Estimating Baryonic Properties of Galaxies from Their Dark Matter - - PowerPoint PPT Presentation

▶
estimating baryonic properties of galaxies from their
SMART_READER_LITE
LIVE PREVIEW

Estimating Baryonic Properties of Galaxies from Their Dark Matter - - PowerPoint PPT Presentation

Apr 04, 2024 39 likes •262 views

Machine Learning in Astrophysics & Cosmology Estimating Baryonic Properties of Galaxies from Their Dark Matter universetoday.com MPA/Springel M51/HST Ji-hoon Kim (Department of Physics & Astronomy, SNU) Special thanks to: Yongseok Jo

slide-1
SLIDE 1

Ji-hoon Kim (Department of Physics & Astronomy, SNU)

Special thanks to: Yongseok Jo (SNU), Juhan Kim (KIAS), et al.

Machine Learning in Astrophysics & Cosmology

Estimating Baryonic Properties of Galaxies from Their Dark Matter

universetoday.com MPA/Springel M51/HST

slide-2
SLIDE 2

Evolution of Our Universe

  • Based on the mankind’s best guess for our Universe, standard

Big Bang LCDM cosmology.

slide-3
SLIDE 3

Kravtsov et al.

150 million light-years

Dark matter

  • interacts only by

gravitational force

  • dominates

structural evolution Evolution for the past ~13.6 billion years

Numerical Cosmology: Structure Formation

www.jihoonkim.org

slide-4
SLIDE 4

Kim et al. NASA/GSFC Governato et al. Governato et al.

slide-5
SLIDE 5

“Galaxies”

Dark matter web (Abel et al.)

Large-Scale Structure (LSS) of Universe

~2 billion light-years

Lemson et al.

slide-6
SLIDE 6

Jonsson/GADGET-2 Teyssier/RAMSES Teyssier/RAMSES Moody/ART-I Wadsley/GASOLINE Kim/ENZO Guedes/GASOLINE Guedes/GASOLINE Guedes/GASOLINE Agertz/RAMSES Kim/GIZMO Vogelsberger/AREPO Hopkins/GIZMO Fujimoto/ENZO Renaud/RAMSES Keller/GASOLINE Agertz/RAMSES Pontzen/CHANGA

Numerical Experiment in Cosmology

  • Numerical experiment essential due to galaxy’s nonlinear nature

→ often the only tool to test a theory on Universe’s evolution

www.jihoonkim.org

Gallery of galaxy-scale cosmological simulations

slide-7
SLIDE 7

Numerical Galaxy Formation: Upcoming Era

Year Mass resolution in stellar disk (M⊙)

109 107 105 101 103 1940 1980 1990 2000 2010 2020 2030

Holmberg 1941 Toomre 1978 Springel+ 2005 Hopkins+ 2015 Barnes 1988

2019 Next Decade

www.jihoonkim.org

slide-8
SLIDE 8

~10-4 pc ~100 pc ~101 pc ~104 pc ~106 pc ~109 pc

* 1 parsec (pc) = 3.26 light-year

Galaxies: Building Blocks of Universe

www.jihoonkim.org

  • Galaxies sit right in the middle of cosmological distance scale.

→ Universe’s building blocks + give contexts for star formation

Stars Planets Star Clusters

Galaxies

Galaxy Clusters Large-scale Structures

slide-9
SLIDE 9

MBH Gas Gas Star

Stars or Molecular Clouds Formation & Feedback

RGMC ~ 10 pc

Massive Black Hole (MBH) Accretion & Feedback

RBondi ~ pc

  • RSchw. ~ 10-5 pc

(Inter-)Galactic Dynamics & Interaction

Rhalo ~ 100 kpc

SFR in M33 M87/VLA Radio M51/HST

Rhalo

Multi-scale, Nonlinear Interactions

www.jihoonkim.org

slide-10
SLIDE 10

Multi-scale Hydrodynamics Simulation Code

——

Adaptive Mesh Refinement Enzo (Bryan et al. 2014)

  • Self-consistent galaxy-SMBH co-evolution from first principles
  • computing techniques like adaptive refinement help include more physics

Adaptive mesh

Insert a star simply by Schmidt law ( ρSFR ~ ρgas1.5 ) Turn off gas cooling or thermal energy Insert a star when a cell of ~103 M⊙ turns Jeans unstable UV photons radiative transfer (photoheating & ionization) + Supernova thermal energy

Previous Work New Approach

Star-forming Clumps Formation & Feedback Massive Black Hole Accretion & Feedback

www.jihoonkim.org

Artificially boosted Bondi accretion Thermal energy Bondi accretion without any boost factor UV/X-ray photons + Winds

