The luminosity-redshift relation in brane-worlds: II. Confrontation with experimental data

The luminosity distance - redshift relation for a wide class of generalized Randall-Sundrum type II brane-world models with Weyl fluid is compared to the presently available supernova data. We find that there is a class of spacially flat models with different amounts of matter $\Omega_{\rho}$ and Weyl fluid $\Omega_{d}$, which have a very similar fitting quality. The best-fit models are equally likely and can be regarded as extensions of the $\Lambda $CDM model, which is also included. We examine three models with different evolutionary history of the Weyl fluid, characterized by a parameter $\alpha =0,~2$ and$~3$. The first model describes a brane which had radiated energy into the bulk some time ago, but in recent times this energy exchange has ceased and only a dark radiation ($\alpha =0$) is left. In the other two models the Weyl-fluid describes a radiating brane throughout the cosmological evolution, up to our days. We find that the trought of the fitting surface extends over a wider $\Omega_{d}$-range with increasing $\alpha $, but the linear correlation of $\Omega_{d}$ and $\Omega_{\rho}$ holds all over the examined $\Omega_{d}$ range.


Introduction
Current observational data [1]- [3] suggest that the cosmological model of a Universe with only baryonic matter has to be modified. In the easiest way, the model can be reconciled with observations by the introduction of a cosmological constant Λ and of considerable amount of dark matter (ΛCDM model). Because the energy densities of both baryonic and dark matter decrease during cosmological evolution, the cosmological constant will dominate the late-time evolution. This process was first suggested for the explanation of Ia supernovae data, which suggest that our Universe has reached an accelerating phase. In a Λ-dominated universe, the luminosity distance increases faster with redshift than in the model without Λ [4], exactly as required by the supernova data.
Generally, the agreement with experiments can be achieved by introducing a dark energy component of the Universe, which replaces Λ. Such a dark energy in general does not clump. A recent analysis [5] shows that a dark energy model with varying dark energy density going through a transition from an accelerating to a decelerating subclass of the BRANE1 models with the effective energy density having a phantomlike behavior due to extra-dimensional effects, see Fig  The first comprehensive study of the generalized Randall-Sundrum type II (RS) brane-worlds tested against astronomical data was presented in Ref. [26]. Agreement with earlier supernova data has been established in the presence of a cosmological constant. In this analysis the dark radiation from the bulk was switched off (Ω d = 0) and the energy-momentum squared term was kept. Under these assumptions, for flat spatial sections and matter parameter Ω ρ = 0.3 the maximum likelihood method gave Ω λ = 0.004 ± 0.016 for the parameter characterizing the source term quadratic in the energy-momentum. This in turn implies a tiny value of the brane tension, which is disfavored by generic brane-world arguments. Moreover, much lower values for Ω λ emerge from both CMB and BBN.
Given the high limits for the values of λ, in any realistic model Ω λ can be safely ignored. This is a crucial difference of our forthcoming analysis as compared to the one presented in Refs. [27] and [26], where the corresponding cosmological parameter Ω λ was kept.
The next question is whether the source term arising from the Weyl curvature of the bulk may be kept, in other worlds, whether Ω d = 0. The Weyl curvature of the bulk gives an energy density ρ d = 6m/κ 2 a 4 , where κ 2 = 8πG is the gravitational coupling constant. In Ref. [32] it was shown that the BBN limits constrained the dark radiation component as −1.23 ≤ ρ d (z BBN ) /ρ γ (z BBN ) ≤ 0.11. Combining this with CMB constraints reduces this range to −0.41 ≤ ρ d (z BBN ) /ρ γ (z BBN ) ≤ 0.105. Here ρ γ is the energy density of the background photons. Another constraint for the value of the dark radiation at BBN was derived in [33] as −1 < ρ d (z BBN ) /ρ ν (z BBN ) < 0.5, where ρ ν is the energy density contributed by a single, two-component massless neutrino. This constraint was derived for high values of the 5-dimensional Plank mass.
