Revisiting the WMAP - NVSS angular cross correlation. A skeptic's view

Abstract

In the context of the study of the ISW, we revisit the angular cross correlation of WMAP CMB data with the NVSS radio survey. We compute 2-point cross functions between the two surveys in real and in Fourier space, paying particular attention on the dependence of results on the flux of NVSS radio sources, the angular scales where correlations arise and the comparison with theoretical expectations. We reproduce previous results that claim an excess of correlation in the angular correlation function (ACF), and we also find some (low significance) similarity between the CMB and radio galaxy data in the multipole range $\el \in $ [10, 25]. However, the S/N in the ACFs increases with higher flux thresholds for NVSS sources, but drops a $\sim$ 30 - 50% in separations of the order of a pixel size, suggesting some residual point source contribution. When restricting our analyses to multipoles $\el \gt $60, we fail to find any evidence for cross correlation in the range $\el \in [2,10]$, where according to the model predictions and our simulations $\sim$ 50% of the S/N is supposed to arise. Also, the accumulated S/N for $\el \lt $60 is below 1, far from the theoretical expectation of S/N$\sim 5$. Part of this disagreement may be caused by an inaccurate modeling of the NVSS source population: as in previous works, we find a level of large scale ($\el \lt $70) clustering in the NVSS catalog that seems incompatible with a high redshift population. This is unlikely to be caused by contaminants or systematics, since it is independent of flux threshold, and hence present for the brightest ($\gt 30 \sigma$) NVSS sources. Either our NVSS catalogs are not probing the high redshift, large scale gravitational potential wells, or there is a clear mismatch between the ISW component present in WMAP data and theoretical expectations.