Galaxy mergers are crucial to understanding galaxy evolution, therefore we must determine their observational signatures to select them from large IFU galaxy samples such as MUSE and SAMI. We employ 24 high-resolution idealized hydrodynamical galaxy merger simulations based on the ‘Feedback In Realistic Environment’ (FIRE-2) model to determine the observability of mergers to various configurations and stages using synthetic images and velocity maps. Our mergers cover a range of orbital configurations at fixed 1:2.5 stellar mass ratio for two gas rich spirals at low redshift. Morphological and kinematic asymmetries are computed for synthetic images and velocity maps spanning each interaction. We divide the interaction sequence into three: (1) the pair phase; (2) the merging phase; and (3) the post-coalescence phase. We correctly identify mergers between first pericentre passage and 500 Myr after coalescence using kinematic asymmetry with 66 per cent completeness, depending upon merger phase and the field of view of the observation. We detect fewer mergers in the pair phase (40 per cent) and many more in the merging and post-coalescence phases (97 per cent). We find that merger detectability decreases with field of view, except in retrograde mergers, where centrally concentrated asymmetric kinematic features enhances their detectability. Using a cut-off derived from a combination of photometric and kinematic asymmetry, we increase these detections to 89 per cent overall, 79 per cent in pairs, and close to 100 per cent in the merging and post-coalescent phases. By using this combined asymmetry cut-off we mitigate some of the effects caused by smaller fields of view subtended by massively multiplexed integral field spectroscopy programmes.
Machine learning is becoming a popular tool to quantify galaxy morphologies and identify mergers. However, this technique relies on using an appropriate set of training data to be successful. By combining hydrodynamical simulations, synthetic observations, and convolutional neural networks (CNNs), we quantitatively assess how realistic simulated galaxy images must be in order to reliably classify mergers. Specifically, we compare the performance of CNNs trained with two types of galaxy images, stellar maps and dust-inclusive radiatively transferred images, each with three levels of observational realism: (1) no observational effects (idealized images), (2) realistic sky and point spread function (semirealistic images), and (3) insertion into a real sky image (fully realistic images). We find that networks trained on either idealized or semireal images have poor performance when applied to survey-realistic images. In contrast, networks trained on fully realistic images achieve 87.1 per cent classification performance. Importantly, the level of realism in the training images is much more important than whether the images included radiative transfer, or simply used the stellar maps ($87.1{{\ \rm per\ cent}}$ compared to $79.6{{\ \rm per\ cent}}$ accuracy, respectively). Therefore, one can avoid the large computational and storage cost of running radiative transfer with a relatively modest compromise in classification performance. Making photometry-based networks insensitive to colour incurs a very mild penalty to performance with survey-realistic data ($86.0{{\ \rm per\ cent}}$ with r-only compared to $87.1{{\ \rm per\ cent}}$ with gri). This result demonstrates that while colour can be exploited by colour-sensitive networks, it is not necessary to achieve high accuracy and so can be avoided if desired. We provide the public release of our statistical observational realism suite, RealSim, as a companion to this paper.
more » « less- NSF-PAR ID:
- 10124105
- Publisher / Repository:
- Oxford University Press
- Date Published:
- Journal Name:
- Monthly Notices of the Royal Astronomical Society
- Volume:
- 490
- Issue:
- 4
- ISSN:
- 0035-8711
- Page Range / eLocation ID:
- p. 5390-5413
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
ABSTRACT -
more » « less
ABSTRACT Supermassive black holes require a reservoir of cold gas at the centre of their host galaxy in order to accrete and shine as active galactic nuclei (AGN). Major mergers have the ability to drive gas rapidly inwards, but observations trying to link mergers with AGN have found mixed results due to the difficulty of consistently identifying galaxy mergers in surveys. This study applies deep learning to this problem, using convolutional neural networks trained to identify simulated post-merger galaxies from survey-realistic imaging. This provides a fast and repeatable alternative to human visual inspection. Using this tool, we examine a sample of ∼8500 Seyfert 2 galaxies ($L[\mathrm{O\, {\small III}}] \sim 10^{38.5 - 42}$ erg s−1) at z < 0.3 in the Sloan Digital Sky Survey and find a merger fraction of $2.19_{-0.17}^{+0.21}$ per cent compared with inactive control galaxies, in which we find a merger fraction of $2.96_{-0.20}^{+0.26}$ per cent, indicating an overall lack of mergers among AGN hosts compared with controls. However, matching the controls to the AGN hosts in stellar mass and star formation rate reveals that AGN hosts in the star-forming blue cloud exhibit a ∼2 × merger enhancement over controls, while those in the quiescent red sequence have significantly lower relative merger fractions, leading to the observed overall deficit due to the differing M*–SFR distributions. We conclude that while mergers are not the dominant trigger of all low-luminosity, obscured AGN activity in the nearby Universe, they are more important to AGN fuelling in galaxies with higher cold gas mass fractions as traced through star formation.
