Sunday, May 20, 2012

What makes a supernova?

At the beginning of the month we had a short cosmology meeting here in Bielefeld, known as Kosmologietag (though in fact it was spread over two days). One of the main talks was given by Fritz Röpke, of the Universität Würzburg. Fritz spoke about a very important sub-class of supernova explosions, known as Type Ia, and our current understanding of what causes them to explode in the first place.

Although I have worked in the past with supernovae data in a cosmological context, much of what he said during his talk was new to me, and also was very interesting. Supernovae are particularly amazing events, after all! The slides of his talk are available here though unfortunately some of the movies he showed will not work (incidentally, all the other talks from the meeting — including my own — can be found here, in case anyone is interested).

This post is somewhat experimental because I'm writing about a subject I don't have much expertise in, and also because I will try to convey some of the new things I learned from Fritz's research-level talk while assuming little or no prior knowledge about the subject among my readers. So please bear with me. I suppose I'd better start by briefly explaining what a supernova is, and what the 'Type Ia' business refers to.

Supernovae are explosions that occur at the end of the life cycle of some types of massive stars, that cause the exploding star to increase rapidly in brightness for a short period of time. The first thing everyone should know about supernovae is that they are truly catastrophic explosions, the most dramatic events that occur in the Universe today, which cause a single star to become as bright as an entire galaxy containing about a hundred billion stars. Indeed when the star whose remnants are now known as the Crab Nebula went supernova in 1054 AD, it was bright enough to be seen in the daytime, and was recorded by Chinese astronomers of the time as a 'guest star'.

Here's a picture from the Hubble Space Telescope of Supernova 1994D (this means it was the fourth supernova observed in 1994) in Galaxy NGC 4526:

Supernova 1994D exploding in galaxy NGC 4526

The second thing you should know is that all the metal in the computer that you are reading this on was formed in such a gigantic supernova several billion years ago. We, and everything around us, are really composed of the debris of one of these stellar explosions. Now that should be enough to capture your interest!

Once a supernova is detected — usually this is done by painstakingly comparing images of the same area of the sky before and after — astronomers get busy with more precise measurements. The two most important things to measure are the change in the amount of light received from the supernova (the flux) with time, and the spectrum of that light. Different elements present in the debris of the exploding star lead to specific spectral features, so careful analysis of the spectrum gives us important clues about what the star that went supernova consisted of. Based on this analysis, supernovae are sorted into different categories: Type I supernovae don't have a strong hydrogen absorption line and Type II do. There are also further subcategories: Type Ia show a strong silicon (Si II) line and lack helium, Types Ib and Ic don't have the  Si II line and Ib also has some helium. Slightly more than half of all observed supernovae are of Type II. Type Ia account for about 24%.

Characteristic spectral features of the different types of supernovae explosions
Spectral features seen in the light from supernovae of different types. The y-axis shows the relative flux. I nicked this image from Fritz's talk, but he clearly took it from somewhere else himself! 
Cosmologists are normally more interested in Type Ia supernovae than any of the other types. This is because it has been discovered empirically that amongst Type Ia supernovae, intrinsically faint supernovae increase and decrease in brightness quite quickly, whereas intrinsically brighter ones change relatively slowly. Of course supernovae appear brighter or fainter depending on how far away they are from us, but this relationship makes it possible to tell how bright a given supernova actually was, and therefore how far away from us it is — in the jargon, they are known as "standard candles" (strictly speaking "standardizable candles" since they aren't actually all the same brightness). Following through this chain of reasoning and making very careful and clever measurements is what allowed supernova search teams to deduce that the expansion of the universe is currently accelerating, a discovery which earned Saul Perlmutter, Brian Schmidt and Adam Riess the Nobel Prize last year.

But that isn't what I'm going to talk about this time. Fritz's talk was to do with what causes Type Ia supernovae to occur in the first place. You'd think that given that we can use these supernovae as standard candles to measure the acceleration of the universe so precisely, we understood them very well. But actually that isn't true. And clearly it is a really important thing to study!

The first clue we have about the nature of Type Ia progenitors comes from the fact that their spectra show no hydrogen or helium. The other clue is that if you look at images of the region where a supernova occurred from before it occurred, you don't see any star there. So the star that exploded must have been very compact and not very bright, and lacking hydrogen and helium. This suggests the progenitors are carbon-oxygen white dwarfs.

A white dwarf is the end-point in the life-cycle of a star. During their active lives, the gravitational infall of material in stars raises the temperature and density at their centres until such point as nuclear fusion reactions can occur there. While the energy released by the nuclear fusion provides an outward pressure to balance the force of gravity the star can remain in equilibrium. Where this point of equilibrium is depends on which fusion reaction is taking place. Once the supply of hydrogen in the core of the star is exhausted it must burn helium, but the helium fusion is only ignited at significantly higher temperatures and densities. The same thing happens when the helium supply runs out: to achieve carbon burning, gravity must work to raise the central temperature above a minimum threshold.

Some stars simply aren't massive enough for gravity to be able to produce the temperatures necessary for carbon burning. What happens is that the star collapses under gravity until electron degeneracy pressure (arising due to the fact that electrons, being fermions, cannot be compressed into too small a volume of space) halts it, before carbon fusion starts. If the mass of the collapsing core is below the famous Chandrashekhar limit of about 1.4 times the mass of the Sun, gravity isn't enough to overcome this pressure and so this is where the story ends: a small, dense core composed of carbon and oxygen, that just quietly cools until it essentially emits no light at all. A white dwarf of this type is essentially eternally stable.

