The data for "Constructing a Mock Galaxy Catalog for the All-Sky Spectroscopic Survey (A-SPEC) Using Machine-assisted Semi-Simulation Model"

This dataset provides the mock galaxy catalog for the All-Sky Spectroscopic Survey (A-SPEC), constructed using the Machine-assisted Semi-Simulation Model (MSSM; Jo & Kim 2019). The MSSM is trained on the IllustrisTNG cosmological magnetohydrodynamical simulation to predict baryonic properties of galaxies from dark-matter-only input features. The model is applied here to NASIM, a custom N-body simulation tailored to the A-SPEC requirements. NASIM is composed of four zoom-in simulations of progressively higher resolution covering the redshift range 0 < z < 0.26, with an observer placed at the center of the simulation box (see Table 1 and Figure 5 of the associated paper). Catalog contents. Each entry is a subhalo with dark matter and predicted baryonic properties defined identically to those in the IllustrisTNG simulation. Subhalos are identified using the SUBFIND algorithm. Dark matter properties Total mass [$h^{-1}\,\rm{M}_\odot$] Redshift (with and without peculiar velocity contribution) Right Ascension, Declination [degree] Subhalo half-mass radius [$h^{-1}\,\rm{Mpc}$] Velocity [$\rm{km}\,\rm{s}^{-1}$] Maximum circular velocity, Velocity dispersion [$\rm{km}\,\rm{s}^{-1}$] Spin [$h^{-1}\,\rm{Mpc} ~ \rm{km}\,\rm{s}^{-1}$] Host halo-related properties: Group number (GroupNr), Rank in group (RankInGr) Baryonic properties Stellar mass [$h^{-1}\,\rm{M}_\odot$] Gas mass [$h^{-1}\,\rm{M}_\odot$] Absolute magnitude in eight photometric bands (U, B, V, K, g, r, i, z) [Vega] Star formation rate averaged over 100 Myr [$\rm{M}_\odot \, \rm{yr}^{-1}$] Mass-weighted gas metallicity Host friends-of-friends halo information (all units are identical to those of the subhalo properties) Halo ID Redshift (with and without peculiar velocity contribution) Right Ascension, Declination Total mass, $M_{200c}$, $M_{500c}$, $M_{200m}$, $R_{200c}$, $R_{500c}$, $R_{200m}$ Number of subhalos File format. Catalogs are provided as HDF5 files (zoom1_mock.hdf5 through zoom4_mock.hdf5), corresponding to the four zoom-in boxes covering different redshift ranges as listed in Table 1 of the associated paper. Notes. We release two versions of the catalog, distinguished by the hyperparameter settings used in the MSSM training. The default version (ASPEC_MSSM_Mock.tar.gz) uses hyperparameters tuned via cross-validation to prevent overfitting (see Section 3.1 of the associated paper). This version is identical to the catalog presented in the paper and is sufficient for general purposes. The second version (ASPEC_MSSM_Mock_NoHyperTuning.tar.gz) uses the default scikit-learn ERT (Extremely Randomized Trees) settings without any hyperparameter tuning. This variant is overfit to the TNG100-1 training data, but produces smoother predictions at the high-luminosity end where the tuned model exhibits an abrupt cutoff. We caution that smoother predictions do not necessarily imply better agreement with observations. Users may wish to apply small offsets or additional Gaussian scatter to the predicted properties depending on their science goal. To reproduce the analysis presented in the paper, use the default version. Contact. For questions or requests for additional simulation outputs, please contact Dongkok Kim ([e-mail]). References. Jo & Kim (2019), MNRAS, 489, 3565. doi: 10.1093/mnras/stz2304 Kim et al. (2026), ApJS. [DOI to be added upon publication]