Nick Gnedin

Epigraph

Why is reionization interesting? I think the way to think about it is that

it was the last time when most baryons got together and did something together.

After that they kind of did their own thing.

Peng Oh

Outline

The brief history of time. How to reionize the universe. Simulations of yesterday: successes and

limitations. Homework: improving physics of star formation.

CROC Project: ultimate reionization simulations.

It pays off to do things right…

The Brief History of Time

Today:z=0

t=13.7 Gyr

End of inflation:z=1027

t=10-36s

Recombination:z=1089

t=379 kyr

The Brief History of Time

ionized neutral ionized

RE-IONIZATION

How To Reionize Universe

Ionizing radiation must be produced by some sources (galaxies, quasars, gamma-ray bursts, dark matter annihilation, laser-armed alien spaceships).

As ionizing photons propagate through space, they will be consumed in ionizing neutral atoms.

They can also be absorbed by other sinks – hence, wasted from the point of view of reionization.

Sources: Galaxies

To the best of our knowledge today, reionization is produced by stars.

Stars only form inside galaxies.

Galaxy LF at z=6(Bouwens et al. 2006)

Need ~50 Mpc box(do not forget bias!)

Sources: Quasars

QSOs are efficient producers of ionizing radiation. The only problem – they are too rare at z>3.

Sinks: Lyman-Limit Systems

Sinks: Lyman-Limit Systems

LL systems limit the Mean Free Path of an ionizing photon (a source cannot ionize a very distant atom).

Simulating Reionization: Scales

Capturing most of star-forming galaxies requires boxes > 50 Mpc (comoving).

Resolving LL system requires resolution ~ 1 kpc. Resolving star formation requires resolution

~100-200 pc (depends on your sub-grid model). Resolving all star-forming galaxies requires mass

resolution of about 107 M8.

Conclusion: we need simulations with spatial dynamic range L/Dx > 100,000 mass dynamic range Mbox/Mcell > 1 billion

Simulating Reionization: Physics

Dark matter Gas dynamics Atomic processes (cooling, H2 chemistry)

Radiative transfer of ionizing+UV radiation Star formation and stellar feedback

• Problem is not only in the lack of complete understanding, but also in the huge spatial dynamic range that needs to be modeled.

• Star formation today can only be modeled as a sub-grid model.

Simulations: The First Decade (2000-2010)

Small box Large box

physics full incomplete

spatial resolution high low

mass resolution high high

dynamic range low high

volume small large

Simulations: The First Decade (2000-2010)

Refresher: Ly-a Forest

Intrinsic spectrum

Absorbed flux

Transmitted flux

Quiz: find the SDSS Quasar

Where Do We Go Next?

One can run bigger simulations today than yesterday, but what’s the point if we do not model physics right?

We have homework! #1: Figure out how to model star formation

(sufficiently accurately for our purposes). #2: Figure out how to model stellar feedback.

Homework #1: Star Formation

2000s SF:

2010s SF:

Gas Stars

Atomic Gas StarsMolecular Gas

Modeling(calibrated with observations)

Observations

Kennicutt-Schmidt Relation

(Bigiel et al 2011)

Works on scales > 100 pc

may or may not be the same at z=0 and z=2.

Homework #2: Stellar Feedback

2010s feedback:

We were looking in the dark room for an exit, and we grabbed something that felt like a door knob…

“Blastwave” Model

All one needs to do to make a realistic galaxy in a cosmological simulation is to switch off cooling in a star forming region for ~10 Myr.

It is a numerical trick, but several real physical processes that operate on <1 pc scales manifest themselves on ~100 pc scale as if cooling was switched off:• Radiation pressure• Thermal pressure of coronal gas• Dynamic pressure of tubulence• Pressure of cosmic rays

Moore’s Law

Summer 2013 Fall 2014

Covering The Gap

With peta-scale computing power we can run large-box simulations with full physics.

Small box Large box

physics full incomplete fullspatial resolution high low highmass resolution high high highdynamic range low high highvolume small large large

Yesterday Today

The CROC Project:Simulations Dx = 100/200 pc with AMR (Deep/Shallow) M1 < 106 M8

Sets of boxes: Low/ Med/High• Small 20 CHIMP, 2563/ 5123/10243

• Medium 40 CHIMP, 10243/20483

• Large 80 CHIMP, 20483/40963

“Ultimate” simulation

Small box size Medium mass resolution Shallow spatial resolution

The CROC Project:Simulations

Mass

Space

Large box size High mass resolution Deep spatial resolution

100 pc 200 pc 20 Mpc/h 80 Mpc/h

106 M8

107 M8

The CROC Project:Convergence

Numerical

convergence: Medium resolution

runs converge to 70% at z=6

High resolution runs converge to 90% at z=6 and to 70% at z=9

The CROC Project:Survival Test #1

Galaxy UV luminosity functions:

Sources are modeled correctly (at least at z>5).

The CROC Project:Survival Test #2

Gunn-Peterson optical depth:

Sinks are modeled correctly.

(Becker at al 2014)

Example 1: Backreaction of Reionization on Galaxies Reionization suppresses gas accretion on low

mass halos (“photoevaporation”). Reionization may affect global star formation rate

(“Barkana & Loeb effect”).

(Barkana & Loeb 2000)

One of JWST science goals.

Backreaction:Gas Fractions Match Okamoto et al (2008) results

exactly (after reionization, of course).

C

Backreaction:Barkana-Loeb Effect There is no feature at reionization:

“Barkana-Loeb” effects does not exist.

D

Backreaction:Faint-End Slope

Faint-end slope of UV luminosity function varies by           for             .

Backreaction:Why? Galaxies affected by photoionization

contain no molecular gas.

Gas fractions

Molecular gas

Example 2. Ionizations from DM Annihilation If the g-ray signal from Galactic Center is dark

matter annihilation, annihilation also contributed ~20% of ionizing photons during reionization.

(Cirelli et al 2009)

And Many More…

The quintessential question: does reionization proceed inside-out or outside-in? Answer: both (first inside-out, later outside-in).

Evidence that dynamical processes of cosmic dust formation and evolution affect (read: can be studied via) reionization.

Clear demonstration that SDSS is the best (i.e. one cannot improve much on constraints from high-z SDSS quasars, no matter how much telescope time one throws at it).

…

Observations Are Coming

The next observational breakthroughs in studying reionization will happen soon:• Radio observations of the redshifted 21cm

emission from neutral hydrogen during reionization will map the structure of ionized bubbles.

• JWST/GMT/TMT will provide detailed view on reionization sources.

Redshifted 21 cm in 2013

Precision Array for Probing the Epoch of Reionization (PAPER)

Murchison Widefield Array (MWA) …

Redshifted 21 cm in 2013

Redshifted 21 cm in 2013

3 orders of magnitude passed.

1 to go…

Redshifted 21 cm Tomorrow

Progress in measuring the redshifted 21 cm is mind-blowing.

¾ of the way already passed with ~50% reduced funding.

The actual detection by ~2016-17 is likely.

Conclusions

Supercomputing power has passed the “sustained peta-scale” mark.

That power is just right for modeling cosmic reionization numerically.

The first realistic (i.e. modeling both sources and sinks correctly) simulations of reionization are currently being worked on by several groups (CROC including).

These simulations help us learn about the diverse range of physical phenomena: from cosmic dust to dark matter.

Reionization

The End