Error analysis of quadratic power spectrum estimates for CMB polarization: sampling covariance

Abstract

Quadratic methods with heuristic weighting (e.g. pseudo-C_l or correlation function methods) represent an efficient way to estimate power spectra of the cosmic microwave background (CMB) anisotropies and their polarization. We construct the sample covariance properties of such estimators for CMB polarization, and develop semi-analytic techniques to approximate the pseudo-C_l sample covariance matrices at high Legendre multipoles, taking account of the geometric effects of mode coupling and the mixing between the electric (E) and magnetic (B) polarization that arise for observations covering only part of the sky. The E-B mixing ultimately limits the applicability of heuristically-weighted quadratic methods to searches for the gravitational-wave signal in the large-angle B-mode polarization, even for methods that can recover (exactly) unbiased estimates of the B-mode power. We show that for surveys covering one or two per cent of the sky, the contribution of E-mode power to the covariance of the recovered B-mode power spectrum typically limits the tensor-to-scalar ratio that can be probed with such methods to around 0.05.