Correlation of iteratively reconstructed lensing templates to ideal for 06b III
Oct 13 14 2020, Clem Pryke
This post follows from Correlation of iteratively reconstructed lensing templates to ideal for 06b II.
Julien Carron provided additional iterative lensing template reconstructions from the 06b.07 and 06b.09 95GHz "comb" maps. The 07 foreground model is the amplitude modulated Gaussian dust+sync introduced in Amplitude modulated Gaussian dust sims. The 09 foreground model is the Vansyngel introduced in Vansyngel Model and based on arxiv/1611.02577.
The plot below shows that the lensing template reconstructions from the maps containing foreground correlate with the ideal lensing template slightly less well. The right plot is a zoom in of the left plot. It is the ell=80 region that we care about. We see that the correlation drops more for the amplitude modulated Gaussian model (07) than for the Vansyngel model (09). This will likely make a noticeable difference to sigma(r) - Caterina already reported results for no foreground and model 00 in [20200928_ML_dc06] and model 00 was slightly worse.
The plot below shows the differences of the reconstruction from maps containing foreground versus those without for the first realization. These are all Q maps. The color range for the left column is +/-0.1uK and in the right column +/-0.3uK. This plot is really interesting - we can see that where the foreground is bright the lensing reconstruction is degraded. I think this is also true for the Vansyngel model but the effect is a bit obscured by blobs which contain only large scale structure which I think are artifacts from the ell<30 cutoff I am imposing to simulate the fact that small patch observations are not sensitive to low ell. I am intrigued by the region at the left and right edges in the model 00 case - somehow the introduction of uniform amplitude foreground causes degradation in higher noise regions?
Next step is to see if we can do better using the ILC cleaned maps Caterina has been working on.