Non-Gaussianities in the local curvature of the 5-year WMAP data

TL;DR

This study re-examines non-Gaussianities in WMAP data using improved methods and finds that previous anomalies disappear when using the full covariance matrix, suggesting earlier detections may have been artifacts.

Contribution

It applies the full covariance matrix to local curvature statistics in WMAP data, clarifying the nature of non-Gaussianities and ruling out weak lensing as their cause.

Findings

Anomalies persist with 1-year data but are not confirmed with 5-year data.

Using the full covariance matrix removes the non-Gaussian signals.

Weak lensing does not account for the observed non-Gaussianities.

Abstract

Using the 5 year WMAP data, we re-investigate claims of non-Gaussianities and asymmetries detected in local curvature statistics of the 1 year WMAP data. In Hansen et al 2004, it was found that the northern ecliptic hemisphere was non-Gaussian at the ~1% level testing the densities of hill-, lake and saddle points based on the second derivatives of the CMB temperature map. The 5 year WMAP data has a much lower noise level and better control of systematics. Using these, we find that the anomalies are still present at a consistent level. Also the direction of maximum non-Gaussianity remains. Due to limited availability of computer resources, Hansen et al. 2004 were unable to calculate the full covariance matrix for the chi^2 test used. Here we apply the full covariance matrix instead of the diagonal approximation and find that the non-Gaussianities disappear and there is no preferred non-Gaussian direction. We compare with simulations of weak lensing to see if this may cause the observed non-Gaussianity when using diagonal covariance matrix. We conclude that weak lensing does not produce non-Gaussianity in the local curvature statistics at the scales investigated in this paper. The cause of the non-Gaussian detection in the case of a diagonal matrix remains unclear.