Séminaire Univers /
Seminar Universe

Non-Gaussianity in cosmology refers to deviations from Gaussian statistics in cosmological observables and is commonly probed through the three-point correlator, the bispectrum. Primordial non-Gaussianity (PNG), i.e. the bispectrum of primordial fluctuations, provides a powerful probe of inflationary physics and a means of discriminating between competing inflationary models.

I will present how our analysis of the Planck Release 4 (PR4) data has yielded the most stringent constraints to date on scalar PNG. I will then discuss how future missions, such as LiteBIRD, will further enhance these measurements by accessing the tensor sector.

Next, I will introduce a new formalism for the Spectral Matching Independent Component Analysis (SMICA) method that incorporates higher-order statistical information from foregrounds. This framework enables direct bispectrum estimation from frequency-channel maps, allowing for the simultaneous recovery of the bispectra of multiple components. By shifting non-Gaussianity estimation from the cleaned-map stage to the component-separation step, this approach more effectively propagates and controls foreground-related uncertainties.

Finally, I will briefly introduce my current research on novel probes of non-Gaussianity and on studying correlations across different scales.