Tensors, BICEP2, prior dependence, and dust

TL;DR

This paper examines how prior choices and dust foregrounds influence the interpretation of BICEP2 and Planck data on primordial tensor perturbations, highlighting the importance of parameterization and foreground modeling.

Contribution

It introduces an optimized parameterization of the tensor spectrum and analyzes the impact of dust foregrounds on inflationary parameter constraints.

Findings

Without foregrounds, BICEP2 and Planck results suggest a blue-tilted tensor spectrum, contradicting standard inflation.

Accounting for dust foregrounds aligns BICEP2 results with standard inflation predictions.

Sufficient dust can mask the primordial signal, making detection by BICEP2 unlikely under current limits.

Abstract

We investigate the prior dependence on the inferred spectrum of primordial tensor perturbations, in light of recent results from BICEP2 and taking into account a possible dust contribution to polarized anisotropies. We highlight an optimized parameterization of the tensor power spectrum, and adoption of a logarithmic prior on its amplitude $A_T$, leading to results that transform more evenly under change of pivot scale. In the absence of foregrounds the tension between the results of BICEP2 and Planck drives the tensor spectral index $n_T$ to be blue-tilted in a joint analysis, which would be in contradiction to the standard inflation prediction ($n_T<0$). When foregrounds are accounted for, the BICEP2 results no longer require non-standard inflationary parameter regions. We present limits on primordial $A_T$ and $n_T$, adopting foreground scenarios put forward by Mortonson & Seljak and motivated by Planck 353 GHz observations, and assess what dust contribution leaves a detectable cosmological signal. We find that if there is sufficient dust for the signal to be compatible with standard inflation, then the primordial signal is too weak to be robustly detected by BICEP2 if Planck+WMAP upper limits from temperature and E-mode polarization are correct.