A Bayesian method for point source polarisation estimation

¹ Instituto de Física de Cantabria, CSIC-UC, Av. de Los Castros s/n, 39005 Santander, Spain
e-mail: herranz@ifca.unican.es
² Departamento de Matemáticas, Universidad de Oviedo, C. Federico García Lorca 18, 33007 Oviedo, Spain
³ Departamento de Física, Universidad de Oviedo, C. Federico García Lorca 18, 33007 Oviedo, Spain
⁴ Instituto Universitario de Ciencias y Tecnologías Espaciales de Asturias (ICTEA), Escuela de Ingeniería de Minas, Materiales y Energía de Oviedo, C. Independencia 13, 33004 Oviedo, Spain
⁵ Departamento de Física Moderna, Universidad de Cantabria, 39005 Santander, Spain
⁶ ESAC, Camino Bajo del Castillo s/n, 28692 Villafranca del Castillo, Madrid, Spain

Received: 22 October 2020
Accepted: 2 March 2021

Abstract

The estimation of the polarisation P of extragalactic compact sources in cosmic microwave background (CMB) images is a very important task in order to clean these images for cosmological purposes –for example, to constrain the tensor-to-scalar ratio of primordial fluctuations during inflation– and also to obtain relevant astrophysical information about the compact sources themselves in a frequency range, ν ∼ 10–200 GHz, where observations have only very recently started to become available. In this paper, we propose a Bayesian maximum a posteriori approach estimation scheme which incorporates prior information about the distribution of the polarisation fraction of extragalactic compact sources between 1 and 100 GHz. We apply this Bayesian scheme to white noise simulations and to more realistic simulations that include CMB intensity, Galactic foregrounds, and instrumental noise with the characteristics of the QUIJOTE (Q U I JOint TEnerife) experiment wide survey at 11 GHz. Using these simulations, we also compare our Bayesian method with the frequentist filtered fusion method that has been already used in the Wilkinson Microwave Anisotropy Probe data and in the Planck mission. We find that the Bayesian method allows us to decrease the threshold for a feasible estimation of P to levels below ∼100 mJy (as compared to ∼500 mJy which was the equivalent threshold for the frequentist filtered fusion). We compare the bias introduced by the Bayesian method and find it to be small in absolute terms. Finally, we test the robustness of the Bayesian estimator against uncertainties in the prior and in the flux density of the sources. We find that the Bayesian estimator is robust against moderate changes in the parameters of the prior and almost insensitive to realistic errors in the estimated photometry of the sources.