Using Morphology to understand Galaxy Evolution across the Dwarf and Massive-Galaxy Regimes

The morphological properties of galaxies are important tracers of the physical processes, e.g. minor/major
\nmergers, gas accretion and tidal interactions, that have shaped their evolution. This thesis first studies
\nthe morphological and structural properties of dwarf galaxies (which remain relatively unexplored
\ncompared to massive galaxies). It then explores the measurement of morphology using unsupervised
\nmachine learning (UML) algorithms, which will become important in the forthcoming era of peta and
\nexascale surveys that will dominate the astrophysical landscape in the future.
\nI use a complete, unbiased sample of 257 dwarf (108 M⊙ < M⋆ < 109.5 M⊙) galaxies at z < 0.08, in the
\nCOSMOS field, to study the morphological mix of the dwarf population in low-density environments.
\nVisual inspection of extremely deep optical images reveals three principal morphological classes: 43
\nand 45 per cent of dwarfs exhibit the traditional ‘early-type’ (elliptical/S0) and ‘late-type’ (spiral) morphologies
\nrespectively, while 10 per cent populate a ‘featureless’ class, that lacks both the central light
\nconcentration seen in early-types and any spiral structure. This class is missing in the massive-galaxy
\nregime. Compared to their massive counterparts, dwarf early-types show a much lower incidence of
\ninteractions, are significantly less concentrated and share similar rest-frame colours as dwarf late-types.
\nThis suggests that the formation histories of early-types is different in the dwarf and massive regimes.
\nWhile massive early-types are likely shaped by interactions, their dwarf counterparts are shaped more
\nby secular processes. This study suggests that featureless dwarfs in low-density environments are created
\nvia internal baryonic feedback, rather than by environmental processes. Finally, I show that while
\ninteracting dwarfs can be identified using the asymmetry parameter, it is challenging to cleanly separate
\nearly and late-type dwarfs using traditional morphological parameters, such as ‘CAS’, M20 and the Gini
\ncoefficient (unlike in the massive-galaxy regime).
\nI proceed by using 211 out of the 257 dwarf described above to study their structural properties: effective
\nradii (Re), effective surface brightnesses (μe) and colour profiles and gradients. I explore these properties
\nas a function of stellar mass and the three principal dwarf morphological types (early-type galaxies
\n(ETGs), late-type galaxies (LTGs) and featureless systems). The median effective radii of LTGs and
\nfeatureless galaxies are factors of ∼2 and ∼1.2 larger than the ETGs. While the median effective brightness
\nof the ETGs and LTGs is similar, the featureless class is much fainter. Dwarfs residing within the
\n‘UDG’ region in μe vs Re space are a continuous extension of the galaxy population towards lower values
\nof effective brightness, indicating that UDGs are not a distinct class of objects. While massive ETGs
\ntypically exhibit negative or flat colour gradients, dwarf ETGs show positive colour gradients (bluer
\ncentres). The growth of ETGs changes from being ‘outside-in’ to ‘inside out’ as I move from the dwarf
\nto the massive regime. The colour gradients of dwarf and massive LTGs are, however, similar. I show
\nthat, compared to their non-interacting counterparts, interacting dwarfs are larger, bluer at all radii and
\nexhibit similar median effective surface brightness, suggesting that interactions enhance star formation
\nacross the entire galaxy.
\nForthcoming ‘Big data’ surveys (e.g. LSST/SKA), which will produce peta and exabyte volumes of
\ndata will become the new ‘normal’ in the coming decades. These volumes will make morphological
\nclassification using traditional methods (e.g. direct visual inspection, as used in the first two chapters)
\nimpractical. Even semi-automated techniques, e.g. supervised machine learning with training sets built
\nvia visual inspection, may be difficult, because of the time-consuming nature of creating these sets.
\nHowever, UML, does not require training on labelled data, making it ideal for galaxy morphological
\nclassification for large surveys.
\nHere I present an UML algorithm, that utilizes hierarchical clustering and growing neural gas networks
\nto group together survey image patches with similar visual properties, followed by a clustering of objects
\n(e.g. galaxies) that are reconstructed from these patches. I implement the algorithm on the Deep layer
\n(27 deg2) of the Hyper Suprime-Cam Subaru-Strategic-Program, to reduce a population of hundreds of
\nthousands of galaxies to a small number (∼150) of morphological clusters, which exhibit high purity.
\nThese clusters, rather than individual galaxies, can then be rapidly benchmarked via visual inspection and
\nclassified into typical morphological types. Using these morphological clusters, I successfully reproduce
\nknown trends of galaxy properties as a function of morphological type, which demonstrates the efficacy
\nof the method.
\nFinally, I study 108 blue, low-mass ellipticals (which have a median stellar mass of 108.7 M⊙) at z <
\n0.3 in the COSMOS field, which have been autonomously detected by the algorithm described above.
\nHow elliptical galaxies form is a key question in observational cosmology. While the formation of
\nmassive ellipticals is strongly linked to mergers, the low mass (M⋆/M⊙ < 109.5) regime remains less
\nwell explored. In particular, studying elliptical populations when they are blue, and therefore rapidly
\nbuilding stellar mass, offers strong constraints on their formation. Visual inspection of the deep optical
\nHSC images indicates that less than 3 per cent of these systems have visible tidal features, a factor of 2
\nless than the incidence of tidal features in a control sample of galaxies with the same distribution of stellar
\nmass and redshift. This suggests that the star formation activity in these objects is not driven by mergers
\nor interactions but by secular gas accretion (i.e. steady gas accretion over Gyr timescales). I show
\nthat these blue ellipticals reside in low-density environments, further away from nodes and large-scale
\nfilaments than other galaxies. At similar stellar masses and environments, blue ellipticals outnumber
\ntheir normal (red) counterparts by a factor of 2. Thus, these systems are likely progenitors of not only
\nnormal ellipticals at similar stellar mass but, given their high star formation rates, also of ellipticals at
\nhigher stellar masses. Secular gas accretion, therefore, likely plays a significant (and possibly dominant)
\nrole in the stellar assembly of elliptical galaxies in the low mass regime.