skip to main content
Explore Scholarly Publications and Datasets in the NSF-PAR

This content will become publicly available on August 1, 2023

Title: Inferring Halo Masses with Graph Neural Networks
Abstract Understanding the halo–galaxy connection is fundamental in order to improve our knowledge on the nature and properties of dark matter. In this work, we build a model that infers the mass of a halo given the positions, velocities, stellar masses, and radii of the galaxies it hosts. In order to capture information from correlations among galaxy properties and their phase space, we use Graph Neural Networks (GNNs), which are designed to work with irregular and sparse data. We train our models on galaxies from more than 2000 state-of-the-art simulations from the Cosmology and Astrophysics with MachinE Learning Simulations project. Our model, which accounts for cosmological and astrophysical uncertainties, is able to constrain the masses of the halos with a ∼0.2 dex accuracy. Furthermore, a GNN trained on a suite of simulations is able to preserve part of its accuracy when tested on simulations run with a different code that utilizes a distinct subgrid physics model, showing the robustness of our method. The PyTorch Geometric implementation of the GNN is publicly available on GitHub ( https://github.com/PabloVD/HaloGraphNet ).
Authors:
; ; ; ; ; ; ; ; ;
Award ID(s):
2108944
Publication Date:
NSF-PAR ID:
10349264
Journal Name:
The Astrophysical Journal
Volume:
935
Issue:
1
Page Range or eLocation-ID:
30
ISSN:
0004-637X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; ( , Monthly Notices of the Royal Astronomical Society)
    ABSTRACT

    Cosmological simulations are reaching the resolution necessary to study ultra-faint dwarf galaxies. Observations indicate that in small populations, the stellar initial mass function (IMF) is not fully populated; rather, stars are sampled in a way that can be approximated as coming from an underlying probability density function. To ensure the accuracy of cosmological simulations in the ultra-faint regime, we present an improved treatment of the IMF. We implement a self-consistent, stochastically populated IMF in cosmological hydrodynamic simulations. We test our method using high-resolution simulations of a Milky Way halo, run to z = 6, yielding a sample of nearly 100 galaxies. We also use an isolated dwarf galaxy to investigate the resulting systematic differences in galaxy properties. We find that a stochastic IMF in simulations makes feedback burstier, strengthening feedback, and quenching star formation earlier in small dwarf galaxies. For galaxies in haloes with mass ≲ 108.5 M⊙, a stochastic IMF typically leads to lower stellar mass compared to a continuous IMF, sometimes by more than an order of magnitude. We show that existing methods of ensuring discrete supernovae incorrectly determine the mass of the star particle and its associated feedback. This leads to overcooling of surrounding gas, with at leastmore »∼10 per cent higher star formation and ∼30 per cent higher cold gas content. Going forwards, to accurately model dwarf galaxies and compare to observations, it will be necessary to incorporate a stochastically populated IMF that samples the full spectrum of stellar masses.

    « less
  2. ; ; ; ; ; ( , Monthly Notices of the Royal Astronomical Society)
    ABSTRACT

    In order to prepare for the upcoming wide-field cosmological surveys, large simulations of the Universe with realistic galaxy populations are required. In particular, the tendency of galaxies to naturally align towards overdensities, an effect called intrinsic alignments (IA), can be a major source of systematics in the weak lensing analysis. As the details of galaxy formation and evolution relevant to IA cannot be simulated in practice on such volumes, we propose as an alternative a Deep Generative Model. This model is trained on the IllustrisTNG-100 simulation and is capable of sampling the orientations of a population of galaxies so as to recover the correct alignments. In our approach, we model the cosmic web as a set of graphs, where the graphs are constructed for each halo, and galaxy orientations as a signal on those graphs. The generative model is implemented on a Generative Adversarial Network architecture and uses specifically designed Graph-Convolutional Networks sensitive to the relative 3D positions of the vertices. Given (sub)halo masses and tidal fields, the model is able to learn and predict scalar features such as galaxy and dark matter subhalo shapes; and more importantly, vector features such as the 3D orientation of the major axismore »of the ellipsoid and the complex 2D ellipticities. For correlations of 3D orientations the model is in good quantitative agreement with the measured values from the simulation, except for at very small and transition scales. For correlations of 2D ellipticities, the model is in good quantitative agreement with the measured values from the simulation on all scales. Additionally, the model is able to capture the dependence of IA on mass, morphological type, and central/satellite type.

