the W. M. Keck Observatory. We acknowledge support from NSERC, NERSC, CIAR, CNRS/IN2P3, CNRS/INSU, CEA and the DOE.
This file contains Supplementary Methods, Supplementary Figure, Supplementary Table and additional references. (PDF 141 kb)
About this article
General Relativity and Gravitation (2018)
Space Science Reviews (2018)
Foundations of Physics (2017)