paper

A universal and physically motivated threshold for Hessian-based cosmic web identification: V-Web case

arXiv:2412.09531 · doi:10.1051/0004-6361/202453376

Abstract

The study of large-scale structure can benefit from accurate and robust identification of the cosmic web. Having such classification can facilitate a more complete extraction of cosmological information encoded therein. Classification methods like T-web and V-web, based on the Hessian matrix, are widely used to signal-out voids, sheets, filaments, and knots. However, these techniques depend on a threshold parameter which value is chosen without physical justification, usually relying on a user visual impression, thus limiting the universality of results. In this paper we focus on the V-web method. Our aim is to find a physical motivation for deriving an universal threshold that can be applied across different cosmic scales and epochs. V-web classify the large-scale structure using the eigenvalues of the velocity shear tensor. Using a set of gravity-only simulations we introduce a normalization that incorporates the standard deviation of the velocity divergence field, isolating the beyond Gaussian evolution of cosmic web elements. In the Zeldovich's approximation, the probability presence of each cosmic web element remains constant at a threshold equal to 0. For the first time, we reveal that this behavior also holds in the non-linear regime for a normalized positive 'constant volume threshold' that depends on both the redshift and the applied smoothing scale. The conservation of volume fractions is valid for the studied redshifts between 0 and 2, regardless of cosmic variance, and is most precise for intermediate smoothing scales around 3 Mpc/h. The properties of the cosmic web derived using this approach in the V-web align with expectations from other methods, including visual impressions. We provide a general fit formula to compute the constant volume threshold for any standard cosmological simulation, regardless of its specific properties.

Submitted for publication in A&A

References in corpus (44)

Cited by in corpus (1)