    « less
  3. ; ( , Monthly Notices of the Royal Astronomical Society)
    ABSTRACT

    We present a machine learning (ML) approach for the prediction of galaxies’ dark matter halo masses which achieves an improved performance over conventional methods. We train three ML algorithms (XGBoost, random forests, and neural network) to predict halo masses using a set of synthetic galaxy catalogues that are built by populating dark matter haloes in N-body simulations with galaxies and that match both the clustering and the joint distributions of properties of galaxies in the Sloan Digital Sky Survey (SDSS). We explore the correlation of different galaxy- and group-related properties with halo mass, and extract the set of nine features that contribute the most to the prediction of halo mass. We find that mass predictions from the ML algorithms are more accurate than those from halo abundance matching (HAM) or dynamical mass estimates (DYN). Since the danger of this approach is that our training data might not accurately represent the real Universe, we explore the effect of testing the model on synthetic catalogues built with different assumptions than the ones used in the training phase. We test a variety of models with different ways of populating dark matter haloes, such as adding velocity bias for satellite galaxies. We determinemore »that, though training and testing on different data can lead to systematic errors in predicted masses, the ML approach still yields substantially better masses than either HAM or DYN. Finally, we apply the trained model to a galaxy and group catalogue from the SDSS DR7 and present the resulting halo masses.

    « less
  4. ; ; ; ; ; ( , Monthly Notices of the Royal Astronomical Society)
    null (Ed.)
    ABSTRACT We present the first set of cosmological baryonic zoom-in simulations of galaxies including dissipative self-interacting dark matter (dSIDM). These simulations utilize the Feedback In Realistic Environments galaxy formation physics, but allow the dark matter to have dissipative self-interactions analogous to standard model forces, parametrized by the self-interaction cross-section per unit mass, (σ/m), and the dimensionless degree of dissipation, 0 < fdiss < 1. We survey this parameter space, including constant and velocity-dependent cross-sections, and focus on structural and kinematic properties of dwarf galaxies with $M_{\rm halo} \sim 10^{10-11}{\, \rm M_\odot }$ and $M_{\ast } \sim 10^{5-8}{\, \rm M_\odot }$. Central density profiles (parametrized as ρ ∝ rα) of simulated dwarfs become cuspy when $(\sigma /m)_{\rm eff} \gtrsim 0.1\, {\rm cm^{2}\, g^{-1}}$ (and fdiss = 0.5 as fiducial). The power-law slopes asymptote to α ≈ −1.5 in low-mass dwarfs independent of cross-section, which arises from a dark matter ‘cooling flow’. Through comparisons with dark matter only simulations, we find the profile in this regime is insensitive to the inclusion of baryons. However, when $(\sigma /m)_{\rm eff} \ll 0.1\, {\rm cm^{2}\, g^{-1}}$, baryonic effects can produce cored density profiles comparable to non-dissipative cold dark matter (CDM) runs but at smaller radii. Simulated galaxies withmore »$(\sigma /m) \gtrsim 10\, {\rm cm^{2}\, g^{-1}}$ and the fiducial fdiss develop significant coherent rotation of dark matter, accompanied by halo deformation, but this is unlike the well-defined thin ‘dark discs’ often attributed to baryon-like dSIDM. The density profiles in this high cross-section model exhibit lower normalizations given the onset of halo deformation. For our surveyed dSIDM parameters, halo masses and galaxy stellar masses do not show appreciable difference from CDM, but dark matter kinematics and halo concentrations/shapes can differ.« less
  5. ; ; ; ; ; ; ; ; ; ; et al ( , Monthly Notices of the Royal Astronomical Society)
    ABSTRACT Galaxy–galaxy lensing is a powerful probe of the connection between galaxies and their host dark matter haloes, which is important both for galaxy evolution and cosmology. We extend the measurement and modelling of the galaxy–galaxy lensing signal in the recent Dark Energy Survey Year 3 cosmology analysis to the highly non-linear scales (∼100 kpc). This extension enables us to study the galaxy–halo connection via a Halo Occupation Distribution (HOD) framework for the two lens samples used in the cosmology analysis: a luminous red galaxy sample (redmagic) and a magnitude-limited galaxy sample (maglim). We find that redmagic (maglim) galaxies typically live in dark matter haloes of mass log10(Mh/M⊙) ≈ 13.7 which is roughly constant over redshift (13.3−13.5 depending on redshift). We constrain these masses to ${\sim}15{{\ \rm per\ cent}}$, approximately 1.5 times improvement over the previous work. We also constrain the linear galaxy bias more than five times better than what is inferred by the cosmological scales only. We find the satellite fraction for redmagic (maglim) to be ∼0.1−0.2 (0.1−0.3) with no clear trend in redshift. Our constraints on these halo properties are broadly consistent with other available estimates from previous work, large-scale constraints, and simulations. The framework built in this paper willmore »be used for future HOD studies with other galaxy samples and extensions for cosmological analyses.« less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email