Squeezing the window on isocurvature modes with the Lyman-alpha forest

TL;DR

This work strengthens bounds on cold dark matter isocurvature perturbations by incorporating high-resolution Ly-α forest data, which probes small-scale matter fluctuations at redshifts z~2–3. Using a Bayesian framework and hydrodynamical calibration of the flux-to-matter power relation, the authors constrain the isocurvature fraction to α<0.4 (95% CL) and find an isocurvature tilt n_iso ≈ 1.9±1.0, with a tendency toward uncorrelated modes. The analysis also evaluates curvaton and double-inflation scenarios, finding a highly constrained curvaton case (f_iso<0.05 at 95% CL) requiring near-total curvaton domination at decay, and a stringent bound R<3 for the double inflation model. Overall, Ly-α data reduce the allowed parameter space for isocurvature contributions, strengthening the case for predominantly adiabatic initial conditions and guiding future explorations of multi-field inflationary scenarios.

Abstract

Various recent studies proved that cosmological models with a significant contribution from cold dark matter isocurvature perturbations are still compatible with most recent data on cosmic microwave background anisotropies and on the shape of the galaxy power spectrum, provided that one allows for a very blue spectrum of primordial entropy fluctuations (n_iso > 2). However, such models predict an excess of matter fluctuations on small scales, typically below 40 Mpc/h. We show that the proper inclusion of high-resolution high signal-to-noise Lyman-alpha forest data excludes most of these models. The upper bound on the isocurvature fraction alpha=f_iso^2/(1+f_iso^2), defined at the pivot scale k_0=0.05/Mpc, is pushed down to alpha<0.4, while n_iso=1.9+-1.0 (95% confidence limits). We also study the bounds on curvaton models characterized by maximal correlation between curvature and isocurvature modes, and a unique spectral tilt for both. We find that f_iso<0.05 (95% c.l.) in that case. For double inflation models with two massive inflatons coupled only gravitationally, the mass ratio should obey R < 3 (95% c.l.).

Figures (8)

• Figure 1: (Left) Matter power spectrum for a family of mixed AD+CDI models, with all parameter fixed except $\alpha$ and $\beta$ (the global normalization also varies in order to mantain a fixed amplitude on large scales). In particular, in all models we kept $n_{\rm ad}=0.95$, $n_{\rm iso}=3$ and $n_{\rm cor}=0$. The thick line stands for the pure adiabatic case ($\alpha=0$). The thin solid (red) lines show uncorrelated models ($\beta=0$) with $\alpha=0.1,0.2,0.3,0.4,0.5,0.6$. The lower blue (upper green) dashed lines show the maximally correlated models with $\beta=1$ (anti-correlated with $\beta=-1$) for the same values of $\alpha$. (Right) Same as the left plot, but with $n_{\rm iso}$ reduced to $2.2$.
• Figure 2: Likelihood for the AD+CDI model, using all our data set. The first eleven parameters are independent, while the last four are related parameters (with non-flat priors).
• Figure 3: Two-dimensional likelihood for the amplitude of the isocurvature mode and of the cross-correlation component, near the pivot scale. We adopted a flat prior within the ellipse (which appears here as a circle) in which these parameters are defined.
• Figure 4: Likelihood of the isocurvature-related parameters, for the three combinations of data sets described in section \ref{['impact']}: "Lyman-$\alpha$" (red), "2dF bias prior" (green) and "none" (blue). (Left) Marginalized 1$\sigma$ and 2$\sigma$ confidence levels in the $(\alpha, 2 \beta [\alpha (1- \alpha)]^{1/2})$ space. (Right) Marginalized probability distribution for $n_{\rm iso}$.
• Figure 5: Favored ranges for the matter power spectrum $P(k)$ in the three runs "Lyman-$\alpha$" (dark), "2dF bias prior" (medium) and "none" (light), compared with our Lyman-$\alpha$ data, from the LUQAS quasar spectra (left) and from the re-analyzed Croft et al. spectra (right). The bands represent the envelope of all the matter power spectra in the Markov chains (after eliminating models with the worse likelihood). Each power spectrum has been computed at the median redshift of the data and re-expressed in units of km/s. In addition to the statistical errors, the data points share an overall effective calibration error, whose standard deviation is displayed in the top right corners. For the run including the Lyman-$\alpha$ data, each power spectrum has been divided by the value of the calibration parameter. The red dashed curves show the particular power spectrum discussed in section \ref{['posteriori']}.