In the simplest case the Weyl source term evolves as a radiation, thus its present value is obviously tiny. This is the reason why all mentioned references [18] and [26] comparing RS brane-worlds with observations disregard dark radiation. But is this a necessary assumption? Formulating the question the other way around: if we include even a small component of dark radiation into the late-time universe model we face a serious problem. Due to the fact that the energy density of dark radiation decreases as a −4 (compared to that of matter which is a −3 ), even an amount of dark radiation of the same order as the amount of baryonic matter nowadays implies dark radiation dominance in the past, for example during structure formation. This conclusion is contradicted by numerical simulations, which favorize cold dark matter as the dominant component of the Universe during structure formation [34].
However we can generalize the validity of the model by lifting the requirement of a constant mass m in the dark radiation energy density. A constant m implies a static Schwarzschild-anti de Sitter bulk and no energy exchange between the brane and the bulk. Therefore dark radiation is a manifestation of an equilibrium configuration with a static bulk, and it may be well possible that such a situation is reached only at the latest stages of the evolution of the brane-world Universe. Whenever m depends on a certain, non-zero power of a, the evolution of the energy density of the Weyl source term evolves in a non-standard way, allowing to escape from the argument of a small Weyl fluid left nowadays.
We propose here the LWRS (Lambda-Weyl fluid-Randall-Sundrum) model, a specific RS model with i) cosmological constant, ii) the brane radiating away energy during various stages of the cosmological evolution, characterized by the index α and iii) a Weyl fluid depending on the actiul value of α during the latest stage of cosmological evolution, which can be tested by supernova observations. For the inclusion of the LWRS model in the classification of brane-world models, see The LWRS model takes into account the possibility of an energy exchange between the brane and the bulk. This idea is not new. Indeed, it was already proposed that during an inflationary phase on the brane radiation is emitted and black holes thermally nucleate in the bulk [35]. Later on, but still in the high energy regime, the brane radiates such that the mass function of the bulk black hole increases with a 4 [36]. This means that the Weyl source term becomes a constant in this era. The brane continues to radiate away energy during structure formation [37], a process leading to a bulk black hole mass function m ∝ a α , with 1 ≤ α ≤ 4. (Other models with the brane radiating energy into the bulk are also known [38].) For α = 0 the Weyl fluid is known as dark radiation, for α = 2, 3 it gives the correct growth factor during structure formation. For α = 1, 4 it is indistinguishable from dark matter and a cosmological constant, respectively. Therefore pure dark radiation can emerge only in the low-z limit, while at earlier times a dynamic bulkbrane interaction governed by energy exchange should be present.

Confronting the models with the selected supernova data
As type Ia supernovae result from the explosion of white dwarf stars with identical mass, they show remarkable similarities. By employing well established calibration methods, one can calculate the maximal luminosity of the object (in the reference system of the explosion). This is done by analyzing the time-dependent variation of the emitted luminosity and the spectrum, a method known as the Multi-Color Light Curve analysis [39], [23]. In this process the observed parameters, the shape of the light curve and the spectral distribution of the emission have to be converted into the reference system of the host galaxy. For distant supernovae this translates to take into account the time dilation and the so-called K-correction [40]. While these methods depend on z, they are independent on the specific cosmological model. After performing these corrections, we have well-calibrated maximal luminosities for the supernovae of type Ia and in consequence they are considered as standard candles.
In 2003 a list of d L -z data pairs were published for 230 supernovae of type Ia [41], and 60 of them had low absoprtion (i.e. A V .1) and cosmological redshift (z > 0.01). To give result which can easily be compared to earlier works, we also involve this selected low-absorption data set for the most of the examinations. The basic supernova data we use here is the improved Gold set [42], which was released in 2006.
