Overview
Modeling galaxy formation is a multi-scale, multi-physics computational problem that must account for a wide range of astrophysical phenomena, scales, and interactions. These models require cutting-edge numerical simulations using vast computational resources. We have performed state-of-the-art adaptive mesh refinement (AMR) hydrodynamic simulations of galaxy formation and evolution of the intergalactic medium.
Project Details
Our three-dimensional AMR simulations are the current state-of-the-art in galactic modeling. The spatial dynamic range spans more than 6 orders of magnitude, while the mass dynamic range exceeds 10 orders of magnitude. These advanced simulations are, for the first time, enabling the formation and evolution of thousands of galaxies to be followed simultaneously. Detailed analysis of such a large sample of galaxies provides an unprecedented opportunity to gain deep physical understanding of the fundamental mechanisms underlying galaxy formation and cosmological phenomena.

Results and Impact
In contrast to conventional wisdom, our new simulations show that galaxies with very high redshifts (redshifts greater than 6) are not only dusty but contain a significant number of old stars. Our simulation results for these dusty galaxies show excellent agreement with observations from the Hubble Space Telescope in terms of stellar mass function, ultraviolet (UV) luminosity function, far-UV to near-UV color, optical to UV color, and other properties. In the near future, data from the Atacama Large Millimeter Array will also enable us to verify additional simulation predictions about the far-infrared properties of these high-redshift galaxies.

Role of High-End Computing Resources
NASA’s high-end supercomputing resources have been indispensable to this project. The Pleiades and Endeavour supercomputers provide the combination of a large-scale parallel cluster platform and a high-RAM symmetric multiprocessing platform needed to perform both our intensive production simulations and post-simulation data mining.Renyue Cen, Princeton University
cen@astro.princeton.edu