Detectability of CMB tensor B modes via delensing with weak lensing galaxy surveys

Abstract

We analyze the possibility of delensing CMB polarization maps using foreground weak lensing (WL) information. We build an estimator of the CMB lensing potential out of optimally combined projected potential estimators to different source redshift bins. Our estimator is most sensitive to the redshift depth of the WL survey, less so to the shape noise level. Estimators built using galaxy surveys like LSST and SNAP yield a 30-50% reduction in the lensing B-mode power. We illustrate the potential advantages of a 21-cm survey by considering a fiducial WL survey for which we take the redshift depth zmax and the effective angular concentration of sources n as free parameters. For a noise level of 1 muK arcmin in the polarization map itself, as projected for a CMBPol experiment, and a beam with FWHM=10 arcmin, we find that going to zmax=20 at n=100 gal/sqarcmin yields a delensing performance similar to that of a quadratic lensing potential estimator applied to small-scale CMB maps: the lensing B-mode contamination is reduced by almost an order of magnitude. In this case, there is also a reduction by a factor of ~4 in the detectability threshold of the tensor B-mode power. At this CMB noise level, there is little gain from sources with zmax>20. The delensing gains are lost if the CMB beam exceeds ~20 arcmin. The delensing efficiency and useful zmax depend acutely on the CMB map noise level, but beam sizes below 10 arcmin do not help. Delensing via foreground sources does not require arcminute-resolution CMB observations, a substantial practical advantage over the use of CMB observables for delensing.