Beyond the hubble sequence – exploring galaxy morphology with unsupervised machine learning

ABSTRACT We explore unsupervised machine learning for galaxy morphology analyses using a combination of feature extraction with a vector-quantized variational autoencoder (VQ-VAE) and hierarchical clustering (HC). We propose a new methodology that includes: (1) consideration of the clustering performance simultaneously when learning features from images; (2) allowing for various distance thresholds within the HC algorithm; (3) using the galaxy orientation to determine the number of clusters. This set-up provides 27 clusters created with this unsupervised learning that we show are well separated based on galaxy shape and structure (e.g. Sérsic index, concentration, asymmetry, Gini coefficient). These resulting clusters also correlate well with physical properties such as the colour–magnitude diagram, and span the range of scaling relations such as mass versus size amongst the different machine-defined clusters. When we merge these multiple clusters into two large preliminary clusters to provide a binary classification, an accuracy of $\sim 87{{\ \rm per\ cent}}$ is reached using an imbalanced data set, matching real galaxy distributions, which includes 22.7 per cent early-type galaxies and 77.3 per cent late-type galaxies. Comparing the given clusters with classic Hubble types (ellipticals, lenticulars, early spirals, late spirals, and irregulars), we show that there is an intrinsic vagueness in visual classification systems, in particular galaxies with transitional features such as lenticulars and early spirals. Based on this, the main result in this work is not how well our unsupervised method matches visual classifications and physical properties, but that the method provides an independent classification that may be more physically meaningful than any visually based ones.

Structural diversity of disc galaxies originating in the cold gas inflow from cosmic webs

ABSTRACT Disc galaxies show a large morphological diversity with varying contribution of three major structural components: thin discs, thick discs, and central bulges. Dominance of bulges increases with the galaxy mass (Hubble sequence), whereas thick discs are more prominent in lower mass galaxies. Because galaxies grow with the accretion of matter, this observed variety should reflect diversity in accretion history. On the basis of the prediction by the cold-flow theory for galactic gas accretion and inspired by the results of previous studies, we put a hypothesis that associates different accretion modes with different components. Namely, thin discs form as the shock-heated hot gas in high-mass haloes gradually accretes to the central part, thick discs grow by the direct accretion of cold gas from cosmic webs when the halo mass is low, and finally bulges form by the inflow of cold gas through the shock-heated gas in high-redshift massive haloes. We show that this simple hypothesis reproduces the mean observed variation of galaxy morphology with the galaxy mass. This scenario also predicts that thick discs are older and poorer in metals than thin discs, in agreement with the currently available observations.

The CALIFA view on stellar angular momentum across the Hubble sequence

We present the apparent stellar angular momentum over the optical extent of 300 galaxies across the Hubble sequence using integral-field spectroscopic (IFS) data from the CALIFA survey. Adopting the same λR parameter previously used to distinguish between slow and fast rotating early-type (elliptical and lenticular) galaxies, we show that spiral galaxies are almost all fast rotators, as expected. Given the extent of our data, we provide relations for λR measured in different apertures (e.g. fractions of the effective radius: 0.5 Re, Re, 2 Re), including conversions to long-slit 1D apertures. Our sample displays a wide range of λRe values, consistent with previous IFS studies. The fastest rotators are dominated by relatively massive and highly star-forming Sb galaxies, which preferentially reside in the main star-forming sequence. These galaxies reach λRe values of ∼0.85, and they are the largest galaxies at a given mass, while also displaying some of the strongest stellar population gradients. Compared to the population of S0 galaxies, our findings suggest that fading may not be the dominant mechanism transforming spirals into lenticulars. Interestingly, we find that λRe decreases for late-type Sc and Sd spiral galaxies, with values that occasionally set them in the slow-rotator regime. While for some of them this can be explained by their irregular morphologies and/or face-on configurations, others are edge-on systems with no signs of significant dust obscuration. The latter are typically at the low-mass end, but this does not explain their location in the classical (V/σ, ε) and (λRe, ε) diagrams. Our initial investigations, based on dynamical models, suggest that these are dynamically hot disks, probably influenced by the observed important fraction of dark matter within Re.

Automatic detection of full ring galaxy candidates in SDSS

ABSTRACT A full ring is a form of galaxy morphology that is not associated with a specific stage on the Hubble sequence. Digital sky surveys can collect many millions of galaxy images, and therefore even rare forms of galaxies are expected to be present in relatively large numbers in image data bases created by digital sky surveys. Sloan Digital Sky Survey (SDSS) data release (DR) 14 contains ∼2.6 × 106 objects with spectra identified as galaxies. The method described in this paper applied automatic detection to identify a set of 443 ring galaxy candidates, 104 of them were already included in the Buta + 17 catalogue of ring galaxies in SDSS, but the majority of the galaxies are not included in previous catalogues. Machine analysis cannot yet match the superior pattern recognition abilities of the human brain, and even a small false positive rate makes automatic analysis impractical when scanning through millions of galaxies. Reducing the false positive rate also increases the true negative rate, and therefore the catalogue of ring galaxy candidates is not exhaustive. However, due to its clear advantage in speed, it can provide a large collection of galaxies that can be used for follow-up observations of objects with ring morphology.

