On the Coherence of WMAP and Planck Temperature Maps

TL;DR

This paper introduces Generalized Phases to compare WMAP and Planck CMB maps, demonstrating high coherence and providing insights into their Gaussianity, noise, and potential for non-Gaussianity detection.

Contribution

It presents a novel phase-based method for assessing map coherence, revealing strong agreement between WMAP and Planck data beyond power spectrum analysis.

Findings

High coherence between WMAP and Planck maps up to l<1100

WMAP power spectrum is about 2.6% higher than Planck's at certain scales

Phase coherence can be used to constrain non-Gaussianity and noise models

Abstract

The recent data release of ESA's Planck mission together with earlier WMAP releases provide the first opportunity to compare high resolution full sky Cosmic Microwave Background temperature anisotropy maps. To quantify the coherence of these maps beyond the power spectrum we introduce Generalized Phases, unit vectors in the (2l+1) dimensional representation spaces. For a Gaussian distribution, Generalized Phases are random and if there is non-Gaussianity, they represent most of the non-Gaussian information. The alignment of these unit vectors from two maps can be characterized by their angle, 0 deg expected for full coherence, and 90 deg for random vectors. We analyze maps from both missions with the same mask and Nside=512 resolution, and compare both power spectra and Generalized Phases. We find excellent agreement of the Generalize Phases of Planck Smica map with that of the WMAP Q,V,W maps, rejecting the null hypothesis of no correlations at 5 sigma for l's l<700, l<900 and l<1100, respectively, except perhaps for l<10. Using foreground reduced maps for WMAP increases the phase coherence. The observed coherence angles can be explained with a simple assumption of Gaussianity and a WMAP noise model neglecting Planck noise, except for low-intermediate l's there is a slight, but significant off-set, depending on WMAP band. On the same scales WMAP power spectrum is about 2.6% higher at a very high significance, while at higher l's there appears to be no significant bias. Using our theoretical tools, we predict the phase alignment of Planck with a hypothetical perfect noiseless CMB experiment, finding decoherence at l > 2900; below this value Planck can be used most efficiently to constrain non-Gaussianity.