On the Impact of Random Residual Calibration Error on the Gibbs ILC Estimates over Large Angular Scales

Abstract

Residual error in calibration coefficients corresponding to observed CMB maps is an important issue while estimating a pure CMB signal. A component separation method, if these errors in the input foreground contaminated CMB maps are not properly taken into account, may lead to bias in the cleaned CMB map and estimated CMB angular power spectrum. But the inability to exactly determine the calibration coefficients corresponding to each observed CMB map from any CMB experiment makes it very difficult to incorporate their exact and actual values in a component separation analysis. Hence the effect of any random and residual calibration error on the cleaned CMB map and its angular power spectrum of a component separation problem can only be understood by performing detailed Monte Carlo simulations. In this paper, we investigate the impact of using input foreground contaminated CMB maps with random calibration errors on posterior density of cleaned CMB map and theoretical CMB angular power spectrum over large angular scales of the sky following the Gibbs ILC method proposed by \cite{Sudevan:2018qyj}. By performing detailed Monte Carlo simulations of WMAP and Planck temperature anisotropy observations with calibration errors compatible with them we show that the best-fit map corresponding to posterior maximum is minimally biased in Gibbs ILC method by a CMB normalization bias and residual foreground bias. The bias in best-fit CMB angular power spectrum with respect to the case where no calibration error is present are $\sim 28 μK^2$ and $-4.7 μK^2$ respectively between $2 \le \ell \le 15$ and $16 \le \ell \le 32$. The calibration error induced error in best-fit power spectrum causes an overall $6\%$ increase of the net error when added in quadrature with the cosmic variance induced error.