Structure in the 3D Galaxy Distribution: I. Methods and Example Results

TL;DR

This paper introduces three novel methods for detecting and characterizing complex structures in 3D galaxy distributions, enabling analysis of clusters, filaments, and sheets across multiple scales.

Contribution

It presents two new methods—classification with self-organizing maps and segmentation with Bayesian blocks—for structure detection in galaxy data, along with a density estimation approach using adaptive kernels.

Findings

Methods successfully identify structures in SDSS data

Effective in detecting filaments and sheets

Demonstrated on simulation and Poisson data

Abstract

Three methods for detecting and characterizing structure in point data, such as that generated by redshift surveys, are described: classification using self-organizing maps, segmentation using Bayesian blocks, and density estimation using adaptive kernels. The first two methods are new, and allow detection and characterization of structures of arbitrary shape and at a wide range of spatial scales. These methods should elucidate not only clusters, but also the more distributed, wide-ranging filaments and sheets, and further allow the possibility of detecting and characterizing an even broader class of shapes. The methods are demonstrated and compared in application to three data sets: a carefully selected volume-limited sample from the Sloan Digital Sky Survey redshift data, a similarly selected sample from the Millennium Simulation, and a set of points independently drawn from a uniform probability distribution -- a so-called Poisson distribution. We demonstrate a few of the many ways in which these methods elucidate large scale structure in the distribution of galaxies in the nearby Universe.