Probing the neutral intergalactic medium during cosmic reionization using the 21cm line of hydrogen

KIAA-PKU Summer School, Beijing, China

Chris Carilli, NRAO, August 10, 2007

Introduction: What is Cosmic Reionization?

Current constraints on the IGM neutral fraction with cosmic epoch

Neutral Intergalactic Medium (IGM) – HI 21cm signals

Low frequency telescopes and observational challenges

References

Reionization and HI 21cm studies of the neutral IGM

“Observational constraints on cosmic reionization,” Fan, Carilli, Keating 2006, ARAA, 44, 415

“Cosmology at low frequencies: the 21cm transition and the high redshift universe,” Furlanetto, Oh, Briggs 2006, Phys. Rep., 433, 181

Early structure formation and first light

“The first sources of light and the reionization of the universe,” Barkana & Loeb 2002, Phys.Rep., 349, 125

“The reionization of the universe by the first stars and quasars,” Loeb & Barkana 2002, ARAA, 39, 19

“Observations of the high redshift universe,” Ellis 2007, Saas-Fe advanced course 36

Ionized f(HI) ~ 0

Neutral f(HI) ~ 1

Reionized f(HI) ~ 1e-5

History of Baryons in the Universe

Chris Carilli (NRAO)

Berlin June 29, 2005

WMAP – structure from the big bang

Hubble Space Telescope Realm of the Galaxies

Dark Ages

Twilight Zone

Epoch of Reionization

• Last phase of cosmic evolution to be tested • Bench-mark in cosmic structure formation indicating the first luminous structures

Dark Ages

Twilight Zone

Epoch of Reionization

• Epoch?• Process?• Sources?

Cen 2002

Some basics

• Stellar fusion produces 7e6eV/H atom.

• Reionization requires 13.6eV/H atom

=>Need to process only 1e-5 of baryons through stars to reionize the universe

At z>6:

tuniv ~ 0.55 [(1+z)/10]-3/2Gyr

z > 8: trecomb < tuniv

Age of the Universe IGM recombination time

QuickTime™ and aYUV420 codec decompressor

are needed to see this picture.

Gnedin 03

Reionization: the movie

8Mpc comoving

Barkana and Loeb 2001

Constraint I: Gunn-Peterson Effect

z

Gunn-Peterson Effect toward z~6 SDSS QSOs

Fan et al 2006

Gunn-Peterson limits to f(HI)

• to f(HI) conversion requires ‘clumping factor’

• >>1 for f(HI)>0.001 => low f() diagnostic

• GP => Reionization occurs in ‘twilight zone’, opaque for obs <0.9 m

GP = 2.6e4 f(HI) (1+z)3/2

End of reionization?

f(HI) <1e-4 at z= 5.7

f(HI) >1e-3 at z= 6.3

But…

Contraint II: The CMB

Temperature fluctuations due to density inhomogeneities at the surface of last scattering (z ~ 1000)

Angular power spectrum ~ variance on given angular scale ~ square of visibility function

Baryon Acoustic Oscillations: Sound horizon at recombination

Sachs-Wolfe

Thomson scatting during reionization (z~10)

Acoustics peaks are ‘fuzzed-out’ during reionization.

Problem: degenerate with intrinsic amplitude of the anisotropies.

Reionization and the CMB

Constraint II: CMB large scale polarization -- Thomson scattering during reionization

Scattering CMB local quadrapole => polarized

Large scale: horizon scale at reioniz ~ 10’s deg

Signal is weak:

TE ~ 10% TT

EE ~ 1% TT

e = 0.084 +/- 0.016

~ L/mfp ~ Lne e (1+z)2

Hinshaw et al 2008

Constraint II: CMB large scale polarization -- Thomson scattering during reionization

Rules-out high ionization fraction at z> 15

Allows for finite (~0.2) ionization to high z

Most action occurs at z ~ 8 to 14

Dunkley et al. 2008

e = integral measure to recombination=> allows many IGM histories

Combined CMB + GP constraints on reionization

But…

• tuniv = 0.87Gyr• Lbol = 1e14 Lo

• Black hole: ~3 x 109 Mo (Willot etal.)• Gunn Peterson trough (Fan etal.)

