High-redshift 21cm

and redshift distortions

Antony LewisInstitute of Astronomy, Cambridge

http://cosmologist.info/

Anthony Challinor (IoA, DAMTP); astro-ph/0702600Richard Shaw (IoA), arXiv:0808.1724

Following work by Scott, Rees, Zaldarriaga, Loeb, Barkana, Bharadwaj, Naoz, Scoccimarro + many more …

Work with:

Hu & White, Sci. Am., 290 44 (2004)

Evolution of the universe

Opaque

Transparent

Easy

Hard

Dark ages

30<z<1000

CMB temperature

Why the CMB temperature (and polarization) is great

Why it is bad

- Probes scalar, vector and tensor mode perturbations

- The earliest possible observation (bar future neutrino anisotropy surveys etc…)

- Includes super-horizon scales, probing the largest observable perturbations

- Observable now

- Only one sky, so cosmic variance limited on large scales

- Diffusion damping and line-of-sight averaging: all information on small scales destroyed! (l>~2500)

- Only a 2D surface (+reionization), no 3D information

If only we could observe the CDM perturbations…

- not erased by diffusion damping (if cold): power on all scales - full 3D distribution of perturbations

What about the baryons?

- fall into CDM potential wells; also power on small scales- full 3D distribution- but baryon pressure non-zero: very small scales still erased

How does the information content compare with the CMB?

CMB temperature, 1<l<~2000: - about 106 modes - can measure Pk to about 0.1% at l=2000 (k Mpc~ 0.1)

Dark age baryons at one redshift, 1< l < 106: - about 1012 modes - measure Pk to about 0.0001% at l=106 (k Mpc ~ 100)

About 104 independent redshift shells at l=106

- total of 1016 modes - measure Pk to an error of 10-8 at 0.05 Mpc scales

What about different redshifts?

e.g. running of spectral index:

If ns = 0.96 maybe expect running ~ (1-ns)2 ~ 10-3

Expected change in Pk ~ 10-3 per log k

- measure running to 5 significant figures!?

So worth thinking about… can we observe the baryons somehow?

• How can light interact with the baryons (mostly neutral H + He)?

- after recombination, Hydrogen atoms in ground state and CMB photons have hν << Lyman-alpha frequency

* high-frequency tail of CMB spectrum exponentially suppressed * essentially no Lyman-alpha interactions * atoms in ground state: no higher level transitions either

- Need transition at much lower energy

* Essentially only candidate for hydrogen is the hyperfine spin-flip transition

Credit: Sigurdson

triplet

singlet

Define spin temperature Ts

Spontaneous emission: n1 A10 photons per unit volume per unit proper time

Rate: A10 = 2.869x10-15 /s

Stimulated emission: net photons (n1 B10 – n0 B01)Iv

0

1

h v = E21

Net emission or absorption if levels not in equilibrium with photon distribution- observe baryons in 21cm emission or absorption if Ts <> TCMB

What can we observe?

In terms of spin temperature:

Total net number of photons:

What determines the spin temperature?

• Interaction with CMB photons: drives Ts towards TCMB

• Collisions between atoms: drives Ts towards gas temperature Tg

• (+Interaction with Lyman-alpha photons - not important during dark ages)

TCMB = 2.726K/a

At recombination, Compton scattering makes Tg=TCMB

Later, once few free electrons, gas cools: Tg ~ mv2/kB ~ 1/a2

Spin temperature driven below TCMB by collisions: - atoms have net absorption of 21cm CMB photons

What’s the power spectrum?

Use Boltzmann equation for change in CMB due to 21cm absorption:

Background: Perturbation:

Fluctuation in density of H atoms,+ fluctuations in spin temperature

CMB dipole seen by H atoms:more absorption in direction of gas motion relative to CMB

l >1 anisotropies in TCMB

Doppler shiftto gas rest frame

+ reionization re-scattering terms

Solve Boltzmann equation in Newtonian gauge:

Redshift distortionsMain monopolesource Effect of local

CMB anisotropy

Tiny Reionization sourcesSachs-Wolfe, Doppler and ISW change to redshift

For k >> aH good approximation is

(+re-scattering effects)

21cm does indeed track baryons when Ts < TCMB

So can indirectly observe baryon power spectrum at 30< z < 100-1000 via 21cm

z=50

Kleban et al. hep-th/0703215

Observable angular power spectrum:

‘White noise’from smaller scales

Baryonpressuresupport

1/√N suppressionwithin window

baryon oscillations

z=50

Integrate over window in frequency

Small scales:

What about large scales (Ha >~ k)?

<~ 1% effect at l<50 Extra terms largely negligible

Narrow redshift window

14 000 Mpc

Dark ages~2500Mpc

Opaque ages ~ 300MpcComoving distance

z=30

z~1000

New large scaleinformation?- potentials etccorrelated with CMB

l ~ 10

Non-linear evolution

Small scales see build up of power from many larger scale modes - important

But probably accurately modelled by 3rd order perturbation theory:

On small scales non-linear effects many percent even at z ~ 50

+ redshift distortions, see later.

