Reionization constraints from five-year WMAP data

TL;DR

This paper uses five-year WMAP data to improve constraints on the reionization history, showing that the total optical depth is better determined and that future polarization measurements can further refine our understanding of reionization.

Contribution

The study introduces a principal component approach to model reionization, providing a model-independent estimate of optical depth with improved precision over previous data.

Findings

Optical depth estimate is nearly twice as precise as 3-year WMAP results.

Scale invariant n_s=1 spectra are no longer strongly disfavored.

Better measurements of E-mode polarization are needed to further constrain reionization history.

Abstract

We study the constraints on reionization from five years of WMAP data, parametrizing the evolution of the average fraction of ionized hydrogen with principal components that provide a complete basis for describing the effects of reionization on large-scale E-mode polarization. Using Markov Chain Monte Carlo methods, we find that the resulting model-independent estimate of the total optical depth is nearly twice as well determined as the estimate from 3-year WMAP data, in agreement with simpler analyses that assume instantaneous reionization. The mean value of the optical depth from principal components is slightly larger than the instantaneous value; we find tau=0.097+/-0.017 using only large-scale polarization, and tau=0.101+/-0.019 when temperature data is included. Likewise, scale invariant n_s=1 spectra are no longer strongly disfavored by WMAP alone. Higher moments of the ionization history show less improvement in the 5-year data than the optical depth. By plotting the distribution of polarization power for models from the MCMC analysis, we show that extracting most of the remaining information about the shape of the reionization history from the CMB requires better measurements of E-mode polarization on scales of l~10-20. Conversely, the quadrupole and octopole polarization power is already predicted to better than cosmic variance given any allowed ionization history at z<30 so that more precise measurements will test the LambdaCDM paradigm.