skip to main content


Title: Using Machine Learning to Determine Morphologies of z < 1 AGN Host Galaxies in the Hyper Suprime-Cam Wide Survey
Abstract

We present a machine-learning framework to accurately characterize the morphologies of active galactic nucleus (AGN) host galaxies withinz< 1. We first use PSFGAN to decouple host galaxy light from the central point source, then we invoke the Galaxy Morphology Network (GaMorNet) to estimate whether the host galaxy is disk-dominated, bulge-dominated, or indeterminate. Using optical images from five bands of the HSC Wide Survey, we build models independently in three redshift bins: low (0 <z< 0.25), mid (0.25 <z< 0.5), and high (0.5 <z< 1.0). By first training on a large number of simulated galaxies, then fine-tuning using far fewer classified real galaxies, our framework predicts the actual morphology for ∼60%–70% of the host galaxies from test sets, with a classification precision of ∼80%–95%, depending on the redshift bin. Specifically, our models achieve a disk precision of 96%/82%/79% and bulge precision of 90%/90%/80% (for the three redshift bins) at thresholds corresponding to indeterminate fractions of 30%/43%/42%. The classification precision of our models has a noticeable dependency on host galaxy radius and magnitude. No strong dependency is observed on contrast ratio. Comparing classifications of real AGNs, our models agree well with traditional 2D fitting with GALFIT. The PSFGAN+GaMorNetframework does not depend on the choice of fitting functions or galaxy-related input parameters, runs orders of magnitude faster than GALFIT, and is easily generalizable via transfer learning, making it an ideal tool for studying AGN host galaxy morphology in forthcoming large imaging surveys.

 
more » « less
NSF-PAR ID:
10397505
Author(s) / Creator(s):
; ; ; ; ; ; ; ; ; ; ;
Publisher / Repository:
DOI PREFIX: 10.3847
Date Published:
Journal Name:
The Astrophysical Journal
Volume:
944
Issue:
2
ISSN:
0004-637X
Format(s):
Medium: X Size: Article No. 124
Size(s):
["Article No. 124"]
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    The COSMOS field has been extensively observed by most major telescopes, including Chandra, HST, and Subaru. HST imaging boasts very high spatial resolution and is used extensively in morphological studies of distant galaxies. Subaru provides lower spatial resolution imaging than HST but a substantially wider field of view with greater sensitivity. Both telescopes provide near-infrared imaging of COSMOS. Successful morphological fitting of Subaru data would allow us to measure morphologies of over 104known active galactic nucleus (AGN) hosts, accessible through Subaru wide-field surveys, currently not covered by HST. The morphological parameters indicate the types of galaxies that host AGNs. For 4016 AGNs between 0.03 <z< 6.5, we study the morphology of their galaxy hosts using GALFIT, fitting components representing the AGN and host galaxy simultaneously using thei-band imaging from both HST and Subaru. Comparing the fits for the differing telescope spatial resolutions and image signal-to-noise ratios, we identify parameter regimes for which there is strong disagreement between distributions of fitted parameters for HST and Subaru. In particular, the Sérsic index values strongly disagree between the two sets of data, including sources at lower redshifts. In contrast, the measured magnitude and radius parameters show reasonable agreement. Additionally, large variations in the Sérsic index have little effect on theχν2of each fit, whereas variations in other parameters have a more significant effect. These results indicate that the Sérsic index distributions of high-redshift galaxies that host AGNs imaged at ground-based spatial resolution are not reliable indicators of galaxy type and should be interpreted with caution.

     
    more » « less
  2. Abstract

    Radio wave scattering can cause severe reductions in detection sensitivity for surveys of Galactic and extragalactic fast (∼ms duration) transients. While Galactic sources like pulsars undergo scattering in the Milky Way interstellar medium (ISM), extragalactic fast radio bursts (FRBs) can also experience scattering in their host galaxies and other galaxies intervening in their lines of sight. We assess Galactic and extragalactic scattering horizons for fast radio transients using a combination of NE2001 to model the dispersion measure and scattering time (τ) contributed by the Galactic disk, and independently constructed electron density models for the Galactic halo and other galaxies’ ISMs and halos that account for different galaxy morphologies, masses, densities, and strengths of turbulence. For source redshifts 0.5 ≤zs≤ 1, an all-sky, isotropic FRB population has simulated values ofτ(1 GHz) ranging from ∼1μs to ∼2 ms (90% confidence, observer frame) that are dominated by host galaxies, althoughτcan be ≫2 ms at low Galactic latitudes. A population atzs= 5 has 0.01 ≲τ≲ 300 ms at 1 GHz (90% confidence), dominated by intervening galaxies. About 20% of these high-redshift FRBs are predicted to haveτ> 5 ms at 1 GHz (observer frame), and ≳40% of FRBs betweenzs∼ 0.5–5 haveτ≳ 1 ms forν≤ 800 MHz. Our scattering predictions may be conservative if scattering from circumsource environments is significant, which is possible under specific conditions. The percentage of FRBs selected against from scattering could also be substantially larger than we predict if circumgalactic turbulence causes more small-scale (≪1 au) density fluctuations than observed from nearby halos.

