Improved constraints on primordial non-Gaussianity for the Wilkinson Microwave Anisotropy Probe 5-yr data

TL;DR

This paper improves constraints on primordial non-Gaussianity using WMAP 5-year data and an advanced wavelet-based method, finding no significant non-Gaussian signal and ruling out previous claims of high fnl values.

Contribution

It introduces an enhanced wavelet-based analysis with all third order statistics across multiple scales, providing tighter constraints on fnl from WMAP data.

Findings

Constrained fnl to -18 < fnl < 80 at 95% CL.

Excluded the previously reported fnl=87 at ~99% CL.

Found no significant North-South asymmetry in fnl.

Abstract

We present new constraints on the non-linear coupling parameter fnl with the Wilkinson Microwave Anisotropy Probe (WMAP) data. We use an updated method based on the spherical Mexican hat wavelet (SMHW) which provides improved constraints on the fnl parameter. This paper is a continuation of a previous work by Curto et al. where several third order statistics based on the SMHW were considered. In this paper, we use all the possible third order statistics computed from the wavelet coefficient maps evaluated at 12 angular scales. The scales are logarithmically distributed from 6.9 arcmin to 500 arcmin. Our analysis indicates that fnl is constrained to -18 < fnl < +80 at 95% confidence level (CL) for the combined V+W WMAP map. This value has been corrected by the presence of undetected point sources, which adds a positive contribution of Delta_fnl = 6 +- 5. Our result excludes at ~99% CL the best-fitting value fnl=87 reported by Yadav & Wandelt. We have also constrained fnl for the Q, V and W frequency bands separately, finding compatibility with zero at 95 % CL for the Q and V bands but not for the W band. We have performed some further tests to understand the cause of this deviation which indicate that systematics associated to the W radiometers could be responsible for this result. Finally we have performed a Galactic North-South analysis for fnl. We have not found any asymmetry, i.e. the best-fitting fnl for the northern pixels is compatible with the best-fitting fnl for the southern pixels.