The performance of spherical wavelets to detect non-Gaussianity in the CMB sky

Abstract

We investigate the performance of spherical wavelets in discriminating between standard inflationary models (Gaussian) and non-Gaussian models. For the later we consider small perturbations of the Gaussian model in which an artificially specified skewness or kurtosis is introduced through the Edgeworth expansion. By combining all the information present in all the wavelet scales with the Fisher discriminant, we find that the spherical Mexican Hat wavelets are clearly superior to the spherical Haar wavelets. The former can detect levels of the skewness and kurtosis of ~1% for 33' resolution, an order of magnitude smaller than the later. Also, as expected, both wavelets are better for discriminating between the models than the direct consideration of moments of the temperature maps. The introduction of instrumental white noise in the maps, S/N=1, does not change the main results of this paper.