Cosmology with HI Intensity Mapping: Meerkat and the SKA

Cosmology with HI intensity mapping: MeerKAT and the SKA

Mário G. Santos, University of the Western Cape/SKA South Africa GR Effects in Cosmological LSS, Sesto, 2018 The Square Kilometre Array (SKA)

• Built in 2 phases • Phase 1: 2025? (first light) • Phase 2: 2030? – 10x phase 1… • Built in 2 countries: • South Africa (maybe also in Botswana, Ghana, Kenya, Madagascar, Mauritius, Mozambique, Namibia and Zambia) • Australia (maybe also New Zealand) • Phase 1 (went through a “re-scoping” - cost cap at 650M Euros) • Phase 1 with 2 different instruments: SKA1-MID

• South Africa • 130, 15m dishes plus 64, 13.5m dishes (MeerKAT) • Single pixel feeds • Up to 120 Km baselines • From 350 MHz to 14 GHz • Cosmology: 0 < z < 3.0 SKA1-LOW

• Australia • About 121,000 dipole antennas over 476 stations • up to 40 Km baselines • 50 – 350 MHz • 3.0 < z < 27 • Mostly to probe the Epoch of Reionization… MeerKAT: The SKA percursor MeerKAT: a radio interferometer

Correlate (multiply) radio waves from each dish in the array to make a single image. The larger the distance (baseline) the better the angular resolution. 8 Km ~ 6 arcsec Milky way center, MeerKAT official image

Credit: SARAO - www.ska.ac.za 7 Cosmology?

• Cosmic Microwave Background – great achievements but… • Dark energy and dark matter? • Nature of primordial density fluctuations? • Is the Universe really statistically isotropic and homogeneous? • Does General relativity fail on large Cosmological scales?

• We need 3 dimensional maps of the Large Scale Structure of the Universe!

2dF sky survey 8 Proposed SKA1-MID Cosmology Surveys

• 3 types of surveys:

• Continuum galaxy survey (both number counts and weak lensing)

• HI intensity mapping

• HI galaxy survey (to a lesser extent – see later)

• All very large (5,000 to 20,000 deg2). Try to do simultaneously to save time…

9 SKA1 Continuum survey

• Advantage: Strong signal - large number of galaxies • Disadvantage: no redshift information • Survey ~ 20,000 deg2, band 1 • Resolution? Need ~ 0.5 arcsec resolution for morphological classification of sources • flux sensitivity ~ 5 uJy rms (~10,000 hours) • Covers huge volume – up to z~4 • ~ 0.5 billion galaxies! – allows for strong tests of isotropy, e.g. dipole

NVSS (VLA) continuum survey Each point represents a galaxy 10 SKA1 Continuum Survey

• Check consistency between galaxies and the CMB!

} In standard cosmology, the dipole of the large-scale matter distribution should agree with the dipole of the CMB

} SKA1 will measure the amplitude and direction of the dipole (~few degrees accuracy)

} So far, directions agree but not the amplitude (Bengaly et al. 2018, Singal 2011, Rubart & Schwarz 2013, etc)

Schwarz, et al. 2015 SKA1-MID weak lensing survey

• SKA1-MID band 2 • ~ 5,000 deg2 • ~ 2.7 sources/arcmin2 • Cross correlation will be crucial to remove systematics for cosmology with weak lensing!

Camera et al. 2016 12 Galaxy redshift surveys

• High precision 3d positions – but we need redshifts! • For the radio: HI 21cm line from each galaxy • Very expensive! • SKA1: ~ 10 million HI galaxies over 5,000 deg2 at z<0.4 • Not enough for our cosmology needs…

13 But do we need to detect galaxies?

• Cosmology only cares about scales much larger than a galaxy • Using only detected galaxies 'wastes' photons…

High SNR: ignored information detected galaxies

14 Solution: intensity mapping

• Intensity mapping is very fast → uses all the photons • Provides high frequency/redshift resolution (in the radio…)

galaxies Intensity map

Battye, et al. 2004, MNRAS, 355, 1339 Bharadwaj, et al. 2001, Journal of Astrophysics and Astronomy, 22, 21 Loeb, A. & Wyithe, J. S. B. 2008, Physical Review Letters, 100, 15161301 ……… Other lines?

