Optimized supernova constraints on dark energy evolution

A model-independent method to study the possible evolution of dark energy is presented. Optimal estimates of the dark energy equation of state w are obtained from current supernovae data from Riess et al. (2004) following a principal components approach. We assess the impact of varying the number of piecewise constant estimates of w using a model selection method, the Bayesian information criterion, and compare the most favored models with some parametrizations commonly used in the literature. Although data seem to prefer a cosmological constant, some models are only moderately disfavored by our selection criterion: a constant w, w linear in the scale factor, w linear in redshift and the two-parameter models introduced here. Among these, the models we find by optimization are slightly preferred. However, current data do not allow us to draw a conclusion on the possible evolution of dark energy. Interestingly, the best fits for all varying-w models exhibit a w<-1 at low redshifts.