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.

We present an analysis of the constraining power of future measurements of the Integrated Sachs-Wolfe (ISW) effect on models of the equation of state of dark energy as a function of redshift, w(z). To achieve this, we employ a new parameterization of w, which utilizes the mean value of w(z) (<w>) as an explicit parameter. This helps to separate the information contained in the estimation of the distance to the last scattering surface (from the CMB) from the information contained in the ISW effect. We then use Fisher analysis to forecast the expected uncertainties in the measured parameters from future ISW observations for two models of dark energy with very different time evolution properties. For example, we demonstrate that the cross-correlation of Planck CMB data and LSST galaxy catalogs will provide competitive constraints on w(z), compared to a SNAP-like SNe project, for models of dark energy with a rapidly changing equation of state (e.g. 'Kink' models). Our work confirms that, while SNe measurements are more suitable for constraining variations in w(z) at low redshift, the ISW effect can provide important independent constraints on w(z) at high z.

We study the cosmological evolution of scalar fields that arise from a phase transition at some energy scale $\Lm_c$. We focus on negative power potentials given by $V=c\Lm_c^{4+n}\phi^{-n}$ and restrict the cosmological viable values of $\Lm_c$ and $n$. We make a complete analysis of $V$ and impose $SN1a$ conditions on the different cosmological parameters. The cosmological observations ruled out models where the scalar field has reached its attractor solution. For models where this is not the case, the analytic approximated solutions are not good enough to determine whether a specific model is phenomenologically viable or not and the full differential equations must be numerically solved. The results are not fine tuned since a change of 45% on the initial conditions does not spoil the final results. We also determine the values of $N_c, N_f$ that give a condensation scale $\Lm_c$ consistent with gauge coupling unification, leaving only four models that satisfy unification and SN1a constraints.

We show that a positive accelerating universe can be obtained simply by the dynamics of a non-abelian gauge group. It is the condensates of the chiral fields that obtain a negative power potential, below the condensation scale, and allow for a quintessence interpretation of these fields. The only free parameters in this model are $N_c$ and $N_f$ and the number of dynamically gauge singlet bilinear fields $\phi$ generated below the condensation scale. We show that it is possible to have unification of all coupling constants, including the standard and non standard model couplings, while having an acceptable phenomenology of $\phi$ as the cosmological constant. This is done without any fine tuning of the initial conditions. The problem of coincidence (why the universe has only recently started an accelerating period) is not solved but it is put at the same level as what the particle content of the standard model is.