Data Products

Overview

All simulations in the Kun Universe are generated by the modified version of the Gadget-4 code with 129 different cosmologies. Each simulation evolves \(3072^3\) particles in a box of \(1~h^{-1}\mathrm{Gpc}\) on a side. The massive neutrinos are incoporated through the Newtonian-motion gauge (detailed in Heuschling et al. 2022). All particles are initialized with the second-order Lagrangian Perturbation Theory (2LPT) at the fixed redshift \(z = 127\). An isotropic glass-like distribution is applied to pre-initial loads. At small scales, we fix the equivalent Plummer softening length to \(\epsilon = 8~h^{-1}\mathrm{kpc}\) in comoving coordinates. At large scales, the grid size for Particle-Mesh (PM) force calculation is fixed at \(N_\mathrm{grid}=6144\). The maximum allowed time step is set to \(\mathrm{max}(\Delta \ln a) = 0.04\).

We output 12 snapshots at fixed redshifts of z = {3.00, 2.50, 2.00, 1.75, 1.50, 1.25, 1.00, 0.80, 0.50, 0.25, 0.10, 0.00}. The total storage size for the whole simulation suite is about 3.1 PB. The directory of each simulation output is named as csst/simulation/cXXXX/output/, where XXXX is the cosmology index from 0000 - 0128. In the KunShan cluster, the data is stored in the /public/home/chenzhao/ directory. In the Gravity cluster in the Department of Astronomy, SJTU, the data is stored in the home/cossim/yuyu22/chenzhao/ directory.

The key data products are listed as follows. The corresponding Python analysis scripts is available in the github repo.

Particle Data

Snapshot

Full particle samples at 12 redshifts are saved in the HDF5 format, including positions, velocities, and particle IDs. The positions and velocities are stored with the 32-bit float, and the particle IDs are stored with the 64-bit integer. The directory is named as snapdir_xxx/snapshot_xxx.*.hdf5, where xxx is the snapshot index from 000 - 011 and * is the file index from 0 - 1023. The 1/64 downsampled particle samples are also available in the snapdir_xxx_down64/snapshot_xxx_down64.*.hdf5 files, where xxx is the snapshot index from 000 - 011 and * is the file index from 0 - 1023.

Particle Lightcone

We also save two on-the-fly pyramid particle light-cones up to \(z\leq 3\) by tiling simulation boxes along the three main axes. Each cone covers \(2116\,\mathrm{deg^2}\). Two special line-of-sight directions \(\vec{n}_1 = \{ 0.8678,\ 0.3595,\ 0.3431 \}\) and \(\vec{n}_2 = \{ 0.3595,\ 0.8678,\ 0.3431\}\) are selected to dilute the structure repetitions caused by box replication. Files for particle lightcone are saved in the HDF5 format. The directory is named as lightcone_xx/conedir_yyyy/conesnap_yyyy.*.hdf5, where xx is the lightcone index from 00 - 01, yyyy is the lightcone index starting from 0000, and * is the file index from 0 - 1023.

Halo Data

FOF-SubFind Halo

The Friends-of-Friends (FOF) and SubFind halo/subhalo catalogs are saved in the HDF5 format. We utilize the algorithm integrated in Gadget-4. It is stored in groups_xxx/fof_subhalo_tab_xxx.*.hdf5, where xxx is the snapshot index from 000 - 011 and * is the file index from 0 - 1023.

Rockstar Halo

The Rockstar halo catalogs are saved in both ASCII format and HDF5 format. We utilize the algorithm integrated in Rockstar. The ASCII format is stored in groups_all_rockstar/out_x.list, where x` is the snapshot index from 0 - 11. The HDF5 format is stored in ``groups_xxx_rockstar/rockstar_tab_xxx.*.hdf5, where xxx is the snapshot index from 000 - 011 and * is the file index from 0 - 63. For further information about each halo property (e.g., Mvir_all), we recommand to check the cosmosim site and Rockstar Paper.

Field Data

Matter Density Field

The matter density field is saved in the HDF5 format for each snapshot. The number of mesh grids is \(1536^3\) and window function is the Cloud-in-Cell (CIC) scheme. The files are named as mesh_xxx/mesh_xxx_CIC_Nmesh1536.*.hdf5, where xxx is the snapshot index from 000 - 011 and * is the file index from 0 - 191.

Full-sky Mass Map

The full-sky mass map is saved in the HDF5 format with thickness of \(50~h^{-1}\mathrm{Mpc}\). The files are named as mapsdir_xxx/maps_xxx.*.hdf5, where xxx is the snapshot index starting from 000 but ending at different number for different cosmologies. And * is the file index from 0 - 1023. The map is schemed with the Healpix pixelization with the resolution of \(N_{\mathrm{side}} = 8192\).

Weak Lensing Convergence Map (Born Approximation)

The weak lensing convergence map is saved in the binary format for \(z_s \in [0.5, 3.0]\) with distance interval of \(50~h^{-1}\mathrm{Mpc}\). The files are named as kappas/kappa-nbox*-thickness50-8192.bin, where * represents the used number of shells. The corresponding source redshift is nbox * thickness.

Table of Data Products

Here, we summarize the key data products in the Kun Universe. The Server column indicates the storage location of the data. K represents the KunShan cluster, and G represents the Gravity cluster in SJTU. The data for each simulation is stored in the /public/home/chenzhao/ directory in the KunShan cluster and the home/cossim/yuyu22/chenzhao/ directory in the Gravity cluster. The meanings of x and y in the File Name are shown in the previous sections. The Storage Size column represents the probable storage size for each cosmological model. According to the different cosmologies, the storage size may vary.

Data Products	Server	File Name	Storge Size
Snapshot	K	snapdir_xxx/snapshot_xxx.*.hdf5	~ 6.6 TB
Snapshot 1/64 subsample	K & G	snapdir_xxx_down64/snapshot_xxx_down64.*.hdf5	~ 132 GB
Particle Lightcone	K	lightcone_xx/conedir_yyyy/conesnap_yyyy.*.hdf5	~ 18 TB
FOF-SubFind Halo/Subhalo	K & G	groups_xxx/fof_subhalo_tab_xxx.*.hdf5	~ 100 GB
Rockstar Halo/Subhalo (TXT)	K & G	groups_all_rockstar/out_x.list	~ 300 GB
Rockstar Halo/Subhalo (HDF5)	G	groups_xxx_rockstar/rockstar_tab_xxx.*.hdf5	~ 100 GB
Matter Density Field	K & G	mesh_xxx/mesh_xxx_CIC_Nmesh1536.*.hdf5	~ 156 GB
Full-sky Mass Map	K & G	mapsdir_xxx/maps_xxx.*.hdf5	~ 225 GB
\(\kappa\) Map (Born)	K & G	kappas/kappa-nbox*-thickness50-8192.bin	~ 250 GB

All analysis scripts (Python) is available in the github repo.