We provide a generic but physically clear discussion of the clustering properties of dark energy models. We explicitly show that in quintessence-type models the dark energy fluctuations, on scales smaller than the Hubble radius, are of the order of the perturbations to the Newtonian gravitational potential, hence necessarily small on cosmological scales. Moreover, comparable fluctuations are associated with different gauge choices. We also demonstrate that the often used homogeneous approximation is unrealistic, and that the so-called dark energy mutation is a trivial artifact of an effective, single fluid description. Finally, we discuss the particular case where the dark energy fluid is coupled to dark matter.

In this paper we show that a \emph{one-to-one} correspondence exists between any dark energy model and an equivalent (from a cosmological point of view, in the absence of perturbations) quartessence model in which dark matter and dark energy are described by a single perfect fluid. We further show that if the density fluctuations are small, the evolution of the sound speed squared, $c_s^2$, is fully coupled to the evolution of the scale factor and that the transition from the dark matter to the dark energy dominated epoch is faster (slower) than in a standard $\Lambda$CDM model if $c_s^2 > 0$ ($c_s^2 < 0$). In particular, we show that the mapping of the simplest quintessence scenario with constant $w_Q \equiv p_Q/ \rho_Q$ into a unified dark energy model requires $c_s^2 < 0$ at the present time (if $w_Q > -1$) contrasting to the Chaplygin gas scenario where one has $c_s^2 > 0$. However, we show that non-linear effects severely complicate the analysis, in particular rendering linear results invalid even on large cosmological scales. Although a detailed analysis of non-linear effects requires solving the full Einstein field equations, some general properties can be understood in simple terms. In particular, we find that in the context of Chaplygin gas models the transition from the dark matter to the dark energy dominated era may be anticipated with respect to linear expectations leading to a background evolution similar to that of standard $\Lambda$CDM models. On the other hand, in models with $c_s^2 > 0$ the expected transition from the decelerating to the accelerating phase may never happen.

We explicitly demonstrate that, contrary to recent claims, the dynamics of a generalized Chaplygin gas model with an equation of state $p=-C$ (where $C$ is a positive constant) is equivalent to that of a standard $\Lambda$CDM model to first order in the metric perturbations. We further argue that the analogy between the two models goes well beyond linear theory and conclude that they cannot be distinguished based on gravity alone.

We discuss the onset of the non-linear regime in the context of unified dark matter models involving a generalised Chaplygin gas. We show that the transition from dark matter-like to dark energy-like behaviour will never be smooth. In some regions of space the transition will never take place while in others it may happen sooner or later than naively expected. As a result the linear theory used in previous studies may break down late in the matter dominated era even on large cosmological scales. We study the importance of this effect showing that its magnitude depends on the exact form of the equation of state in the low density regime. We expect that our results will be relevant for other unified dark matter scenarios particularly those where the quartessence candidate is a perfect fluid.

We discuss the importance of including baryons in analyses of unified dark matter scenarios, focusing on toy models involving a generalized Chaplygin gas. We determine observational constraints on this unified dark matter scenario coming from large scale structure, type Ia Supernovae and CMB data showing how this component can bring about a different behaviour from classical $\Lambda$CDM and thus motivate further studies of this type of models. We also speculate on interesting new features which are likely to be important on non-linear scales in this context.

We discuss the issue of toy model building for the dark energy component of the universe. Specifically, we consider two generic toy models recently proposed as alternatives to quintessence models, known as Cardassian expansion and the Chaplygin gas. We show that the former is enteriely equivalent to a class of quintessence models. We determine the observational constraints on the latter, coming from recent supernovae results and from the shape of the matter power spectrum. As expected, these restrict the model to a behaviour that closely matches that of a standard cosmological constant $\Lambda$.