All-sky component separation in the presence of anisotropic noise and dust temperature variations

We present an extension of the harmonic-space maximum-entropy component separation method (MEM) for multi-frequency CMB observations that allows one to perform the separation with more plausible assumptions about the receiver noise and foreground astrophysical components. Component separation is considered in the presence of spatially-varying noise variance and spectral properties of the foreground components. It is shown that, if not taken properly into account, the presence of spatially-varying foreground spectra, in particular, can severely reduce the accuracy of the component separation. Nevertheless, by extending the basic method to accommodate such behaviour and the presence of anisotropic noise, we find that the accuracy of the component separation can be improved to a level comparable with previous investigations in which these effects were not present.