The observational constraints on the flat \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\phi $$\end{document}ϕCDM models

Most dark energy models have the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varLambda $$\end{document}ΛCDM as their limit, and if future observations constrain our universe to be close to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varLambda $$\end{document}ΛCDM Bayesian arguments about the evidence and the fine-tuning will have to be employed to discriminate between the models. Assuming a baseline \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varLambda $$\end{document}ΛCDM model we investigate a number of quintessence and phantom dark energy models, and we study how they would perform when compared to observational data, such as the expansion rate, the angular distance, and the growth rate measurements, from the upcoming Dark Energy Spectroscopic Instrument (DESI) survey. We sample posterior likelihood surfaces of these dark energy models with Monte Carlo Markov Chains while using central values consistent with the Planck \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varLambda $$\end{document}ΛCDM universe and covariance matrices estimated with Fisher information matrix techniques. We find that for this setup the Bayes factor provides a substantial evidence in favor of the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varLambda $$\end{document}ΛCDM model over most of the alternatives. We also investigated how well the CPL parametrization approximates various scalar field dark energy models, and identified the location for each dark energy model in the CPL parameter space.


I. INTRODUCTION
It is well established that our universe is undergoing an accelerating expansion today, Refs.[1][2][3].Several observations suggest that this accelerated expansion started relatively recently at z ∼ 0.7, Refs.[4,5].One of the possible explanations is to assume the presence of dark energy as a dominant component of the total energy density budget in the universe today (i.e.around 70% of the universe matter-energy content today is a substance with negative pressure that drives today's accelerated expansion).Dark energy is characterized by an equation of state (EOS) parameter w defined by as a ratio between the pressure (p) and the energy density (ρ), w ≡ p/ρ.The accelerated expansion requires that w < −1/3.Generally speaking w parameter is time dependent.In the framework of the standard cosmological (concordance) model, dark energy is represented by the cosmological constant Λ (that was originally introduced by Albert Einstein, and it is assumed to be associated with the vacuum energy density).This cosmological model is referred as ΛCDM model, in this case the EOS parameter is constant, w = −1.The ΛCDM model is simple and easy to constrain through observations, but besides good agreements with existing observational data, the model has a number of shortcomings (the cosmological constant problem, the coincidence problem, the matter -anti-matter asymmetry, the weakness of gravity compared to other forces, etc.), Refs.[6][7][8][9].The most notable of these puzzles is the cosmological constant problem which stems from the fact that the theoretically expected value (based on quantum field theory approach and on dimensional arguments) of the cosmological constant associated energy density is determined by M 4 pl (where M pl is the Planck mass), while the actual value (suggested through observational data) is order of 120 magnitudes lower, Refs.[10][11][12].In order to overcome this (and other) difficulties (the coincidence problem, for example), dynamical dark energy models were proposed, Refs.[13,14], and see Ref. [15] for a recent review.
In this paper we investigate a representative family of dark energy models that are based on the idea of a cosmological scalar field, Refs.[16][17][18][19].If the scalar field φ has a slowly rolling stage, the energy density associated with this field can mimic the presence of the cosmological constant at late stages.There are many proposals for the functional form of the self-interacting potential of the scalar field that are allowed by the current observational data, Refs.[20][21][22][23][24][25][26][27][28][29][30][31][32][33].In this paper we consider two types models: the quintessence (dark energy from canonical scalar fields) and the phantom models (dark energy from non-canonical scalar fields).As of now, there is no consensus on which of these models is preferable based on the results obtained from the different observations, Refs.[34][35][36][37][38].We study the scalar field models with 10 quintessence and 7 phantom potentials in Bayesian framework, Refs.[39][40][41].We also limit ourselves by considering flat φCDM models, and we compare these models with a baseline ΛCDM model.We have found that under these assumptions a vast majority of the scalar field dark energy models will be characterized by low enough Bayes factors to suggest a substantial preference for the ΛCDM model.
Several large scale structure surveys missions, such as e.g.Dark Energy Spectroscopic Instrument (DESI), Wide-Field Infrared Survey Telescope (WFIRST) and Euclid are scheduled to start operating within the next decade.Once these missions are completed we will have very accurate measurements of the expansion rate, the angular distance, and the growth rate in the universe up to the redshifts of z ∼ 2.0, Refs.[42][43][44][45][46].These measurements cumulatively have a very strong constraining power on the behavior of both dark energy and gravity on large length scales.If the ΛCDM model is not the correct cosmological model, we will be able to see this in upcoming data.If, however, the ΛCDM model or a model very close to it, is the correct model, the interpretation of the data will be less straightforward.One reason for this is that the most viable dark energy models have the ΛCDM model as their limit so the Bayesian arguments about the fine-tuning of the extra parameters will have to be employed.In this work we refer to a simulated DESI data and study how these models would perform when compared to the baseline ΛCDM model.
The paper is organized as follows: in Sec.II we review the dark energy models (including the scalar field quintessence and phantom models); in Sec.III we describe observational tests, our results are presented in Sec.IV, and we conclude in Sec.V. We use natural units: c = = k B = 1 throughout the paper.