Pushing into reionization: QSO 1148+52 at z=6.4

1148+52 z=6.42: Gas detection

Off channelsRms=60uJy

46.6149 GHzCO 3-2

• M(H2) ~ 2e10 Mo

• zhost = 6.419 +/- 0.001

(note: zly = 6.37 +/- 0.04)

VLA

IRAM

VLA

Constrain III: Cosmic Stromgren Sphere

• Accurate zhost from CO: z=6.419 +/- 0.001

• Proximity effect: photons leaking from 6.32<z<6.419

z=6.32White et al. 2003

‘time bounded’ Stromgren sphere: Rphys = 4.7 Mpc

Loeb & Rybicki 2000

‘time bounded’ Stromgren sphere: R = 4.7 Mpc

# ionizing photons = # atoms in volume Luv • tqso = 4/3R3 • <nHI>

=> tqso = 1e5 R3 f(HI)~ 1e7yrs or f(HI) ~ 1 (tqso/1e7 yr)

CSS: Constraints on neutral fraction at z~6

Nine z~6 QSOs with CO or MgII redshifts: <R> = 4.4 Mpc (Wyithe et al. 05; Fan et al. 06; Kurk et al. 07)

GP => f(HI) > 0.001

If f(HI) ~ 0.001, then <tqso> ~ 1e4 yrs – implausibly short given QSO fiducial lifetimes (~1e7 years)?

Probability arguments: f(HI) > 0.05

Wyithe et al. 2005

= tqso/4e7 yrs

90% probability x(HI) > curve

P(>xHI)

Difficulties for Cosmic Stromgren Spheres

(Lidz + 07, Maselli + 07)

Requires sensitive spectra in difficult near-IR band

Sensitive to R: f(HI) R^-3

Clumpy IGM => ragged edges

Pre-QSO reionization due to star forming galaxies, early AGN activity

ESO

OI

Not ‘event’ but complex process, large variance: zreion ~ 6 to 14

Good evidence for qualitative change in nature of IGM at z~6

OI

Saturates, HI distribution function, pre-ionization?

Abundance?

Integral measure?

Local ionization?

Geometry, pre-reionization?

Current probes are all fundamentally limited in diagnostic power

Need more direct probe of process of reionization = HI 21cm line

Local ioniz.?

Low frequency radio astronomy: Most direct probe of the neutral IGM during, and prior to, cosmic reionization, using the redshifted HI 21cm line: z>6 => 100 – 200 MHz

Square Kilometer Array

Advantages to the HI 21cm line

1. Spectral line signal => full three dimensional (3D) diagnostic of structure formation.

2. Direct probe of IGM = dominant component of baryons during reinization/dark ages

3. Hyperfine transition = forbidden (weak) => avoid saturation (cf. Ly ), can study the full redshift range of reionization.

Diagnostics:

1. When: direct measure of epoch of reionization

2. Process: Inside-out vs. outside-in reionization

3. Sources: Xrays vs. UV vs. shocks

4. Feedback due to galaxy/AGN formation

5. Exotic mechanisms: Very high z particle decay

1e13 Mo

1e9 Mo

HI mass limits => large scale structure

Reionization

Brightness Temperature• Brightness temp = measure of surface brightness (Jy/SR, Jy/beam, Jy/arcsec2)

•TB = temp of equivalent black body, B, with surface brightness = source surface brightness at : I = S / = B= kTB/ 2

• TB = 2 S / 2 k

• TB = physical temperature for optically thick thermal object

• TA <= TB always

Source size > beam TA = TB (2nd law therm.)

Source size < beam TA < TB

beam

source

telescope

TB

[Explains the fact that temperature in focal plane of optical telescope cannot exceed TB of a source]

HI 21cm radiative transfer: large scale structure of the IGM

LSS: Neutral fraction / Cosmic density / Temperature: Spin, CMB

Furlanetto et al. 2006

Spin Temperature

€

n1

n2

= 3e−hν 21 / kTs

1. Collisions w. e- and atoms

2. Ambient photons (predominantly CMB)

3. Ly resonant scattering: Wouthuysen-Field effect = mixing of 1S HF levels through resonant scattering of Ly drives Ts to Tkin

Ly

21cm

Each Ly photon scatters ~ 1e5 times in IGM before redshifting out of freq window.

