Abacus Cosmos

A suite of cosmological N-body simulations


Overview

The file formats for all the data products are documented below. Python code is also provided to load the friends-of-friends and Rockstar halos, but using it is completely optional. If nothing else, it provides a concrete example of how to read the files, in case you have any questions after reading the documentation below.

Jump to:

Friends-of-friends halos

Directory structure

The friends of friends (FoF) halos are located in the SimName_FoF_halos subdirectory of each simulation. An example directory structure follows:

  • AbacusCosmos_1100box_products
    • AbacusCosmos_1100box_00_products
      • AbacusCosmos_1100box_00_FoF_halos
        • info (see Info directory)
        • z0.300
          • halos.tar.gz: The main halo catalogs
          • fof.cfg: FoF configuration file
          • header: the Abacus parameter file describing this redshift slice
          • halo_subsamples.tar.gz: a 10% subsample of particles inside halos
          • halo_subsample_ids.tar.gz: the corresponding particle IDs
          • field_subsamples.tar.gz: a 10% subsample of particle outside of halos
          • field_subsample_ids.tar.gz: the corresponding particle IDs

filename: halos_N

These files form the main halo catalog (i.e. all halo information except for particle subsamples). The number of these files is arbitrary. The file format is two integers, followed by the list of halo records:

Name Type Units Description
num_groups uint64_t   Number of halo records in this file
n_largest_subhalos uint64_t   Number of subhalos records per halo
halos struct halo   The halos

The halo records are unpacked C structs with the following fields:

Name Type Units Description
id int64_t   Halo ID
npstart uint64_t   Particle subsample starting index in corresponding particles_N file
npout uint32_t   Number of subsampled particles
N uint32_t   Number of halo particles
subhalo_N uint32_t[N_LARGEST_SUBHALOS]   Number of particles in the few most massive subhalos
x float[3] Mpc/h Center of mass of the halo
v float[3] km/s Velocity of the center of mass
sigmav float[3] km/s Velocity dispersion
r25, r50, r75, r90 floats Mpc/h Radial percentiles of the particle distribution
vcirc_max float km/s The maximum circular velocity of the halo
rvcirc_max float Mpc/h The radius at which vcirc_max occurs
subhalo_x float[3] Mpc/h Center of mass of the most massive subhalo
subhalo_v float[3] km/s Velocity of the center of mass
subhalo_sigmav float[3] km/s Parent halo velocity dispersion, centered on the most massive subhalo
subhalo_r25, subhalo_r50, subhalo_r75, subhalo_r90 floats Mpc/h Parent halo radial percentiles, centered on the most massive subhalo
subhalo_vcirc_max float km/s The maximum circular velocity of the parent halo, centered on the most massive subhalo
subhalo_rvcirc_max float Mpc/h The radius at which subhalo_vcirc_max occurs

filename: particles_N

These files hold the particle subsamples for the halos in file halos_N. The format is a list of particles stored as binary C structs. Each struct has the following format:

Name Type Units Description
pos float[3] Mpc/h The particle position
vel float[3] km/s The particle velocity

Particles can be associated with halos by the npstart and npout fields of the corresponding halos_N file.

The corresponding particle IDs can be found in particle_ids_N. They appear in the same order as this particles_N file.

filename: particle_ids_N

These files hold the particle IDs corresponding to the particle subsamples for the halos in file halos_N. The format is a binary list of particle IDs:

Name Type Units Description
pid uint64_t   The particle ID

Particle IDs appear in the same order as particle positions in the particles_N file. Particles can be associated with halos by the npstart and npout fields of the corresponding halos_N file.

filename: field_particles_N

These files hold the field particle subsamples. Field particles are particles that do not fall in any FoF halo. The format is a list of particles stored as binary C structs. Each struct has the following format:

Name Type Units Description
pos float[3] Mpc/h The particle position
vel float[3] km/s The particle velocity

