Residual foreground contamination in the WMAP data and bias in non-Gaussianity estimation

TL;DR

This study investigates residual foreground contamination in WMAP 7-year data and demonstrates its significant impact on non-Gaussianity estimates, highlighting the importance of proper foreground removal for accurate CMB analysis.

Contribution

The paper provides a detailed analysis of residual foreground contamination in WMAP data and quantifies its effect on non-Gaussianity estimations using Minkowski Functionals.

Findings

Positive correlations indicate residual Galactic and point source contamination.

Masking extended point sources reduces contamination levels.

Residual foregrounds significantly influence non-Gaussianity measurements.

Abstract

We analyze whether there is any residual foreground contamination in the cleaned WMAP 7 years data for the differential assemblies, Q, V and W. We calculate the correlation between the foreground map, from which long wavelength correlations have been subtracted, and the foreground reduced map for each differential assembly after applying the Galaxy and point sources masks. We find positive correlations for all the differential assemblies, with high statistical significance. For Q and V, we find that a large fraction of the contamination comes from pixels where the foreground maps have positive values larger than three times the rms values. These findings imply the presence of residual contamination from Galactic emissions and unresolved point sources. We redo the analysis after masking the extended point sources cataloque of Scodeller et al. [7] and find a drop in the correlation and corresponding significance values. To quantify the effect of the residual contamination on the search for primordial non-Gaussianity in the CMB we add estimated contaminant fraction to simulated Gaussian CMB maps and calculate the characteristic non-Gaussian deviation shapes of Minkowski Functionals that arise due to the contamination. We find remarkable agreement of these deviation shapes with those measured from WMAP data, which imply that a major fraction of the observed non-Gaussian deviation comes from residual foreground contamination. We also compute non-Gaussian deviations of Minkowski Functionals after applying the point sources mask of Scodeller et al. and find a decrease in the overall amplitudes of the deviations which is consistent with a decrease in the level of contamination.