Evidence for Feedback: Cosmological Evidence for Feedback: Cosmological Reionization and High-z SMBH GrowthReionization and High-z SMBH Growth

Columbia University

Reionization Workshop Beijing, China July 9-11, 2008

ZoltZoltáán Haimann Haiman

Outline of TalkOutline of Talk

• Reionization History

– observational clues from quasars & CMB

– no-feedback model over-produces e from WMAP3

– processes that can suppress earliest galaxies

• Growth of Supermassive BHs

– contribution of accreting BHs to reionization and feedback

– assembly of billion-solar mass SMBHs by z=6

• Metal Enrichment: intergalactic gas enriched by heavy elements out to z~6 (ubiquitous but cold)

• Reionization: intergalactic gas highly ionized at z~6 (and possibly already at z~13)

Stars and Black Holes at z>6Stars and Black Holes at z>6

• Massive Black Holes: >109M holes known to exist already at z~6 (their seeds must be in place at z>10)

chemical/thermal/ionization state of IGM and star/BH formation are governed by global feedback processes ? ( no analog at low redshift )

Distant Quasars: Reionization at z~6?Distant Quasars: Reionization at z~6?

xH 210-4

xH 10-4

xH ≳ 10-3

Fan et al. 2002

z=5.82

z=5.99

z=6.28

Gunn-Peterson trough:

Observational Breakthroughin 2002: SDSS quasars at z~6

Spectrum of a Pre-Reionization SourceSpectrum of a Pre-Reionization Source

HIIHII HIHI∢

Z=0 Z=6 Z=7

IGMIGM

Observed Wavelength (Å)

Op

acit

y (l

og

)

6

0

-6

HII region (xHII region (xHIHI ~ r ~ r22))

Two contributions to Ly absorption:

HII region (xHI ~ r2)Gunn-Peterson wing

The Hunt for the GP Damping WingThe Hunt for the GP Damping Wing

redshifted wavelength (Å)redshifted wavelength (Å)

opti

cal d

epth

opti

cal d

epth

- Create mock spectra:Create mock spectra: 300 lines of sight in a300 lines of sight in a hydro simulation box hydro simulation box from Renyue Cen

- Match observed statistic: Match observed statistic: (1) rapid loss of flux(1) rapid loss of flux Mesinger & Haiman (2004)

(2) pixel opt.depth PDF(2) pixel opt.depth PDF Mesinger & Haiman (2007)

- Fit 3 free parameters:Fit 3 free parameters:

RRss (HII region size)(HII region size)

xxHIHI (neutral fraction)(neutral fraction)

LLionion (ionizing luminosity)(ionizing luminosity)

Pixel Optical Depth for z=6.28 quasarPixel Optical Depth for z=6.28 quasar(Becker et al. 2001; White et al. 2003; 2004)

observed wavelength (Å)observed wavelength (Å)

flu

x (a

rbit

rary

un

its)

flu

x (a

rbit

rary

un

its)

modelingmodeling

(1)(1) Fit for intrinsic Fit for intrinsic emission from redemission from red side of Lya lineside of Lya line (double Gaussian + NV)

(2) Derive opt. depth(2) Derive opt. depth in each pixel on in each pixel on blue side of lineblue side of line (~30 pixels available)

Pixel Optical Depth PDF Results (z=6.22)Pixel Optical Depth PDF Results (z=6.22)(Mesinger & Haiman 2007)

Inferred optical depth (Inferred optical depth ( ))

Fra

ctio

n o

f p

ixel

sF

ract

ion

of

pix

els

Rs HI fion Prob

29 1.0 0.7 0.39

29 .04 0.7 0.01

33 1.0 0.7 0.01

29 1.0 0.1 0.01

Detection of a Cosmic Strömgren SphereDetection of a Cosmic Strömgren Sphere

- R- RSS = 6 = 6 0.3 Mpc (~40 comoving Mpc) 0.3 Mpc (~40 comoving Mpc)- X- XHIHI ≳≳ 0.04 - 0.1 0.04 - 0.1- L- Lionion = (0.8-1.6)×10 = (0.8-1.6)×105757ss-1-1

