AbacusSummit DR1 Mock Catalogues

• AbacusSummit DR1 Mock Catalogues are a comprehensive simulation suite offering high-resolution synthetic datasets for galaxies, halos, and large-scale structures.
• They employ advanced N-body techniques, rigorous halo finding, and merger tree construction to model nonlinear cosmic evolution with sub-percent accuracy.
• The catalogues support calibration of next-generation surveys by accurately capturing observables such as clustering, BAO, lensing, and the Lyman α forest.

The AbacusSummit DR1 Mock Catalogues are a comprehensive suite of high-fidelity synthetic galaxy, halo, and large-scale structure datasets generated from the AbacusSummit $N$-body simulation campaign. These catalogues are designed for large-scale cosmological analyses, especially targeting the calibration and validation needs of next-generation spectroscopic surveys such as DESI, Euclid, and LSST. They provide multi-epoch, multi-cosmology realizations of dark matter evolution, halo populations, merger histories, light cone products, and derived mocks of galaxies and the Lyman $\alpha$ forest. The datasets underpin the precision modeling of observables such as clustering multipoles, BAO measurements, cosmic shear, and lensing signals, while offering sub-percent accuracy and exceptional scalability.

1. Simulation Construction and Fundamental Properties

AbacusSummit DR1 consists of over 60 trillion particles evolved across 139 “base” simulations ($L = 2\,h^{-1}$Gpc, $N_\mathrm{part} = 6912^3$, $m_p = 2 \times 10^9\,h^{-1}M_\odot$) and several auxiliary suites in varied box sizes and cosmological models (Maksimova et al., 2021). The code employs exact near-field summation and Taylor-expanded far-field components, achieving median fractional force errors of $\mathcal{O}(10^{-5})$. Simulations span 97 cosmological models including the Planck 2018 $\Lambda$CDM baseline, $w_0w_a$CDM variants, altered neutrino physics, thawing dark energy, and reduced clustering amplitude scenarios (Gsponer et al., 9 Sep 2025).

Time-stepping employs a global second-order leapfrog scheme ($\Delta\ln a$), with output epochs (33 per box) ranging from $z=8$ to $z=0.1$. Products include halo catalogues (CompaSO/L1+L2), merger trees, particle subsamples, and full and light cone snapshots. A covariance sub-suite of 1883 small boxes ($L = 500\,h^{-1}$Mpc) is provided for precise error estimation. Data access is managed via OLCF/Constellation portal (DOI: 10.13139/OLCF/1811689).

2. Halo Catalogues and Merger Trees

Halo identification in AbacusSummit DR1 is performed by the CompaSO finder, which robustly tracks disjoint parent halos and associated subhalos. Merger trees are constructed by matching core particle subsets across epochs, using center-of-mass positions and unique particle IDs. The HaloIndex assigns a persistent identifier:

$$\text{HaloIndex} = 10^{12} \times \text{FullStepNumber} + 10^{9} \times \text{SuperslabNumber} + \text{IndexInSuperslabFile}$$

Allowing for robust progenitor-descendant relationships, trees reveal mass accretion histories (MAH), conditional mass functions, and formation redshifts within a few percent agreement over variable snapshot cadences (Bose et al., 2021). Post-processing “cleaning via merger trees” removes halos with non-monotonic mass evolution—using a threshold $\kappa$ ($M_\text{peak}/M(z_i) > 2$)—and reduces artificial clustering excesses and unphysical satellite populations, thus improving mock fidelity for HOD population models.

3. Light Cone Interpolation, Weak Lensing, and Galaxy Mock Catalogues

Halo light cone catalogues are constructed by interpolating halo trajectories using merger trees, aligning halo positions and velocities to the instant they cross the observer’s light cone. Interpolation is performed using: $$q_\text{interp} = q_i + \left(\frac{\chi_* - \chi_i}{\chi_{i+1} - \chi_i}\right) (q_{i+1} - q_i)$$ for properties $q$ (position, velocity, mass) between snapshots. For halo masses $M_\text{halo} > 2.1 \times 10^{11}\,M_\odot/h$ ($\geq 100$ particles), tracking and interpolation accuracy is robust (Hadzhiyska et al., 2021).

Once light cone haloes are generated, galaxies are assigned via an extended AbacusHOD model featuring additional parameters for satellite distributions, velocity and central galaxy bias, and assembly bias. HOD parameters are empirically calibrated using fits to simulation snapshots, hydrodynamical results, and survey clustering data. Weak lensing maps (shear, convergence, deflection) are constructed by projecting density fields onto HEALPix grids ($N_\mathrm{side} = 16384$) and integrating using the Born approximation: $$\kappa(\theta) = \frac{3H_0^2\Omega_m}{2c^2} \int d\chi\,\delta(\chi,\theta) \frac{(\chi_s-\chi)\chi}{\chi_s a}$$ CMB convergence maps use $z_\mathrm{source} \approx 1090$ and match theoretical predictions at the sub-percent level (Hadzhiyska et al., 2023).

4. Statistical Completeness and Covariance Suite

The covariance sub-suite facilitates robust error estimation for clustering statistics. Each realization is independent (periodic boundaries) and matches the main suite in mass resolution and cosmology, enabling construction of sample covariance matrices for power spectrum $P(k)$, $N$-point statistics, lensing observables, and BAO peak calibrations.

