Lensing reconstructions 02.00

From CMB-S4 wiki
Jump to: navigation, search

J.Carron 27 Sept. 2018

This page documents the 600 lensing maps (healpy alm fits files, lmax=3000, temperature, polarization and MV reconstruction) available at

     /project/projectdirs/cmbs4/reanalysis/phi_recons/02.00_carron_180920/plms_sepTP_apo_lmin200lmax3000_fsky5pc_cut10_180925

These lensing maps were built from these CMB + noise + foreground maps (described at Sim map sets to demonstrate "real_delensing" (02.00 and 02.09)):

     /project/projectdirs/cmbs4/data_xx.yy/02.00/cmbs4_02p00_comb_f145_b04_ellmin30_map_2048_mc_{0000 to 0099}.fits

The lensing maps have the usual quadratic esimator normalization (unbiased with respect to the input lensing, to a good approximation, with a lot of noise on small scales), and have contribution in them from the mask and noise anisotropies (mean-field). Maps further processed which are ready to use for delensing are described in Ready for delensing use lensing maps 02.00

For illustration, the temperature (TT), polarization (PP) and Minimum Variance (MV) Wiener-filtered displacement reconstructions, together with the input, for realization 99:

Recmaps cmbs4 apo.png

There are 6 type of lensing maps, labeled

  • sim_ptt_????_lmax3000.fits

lensing map from temperature only, simulation index ???? 0000 to 0099

  • sim_p_p_????_lmax3000.fits

lensing map from polarization only

  • sim_p_????_lmax3000.fits

lensing map from temperature and polarization (MV)

  • sim_xtt_????_lmax3000.fits

lensing curl map from temperature, useful for null tests

  • sim_x_p_????_lmax3000.fits

lensing curl map from polarization

  • sim_x_????_lmax3000.fits

lensing curl from temperature and polarization (MV)

The maps were built using the quadratic estimator implementation described in the Planck 2018 CMB lensing paper, with the exception of the filtering, which is isotropic after application of a slightly apodized mask.

Map auto-spectra and mean-fields

The raw auto-spectra of the map typically look like the left panel of this figure:

Autospecs rec2.png

On large scales there is a very srong contribution from the mask and noise variance map anisotropies (the mean-field). On small scales the recontruction noise N0 takes over.

The mean-field has not been subtracted out from the maps on disk. This can be performed subtracting the lensing reconstructed maps averaged over a subset of the simulations. The right panel of the figure above shows the spectra of the same realization, after mean-field subtraction.

The mean-field spectra themselves look like this:

MF cmbs4.png

On small scales there is large Monte-Carlo noise in the spectrum estimate.

Fidelity to the input, delensing efficiency

The empirical cross-correlation correlation to the input maps, as calculated across the entire patch is like this:

Cmbs4rho.png

The square of this is the expected delensing efficiency. The true delensing efficiency will be bit higher in the center, and a bit lower in the edges.

In order to use one of these maps for delensing, it is necessary to:

  • Subtract a mean-field estimate (for instance obtained as the average of the remaining simulations)
  • Wiener-filter the map. This can be performed multiplying the alm by the ratio C / C + N0 estimates for these maps provided in the last column of the files
     /project/projectdirs/cmbs4/reanalysis/phi_recons/02.00_carron_180920/plms_sepTP_apo_lmin200lmax3000_fsky5pc_cut10_180925/nlpp_{ptt, p_p, p}.dat

Lensing reconstruction fiducial parameters

  • CMB multipoles from 200 to 3000
  • Analysis mask: fmask.fits in same folder
  • Gaussian beam of 4 arcmin FWHM
  • Flat noise of 1.5 (T) and 2.12 (Pol.) muK-arcmin
  • FFP10 fiducial CMB spectra
  • Separate temperature and polarization filtering