Friday, June 8, 2012

Galaxy Masses and the "Parking Lot" Problem

I recently helped to organize a small workshop on galaxy formation at the Leiden Observatory on May 14-15, 2012, entitled "Theory Goes out on a Limb: Theoretical Predictions for z > 1 Galaxies" (Harry Ferguson of CANDELS and Marijn Franx of Leiden were co-organizers). The goal of the workshop was to feature theory more than is usual at astronomy conferences, in an effort to stimulate theoreticians to make concrete predictions about what properties of galaxies should be observed at moderate to high redshifts.

Illustration of a spiral galaxy embedded in a dark
matter halo, with infalling gas:
Credit: NASA; ESA; A. Feild, STScI
I was struck by one of the major themes that emerged at the workshop, namely, the persistence of a problem that has been with us for several years now and refuses to go away. The problem is the fact that our theoretical predictions for matching the number of galaxies of different masses are increasingly bad back in time -- models tend to UNDERpredict the number of massive galaxies but OVERpredict the number of small galaxies. The discrepancy is at the level of factors of 2-10 depending on mass, big enough to be quite disturbing.

A new insight (for me) was that the fault can plausibly be attributed to the fact that most models take gas falling into halos of galaxies (drawn in by gravity) and rather quickly and efficiently turn it into stars. We understand this pattern of gaseous infall quite well from theory, and there is little opportunity to alter it -- it is driven by the growth of halo structure in the universe, which is well understood. But this pattern of gaseous infall is failing badly to reproduce the mass growth of galaxies. To save the theory seemingly requires major modifications to the sequence of events whereby galaxies take infalling gas and turn it into stars.

Though details are lacking, we understand in a general way the fix that is required.  Essentially, we have to save up a lot of the gas early on in some kind of inert form and then turn it into stars a few billion years later. My favorite name for this solution is the "parking lot", wherein gas has to be put into a reservoir where it does nothing for some time but then becomes available for star formation after a few billion years. This latter part is the tricky thing -- it's pretty easy to drive gas out of the galaxy entirely by using energy produced by the first generation of stars in "feedback" mode. Early stars can produce a lot of energy (or momentum) to drive gas out in the form of intense radiative from young hot, stars and blastwaves from exploding supernovae. Such feedback has been known for some time.  But feedback is very easily overdone -- if too intense, it can drive gas out of the galaxy entirely and render it permanently unavailable. In other words, successful feedback has to walk a narrow knife edge in which there is sufficient feedback to eject the gas and keep it ejected for some time but not so much as to lose it altogether. And then the gas has to return on just the right timescale to make stars later and match the observations.

It was my impression from the meeting that there is no generally agreed-on method to accomplish this. Without a solution to this problem, we are lacking a fundamental theory for galaxy masses, and, without a theory for masses, we cannot predict the galaxy content of the distant universe to any useful accuracy.

Perhaps a better title for the workshop  would have been:
         "Theory Goes Out on a Limb...and Saws It Off."


  1. A really interesting post, and it sounds like a great meeting!

    Is the problem perhaps not even more complex though? The observations seem to suggest the high mass end of the stellar mass function is in place by redshift 1. So would you not need a mechanism to increase the star formation efficiency in high mass galaxies (relative to their low mass counterparts) at early times in particular? However, this mechanism must become less effective at late times. If it didn't, the high mass end of the stellar mass function would continue to evolve and we would over-predict the abundance of the most massive galaxies. So the mechanism evoked must require a time and mass dependent evolution to it.

    Alternatively the observations could be always be inaccurate. The steepness of the stellar mass function at the high mass end makes this the regime which is most susceptible to systematics in the stellar mass determinations. Then there is always the possibility of a varying initial mass function complicating things...

    1. I agree that there are problems at the high mass end as well. Most of the discussion at the meeting focused on "quenching" of star formation at the high mass end and whether it was internally triggered (e.g. if a galaxy gets above a certain mass) or externally triggered. Of course the quenching process might be different for galaxies that are at the centers of their dark-matter halos than ones that are satellites within a larger halo (e.g. a group or cluster of galaxies).

      And I completely agree that the observations can be inaccurate. It's great to compare models to "stellar mass functions," but also important to compare directly to the observed distribution of luminosities and colors, where it's much easier to understand the observational errors.