Variability search in M 31 using Principal Component Analysis and the Hubble Source Catalog

TL;DR

This study demonstrates that Principal Component Analysis can effectively identify and classify variable stars in large astronomical datasets, recovering most known variables and discovering new ones in M 31.

Contribution

The paper introduces the use of PCA on variability indices for large photometric datasets to detect and classify variable stars, showing high recovery and discovery rates.

Findings

Recovered over 90% of known variables

Discovered 38 new variable stars, mainly LPVs

Successfully distinguished between different variable star types

Abstract

Principal Component Analysis (PCA) is being extensively used in Astronomy but not yet exhaustively exploited for variability search. The aim of this work is to investigate the effectiveness of using the PCA as a method to search for variable stars in large photometric data sets. We apply PCA to variability indices computed for light curves of 18152 stars in three fields in M 31 extracted from the Hubble Source Catalogue. The projection of the data into the principal components is used as a stellar variability detection and classification tool, capable of distinguishing between RR Lyrae stars, long period variables (LPVs) and non-variables. This projection recovered more than 90% of the known variables and revealed 38 previously unknown variable stars (about 30% more), all LPVs except for one object of uncertain variability type. We conclude that this methodology can indeed successfully identify candidate variable stars.