We confront with supernova observations several models from paper I. In Fig 1 we represent graphically on both logarithmic and linear scales their luminosity distanceredshift relations up to z = 2.5 . The plots are for k = 0 and Ω ρ = 0.27 (according to the combined analysis of the SDSS and WMAP 1-year data in Ref [2]). In particular, the luminosity distance -redshift relation is shown for the following models: . The latter models contain a brane which radiates energy at early times (for Ω d > 0) and during structure formation, such that a bulk black hole is formed and its mass increases continuously. As this process slows down, the Weyl curvature of the bulk induces the late-time dark radiation on the brane. • The ΛCDM model, given by Eqs. (60)-(61) of paper I (curve 2).
• The solutions with brane tension λ = 2Λ/κ 2 and no dark radiation (given by Eq. (52) of paper I) for both admissible values for this model, at Ω Λ = 0.704 (curve 4) and Ω Λ = 0.026 (curve 6). The former is similar to the class of models discussed in [26]. • The late-time universe Ω λ = 0 limit of the RS model with Randall-Sundrum finetuning, containing a huge amount of dark radiation Ω d = 0.73, given by Eq. (44) of paper I (curve 5).
In Fig. 1 we plot these models in a comparison to low-absorption supernova data from Ref. [41] (red triangles) together with the Gold set [23] (black dots). The error bars are indicated in the respective colors. The diagrams with linear scale are more instructive, as they emphasize the difference among the predictions of the chosen models and how they fit data, while the logarithmic scale better disseminate between the low z points.
The models represented by the curves 1, 3 and 4 by eye seem to compare as well with the supernova observations as the ΛCDM model (curve 2). By contrast, the models represented by the curves 5 and 6 seem to be not supported by observations. The model with no cosmological constant and significant dark radiation Ω d = 0.73, Ω λ = 0 (curve 6) and the model with Λ = κ 2 λ/2 and Ω Λ = 0.025 (curve 5) are significantly inconsistent with the observations, as they give χ 2 = 213 and 395, respectively ‡. All other models shown on Fig 1 are comparable with the supernova observations, as it was expected by a simple glance.
The χ 2 = 50 value found for the Λ = κ 2 λ/2-model with Ω Λ = 0.74 (curve 4) is slightly better than χ 2 found the ΛCDM model. However, as mentioned earlier, the tiny brane tension λ = 38.375 × 10 −60 TeV 4 , several order of magnitudes lower than all existing lower limits rules out this model as well.
The best fitting models are the models with brane cosmological constant; a high value of the brane tension (leading to Ω λ ≈ 0) and a small contribution of dark radiation, Ω d = ±0.05 (the curves 1 and 3). For Ω d = −0.05 we find χ 2 = 65, which is still acceptable. For Ω d = 0.05 we get χ 2 = 49. ‡ We mention here that we have also excluded several other models with Randall-Sundrum finetuning (not shown on Fig 1), which have either a very low value of the brane tension or a significant dark radiation. For example, the models with Ω d = 0.0258835, Ω λ = 0.70412 and Ω d = 0.70412, Ω λ = 0.0258835 gave χ 2 = 246 and 415, respectively. Selected low absorption supernovae from Ref. [41] are plotted with red, black dots represent the Gold set [23]. Both sets are represented with the corresponding error bars on the log-scaled diagrams. For the sake of perspicuity, the error bars of low absorption supernovae are not represented on the linearly scaled diagrams. The plotted models are the ΛCDM model (2); the brane models with cosmological constant and late-time dark radiation (1 and 3); without cosmological constant but with dark radiation (5); with cosmological constant satisfying Λ = κ 2 λ/2, thus low brane tension (4 and 6) and no dark radiation.
Values of Ω d between these limits are also admissible. It is likely that by increasing Ω d towards higher positive values, χ 2 remains compatible, however the accuracy of the perturbative solution is deteriorated with increasing Ω d , therefore higher orders in the expansion would be necessary to take into account.

The Gold2006 set of supernovae
More recently, Riess et al. [42] have published a new set of 182 gold supernovae, including new HST observations and recalibrations of the previous measurements. It is an interesting question how this recalibration influenced the above conclusions for the well-fitting models with dark radiation.