• HI (21 cm signal – 1.4 GHz) • At z>6, HI signal dominated by IGM emission • At z<3 (e.g. Cosmology), HI inside galaxies – value ~ 0.2 mK -14 2 (~ 10 erg/s/cm /Sr) 16 Current measurements…

Chang et al., Nature 2010 (GBT) Anderson et al., MNRAS Masui, et al., ApJ 2012 (GBT) 2018 (Parkes) HI cross-correlation with Wigglez Cross-correlation with 2dF (0.057 < z < 0.098)

Detections in cross-correlations only -3 GBT (Switzer et al., 2013): ΩHibias » 0.6£ 10 at z~0.8 17 Experiments…

BINGO (Battye, et al., - GBT http://arxiv.org/abs/1209.1041) - Parkes

- CHIME (Canada) - Tianlai (China)

18 SKA1-MID as an intensity mapping “machine”

SKA1-MID (~200 dishes by 2025)

• Interferometer: baselines not small enough to probe BAO scales and above • Main idea: use each dish in “single observation mode” • Save interferometer data for other science/calibration • SKA1-MID HI intensity mapping survey will turn SKA into a state of the art cosmology machine • Only way to really go after the unexplored very large scales (specially in combination with LSST) • Papers: Santos et al., arXiv:1501.03989; Bull et al., arXiv:1405.1452; arXiv: 1509.07562

Note: Also going to be done with SKA1-LOW in interferometer mode – EoR and z>3 cosmology! BAO scales 19 High precision cosmology with SKA1-MID HI intensity mapping survey

Hubble rate Growth rate

P. Bull, arXiv:1509.07562, 2015 } Use SKA1-MID band 2 (z < 0.8) } Low z measurements will be unmatched and surpass contemporary spectroscopic galaxy surveys such as DESI and Euclid in terms of constraints on modified gravity parameters

20 What about really large scales?

• ? ? ? Scales past the equality peak? ? ? (never measured before) ? • “Smoking gun” for new physics? • Require general relativistic Corrections • Probe of primordial fluctuations

SDSS/BOSS 21 Primordial Non-Gaussianity?

• Primordial � is usually assumed Guassian – fNL = 0 • Planck: fNL < 6.5 (using bispectrum) • Dark matter NG affects the clustering of object -> scale-dependent biasing

Correction to halo bias: Z=2.5

(bias 1) f z=0.4 f =10 / k2 NL NL

SKA1-MID using power spectrum: fNL < 3 (assuming a conservative bias for HI IM) (Camera et al., PRL, 2013; Camera et al. Afonso et al. ApJ 2015) Matarrese and Verde, Astrophys.J. 2008, Dalal et al., PRD 2008, … Can we do better? Beating cosmic variance on large scales

• The multi-tracer technique: 1 realization of dark matter field (our Universe)

Galaxy survey 1 Galaxy survey 2 (Dark matter halo of (Dark matter halo of mass M1) mass M2)

± » b ± 1 1 m ±2 » b2±m

Can measure directly b1/b2 (plus shot noise) – no cosmic variance! This can bring a new take on the way we design surveys… SKA1-MID HI intensity mapping and the multi-tracer technique

Fonseca et al., ApJ Letters, 2015

• Combining a HI intensity mapping survey using SKA1-MID with Euclid or LSST will detect fNL < 1 as well as GR corrections! • A powerful test of GR on large scales! Alonso and Ferreira, PRD, 2015 • A nice way to “fight” systematics Real life estimators?

