1211.3358 (S. Khedekar et al.)

Bias from gas inhomogeneities in the pressure profiles as measured from X-ray and SZ observations    [PDF]

S. Khedekar, E. Churazov, A. Kravtsov, I. Zhuravleva, E. T. Lau, D. Nagai, R. Sunyaev

X-ray observations of galaxy clusters provide emission measure weighted spectra, arising from a range of density and temperature fluctuations in the intra-cluster medium (ICM). This is fitted to a single temperature plasma emission model to provide an estimate of the gas density and temperature, which are sensitive to the gas inhomogeneities. Therefore, X-ray observations yield a potentially biased estimate of the thermal gas pressure, P_X. At the same time Sunyaev-Zeldovich (SZ) observations directly measure the integrated gas pressure, P_SZ. If the X-ray pressure profiles are strongly biased with respect to to the SZ, then one has the possibility to probe the gas inhomogeneities, even at scales unresolved by the current generation of telescopes. At the same time, a weak bias has implications for the use of mass proxies like Y_SZ and Y_X as cosmological probes. In this paper we investigate the dependence of the bias, P_X(r)/P_SZ(r)-1, on the characteristics of fluctuations in the ICM taking into account the correlation between temperature and density fluctuations. We made a simple prediction of the irreducible bias in idealised X-ray vs SZ observations using multi-temperature plasma emission model. We also provide a simple fitting form to estimate the bias given the distribution of fluctuations. Analysing a sample of 16 simulated clusters extracted from hydrodynamical simulations, we find that the median value of bias is within +/-3% within R_500, it decreases to -5% at R_500 < r < 1.5R_500 and then rises back to ~0% at > 2R_500. The scatter of b_P(r) between individual relaxed clusters is small -- at the level of <0.03 within R_500, but turns significantly larger (0.25) and highly skewed at r > 1.5 R_500. Unrelaxed clusters display larger scatter (both from radius to radius and from cluster to cluster). Nevertheless, the bias remains within +/-20% within 0.8R_500 for all clusters.

View original: http://arxiv.org/abs/1211.3358