Improving Foreground Models for CMB B-mode Observations

One of the main challenges in detecting the faint B-mode polarization signal of the cosmic microwave background (CMB) is the modeling of polarized foregrounds. Thermal dust emission and synchrotron radiation dominate over the primordial B-mode signal across all frequencies, and must be modeled at sub-percent accuracy to avoid biasing the primordial signal. We use Planck data to estimate the variation in dust temperature and spectral index along individual lines of sight and show that if a single modified blackbody model is used, the expected bias in the tensor-to-scalar ratio will be significant compared to the target sensitivities of planned observations from increasingly sensitive CMB experiments. We discuss methods for fitting more complicated models while making minimal assumptions about the underlying distribution of physical parameters and quantify the extent to which planned CMB missions can differentiate between models of dust emission.