The quadrupole power of cosmic microwave background (CMB) temperature anisotropies seen in the WMAP data is puzzlingly low. In this paper we demonstrate that Minimum Variance Optimization (MVO), a technique used by many authors (including the WMAP science team) to separate the CMB from contaminating foregrounds, has the effect of forcing the extracted CMB map to have zero statistical correlation with the foreground emission. Over an ensemble of universes the true CMB and foreground are indeed expected to be uncorrelated, but any particular sky pattern (such as the one we happen to observe) will generate non-zero measured correlations simply by chance. We call this effect cosmic covariance and it is a possible source of bias in the CMB maps cleaned using the MVO technique. We show that the presence of cosmic covariance is expected to artificially suppress the variance of the Internal Linear Combination (ILC) map obtained via MVO. It also propagates into the multipole expansion of the ILC map, generating a quadrupole deficit with more than 90% confidence. Since we do not know the CMB and the foregrounds a priori, there is therefore an unknown contribution to the uncertainty in the measured quadrupole power, over and above the usual cosmic variance contribution. Using the MVO on a series of Monte Carlo simulations that assume Gaussian CMB fluctuations, we estimate that the real quadrupole power of the CMB lies in the range [305.16,400.40] microKelvin^2 (at the 1-sigma level).
Download