The initial conditions of the universe: how much isocurvature is allowed?

TL;DR

It is discovered that subtle flat directions in parameter space that tolerate large fractions of nonadiabatic fluctuations and larger values of the baryon density and a spectral tilt are allowed.

Abstract

We investigate the constraints imposed by the current data on correlated mixtures of adiabatic and non-adiabatic primordial perturbations. We discover subtle flat directions in parameter space that tolerate large (~60%) contributions of non-adiabatic fluctuations. In particular, larger values of the baryon density and a spectral tilt are allowed. The cancellations in the degenerate directions are explored and the role of priors elucidated.

Figures (4)

• Figure 1: Marginalized distributions for the 'cosmological' quantities, for CMB (dashed), CMB+LSS (solid). Results for pure adiabatic models with CMB+LSS are dot-dashed.
• Figure 2: Top four panels: marginalized distributions of relative power contributions $z_{ii}$ of the four auto-correlated modes: adiabatic (AD), cold dark matter density isocurvature (${\rm CI}$), neutrino velocity isocurvature (${\rm NIV}$) and neutrino density isocurvature (${\rm NID}$) for CMB (dashed) and CMB+LSS (solid). Bottom six panels: marginalized distributions for relative power contributions $z_{ij}$ of cross-correlated modes, for CMB (dashed) and CMB+LSS (solid).
• Figure 3: The CMB temperature spectrum (panel 1) and the temperature-polarization cross-correlation spectrum (panel 2) are plotted for the maximum likelihood adiabatic model (dashed line) with parameter values ($\omega_b$, $\omega_c$, $\Omega_\Lambda$, $n_s$, $\tau$, $\beta$)$=$(0.023, 0.12, 0.72, 0.97, 0.14, 0.49) and a high-likelihood mixed model (solid line) with parameter values (0.041, 0.13, 0.75, 1.06, 0.28, 0.37), with the CMB data overplotted. The galaxy power spectrum $P(k)$ for these models along with the pure mode contributions to the mixed model is shown in panel 3, including sizes of LSS error bars.
• Figure 4: Top panel: The total derivative along the degenerate direction (dot-dashed) described in the text is decomposed into variations in four parameters ($\omega_b$, $n_s$, AD, NIV), which cancel to within the experimental error bars (shaded). Bottom panel: (left) marginalized prior for the adiabatic mode contribution $z_{\hbox{\tiny AD}}$, no data; (centre) marginalized posterior for $z_{\hbox{\tiny AD}}$ with uniform prior on all models (dashed) and with uniform prior on $z_{\hbox{\tiny AD}}$ (solid); (right) marginalized posterior for $\omega_b$ with priors as in centre panel.