Primordial and Doppler modulations with Planck

22
Primordial and Doppler modulations with Planck Antony Lewis On behalf of the Planck collaboration http://cosmologist.info/

Transcript of Primordial and Doppler modulations with Planck

Page 1: Primordial and Doppler modulations with Planck

Primordial and Doppler modulations with Planck

Antony Lewis

On behalf of the Planck collaboration

http://cosmologist.info/

Page 2: Primordial and Doppler modulations with Planck

• Primordial modulations and power asymmetry• 𝜏𝑁𝐿 trispectrum• Kinematic Doppler dipoles

Outline

𝑇 ∼ 𝑃(𝑇|Ω) is statistically anisotropic in direction Ω

⇒ 𝑇 ∼ ∫ 𝑃(𝑇, Ω)𝑑Ω is statistically isotropic and non-Gaussian

Note: statistical anisotropy ≡ trispectrum

Page 3: Primordial and Doppler modulations with Planck

z~1000

14 000 Mpc

z=0θ

Squeezed shape ⇒ large-scale modulations⇒

Primordial curvature modulation:

Page 4: Primordial and Doppler modulations with Planck

e.g. local NG

Long + short modes: 𝜁0 = 𝜁𝑠 + 𝜁𝑙

i.e. modulated 𝜁 ∼ 𝜁𝑠 1 + 𝜙 with 𝜙 =6𝑓𝑁𝐿

5𝜁𝑙

Large-scale modulations ⇒

CMB bispectrum ∼6

5𝑓𝑁𝐿𝐶𝐿

𝑇𝜁∗(𝐶𝑙2 + 𝐶𝑙3)

CMB trispectrum ∼6

5𝑓𝑁𝐿

2𝐶𝐿𝜁∗𝜁∗(𝐶𝑙1 + 𝐶𝑙2) (𝐶𝑙3 + 𝐶𝑙4)

𝑙2

𝑙3𝐿

𝑙4

𝐿

𝑙3

𝑙1𝑙2

Page 5: Primordial and Doppler modulations with Planck

Define 𝜏𝑁𝐿 trispectrum by (almost all S/N at 𝐿 < 10, half in dipole)

Pearson, Lewis, Regan arXiv: 1201.1010

𝑓 =6𝑓𝑁𝐿

5𝜁𝑙 + 𝜒

(optimal to percent level)

Combined estimator for nearly scale-invariant modulation

Note 𝑓 ∼ 𝑂(10−3) ⇒ 𝜏𝑁𝐿 ∼ 500

𝑓 =6𝑓𝑁𝐿5

𝜁𝑙

Page 6: Primordial and Doppler modulations with Planck

Pipeline almost identical to CMB lensing, but with different weight functions.General anisotropy estimator is

Just need to reconstruct 𝑓( 𝐧) to find its power spectrum

QML estimator for f:

Hanson & Lewis 2009

Mean field(estimate from sims)

Optimally filtered temperature

Page 7: Primordial and Doppler modulations with Planck

Modulation mean field: mainly noise + mask

143x143

143x217

Avoids uncertainty in noise modelling(mask is well known)

Page 8: Primordial and Doppler modulations with Planck

Kinematic dipole signal

Aberration

𝒏 → 𝒏 + ∇( 𝒏 ⋅ 𝒗)

- just like a dipole lensingconvergence

Modulation

Illustrated for 𝑣

𝑐= 0.85

Challinor & van Leeuwen 2002

Page 9: Primordial and Doppler modulations with Planck

known dipole amplitude and direction

𝑣

𝑐≈ 1.23 × 10−3

Modulation 𝑓 = 𝑥 coth𝑥

2

2− 1 𝒏 ⋅ 𝒗 ≡ 𝑏𝜈 𝒏 ⋅ 𝒗

100 GHz143 GHz217 GHz353 GHz545 GHz857 GHz

Planck maps: 1.5235914

Approx boost factor Map modulation amplitude

0.18%0.24%0.37%0.64%1.1%1.7%

𝑏𝜈 = 𝑥 coth𝑥

2/2 − 1

Note: SMICA maps are a complicated mixture; modulation effect not currently included in FFP6 sims

Use 143, 217 only (with dust subtraction from 857)

Page 10: Primordial and Doppler modulations with Planck

Dipole kinematic effect using appropriate quadratic estimators

Note: not included in parameter analysis

- bias due to aberration average over mask ∼ 0.25𝜎

Page 11: Primordial and Doppler modulations with Planck

143x217 modulation reconstruction (𝐿 ≤ 5)- map of estimated modulation field 𝑓

Kinematics not subtracted

Kinematics subtractedin mean field from sims

Page 12: Primordial and Doppler modulations with Planck

Modulation pseudo-power spectrum

𝑓

Consistent with zero except for anomalous octopole

Page 13: Primordial and Doppler modulations with Planck

Octopole signal varies between frequencies:(large auto-quadrupole expected from noise bias)

Anomalous signal seems to be mostly due to 217

- may be related to 𝛽× (frequency dependence from dust?)

Page 14: Primordial and Doppler modulations with Planck

Estimator result

Gaussian simulations:

Note: signal most L<5 - small number of modes

Skewed distribution

Upper limits weaker than you might expect

Conservative upper limit, allowing octopole to be physicalusing Bayesian posterior

Consistent with Gaussian null hypothesis (octopole has small weight)

Planck 𝝉𝑵𝑳 trispectrum constraint 𝐿max = 2 𝐿max = 10

Page 15: Primordial and Doppler modulations with Planck

Scale-dependent dipole modulation and power asymmetries

Full analysis suggests no non-kinematic dipole power asymmetry

Can also look for scale-dependent effect: filter range of scales used in quadratic estimator

(new results, thanks Duncan)

Page 16: Primordial and Doppler modulations with Planck

Planck 217x143 (kinematic subtracted)WMAP 5 (Hanson & Lewis 2009)

Power modulation dipole amplitude for 𝑙 ≤ 𝑙max

Modulation < 1% for 𝑙 ≤ 𝑙max = 500 Modulation < 0.2% for 𝑙max = 1500 − 2000

1%

0.1%

Page 17: Primordial and Doppler modulations with Planck

Kinematic subtracted

Kinematics not subtracted

(as in Doppler paper but here pure modulation estimator)

Power dipole directions (𝑙 ≤ 𝑙max)

Page 18: Primordial and Doppler modulations with Planck

Power dipoles in Δl = 100 bands

Kinematic subtracted

Kinematics not subtracted

Page 19: Primordial and Doppler modulations with Planck

Conclusions• 5𝜎 detection of kinematic dipole effects in Planck maps

• Large-scale modulation power “nearly” consistent with zero after kinematic subtraction (foreground octopole?)

• Conservative limit 𝜏𝑁𝐿 < 2800 (95% CL)

• Power at 𝑙 ≤ 400 consistent with WMAP and previous analyses(must be – maps looks the same)

• Dipole power modulations at low L do not persist to high L after kinematic subtraction: |f| < 0.2% at 𝑙max = 2000.(but possible foreground issues, ongoing work..)

• Kinematic effects currently not included in Planck isotropy paper results, e.g. hemisphere and patch anisotropy constraints.Different model, mask, maps, filtering… so not directly comparable; ongoing work…

Page 20: Primordial and Doppler modulations with Planck
Page 21: Primordial and Doppler modulations with Planck

𝑙 ≥ 1100 only

Page 22: Primordial and Doppler modulations with Planck

SMICA (sims subtract aberration but not modulation)

Power in Δl = 100 bands

Cumulative