1. Constraining the redshift of reionization using a “modest” array
Jonathan Bittner
Advisor: Avi Loeb
MWA Meeting June 5th 2011

2. Global experiments can constrain the redshift of reionization
Bowman and Rogers, Nature 2011

3. Generically, the EOR has peak variance when universe is half-ionized
150 kHz 1’ 2’
300 kHz 4’
600 kHz
0.6° 1°
1.2 MHz
@150kHz 2°
@ 1arcmin

4. Why think about this? A simple constraint to compare to CMB and EDGES
Not an integrated constraint (like CMB), one which actually depends on rate at which reionization proceeded
First detection of cosmological 21-cm signal

5. The excess variance at z_reion can be measured over noise
signal
total
Ratio of signal / noise (z)
w/o
signal +
Signal profile and realization
“Effective noise”
For tint=500, Δν=2.4 MHz, θw = 1.2º

6. The redshift of reionization can be detected (in principle) with 32T

7. MWA 32T's dense uv-coverage would help overcome many issues
Without synthesis
With rotation synthesis
Judd Bowman, personal communication

8. Foreground subtraction works best at scales with good UV coverage
Liu, Tegmark, Zaldarriaga 2008
Bowman, Morales, Hewitt 2009

9. More information at: JCAP04(2011)038 arxiv:1006:5460
Thanks!
More information at:
JCAP04(2011)038
arxiv:1006:5460

10. Dense UV coverage should lessen problem of point-source “frizz”
Liu 2008
… the small-scale synthesized beam frizz seen in Figure 1 is largely averaged away when expanding the sky into long-wavelength Fourier modes, whereas the small-scale modes are severely affected. The characteristic scale separating “short" and “long" Fourier modes is determined by the longest baseline radius for which u-v coverage is complete.
Bowman 2009
However, after the residual map is transformed back to the Fourier
domain, it becomes evident that the polynomial fit has actually done an excellent job of subtracting the foreground contamination from baselines within a radius of u < 500λ, and only a poor job for baselines beyond this radius… where visibility measurements with the MWA become sufficiently sparse that there is no longer complete coverage.

11. We use 21-cm FAST to test this idea
Large cosmological volume (600 Mpc)
Low computational requirements – will run on workstation in < 1 day
Close match to simulations on large scales
z=9.25, xHI=0.41

12. We create simulated beams through data cube
Constant comoving width
Gaussian or square top-hat averaging
Random periodic boundary conditions
z=7 to z=15, Δz=0.25
Vary frequency and beam-width resolution
[For illustration only
Not a real 3D image]

13. We assume a very modest instrument
32 tiles, 32x16 dipoles (fixed)
Tile layout as given for MWA 32T by J. Bowman (private communication)
Atot = 460 m² at 158 MHz
Integration: Hours
(about 1000 per year is usable)
Varying:
Beam Width (θw): 4 arcmin to 2.4º
Frequency resolution: 150 KHz to 2.4 MHz θw
Furlanetto, Peng, Oh (2006).

14. δTb(z) and xHII in our realization

15. The effective thermal noise is the sampling error in the noise
Red line will differ due to finite resolution
Excess due to reionization

16. New definitions

17. The peaks occur roughly at xi=0.5 if taken at large scales