We applied the same tests to the Gold2006 data set as described in the previous section. First we assumed that Ω ρ = 0.27, as before, cf. Ref [2]. In this case the  [23]. The best fit is obtained for the brane-world model (3), with 5% dark radiation.
critical value of χ 2 is 197 at 80% level and 209 at 90% confidence level. Then the models represented by the curves 1-4 of Fig 2 behave as follows. The model with a small amount of negative dark radiation is disfavored at 80% confidence (χ 2 = 204). The models with λ = 2Λ/κ 2 and Ω Λ = 0.704 are ruled out at 90% confidence level, too (as χ 2 = 221). As expected from the previous analysis, the ΛCDM model (χ 2 = 192) and the LWRS model with Ω d = 0.05 (giving χ 2 = 194) compete closely. We also remark that varying Ω d between −0.03 and 0.07, the χ 2 remains under the critical value.
For gaining a deeper insight we have then calculated the predictions of the models between Ω d = −0.10 ÷ 0.10 with a stepsize of 0.01 in Ω d , with Ω ρ allowed to freely vary in the domain 0.15 ÷ 0.35 and z in the range 0 ÷ 3. Then we looked for the best fit of the Gold2006 set in the Ω d − Ω ρ space. This is represented on Fig 3. The global minimum of the surface is at Ω d = 0.040, Ω ρ = 0.225 (χ 2 = 190.52), which suggests an interesting opportunity for a Universe with less baryonic density and with dark radiation, compatible with the Gold2006 supernova data. The 1-σ confidence interval is centered about this value. The ΛCDM model (where Ω d is exactly 0) has the local minimum of Ω ρ = 0.275 (χ 2 = 195.8), but this is outside the 1-σ confidence interval. Similar conclusions emerge from the plot in the Ω Λ − Ω ρ plane, Fig 4. Here the global minimum of the surface is at Ω Λ = 0.735, Ω ρ = 0.225. The local minimum of the ΛCDM model is at Ω Λ = 0.725.
We note that there is a forbidden parameter range in both planes Ω d − Ω ρ and Ω Λ − Ω ρ , represented by white regions on Figs 3 and 4. This is because the Friedmann equation for these brane-world models combined with Ω Λ + Ω ρ + Ω d = 1 gives the constraints in the Ω d − Ω ρ plane and in the Ω Λ − Ω ρ plane. The forbidden region increases in both cases with z. If we would like to extend the limits to z → ∞, we obtain the limiting curves lim z→∞ Ω min d (z, Ω ρ ) = 0 in the Ω d − Ω ρ plane and lim z→∞ Ω max Λ (z, Ω ρ ) = 1 − Ω ρ in the Ω Λ − Ω ρ plane. However the LWRS model being valid only for low values of z, we represent on the graphs only the forbidden range for z = 3. The energy density of dark radiation decreases too fast during cosmological evolution to result in a considerable amount nowadays. We discuss this problem and its possible remedy in detail here. First we comment on the problem, then we show how an energy exchange between the brane and the bulk can leave a considerable amount of dark radiation. The constraint derived in [32] for the energy density of the dark radiation: where ρ γ (z BBN ) = βT 4 BBN is the energy density of the background photons at the beginning of BBN. The coefficient contains [11], [43] the effective number g * of relativistic degrees of freedom, which depends on the temperature. According to [44] g * = 10.75 at the beginning of BBN, when T BBN = 1.16 × 10 10 K. Thus ρ γ (z BBN ) = 7.37 × 10 25 J m −3 emerges, giving the constraint Note, that the domain of allowable negative values is larger than the one for positive values.