The corresponding particle IDs can be found in field_ids_N. They appear in the same order as this field_particles_N file.

filename: field_ids_N

These files hold the particle IDs corresponding to the field particle subsamples. Field particles are particles that do not fall in any FoF halo. The format is a binary list of particle IDs:

Name Type Units Description
pid uint64_t   The particle ID

Field particle IDs appear in the same order as field particle positions in the field_particles_N file.

filename: fof.cfg

The parameters that FoF was invoked with. The most important parameters are documented here:

Name Type Units Description
linklen float interparticle spacing Level 1 FoF linking length for finding halos
linklen_L2 float interparticle spacing Level 2 FoF linking length for finding subhalos
n_largest_subhalos int   How many subhalo masses to record in subhalo_N
min_members int Particle count Minimum halo size to output
n_block int   The number of blocks to divide the domain into. Should equal the number of halo_N output files.

Units

All distances are comoving in the domain [-BoxSize/2, BoxSize/2), and all velocities are proper. See the Rockstar unit notes for information about comparing FoF positions to Rockstar positions.

Rockstar halos

The Rockstar halos are located in the SimName_Rockstar_halos subdirectory of each simulation. An example directory structure follows:

  • AbacusCosmos_1100box_products
    • AbacusCosmos_1100box_00_products
      • AbacusCosmos_1100box_00_rockstar_halos
        • info (see Info directory)
        • z0.300
          • halos.tar.gz: the main halo catalogs
          • rockstar.cfg: the Rockstar configuration file
          • header: the Abacus parameter file describing this redshift slice
          • halo_subsamples.tar.gz: a 10% subsample of particles inside halos

filename: halos_M.N.h5

The Rockstar halos are stored in HDF5 files. There is one dataset called halos. The fields are identical to default Rockstar, with the following modifications:

  • We split pos[6] into pos[3] and vel[3]
  • We add the parent_id field. For subhalos, this is the id of the parent halo. For parent halos, this is -1.
  • We add the subsamp_start and subsamp_len fields, corresponding to the starting index and count of the halo particle subsample.
  • We add m_SO and alt_m_SO[4] for the spherical overdensity masses.
  • We add N, alt_N[4], N_SO, and alt_N_SO[4]: the particle counts corresponding to the mass values.

The header file is also stored as an HDF5 dataset attribute, but this is purely for convenience; it contains the same information as the actual header file.

filename: particles_M.N.h5

The particle subsamples corresponding to halos in halos_M.N.h5. There is one dataset called subsamples (in AbacusCosmos) or particles (in emulator_planck). The dataset contains the positions, velocities, and, in AbacusCosmos, the PIDs.

The header file is also stored as an HDF5 dataset attribute, but this is purely for convenience; it contains the same information as the actual header file.

filename: rockstar.cfg

The Rockstar configuration file. See the Rockstar documentation for definitions.

Units

The mass definitions corresponding to the m and alt_m fields can be read from the MASS_DEFINITION fields in rockstar.cfg. In almost all cases, we have left the mass definitions at the default, so m corresponds to the virial mass.

For convenience, we copy the Rockstar notes about units from the ASCII header here. You may need to consult the Rockstar documentation or source code for the exact definition of a quantity, however.

  • Masses in Msun / h
  • Positions in Mpc / h (comoving)
  • Velocities in km / s (physical, peculiar)
  • Halo Distances, Lengths, and Radii in kpc / h (comoving)
  • Angular Momenta in (Msun/h) * (Mpc/h)*km/s (physical)
  • Spins are dimensionless
  • Total energy in (Msun/h) * (km/s)^2 (physical)
  • idx, i_so, and i_ph are internal debugging quantities
  • Np is an internal debugging quantity.

The Rockstar halo positions are in the domain [0, BoxSize). Rockstar subsample particle positions may be negative if the halo is near the edge of the box. Applying a periodic wrap (pos % BoxSize) will bring the particle positions back to the same domain as the halos.

