Large-scale polarized foreground component separation for Planck

We use Bayesian component estimation methods to examine the prospects of large-scale polarized map and cosmological parameter estimation with simulated Planck data assuming simplified white noise properties. The sky signal is parametrized as the sum of the cosmic microwave background (CMB), synchrotron emission, and thermal dust emission. The synchrotron and dust emission components are modelled as power laws in frequency, with a spatially varying spectral index for synchrotron and a uniform index for dust. Using the Gibbs sampling technique, we estimate the linear polarization Q and U posterior amplitudes of the CMB, synchrotron and dust maps as well as the two spectral indices in ~4°pixels. We use the recovered CMB map and its covariance in an exact pixel likelihood algorithm to estimate the optical depth to reionization τ, the tensor-to-scalar ratio r, and to construct conditional likelihood slices for C^EE_ℓ and C^BB_ℓ. Given our foreground model, we find σ(τ) ≈ 0.004 for τ= 0.1, σ(r) ≈ 0.03 for a model with r= 0.1, and a 95 per cent upper limit of r < 0.02 for r= 0.0.