Also lensing

Lensing potential power spectrum

Lensed

Unlensed

Lewis, Challinor: astro-ph/0601594

Modified Bessel function

Wigner functions

c.f. Madel & Zaldarriaga: astro-ph/0512218

like convolution with deflection angle power spectrum;generally small effect as 21cm spectrum quite smooth

Cl(z=50,z=52)Cl(z=50,z=50)

Lots of information in 3-D (Zahn & Zaldarriaga 2006)

Observational prospects

No time soon…

- (1+z)21cm wavelengths: ~ 10 meters for z=50

- atmosphere opaque for z>~ 70: go to the moon?

- fluctuations in ionosphere: phase errors: go to moon?

- interferences with terrestrial radio: far side of the moon?

- foregrounds: large! use signal decorrelation with frequency

See e.g. Carilli et al: astro-ph/0702070, Peterson et al: astro-ph/0606104

But: large wavelength -> crude reflectors OK

Current 21cm:

LOFAR, PAST, MWA: study reionization at z <20SKA: still being planned, z< 25

Things you could do with precision dark age 21cm

• High-precision on small-scale primordial power spectrum(ns, running, features [wide range of k], etc.)e.g. Loeb & Zaldarriaga: astro-ph/0312134, Kleban et al. hep-th/0703215

• Varying alpha: A10 ~ α13 (21cm frequency changed: different background and perturbations)Khatri & Wandelt: astro-ph/0701752

• Isocurvature modes(direct signature of baryons; distinguish CDM/baryon isocurvature)Barkana & Loeb: astro-ph/0502083

• CDM particle decays and annihilations(changes temperature evolution)Shchekinov & Vasiliev: astro-ph/0604231, Valdes et al: astro-ph/0701301

• Primordial non-Gaussianity(measure bispectrum etc of 21cm: limited by non-linear evolution)Cooray: astro-ph/0610257, Pillepich et al: astro-ph/0611126

• Lots of other things: e.g. cosmic strings, warm dark matter, neutrino masses, early dark energy/modified gravity….

Back to reality – after the dark ages?

• First stars and other objects form• Lyman-alpha and ionizing radiation present:

Wouthuysen-Field (WF) effect: - Lyman-alpha couples Ts to Tg

- Photoheating makes gas hot at late times so signal in emission

Ionizing radiation: - ionized regions have little hydrogen – regions with no 21cm signal Both highly non-linear: very complicated physics

• Lower redshift, so less long wavelengths:- much easier to observe! GMRT (z<10), MWA, LOFAR (z<20), SKA (z<25)….

• Discrete sources: lensing, galaxy counts (~109 in SKA), etc.

How do we get cosmology from this mess?

1. Do gravitational lensing: measure source shears

- probes line-of-sight transverse potential gradients (independently of what the sources are)

Would like to measure dark-matter power on nearly-linear scales.

Want to observe potentials not baryons

2. Measure the velocities induced by falling into potentials

- probe line-of-sight velocity, depends on line-of-sight potential gradients

Redshift distortions

A closer look at redshift distortions

x

y

x

y(z)

Real space Redshift space

+

Density perturbed too. In redshift-space see

More power in line-of-sight direction -> distinguish velocity effect

Both linear (+higher) effects – same order of magnitude. Note correlated.

Linear-theory

Redshift-space distance:

Actual distance:

n

Define 3D redshift-space co-ordinate

Fourier transformed:

Transform using Jacobian: Redshift-space perturbation

Linear power spectrum:

Messy astrophysics Depends only on velocities -> potentials

(n.k)4 component can be used for cosmology

Barkana & Loeb astro-ph/0409572

Is linear-theory good enough?

RMS velocity 10-4-10-3 Radial displacement ~ 0.1-1 MPc

Megaparsecscale

Looks like big non-linear effect!

Real Space Redshift space

BUT: bulk displacements unobservable. Need more detailed calculation.

M(x + dx) ≠ M(x) + M’(x) dx

Non-linear redshift distortions

Power spectrum from:

Exact non-linear result (for Gaussian fields on flat sky):

Assume all fields Gaussian. Shaw & Lewis, 2008 also Scoccimarro 2004

Significant effectDepending on angle

z=10

Small scale power boostedby superposition of lots oflarge-scale modes

More important at lower redshift.Not negligible even at high z. Comparable to non-linear evolution

z=10

Similar effect on angular power spectrum:

(sharp z=z’ window, z=10)

(A ~ velocity covariance), u= (x-z,y-z’)

What does this mean for component separation?Angular dependence now more complicated – all μ powers, and not clean.

Assuming gives wrong answer…

Need more sophisticated separation methods to measure small scales.

z=10

Also complicates non-Gaussianity detectionRedshift-distortion bispectrum

• Mapping redshift space -> real space nonlinear, so non-Gaussian

Just lots of Gaussian integrals (approximating sources as Gaussian)..

Linear-theory source

Can do exactly, or leading terms are:

Not attempted numerics as yet

Zero if all k orthogonal to line of sight.

Also Scoccimarro et al 1998, Hivon et al 1995

Conclusions• Huge amount of information in dark age perturbation spectrum

- could constrain early universe parameters to many significant figures

• Dark age baryon perturbations in principle observable at 30<z< 500 up to l<107 via observations of CMB at (1+z)21cm wavelengths.

• Dark ages very challenging to observe (e.g. far side of the moon)

• Easier to observe at lower z, but complicated astrophysics

• Redshift-distortions probe matter density – ideally measure cosmology separately from astrophysics by using angular dependence

• BUT: non-linear effects important on small scales - more sophisticated non-linear separation methods may be required

Correlation function

z=10