Ultra-High Resolution Galaxy Formation



Galaxies emit nearly all optical light in the Universe and, as a consequence, comprise the overwhelming majority of astronomical observations of objects beyond our own Milky Way. On the other hand, cosmological theories most easily model the overall mass distribution of the Universe which is dominated by so-called dark matter which, by definition, cannot be observed. These theories include such parameters as the total mass in the Universe, the fraction of the total mass in baryons and dark-matter, the age of the Universe, the visible size of the Universe, and the ultimate fate of the Universe. Consequently, the understanding of galaxy formation, evolution, and interaction is essential in linking observational constraints, which come from baryonic matter in the form of galaxies, to cosmological theories, which describe dark matter. Despite this overwhelming importance, the uncertainty in modeling galaxies and thus their connection to the underlying dark-matter mass distribution remains the weakest link between astrophysical theory and astronomical observation.

Our ignorance stems from the fact that galaxy formation and evolution are extremely difficult for numerical studies to probe. Galaxies themselves are extraordinarily complex systems where dynamical processes on scales less than a light-year have ramifications on scales upwards of 10,000 light-years. A minimum resolution of order 100,000 mass elements per galaxy is required to have any hope of realizing realistic galaxy behavior. Consequently, some researchers have attempted to simulate an individual galaxy in isolation by allocating all of their computational power to realizing the galaxy itself. Unfortunately, the preponderance of observational evidence suggests that galaxies do not exist in isolation, but rather experience ongoing interaction with their environment. Therefore, these studies are unrealistic to some degree. Another method is to perform large-scale simulations, typically 100 million or so light-years across, in order to realize all possible environmental effects. Unfortunately, due to computational limitations, these studies only allocate a few hundred resolution elements to an individual galaxy and fail to recreate even the most basic aspects of galaxy dynamics. Our simulation methodology combined with numerous algorithmic enhancements plus the overall computational sophistication and scalability of our code allows us to achieve the maximal dynamic range necessary to model the (100 million light-year) environment surrounding a galaxy and still allocate 100,000 to 1 million resolution elements to the (10 thousand light-year) galaxy itself. Consequently, in tackling this problem we are providing the most realistic simulations of galaxy formation, evolution and interaction to date.

Presented on this page is our first simulation of a "realistic" galaxy. It was performed on the Cray T3E at the Arctic Region Supercomputing Center. It contains over 100,000 particles in the disk, which is roughly 10 kpc wide (~30,000 light-years), yet is simulated in a volume which is 100 Mpc (or 326 million light-years) on a side.

The Simulation Volume:

The image below is the total simulation volume, 100 Mpc (326 million light-years) on one side. The small green box in the center is the region in which the galaxy resides. The actual galaxy (shown in the following pictures) is still only a fraction of the width green box, but if I'd made the box any smaller, you would not have been able to see it! The mass density is roughly the same thoughout the whole box. The outer areas look grainier because only a small number of extremely massive particles are used. As you look closer towards the center, increased numbers of lower mass particles provide greater detail. Each region of space in the box is only resolved at the level necessary to accurately model it's effect on the galaxy at the very center.

The Galaxy:

Now, we zoom in to the very center of the run where the galaxy itself resides. Shown below are the edge-on and face-on views of the stars (upper panels) and the gas (lower panels). The stars are color-coded by ago, with red being the oldest and yellow the youngest. In the edge-on view you can clearly see a older bulge component and a younger disk.

The simulation begins at an epoch well before the first stars in the Universe form. Consequently, all of the star particles that you see here were converted from nearby gas particles as the simulation progressed. There are many more star particles because at every timestep, only a fraction of a gas particle's mass is converted into stars, and the resulting star particle is separated. Hence, the star particles are much less massive than the gas.

Click on an image for a higher resolution version:




Movies:

Movies showing the evolution of the galaxy (QuickTime format). In the first three movies below, the star particles are color-coded by age (red=old, white/yellow=young) and the gas particles are color-coded by density (red=sparse, white/yellow=dense):

Quicktime movie showing just the stars (57MB).

Much larger Quicktime movie showing just the stars (92MB).

QuickTime movie showing the stars and gas (191MB).

QuickTime movie rendered using Joel Welling's StarSplatter package (293MB). Here, the stars are blue/white and the gas is red.

HI Column Density at z=2:

Here we see the galaxy at redshift 2 (approximately 1/8 it's age at the present day) face-on in neutral hydrogen (HI). Note the high degree of substructure as well as the numerous pockets of absorption in small satellites surrounding the man spiral.