Dedicated DR1 mocks, including galaxy samples for DESI-BGS and Euclid, employ HOD fitting against number density and correlation function benchmarks. For DESI-BGS, the HOD for central/satellite populations is parameterized as a function of absolute magnitude threshold $M_\mathrm{r}$, fitted against MXXL-based mocks using likelihood: $$\mathcal{L}(M) = -\frac{1}{2} \sum_r \left[ \left(\frac{\xi(r, M) - \xi_\mathrm{target}(r)}{\sigma_\xi(r)}\right)^2 + \left(\frac{n(M) - n_\mathrm{target}(M)}{\sigma_n}\right)^2 \right]$$ Rapid evaluation via halo tabulation of pair counts enables efficient HOD parameter emulation across cosmology grid points (Smith et al., 2023).

5. Lyman $\alpha$ Forest and QSO Mock Catalogues

For Lyman $\alpha$ forest analyses, AbacusSummit facilitates the planting of high-resolution Ly$\alpha$ skewers through large simulation volumes. Optical depth is assigned via the modified Fluctuating Gunn–Peterson Approximation (FGPA),

$\tau = A \left(\frac{\rho}{\bar{\rho}}\right)^\eta$

calibrated to hydrodynamic statistics (IllustrisTNG) to reproduce 1D/3D flux power spectra within ${\lesssim}10\%$ up to $k < 2\,h/$Mpc (Hadzhiyska et al., 2023). QSO mock catalogues are drawn to match DESI selection, supporting cross-correlation studies with Ly$\alpha$ forest skewers. The mocks exhibit non-linear broadening of the BAO peak beyond the Kaiser approximation, highlighting the impact of redshift-space distortion modelling for cross-correlations and the need for refined theoretical templates in full-shape analyses.

6. Fiducial Cosmology Systematics and Impact on Cosmological Inference

The suite is pivotal in quantifying systematics in cosmological parameter inference—arising when measured redshifts and angles are converted to comoving separation under an assumed fiducial cosmology. The paper of (Gsponer et al., 9 Sep 2025) employs the formulae: $$q_\perp(z) = \frac{D_m(z)}{D_m^\mathrm{grid}(z)},\quad q_\parallel(z) = \frac{D_H(z)}{D_H^\mathrm{grid}(z)}$$ in FM approaches, and,

$$\alpha_\perp(z) = \frac{D_m(z) r_d^\mathrm{temp}}{D_m^\mathrm{grid}(z) r_d},\quad \alpha_\parallel(z) = \frac{D_H(z) r_d^\mathrm{temp}}{D_H^\mathrm{grid}(z) r_d}$$

in SF approaches, to correct for scaling mismatches. The systematic shift $$N_{\sigma_\mathrm{DR1}} = |\Delta x|/\sigma_\mathrm{DR1}$$ quantifies parameter deviation. Across $\Lambda$CDM, $w_0w_a$CDM, thawing-DE, high-$N_\mathrm{eff}$, and low-$\sigma_8$ models, recovered shifts are $\lesssim 0.22\sigma_{\mathrm{DR1}}$ (FM) and $\lesssim 0.45\sigma_{\mathrm{DR1}}$ (SF), ensuring full-shape cosmological inference is robust to moderate fiducial cosmology mismatches.

7. Comparison With Perturbative and Approximate Mock Techniques

AbacusSummit DR1 mocks are distinguished from fast approximate methods such as EZmocks (Chuang et al., 2014), PINOCCHIO (Collaboration et al., 16 Jul 2025), and halo catalogues based on 2LPT or parametric prescriptions (Yapici et al., 2019) by their ability to capture detailed nonlinear structure evolution, redshift-space distortions, and halo assembly bias. While EZmocks and similar approaches achieve percent-level accuracy in two-point and three-point statistics up to $k \sim 0.55\,h/$Mpc and on scales $r \gtrsim 10\,h^{-1}$Mpc, they sacrifice small-scale fidelity and velocity correlation accuracy. AbacusSummit provides a reference for the validation and calibration of these fast mocks, which remain critical for covariance estimation and pipeline development due to their computational efficiency.

8. Scientific Applications and Public Availability

The catalogues support a range of analyses: galaxy clustering, BAO fitting, cosmic shear/dark energy constraints, lensing forecasts, Ly$\alpha$ forest and QSO cross-correlations, systematics studies, and emulator training. Data products (halo catalogues, light cones, merger trees, lensing maps) are hierarchically organized and compressed for efficient access, with open release via OLCF and Globus documented in (Maksimova et al., 2021, Hadzhiyska et al., 2021, Hadzhiyska et al., 2023).

Validation across particle subsamples, clustering statistics, and lensing angular power spectra shows excellent agreement with theoretical predictions and compatible results against hydrodynamical and Flagship mocks (Collaboration et al., 16 Jul 2025). Systematic effects induced by numerical artifacts, halo finder choices, or cosmology conversion mismatches have been thoroughly investigated and found to be subdominant compared to the statistical precision expected in DESI DR1 and Euclid DR1.

9. Summary

AbacusSummit DR1 Mock Catalogues offer a reference-grade, high-resolution $N$-body simulation dataset that sets the standard for multi-epoch, multi-cosmology, synthetic structure catalogues. By capturing the full nonlinear dynamics of cosmic structure formation and enabling calibrated, validated mocks for galaxies, Ly$\alpha$ forest skewers, and lensing observables, AbacusSummit DR1 is an essential resource for precision cosmological analyses. Its extensive ensemble, accuracy, public availability, and compatibility with both full-modelling and compressed analysis techniques ensure ongoing relevance for the cosmological survey community.