     
    more » « less
  3. ABSTRACT

    We aim to determine the intrinsic far-Infrared (far-IR) emission of X-ray-luminous quasars over cosmic time. Using a 16 deg2 region of the Stripe 82 field surveyed by XMM-Newton and Herschel Space Observatory, we identify 2905 X-ray luminous (LX > 1042 erg/s) active galactic nuclei (AGN) in the range z ≈ 0–3. The IR is necessary to constrain host galaxy properties such as star formation rate (SFR) and gas mass. However, only 10 per cent of our AGN are detected both in the X-ray and IR. Because 90 per cent of the sample is undetected in the far-IR by Herschel, we explore the mean IR emission of these undetected sources by stacking their Herschel/SPIRE images in bins of X-ray luminosity and redshift. We create stacked spectral energy distributions from the optical to the far-IR, and estimate the median SFR, dust mass, stellar mass, and infrared luminosity using a fitting routine. We find that the stacked sources on average have similar SFR/Lbol ratios as IR detected sources. The majority of our sources fall on or above the main sequence line suggesting that X-ray selection alone does not predict the location of a galaxy on the main sequence. We also find that the gas depletion time scales of our AGN are similar to those of dusty star forming galaxies. This suggests that X-ray selected AGN host high star formation and that there are no signs of declining star formation.

     
    more » « less
  4. Abstract

    We measure the correlation between black hole massMBHand host stellar massM*for a sample of 38 broad-line quasars at 0.2 ≲z≲ 0.8 (median redshiftzmed= 0.5). The black hole masses are derived from a dedicated reverberation mapping program for distant quasars, and the stellar masses are derived from two-band optical+IR Hubble Space Telescope imaging. Most of these quasars are well centered within ≲1 kpc from the host galaxy centroid, with only a few cases in merging/disturbed systems showing larger spatial offsets. Our sample spans two orders of magnitude in stellar mass (∼109–1011M) and black hole mass (∼107–109M) and reveals a significant correlation between the two quantities. We find a best-fit intrinsic (i.e., selection effects corrected)MBHM*,hostrelation oflog(MBH/M)=7.010.33+0.23+1.740.64+0.64log(M*,host/1010M), with an intrinsic scatter of0.470.17+0.24dex. Decomposing our quasar hosts into bulges and disks, there is a similarMBHM*,bulgerelation with slightly larger scatter, likely caused by systematic uncertainties in the bulge–disk decomposition. TheMBHM*,hostrelation atzmed= 0.5 is similar to that in local quiescent galaxies, with negligible evolution over the redshift range probed by our sample. With direct black hole masses from reverberation mapping and the large dynamical range of the sample, selection biases do not appear to affect our conclusions significantly. Our results, along with other samples in the literature, suggest that the locally measured black hole mass–host stellar mass relation is already in place atz∼ 1.

     
    more » « less
  5. Abstract

    We investigate spatially resolved emission-line ratios in a sample of 219 galaxies (0.6 <z< 1.3) detected using the G102 grism on the Hubble Space Telescope Wide Field Camera 3 taken as part of the CANDELS LyαEmission at Reionization survey to measure ionization profiles and search for low-luminosity active galactic nuclei (AGN). We analyze [Oiii] and Hβemission-line maps, enabling us to spatially resolve the [Oiii]/Hβemission-line ratio across the galaxies in the sample. We compare the [Oiii]/Hβratio in galaxy centers and outer annular regions to measure ionization differences and investigate the potential of sources with nuclear ionization to host AGN. We investigate some of the individual galaxies that are candidates to host strong nuclear ionization and find that they often have low stellar mass and are undetected in X-rays, as expected for low-luminosity AGN in low-mass galaxies. We do not find evidence for a significant population of off-nuclear AGN or other clumps of off-nuclear ionization. We model the observed distribution of [Oiii]/Hβspatial profiles and find that most galaxies are consistent with a small or zero difference between their nuclear and off-nuclear line ratios, but 6%–16% of galaxies in the sample are likely to host nuclear [Oiii]/Hβthat is ∼0.5 dex higher than in their outer regions. This study is limited by large uncertainties in most of the measured [Oiii]/Hβspatial profiles; therefore, deeper data, e.g., from deeper HST/WFC3 programs or from JWST/NIRISS, are needed to more reliably measure the spatially resolved emission-line conditions of individual high-redshift galaxies.

     
    more » « less