-
Constraints on the in-situ and ex-situ stellar masses in nearby galaxies with Artificial Intelligence
Authors:
Eirini Angeloudi,
Jesús Falcón-Barroso,
Marc Huertas-Company,
Alina Boecker,
Regina Sarmiento,
Lukas Eisert,
Annalisa Pillepich
Abstract:
The hierarchical model of galaxy evolution suggests that the impact of mergers is substantial on the intricate processes that drive stellar assembly within a galaxy. However, accurately measuring the contribution of accretion to a galaxy's total stellar mass and its balance with in-situ star formation poses a persistent challenge, as it is neither directly observable nor easily inferred from obser…
▽ More
The hierarchical model of galaxy evolution suggests that the impact of mergers is substantial on the intricate processes that drive stellar assembly within a galaxy. However, accurately measuring the contribution of accretion to a galaxy's total stellar mass and its balance with in-situ star formation poses a persistent challenge, as it is neither directly observable nor easily inferred from observational properties. Here, we present theory-motivated predictions for the fraction of stellar mass originating from mergers in a statistically significant sample of nearby galaxies, using data from MaNGA. Employing a robust machine learning model trained on mock MaNGA analogs (MaNGIA) in turn obtained from a cosmological simulation (TNG50), we unveil that in-situ stellar mass dominates almost across the entire stellar mass spectrum (1e9Msun < M* < 1e12Msun). Only in more massive galaxies (M* > 1e11Msun) does accreted mass become a substantial contributor, reaching up to 35-40% of the total stellar mass. Notably, the ex-situ stellar mass in the nearby universe exhibits significant dependence on galaxy characteristics, with higher accreted fractions favored by elliptical, quenched galaxies and slow rotators, as well as galaxies at the center of more massive dark matter halos.
△ Less
Submitted 28 June, 2024;
originally announced July 2024.
-
AstroPT: Scaling Large Observation Models for Astronomy
Authors:
Michael J. Smith,
Ryan J. Roberts,
Eirini Angeloudi,
Marc Huertas-Company
Abstract:
This work presents AstroPT, an autoregressive pretrained transformer developed with astronomical use-cases in mind. The AstroPT models presented here have been pretrained on 8.6 million $512 \times 512$ pixel $grz$-band galaxy postage stamp observations from the DESI Legacy Survey DR8. We train a selection of foundation models of increasing size from 1 million to 2.1 billion parameters, and find t…
▽ More
This work presents AstroPT, an autoregressive pretrained transformer developed with astronomical use-cases in mind. The AstroPT models presented here have been pretrained on 8.6 million $512 \times 512$ pixel $grz$-band galaxy postage stamp observations from the DESI Legacy Survey DR8. We train a selection of foundation models of increasing size from 1 million to 2.1 billion parameters, and find that AstroPT follows a similar saturating log-log scaling law to textual models. We also find that the models' performances on downstream tasks as measured by linear probing improves with model size up to the model parameter saturation point. We believe that collaborative community development paves the best route towards realising an open source `Large Observation Model' -- a model trained on data taken from the observational sciences at the scale seen in natural language processing. To this end, we release the source code, weights, and dataset for AstroPT under the MIT license, and invite potential collaborators to join us in collectively building and researching these models.
△ Less
Submitted 23 May, 2024;
originally announced May 2024.
-
ERGO-ML: Comparing IllustrisTNG and HSC galaxy images via contrastive learning
Authors:
Lukas Eisert,
Connor Bottrell,
Annalisa Pillepich,
Rhythm Shimakawa,
Vicente Rodriguez-Gomez,
Dylan Nelson,
Eirini Angeloudi,
Marc Huertas-Company
Abstract:
Modern cosmological hydrodynamical galaxy simulations provide tens of thousands of reasonably realistic synthetic galaxies across cosmic time. However, quantitatively assessing the level of realism of simulated universes in comparison to the real one is difficult. In this paper of the ERGO-ML series (Extracting Reality from Galaxy Observables with Machine Learning), we utilize contrastive learning…
▽ More
Modern cosmological hydrodynamical galaxy simulations provide tens of thousands of reasonably realistic synthetic galaxies across cosmic time. However, quantitatively assessing the level of realism of simulated universes in comparison to the real one is difficult. In this paper of the ERGO-ML series (Extracting Reality from Galaxy Observables with Machine Learning), we utilize contrastive learning to directly compare a large sample of simulated and observed galaxies based on their stellar-light images. This eliminates the need to specify summary statistics and allows to exploit the whole information content of the observations. We produce survey-realistic galaxy mock datasets resembling real Hyper Suprime-Cam (HSC) observations using the cosmological simulations TNG50 and TNG100. Our focus is on galaxies with stellar masses between $10^9$ and $10^{12} M_\odot$ at $z=0.1-0.4$. This allows us to evaluate the realism of the simulated TNG galaxies in comparison to actual HSC observations. We apply the self-supervised contrastive learning method NNCLR to the images from both simulated and observed datasets (g, r, i - bands). This results in a 256-dimensional representation space, encoding all relevant observable galaxy properties. Firstly, this allows us to identify simulated galaxies that closely resemble real ones by seeking similar images in this multi-dimensional space. Even more powerful, we quantify the alignment between the representations of these two image sets, finding that the majority ($\gtrsim 70$ per cent) of the TNG galaxies align well with observed HSC images. However, a subset of simulated galaxies with larger sizes, steeper Sersic profiles, smaller Sersic ellipticities, and larger asymmetries appears unrealistic. We also demonstrate the utility of our derived image representations by inferring properties of real HSC galaxies using simulated TNG galaxies as the ground truth.
