Investigation of the Effect of Bars on the Properties of Spiral Galaxies: A Multivariate Statistical Study

TL;DR

This study uses multivariate statistical methods to classify a large sample of spiral galaxies, revealing correlations between bars and galaxy properties, and identifying distinct age-based groups with recurrent bar formation evidence.

Contribution

Introduces an objective, multivariate classification of over 26,000 spiral galaxies using ICA and clustering, uncovering new insights into bar effects and galaxy groupings.

Findings

Identification of 12 distinct galaxy groups with age and bar characteristics.

Evidence of recurrent bar formation phenomena in multiple groups.

Robustness of clustering confirmed by Gaussian Mixture Modeling.

Abstract

Subjective classification of spiral galaxies is not sufficient for studying the effect of bars on their physical characteristics. In reality the problem is to comprehend the complex correlations in a multivariate parametric space. Multivariate tools are the best ones for understanding this complex correlation. In this work an objective classification of a large set (26,089) of spiral galaxies was compiled as a value added galaxy catalogue from sdss DR 15 virtual data archive. Initially for dimensionality reduction, Independent Component Analysis is performed to determine a set of Independent Components that are linear combinations of 48 observed features (namely ionised lines, Lick indices, photometric and morphological properties). Subsequently a K-means cluster analysis is carried out on the basis of the 14 best chosen Independent Components to obtain 12 distinct homogeneous groups of spiral galaxies. Amongst these, 3 groups are the oldest ones (1.6 Gyr - 5.9 Gyr), while 5 groups fall in the medium aged category (1.4 Gyr - 1.6 Gyr), 2 groups consist of only unbarred spirals, 1 group is the youngest one and the remaining one is an outlier. In many groups there are clear indication of recurrent bar formation phenomena which is consistent with few previous simulation works. In order to study the robustness a second method of clustering by Gaussian Mixture Modeling Method (GMMBC) is applied.