BK-style power spectra of 06b with iterative lensing templates

— C. Pryke

2020 Sept 25: Pager plots for "ideal LT" updated to fix previously broken bpwfs.

2020 Sept 28: Pager plots below updated to empirical supfac on all spectra, rather than just on LT. This makes almost no visible difference here but fixes the bias problems seen in ML searches here.

Summary

Spectra of the 06b sims with iterative lensing templates are finally ready for BK-style reanalysis. A couple of problems have been found and fixed. The agreement between the bandpower window function (BPWF) derived model expectation values and the mean of the simulations is much improved and no biases are to be expected in recovered model parameters (including $$r$$). A method has been found to include the provided iterative lensing templates in the BK-style analysis which is likely adequate. However, it might be possible to improve the template production algorithm from the point of view of this style of reanalysis.

This posting is similar to 20181207_bkfinal_02lt. Analysis of those 02 sims with quadratic estimator derived lensing templates was presented in 20190120_dc02_ML_LT and showed around $$A_L=0.5$$ equivalent - far from what we are looking for.

In 20191016_dc06_dsr we defined the 06 sim set with direct scaling from BK15 achieved sensitivity to DSR detector-years and sky coverage patterns. In 20191101_dc06_specsmaps some check plots etc were presented. In this April 10 2020 posting Julien Carron presented new technology iterative lensing reconstructions. It should be noted that Julien has run his iterative reconstruction on the high resolution 95GHz lensed-LCDM+noise only. i.e. He did not include any foreground.

The BK auto/cross spectrum analysis method requires to know the noise bias of the auto spectra, and the degree of filtering which has been applied to the maps (aka the "filter/beam" suppression factor). For the quadratic estimator templates we ran the algorithm on the signal only and signal+noise realizations and took the map difference as the noise realization - and this worked OK. For the iterative method Julien reports that trying to run on signal only does not work. So instead I have used the method described in 20200708_itrecon_noise to infer signal and noise maps. There are outstanding issues but I think things should work for the present.

This posting is after a tweak to the 06b apodization mask to improve E/B purity as described in this post.