Using Machine Learning to Study the Relationship Between Galaxy Morphology and Evolution
Embargo until
Date
2016-07-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Johns Hopkins University
Abstract
We can track the physical evolution of massive galaxies over time by characterizing the morphological signatures inherent to different mechanisms of galactic assembly. Structural studies rely on a small set of measurements to bin galaxies into disk, spheroid and irregular classifications. These classes are correlated with colors, SF history and stellar masses. Rare and subtle features that are lost in such a generic classification scheme are important for characterizing the evolution of galaxy morphology. We can connect the Hubble sequence observed for local galaxies to their high redshift progenitors to determine the full distribution of galaxy morphologies as a function of time over the entire lifetime of the Universe. To fully capture the complex morphological transformation of galaxies we need more useful classifications. To accomplish such a feat in a computationally tractable way we will need to convert galaxy images to low-dimensional representations of only a few parameters.
To overcome the limitations of the Hubble sequence, we use a principal component analysis of non-parametric morphological indicators (concentration, asymmetry, Gini coefficient, M20, multi-mode, intensity and deviation) measured at rest-frame B-band (corresponding to HST/WFC3 F125W at 1.4 < z < 2) to trace the natural distribution of massive (>10^10 Msun) galaxy morphologies. Principal component analysis (PCA) quantifies the correlations between these morphological indicators and determines the relative importance of each. The first three principal components (PCs) capture ~75% of the variance inherent to our sample. We interpret the first principal component (PC) as bulge strength, the second PC as dominated by concentration and the third PC as dominated by asymmetry. PC1 is a better predictor of quenching than stellar mass, as good as other structural indicators (Sersic-n or compactness). We divide the PCA results into groups using an agglomerative hierarchical clustering method. Distinguishing between these galaxy structural types in a quantitative manner is an important step towards understanding the connections between morphology, galaxy assembly and star-formation.
Using a random forest classification technique, we are able to distinguish mergers from non-merger galaxies in Pan-STARRS imaging using a variety of input features (PCs, non-parametric morphologies, sSFR, M*, rest-frame color). Determining if a galaxy is a merger is important to understand how influential mergers are in building bulges and assembling galaxies. The galaxies were initially visually classified by users of Galaxy Zoo. Asymmetry is by far the most important indicator of whether a galaxy is experiencing a merger. The next most important features include: PC7, PC5, PC3, deviation and d(G,M20). The importance of PC7 represents a very interesting result because PC7 is the least important PC but plays a huge role in determining whether a galaxy is a merger.
Galaxy simulations can provide valuable insight into the mechanisms behind galaxy evolution. The VELA simulations and subsequent non-parametric morphological measurements provide a resource to study the connection between morphology (through the use of PC results) and physical properties (such as sSFR, gas fraction, etc.). We stack the results of a discrete cross correlation between PCs and physical parameters from 9 VELA galaxies. Each of the first three PCs correlates differently with these physical parameters: PC1 is correlated strongly with ex-situ stellar mass, the gas fraction and sSFR; PC2 is weakly anti-correlated with all physical properties; PC3 is strongly correlated with sSFR at all length scales and with gas fraction in the central kpc. The process of star-formation, gas accretion and bulge assembly is a messy picture that will require more simulate galaxies to further understand the process of galaxy evolution.
Description
Keywords
extragalactic astronomy, machine learning, random forest, galaxy morphology, galaxy evolution