Main

The progenitor of any type Ia supernova has never been identified. Various candidate classes have been proposed (see Table 1 and Supplementary Information section 1), although arguments and counterarguments have resulted in no decisive solution. It is possible that the observed type Ia supernovae might have two comparable-sized progenitor classes21. In double-degenerate systems, the two white dwarfs will both be completely destroyed by the supernova explosion. In single-degenerate systems, the mass-donor star (orbiting the doomed white dwarf) will survive the explosion, and shine at near its pre-explosion brightness from the middle of the expanding supernova remnant. (During the explosion, portions of the outer envelope of the companion star will be stripped off22,23, but its location on the colour–magnitude diagram will not change greatly24.) An observational programme to distinguish between the progenitor models by looking for an ex-companion star inside a known type Ia supernova remnant has been attempted only once14, for Tycho’s supernova of 1572. A particular G-type subgiant star has been identified as being the ex-companion: if this is correct, it would point to a recurrent nova as the progenitor for Tycho’s supernova14. Several concerns have been raised15,17 about this identification, and these have been answered18, although the case remains unresolved.

Table 1 Candidate progenitor classes
Full size table

To break this impasse, we have looked at a supernova remnant in the nearest galaxy to our own, the Large Magellanic Cloud (LMC). We consider the case of SNR 0509−67.5, which was an type Ia supernova (of the SN 1991T class) 400 ± 50 years ago19,20,25,26. SNR 0509−67.5 has excellent images in the public domain that were taken by the Hubble Space Telescope (HST). All of the stars in the field have been measured for B, V and I magnitude with standard IRAF aperture photometry and converted to Vega magnitudes with the standard calibration (see Table 2). The faintest visible star (at the 5σ detection level) is at V = 26.9 mag.

Table 2 Objects near the centre of SNR 0509−67.5
Full size table

If any ex-companion still exists after the explosion ∼400 years ago, then it must be located near the centre of the remnant. We have measured the geometric centre of the shell with three independent methods (see Supplementary Information section 2): using the edge of the Hα shell, the edge of the X-ray shell, and the minimum of the Hα light in the interior of the remnant. Each of these three derived centres are from different gas masses and regions, so they are independent and provide a measure of the statistical and systematic uncertainties in the centre position. Our combined geometric centre is at right ascension 05 h 09 min 31.208 s, declination −67° 31′ 17.48″ (J2000), with 1σ uncertainties of 0.14″ along the short axis (roughly ENE to WSW) and 0.20″ along the long axis (tilted 18° ± 3° to the west of north).

The position of any ex-companion star will be offset from the estimated geometric centre of the shell, owing to proper motion of the star, asymmetries in the shell, and measurement errors of the centre position. We now consider these factors in order. First, the proper motion of the star will depend on its orbital velocity and the kick onto the star from the supernova explosion. This distribution of the offsets from the centre resulting from the proper motion of the star does not have a Gaussian profile, so we express the allowed positions as ellipses with a 99.73% probability (that is, 3σ) of containing the position of the ex-companion star. As the proper motion depends on the nature of the companion, we report ellipses for red giants, subgiants and main-sequence stars. Second, for SNR 0509−67.5 in particular, the shell expansion is uniform in all directions except for one quadrant where the interstellar medium is more dense (as shown by the excess 24-μm emission seen in the Spitzer image27 from pre-existing dust swept up by the shell) and so the expansion has recently slowed down28. This slowing in only one quadrant accounts for the small observed ellipticity of the shell, from which we can derive the apparent offset (1.39″ ± 0.14″ along a line 18° ± 3° south of west) between the observed geometric centre of the shell and the site of the supernova explosion. Last, our derived best estimate for the site of the explosion is right ascension 05 h 09 min 30.976 s, declination −67° 31′ 17.90″ (J2000). The error ellipse is nearly circular, with a conservative radius of 1.43″ for a maximal proper motion (390 km s−1), a maximal age for the remnant (550 years) and for 99.73% (3σ) containment. (See Supplementary Information section 3 for details.)

The error circle is completely empty of all visible point sources down to the deep limits of HST. Importantly, there are no red giant or subgiant stars in or near the circle. (Red giants and subgiants can be confidently recognized by their position above the main sequence in the colour–magnitude diagram.) The nearest red giant (star O in Fig. 1) is 7.4″ from the centre, while the nearest subgiant star (star N) is 5.8″ from the centre. The nearest star brighter than V = 22.7 mag (star K), that is, the nearest possible ex-companion of any type, is 2.9″ from the centre. The only source in the circle is an extended faint nebula, and the excellent angular resolution of the HST allows us to see that no point source is hidden within the nebula. (This nebula is probably an irregular galaxy of moderate redshift, but the coincidence of this nebula with the site of the supernova suggests that its origin might be associated with the explosion, as discussed in Supplementary Information section 4.) The error circle is empty of point sources to a limiting magnitude of V = 26.9 mag (at the 5σ level). This requires that any ex-companion be less luminous than MV = +8.4 mag.

Figure 1: SNR 0509−67.5 and the extreme 99.73% error circle.
figure 1

This is a composite HST image: the Hα image was taken with the WFPC2 over three orbits in November 2007 with a total of 5,000 s of exposure; the B, V and I images were taken with the WFC3 over two orbits in November 2010 with 1,010, 696 and 800 s exposure, respectively. North is up and east is to the left. These HST data were processed and combined with standard PYRAF and IRAF procedures. The figure shows a combination of all four filters, with the remarkably smooth Hα shell visible. The error circle (solid line, at centre of image; with 1.43″ radius) is the extreme 99.73% region (3σ), where to be on the edge the ex-companion star must be a main sequence star with the minimum possible mass for any published model (1.16 solar masses), the velocity must be entirely perpendicular to the line of sight, the age of the supernova remnant must be pushed to the 3σ highest possible value (550 years), and the measurement error for the remnant’s geometric centre must be pushed to the 3σ extreme. The only source inside the error circle is a nebulous object that looks like a background galaxy, however the location of this object at the centre suggests it might be related to the supernova event (see Supplementary Information section 4). There are no stars within the extreme error circle to V = 26.9 mag, which corresponds to an absolute magnitude of MV = +8.4 mag in the LMC. All published models for single-degenerate progenitors have the ex-companion star appearing more luminous than MV = +4.2 mag (V = 22.7 mag in the LMC). In all, our extreme 99.73% error circle is very conservative, and there is no point source to limits 4.2 mag deeper than possible for any published model of single-degenerate systems.

PowerPoint slide

Full size image

Our new limit can be compared to the expected presence of ex-companion stars for the various single-degenerate models (see Table 1). There is no red giant star in or near the error circle, and this is strongly inconsistent with the symbiotic progenitor model. There is no red giant or sub-giant star in or near the error circle, and this is strongly inconsistent with the recurrent nova, helium star and spin-up/spin-down progenitor models. There is no star brighter than V = 22.7 mag in or near the error circle, and this is strongly inconsistent with the supersoft source progenitor model. The lack of any possible ex-companion star to MV = +8.4 mag rules out all published single-degenerate progenitor models. With all single-degenerate models eliminated, the only remaining progenitor model for SNR 0509−67.5 is the double-degenerate model.