Observational constraints on low redshift evolution of dark energy: how consistent are different observations?

Abstract

The dark energy component of the Universe is often interpreted either in terms of a cosmological constant or as a scalar field. A generic feature of the scalar field models is that the equation of state parameter w≡P/ρ for the dark energy need not satisfy w=-1 and, in general, it can be a function of time. Using the Markov chain Monte Carlo method we perform a critical analysis of the cosmological parameter space, allowing for a varying w. We use constraints on w(z) from the observations of high redshift supernovae (SN), the Wilkinson Microwave Anisotropy Probe (WMAP) observations of cosmic microwave background (CMB) anisotropies, and abundance of rich clusters of galaxies. For models with a constant w, the ΛCDM(cold dark matter) model is allowed with a probability of about 6% by the SN observations while it is allowed with a probability of 98.9% by WMAP observations. The ΛCDM model is allowed even within the context of models with variable w: WMAP observations allow it with a probability of 99.1% whereas SN data allows it with 23% probability. The SN data, on its own, favors phantom-like equation of state (w<-1) and high values for Ω_NR. It does not distinguish between constant w (with w<-1) models and those with varying w(z) in a statistically significant manner. The SN data allows a very wide range for variation of dark energy density, e.g., a variation by factor ten in the dark energy density between z=0 and z=1 is allowed at 95% confidence level. WMAP observations provide a better constraint and the corresponding allowed variation is less than a factor of 3. Allowing for variation in w has an impact on the values for other cosmological parameters in that the allowed range often becomes larger. There is significant tension between SN and WMAP observations; the best fit model for one is often ruled out by the other at a very high confidence limit. Hence results based on only one of these can lead to unreliable conclusions. Given the divergence in models favored by individual observations, and the fact that the best fit models are ruled out in the combined analysis, there is a distinct possibility of the existence of systematic errors which are not understood.