NOTE:  These data have been removed, since they are now so old as to be
       essentially obsolete.

NOTE:  We discovered a file system error on 12 July 2002 which had caused
       some of the files here to be corrupted at some indeterminate time
       in the past.  If you downloaded data before 12 July 2002 you will
       need to reload the data from the current version.

Simulation data.
================

The data represent dumps at constant time of a simulation employing 512^3
particles in a periodic volume 300Mpc/h on a side.  The cosmology assumed
OmegaM=0.3, OmegaL=0.7, h=0.7, n=1, OmegaB.h^2=0.02, sigma_8=1.  The
initial conditions were generated by displacing particles from a regular
grid using the Zel'dovich approximation at z=60.  The transfer function was
taken to be the fit of Eisenstein & Hu with the baryon oscillations `smoothed'
over.  The simulation was evolved to z=0 using a TreePM code with a Plummer
equivalent force softening of 20kpc/h (fixed in comoving coordinates).  The
particles have equal masses M~1.7e10 Msun/h.

For each output I ran a FoF group finder with a linking length of 0.15 times
the mean interparticle spacing.  All groups with fewer than 8 particles were
assigned to the special group "0" which stands for "not in a group".  Other
groups were numbered from 1 to N with no particular correlation between group
number and number of members.

All of the files were written for a PC running Linux, so have the opposite
endianness to a Sun or IBM.  The files employ only 4-byte fields so you can
swap the endianness trivially if you need to.  The program "swap4endian.c"
will do this if you do not have code to do it yourself.

The group data is stored in tpmsph_?.????.grp where ?.???? identifies the
scale factor of the output.  This file is simply 512^3 integers, with each
integer giving the group number to which the particle in the dump corresponds.

The phase space information is split across 16 files (which are written in
parallel to increase throughput during the run).  The file format for the
phase space distribution is some header data:

Type	Name	Comments
----	----	--------
int	eflag	An integer which should be 1.  Used to check the endianness
		is correct for the machine reading the data.
int	hsize	The size of the header information, in bytes.  This is 20
		for the current run.
int	Np	The number of particles in this file.
int	Nsph	The number of gas particles in this file, should be 0 for
		a pure DM run like this one.
int	Nstar	The number of star particles in the file, again 0.
float	aa	The exact scale factor of the dump.
float	eps	The softening, in units of the box length, where the force
		becomes exactly 1/r^2.


then the x,y,z positions of all the particles, listed sequentially, each stored
as a single precision real number.  The positions are in units of the box
length L=300Mpc/h.  After the positions come the velocities, vx,vy,vz, for each
particle, also single precision real numbers.  These are in units of the
expansion velocity across the box.  To convert to km/s you multiply by aHL
which at z=0 is 30,000km/s.  In these units it is very easy to move into
redshift space from real space: simply add the relevant velocity coordinate
to the position coordinate.
After the position and velocity coordinates come the particle ID numbers.  The
particles are not stored in the different time dumps in the same order always,
so the PID is listed to help in making merger trees etc.

Also in this directory is a program "readhead.c" which will read and print the
header information for each of the .bin files and a program "bin2unf.c" which
will convert each of the files into a set of seven f77 unformatted files
containing arrays of x, y, z-position, x, y, z-velocity and pid for those who
use Fortran.

Martin White
mwhite@berkeley.edu