Welcome experts!

The Models

• Motivated by simple inflation theories, we investigate structure formation in two different cosmologies - tCDM and LCDM - with flat space-time metrics. The former is a traditional universe dominated by cold dark matter (CDM) which uses an empirically motivated initial spectrum of density fluctuations. The latter is a universe dominated by vacuum energy density (aka cosmological constant) with a minority CDM component and a self-consistent spectrum of initial fluctuations.

• The simulations for each model are designed to resolve with 500 mass element ("particles") a structure containing mass equal to that of the nearby Coma cluster of galaxies. This constraint sets the mass associated with each simulation particle to be 2.2*10¹² h^-1 Msun and sets the length L of the periodic, cubic region of space simulated to the values listed below.

• Each simulation:
  • employs one billion mass elements and 1024³ Fourier grid cells
  • generates nearly 0.5 terabytes of raw output (later compressed to about 200 Gb)
  • requires roughly 70 hours of CPU on 512 processors (four years of a single processor!)

• Some details of the tCDM model :
  • W_m = 1, s₈ = 0.6, spectral shape parameter G = 0.21
  • simulated cube of comoving length 2/h gigaparsecs (2000/h Mpc)
  • simulation begun at redshift z = 29
  • force resolution is 0.1/h Mpc

• Some details of the LCDM model :
  • W_m = 0.3, W_L=0.7, s₈ = 0.9, power spectrum from CMBFAST
  • simulated cube of comoving length 3/h gigaparsecs (3000/h Mpc)
  • simulation begun at redshift z = 35
  • force resolution is 0.1/h Mpc

menu

Light cone output

• The act of hypothetical "observers" in the volume collecting data along their past light cones is created by propagating output filters at the (virtual) speed of light through the simulated volume.  Several such virtual surveys, differing in depth and geometry, are created during each simulation, including
  • a 100 square degree, deep wedge survey oriented along the cube diagonal extending to z=4.6, 4.8 (tCDM, LCDM)
  • two deep octant surveys extending the length of the simulated cube, z=1.25, 1.46
  • two "all-sky" surveys extending to half the cube length, z=0.44, 0.58
  • one deep slice survey (15 x 75 degrees) extending to twice the simulation length using periodic images (LCDM only)
  • one planar survey (collection of pencil beams) extending to twice the simulation length using one periodic image (tCDM only)

  Use the navigation tool or simply click here to see light cone images.

menu

The Computer

• The simulations were created on a 512 (now 784) processor SGI/Cray T3E massively parallel computer at the Rechenzentrum Garching (Computing Center at Garching, or RZG) with 64 Gb of main memory (128 Mb per processor).

  Each simulation was broken into roughly 10 sequential time evolution sequences, with data archiving performed at the end of each sequence. This process was administered by T. MacFarland, J. Colberg and N. Yoshida.

menu

The Software

• We use the Hydra N-body code of Couchman, Thomas & Pearce (1995, ApJ, 452, 797) and Pearce & Couchman (1997, NewA, 2, 411) parallelized for the T3E by T. MacFarland using shmem library calls to implement explicit message passing
• . The initial condition generator was written by A. Jenkins and light cone generation routines were written by A. Evrard and A. Kudlicki.

menu

The People

• Munich (MPA) :  Tom MacFarland,  Joerg Colberg,  Simon White
• Durham :  Adrian Jenkins,  Frazer Pearce,  Carlos Frenk
• Western Ontario : Hugh Couchman
• Michigan : August Evrard
• Warsaw : Andrzej Kudlicki
• Sussex : Peter Thomas
• Edinburgh : John Peacock
• Cambridge : George Efstathiou
• For more information, see the Virgo Consortium homepage.

menu