A related issue is that Rockstar takes the Abacus particle positions modulo the box size, since the original particles span [-BoxSize/2, BoxSize/2) but Rockstar wants to work in [0, BoxSize). FoF works in native Abacus units and does not do this conversion. If you want to compare halos between Rockstar and FoF, you will need to apply the same unit conversion. In the forward direction, one way to do this is as follows:

\( \mathbf{x}_\mathrm{Rockstar} = \mathbf{x}_\mathrm{FoF} \bmod L \)

In the backwards direction:

\( \mathbf{x}_\mathrm{FoF} = \mathbf{x}_\mathrm{Rockstar} - L\, \mathrm{round}(\mathbf{x}_\mathrm{Rockstar}/L)\)

SO halos

Rockstar can also produce spherical overdensity masses for the same set of halo centers that the main code uses. That is, no new halo finding is done, but new masses are computed that include all unbound particles. We compute these masses and store them in the m_SO and alt_m_SO fields.

Caution should be used when interpreting SO halo masses of less than 100 particles. About 2% of halos in this regime are so close to another halo that their spherical overdensities do not fall below the threshold before encompassing the larger halo. To prevent small halos from artificially inflating their masses this way, Rockstar limits the SO search radius to 10% beyond the nominal halo radius (this is the BCG2_R=1.1e-3 factor). In these cases, the reported mass is not a “true” SO mass in the sense that the particle counting was truncated before the SO threshold was reached. If nothing else, this is yet another reason not to trust small halos.

Particle subsamples

Both the friends of friends and Rockstar catalogs contain 10% halo particle subsamples for use, e.g. in populating halos with HOD galaxies. The subsample particles are selected based on their particle ID, so the set of “subsample-able” particles is the same across time slices, and even between FoF and Rockstar. The fact that the same particles are subsampled in every time slice (if they are in a halo) could allow one to construct crude merger trees, for instance.

The FoF catalogs also include a 10% subsample of particles that fall outside of halos, called “field” particles. Thus, the union of the halo particles and the field particles is a uniform 10% subsample of the whole matter density field.

We do not include a field subsample from Rockstar, as it would increase the data volume significantly and require internal Rockstar changes.

The Rockstar particle subsamples are taken from the FoF group belonging to each halo (corresponding to the num_p field, not N, which takes into account substructure and mass unbinding, among other things). Thus, the subsamples will be 10% of num_p, not N and will not include substructure. To fetch substructure particles, use the parent_id field to find the subhalos for a given halo.

Power spectra

Power spectra from each redshift slice are located in the SimName_power subdirectory of each simulation. An example directory structure follows:

  • AbacusCosmos_1100box_products
    • AbacusCosmos_1100box_00_products
      • AbacusCosmos_1100box_00_power
        • info (see Info directory)
        • z0.300
          • power_nfft2048.csv: The measured power spectrum
          • header: the Abacus parameter file describing this redshift slice

The power spectrum file power_nfft2048.csv is a comma separated values file with three columns: wavenumber \(k\) [h/Mpc], power \(P(k)\) [(Mpc/h)^3], and number of modes \(N_\mathrm{modes}\). The suffix _nfft2048 in the filename indicates that the power spectrum computation used a \(2048^3\) mesh.

We compute the matter power spectra by gridding the particles onto a mesh (\(2048^3\) or finer) with triangle-shaped cloud mass assignment. We then Fourier transform the density field, convert the result to a power spectrum, de-convolve the TSC-aliased window function from Jeong (2010), and bin in spherical annuli. The number of mesh cells that fall into each annulus is recorded as N_modes in the third column of the csv file.

info directory

Every data product directory contains an info directory (alongside the redshift directories) with various input and output files. An example info directory structure is shown here:

  • info
    • abacus.par: the main Abacus parameter file
    • abacus_params.par2: level 2 parameter file that gets processed into the level 1 abacus.par file. Useful for just seeing the parameters that vary among sims.
    • camb_derived.out: quantities that CAMB outputs as it runs
    • camb_matterpower.dat: the main CAMB matter power spectrum; this is the input power spectrum to the zeldovich-PLT code.
    • camb_params.ini: the CAMB input file
    • camb_transfer_out.dat: various transfer functions that CAMB produces

Abacus parameter and header files

The parameter file info/abacus.par is the input configuration file to Abacus for the simulations. It contains settings that do not vary with time, like the box size and Omega_M.