• If: • Then:

No foreground residuals… 25 Multi-tracer simulations

• Full sky simulations for HI IM and LSST galaxies • Include all instrumental effects for SKA and LSST (noise, sky mask, beam, etc) • Include HI IM foregrounds and cleaning method • Generate many simulations to test the error and bias

• We show constraints on the bias ratio –> fNL or GR effects will generate an l dependence for small l

Witzemann, Alonso, Fonseca and Santos, in prep.

26 Simulation results (preliminary)

z=0.8, no noise, fg free z=0.8

102 102 N N

/ 1

/ 1 10

10 S S auto auto cross cross CV CV 100 opt 100 opt

101 102 101 102 ` ` A. Witzemann et al.

CV – best we could do without multi-tracer (zero noise)

27 Do foregrounds bias the results?

A. Witzemann et al.

The cross-correlation estimator is unbiased!

28 The near future: an SKA cosmology survey precursor?

• MeerKLASS: MeerKAT Large Area Synoptic Survey, http://arxiv.org/abs/1709.06099 • Aim: Cosmology (IM/continuum) and lots of other stuff • ~4,000 deg2 down to ~ 5 uJy continuum in L-band (most sensitive radio survey on these scales) • Take advantage of multi-wavelength data for cross-correlation • Overlap with DES: Cross-correlations will help clean foregrounds but also improve redshifts of DES!

+10r

0r HH56 682 HHLM6 ACTPRl D56

Dec. (-2000) Dec. CFHT W1 -10r

04h 03h 02h 01h 00h 23h 22h R.A. (-2000)

0.0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0 TemperDture (m.) 29 BAO and Redshift Space distortions with MeerKAT

S/N DA/σ H/σ fσ8/σ z=0.4 25 33 36

z=1.0 12 20 17

} Also, primdordial fNL constraints: ◦ Planck ~ 6.5 ◦ DES ~ 12 ◦ MeerKAT L-bandxDES ~ 3.5 ◦ MeerKAT UHFxDES ~ 2.3 • Strong detections of correlations with DES Fonseca et al., arXiv1611.01322 • Strong detection of baryon acoustic oscillations using the intensity mapping technique… • Constraints on growth rate will surpass the BOSS survey and be unmatched at low z on near future

30 Challenges with IM observations…

} Need to remove everything else that falls in our pixel! } Main contaminant: galactic synchrotron (about 1000 times stronger) } Other lines (OH, CH) not a concern } Note: ionosphere not really a problem at these frequencies…

Alonso, Ferreira and Santos, 2014, arXiv:1405.1751 T [mK] Simulations: http://intensitymapping.physics.ox.ac.uk/CRIME.html 31 Foreground cleaning… Foregrounds frequency dependency

• We have lots of frequency channels (compare to CMB!) • Take advantage of extra information: foregrounds are smooth across frequency while signal fluctuates • Results show that signal can be extracted with great accuracy (e. g. Alonso, et al., MNRAS 2014) • Polarisation leakage might be a problem…

} Methods: Polynomial fitting (Gleser et al. 2008, Liu et al. 2009), CCA (Ricciardi et al., 2010), Wp smoothing (Harker et al. 2009), FastICA (Hyvärinen et al. 1999, Chapman et al. 2012, Wolz et al. 2013), GMCA (Chapman et al. 2013), GNILC (Olivari et al.), Bigot-Sazy et al.

32 Foregrounds and other complications...

• Real problem is instrument fluctuations+foregrounds (primary beam effects, gain stability) • Most of the 1/f noise can be removed in frequency • Cross-correlating the “auto-correlation” between different dishes might help

33 Summary

• Observations at radio wavelengths will soon produce a wealth of data with profound implications for cosmology (Continuum, Weak Lensing, HI galaxies, HI intensity mapping, and combinations…)

• HI intensity mapping opens up a new possibility for SKA1: tests of dark energy and modified gravity and in particular, testing the cosmological model on very large scales

• Huge potential in cross-correlating with optical surveys!

Ø The future is now: an HI intensity mapping survey with MeerKAT – on the path to the SKA!