Constraining the giant radio galaxy population with machine learning and Bayesian inference

Abstract: Large-scale sky surveys at low frequencies, like the LOFAR Two-metre Sky Survey (LoTSS), allow for the detection and characterisation of unprecedented numbers of giant radio galaxies (GRGs, or 'giants'). In this work, by automating the creation of radio--optical catalogues, we aim to significantly expand the census of known giants. We then combine this sample with a forward model to constrain GRG properties of cosmological interest. In particular, we automate radio source component association through machine learning and optical host identification for resolved radio sources. We create a radio--optical catalogue for the full LoTSS Data Release 2 (DR2) and select all possible giants. We combine our candidates with an existing catalogue of LoTSS DR2 crowd-sourced GRG candidates and visually confirm or reject them. To infer intrinsic GRG properties from GRG observations, we develop further a population-based forward model that takes into account selection effects and constrain its parameters using Bayesian inference. We confirm 5,647 previously unknown giants from the crowd-sourced catalogue and 2,597 previously unknown giants from the ML-driven catalogue. Our confirmations and discoveries bring the total number of known giants to at least 11,585. We predict a comoving GRG number density $$n_\mathrm{GRG} = 13 \pm 10\ (100\ \mathrm{Mpc})<sup>{-3}$$, close to a recent estimate of the number density of luminous non-giant radio galaxies. We derive a current-day GRG lobe volume-filling fraction $$V_\mathrm{GRG-CW}(z = 0) = 1.4 \pm 1.1 \cdot 10<sup>{-5}$$ in clusters and filaments of the Cosmic Web. Our analysis suggests that giants are more common than previously thought. Moreover, tentative results imply that it is possible that magnetic fields once contained in giants pervade a significant ($\gtrsim 10\%$) fraction of today's Cosmic Web.

• The paper introduces an ML-driven radio-optical catalogue and Bayesian model to robustly constrain the giant radio galaxy population from LoTSS data.
• The paper’s methodology integrates five algorithms to optimize host identification, achieving a 50% confirmation rate for GRG candidates.
• The paper reveals a curved power law in GRG length distribution and suggests GRGs play a significant role in cosmic magnetogenesis.

Overview of "Constraining the Giant Radio Galaxy Population with Machine Learning and Bayesian Inference"

This study aims to enhance the understanding of giant radio galaxies (GRGs), specifically those with lengths reaching at least 0.7 Mpc, identified through large-scale sky surveys at low frequencies such as the LOFAR Two-metre Sky Survey (LoTSS). By utilizing ML and Bayesian inference, the researchers have devised methods to efficiently identify GRG populations and analyze their properties.

Key Contributions

1. Automated Radio-Optical Catalogues:
   • The authors have constructed an ML-driven pipeline integrating five existing codes, optimizing the association of radio source components and optical host identification. This approach offers a significant expansion of the known census of GRGs by creating a comprehensive radio-optical catalogue covering the LoTSS DR2 footprint.
2. Population Model and Bayesian Inference:
   • The paper introduces a population-based forward model, factoring in selection effects to infer intrinsic GRG properties, such as length distribution, number density, and lobe volume-filling fraction (VFF) in the Cosmic Web.
   • The model employs a Bayesian framework to derive posterior probability distributions for these parameters.
3. Results and Performance:
   • The ML pipeline exhibited a success rate where approximately half of the identified GRG candidates were confirmed as giants after visual inspection, indicating a significant improvement over previous models, where success rates were markedly lower.
   • Through the analysis of the LoTSS DR2-based samples and existing literature, the authors increased the known total number of GRGs to at least 11,585.

Findings

• The intrinsic length distribution for GRGs aligns with a curved power law probability density function, indicating a dynamically changing slope with increasing length.
• The estimated comoving GRG number density was determined to be close to that of luminous non-giant radio galaxies.
• Preliminary results regarding GRG lobe VFF suggest a potential substantial contribution to magnetic fields within the Cosmic Web, implying a non-negligible cosmic magnetogenesis role.

Implications

This research offers a refined methodology to explore the GRG population, leveraging advancements in ML and statistical methods. The stronger predictive power and precision of the ML pipeline could catalyze advancements in radio astronomy, especially in understanding the large-scale impact of GRGs on their environments. The Bayesian approach enhances confidence in the GRG population model, yielding robust predictions about galactic dynamics and their role in cosmic magnetic fields.

Future Directions

The ongoing enhancement of ML algorithms and data processing capabilities suggests a path forward in increasing the precision of GRG detection and characterization. With further developments in observational technologies and data handling, coupled with refined computational models, the scope for systematic cosmic studies based on GRG populations appears promising. Integrating more observational data from future surveys might continuously update and verify the population models outlined in the study.

This investigation stands as a comprehensive application of state-of-the-art analytics in an astrophysical context, representing a formidable step towards solving questions about the origins and impacts of cosmic magnetic fields.