- IGM significantly neutralIGM significantly neutral Follows independently from (1) spectrum and (2) size of HII region

- LyLy and Ly and Ly can yield dynamic range to can yield dynamic range to find ionization topologyfind ionization topology

- constrain spectral hardnessconstrain spectral hardness EEionion < 0.2 keV < 0.2 keV

Implications:Implications:

Results:Results: (“2(“2”)”)

- 4 quasars with full GP troughs show diversity4 quasars with full GP troughs show diversityOther quasars:Other quasars:

Quasars vs. GRB afterglowsQuasars vs. GRB afterglows

Conventional Wisdom: GRBs are better probes because

1. bright and seen to high-z 2. No HII region present 3. power-law spectrum easier interpretation

But… quasar spectra ultimately better probes

1. Statistical sample (>100 spectra) will be needed (GRBs appear to be too rare) 2. Large QSO HII region moves I-front to mean IGM (GRBs have intrinsic HI absorption) 3. Lyman line modeling is not a limitation

Intrinsic LyIntrinsic Ly Line Shapes Line Shapes

Can we predict flux on blue side of Ly, using red side?

Repeat fitting procedure using low-redshift templates - 46 HST spectra of 42 z<1 quasars - select spectra showing no absorption - use observed spectrum + model IGM to create mock z=6 spectrum - 6-parameter fit of 3 Gaussians (fix central wavelength) to red side - find best-fit xHI , RHI , Lion from blue side

Scatter and bias in the inferred mean xHI in the IGM?

Kramer & Haiman (2008, in preparation)

Recovered Fit ParametersRecovered Fit ParametersKramer & Haiman (2008, in preparation)

Recovered Neutral FractionsRecovered Neutral FractionsKramer & Haiman (2008, in preparation)

Recovered Neutral Fractions (xRecovered Neutral Fractions (xHIHI=0.04)=0.04)Kramer & Haiman (2008, in preparation)

CMB: Electron Scattering Optical DepthCMB: Electron Scattering Optical Depth

WMAP5 Polarization AnisotropyDunkley et al. (2008)

0.087 0.017

zr 10.9 2.5

c.f. =0.04 for zr= 6

1. early reionization1. early reionization oror

2. “tail” of partial2. “tail” of partialIonization to higher zIonization to higher z

Primordial Gas CoolingPrimordial Gas Cooling

log( Temperature / K )

log(

coo

lin

g ra

te /

erg

s-1 c

m3 )

COSMIC TIME

MASS SCALE

Tvir 104 M

108 M

2

3 1 z

11

K

Gas Phase Chemistry: H + e- H- + H- + H H2+ e-

Virial Temperature

Minihalo: Tvir < 104 K

What forms in the early minihalos What forms in the early minihalos

• STARS: FIRST GENERATION METAL FREE - massive stars with harder spectra - boost in ionizing photon rate by a factor of ~ 20 - return to “normal” stellar pops at Z≳10-3.5 Z⊙

(Tumlinson & Shull 2001 ; Bromm, Kudritzki & Loeb 2001; Schaerer 2002)

• SEED BLACK HOLES: (~102-6 M ⊙ ) - boost by ~10 in number of ionizing photons/baryon - harder spectra up to hard X-rays: “pre-ionization” (Ricotti & Ostriker; Oh; Venkatesan & Shull; Madau et al.)

- must eventually evolve to quasars and remnant holes - accreting BHs would overproduce unresolved

soft XRB at 1 keV if they dominated reionization at z~6 (Dijkstra, Haiman & Loeb 2005)

Vanilla Reionization History (no minihalos)Vanilla Reionization History (no minihalos)

Wyithe & Loeb; Ciardi et al.Somerville et al.; Sokasian et al. Fukugita & Kawasaki; Cen; ….

