Improved CLEAN reconstructions for rotation measure synthesis with maximum likelihood estimation

Abstract

The CLEAN deconvolution algorithm has well-known limitations due to the restriction of locating point source model components on a discretized grid. In this Letter, we demonstrate that these limitations are even more pronounced when applying CLEAN in the case of rotation measure (RM) synthesis imaging (known as RMCLEAN in this context). We suggest an approach that uses maximum likelihood estimation to adjust the RMCLEAN-derived model. We demonstrate through the use of mock, one-dimensional RM synthesis observations that this technique improves significantly over standard RMCLEAN and gives results that are independent of the chosen pixelization. We suggest using this simple modification to RMCLEAN in upcoming polarization sensitive sky surveys. © 2013 ESO.