Self-consistent Simulation Framework

slide-11
SLIDE 11

Adaptive Mesh Refinement Cosmology

Abel et al. (2002)

level=0 level=1 level=2

  • Adaptively focus on regions of high interests
  • refines cells if found interesting; e.g., dense, collapsing, unstable, shocked
  • More easily include relevant physics for galaxy-SMBH evolution
  • e.g., gravity, hydrodynamics, radiation, cooling, chemistry, stellar physics

Enzo AMR

www.jihoonkim.org

slide-12
SLIDE 12

Numerical Galaxy Formation: Upcoming Era

Year Mass resolution in stellar disk (M⊙)

109 107 105 101 103 1940 1980 1990 2000 2010 2020 2030

Holmberg 1941 Toomre 1978 Springel+ 2005 Hopkins+ 2015 Barnes 1988

2019 Next Decade

www.jihoonkim.org

slide-13
SLIDE 13

Gravity and Hydrodynamics

  • Gravity (dominated by dark matter) drives Universe’s evolution.
  • Hydrodynamics and others small-scale physics (dominated by

baryon or ordinary matter) affect detailed properties of galaxies.

MPA/Springel ESA/Planck (2013) www.jihoonkim.org

slide-14
SLIDE 14

Gravity and Hydrodynamics

  • There is a hint that we may reliably correlate baryonic (normal

matter) properties of a galaxy only with dark matter properties.

www.jihoonkim.org

Galaxies Baryonic interactions inside a galaxy Large-scale structure sculpted by dark matter

slide-15
SLIDE 15

Machine-assisted Semi-Simulation Model (MSSM)

www.jihoonkim.org

slide-16
SLIDE 16

Machine-assisted (Semi-)Simulation

Dark Matter Only Simulation Halo Catalog APPLICATION MACHINE LEARNING MODEL TRAINING Hydrodynamic Simulation Halo Catalog

DARK MATTER DATA

BARYON DATA MACHINE LEARNING MODEL

DARK MATTER DATA

BARYON DATA

Yongseok Jo, kerex@snu.ac.kr

arXiv:1908.09844 (MNRAS, 2019)

  • Predict ordinary matter from dark matter using ERT
slide-17
SLIDE 17

Machine-assisted (Semi-)Simulation

Dark matter of IllustrisTNG IllustrisTNG colored with baryons This work

Left: DM density of IllutrisTNG with circles representing virial radii. Middle/right: circles representing stellar masses colored by baryon fraction

  • Done in negligible time (~hours) compared to

full-scale hydrodynamic simulation (~months).

  • Can “paint” a large DM-only simulation with the

result from a smaller, high-resolution simulation.

Yongseok Jo, kerex@snu.ac.kr

arXiv:1908.09844 (MNRAS, 2019)

slide-18
SLIDE 18

Improvements In Machine Design

Previous work This work

Considers 3 input properties of

  • nly the target halo itself

Considers 11 inputs including the halo’s growth history and its local environment

Error = 1 N

N

∑

i

(yi

pred − ̂

yi

hydro)2

Error = 1 n

n

∑

i

(log yi

pred −

̂ log yi

hydro)2

Refined error function By building a two-stage tree, model grows linearly, resulting in smaller but deeper ERT Model grows exponentially Default error function in ERT

slide-19
SLIDE 19

Improved Machine Performance

  • More accurate prediction compared to previous attempts.

www.jihoonkim.org

Stellar mass in each galaxy from a full simulation Predicted stellar mass based only on dark matter

MSE: ~0.02 MSE: ~0.0002

slide-20
SLIDE 20

Machine Better Than Human Parametrization

  • Machine (red line) predicts galaxy’s baryonic properties better

than a human-devised parametrization model (blue line) does.

www.jihoonkim.org

Galactic stellar mass distribution in the Universe Massive black hole distribution in the Universe

slide-21
SLIDE 21

sdss.kias.re.kr/astro/Horizon-Runs/ www.jihoonkim.org illustris-project.org

  • Viable to generate a large, yet detailed mock Universe with ML.
  • a large volume to enclose variations at a “cosmological” scale
  • and yet a detailed physics and structures at a very small scale

NASA

Machine-assisted (Semi-)Simulation

slide-22
SLIDE 22
  • 1. New Opportunity:
  • Ever-increasing Importance of Numerical Cosmology
  • Advances in Machine Learning
  • II. Newly Possible Research:
  • Possibility of “Machine-assisted” (Semi-)Simulation

www.jihoonkim.org

Machine Learning in Astrophysics & Cosmology

Estimating Baryonic Properties of Galaxies from Their Dark Matter

arXiv:1908.09844 (MNRAS, 2019)