Haiman & Bryan (2006)

NN = 4000= 4000ff* * = 15%= 15%ffesc esc = 20%= 20%C = 10C = 10

/C = N/C = Nff**ffescesc/C = 12/C = 12

0.090.09

redshiftredshift

ion

ized

vol

um

e fr

acti

onio

niz

ed v

olu

me

frac

tion

Fixed by requirementFixed by requirementthat z(percolation) ~ 6that z(percolation) ~ 6

TTminmin= 10= 1044 K K

Expectations for MinihalosExpectations for Minihalos

• Ionizing sources: - ff* * = 0.0025 (vs 0.15)= 0.0025 (vs 0.15) - NN = 80,000 (vs 4,000)= 80,000 (vs 4,000) - f fesc esc = 1 (vs 0.2)= 1 (vs 0.2)

• Clumping evolution

• Excluded from from ionized regions - cannot complete reionization at z~6 cannot complete reionization at z~6

Simulations (e.g. Iliev et al. 2006)

C(z) 1 91 z

7

NNff**ffesc esc = 200 vs 120 = 200 vs 120

C(z~15)=3 vs C(z=6)=10C(z~15)=3 vs C(z=6)=10

in absence of feedback, minihalos could contribute to reionization:

Reionization History with MinihalosReionization History with Minihalos

redshiftredshift

ion

ized

vol

um

e fr

acti

onio

niz

ed v

olu

me

frac

tion

efficiencyefficiency

Op

tica

l dep

thO

pti

cal d

epth

0.190.19

0.090.09

Haiman & Bryan (2006)

Minihalo contribution suppressed by a factor of ~10 (2)

Feedback ProcessesFeedback Processes• INTERNAL TO SOURCES - UV flux unbinds gas - supernova expels gas, sweeps up shells - H2 chemistry (positive and negative) - metals enhance cooling

• GLOBAL (FAR REACHING OR LONG LASTING) - H2 chemistry (LW: negative X-rays: positive) - photo-evaporation (minihalos with < 10 km/s) - photo-heating (halos with 10 km/s < < 50 km/s) - entropy floor (inactive fossil HII regions or X-rays) - global dispersion of metals (pop III pop II) - mechanical (SN blast waves)

Do most minihalos fail to form stars or black holes?

SF/Reionization History Self-Regulates?SF/Reionization History Self-Regulates?

Case 1 : No net feedbackCase 1 : No net feedback reionization completed early reionization completed early small halossmall halos closely spacedclosely spaced smooth smooth He/H close in timeHe/H close in time

Case 2 : Negative feedbackCase 2 : Negative feedback reionization completed laterreionization completed later larger halos,larger halos, farther spacedfarther spaced more patchymore patchy He/H farther in timeHe/H farther in time

*IF* feedback regulates reionization history, then there*IF* feedback regulates reionization history, then therewill be a period with a robust ‘steady state’ solution will be a period with a robust ‘steady state’ solution for the star formation history - need to know Jfor the star formation history - need to know Jcritcrit(M(Mhalohalo,z),z)

log

[dlo

g [d

* * /d

t/d

t // MM

y

ryr-1-1 M

pc M

pc-3-3]]

redshift

Haiman, Abel & Rees (2000)

Outline of TalkOutline of Talk

• Reionization History

– observational clues from quasars & CMB

– no-feedback model over-produces e from WMAP3

– processes that can suppress earliest galaxies

• Growth of Supermassive BHs

– contribution of accreting BHs to reionization and feedback

– assembly of billion-solar mass SMBHs by z=6

Growth of High-z Supermassive BHsGrowth of High-z Supermassive BHs

z=6.43

z=20

CDM mergertree

1. Can growing BHs contribute significantly to reionization?2. How are the z~6 SMBHs with Mbh = few × 109 M assembled?

min ~ 10 km/s

Mbh= few × 109 M

Is Super-Eddington Growth Required?Is Super-Eddington Growth Required?

