Supernovae are spectacular but fairly rare events, at least on the human time scale. In our own galaxy, only 5 have been seen (necessarily by the naked eye, before telescopes were invented in 1608) in the last 2000 years. Since there have been none in our galaxy when any telescopes were available to study them, let alone the sophisticated instruments we have now, it’s not very surprising that there’s a lot that isn’t known about how these stellar explosions occur.
We do know that the are several types of supernova that can be distinguished by properties of their spectra (if they are close enough to Earth for spectra to be obtained). The most basic distinction is between supernovae that have hydrogen lines in their spectra (Type II), and those that don’t (Type I). Further subdivisions are possible, but the most important is between subtypes of Type I that have a prominent line due to silicon (Type Ia), and those that don’t (Type Ib and Type Ic).
For various reasons, Type Ia supernovae are especially interesting. One is that at their peak they have a fairly uniform intrinsic brightness (absolute magnitude -19.3, in the somewhat peculiar way that astronomers measure brightness). All other types have absolute magnitudes that vary over a range of about -17 to -18.5. One unit in this scale corresponds to a factor of 2.51, so for example a supernova of absolute magnitude -17 is only about 10% as bright as the standard Type Ia supernova. The fact that all Type Ia supernovae have nearly the same absolute magnitude is what makes them useful as “standard candles” to measure cosmic distances.
That Type Ia supernovae have this uniform intrinsic brightness has been explained by a model in which a white dwarf star of mass less than 1.38 solar masses (the Chandrasekhar limit) gradually (or perhaps suddenly) exceeds this limit, resulting in an explosion that destroys the star. Almost all other types of supernovae can be explained as occurring when a very massive star (more than 9 solar masses) runs out of nuclear fuel, resulting in a “core collapse” since the whole mass of the star can no longer be supported by the pressure of nuclear reactions.
Two more detailed scenarios have been proposed for circumstances leading to a Type Ia supernova. One involves a binary system with one white dwarf and one normal star, portions of whose matter are gradually drawn off by the white dwarf. The second involves a binary system of two white dwarfs that eventually merge, as gravitational waves slowly dissipate their orbital kinetic energy. Historically, astrophysicists have favored the first scenario, though with rather little actual evidence.
Surprisingly (or perhaps not, since white dwarfs aren’t very luminous stars), it has never been possible to identify a progenitor for a Type Ia supernova – all that have been observed with telescopes are in other galaxies. Neither has it been possible to detect another star, afterwards, at the event location, which would be the remaining member of the binary pair. But now research has just been published that provides evidence, rather indirectly, for the second scenario, in which Type Ia supernovae in fact result from the merger of two white dwarfs.
A new survey of distant Type Ia supernovae suggests that many if not most of these supernovae – key to astronomers’ conclusion that dark energy is accelerating the expansion of the universe – result when two white dwarf stars merge and annihilate in a thermonuclear explosion.
Evidence that Type Ia supernovae are caused by the merger of two white dwarfs – the so called double-degenerate theory – has been accumulating over the past two years, based on surveys by the Hubble Space Telescope and others. Before, astronomers favored the single-degenerate model: the idea that Type Ia’s result from the explosion of a white dwarf grown too fat by feeding on its normal stellar companion.
So, if all Type Ia events look so much alike, in spite of different models that could explain them, and if it’s difficult or impossible to observe directly what the system was like before the event, just how does the research reach its conclusions? The reasoning is somewhat indirect, but ingenious and not that hard to follow.
The key is to survey the relative frequency of Type Ia supernova events in the early universe, compared to the rate of overall star formation before the event. White dwarfs are often thought of as old stars, but in fact they may be the final stage of stars that aren’t all that old. Any star of less then 9 solar masses (those not big enough to suffer a core collapse) will eventually become a white dwarf. If the star was initially much more massive than the Sun, it will reach the white dwarf stage in perhaps one or two billion years.
