Towards ending the partial sky E-B ambiguity in CMB observations

A crucial problem for partial sky analysis of CMB polarization is the $E$-$B$ leakage problem. Such leakage arises from the presence of `ambiguous modes that satisfy properties of both $E$ and $B$ modes. Solving this problem is critical for primordial polarization $B$ mode detection in partial sky CMB polarization experiments. In this work we introduce a new method for reducing the leakage. We demonstrate that if we complement the $E$-mode information outside the observation patch with ancillary data from full-sky CMB observations, we can reduce and even effectively remove the $E$-to-$B$ leakage. For this objective, we produce $E$-mode Stokes $QU$ maps from Wiener filtered full-sky intensity and polarization CMB observations. We use these maps to fill the sky region that is not observed by the ground-based experiment of interest, and thus complement the partial sky Stokes $QU$ maps. Since the $E$-mode information is now available on the full sky we see a significant reduction in the $E$-to-$B$ leakage. We evaluate on simulated data sets the performance of our method for a `shallow $f_text{sky}=8%$, and a `deep $f_text{sky}=2%$ northern hemisphere sky patch, with AliCPT-like properties, and a LSPE-like $f_text{sky}=30%$ sky patch, by combining those observations with Planck-like full sky polarization maps. We find that our method outperforms the standard and the pure-$B$ method pseudo-$C_ell$ estimators for all of our simulations. Our new method gives unbiased estimates of the $B$-mode power spectrum through-out the entire multipole range with near-optimal pseudo-$C_ell$ errors for $ell>20$. We also study the application of our method to the CMB-S4 experiment combined with LiteBIRD-like full sky data, and show that using signal-dominated full sky $E$-mode data we can eliminate the $E$-to-$B$ leakage problem.