Example: SDSS 1114-5251 (Fan et al. 2003)

z=6.43 Mbh 4 x 109 M

e-folding (Edd) time:

4 x (/0.1) 107yrΩ

Age of universe (z=6.43)

8 x 108 yr

How did this SMBH grow so massive? (Haiman & Loeb 2001)

No. e-foldings needed

ln(Mbh/Mseed) ~ 20 Mseed ~100 M

Strong beaming? No. (Haiman & Cen 2002)

Gravitational lensing? No. (Keeton, Kuhlen & Haiman 2004)

• Gravitational radiation produces sudden recoil

— kick velocity depends on mass ratio and on spin vectors — typical v(kick) ~ few 100 km/s (Baker et al. 2006, 2007

— maximum v(kick) ~ 4,000 km/s Gonzalez et al. 2007)

• Most important at high redshift when halos are small — escape velocities from z>6 halos is few km/s

Major obstacle: gravitational recoil

Modeling ProcedureModeling Procedure

• Construct Monte-Carlo DM halo merger tree from z=6 to z>40 - 108M ⊙≤ Mhalo ≤ 1013M ⊙ (Mres =few 105M⊙; N~105 trees)

- seed fraction focc of new halos with BHs (Mseed =100 M⊙)

• Gravitational Recoil - at merger, draw random vkick (Baker et al. 2008) - spin orientation: random or aligned - follow kicked BH trajectory -- damped oscillation (gas drag) - profile either r∝ -2.2 (cool gas) or flat core

• BH growth by accretion - merger delayed by dynamical friction time - seed initially in empty halo - duty cycle for accretion between 0.6-1.0 - maximum of Bondi and Eddington rate

Tanaka & Haiman (2008, in preparation)Tanaka & Haiman (2008, in preparation)

Trajectory of kicked BHTrajectory of kicked BHTanaka & Haiman (2008, in preparation)Tanaka & Haiman (2008, in preparation)

• DM halo NFWDM halo NFW

• gas with gas with flat coreflat core (Shapiro et al)(Shapiro et al)

• gas with gas with steep cuspsteep cusp (Abel & Bryan)(Abel & Bryan)

SMBH mass function at z=6SMBH mass function at z=6Tanaka & Haiman (2008, in preparation)Tanaka & Haiman (2008, in preparation)

Total mass in >10Total mass in >105 5 MMSMBHs: SMBHs: overproduced by a factor of 100-1000 !overproduced by a factor of 100-1000 !

Tanaka & Haiman (2008, in preparation)Tanaka & Haiman (2008, in preparation)

Results from SMBH assemblyResults from SMBH assembly

• (i) spin alignment (ii) focc ≳10-3 optimistic assumptions required (iii) fduty ≳0.8

• The 109 M BHs result from runaway early seeds (z>25) that avoided ejection at merger: asymmetric mass ratio

• Making few 109 M BHs by z=6 without overproducing few 105 M BHs (BH ≦ 4 104 MMpc-3 ) suggests focc ≈ 10-3 and negative feedback at z~20-30

• Growing BHs: X-ray pre-ionization (10-20%) and heating

• Alternative : a rapid (super-Eddington) growth phase

ConclusionsConclusions

1. WMAP+SDSS quasars together constrain early reionization:

mini-halo contribution at z ≳15 suppressed by factor of ~10

2. Suppression can be caused by H2 photo-dissociation by early

LW background, or by X-ray pre-heating by BHs

3. GR kick major obstacle for early BH growth: few x 109 M by

z=6 requires uninterrupted Eddington accretion

4. Difficult to build 109 M BHs without overproducing 105-6 M

BHs: independent evidence for negative feedback at z~20

Remnants of Massive StarsRemnants of Massive Stars Heger et al. 2003 (for single, non-rotating stars)

10M 25M 40M 140M 260M

ZZ=Z=Z

ZZ=0=0

met

alic

ity

met

alic

ity

SMBHs in TSMBHs in Tvir vir >10>1044K halos?K halos?

