Band Power Spectra and Parameters estimated from cleaned CMB maps

G. Rocha, 2018 September 19th

We report here the Band Power Spectra and parameters estimated for the component separation methods applied to Simulated data. For each method we estimate the full sky band power spectra with Xfaster, propagate to parameters with a Gaussian correlated Likelihood, and sample parameters with CosmoMC, marginalizing over tailored foreground residual templates. In addition we consider a hybrid likelihood: BFlike (pixel based likelihood) at low-l and XFlike (Xfaster likelihood) at high-l.

red|Updates (on October 4th, 2018)
- added BandPowers for Sevem maps all frequencies, case17 (90.00).
- added bandPowers (and pseudo-CL for TT) for NILC case 90.0, COMB - to exemplify the issues found for T and E, Currently trying to understand whether the issue is in the component separation step or misinterpretation of the inputs in the BandPower estimation.

red|Updates (on October 18th, 2018)
- added the marginal distribution for parameter r for Sevem map, case 17 (90.00), 155GHz.

red|Updates (on October 22th, 2018)
- added the BandPowers for Sevem maps at 62GHz, case17 (90.00) for full-sky - to investigate the lack of BB power in the cut-sky case (with mask not apodised).

red|Updates (on October 24th, 2018)
added the BandPowers for Sevem maps at 62GHz, case17 (90.00) for apodized mask - to investigate the lack of BB power in the cut-sky case with a mask not apodised.

red|Updates (on October 25th, 2018)
added the Bandpowers for the combined map, combined_90_108_129_155, Model 90, for 2 apodized masks and one non-apodized mask. The BB-spectra exhibits a lack of power at low-ell, for all masks. This is currently being investigated.

Band Power Spectra

SEVEM

155Ghz - case 90.00

Band Powers (BP) and uncertainties for SEVEM map at 155GHz, case 90.00 (first panel).
Also shown are the residuals of the estimated BP and the input spectra used to simulate the input maps (second panel).
The BP was estimated with XFpipe.

Inputs:

maps and mask are located at:

observation: /project/projectdirs/pico/barreiro/SEVEM/CASE17/ clean_155.fits
noise: /project/projectdirs/pico/barreiro/SEVEM/CASE17_NOISE/ clean_155.fits
mask: /project/projectdirs/pico/barreiro/SEVEM/mask0.15_pol_sm7deg_nu0p45_up512.fits
beam window function: /project/projectdirs/pico/XFpipe/Bl/ bl_gauss_6.2arcmin_155_TP.dat
Binning_scheme: CTP: /project/projectdirs/pico/XFpipe/bins/ CTP_bin_TT; CTP_bin_TE; CTP_bin_EE; CTP_bin_BB

input spectra: /project/projectdirs/pico/sky_yy/cmb/cls/ ffp10_lensedCls.dat

Outputs:

Spectra is located at /project/projectdirs/pico/XFpipe/spectra/
sevem_case17_155_6.2arcmin_uK_mask_l1500_Gfg.newdat
sevem_case17_155_6.2arcmin_uK_mask_l1500_Gfg.save

Band Power Spectra, Uncertainties and Residuals

The plots shown below are located at: /project/projectdirs/pico/XFpipe/spectra/plots/
sevem_155_TT.png; sevem_155_TE.png; sevem_155_EE.png; sevem_155_BB.png

Sevem Bandpowers for case17, 155Ghz, vs the input model (solid orange line) for lmax=1000


TT		TE


EE		BB

All frequencies - case 90.00

For Case17 - for all frequencies, xmax=1000

Sevem Bandpowers for all frequencies vs the input model (solid orange line) for lmax=1000


TT		TE


EE		BB

For Case17 - for all frequencies, xmax=300

Sevem Bandpowers for all frequencies vs the input model (solid orange line) for lmax=300


TT		TE


EE		BB

Test case for full-sky, 62GHz - case 90.00

Full-sky 62GHz case 90.00 lmax=1000

Sevem Bandpowers for case17, 62Ghz, full-sky (blue data points), vs the input model (solid orange line) for lmax=1000.

