Detection and discrimination of cosmological non-Gaussian signatures by multi-scale methods^*

¹ DAPNIA/SEDI-SAP, Service d'Astrophysique, CEA-Saclay, 91191 Gif-sur-Yvette Cedex, France
² IAS-CNRS, Université Paris Sud, Bâtiment 121, 91405, Orsay Cedex, France

Corresponding author: J.-L. Starck, jstarck@cea.fr

Received: 19 February 2003
Accepted: 6 November 2003

Abstract

Recent Cosmic Microwave Background (CMB) observations indicate that the temperature anisotropies arise from quantum fluctuations in the inflationary scenario. In the simplest inflationary models, the distribution of CMB temperature fluctuations should be Gaussian. However, non-Gaussian signatures can be present. They might have different origins and thus different statistical and morphological characteristics.
In this context and motivated by recent and future CMB experiments, we search for, and discriminate between, different non-Gaussian signatures. We analyse simulated maps of three cosmological sources of temperature anisotropies: Gaussian distributed CMB anisotropies from inflation, temperature fluctuations from cosmic strings and anisotropies due to the kinetic Sunyaev-Zel'dovich (SZ) effect both showing a non-Gaussian character. We use different multi-scale methods, namely, wavelet, ridgelet and curvelet transforms. 
The sensitivity and the discriminating power of the methods is evaluated using simulated data sets. 
We find that the bi-orthogonal wavelet transform is the most powerful for the detection of non-Gaussian signatures and that the curvelet and ridgelet transforms characterise quite precisely and exclusively the cosmic strings. They allow us thus to detect them in a mixture of CMB + SZ + cosmic strings. We show that not one method only should be applied to understand non-Gaussianity but rather a set of different robust and complementary methods should be used.