II. DARK ENERGY MODELS
In the scalar field models, dark energy is represented by the cosmological scalar field φ.We will consider two families of scalar field dark energy models: the quintessence (canonical) and the phantom scalar field (non-canonical) models.These models have opposite properties in their manifestation today: (i) in the range of the EOS parameter values (w < −1 for the phantom field and −1/3 < w < −1 for the quintessence field); (ii) in the sign of the kinetic term in the Lagrangian (the negative sign for the phantom field and the positive one for the quintessence field); (iii) in the dynamics of the scalar fields (the quintessence field rolls to the minimum its potential, the phantom field rolls to the "uphill" its potential); (iv) in the dynamics of the dark energy density (increases over time for the phantom field and almost doesn't change over time for the quintessence field); (v) in the forecast for the future evolution of the universe: for the phantom models violent future events (such as big/little/pseudo rips) are predicted, while in the quintessence models either an eternal expansion or a re-collapse depending on the spatial curvature of the universe is predicted.
The action associated with the scalar field φ is given field by, Refs.[47]: where "+" sign before the kinetic term (g µν ∂ µ φ∂ ν φ/2) refers to the quintessence models, while "−" stands for the phantom models; g µν is the background metric,1 and V (φ) is the self-interacting potential of the scalar field φ.The scalar field is assumed to exhibit the negligible spatial variations, so that the spatial derivatives are small compared to the time derivatives, and thus we assume the scalar field to be an homogeneous field.
Varying the action Eq. ( 1), the dynamical (or so called motion) equation for the scalar field (i.e. the Klein-Gordon equation), Refs.[48]: where again "+/−" sign corresponds to the quintessence/phantom model respectively, the over-dot denotes a derivative with respect to the physical time t.