An aside: what happens if the mass is higher than the Chandrashekhar limit and electron degeneracy pressure cannot withstand gravity? The core collapses, leaving either a neutron star, supported by neutron degeneracy pressure, or perhaps a black hole. The collapsed configuration has much lower gravitational potential energy, so in the process of collapsing the core releases a huge amount of energy, mostly in the form of neutrinos; part of this is transferred to the outer layers of the star causing them to heat up, become very bright and expand very fast. This is a "core-collapse" supernova, and is what is believed to happen in Type Ib, Ic and II supernovae. (I say "believed" because apparently the people who try to produce computer simulations of such supernovae can't get their simulations to explode.)

But Type Ia supernovae — the most important ones — are believed to come from carbon-oxygen white dwarfs, so something must happen to suddenly make them unstable. The conventional story, which is still found in plenty of popular descriptions, is that the white dwarf below the Chandrashekhar limit gains material from a nearby companion star until it goes above the limit (this is called the "single degenerate" scenario). When this happens, the density and temperature to start carbon fusion is achieved. Carbon fusion releases a lot of energy, raising the temperature, and the higher temperature causes the carbon fusion reaction to speed up: the white dwarf achieves thermonuclear runaway. Since the Chandrashekhar limit is always the same, all such supernovae should proceed in pretty much the same way, which is why they should be such good standard candles.

Actually this is the point where, before Fritz's talk, my knowledge ran out. Everything from here on I have only learned recently. The first thing I learned was that the crucial product of carbon fusion is an unstable isotope of nickel, $^{56}$Ni. This isotope then decays to the stable $^{56}$Fe end product emitting $\gamma$-ray photons, neutrinos and positrons along the way; it is these that transfer energy to the ejecta and cause what we see as the supernova.

When the carbon burning starts, it doesn't proceed everywhere at once. Instead you get a combustion front, a thin layer a few millimetres thick where the actual carbon fusion occurs, that travels through the white dwarf. Modelling how this front travels through the degenerate electron plasma of the white dwarf is quite complicated, and the amount of $^{56}$Ni produced at the end is quite sensitive to the details. Since it is this amount of $^{56}$Ni which ultimately determines the brightness of the supernova, it is important to get the modelling right. The state-of-the-art models require a combination of a turbulent flame front advancing relatively slowly through the white dwarf, and a supersonic shock wave or "detonation front". When this is done, it appears that the computer models can reproduce some of the observed light curves reasonably well.

The real difficulty for supernova modellers trying to fit all observations is what the initial conditions for their models should be. This is because it isn't clear that the standard picture of the single degenerate starting scenario is correct. For a start, although the white dwarf is too dim to be seen, its companion star shouldn't be. Yet on archive images it is very rare to find an image of a star in the location where a supernova was subsequently observed. In addition, as the white dwarf strips gas from its partner it should emit X-rays, but these haven't been seen.

Even more importantly, in computer models the force of the supernova explosion knocks large amounts of hot hydrogen- and helium-rich gas off its companion star. This ejected material should then show up in the form of tell-tale spectral lines, but right from the very beginning Type Ia supernovae were classified as those which had no hydrogen or helium spectral lines! The companion star is however not completely destroyed in the computer models, so it should be possible to find survivors at the site of Type Ia explosions, yet they aren't found. And finally, there are in fact several sub-classes of Type Ia supernovae that have somewhat different characteristics — if all explosions started in exactly the same manner then we shouldn't have this variability.

A somewhat more attractive explanation is called the "double degenerate" scenario: two independently stable white dwarfs with masses below the Chandrashekhar limit happen to collide with each other and thus trigger the thermonuclear runaway reaction. In this case there wouldn't be a visible progenitor system before the supernova, nor any surviving star at the end of it. There would be no problem with hydrogen or helium lines in the spectrum, and no expected X-rays, nor would you expect all mergers to have similar characteristics. I would have imagined that the probability for two white dwarfs to encounter each other and merge in this fashion would be too small, but apparently this isn't the case either.

If it is true that the progenitor system for Type Ia supernovae is different to what we used to think it was, this will have important consequences for our understanding of the explosions themselves. And of course while the complicated plasma physics, hydrodynamics and nuclear physics that go into understanding supernovae explosions are fascinating and important goals in themselves, the cosmologists' perspective is to wonder how this affects our understanding of supernovae as standard candles. Do Type Ia supernovae evolve differently now than they did a billion years ago? To answer this conclusively we probably need to properly understand how they evolve at all, and to do that we need to understand what it is that explodes in the first place.

5 comments:

  1. I've known about the standard explanation of type Ia supernovae since my first year astro course, and it's rather extraordinary to hear about the number of problems with it!

    One thing puzzles me about the WD-WD merger suggestion: if these occur frequently, shouldn't there also be mergers between all sorts of other star pairs? If so, what do these events look like? I imagine they would also typically be quite energetic.

    ReplyDelete
  2. This comment has been removed by the author.

    ReplyDelete
  3. (Sorry, technical issues with the previous comment. This is what it said.)

    Good question. I have to admit that I don't really know the answer, though I have seen some suggestions that SN 1987A occurred as the result of two stars (not WDs) colliding. Type Ia supernovae are only <25% of all supernovae, so there's room for other types of violent ends! Also sometimes the collision may not result in a nova or supernova but something less dramatic, leaving you with a star that just seems to have the wrong evolutionary history - I just saw a paper that suggests something like this might be responsible for 'blue stragglers'.

    Perhaps I missed something in Fritz's talk, but now I'm not sure whether the merging WDs are supposed to have been binary companions or a chance encounter between independent systems.

    ReplyDelete
    Replies
    1. Definitely binary companions. Chance encounters between independent systems are exceedingly rare. Even when galaxies collide head-on, the stars don't.

      Delete