Thus the present value of Ω γ is which is quite negligible. If the Weyl source term were to evolve as radiation, its value would be even smaller, cf. Eq. (4). Indeed Eqs. (6) and (8) imply |Ω d | is of the same order of magnitude or smaller as Ω γ . However if the brane is radiating during structure formation, the mass parameter m becomes a function of the scale factor m ∝ a α , with 1 ≤ α ≤ 4 [37]. Then the energy density scales as a 4−α . Now let us suppose that the brane is in an equilibrium (non-radiating) configuration with α = 0 in the domain 0 ≤ z ≤ z 1 . In a preceding era z 1 < z ≤ z * the brane radiates such that α = 0, finally right after the beginning of BBN, at z * < z ≤ z BBN there is equilibrium once more (α = 0). Here z BBN = (T BBN /T 0 ) − 1 = 4.26 × 10 9 . According to this evolution Inserting this in Eq. (6) and employing Eq. (8) we obtain: In the particular case α = 0 we recover the constraint (10) set on pure dark radiation. However for any α > 0 we get Let us specify this result for the best fit value Ω d = 0.04. Depending on α we obtain the following numerical relations between the redshifts characterizing the switching on and off of the radiation leaving the brane: 1527. 80 + 1528. 80 z 1 , α = 1 38. 10 + 39. 10 z 1 , α = 2 10. 52 + 11. 52 z 1 , α = 3 5. 25 + 6. 25 It is evident that the value of z * increases with z 1 (this dependence becoming an approximate scaling for higher values of z 1 ) and decreases with α. The lower limit in the LWRS model is z 1 = 3. Then For the higher values of α the duration of the radiative brane regime necessary to produce a high value of Ω d today is quite short.
The luminosity-redshift relation in brane-worlds:II. Confrontation with experimental data 11

LWRS models with α = 2, 3 confronted with supernova data
In the absence of a known mechanism for changing α, we examine here the cases when α = 2 and α = 3 hold throughout the cosmological evolution, up to nowadays. For this we confront these models with the Gold 2006 set exactly as described before. To preserve the validity of the perturbative solution, the range of Ω d was selected to be −0.1-0.1, and we probed the range 0.15-0.35 of Ω ρ . The assumption for flatness was kept, too.
The results are qualitatively similar to the α = 0 case. The remarkable difference is that the peak of the minimum turned into a "trough", which lies aslope in the Ω ρ -Ω d space. This means that instead of a district solution, a complete model family exists in both cases, which can equally well explain the supernova data. The Ω ρ dependence of Ω d is less in the α = 2 model as compared to α = 0, and is very small if α = 3. The Ω d = 0 case is the ΛCDM model where these model intersect. The steeper slope of the minimum trough thus allows a much lower range of Ω ρ in the α = 2, and especially in the α = 3 models, with a value close to 0.3. On the other hand, the range of Ω d gets more and more wide with increasing α, which results in the conclusion that the presence of Ω d is mathematically plausible, and they have to be accounted for in RS cosmology.
Due to the higher slopes of the 1-σ and 2-σ contours in these α = 2, 3 models, Ω ρ is much less affected by the Weyl fluid, while Ω d can have various values in the detriment of Ω Λ . Therefore the Weyl fluid can explain some of the dark energy.

Conclusions
The luminosity distance given in paper I as function of redshift in terms of elementary functions and elliptical integrals of first and second type for various brane-world models with Weyl fluid was confronted with the available supernova data sets, including the Gold2006 data [42]. The tested models were: The luminosity-redshift relation in brane-worlds:II. Confrontation with experimental data 12 (A) The models with Randall-Sundrum fine-tuning, discussed in section 4 of paper I, with a considerable amount of dark radiation as a bulk effect, and a high value of the brane tension.
(B) The two models discussed in subsection 5.1 of paper I, which obey Λ = κ 2 λ/2, have no dark radiation and were integrable in terms of elementary functions.