h21/k = 0.067K

Dark Ages HI 21cm signal

•z > 200: T = TK = Ts due to collisions + Thomson scattering => No signal

•z ~ 30 to 200: TK decouples from T, but collisions keep Ts ~ TK => absorption signal

•z ~ 20 to 30: Density drops Ts~ T => No signal

TK = 0.026(1+z)^2

T = 2.73(1+z)

Furlanetto et al. 2006

Enlightenment and Cosmic Reionization -- first luminous sources

•z ~ 15 to 20: TS couples to TK via Lya scattering, but TK < T => absorption

•z ~ 6 to 15: IGM is heated (Xrays, Lya, shocks), partially ionized => emission

•z < 6: IGM is fully ionized

TK

T

Signal I: Global (‘all sky’) reionization signature

Signal ~ 20mK < 1e-4 sky

Possible higher z absorption signal via Lya coupling of Ts -- TK due to first luminous objects

Feedback in galaxy formation

No Feedback

Furlanetto, Oh, Briggs 06

Signal II: HI 21cm Tomography of IGM Zaldarriaga + 2003

z=12 9 7.6

TB(2’) = 10’s mK

SKA rms(100hr) = 4mK

LOFAR rms (1000hr) = 80mK

Signal III: 3D Power spectrum analysis

SKA

LOFAR

McQuinn + 06

only

+ f(HI)

PS dependence on neutral fraction

xi = 0.13

xi=0.78

z=10

Furlanetto et al. 2006

‘knee’ ~ characteristic bubble scale ~ 1 to 10Mpc (cm)

Inside-out vs. Outside-in

Furlanetto et al. 2004

z=12z=15

Inside-out Outside-in

Sensitivity of MWA for PS measurements (Lidz et al. 2007)

1yr, 30MHz BW, 6MHz chan

Will measure PS variance over k ~ 0.1 to 1 Mpc-1 (cm)

Sensitivity is maximized with compact array configuration (blue)

Sensitivity of MWA for PS measurements (Lidz et al. 2007)

Constrain amplitude of PS (variance) to 5 to 10

Constrain slope of PS to similar accuracy

• N(HI) = 1e13 – 1e15 cm^-2, f(HI/HII) = 1e-5 -- 1e-6 => before reionization N(HI) =1e18 – 1e21 cm^-2

• Lya ~ 1e7 21cm => neutral IGM opaque to Lya, but translucent to 21cm

Signal IV: Cosmic Web after reionization

Ly alpha forest at z=3.6 ( < 10)

Womble 96

z=12 z=819mJy

130MHz

• radio G-P (=1%)

• 21 Forest (10%)

• mini-halos (10%)

• primordial disks (100%)

Signal IV: Cosmic web before reionization: HI 21Forest

• Perhaps easiest to detect (use long baselines)

• ONLY way to study small scale structure during reionization

159MHz

Radio sources beyond the EOR

sifting problem (1/1400 per 20 sq.deg.)

2240 at z > 6

1.4e5 at z > 6

S120 > 6mJy

Signal V: Cosmic Stromgren spheres around z > 6 QSOs

0.5 mJy

LOFAR ‘observation’:

20xf(HI)mK, 15’,1000km/s

=> 0.5 x f(HI) mJy

Pathfinders: Set first hard limits on f(HI) at end of cosmic reionization

Easily rule-out cold IGM (T_s < T_cmb): signal = 360 mK

Wyithe et al. 2006

5Mpc

Dark age HI 21cm signal: baryon oscillations

Barkana & Loeb: “Richest of all cosmological data sets”

• Three dimensional in linear regime

• Probe to k ~ 103 Mpc-1 vs. CMB limit set by photon diffusion ~ 0.2Mpc-1

•Alcock-Pascinsky effect

•Kaiser effect + peculiar velocites

0.1 1.0 10 Mpc-1

Challenge: sensitivity at very low frequency

PS detection

• 1 SKA, 1 yr, 30MHz (z=50), 0.1MHz

•TBsky = 100(/200MHz)-2.7 K

= 1.7e4 K

At l=3000, k=0.3 Mpc-1

• Signal ~ 2 mK

• Noise PS ~ 1 mK

Requires few SKAs

BREAK

Challenge I: Low frequency foreground – hot, confused sky

Eberg 408 MHz Image (Haslam + 1982)

• Coldest regions: T ~ 100z)-2.6 K

• 90% = Galactic foreground

• 10% = Egal. radio sources ~ 1 source deg-2 with S140 > 1 Jy

Signal < 20mK

Sky > 200 K

DNR > 1e4

Solution: spectral decomposition (eg. Morales, Gnedin…)

Foreground = non-thermal = featureless over ~ 100’s MHz

Signal = fine scale structure on scales ~ few MHz

10’ FoV; SKA 1000hrs

Signal/Sky ~ 2e-5

Cygnus A

500MHz 5000MHz

Simply remove low order polynomial or other smooth function?

