Sahyadri: A simulation suite for the cosmology dependence of the Cosmic Web

TL;DR

Sahyadri addresses the need for high-mass-resolution, cosmology-derivative simulations at low redshift to exploit non-linear clustering in upcoming surveys. It achieves this with a $200\,h^{-1}\mathrm{Mpc}$ box containing $2048^3$ particles, resolving halos down to $M_{\min} = 3.2\times10^9\,h^{-1}M_\odot$ and varying six cosmological parameters around Planck 2018 with seed-matched initial conditions to enable derivatives. The suite demonstrates accurate matter and halo clustering predictions, and explores beyond-2-point statistics such as the Voronoi volume function and $k$NN distributions, finding strong sensitivity to $\Omega_m$ and enabling new insights into assembly bias via tidal anisotropy and halo environment. A custom compression scheme reduces storage by ~3x without sacrificing clustering accuracy, and data products including halo and value-added catalogs will be publicly available, supporting Fisher analyses and cosmological inference across non-linear scales. Overall, Sahyadri fills a critical gap between mass resolution and cosmology coverage, enabling detailed modeling of low-redshift galaxies and the cosmic web for DESI, 4MOST, and similar surveys.

Abstract

We present Sahyadri, a suite of cosmological $N$-body simulations designed to enable precision studies of the low-redshift Universe with next-generation spectroscopic surveys. Sahyadri includes systematic variations of six cosmological parameters around Planck 2018 constraints, with seed-matched initial conditions enabling cosmological parameter derivatives. Each simulation evolves $2048^3$ particles in a periodic box of side length $200$ $h^{-1}$ Mpc, yielding a particle mass of $m_{\rm{p}} = 8.1 \times 10^{7}\,h^{-1}\,M_{\odot}$ in the fiducial Planck 2018 cosmology. This resolution enables robust identification of dark matter halos down to $M_{\rm min} = 3.2 \times 10^{9}$ $h^{-1}$ $M_\odot$, which represents a factor of $\sim$25 improvement over the AbacusSummit suite, and is over two orders of magnitude better than the Quijote and Aemulus suites. We estimate that approximately 40% of DESI BGS galaxies at redshift $z < 0.15$ - roughly 1.6 million objects - reside in halos accessible to Sahyadri but beyond the reach of existing parameter-varying simulation suites. We demonstrate Sahyadri's capabilities through measurements of the matter power spectrum, halo mass function and power spectrum, and beyond 2-point statistics such as the Voronoi volume function and $k^{\rm th}$ nearest neighbour statistics, showing excellent agreement with theoretical predictions and significant sensitivity to $Ω_{\rm m}$ variations. We implement a custom compression scheme reducing storage requirements by a factor of $\sim$3 while maintaining sub-percent clustering accuracy. Key data products will be made publicly available.

Figures (17)

• Figure 1: Stellar mass function using a stellar-to-halo mass relation and a conditional $r$-band magnitude distribution at redshift $z=0.15$ based on Alam et al. (in prep). We show that the default Sahyadri simulation allows us to model much lower stellar masses as compared to AbacusSummit which misses $\sim 40\%$ of galaxies in BGS at these low redshifts.
• Figure 2: Comparison of the Sahyadri simulation suite with other large-scale structure simulation efforts. Grey lines indicate simulations with constant particle number assuming Planck 2018 cosmology. Coloured points highlight suites with cosmology variations. The red (black) star marks the Sahyadri (Sinhagad) simulation configuration, highlighting the state-of-the-art mass resolution achieved by our suite.
• Figure 3: Visualization of the evolution of the dark matter density field as a function of redshift for the fiducial cosmology. The density field is evaluated on a $1024^3$ grid and smoothed with a Gaussian kernel of radius $185\,h^{-1}{\rm kpc}$. The left, middle and right columns correspond to the same single-cell spatial slices at redshift $z=5.6, 1, 0$ respectively. The top panel shows a slice through the entire simulation box containing the most massive halo at $z=0$. The middle panel provides a zoom-in on this halo, while the bottom panel shows the evolution of a region that later develops into a prominent filament. The centers of the slices are indicated in each column. Each panel’s $x$- and $y$-axes span the same length, ensuring uniform tick mark size in the $x$ and $y$ directions.
• Figure 4: Comparison between AbacusSummit-like and Sahyadri halos. Right panel shows the highest number density halo sample used in this paper, with $n=2\times 10^{-2} \rm{Mpc}^{-3}$ (see section \ref{['subsec:halo_samples']} for the selection criterion), from the default Sahyadri simulation at $z=0$. Left panel shows the sample of halos from the same simulation, but after degrading the particle mass to that of AbacusSummit and applying the same selection criterion, leading to $n=6.5\times 10^{-4} \rm{Mpc}^{-3}$. Both panels display the same spatial slice as the top right panel of Figure \ref{['fig:z_comparison']}, including halos in this slice and the two neighboring slices, overplotted on the underlying density field. Circles mark halo positions, with radii equal to $4R_{\rm{200b}}$, and are colored by $\log_{10}(\alpha)$, where $\alpha$ is the tidal anisotropy defined in equation \ref{['eq:alpha-def']}. The $x$- and $y$-axes span the same length, ensuring uniform tick mark size in each direction. The higher resolution of Sahyadri allows significantly more low-mass halos to be resolved, improving the sampling of all cosmic environments.
• Figure 5: Same as Figure \ref{['fig:alpha_comparison']}, but with halos are coloured by the halo-by-halo bias $b_1$. We see that the most massive halo does not necessarily correspond to the largest bias, which instead traces the larger-scale density field, as established in the literature. See text for a discussion.