The properties of magnetized plasmas are always investigated under the hypothesis that the relativistic inhomogeneities stemming from the fluid sources and from the geometry itself are sufficiently small to allow for a perturbative description prior to photon decoupling. The latter assumption is hereby relaxed and predecoupling plasmas are described within a suitable expansion where the inhomogeneities are treated to a given order in the spatial gradients. It is argued that the (general relativistic) gradient expansion shares the same features of the drift approximation, customarily employed in the description of cold plasmas, so that the two schemes are physically complementary in the large-scale limit and for the low-frequency branch of the spectrum of plasma modes. The two-fluid description, as well as the magnetohydrodynamical reduction, is derived and studied in the presence of the spatial gradients of the geometry. Various solutions of the coupled system of evolution equations in the anti-Newtonian regime and in the quasi-isotropic approximation are presented. The relation of this analysis to the socalled separate universe paradigm is outlined. The evolution of the magnetized curvature perturbations in the nonlinear regime is addressed for the magnetized adiabatic mode in the plasma frame.
I. MOTIVATIONSThe analyses of the large-scale galaxy distribution [1,2], of the high-redshift type Ia supernovae [3,4], and of the CMB observables [5,6] seem to converge, these days, on a concordance model sometimes called the ÃCDM scenario, where à stands for the dark energy component and CDM accounts for the dark matter component. The ÃCDM scenario is just the compromise between the number of ascertainable parameters and the quality of the observational data. The quest for a concordance lore is also able to shed some light on the presence of large-scale magnetic fields in nearly all gravitationally bound systems we observe. Since we do see magnetic fields today over large distance scales, it seems natural to scrutinize their impact on the CMB observables. This is the motivation of a program aimed at bringing the unconventional study of magnetized CMB anisotropies to the same standard of the more conventional adiabatic 1 paradigm (see [7][8][9][10] and references therein). While different approaches to the problem are certainly available [11][12][13][14][15] (see [16] for a more complete list of earlier references), the path followed in [7,8] led to the calculation of the temperature and polarization anisotropies induced by the magnetized (adiabatic and entropic) initial conditions. The parameters of the magnetized background have been estimated (for the first time) in [9,10] by using the TT and TE correlations 2 measured by the WMAP Collaboration. The obtained results 3 show that large-scale (comoving) magnetic fields larger than 3.5 nG are excluded to 95% C.L. and for magnetic spectral indices n B ¼ 1:6 0:8 À0:1 . These determinations have been conducted in the context of the minimal mÃCDM, where m stands for magneti...