Comments from Anthony Challinor and Rupert Allison regarding the impact of Galactic Foregrounds on lensing
Below is a statement of the views from Anthony Challinor and Rupert Allison regarding the impact of Galactic Foregrounds on lensing. They led the recent CORE paper, which is at https://arxiv.org/abs/1707.02259.
This is not really a “quick question” but one that warrants further careful follow-up work. We (Rupert Allison and myself) made a start on this in the CORE paper, attempting to bound the impact of Galactic dust on observations at 150 GHz by focusing on about the worst 600 deg^2 patch that lies outside the Planck analysis mask. We have no quantitative results for synchrotron. Our feeling is that, at this stage, it would certainly be prudent to include the ability to subtract dust emission at high resolution while still achieving the target sensitivity in the cleaned map. The main reason for this is that we simply do not know enough about the non-Gaussian nature of dust emission at scales of a few arcmin, particularly in polarization, to be confident that it will not be an issue for lensing aiming at sub-percent-level measurements of the power (and other non-Gaussianity studies). Our results in the CORE paper should be treated only as being indicative of the magnitude of the problem — they are only as reliable as the simulated dust maps that we used. While these maps do capture some of the non-Gaussian structure in polarization, they are certainly missing small-scale effects from variations in the polarization directions.
To summarise what we found in the CORE paper for this bright-dust field assuming no dust subtraction:
(i) For the T-based lens reconstruction and power spectrum estimation, the dust emission leads to power spectrum biases at around the 15% level. Simply down-weighting dusty modes based on their expected *power* (i.e., including dust power when inverse-variance filtering the maps during the lens reconstruction) is ineffective in suppressing the bias. Alternative approaches, such as imposing l cuts by hand, may be more effective (further work is needed here).
(ii) For the EB reconstruction, which is expected to be the most important for S4 and which should be least troubled by extragalactic foregrounds, there is a very large bias if the dust power is not included in the inverse-variance weighting. However, including it does significantly down-weight the larger-scale modes and appears to be rather effective in reducing the bias to levels that are undetectable given the small number of CMB simulations that we employed (limiting our knowledge of any power spectrum bias to around the 2% level). The statistical error on the power spectrum amplitude is increased by around 50% over what one would have for the no-dust case.
Anthony and Rupert