Performance test of RM CLEAN and its evaluation with chi-square value

TL;DR

This study evaluates the effectiveness of RM CLEAN in reconstructing cosmic magnetic field distributions from polarization spectra, focusing on source resolution and parameter estimation, and highlights limitations revealed by chi-square analysis.

Contribution

It systematically assesses RM CLEAN's performance using simulations, revealing conditions under which it fails to resolve sources and the limitations of chi-square as a quality metric.

Findings

Two sources cannot be resolved when closely located in Faraday depth space.

Chi-square values may be low even when RM CLEAN results are incorrect.

RM CLEAN's resolution depends on initial polarization angles and Faraday depths.

Abstract

RM CLEAN is a standard method to reconstruct the distribution of cosmic magnetic fields and polarized sources along the line of sight (LOS) from observed polarization spectrum. This method is similar to the CLEAN algorithm for aperture synthesis radio telescope images but it is rather unclear in what cases RM CLEAN works well. In this paper, we evaluate the performance of RM CLEAN by simulating spectro-polarimetric observations of two compact sources located in the same LOS, varying the relative initial polarization angle and Faraday depth systematically. Especially, we focus on if the two polarized sources can be resolved in the Faraday depth space and how well the source parameters can be estimated. We confirm the previous studies that two sources cannot be resolved when they are closely located in the Faraday depth space for specific values of the relative initial polarization angle. Further, we calculate the chi-square value for the fit between the mock data of polarization spectrum and the one from RM CLEAN. Then we find that the chi-squared value is not always significantly large even when RM CLEAN gives wrong results.