Preliminary ILC maps - part 3

 (C. Umilta)


This posting shows preliminary ILC maps created with 06 sims, and follows this first and this second posting. Information on DC06 maps can be found here. There are three different sky masks, Pole deep and wide, and Chile deep. We consider three foregorund models, a simple Gaussian realization of dust and synchrotron, a Gaussian realization with modulated amplitude anda non-Gaussian realization of foregrunds. There are no extra galactic foregorunds in these simulations. More informations on 06 Data Challenge simulations can be found here. A total of 100 maps exists for each simulation set. I create here ILC maps for the first three realization of each set and compare them with the input CMB maps and 95 GHz maps.

For each map set, I compute and store the alm for each frequency, debeam them and rebeam them to a common resolution of 2.3 arcmin (as this is the resolution of the 95 GHz map). For the Healpix version I use these alm to compute the weights and the reconstructed maps. The mask correction is done through a simple fsky scaling. For the NaMaster version, I compute weights using spectra obtained with namaster, and use them to scale alm computed with Healpix. I use three different mask with Namaster: the apodized mask used for the Healpix analysis, a mask apodized using the internal NaMaster routine with 1 deg smoothing and the hit map. The mask used are shown at the bottom of this posting, in figure 4.

For each map I show the input and reconstructed ILC map, as well as the 95 GHz map. I also show the difference between these maps at fixed scale and free scal range. We can see that residuals are very faint for both the ILC and 95 GHz map. Looking at the "free scale" map, we can see that the pattern of the residual is quite different.

Figure 1:
Fig. 1 shows the input and reconstructed ILC map (and their difference) for the first three realizations. also shows the 95 GHz map for comparison.

I plot here the spectra of the ILC map and the original map, as well as the 95 GHz spectrum and the ILCxinput spectrum. The bottom panel shows the relative or absolute difference in \(D_{\ell}\) In the Healpix version, mask effects are corrected via a simple fsky scaling, while Namaster implements internally the mask correction using the Master algorithm. Also, NaMaster corrects for E->B leakage. All spectra are binned with bin_width=20. For TT spectra, the 95 GHz has slightly lower residuals. It would be interesting to see if this changes with the addition of extragalactic foregrounds. For polarization spectra, the ILC yields lower residuals.

Figure 2:
Fig. 2 shows the input and reconstructed ILC spectra for the first three realizations.

In Figure 3 I show the weights for the ILC in the cases presented above.

Figure 3:
Fig. 3 shows the weights for the reconstructed ILC. The first two rows present the weights for each frequency, and each plot has three different foreground models. In the last row all frequency are plotted together for each foreground model respectively. The legends in the last row refer to all plots in the last row.

In Figure 4 I show the apodized masks and the hit maps for the three sky configurations.

Figure 4:
Fig. 4 shows the three masks.
  1. Looking at map differences, using the hit maps seems to give highest residual. Also, the Pole wide and deep difference maps seem to have some large scale CMB signal.
  2. In TT spectra, performance of ILC and 95GHz map are comparable for Pole wide. For smaller fsky 95GHz has lower residual, while for larger fsky, ILC perfroms better.
  3. In polarization, ILC always has lower residuals then 95 GHz spectra at high-\(\ell\). The only issues appear at low fsky and low multipoles (hit map here seems to give better results)
  4. Polarization spectra obtained with the NaMaster 1 deg smooth mask are the worse.
  5. Looking at the weights, we see that the reconstruction is dominated by 95 and 145 GHz. The NaMaster 1 deg smooth mask seems to give smooother weights, which are in generl preferable for ILC reconstruction. However the spectra are the one with highest residuals (in polarization).