As shown: in the full-sky case the BB bandpowers recover well the BB input spectra. This suggests the lack of power for the cut-sky is caused by the (non-apodized) mask.

Next: estimation the BP for anodized mask


TT		TE


EE		BB

Full-sky, 62GHz, case 90.00 lmax=300

Sevem Bandpowers for case17, 62Ghz, full-sky (blue data points), vs the input model (solid orange line) for lmax=300.


TT		TE


EE		BB

Test case for apodized mask, 62GHz - case 90.00

Apodized mask, 62GHz case 90.00 lmax=1000

Sevem Bandpowers for case17, 62Ghz, with apodized mask (dark blue data points), vs the input model (solid orange line) for lmax=1000, and full-sky (light blue data points) and non apodized mask (black data points).

As shown: The BB bandpowers still exhibit deficit of power at low -ell. This seems to indicate that the mask is not the source of the problem. Note that the same mask kernel is applied to the other frequencies (plots above) which do not show deficit, in fact the input BB spectra is recovered well - this indicates that the kernel should be fine. On the other hand the full-sky case recovers well the input BB spectra indicating that the inputs should be fine as well (ie no mis-estimation of the noise or something else going on here) - hence it is unclear what is causing this lack power. I am currently investigating this and checking the iterations and convergence.


TT		TE


EE		BB

Apodized mask, 62GHz case 90.00 lmax=300

Sevem Bandpowers for case17, 62Ghz, with apodized mask (dark blue data points), vs the input model (solid orange line) for lmax=300, and full-sky (light blue data points) and non apodized mask (black data points).


TT		TE


EE		BB

Combined map - case 90.00

Sevem Bandpowers for the map combined_90_108_129_155, vs the input model (solid orange line) for lmax=1000

inputs in dir: /project/projectdirs/pico/reanalysis/compsepmaps/sevem_04.00_181011/MODEL90.91/

data: data_combined_90_108_129_155_fwf15.0d0.fits
noise: noise_combined_90_108_129_155_fwf15.0d0.fits

masks: /project/projectdirs/pico/reanalysis/compsepmaps/sevem_04.00_180815/
mask0.15_pol_sm7deg_nu0p45_up512_sm7deg_nu0p57_apocos5pix.fits - maskapo
mask0.15_pol_sm7deg_nu0p45_up512_apocos5pix.fits - maskapo_v1
mask0.15_pol_sm7deg_nu0p45_up512.fits - mask

While all spectra is recovered reasonably well for all masks, BB spectra exhibits lack of power at low-ell for both apodized and non-apodized masks. This is currently being investigated.


TT		TE


EE		BB

FG templates

To be added

NILC

Problem shown in the plots below - trying to understand whether the issue is in the component separation step or misinterpretation of the inputs for bandPower estimation.

BandPowers


TT		TE


EE		BB

Pseudo-cl for TT


TT

Parameters

SEVEM

Marginal distribution for parameter r for Sevem map, case 17, 155Ghz

Inputs:

- all parameter but r are fixed to the values of the model used to generate the input CMB simulation.

scalar_spectral_index(1): 0.96
re_optical_depth': 0.06
ombh2: 0.0219
omch2: 0.122
hubble: 68.0
scalar_amp(1): 2.1e-09

- Only BB spectra is used to estimate the posterior distribution of r.

Outputs:

- The distribution of r peaks close to zero with 1σ upper limit = 0.3251875E-02
- The BB best fit model underestimates the BB power spectra when compared to the observed APS for l>~300 - this indicates an underestimation of the r best fit value and lensing.

We are investigating this issue by:
- checking the setup used in the runs for tensor and lensing
- checking the fixed parameters, derived parameters, the covariance of the APS
- running parameter estimation using all spectra and with varying parameter values (for the parameters shown above).