Cross correlation in frequency, or 3D power spectral analysis: different symmetries in frequency space for signal and foregrounds.

Freq

Signal Foreground

Morales 2003

Cygnus A at WSRT 141 MHz 12deg field (de Bruyn)

Frequency differencing ‘errors’ are ‘well-behaved’ ‘CONTINUUM’ (B=0.5 MHz) ‘LINE’ CHANNEL (10 kHz) - CONT

(Original) peak: 11000 Jy noise 70 mJy

dynamic range ~ 150,000 : 1

10 kHz channels

12o

Cygnus A w. PAPER/GB 130 MHz 30deg FoV

Frequency differencing with MHz channels doesn’t work well for far-out sidelobes due to chromatic aberration.

1MHz separation

5MHz separation

10o

‘stochastic noise term’

Calibration requirements for point source removal

20o x 20o FoV => expect brightest source ~ 34Jy at 158MHz

CSS detection: 10mK signal over 10’ = 73 uJy

required DNR = 34Jy/73uJy = 4.6e5

DNR ~ Nant / (21/2 x rms phase error in rad)

500 element array=> required rms phase error ~ 0.045o

Galactic foreground polarization‘interaction’ with polarized beams

frequency dependent residuals!

Solution: good calibration of polarization response

NGP 350 MHz 6ox6o ~ 5 K pol IF Faraday-thin 40 K at 150 MHz

WENSS: Schnitzeler et al A&A Jan07

30o x 30o

Speed of light in ionosphere is dependent function of ne

Phase (path length) variation proportional to wavelength2

Challenge II: Ionospheric phase errors – varying e- content

Sources wander by ~ arcmin, timescales of mins.

‘Isoplanatic patch’ = few deg = few km

Ionospheric storms wash-out sky completely

74MHz Lane 03

Challenge II: Ionospheric phase errors – varying e- content

TID

Arcmin seeing => problem is not fuzzing-out 21cm signal itself (typically larger scale) , but calibration errors leading to large (freq-dependent) sidelobes from bright sources

Solution

Direction dependent calibration: Wide field ‘rubber screen’ phase self-calibration = ‘peeling’

Virgo A 6 hrs VLA 74 MHz Lane + 02

QuickTime™ and aCinepak decompressor

are needed to see this picture.

15’

Ionospheric phase errors: The Movie

Challenge III: Interference

100 MHz z=13

200 MHz z=6

Solutions -- RFI Mitigation (Ellingson06)

Digital filtering: multi-bit sampling for high dynamic range (>50dB)

Beam nulling/Real-time ‘reference beam’

LOCATION!

Beam nulling -- ASTRON/Dwingeloo (van Ardenne)

Factor 300 reduction in power

VLA-VHF: 180 – 200 MHz Prime focus CSS search Greenhill, Blundell (SAO); Carilli, Perley (NRAO)

Leverage: existing telescopes, IF, correlator, operations

$110K D+D/construction (CfA)

First light: Feb 16, 05

Four element interferometry: May 05

First limits: Winter 06/07

Project abandoned: Digital TV

KNMD Ch 9

150W at 100km

RFI mitigation: location, location location…

100 people km^-2

1 km^-2

0.01 km^-2

(Briggs 2005)

Multiple experiments under-way: ‘pathfinders’

MWA (MIT/CfA/ANU) LOFAR (NL)

21CMA (China) SKA

QuickTime™ and aCinepak decompressor

are needed to see this picture.

Complication: ‘Tile’ diffraction beam

3

EDGE/CORE

EDGES (Bowman & Rogers MIT)

All sky reionization HI experiment. Single broadband dipole experiment with (very) carefully controlled systematics + polynomial baseline subtraction (7th order)

rms = 75 mK

VaTech Dipole Ellingson

300K

150K

Limits to the Global Step in the HI 21cm signal

Currently limited by systematics

After ~1/2 hour, rms no longer decreases as root time

GMRT 230 MHz – HI 21cm abs toward highest z (~5.2) radio AGN

0924-220 z=5.2

S230MHz = 0.5 Jy

1”

8GHz Van Breugel et al.

