Pixel domain implementation of the Minimally Informed CMB MAp foreground Cleaning method

TL;DR

This paper introduces a pixel domain implementation of a non-parametric maximum likelihood method for separating CMB signals from foregrounds, improving robustness to complex foreground models in polarization data analysis.

Contribution

The authors develop and validate a pixel space implementation of a minimally informed CMB cleaning method that handles spatially variable foreground properties, enhancing existing non-parametric techniques.

Findings

The pixel domain implementation matches the original method's performance.

It effectively accounts for spatial variability in foreground properties.

The method outperforms parametric techniques in complex foreground scenarios.

Abstract

High fidelity separation of astrophysical foreground contributions from the cosmic microwave background (CMB) signal has been recognized as one of the main challenges of modern CMB data analysis, and one which needs to be addressed in a robust way to ensure that the next generation of CMB polarization experiments lives up to its promise. In this work we consider the non-parametric maximum likelihood CMB cleaning approach recently proposed by some of the authors which has been shown to match the performance of standard parametric techniques for simple foreground models, while superseding it in cases where the foregrounds do not exhibit a simple frequency dependence. We present a new implementation of the method in pixel space, extending its functionalities to account for spatial variability of the properties of the foregrounds. We describe the algorithmic details of our approach and its validation against the original code as well as the parametric method for various experimental set-ups and different models of the foreground components. We argue that the method provides a compelling alternative to other state-of-the-art techniques.