The Hubble Sequence at z ∼ 0 in the IllustrisTNG simulation with deep learning

ABSTRACT We analyse the optical morphologies of galaxies in the IllustrisTNG simulation at z ∼ 0 with a convolutional neural network trained on visual morphologies in the Sloan Digital Sky Survey. We generate mock SDSS images of a mass complete sample of $\sim 12\, 000$ galaxies in the simulation using the radiative transfer code SKIRT and include PSF and noise to match the SDSS r-band properties. The images are then processed through the exact same neural network used to estimate SDSS morphologies to classify simulated galaxies in four morphological classes (E, S0/a, Sab, Scd). The CNN model classifies simulated galaxies in one of the four main classes with the same uncertainty as for observed galaxies. The mass–size relations of the simulated galaxies divided by morphological type also reproduce well the slope and the normalization of observed relations which confirms a reasonable diversity of optical morphologies in the TNG suite. However we find a weak correlation between optical morphology and Sersic index in the TNG suite as opposed to SDSS which might require further investigation. The stellar mass functions (SMFs) decomposed into different morphologies still show some discrepancies with observations especially at the high-mass end. We find an overabundance of late-type galaxies ($\sim 50{{\ \rm per\ cent}}$ versus $\sim 20{{\ \rm per\ cent}}$) at the high-mass end [log(M*/M⊙) > 11] of the SMF as compared to observations according to the CNN classifications and a lack of S0 galaxies ($\sim 20{{\ \rm per\ cent}}$ versus $\sim 40{{\ \rm per\ cent}}$) at intermediate masses. This work highlights the importance of detailed comparisons between observations and simulations in comparable conditions.

A study of stellar orbit fractions: simulated IllustrisTNG galaxies compared to CALIFA observations

ABSTRACT Motivated by the recently discovered kinematic ‘Hubble sequence’ shown by the stellar orbit-circularity distribution of 260 CALIFA galaxies, we make use of a comparable galaxy sample at z = 0 with a stellar mass range of $M_{*}/\mathrm{M}_{\odot }\in [10^{9.7},\, 10^{11.4}]$ selected from the IllustrisTNG simulation and study their stellar orbit compositions in relation to a number of other fundamental galaxy properties. We find that the TNG100 simulation broadly reproduces the observed fractions of different orbital components and their stellar mass dependences. In particular, the mean mass dependences of the luminosity fractions for the kinematically warm and hot orbits are well reproduced within model uncertainties of the observed galaxies. The simulation also largely reproduces the observed peak and trough features at $M_{*}\approx 1\rm {-}2\times 10^{10}\, \mathrm{M}_{\odot }$ in the mean distributions of the cold- and hot-orbit fractions, respectively, indicating fewer cooler orbits and more hotter orbits in both more- and less-massive galaxies beyond such a mass range. Several marginal disagreements are seen between the simulation and observations: the average cold-orbit (counter-rotating) fractions of the simulated galaxies below (above) $M_{*}\approx 6\times 10^{10}\, \mathrm{M}_{\odot }$ are systematically higher than the observational data by $\lesssim 10{{\ \rm per\ cent}}$ (absolute orbital fraction); the simulation also seems to produce more scatter for the cold-orbit fraction and less so for the non-cold orbits at any given galaxy mass. Possible causes that stem from the adopted heating mechanisms are discussed.

A comprehensive examination of the optical morphologies of 719 isolated galaxies in the AMIGA sample

Using images from Sloan Digital Sky Survey Data Release 8, we have re-examined the morphology of 719 galaxies from the Analysis of the interstellar Medium in Isolated GAlaxies (AMIGA) project, a sample consisting of the most isolated galaxies that have yet been identified. The goal is to further improve the classifications of these galaxies by examining them in the context of the Comprehensive de Vaucouleurs revised Hubble-Sandage (CVRHS) system, which includes recognition of features that go beyond the original de Vaucouleurs point of view. Our results confirm previous findings that isolated galaxies are found across the complete revised Hubble sequence, with intermediate- to late-type (Sb-Sc) spirals being relatively more common. Elmegreen Arm Classifications are also presented, and show that more than 50 per cent of the 514 spirals in the sample for which an arm class (AC) could be judged are grand design (AC 8,9,12). The visual bar fraction for the sample is ≈50 per cent, but only 16 per cent are classified as strongly barred (SB). The dominant family classification is SA (non-barred), the dominant inner variety classification is (s) (pure spiral), and the dominant outer variety classification is no outer ring, pseudo-ring, or lens. The Kolmogorov–Smirnov test is used to check for potential biases in the morphological interpretations, and for any possible relation between rings, bars, and ACs with local environment and far-infrared excess. The connection between morphology and stellar mass is also examined for a subset of the sample.

Near-infrared spectroscopic indices for unresolved stellar populations

Context. A new generation of spectral synthesis models has been developed in recent years, but there is no matching set of template galaxy spectra, in terms of quality and resolution, for testing and refining the new models. Aims. Our main goal is to find and calibrate new near-infrared spectral indices along the Hubble sequence of galaxies which will be used to obtain additional constraints to the population analysis based on medium-resolution integrated spectra of galaxies. Methods. Spectra of previously studied and well-understood galaxies with relatively simple stellar populations (e.g., ellipticals or bulge dominated galaxies) are needed to provide a baseline data set for spectral synthesis models. Results. X-shooter spectra spanning the optical and infrared wavelengths (350–2400 nm) of bright nearby elliptical galaxies with a resolving power of R ∼ 4000–5400 were obtained. Heliocentric systemic velocity, velocity dispersion, and Mg, Fe, and Hβ line-strength indices are presented. Conclusions. We present a library of very-high-quality spectra of galaxies covering a large range of age, metallicity, and morphological type. Such a dataset of spectra will be crucial to addressing important questions of the modern investigation concerning galaxy formation and evolution.