Cosmological Simulations


  • Eve Kovacs (ANL)
  • Scott Daniel (U. Washington)

Cosmological simulations occupy key roles in achieving the goals of the DESC. They are required for theoretical predictions (e.g., the effects of different dark energy models), for generating mock data sets to understand observational systematics and selection effects, for data analysis (error estimates, testing and development of methods), and for optimizing observational strategies.

No simulation starts from first principles, includes all known physics, and produces quantities which are directly observed to perfect precision. Different approximations are appropriate for different simulations: the appropriateness must be validated for self-consistency and for consistency with known observations. There are gravity-only simulations ('N-body') and 'hydro' simulations that include gas dynamics and additional processes, such as star formation and AGN feedback. 

Dark matter simulations have converged between methods (Heitmann et al. 2005, 2008) for predictions of key inputs to cosmological probes such as the power spectrum and halo mass functions. Developments continue in the assignment of observables (gas, galaxies and their properties) to this skeleton. Subgrid models based on hydro simulations can be incorporated into large volume N-body simulations. For simulations including gas, which can be more directly compared to observations, development is needed both in simulation techniques and in further steps to obtain observables (Agertz et al. 2007). 

This image shows a 29 Mpc/h x 29 Mpc/h x 58 Mpc/h piece of a (923 Mpc/h)3 volume HACC simulation run on the Titan Supercomputer at Oak Ridge National Laboratory.  The run uses 16384 of Titan's GPU-enhanced nodes (almost 90% of the full machine); the image represents the output from only one node. More than half a trillion particles are involved in the simulation in order to construct detailed synthetic catalogs for LSST.



Featured Projects