How did the stellar disk of the Milky Way form over time? Was it a slow, continuous process, or did it happen through violent mergers and the destruction of smaller galaxies that collided with ours in the remote past? How and when was the black hole at the center of our own galaxy formed? How did energy feedback from supernovae influence the formation of successive generations of stars that now populate our galaxy? From whence did the heavy elements come? Understanding the formation of galaxies like our own is a fundamental problem in present day cosmology.

Fabio Governato, Thomas Quinn, Gregory Stinson and Octavio Valenzuela, members of the University of Washington (UW) N-Body Shop originally started by George Lake (Washington State University), have started an ambitious project: simulating the formation in a cosmological context of a spiral galaxy like our own Milky Way at unprecedented detail.

To this aim, they are using PKDGRAV, the parallel N-body code developed by Quinn, Joachim Stadel (University of Zurich) and Wadsley (McMaster, CA). PKDGRAV has been successfully used on ARSC supercomputers over the last ten years. This multistep tree-code is in continuous evolution and has been recently modified by UW graduate student Greg Stinson to incorporate a detailed description of gas physics, star formation and supernova explosions.

Making accurate predictions of galaxy structure requires millions of mass elements and tens of thousands of timesteps, requiring 150,000 node hours. PKDGRAV scales extremely well to 128 nodes on the ARSC IBM SP, so that the researchers can complete a decade of node time in a month of wall-clock time. The resulting speed-up from a single-node computer to a supercomputer is similar to comparing a Mach 1 fighter jet to a bicyclist. This research would be impossible without a computer of this power, say Governato and Quinn.

Gravitational Clustering

Stars orbit around the galaxy center as planets around the sun, with their rotation speed determined by the amount of mass enclosed by their orbit. Measurements of rotation velocities in spiral galaxies are the best direct evidence for dark matter, as they show the existence of massive halos that extend well beyond the optical disk of galaxies. The masses inferred for these halos are consistent with those predicted by the Cold Dark Matter (CDM) theory, where most of the mass in the Universe is made of as yet undetected exotic particles.

Numerical simulations by Governato and Quinn at UW follow gravitational clustering to an unprecedented level of detail, and the main characteristics of the assembly of dark matter halos within the CDM framework is now well understood.

A new simulation done by Governato, Valenzuela and Quinn at ARSC is the highest resolution to-date of the formation of a small dark matter halo. With this spatial resolution, the researchers can make a robust prediction of the expected gamma ray flux caused by the annihilation of the dark matter particles (possibly neutralinos) as they collide with each other at the center of cosmic halos.

But when do galaxies' stars form? As gas radiatively cools, it assembles at the center of dark matter halos eventually forming stars. If the gas conserves its angular momentum acquired from cosmic tidal torques, it will eventually settle into a rotationally supported disk of stars and cold gas. In the CDM framework, dark matter halos contain hundreds of smaller structures, the central cores of halos that survived the merging process. This abundance of substructure clashes with the observational evidence that visible galaxy satellites around our Milky Way are hundreds of times less numerous than what has been predicted by CDM.

Alternative Dark Matter Models

This evidence, coupled with observations suggesting that galaxies have central dark matter densities smaller than those formed in numerical simulations with a CDM component, has lead to several speculations on alternative dark matter models. Some of them have significantly less power at subgalactic scales like warm dark matter and self interacting dark matter. However, constraints on the above proposed solutions are tight, and simulations with alternative dark matter models have been unsuccessful in reproducing the observed abundance and spatial distribution of galaxies.

The above results encouraged the N-Body Shop to look for solutions to these problems by exploring the still poorly understood processes linked to the assembly of baryons and the subsequent energy feedback into the intergalactic medium.
Preliminary simulations have already achieved remarkable success: the researchers were able to form realistic stellar disks supported by rotation at the center of dark matter halos. This feedback model drastically reduces the number of dark matter subhalos that are able to hold onto their gas, form stars and become visible.

The simulations have been run on Klondike and Iceberg and have used over 200,000 node hours so far. Scientists at the N-Body Shop are now busy writing up these results for submission to international peer reviewed journals. Governato plans to use at least twice as many computer hours before this project is completed!

State and National Resource…

The Arctic Region Supercomputing Center supports high performance computational research in science and engineering with an emphasis on high latitudes and the Arctic.

The center provides high performance computational, visualization, networking and data storage resources for researchers within the University of Alaska, other academic institutions, the Department of Defense and other government agencies. ARSC is located on the UAF main campus in Fairbanks, Alaska.

For more information, visit the N-Body Shop.

Arctic Region Supercomputing Center | PO Box 756020, Fairbanks, AK 99775 | voice: 907-450-8600 | email:

home | search | about | support | news | science | resources