GMRT at 230 MHz = z21cm

RFI = 20 kiloJy !

CO Klamer +

M(H2) ~ 3e10 Mo

GMRT 230 MHz – HI 21cm abs toward highest z radio AGN (z~5.2)

rms(20km/s) = 5 mJy

229Mhz 0.5 Jy232MHz 30mJy

rms(40km/s) = 3mJy

N(HI) ~ 2e20TS cm-2 ?

Limits:

Few mJy/channel

Few percent in optical depth

Focus: Reionization (power spec,CSS,abs)

C. Carilli, A. Datta (NRAO), J. Aguirre (Penn)

PAPER: Staged Engineering• Broad band sleeve dipole + flaps

• 8 dipole test array in GB (06/07) => 32 station array in WA (2008) to 256 (2009)

• FPGA-based ‘pocket correlator’ from Berkeley wireless lab: easily scale-able

• S/W Imaging, calibration, PS analysis: AIPY + Miriad/AIPS => Python + CASA, including ionospheric ‘peeling’ calibration

100MHz 200MHz

BEE2: 5 FPGAs, 500 Gops/s

CygA 1e4Jy

PAPER/WA -- 4 Ant, July 2007

RMS ~ 1Jy; DNR ~ 1e4

Parsons et al. 2008

1e4Jy

Lunatic fringe: probing the dark ages from the dark side of the MoonC. Carilli (NRAO), Sackler Cosmology Conf, Cambridge, MA, 2008

Long History of Lunar Low Freq Telescope

Gorgolewski 1965: Ionospheric opacity

• Ionosphere p ~ 10 MHz

• ISM p ~ 0.1 MHz

• Interstellar scattering => size ~ 1o (/1 MHz)-2

• Faraday rotation => no polarization

• z > 140 => not (very) relevant for HI 21cm studies, ‘beyond dark ages’

New window

Lunar window

ion. cutoff ~ 30m

ISM cutoff ~ 3km

Return to moon is Presidential directive

Summary of STScI Workshop, Mario Livio, Nov. 2006

“The workshop has identified a few important astrophysical observations that can potentially be carried out from the lunar surface. The two most promising in this respect are:

(i) Low-frequency radio observations from the lunar far side to probe structures in the high redshift (10 < z< 100) universe and the epoch of reionization

(ii) Lunar ranging experiments…”

Our consensus: Lunar imperative awaits lessons from ground-arrays

Advantage I: Interference

Lunar shielding of Earth’s auroral emission at low freq (Radio Astronomy Explorer 1975)

Alexander + 1975

12MHz

ITU radio regulation Article 22: Far-side of Moon is a radio protected by international agreement

Clementine (NRL) star tracker

Advantage II: Very thin ionosphere -- critical at very low frequencieso Tenuous photoionized layer extending to 100km

o p = 0.2 to 1 MHz ~ 0.01 to 0.1 x Earth

o Large day/night variation

Other practical advantages

• Easier deployment: robotic or human

• Easier maintenance (no moving parts)

• Less demanding hardware tolerances

• Very large collecting area, undisturbed for long periods (no weather, no animals, not many people)

Apollo 15

• Array data rates (Tb/s) >> telemetry limits, requiring in situ processing, ie. low power super computing (LOFAR/Blue Gene = 0.15MW)

• RFI shielding: How far around limb is required?

• Thermal cycling (mean): 120 K to 380 K

• Radiation environment

• Regolith: dielectric/magnetic properties

Lunar challenges

Lunar shielding at 60kHz

Takahashi + Woan

Radio astronomy – Probing Cosmic Reionization

•‘Twilight zone’: study of first light limited to near-IR to radio

• First constraints: GP, CMBpol => reionization is complex and extended: zreion = 7 to 15

• HI 21cm: most direct probe of reionization

• Low freq pathfinders: All-sky, PS, CSS

• SKA: direct imaging of IGM

• Lunar array to probe the Dark Ages

END

CSS: Constraints on neutral fraction at z~6

Fan et al 2005