Multivariate analysis of cosmic void characteristics

TL;DR

This study develops multivariate analysis methods to distinguish genuine cosmic voids from random under-dense regions, improving the purity of void catalogs in galaxy surveys.

Contribution

It introduces a multivariate classification approach trained on simulations to identify true cosmic voids in observational data.

Findings

Void catalog is nearly free of Poisson noise contamination.

Classification efficiency is affected by tracer sparsity and bias.

Method improves the reliability of cosmic void identification.

Abstract

The aim of this study is to distinguish genuine cosmic voids, found in a galaxy catalog by the void finder ZOBOV-VIDE, from under-dense regions in a Poisson distribution of objects. For this purpose, we perform two multivariate analyses using the following physical void characteristics: volume, redshift, density contrast, minimum density, contrast significance and number of member galaxies of the void. The multivariate analyses are trained on a catalog of voids obtained from a random Poisson distribution of points, used as background, and a catalog of voids identified in a mock galaxy catalog, used as signal. The classifications are then applied to voids extracted from the Data Release 12 sample of Luminous Red Galaxies in the redshift range 0.45 $\leq$ z $\leq$ 0.7 from the Sloan Digital Sky Survey Baryon Oscillation Spectroscopic Survey (SDSS BOSS DR12 CMASS). Our results show that the resulting void catalog is nearly free of contamination by Poisson noise. We also study the effect of tracer sparsity and bias on the classification efficiencies.