III. TESTING DARK ENERGY POTENTIALS A. The basic equations
To see how well we will be able to discriminate between these dark energy scalar field potentials after upcoming dark energy surveys, we generate a set of the simulated data (theoretical model predictions) for the Hubble expansion rate, the angular distance, and the growth rate, in the redshift range of 0.15 < z < 1.85 (with z = 1/a − 1 is the redshift) expected from DESI mission, Ref. [42].The measurements are centered around their true values in our fiducial cosmology with the errorbars based on the Fisher matrix predictions.We fit this synthetic data by using the standard MCMC analysis method to estimate the multidimensional posterior likelihood of the model parameters.For all dark energy models we compute: i)The Hubble parameter H(z): The first Friedmann equation for the flat universe is, Ref. [9]: here E(z) = H(z)/H 0 is the normalized Hubble parameter, and H 0 is the Hubble parameter today; Ω i (z) ≡ ρ i (z)/ρ cr is the energy density parameter for "i"-th component (characterized by the energy density ρ i (z)). 4i)The angular diameter distance Assuming a flat universe, the angular diameter distance is given by, Ref. [47]: iii) The combination of the growth rate and the matter power spectrum amplitude f (a)σ 8 (a) The growth rate is defined as f (a) = dlnD(a)/dlna, where D(a) is the growth function defined through the ratio of overdensities δ(a) at different scale factors, as D(a) = δ(a)/δ(a 0 ), normalized to be unity today, (D(a 0 ) = 1), and it is a solution of the following linear perturbation equation, Ref. [51]: here a prime denotes a derivative with respect to the scale factor a ( ′ = d/da).The matter power spectrum amplitude can be characterized through the σ 8 (a) function, σ 8 (a) ≡ D(a)σ 8 , where σ 8 ≡ σ 8 (a 0 ) is the rms linear fluctuation in the mass distribution on scales 8h −1 Mpc (with h is the today Hubble constant in units of 100km/s/Mpc) today.We fix the value of σ 8 to its current best-fit ΛCDM value of σ 8 = 0.815 from the Plank 2015 data, Ref. [37], (see Ref. [52] for model-independent cosmological constraints on σ 8 from growth and expansion).
The EOS parameter of the dark energy models is often characterized by the Chevallier-Polarsky-Linder (CPL) w 0 − w a parametrization, Ref. [53,54]: where w 0 = w(a = 1) and w a = −a −2 (dw/da)| a=1 .This parametrization fits the EOS parameters for most of the dark energy models well enough for some effective values of w 0 and w a , but may fail to describe the arbitrary dark energy models to a good precision (few percents) over a wide redshift range. 5, In addition, the structure growth (in the most dark energy models) tends to be sensitive (only) to the fractional matter density Ω m (a) = Ω m /E 2 (a), with Ω m = Ω m,0 a −3 and as a consequence, the matter perturbation growth rate f (a) function with high accuracy can be parameterized as, Ref. [56]: where γ(a) is so called the growth index, and in general it is a time-dependent function.In the case of wCDM models (or any dark energy models which are the well approximated by the w 0 − w a parametrization), the γ(a) growth index scale factor dependence on can be determined from Eq. ( 9), see Ref. [57]: On the other hand, the function γ(a) can be parameterized by a scale factor independent manner, so called the Linder γ-parametrization, see Ref. [60]: This parametrization is accurate up to redshift of z = 5 (a = 0.2), Ref. [59].The numerical value of the γ itself depends on the dark energy model characteristics (w-parameter), being equal to 0.55 for the ΛCDM model, Ref. [60].

B. The definition of the starting points for the MCMC chains
To find the starting points for our MCMC chains, we solve jointly the scalar field φ equation for the quintessence/phantom models Eq. ( 2), the Friedmann equation Eq. ( 5) and the linear perturbation equation Eq. ( 7) for a wide range of the free parameters and the initial conditions for matter dominated epoch.For each potential we have found the plausible solutions, for which the following three criteria were simultaneously fulfilled: 1.The transition between the matter and dark energy equality (Ω m = Ω φ ) happens relatively recently z ∈ (0.6 − 0.8), Ref. [61].
2. The matter perturbation growth rate f (a) and the fractional matter density Ω m (a) are parameterized by the Linder γ-parametrization (Eq.( 11)).
3. The EOS parameter predicted by the different dark energy models should be in the agreement with the expected EOS parameter value today (for the phantom models w 0 < −1; for the quintessence models with −1 < w 0 < −0.75: for the freezing type w a < 0 and for the thawing type w a > 0).
For all potentials we found the range for (i) the allowed initial conditions and (ii) the model parameters, which we then used as the starting points for the MCMC chains.

