Reconstructing the CMB lensing on PLANCK

The gravitational weak lensing of the cosmic microwave background (CMB) by the line-of-sight matter provides a unique probe of the gravitational potential at large scales and intermediate redshifts. It induces non-gaussianity within the CMB anisotropy maps, which allows the reconstruction of the lensing potential through four-point function estimates. When this reconstruction is performed on PLANCK-like data, it yields a nice measurement of the lensing potential power spectrum, providing one deals with the various systematics among which the most prominent are the astrophysical foreground residuals -- including point sources -- and inhomogeneous noise. We propose two different methods of lensing extraction targeted to PLANCK-like temperature maps. The first one consists in applying a fullsky minimum-variance quadratic estimator on the CMB map after a foreground cleaning with a component separation algorithm and a data restoration of the most contaminated region of the sky using a sparse inpainting procedure. Extensive tests on PLANCK-like maps, has been proved this method to provide a robust lensing reconstruction showing a small residual bias which can be corrected by a Monte- Carlo afterwards. We developed a complementary approach devoted to apply on maps from an unique frequency channel without having recourse to any component separation, which requires to deal with severe spatial cuts of the maps. We show that the best strategy consists in adopting an hybrid approach: the largest scales are still reconstructed through a full-sky estimator on maps restored by inpainting whereas the small scales are better measured locally using a flat-sky estimator on local overlapping tangent planes extracted from the masked sky. This local method relies on original solutions to solve the issues related to the sphere-to-plane projection. After a brief review of the CMB lensing, I will detail these two proposed approaches, highlighting the adopted solutions to deal with the systematics and present how they perform when applied on realistic PLANCK-like simulations, then I will open a discussion on their respective merits