Modified Gravity
Quijote contains N-body simulations with modified gravity: Quijote-MG. The movie below shows one of these simulations together wits \(\Lambda {\rm CDM}\) counterpart:
If you are interested in using these simulations, please contact us at marco.baldi5@unibo.it or villaescusa.francisco@gmail.com.
General description
Quijote-MG contains 4,048 N-body simulations run with MG-Gadget and using the Hu & Sawicki f(R) model as the modified gravity model. Each simulation follows the evolution of \(512^3\) dark matter plus \(512^3\) neutrinos in a periodic cosmological volume of \((1000~{\rm Mpc}/h)^3\). The initial conditions have been generated using the Zel’dovich approximation at \(z=127\) and the simulations have been run with the appropiate Hubble function \(H(z)\). We have saved 5 snapshots, at redshifts 0, 0.5, 1, 2, and 3. For each simulation we have saved FoF catalogs, Rockstar catalogs, and different power spectra (see below).
The simulations can be classified into two different groups:
Simulations designed for Fisher matrix calculations
Simulations designed for machine learning calculations
Simulations for Fisher matrix
For the first category we have 2,000 simulations. In this category there are four different types:
500 simulations run with \(f_{R_0}=-5\times10^{-7}\)
500 simulations run with \(f_{R_0}=-5\times10^{-6}\)
500 simulations run with \(f_{R_0}=-5\times10^{-5}\)
500 simulations run with \(f_{R_0}=-5\times10^{-4}\)
Note
We refer the reader to Structure and types for details on the value of the cosmological parameters, the initial conditions…etc.
These simulations are designed for Fisher matrix calculations, and therefore, they have matching IDs between themselves and among other Quijote simulations. We note that to compute generic partial derivatives:
where \(\vec{S}\) is a generic summary statistics and \(f_R\) is the modified gravity parameter, we can use methods like this:
where the fiducial value of \(f_R\) is set to zero.
Important
Note that the chosen values of \(f_{R_0}\) are not distributed equally in both linear and log considering that the fiducial value is \(f_{R_0}=0\). Thus, when performing Fisher matrix calculations, we recommend perform the following change of variables: \(Y=(f_{R_0})^{\log_{10}(2)}\). In that way, the values of \(f_{R_0}\) equal to 0, \(-5\times10^{-7}\), \(-5\times10^{-6}\), \(-5\times10^{-5}\), \(-5\times10^{-4}\), map to \(Y\) equal to 0, -0.0127, -0.0254, -0.0507, -0.101, and the above formulae can easily be used to evaluate \(\partial \vec{S}/\partial Y\).
Simulations for machine learning
In this category we have 2,048 simulations. Each simulation has a different value of the initial random seed and of the parameters \(\Omega_{\rm m}\), \(\Omega_{\rm b}\), \(h\), \(n_s\), \(\sigma_8\), \(M_\nu\), \(f_{R0}\). The value of those parameters in the simulations are organized in a Sobol sequence with boundaries:
Note
The actual value of these parameters for the different simulations can be found here.
Organization
The data is split into different folders:
Snapshots. This folder contains 2,048 subfolders, one for each simulation. Inside these subfolders, the user can find the initial conditions, snapshots, simulation parameters, and additional files produced by MG-Gadget.Halos. This folder contains 2 folders:FoFandRockstar. Each of those folders contains 2,048 folders, inside which the halo catalogs at different redshifts are located.Pk. This folder contains 2,048 subfolders, one for each simulation. Inside these subfolders, the user can find the different power spectra.
Snapshots
Every simulation contains 5 snapshots. Each snapshot is stored in a folder called snapdir_00X, where X=0 is \(z=3\), X=1 is \(z=2\), X=2 is \(z=1\), X=3 is \(z=0.5\), X=4 is \(z=0\). The snapshots are stored in hdf5 format, and can be read using Pylians (see details in Snapshots). Note that the snapshots have been compressed to save space, so please take a look at FAQ if you encounter problems reading them.
Note
The initial conditions are located inside a folder called ICs. The initial conditions are also stored as hdf5 files, and can be read in the same way as the simulation snapshots.
The MG-Gadget snapshots contains more blocks than traditional Gadget N-body simulations. The fields stored in the snapshots are:
/CompressionInfo
/Header
/PartType1
/PartType1/Acceleration
/PartType1/Coordinates
/PartType1/ModifiedGravityAcceleration Dataset
/PartType1/ModifiedGravityGradPhi Dataset
/PartType1/ModifiedGravityPhi Dataset
/PartType1/ParticleIDs
/PartType1/Velocities
/PartType2
/PartType2/Acceleration
/PartType2/Coordinates
/PartType2/ModifiedGravityAcceleration Dataset
/PartType2/ModifiedGravityGradPhi Dataset
/PartType2/ModifiedGravityPhi Dataset
/PartType2/ParticleIDs
/PartType2/Velocities
where PartType1 represent cold dark matter and PartType2 correspond to neutrinos.
Halo catalogs
Quijote-MG contains both FoF and Rockstar halo catalogs for every snapshot of each simulation. You can find details about how to read these files in Halo catalogs.
Power spectra
For every snapshot of each Quijote-MG simulation we have computed the following power spectra:
cold dark matter auto-Pk in real-space:
Pk_CDM_z=X.XXX.datcold dark matter auto-Pk in redshift-space:
Pk_CDM_RS_axis=Y_z=X.XXX.datneutrino auto-Pk in real-space:
Pk_NU_z=X.XXX.datneutrino auto-Pk in redshift-space:
Pk_NU_RS_axis=Y_z=X.XXX.dattotal matter auto-Pk in real-space:
Pk_CDM+NU_z=X.XXX.dattotal matter auto-Pk in redshift-space:
Pk_CDM+NU_RS_axis=Y_z=X.XXX.datCDM-neutrino cross-Pk in real-space:
Pk_CDMNU_z=X.XXX.datCDM-neutrino cross-Pk in redshift-space:
Pk_CDMNU_RS_axis=Y_z=X.XXX.dat
Where X.XXX is the redshift and Y (0, 1, or 2) is the axis along which the redshift-space distortions have been placed.
Bispectra
For every snapshot of each Quijote-MG simulation we have computed the full matter bispectrum. We use a grid with \(384^3\) voxels and we measure the bispectrum in more than 7,000 different triangle configurations. The name of the files is Bk_m_z=X.X.txt, where X.X represents the redshift.