Simulating the 4% Universe Hydro-cosmology simulations and data - - PowerPoint PPT Presentation

▶
simulating the 4 universe
SMART_READER_LITE
LIVE PREVIEW

Simulating the 4% Universe Hydro-cosmology simulations and data - - PowerPoint PPT Presentation

Sep 19, 2023 116 likes •574 views

Simulating the 4% Universe Hydro-cosmology simulations and data analysis Michael L. Norman SDSC/UCSD Lecture Plan Lecture 1: Hydro-cosmology simulations of baryons in the Cosmic Web Lyman alpha forest (LAF) Baryon Acoustic

slide-1
SLIDE 1

Simulating the 4% Universe

Hydro-cosmology simulations and data analysis

Michael L. Norman SDSC/UCSD

slide-2
SLIDE 2

Lecture Plan

  • Lecture 1: Hydro-cosmology simulations of

baryons in the Cosmic Web

– Lyman alpha forest (LAF) – Baryon Acoustic Oscillation (BAO)

  • Lecture 2: Radiation hydro-cosmology

simulations of Cosmic Renaissance

– Epoch of Reionization (EOR) – First Galaxies

7/17/2012 ISSAC 2012, SDSC, San Diego USA 2

slide-3
SLIDE 3

7/17/2012 ISSAC 2012, SDSC, San Diego USA 3

slide-4
SLIDE 4

Cosmic Renaissance

  • 1. First Stars
  • 2. First Galaxies
  • 3. Reionization

7/17/2012 ISSAC 2012, SDSC, San Diego USA 4

slide-5
SLIDE 5

When did reionization complete?

7/17/2012 ISSAC 2012, SDSC, San Diego USA 5

slide-6
SLIDE 6

Scientific Goals

  • Connect reionization to first galaxies through

direct numerical simulations

  • Some Questions

– How does reionization proceed? – Is the observed high-z galaxy population sufficient to reionize the Universe? – How is galaxy formation and the IGM modified by reionization? – How good are the analytic and semi-numerical models

  • f reionization?

7/17/2012 ISSAC 2012, SDSC, San Diego USA 6

slide-7
SLIDE 7

Three generations of cosmological reionization simulations

  • 1. Local self-consistent

– (small boxes < 10 Mpc) – CRHD+SF+ionization+heating – e.g., Gnedin 2000, Razoumov et al. 2002

  • 2. Global post-processing

– (large boxes > 100 Mpc) – N-body + RT – e.g., Iliev et al. 2006

  • 3. Global self-consistent

– (large boxes > 100 Mpc) – CRHD+SF+ionization+heating – Norman et al. 2012, in prep.

7/17/2012 ISSAC 2012, SDSC, San Diego USA 7

slide-8
SLIDE 8

Post-processing Approach

  • Pioneered by Sokasian et al. (2003) and “perfected” by

Iliev, Shapiro, et al. (2006+)

  • Recipe:

– Perform high resolution N-body DM simulation in large volume (L>100 Mpc/h) – Assign ionizing flux to every halo by some prescription – Post-process snapshots of the density field, sampled onto a coarse grid, with a ray-tracing radiative transfer code, assuming baryons trace DM – Sources and gas clumping factor “coarse grained” on the mesh – No radiative feedback on source population or intergalactic gas

7/17/2012 ISSAC 2012, SDSC, San Diego USA 8

slide-9
SLIDE 9

Post-processing Approach

  • Key insights

– reionization proceeds from the “inside-out” (i.e., from

  • verdense to underdense

regions) – reionization is “rapid” (∆z~2)

  • However

– redshift of overlap is not predicted, but can be “dialed in” since it depends critically

  • n assumed (Mhalo/Lion) and fesc

– minimum halo mass cutoff a free parameter

7/17/2012 ISSAC 2012, SDSC, San Diego USA 9

slide-10
SLIDE 10

radiation background galaxies IGM photo-ionization photo-heating absorption feedback (energy, metals) SF-recipe self-shielding photo-evaporation infall ionizing flux multi-species hydrodynamics radiative transfer N-body dynamics cosmic expansion self-gravity dark matter dynamics baryonic sector

Self-Consistent Approach

7/17/2012 ISSAC 2012, SDSC, San Diego USA 10

slide-11
SLIDE 11

https://code.google.com/p/enzo

7/17/2012 ISSAC 2012, SDSC, San Diego USA 11

slide-12
SLIDE 12

What does “direct simulation” mean?

