Research Galaxy formationScientific computingEarly universe

Academic papersOutreach & mediaMusic

Andrew Pontzen Galaxy Formation

Our Universe is amazingly structured. Far from being scattered at random, stars (of which our Sun is just one) are organized into galaxies. Not all galaxies have the same number of stars, but ours (which is fairly typical) contains a few hundred billion. In turn, galaxies are themselves sometimes clustered together or, more generally, aligned along massive cosmic filaments.

Although most cosmologists currently believe we understand quite well how the universe behaves on very large scales, and that computer simulations can essentially tell us the answers we need, there remains a great deal of uncertainty on galactic scales. Unlike uncertainty associated with unknown physics in the early Universe, these small-scale uncertainties arise from the sheer complexity of known physical laws.

If we are to understand the meaning of our observations of the Universe, it's essential to make sense of this complexity. The way that galaxies grow and evolve changes our view of the Universe completely, for instance determining the crucial link between the dominant — but invisible — dark matter and the visible cosmos.

To make progress in our understanding, we program computers to simulate some fraction of the physical laws which are thought to be important in explaining the processes of galaxy formation. We create initial conditions which reflect the observed state of the early Universe, then get the computer to evolve the virtual universe forward in time. It's the closest astronomers will ever get to being able to do laboratory tests — with the frustrating complication that we don't really know how accurate our ‘laboratory tools’ are.

Here's my introduction to all this in video form.

Why does dark matter behave like it does?

Despite not knowing what dark matter is, we can model its behaviour using a computer (by assuming the particles feel only gravity, no other forces). This kind of computer simulation is relatively simple; conversely gas and stars (which we temporarily ignore) are extraordinarily complicated because they interact in all sorts of different ways (through light emission and absorption, for instance).

Yet, even if we put to one side the fundamental question of what dark matter is, these ‘dark-matter-only’ simulations throw up their own troubling problems. In particular, we find that dark matter tends to collapse into halos (stable objects which host galaxies when we put the gas and stars back in). That's not strange in itself: what's strange is that these halos look the same, regardless of how we configure the dark matter when we start the simulation. You can put in subtle ripples and waves (like the real universe), or just bung in a solid sphere of the particles, but once everything has settled down you find the same kind of universal halos.

Until we understand that side of things, we are poking around in the dark (excuse the pun) when investigating anything more complicated (like gas and stars). So recently I've had a go at combining a fundamental idea in statistics, maximizing entropy, with new insights into how large numbers of dark matter particles behave on average. It seems to explain what's going on pretty well (paper here). I'm planning to go a stage further and link up the ideas in this thread with those detailed below on the effect of gas — watch this space.

How does gas get out of galaxies? And how does it affect the dark matter?

I'm currently spending a significant amount of time trying to make sense of how gas escapes from galaxies. We know this to be an important process in terms of spreading heavy elements — the byproduct of the lives of stars — through the cosmos. It's also likely to be important in understanding how individual galaxies evolve.

It turns out that, when stars form in dense clumps in our simulations, those clumps are very rapidly destroyed as a fraction of the stars explode in supernovae. The destruction process has two important effects.

First, it expels gas from the galaxy (see movie below which shows hot gas exploding out of galaxies in the simulations; you can also download the movie so long as you attribute it when using it in talks).

Second, it makes life difficult for the dark matter. In particular, it is very efficient in removing dark matter from the central parts of the galaxy. That's significant because it resolves the so-called cusp-core problem, which states that standard theories predict too much dark matter in those central parts. The paper explaining all this for specialists is available on arXiv (1106.0499).

Making best use of observations

I'm interested in finding new ways to exploit what we know about the real universe to put our embryonic understanding to the test. In particular, as we make progress towards forming realistic-looking virtual galaxies, it becomes crucial to test whether the actual mechanisms via which the galaxies are forming are true-to-life.

Much of my previous work has focussed on using observations of the high-redshift Universe — i.e. looking at very distant objects. Because the light takes a long time to travel over the immense distances involved, we see these objects as they were when the Universe was much younger.

I'm particularly interested in what absorption line systems tell us about the formation of galaxies. These are not seen in the traditional way, which would involve looking for light emitted by the objects themselves. Instead, we find these objects by looking for a characteristic ‘thumbprint’ on the light coming from yet more distant sources.

This movie shows an example of that thumbprint building up as the light travels through a virtual Universe. At the start, the spectrum (a graph showing the intensity of light of different colours) for a quasar is shown. As that light travels through the Universe, it redshifts, but also picks up very visible 'bites' where the light of a particular colour has been absorbed by a cloud of gas. (See also a more recent HD version including only the diffuse gas.)

It's like looking for an object's shadow instead of looking for the object itself. Strange as that sounds, it actually gives us an interesting view on the cosmos, because we can see the conditions in very distant clouds of gas — even if those clouds are not themselves sending us much (or any) light. We have good reasons to believe that all galaxies started out as relatively small blobs of gas, and this kind of technique is our best bet of finding those blobs of gas.

See publications page for more information.