Accurate 21cm signal forecasts from simulations of ...

Accurate 21cm signal forecasts from simulations of inhomogeneous reionization

Girish Kulkarni – IoA Cambridge — July 15, 2015

Tirth Roy Choudhury (NCRA Pune), Martin Haehnelt (IoA Cambridge), Ewald Puchwein (IoA Cambridge)

Cosmological 21cm signal modelling

• Accurate simulations are especially important due to needle-in-a-haystack nature of the problem

• On scales , power spectrum of 21cm signal predicted to be ~ flat (McQuinn+ 2007; Zahn+ 2011; Friedrich+ 2011)

• But all simulations have a drawback: limited dynamical range

k = 0.1–0.3 Mpc�1

Lyman Limit Systems• Gas expected to have

rich small-scale structure, at least down to the Jeans scale.

• This acts as a sink for ionisation photons

• But importantly, small-scale structure can self-shield and become 21cm bright.

• This can affect large-scale power, but are missed by simulations

Kulkarni+ 2015

Issue: what is the effect of unresolved self-shielded structure on 21cm predictions?

SPH Simulations of the IGM• P-Gadget-3:

2048^3 gas particles in a 160 Mpc/h box (Bolton et al.)

• Mass resolution is ~4*10^6 Msun

• Mean inter-particle separation is ~53 comoving kpc.

z=7

Hybrid Excursion Set Method

• For “radiative transfer,” use excursion set method on the gas distribution (Mesinger & Furlanetto 2007, Choudhury+ 2009, Mesinger+ 2011)

• A point is ionised if sources/sinks > 1 in a sphere centred on it

• Gives large scale morphology but still misses self-shielding (Majumdar+ 2014)

z=7

Accounting for Sub-structure• To get sub-structure, we calibrate simulation to

given reionization history.

• For a given self-shielding criterion, solve global reionization evolution for photo-ionization rate (Choudhury+ 2015)

dQ

dt=

nion

nH� QM

trec

H II H I H II

Get consistent radiative transfer

Choudhury+ 2015

Self-shielded regions are now resolved

z=7

Self-shielded regions are now resolved

21cm brightness temperature

z=7

21cm-bright self-shielded regions are now resolved in HII bubbles

Tb ⇡ 22 mK xHI�gas

Effect on the power spectrum

x

ion

= 0.7

decrease on

large scales

increase on

large scales

LOFAR

LOFAR

Effect on the power spectrum

LLSs dominate towards end of EoR

LOFARLOFAR

“Redshift” dependence of this effect

zHM12 ~ 14 zHM12 ~ 6.5 zHM12 ~ 14 zHM12 ~ 6.5

Power is reduced by ~15%

“Redshift” dependence of this effect

zHM12 ~ 14 zHM12 ~ 6.5

Implications for Observations

• Look at other 1-point statistics now to consider observability (Watkinson and Pritchard 2014)

• These are important for constraining reionization models (Furlanetto+ 2004, Harker+ 2009)

• Smooth maps over ~ 2 cMpc

• Instrument noise for SKA (1000 hours) is ~3 mK (Watkinson and Pritchard 2014)

�2

Variance and Skewness

zHM12 ~ 14 zHM12 ~ 6.5

“Zero-Tb spike”

zHM12 ~ 14 zHM12 ~ 6.5

LLSs suppress strong evolution of moments

Conclusions

• Presented method to calibrate photon budget in high resolution simulations

• LLSs reduce large scale power throughout most of the reionization history by ~15%.

• This suppresses strong evolution of 21cm moments

• This makes constraining epoch of reionization using LOFAR and SKA more difficult