  • All physical processes are simulated at the same mass

and spatial resolution

– DM, gas dynamics – parameterized star formation and feedbacks – radiation sources and transport – ionization/recombination/photoevaporation

  • Only subgrid model is SF, which is calibrated to
  • bservations (Bouwens et al.)
  • Advantage: sources and sinks of ionizing radiation and

radiative feedback effects are simulated directly

  • Disadvantage: very costly to bridge scales; some still

missing (minihalos)

7/17/2012 ISSAC 2012, SDSC, San Diego USA 12

slide-13
SLIDE 13

Two Simulations Differing only in Volume

ΛCDM, WMAP7

80 Mpc

32003 cells/particles

20 Mpc

8003 cells/particles

Run A and Run B have identical mass and spatial resolution, physics, ICs, etc.

Run A “¼ scale simulation” Run B “Renaissance Simulation”

64x volume

7/17/2012 ISSAC 2012, SDSC, San Diego USA 13

slide-14
SLIDE 14

Mass and Spatial Resolution

  • HMF complete to

~108 Ms to include dwarfs

– Sets “minimal” mass and spatial resolution – Mp = 5x105 Ms – ∆x=25 ckpc

  • Simulate largest

volume possible with available computer resources

GOALS Run B

7/17/2012 ISSAC 2012, SDSC, San Diego USA 14

slide-15
SLIDE 15

Numerical Methods

  • We use Enzo V2.1 in non-AMR mode

http://enzo.googlecode.com

– 6 species fluid dynamics: PPM – Dark matter dynamics: Particle-Mesh – Gravity: FFTs

  • Radiation transport: implicit flux-

limited diffusion, coupled to gas ionization and energy equation (Reynolds et al. 2009)

  • Star formation & SN feedback:

modified Cen & Ostriker 92 with “distributed feedback” (Smith et al. 2011)

– Calibrated to Bouwens et al. (2011) SFRD

  • UV radiative feedback: Pop II SED

from Ricotti, Gnedin & Shull 2002

7/17/2012 ISSAC 2012, SDSC, San Diego USA 15

slide-16
SLIDE 16

Tests of Radiation Solver

Reynolds et al. (2009)

  • Correct I-front

speeds are

  • btained even at

low resolution due to implicit coupling

  • f rad. transfer,

ionization, and gas heating

Shapiro & Giroux ‘87 analytic test problem

7/17/2012 ISSAC 2012, SDSC, San Diego USA 16

slide-17
SLIDE 17

Results

  • Run A (1/4 scale simulation)

– Ionizing photons per H atom – Adequacy of MHR estimate

  • Run B (Renaissance Simulation)

– Role of large scale power – Suppression of star formation in low mass halos due to radiative feedback

7/17/2012 ISSAC 2012, SDSC, San Diego USA 17

slide-18
SLIDE 18

z=12.5 z=9.2 z=8 z=7 z=6 t=362 Myr t=552 Myr t=664 Myr t=792 Myr t=969 Myr

ENZO radiation hydrodynamic cosmic reionization

  • G. So, M. Norman, R. Harkness (UCSD), D. Reynolds (SMU)

Redshift/time evolution of density and temperature 8003/20 Mpc/512 core

density temperature

7/17/2012 ISSAC 2012, SDSC, San Diego USA 18

slide-19
SLIDE 19

ENZO radiation hydrodynamic cosmic reionization

  • M. Norman, R. Harkness, G. So (UCSD), D. Reynolds (SMU)

Redshift/time evolution of density and temperature 8003/20 Mpc/512 core

density temperature

7/17/2012 ISSAC 2012, SDSC, San Diego USA 19

slide-20
SLIDE 20

7/17/2012 ISSAC 2012, SDSC, San Diego USA 20

slide-21
SLIDE 21

Ionized Volume Fraction

7/17/2012 ISSAC 2012, SDSC, San Diego USA 21

slide-22
SLIDE 22

Photons per H atom

  • btw. 3.5-4.5

ionizing photons per H atom

7/17/2012 ISSAC 2012, SDSC, San Diego USA 22

slide-23
SLIDE 23

Visualizing “Inside-Out” Reionization: Z-reion Cube

  • Every cell contains the

redshift when it was first photo-ionized

  • yt script:

– Loop over all redshift

  • utputs (80) and test if

fHII>0.9 – Uses nested parallel

  • bjects to divide up the

work on 256 cores – 56 sec on Gordon including IO

Z-reion

7/17/2012 ISSAC 2012, SDSC, San Diego USA 23

slide-24
SLIDE 24

7/17/2012 ISSAC 2012, SDSC, San Diego USA 24

slide-25
SLIDE 25

Result

7/17/2012 ISSAC 2012, SDSC, San Diego USA 25

slide-26
SLIDE 26

Effective of Large Scale Power

slice

7/17/2012 ISSAC 2012, SDSC, San Diego USA 26

slide-27
SLIDE 27

Effective of Large Scale Power

slice

7/17/2012 ISSAC 2012, SDSC, San Diego USA 28

slide-28
SLIDE 28

Effect of large scale power

7/17/2012 ISSAC 2012, SDSC, San Diego USA 29

slide-29
SLIDE 29

HI going, going, gone….