• Highly super-Eddington growth may be possible if gas remains at 10,000 Kelvin due to lack of H2 and cools via atomic H lines

• Jeans mass MJ T2/1/2 ≈ 105-6M ⊙

• Behavior of gas has not been studied in nearly same detail as for minihalos (no 3D simulations)

• But we have a speculation, based on a MMW disk toy model (Oh & Haiman 2002)

Conclusion: TConclusion: Tvir vir >10>1044K halos cool to ~100KK halos cool to ~100K

Similar to minihalos: Similar to minihalos:

Rely on H2 cooling and fragment on similar (few 100 M ) scales

Main difference:Main difference:

contract to high densities less susceptible to feedback

cf: HD reduces temperature and fragmentation scale?

Uehara & Inutsuka 2000Machida et al. 2005Johnson & Bromm 2005

cf: SMBHs

Volonteri &Rees 2005Bromm & Loeb 2005Begelman et al. 2006Spaans & Silk 2006

(Oh & Haiman 2002)

Direct SMBH formation?Direct SMBH formation?

Omukai, Schneider & Haiman (2008)

Evolution of irradiated, metal-free gas: JEvolution of irradiated, metal-free gas: J2121(crit) ≈ 10(crit) ≈ 1033

Direct SMBH formation: impact of metalsDirect SMBH formation: impact of metals

Omukai, Schneider & Haiman (2008)

Including the effect of (1) irradiation and (2) metalsIncluding the effect of (1) irradiation and (2) metals

Dense star cluster, rather than SMBH?Dense star cluster, rather than SMBH?Omukai, Schneider & Haiman (2008)

• Two stringent conditions needed to avoid fragmentation: (i) J(LW) (i) J(LW) ≳≳ few 10 few 1033 10 10-21-21 erg s cm erg s cm-2-2 Hz Hz-1-1 sr sr-1-1

(ii) Z (ii) Z ≲≲55 1010-6-6 Z Z

• First condition may be satisfied in rare (~10-7) cases of a very close, bright & synchronized neighbor (Dijkstra, Haiman, Wyithe & Mesinger, in preparation)

• Second condition eased by factor of 100 if no dust (CII and OI cooling).

• Most likely case with floor metals will form a dense stellar cluster collapse to IMBH of 102-3 M

Combined EffectsCombined Effects

• AMR Simulations with Enzo (Mesinger, Bryan & Haiman 2007)

- (1 h-1 Mpc)3 , 1283 root grid, run from z=99 to z=15 - re-simulate inner (0.25 h-1 Mpc)3

- 10 levels of refinement - 0.36 h-1 pc resolution at z=20 - biased (2.4) region, yields several hundred DM halos in mass range of 105M<M<107M

• Examine Effects of Transient Photoheating - J(UV) = 0 (test run) - Flash ionization (c.f. O’Shea et al. 2006)

- J(UV) = 0.08 or 0.8 for t = 3106 years (uniform, opt.thin) • Examine Effect of Constant LW background - 10-3 < J(LW) < 10-1 added to J(UV)=0 and 0.8 runs

• Cold gas develops in a cooling time t = tcool ∝ T / ngas xH2

• Works on halo-by-halo basis - near cancelation T: varies little (initial Compton cooling) ngas: depressed by factor of ~ 40 xH2: increased by factor of ~ 10 net delay: factor of ~4

• Adding Lyman-Werner background (11-13.6eV)

t(photo-dissociation) ~ t(formation)

Delayed CoolingDelayed Cooling

Radiative Feedback Simulation SummaryRadiative Feedback Simulation Summary

• HH22 cooling in minihalos is strongly suppressed cooling in minihalos is strongly suppressed

for a soft UV background offor a soft UV background of J(LW) J(LW) ≳≳ 0.01 0.01 10 10-21-21 erg s cm erg s cm-2-2 Hz Hz-1-1 sr sr-1-1

