SHARPENING THE PRECISION OF THE SUNYAEV-ZEL'DOVICH POWER SPECTRUM

Using both halo model calculations and a large sample of simulated SZ maps, we demonstrate that high-mass clusters add significant non-Gaussian variance to measurements of the SZ power spectrum amplitude. The difficulty in correctly accounting theoretically for the contribution of these objects to the uncertainty in C-l leads to a reduced sensitivity to sigma(8). We show that a simple solution is to mask out the brightest clusters in the map before measuring the power spectrum. We demonstrate that fairly conservative masking can reduce the variance and Gaussianize the statistics significantly, thus increasing the sensitivity to cosmological parameters. Choosing which objects to mask is nontrivial; we found that using a fixed sky density produced a well-defined and well-behaved estimate that can easily be applied to real maps. For example, masking the 10 (90) brightest clusters in a 100 deg(2) SZ map will improve the sensitivity to C-l by a factor of two at l = 1000 (2000) and 1.5 at l = 2000 (4000). We show that even in the presence of astrophysical foregrounds (primary cosmic microwave background and point sources) and instrument noise, one can increase the precision on measurements of sigma(8) by masking up to 0.9 clusters deg(-2).