Cubic anomalies in WMAP

TL;DR

This paper analyzes WMAP data's bispectrum, revealing potential non-Gaussian features and asymmetries at specific scales and directions, challenging the assumption of Gaussianity in the early universe.

Contribution

It provides a detailed frequentist bispectrum analysis of WMAP data, identifying anomalies and asymmetries not previously confirmed, and compares these with known large-scale anomalies.

Findings

Low chi-squared region in bispectrum for l=32-62 suggests non-Gaussianity.

Detected North-South asymmetry at 99% confidence level.

No connection found with previously claimed quadrupole anomalies.

Abstract

We perform a frequentist analysis of the bispectrum of WMAP first year data. We find clear signal domination up to l=200, with overall consistency with Gaussianity except for the following features. There is a flat patch (i.e. a low chi-squared region) in the same-l components of the bispectrum spanning the range l=32-62; this may be interpreted as ruling out Gaussianity at the 99.6% confidence level. There is also an asymmetry between the North and South inter-l bispectrum components at the 99% confidence level. The preferred asymmetry axis correlates well with the (l,b)=(57,10) direction quoted in the literature for asymmetries in the power spectrum and three-point correlation function. However our analysis of the quadrupole (its bispectrum and principal axes) fail to make contact with previously claimed anomalies.