Armed with just some physical concepts and a powerful computer, can we build a mathematical model of galaxy evolution that looks like what we see through our telescopes? That was the question on the minds of thirty or so astronomers who met May 14-15, 2012 at the Lorentz center at Leiden university in Holland.
The resounding answer was no. Which is fine. We knew that going into the meeting. Otherwise there wouldn't have been much to talk about. The important thing is to try to understand why the models fail. Of course, if we all agreed on that, then there also wouldn't have been much point in holding the meeting. So the topic of discussion and debate was the stress tests for the models -- which observations were particularly useful, which aspects of the models seemed most dicey, and were there any new ideas for addressing them?
To start cooking galaxies on the computer, everyone starts with a mix of dark matter (a bit more than 25%), dark energy (a bit less than 75%), and a frosting of normal baryonic matter (the stuff atoms are made of). They assume that the universe is expanding, and that the expansion rate was slowing down for the first few billion years, but has been speeding up since then due to the pressure of dark energy. They fill the universe with an almost perfectly uniform soup of dark and baryonic matter, with just tiny fluctuations in density with an amplitude and scale set to match observations of the cosmic microwave background and expectations from inflationary theory. They set the universe expanding and then watch as gravity starts to pull material towards the centers of the densest fluctuations.
Okay, this already sounds pretty dicey. No one knows what dark matter and dark energy are. Nonetheless, these basic assumptions explain the propensity for galaxies to cluster together in groups. And the particular assumption that the dark matter is "cold" -- which means it was moving at speeds much lower than the speed when the universe was about 400,000 years old -- explains the clustering of galaxies (measured using correlation functions or power spectra) in detail. The most magnificent illustration of this recently has been the detection of the "baryon acoustic peak" --- the strongest feature observed in the spatial power spectrum of the cosmic microwave background --- in the clustering of present-day galaxies. That said, one of the reasons that astronomers are really interested in understanding galaxies is to use them to help understand fundamental physics, such as dark matter. It's just frustratingly indirect.
Most of the meeting was focused not so much on the ingredients, but in how the galaxies evolve -- how gas cools, condenses into stars and black holes, and cycles energy back into the gas to prevent further cooling. The basic challenge is that the gas that collects around the densest concentrations of dark matter tends to cool too fast. If this is what actually happened, the universe today would be filled with galaxies much smaller than the Milky Way -- hundreds of times more dwarf galaxies than we actually see. So something must have kept the gas in galaxies from cooling too fast. For galaxies about the size of the Milky Way, it's quite likely that this something was the the energy released by stars (and especially the violent deaths of stars in supernova explosions).
Astronomers generally agree on how much energy stars release, but no one has yet been able to cook up perfect galaxies on a computer (following all of the systematic trends of real galaxies). They tend to be too small and to rotate too rapidly. When these problems are fixed by tuning the feedback recipe, then other problems arise. When models are tuned to match observed galaxies at the present day, they generally run into trouble matching the observations of distant galaxies. In a recent experiment called the Aquilla project, experts in simulating galaxies all started their codes with exactly the same initial conditions. The galaxies that they produced ended up looking completely different. And none of them looked quite like the Milky Way. On the other hand, the highest resolution simulation to date, the Eris simulation, does end up looking a lot like the Milky Way. Perhaps this is because in this simulation, the energy from star formation can be deposited where it is needed, in the dense lumps of gas that formed the stars. But with only one simulation at this resolution, we don't yet know if the trends among galaxies can be reproduced.
Lumps in the sauce?
There was quite a lot of interest and debate at the meeting about giant clumps of star-formation in distant galaxies. Such giant clumps were very common in galaxies when the universe was less than four billion years old, but are quite rare today. We do not know for sure if these clumps represent merging galaxies, or if the clumps formed from gas that had already settled in a rotating disk. Some people at the meeting thought the clumps could be just an artifact of the computer models (due to some of the approximations used). Others thought that the clumps were an essential feature of galaxy evolution, and that their existence could be predicted without a sophisticated computer code. They must be fairly long-lived features or else they would be uncommon. But there was a lot of debate on whether they survive long enough to sink to the centers of galaxies, forming the central bulges we see today. For the observers, just finding clumps in distant galaxies and trying to agree on their properties is proving to be a challenge. On the bright side, the simulations can be used to guide our measurement techniques, even if the simulations themselves don't produce very realistic galaxies.
Is it soup yet?
As is often the case in meetings like this, one is left with both hope and despair. There were some remarkable successes (for example in explaining the evolution in the sizes of certain types of galaxies over the past few billion years), and in reconstructing the average star-formation histories as a function of dark-matter halo mass. But there were also sobering comments from experts who pointed out that different modelers sometimes wildly disagree on the equations that govern the same physical process (such as rates of cooling of gas in dark-matter halos), and that there are physical processes that no one really knows how to model well (such as recycling of gas that was ejected from a galaxy but later falls back in). There is plenty of homework before the next such conference.