A. The Bayesian statistics
All dark energy models considered in this work have the following free parameters, Ω m,0 and H 0 .In addition, the scalar field models have the extra parameters describing the strength and shape of the potential V (φ).These free parameters along with the prior ranges considered in our MCMC runs are presented in the Tables I and II.We have found these priors using the phenomenological method, which is described in the previous section, i.e. they correspond to the three conditions imposed on the solutions for each potential.
The reconstruction (and constraining) of the dark energy potentials is a challenging task, see Ref. [62] for more details.To assess the quality of the different models and to distinguish them from each other, we have applied the Akaike information criterion (AIC), Ref. [63] and the Bayesian (or Schwarz) information criterion (BIC), Ref. [64].The information obtained by these criteria complement each other.
AIC and BIC are defined respectively as, and where L max ∝ exp(−χ 2 min /2) is a maximum value of the likelihood function; N is a number of free parameters; k is a number of data points.
We also computed the evidence integral defined as, where p are all parameters of the model and the boundaries of the integral are given by the prior.We explored how tight the prior on the extra parameters needs to be for them to be competitive (in the sense of the Bayesian evidence) with the standard ΛCDM model.On Fig. (1) we show that some of the dark energy models stay very close to the ΛCDM for a wide range of parameter values within our priors.The range of the EOS parameters for the Ferreira-Joyce, the inverse exponent and the Sugra potentials is very small, it almost coincides with the ΛCDM model EOS parameter (w 0 = −1, w a = 0) consequently the likelihood of these model parameters is relatively flat and they can only be distinguished from ΛCDM model by Occam's razor type arguments.The Chang-Scherrer, the Urẽna-López-Matos, and the Barreiro potentials can result in up to 3σ offsets from ΛCDM for some parameter values; and the RP, the ZWS, the Albrecht-Skordis, and the Sahni-Wang potentials even extend beyond 3σ confidence level.This suggests that a significant fraction of the parameter space can be distinguished based on posterior likelihood.All phantom potentials on Fig.( 2), except the quadratic potential, exhibit a similar behaviour.The quadratic potential lies outside the 3σ contours of projected DESI constraints.This happens because in this model it is difficult to get a ΛCDM limit with a natural choice of parameter values and initial conditions.

V. CONCLUSIONS
We have derived projected constraints on a number of dark energy models by fitting them to a mock H(z), d A (z), f (a)σ 8 (z) data generated in a fiducial ΛCDM model.
On Figs. (3)(4) and (5) we show examples of the constraints that we obtain for the quintessence RP, the Sahni-Wang potentials and for the phantom inverse hyperbolic cosine potential.Since all models have the ΛCDM model as their limit, strictly speaking it is impossible to rule them out based on the likelihood arguments alone.Therefore we also used commonly cited model comparison criteria in the Bayesian statistics such as the Bayes factor, the AIC and w 0 BIC information criteria.Computing AIC and BIC in our setup is straightforward.Since all models have the same maximum likelihood by the construction the AIC and the BIC become simply functions of the number of the extra parameters.To compute the Bayes factors we integrated the posterior within the bounds given in the Tables I and II.
The results of the AIC, BIC, and Bayes factors for all the dark energy models are summarized in the Tables III and IV.These numbers clearly demonstrated that if the ΛCDM model is the true description of dark energy, the full DESI data will be able to strongly discriminate most scalar field dark energy models currently under consideration.These results however need to be taken with a grain of salt.The evidence values are very sensitive to the prior ranges.We only restricted the prior range based on constraints, by using the phenomenological method developed by us.Further restriction of the parameter ranges could significantly increase the evidence value.The results were derived assuming a fiducial ΛCDM model and the low value of evidence simply means that the model would be easier to discriminate if ΛCDM was the true model.The flip side of this is that if instead the dynamic dark energy models were true that would also show up more obviously in the data.We also explored how the dark energy models are mapped to the CPL parameter surface.For the models considered in our work this parametrization seems to work reasonably well even for the wide redshift range in a sense that the model predictions are always within one percent of of corresponding CPL predictions.

TABLE IV :
The list of the dark energy phantom potentials, with corresponding AIC, BIC, and Bayes factor values.