△ Less
Submitted 11 April, 2024; v1 submitted 30 October, 2023;
originally announced October 2023.
-
ERGO-ML: Towards a robust machine learning model for inferring the fraction of accreted stars in galaxies from integral-field spectroscopic maps
Authors:
Eirini Angeloudi,
Jesús Falcón-Barroso,
Marc Huertas-Company,
Regina Sarmiento,
Annalisa Pillepich,
Daniel Walo-Martín,
Lukas Eisert
Abstract:
Quantifying the contribution of mergers to the stellar mass of galaxies is key for constraining the mechanisms of galaxy assembly across cosmic time. However, the mapping between observable galaxy properties and merger histories is not trivial: cosmological galaxy simulations are the only tools we have for calibration. We study the robustness of a simulation-based inference of the ex-situ stellar…
▽ More
Quantifying the contribution of mergers to the stellar mass of galaxies is key for constraining the mechanisms of galaxy assembly across cosmic time. However, the mapping between observable galaxy properties and merger histories is not trivial: cosmological galaxy simulations are the only tools we have for calibration. We study the robustness of a simulation-based inference of the ex-situ stellar mass fraction of nearby galaxies to different observables -- integrated and spatially-resolved -- and to different galaxy formation models -- IllustrisTNG and EAGLE -- with Machine Learning. We find that at fixed simulation, the fraction of accreted stars can be inferred with very high accuracy, with an error $\sim5$ per cent (10 per cent) from 2D integral-field spectroscopic maps (integrated quantities) throughout the considered stellar mass range. A bias (> 5 per cent) and an increase in scatter by a factor of 2 are introduced when testing with a different simulation, revealing a lack of generalization to distinct galaxy-formation models. Interestingly, upon using only stellar mass and kinematics maps in the central galactic regions for training, we find that this bias is removed and the ex-situ stellar mass fraction can be recovered in both simulations with < 15 per cent scatter, independently of the training set's origin. This opens up the door to a potential robust inference of the accretion histories of galaxies from existing Integral Field Unit surveys, such as MaNGA, covering a similar field of view (FOV) and containing spatially-resolved spectra for tens of thousands of nearby galaxies.
△ Less
Submitted 1 June, 2023;
originally announced June 2023.
-
Galaxy Morphology from $z\sim6$ through the eyes of JWST
Authors:
M. Huertas-Company,
K. G. Iyer,
E. Angeloudi,
M. B. Bagley,
S. L. Finkelstein,
J. Kartaltepe,
R. Sarmiento,
J. Vega-Ferrero,
P. Arrabal Haro,
P. Behroozi,
F. Buitrago,
Y. Cheng,
L. Costantin,
A. Dekel,
M. Dickinson,
D. Elbaz,
N. A. Grogin,
N. P. Hathi,
B. W. Holwerda,
A. M. Koekemoer,
R. A. Lucas,
C. Papovich,
P. G. Pérez-González,
N. Pirzkal,
L-M. Seillé
, et al. (4 additional authors not shown)
Abstract:
We analyze the Near Infrared ($\sim0.8-1μ$m) rest-frame morphologies of galaxies with $\log M_*/M_\odot>9$ in the redshift range $0<z<6$, compare with previous HST-based results and release the first JWST-based morphological catalog of $\sim20,000$ galaxies in the CEERS survey. Galaxies are classified into four main broad classes -- spheroid, disk+spheroid, disk, and disturbed -- based on imaging…
▽ More
We analyze the Near Infrared ($\sim0.8-1μ$m) rest-frame morphologies of galaxies with $\log M_*/M_\odot>9$ in the redshift range $0<z<6$, compare with previous HST-based results and release the first JWST-based morphological catalog of $\sim20,000$ galaxies in the CEERS survey. Galaxies are classified into four main broad classes -- spheroid, disk+spheroid, disk, and disturbed -- based on imaging with four filters -- $F150W$, $F200W$, $F356W$, and $F444W$ -- using Convolutional Neural Networks trained on HST/WFC3 labeled images and domain-adapted to JWST/NIRCam. We find that $\sim90\%$ and $\sim75\%$ of galaxies at $z<3$ have the same early/late and regular/irregular classification, respectively, in JWST and HST imaging when considering similar wavelengths. For small (large) and faint objects, JWST-based classifications tend to systematically present less bulge-dominated systems (peculiar galaxies) than HST-based ones, but the impact on the reported evolution of morphological fractions is less than $\sim10\%$. Using JWST-based morphologies at the same rest-frame wavelength ($\sim0.8-1μ$m), we confirm an increase in peculiar galaxies and a decrease in bulge-dominated galaxies with redshift, as reported in previous HST-based works, suggesting that the stellar mass distribution, in addition to light distribution, is more disturbed in the early universe. However, we find that undisturbed disk-like systems already dominate the high-mass end of the late-type galaxy population ($\log M_*/M_\odot>10.5$) at $z\sim5$, and bulge-dominated galaxies also exist at these early epochs, confirming a rich and evolved morphological diversity of galaxies $\sim1$ Gyr after the Big Bang. Finally, we find that the morphology-quenching relation is already in place for massive galaxies at $z>3$, with massive quiescent galaxies ($\log M_*/M_\odot>10.5$) being predominantly bulge-dominated.
△ Less
Submitted 3 May, 2023;
originally announced May 2023.