(C) The LWRS models (subsection 5.2 of paper I), with a brane cosmological constant, for which the luminosity distance could be given analytically as function of redshift to first order accuracy in the dark radiation. (Due to its smallness, the source term Ω λ quadratic in the energy density was suppressed in the perturbative models of paper I.) The brane-world models (A) although interesting for historical reasons, do not comply with observations. Even if we introduce an extremely high amount of dark radiation Ω d = 0.73, tentatively replacing the cosmological constant in the energy balance Ω Λ + Ω ρ + Ω d + Ω λ = 1, these models are quickly outruled by supernova data (curve 5 of Fig 1). Dark radiation is not capable to replace the cosmological constant in producing a late-time acceleration, since it scales as usual radiation. The more we go back in the past, the higher becomes its domination over matter. Therefore a cosmological constant or dark energy is still needed in the generalized Randall-Sundrum type II models.
Our analysis has also dismissed immediately the model (B) with Ω Λ = 0.026. Surprisingly, the other toy model (B) with Ω Λ = 0.704 was in good agreement with the Gold2006 data, but ruled out by its low value of the brane tension, similarly as the models discussed in Ref. [26]. A low brane tension is in disagreement with various upper limits set by cosmological and astrophysical tests.
The perturbative approach of subsection 5.2 of paper I can be considered valid for a Weyl fluid with −0.1 < Ω d < 0.1. In this range the LWRS brane-world models (C) were confronted with supernova data and for α = 0 the dark radiation with significant negative energy density ruled out. The fact that a positive dark radiation (corresponding to a bulk black hole rather than to a bulk naked singularity) is favoured by the presently available best supernova data is in accordance with the early behavior of the RS model with late-time dark radiation, where the brane radiating away energy in early times leads to a black hole, which can further grow during structure formation.
The remaining LWRS brane-world models with α = 0 and Ω d between −0.03 and 0.07 (and Ω Λ changed accordingly) turned out to be excellent candidates for describing our universe, as they show remarkable agreement with the Gold2006 supernova data sets. If Ω ρ is allowed to vary in the range (0.15, 0.35), the preferred values are Ω d = 0.040, Ω ρ = 0.225, Ω Λ = 0.735.
The preferred cosmological parameters determined by comparing the LWRS model with α = 0 with supernova data alone are in perfect accordance with the WMAP 3-year data. Indeed according to Ref. [3] Ω ρ h 2 = 0.127 +0.007 −0.013 and h = 0.73 +0.03 −0.03 from which Ω ρ = 0.238 +0.035 −0.041 emerge. The value of Ω ρ determined by comparing the LWRS model with the supernova data alone is well in the middle of the domain allowed by the WMAP 3 year data.
We have then proved that the preferred value of Ω d = 0.04 is compatible with the known history of the Universe if the brane radiates away energy into the bulk during a relatively short period of the cosmological evolution. Such a process occurring between z = 24 and z = 3 could increase the amount of dark energy today with a factor of 10 3 as compared to the non-radiating brane, exactly as required by the LWRS model with α = 0.
The LWRS models with α = 1 (α = 4) are identical with the ΛCDM model with the only difference that some fraction of the dark matter (of the cosmological constant) has geometric origin.
Finally, the LWRS models with α = 2 and α = 3 do not present a sharp minimum, but rather an elongated trought shape in the parameter space, with the slope increasing with the value of α and Ω ρ ≈ 0.3. This means that in this class of models a wide range of values for Ω d (with a slight preference for negative values) and corresponding values for Ω Λ are fitting to the supernova data.
We must note that the reliability of these values is somehow deteriorated by the relatively small number of high-z supernova and by the inherent difficulties in the calibration of the available data. An obvious source of error is that data from the Gold2006 set is a combination of measurements taken on different instruments [45] and in fact it has been already signaled that the Gold2006 data set is not statistically homogeneous [46].
The conclusion of this paper is somewhat similar to that of Ref. [22]: the presently available supernova data are not enough to discern among several cosmological models. However the difference between the predictions of the acceptable models of our analysis (the ΛCDM model, the LWRS brane-world with α = 0 and Ω d = 0.04 and the models with Weyl fluid and α = 2, 3) are increasing with z. One may reasonably hope that the very far (z > 2) supernovae, which will be discovered for sure in the following decade, will improve their comparison.