No matter how large a star was before it reached the white dwarf stage, it can’t be more than the Chandrasekhar limit of about 1.4 solar masses as a white dwarf, in order to be supported by “electron degeneracy pressure“. In a binary system of a white dwarf and a larger normal star, the scenario where the white dwarf reaches the size limit to become a Type Ia supernova can happen only if both the dwarf and its companion are large enough. Given the average rate at which stars are forming in the universe as a whole – the cosmic star forming history (SFH) – it’s possible to calculate the expected rate of subsequent Type Ia supernova events at different times.
The same calculation can be made for Type Ia supernova events that result from the merger of a binary system of two white dwarfs. In this kind of system, the distribution of supernova events in comparison to the (roughly known) SFH should be different in specific ways from the distribution resulting from the other type of binary system. This can be stated in terms of the “delay-time distribution” (DTD) of Type Ia supernovae relative to rate of star formation, the SFH. If there are two white dwarfs in a binary system, it’s more likely that their combined mass will exceed the Chandrasekhar limit than if there’s just a single white dwarf accreting mass from its companion. Hence the DTD curve will be different. Specifically, it won’t decline as rapidly as a function of time.
As it happens, observations of the rate of star formation in the universe at different times point toward a broad peak in this rate somewhere around 3.5 billion years after the big bang. That corresponds to a redshift z≈2. Conveniently, it’s possible to detect Type Ia supernovae at that distance (about 10 billion light-years) with today’s largest ground-based telescopes, such as the Subaru Telescope on Mauna Kea, which was used in the research under discussion here.
Even more conveniently, core collapse supernovae (i. e. all except Type Ia) are usually too faint to be observed beyond redshift z≈1 corresponding to a distance of about 8 billion light-years. At such distances, it’s difficult to use measured redshifts and other observational characteristics to distinguish Type Ia supernovae from other kinds. There are complicated statistical algorithms that can be applied, but fortunately there shouldn’t be that many supernovae detectable at all, except Type Ia, when z>1.
The peak brightness of a Type Ia supernova doesn’t last very long, so there’s only a fairly brief window, perhaps a month and a half, in which to catch it before it’s too dim to see. Initially the brightness increases rapidly for about 2 weeks. It then starts to decline, at a rate of about .087 in magnitude per day for another 3-4 weeks (about 2 magnitudes total), and thereafter continues to decline further, but more slowly.
The survey covered galaxies in a small portion of the sky called the Subaru Deep Field (SDF). Images of each galaxy were taken at different times, and any difference in the apparent brightness of a galaxy between different images indicated a possible supernova event. After eliminating events that probably had other causes (such as active galactic nucleus flares), a total of 150 probable supernovae were detected out to z≈2. 110 of these had redshift z≤1. There were 28 with 1<z<1.5, and 12 with 1.5<z<2. All the events with z>1 were judged to be Type Ia.
Although it would seem there were many fewer events with z>1.5, it is known that galaxies in that range were forming stars much more rapidly than with z<1.5, and so the former group also contained much more gas and dust that would have obscured supernovae. If allowance is made for this, then the supernovae rates in the two groups should have been pretty close.
Finally, when supernova rates were compared with star-forming rates, the distribution of lag times between star formation and supernova events (the DTD) was a lot closer to what would be predicted by a model where two white dwarfs in a binary system merge than a model of a binary with a white dwarf and a normal star.
|Graur, O., Poznanski, D., Maoz, D., Yasuda, N., Totani, T., Fukugita, M., Filippenko, A., Foley, R., Silverman, J., Gal-Yam, A., Horesh, A., & Jannuzi, B. (2011). Supernovae in the Subaru Deep Field: the rate and delay-time distribution of Type Ia supernovae out to redshift 2 Monthly Notices of the Royal Astronomical Society, 417 (2), 916-940 DOI: 10.1111/j.1365-2966.2011.19287.x|