Effect of component separation on the temperature distribution of the cosmic microwave background

We present a study of the effect of component separation on the recovered cosmic microwave background (CMB) temperature distribution. First, we extract the CMB component from simulated multifrequency Planck data (in small patches of the sky) using the maximum-entropy method (MEM), Wiener filter (WF) and a linear combination of the frequency channels (LCFC). We then apply a wavelet-based method to study the Gaussianity of the recovered CMB and compare it with the same analysis for the input map. When the original CMB map is Gaussian (and assuming that point sources have been removed), we find that neither MEM nor WF introduce non-Gaussianity in the CMB reconstruction. Regarding the LCFC, the Gaussian character is also preserved provided that the appropriate combination of frequency channels is used. On the contrary, if the input CMB map is non-Gaussian, all the studied methods produce a reconstructed CMB with lower detections of non-Gaussianity than the original map. In this case, MEM produces in general the highest non-Gaussian detections in the recovered map and, therefore, provides the best reconstruction from the considered methods to search for non-Gaussianity. We have also studied the effect of point sources in the MEM reconstruction. If no attempt of removing point sources is performed, they clearly contaminate the CMB reconstruction, introducing spurious non-Gaussianity. However, if the brightest point sources are removed from the data using the Mexican Hat Wavelet, the Gaussian character of the CMB is preserved in the reconstruction.