Models for the latest stages of the cosmological evolution rely on a less solid theoretical and observational ground than the description of earlier stages like BBN and recombination. As suggested in a previous work by Vonlanthen et al., it is possible to tweak the analysis of CMB data in such way to avoid making assumptions on the late evolution, and obtain robust constraints on ''early cosmology parameters''. We extend this method in order to marginalise the results over CMB lensing contamination, and present updated results based on recent CMB data. Our constraints on the minimal early cosmology model are weaker than in a standard ΛCDM analysis, but do not conflict with this model. Besides, we obtain conservative bounds on the effective neutrino number and neutrino mass, showing no hints for extra relativistic degrees of freedom, and proving in a robust way that neutrinos experienced their non-relativistic transition after the time of photon decoupling. This analysis is also an occasion to describe the main features of the new parameter inference code MONTE PYTHON, that we release together with this paper. MONTE PYTHON is a user-friendly alternative to other public codes like COSMOMC, interfaced with the Boltzmann code CLASS.

We present forecasts for the accuracy of determining the parameters of a minimal cosmological model and the total neutrino mass based on combined mock data for a future Euclid-like galaxy survey and Planck. We consider two different galaxy surveys: a spectroscopic redshift survey and a cosmic shear survey. We make use of the Monte Carlo Markov Chains (MCMC) technique and assume two sets of theoretical errors. The first error is meant to account for uncertainties in the modelling of the effect of neutrinos on the non-linear galaxy power spectrum and we assume this error to be fully correlated in Fourier space. The second error is meant to parametrize the overall residual uncertainties in modelling the non-linear galaxy power spectrum at small scales, and is conservatively assumed to be uncorrelated and to increase with the ratio of a given scale to the scale of non-linearity. It hence increases with wavenumber and decreases with redshift. With these two assumptions for the errors and assuming further conservatively that the uncorrelated error rises above 2% at k = 0.4 h/Mpc and z = 0.5, we find that a future Euclid-like cosmic shear/galaxy survey achieves a 1-σ error on M ν close to 32 meV/25 meV, sufficient for detecting the total neutrino mass with good significance. If the residual uncorrelated errors indeed rises rapidly towards smaller scales in the non-linear regime as we have assumed here then the data on non-linear scales does not increase the sensitivity to the total neutrino mass. Assuming instead a ten times smaller theoretical error with the same scale dependence, the error on the total neutrino mass decreases moderately from σ(M ν ) = 18 meV to 14 meV when mildly non-linear scales with 0.1 h/Mpc < k < 0.6 h/Mpc are included in the analysis of the galaxy survey data.

Dark Matter is essential for structure formation in the late Universe so it must be stable on cosmological time scales. But how stable exactly? Only assuming decays into relativistic particles, we report an otherwise model independent bound on the lifetime of Dark Matter using current cosmological data. Since these decays affect only the low-multipoles of the CMB, the Dark Matter lifetime is expected to correlate with the tensor-to-scalar ratio r as well as curvature Ω k . We consider two models, including r and r + Ω k respectively, versus data from Planck, WMAP, WiggleZ and Baryon Acoustic Oscillations, with or without the BICEP2 data (if interpreted in terms of primordial gravitational waves). This results in a lower bound on the lifetime of CDM given by 160 Gyr (without BICEP2) or 200 Gyr (with BICEP2) at 95% confidence level.

The annihilation or decay of Dark Matter (DM) particles could affect the thermal history of the universe and leave an observable signature in Cosmic Microwave Background (CMB) anisotropies. We update constraints on the annihilation rate of DM particles in the smooth cosmological background, using WMAP7 and recent small-scale CMB data. With a systematic analysis based on the Press-Schechter formalism, we also show that DM annihilation in halos at small redshift may explain entirely the reionization patterns observed in the CMB, under reasonable assumptions concerning the concentration and formation redshift of halos. We find that a mixed reionization model based on DM annihilation in halos as well as star formation at a redshift z 6.5 could simultaneously account for CMB observations and satisfy constraints inferred from the Gunn-Peterson effect. However, these models tend to reheat the inter-galactic medium (IGM) well above observational bounds: by including a realistic prior on the IGM temperature at low redshift and allowing most of the reionization to be due to star formation, we find stronger cosmological bounds on the annihilation cross-section than with the CMB alone.

The role of Lorentz invariance as a fundamental symmetry of nature has been lately reconsidered in different approaches to quantum gravity. It is thus natural to study whether other puzzles of physics may be solved within these proposals. This may be the case for the cosmological constant problem. Indeed, it has been shown that breaking Lorentz invariance provides Lagrangians that can drive the current acceleration of the universe without experiencing large corrections from ultraviolet physics. In this work, we focus on the simplest model of this type, called ΘCDM, and study its cosmological implications in detail. At the background level, this model cannot be distinguished from ΛCDM. The differences appear at the level of perturbations. We show that in ΘCDM, the spectrum of CMB anisotropies and matter fluctuations may be affected by a rescaling of the gravitational constant in the Poisson equation, by the presence of extra contributions to the anisotropic stress, and finally by the existence of extra clustering degrees of freedom. To explore these modifications accurately, we modify the Boltzmann code class. We then use the parameter inference code Monte Python to confront ΘCDM with data from WMAP-7, SPT and WiggleZ. We obtain strong bounds on the parameters accounting for deviations from ΛCDM. In particular, we find that the discrepancy between the gravitational constants appearing in the Poisson and Friedmann equations is constrained at the level 1.8%.