Z=7 Z=6.5 Z=6.05 80 cMpc Projected HI fraction Large-scale neutral patches before overlap

7/17/2012 ISSAC 2012, SDSC, San Diego USA 30

slide-30
SLIDE 30

Effect of large scale power

7/17/2012 ISSAC 2012, SDSC, San Diego USA 31

slide-31
SLIDE 31

Where is the star formation happening?

Z=7.3

7/17/2012 ISSAC 2012, SDSC, San Diego USA 32

slide-32
SLIDE 32

Where is the star formation happening?

Star formation strongly suppressed at Mh < 5x109 Ms

7/17/2012 ISSAC 2012, SDSC, San Diego USA 33

slide-33
SLIDE 33

Is this a resolution effect? NO

adiabatic hydro SF + SN feedback SF + SN feedback + radiative feedback

7/17/2012 ISSAC 2012, SDSC, San Diego USA 34

slide-34
SLIDE 34

Is this a resolution effect? NO

Ratio of Halo Gas Masses Depletion of baryons due to SN feedback Depletion of baryons due to radiative feedback

7/17/2012 ISSAC 2012, SDSC, San Diego USA 35

slide-35
SLIDE 35

Visualizing Jeans Smoothing

  • M. Norman, G. So, R. Harkness (UCSD), D. Reynolds (SMU)

Density fields from RHD and non-RHD models

z=8, RHD z=8, HD Visualization by J. Insley (ANL) & R. Wagner (SDSC) Z=8

7/17/2012 ISSAC 2012, SDSC, San Diego USA 36

slide-36
SLIDE 36

Visualizing Jeans Smoothing

Normailzed density difference between RHD and non-RHD models

z=8, RHD z=8, HD Visualization by J. Insley (ANL) & R. Wagner (SDSC) Z=8

7/17/2012 ISSAC 2012, SDSC, San Diego USA 37

slide-37
SLIDE 37

Visualizing Jeans Smoothing

Normailzed density difference between RHD and non-RHD models

z=8, RHD z=8, HD Visualization by J. Insley (ANL) & R. Wagner (SDSC) Z=8 non-radiative radiative no difference

7/17/2012 ISSAC 2012, SDSC, San Diego USA 38

slide-38
SLIDE 38

radiative density dist. non-radiative density dist. normalized density difference

ρ2-ρ1 ρ2+ρ1

yellow

red red minus

ρ2 ρ1

7/17/2012 ISSAC 2012, SDSC, San Diego USA 39

slide-39
SLIDE 39

7/17/2012 ISSAC 2012, SDSC, San Diego USA 40

slide-40
SLIDE 40

Jeans Smoothing

7/17/2012 ISSAC 2012, SDSC, San Diego USA 41

slide-41
SLIDE 41

Effect on Dark Matter Power

7/17/2012 ISSAC 2012, SDSC, San Diego USA 42

slide-42
SLIDE 42

7/17/2012 ISSAC 2012, SDSC, San Diego USA 43

slide-43
SLIDE 43

SAN DIEGO SUPERCOMPUTER CENTER at the UNIVERSITY OF CALIFORNIA; SAN DIEGO

Cosmology simulation matter power spectrum measurement using vSMP

Source: Rick Wagner, Michael L. Norman. SDC. Used by permission. 2012

We have run two large (32003 uniform grid) simulations, with and without radiation hydrodynamics, to measure the effect of the light from the first stars on the evolution of the universe. To quantitatively compare the matter distribution of each simulation, we use radially binned 3D power spectra.

  • 2 simulations
  • 32003 uniform 3D grids
  • 244GiB+ per field
  • 15k+ files each

Individual simulations Power spectra

  • Ran existing OpenMP-

threaded code

  • ~256GiB memory used
  • ~5 ½ hours per field
  • 0 development effort

Difference

slide-44
SLIDE 44

Summary: by the numbers

  • Direct RHD simulation of reionization now feasible in reasonably large

volumes

  • Reionization completes at z ~ 6 using the observed SFRD (Bouwens et
  • al. 2011)
  • Larger box begins reionization sooner, because of rare peaks, but

completes reionization at the same redshift (self-regulation?)

  • Full reionization requires ~ 4 photons/H atom
  • MHR formula provides a good estimator of when reionization will
  • ccur provided global HII clumping factor is used (dense gas not

excluded)

  • Radiative feedback suppresses star formation in halos Mh < 5x109 Ms

due to baryon depletion arising from Jeans smoothing

  • Large-scale patches (>10 Mpc) of HI remain as late as z=5.8, which

may be observable in LAE correlation function

7/17/2012 ISSAC 2012, SDSC, San Diego USA 45