The header file zZ.ZZZ/header is an Abacus output file that describes the current redshift slice. It contains the entire abacus.par file, followed by time-varying quantities, like the scale factor and OmegaNow_m, the \(\Omega_M\) that an observer at this epoch would measure.

You will usually want to parse the header file for analysis tasks. If you are working in Python, InputFile.py can do this for you.

Important parameters

We collect the most important simulation parameters here for convenience. These are values can all be found in info/abacus.par or zZ.ZZZ/header. If something isn’t covered here, the Abacus user guide contains detailed descriptions of all parameters.

Name Type Units Description
BoxSize float Mpc/h Comoving side length of the simulation cube
H0 float km/s/Mpc The Hubble constant at z=0
InitialRedshift float   The starting redshift of the simulation
NP int64   The number of particles in the simulation. Equal to ppd^3.
N_eff float   The effective number of neutrino species. Only used by CAMB, not Abacus.
Omega_DE float   Dark energy density parameter at z=0
Omega_M float   Matter density parameter at z=0
SofteningLength float Mpc/h in AbacusCosmos sims; Unit-box in emulator_planck sims Plummer-equivalent softening length. Unit-box units can be converted to Mpc/h by multipling by BoxSize.
ns float   Scalar spectral index. Only used by CAMB, not Abacus.
ombh2 float   Physical baryon density parameter. Only used by CAMB, not Abacus.
omch2 float   Physical CDM density parameter. Only used by CAMB, not Abacus.
simga_8 float   Amplitude of density fluctuations at 8 Mpc/h. Technically not read by CAMB or Abacus. This is converted to ZD_Pk_sigma before the zeldovich-PLT IC code is invoked.
w0 float   Dark energy equation of state parameter at z=0
wa float   Dark energy equation of state evolution parameter: \(w(a) = w_0 + (1-a)w_a\)
ppd float   “Particles per dimension”; NP = ppd^3
ScaleFactor float   The current scale factor a = 1 / (1 + z)
BoxSizeMpc and BoxSizeHMpc floats Mpc and Mpc/h The comoving size of the box in Mpc and Mpc/h
HubbleTimeGyr and HubbleTimeHGyr floats Gyr and Gyr/h The value of 1/H0 in Gyr and Gyr/h; note that all times below are reported in units of H0 = 1.
ParticleMassMsun and ParticleMassHMsun floats M_sun and M_sun/h The particle mass in M_sun and M_sun/h units
VelZSpace_to_kms float   Conversion factor of output particle velocities to km/s
Redshift float   The redshift z at this epoch; 1 + z = 1/a
Time float   Proper time in 1/H0 units
Growth float   The linear growth function, normalized to \(a\) at early times (EdS)
Growth_on_a float   The linear growth function divided by \(a\), which is unity in Einstein-de Sitter.
f_growth float   The growth rate \(d \ln D / d \ln a \)
w float   The dark energy equation of state parameter at this epoch
HubbleNow float   The Hubble parameter at the current epoch, in H0 units, i.e., H(z)/H0
Htime float   The product of the Hubble parameter and the current time, H(z)t(z). This 2/3 in Einstein-de Sitter.
OmegaNow_m float   The \(\Omega_M\) that an observer at this epoch would measure
OmegaNow_DE float   The \(\Omega_{DE}\) that an observer at this epoch would measure
SofteningType string   The softening technique employed. See the Abacus Cosmos paper for softening details.