ABSTRACT
The Internal Linear Combination (ILC) method has been extensively used to extract the cosmic microwave background (CMB) anisotropy map from foreground contaminated multifrequency maps. However, the performance of simple ILC is limited and can be significantly improved by heavily constraint equations, dubbed constrained ILC (cILC). The standard ILC and cILC work on spin-0 fields. Recently, a generalised version of ILC has been developed, named polarization ILC (PILC), in which Q ± iU at multiple frequencies are combined using complex coefficients to estimate Stokes Q and U maps. A statistical moment expansion method has recently been developed for high-precision modelling of the galactic foregrounds. This paper develops a semiblind component separation method combining the moment approach of foreground modelling with a generalised version of the PILC method for heavily constraint equations. The algorithm is developed in pixel space over a spin-2 field. We demonstrate the performance of the method on three sets of absolutely calibrated simulated maps at WMAP and Planck frequencies with varying foreground models. We apply this component separation technique in simultaneous estimation of Stokes Q and U maps of the thermal dust at 353 GHz and synchrotron at 30 GHz. We also recover both dust and synchrotron maps at 100 and 143 GHz, where separating two components is challenging.