-
more » « less
Abstract Giant star-forming clumps (GSFCs) are areas of intensive star-formation that are commonly observed in high-redshift (z ≳ 1) galaxies but their formation and role in galaxy evolution remain unclear. Observations of low-redshift clumpy galaxy analogues are rare but the availability of wide-field galaxy survey data makes the detection of large clumpy galaxy samples much more feasible. Deep Learning (DL), and in particular Convolutional Neural Networks (CNNs), have been successfully applied to image classification tasks in astrophysical data analysis. However, one application of DL that remains relatively unexplored is that of automatically identifying and localizing specific objects or features in astrophysical imaging data. In this paper, we demonstrate the use of DL-based object detection models to localize GSFCs in astrophysical imaging data. We apply the Faster Region-based Convolutional Neural Network object detection framework (FRCNN) to identify GSFCs in low-redshift (z ≲ 0.3) galaxies. Unlike other studies, we train different FRCNN models on observational data that was collected by the Sloan Digital Sky Survey and labelled by volunteers from the citizen science project ‘Galaxy Zoo: Clump Scout’. The FRCNN model relies on a CNN component as a ‘backbone’ feature extractor. We show that CNNs, that have been pre-trained for image classification using astrophysical images, outperform those that have been pre-trained on terrestrial images. In particular, we compare a domain-specific CNN – ‘Zoobot’ – with a generic classification backbone and find that Zoobot achieves higher detection performance. Our final model is capable of producing GSFC detections with a completeness and purity of ≥0.8 while only being trained on ∼5000 galaxy images.
-
more » « less
ABSTRACT Hydrogen emission lines can provide extensive information about star-forming galaxies in both the local and high-redshift Universe. We present a detailed Lyman continuum (LyC), Lyman-α (Lyα), and Balmer line (Hα and Hβ) radiative transfer study of a high-resolution isolated Milky Way simulation using the state-of-the-art Arepo-RT radiation hydrodynamics code with the SMUGGLE galaxy formation model. The realistic framework includes stellar feedback, non-equilibrium thermochemistry accounting for molecular hydrogen, and dust grain evolution in the interstellar medium (ISM). We extend our publicly available Cosmic Lyα Transfer (COLT) code with photoionization equilibrium Monte Carlo radiative transfer and various methodology improvements for self-consistent end-to-end (non-)resonant line predictions. Accurate LyC reprocessing to recombination emission requires modelling pre-absorption by dust ($f_\text{abs} \approx 27.5\,\rm{per\,\,cent}$), helium ionization ($f_\text{He} \approx 8.7\,\rm{per\,\,cent}$), and anisotropic escape fractions ($f_\text{esc} \approx 7.9\,\rm{per\,\,cent}$), as these reduce the available budget for hydrogen line emission ($f_\text{H} \approx 55.9\,\rm{per\,\,cent}$). We investigate the role of the multiphase dusty ISM, disc geometry, gas kinematics, and star formation activity in governing the physics of emission and escape, focusing on the time variability, gas-phase structure, and spatial spectral, and viewing angle dependence of the emergent photons. Isolated disc simulations are well-suited for comprehensive observational comparisons with local Hα surveys, but would require a proper cosmological circumgalactic medium (CGM) environment as well as less dust absorption and rotational broadening to serve as analogs for high-redshift Lyα emitting galaxies. Future applications of our framework to next-generation cosmological simulations of galaxy formation including radiation-hydrodynamics that resolve ≲10 pc multiphase ISM and ≲1 kpc CGM structures will provide crucial insights and predictions for current and upcoming Lyα observations.
-
more » « less
ABSTRACT At fixed galaxy stellar mass, there is a clear observational connection between structural asymmetry and offset from the star-forming main sequence, ΔSFMS. Herein, we use the TNG50 simulation to investigate the relative roles of major mergers (stellar mass ratios μ ≥ 0.25), minor (0.1 ≤ μ < 0.25), and mini mergers (0.01 ≤ μ < 0.1) in driving this connection amongst star-forming galaxies (SFGs). We use dust radiative transfer post-processing with SKIRT to make a large, public collection of synthetic Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP) images of simulated IllustrisTNG (TNG) galaxies over 0.1 ≤ z ≤ 0.7 with log (M⋆/M⊙) ≥ 9 (∼750 k images). Using their instantaneous star formation rates (SFRs), known merger histories/forecasts, and HSC-SSP asymmetries, we show (1) that TNG50 SFGs qualitatively reproduce the observed trend between ΔSFMS and asymmetry and (2) a strikingly similar trend emerges between ΔSFMS and the time-to-coalescence for mini mergers. Controlling for redshift, stellar mass, environment, and gas fraction, we show that individual mini merger events yield small enhancements in SFRs and asymmetries that are sustained on long time-scales (at least ∼3 Gyr after coalescence, on average) – in contrast to major/minor merger remnants which peak at much greater amplitudes but are consistent with controls only ∼1 Gyr after coalescence. Integrating the boosts in SFRs and asymmetries driven by μ ≥ 0.01 mergers since z = 0.7 in TNG50 SFGs, we show that mini mergers are responsible for (i) 55 per cent of all merger-driven star formation and (ii) 70 per cent of merger-driven asymmetric structure. Due to their relative frequency and prolonged boost time-scales, mini mergers dominate over their minor and major counterparts in driving star formation and asymmetry in SFGs.
