RM-Tools: Software for Analyzing Polarized Radio Spectra

Polarization observations using modern radio telescopes cover large numbers of frequency channels over broad bandwidths, and require advanced techniques to extract reliable scientific results. We present RM-Tools, analysis software for deriving polarization properties, such as Faraday rotation measures, from spectropolarimetric observations of linearly polarized radio sources. The software makes use of techniques such as rotation measure synthesis and QU-model fitting, along with many features to simplify and enhance the analysis of radio polarization data. RM-Tools is currently the main software that large-area polarization sky surveys such as POSSUM and VLASS deploy for science-ready data processing. The software code is freely available online and can be used with data from a wide range of telescopes.

💡 Research Summary

RM‑Tools is an open‑source Python package that brings together the full suite of techniques required to analyse modern broadband spectropolarimetric data from radio telescopes. The authors begin by reviewing the physics of radio polarisation, defining the Stokes parameters (I, Q, U, V) and the complex linear polarisation \tilde{P}=Q+iU, and then introduce Faraday rotation, Faraday depth φ, and the special case of a Faraday‑thin source where the depth is called the rotation measure (RM).

The traditional method for extracting Faraday information, RM‑synthesis, is described in detail. The algorithm performs a weighted Fourier transform of the complex polarisation as a function of λ², producing the Faraday dispersion function (FDF). The paper presents the full discrete formulation, including channel weights w_k, a reference wavelength λ₀², normalisation, and the construction of the rotation‑measure spread function (RMSF). The authors explain how the RMSF’s full‑width‑half‑maximum (FWHM) sets the Faraday resolution and how the “dirty” FDF is deconvolved using an RM‑CLEAN algorithm. They also discuss the importance of normalising Q and U by Stokes I to avoid spectral‑index‑induced artefacts, and they provide an analytic expression for the theoretical noise in the FDF.

The key innovation of RM‑Tools is the integration of QU‑fitting alongside RM‑synthesis. QU‑fitting directly models the complex polarisation spectrum with physically motivated parametric forms (e.g., multiple Faraday‑thin components, Faraday‑thick slabs, uniform or turbulent magnetic fields). The fitting is performed with robust non‑linear optimisation techniques such as Markov Chain Monte Carlo (MCMC) and nested sampling, which yield posterior distributions for all model parameters. Model comparison is handled through a suite of statistical criteria: raw χ², Akaike and Bayesian information criteria (AIC/BIC), and full Bayesian evidence. This framework allows users to select the most appropriate physical description of a source while penalising over‑parameterisation.

A novel “Faraday complexity metric” is introduced to quantify how many distinct Faraday components are required to explain a given dataset. The metric evaluates the number of significant peaks in the deconvolved FDF relative to the RMSF and provides an automatic flag for sources that are likely Faraday‑complex. This aids in deciding whether a simple RM measurement or a full QU‑fit is warranted.

From a software engineering perspective, RM‑Tools is built as a modular Python library. Core modules handle data I/O (including FITS and CASA measurement sets), channel masking, Stokes I normalisation, RM‑synthesis, RMS‑CLEAN, QU‑fitting, visualisation, and batch processing. Both a command‑line interface and Jupyter‑notebook widgets are provided, making the package accessible to novices and power users alike. The library integrates with Dask for parallel execution, enabling the processing of millions of sources in large surveys such as POSSUM (ASKAP) and VLASS (VLA).

The authors validate the package using both simulated data and real observations. Simulations span a range of Faraday configurations: single thin components, multiple thin components, and continuous thick slabs. Comparisons show that for low‑complexity cases the RM‑synthesis peak and QU‑fit parameters agree, while for high‑complexity cases the complexity metric correctly flags the need for QU‑fitting, which then recovers the underlying physical parameters (e.g., foreground electron density and magnetic field strength) more reliably than the non‑parametric RM‑synthesis alone. Real‑world tests on AGN and diffuse Galactic emission from POSSUM and VLASS demonstrate that RM‑Tools can automatically generate science‑ready RM catalogs, deconvolve complex Faraday structures, and provide robust uncertainty estimates.

Future development plans include multi‑band RM‑synthesis for ultra‑wideband data (0.5–15 GHz), machine‑learning classifiers to automate model selection based on the complexity metric, GPU‑accelerated kernels for real‑time processing on SKA‑Low, and a plug‑in architecture that allows the community to contribute new physical models or visualisation tools.

In summary, RM‑Tools delivers a comprehensive, reproducible, and scalable solution for radio polarisation analysis. By unifying RM‑synthesis, RM‑CLEAN, QU‑fitting, statistical model selection, and a new Faraday‑complexity diagnostic within a single, well‑documented code base, it substantially lowers the barrier for extracting reliable magnetic‑field information from the massive spectropolarimetric datasets that modern and upcoming radio surveys will produce.