• Transient UV photo-heating strengthens negativeTransient UV photo-heating strengthens negative feedback near sources, where flux isfeedback near sources, where flux is J(UV) J(UV) ≳≳ 0. 1 0. 1 10 10-21-21 erg s cm erg s cm-2-2 Hz Hz-1-1 sr sr-1-1

• Smallest halos with MSmallest halos with Mhalo ~ 10 ~ 1066 M⊙ most vulnerable most vulnerable

• Feedback switches from UV to LW at ~100 MyrFeedback switches from UV to LW at ~100 Myr • For comparison, flux needed to ionize universe isFor comparison, flux needed to ionize universe is J(ion) J(ion) 10 10 10 10-21-21 erg s cm erg s cm-2-2 Hz Hz-1-1 sr sr-1-1

Feedback retards reionization when fion ≳ 0.1-1%

• Many pieces of evidence that bright quasar phase is short:

— 107 years ≲ tQ ≲108 years (e.g. Martini 2004)

• Fiducial recombination time in z>6 IGM:

— trec ≈ tHubble ≈ 5 108 years at mean density at z=8

— fossils outnumber active bubbles by factor trec /tQ ≈ 5-50

• Fossils affect the IGM, and are useful probes:— large (40-50 comoving Mpc), prime targets for 21cm imaging

— probe quasar properties (Wyithe, Loeb & Barnes 2005;

Zaroubi & Silk 2005; Kramer & ZH 2007)

— probe IGM properties (Lidz et al., Alvarez & Abel, Geil & Wyithe)

— entropy floor even in recombined fossils (Oh & ZH 2003)

— H2 formation (Ricotti et al. 02; Kuhlen & Madau 05; Mesinger et al. 07)

Fossil Quasar Bubbles

• Recombination must be inhomogeneous:

— over-dense regions recombine quickly

— under-dense regions remain ionized for longer than trec

• Pre-existing galaxies:

— mean free path in fossil starts much higher than outside

— can pre-existing galaxies keep most of the fossil ionized?

(easier than to ionize the region to begin with)

• How do we distinguish fossils?— “grey” bubble: reduced contrast relative to active bubbles

but ionization nearly uniform— large size distinguishes them from rare large galaxy-bubbles

How does HII region recombine?Furlanetto, Haiman & Oh (2008)

Fossil Recombination With Zero Flux

• Assume bg=0

• Follow -dependent recombination

• cf: equivalent crit

with P() from MHR00 Miralda-Escude, Haehnelt& Rees (2000)

• Compute m.f.p. (including the under-dense voids)

• m.f.p. remains ~Mpc if xHI ≲10-3 (at ~1)

z=14

z=6

z=6

z=9

Fossil Evolution vs Global Reionization• Semi-analytical reionization model

• Follow mean xHI in z=10 fossil vs globally

• Additionally: follow -dependent xHI inside fossils

• Compute m.f.p. using MHR00

• c.f. clustering length of pre-existing galaxies • uniform, high ionization

An Early Fossil (z=15)

• Probably much rarer than fossils from z=10

• Still remains highly ionized

• different from z=10 fossil: m.f.p. drops below galaxy clustering length

• will develop (reduced contrast) swiss-cheese

Check validity of MHR00 m.f.p.

• Assume uniform bg

• neglect self-shielding

• Compute optical depth across one MHR00 m.f.p., including the under-dense voids

• z=10 fossil: xHI ≲10-3 =0.9, 0.5, 0.4

• z=15 fossil: =13, 4.5, 1.5, 0.7, 0.1

z=14

z=7

z=7

z=9

ImplicationsImplications

• Fossils outnumber active bubbles, last longer than t(rec)

• Fossils produced at z ≲10 remain highly and

uniformly ionized “grey zones”: look similar in 21cm to

active bubbles, but with a reduced contrast

• Example: XHI~ 10-20% in fossil, 70-80% outside.

• Nearly uniform ionization in fossil, swiss-cheese outside.

• Analogous fossils expected during helium reionization

• Makes “double-reionization” difficult to arrange