Speaker
Description
Radio continuum emission at 1.4\,GHz is widely used as a tracer of star formation rate (SFR) in radio-based galaxy evolution studies. However, the standard relation is highly uncertain, due to complications such as active galactic nucleus (AGN) contamination, synchrotron suppression, and high intrinsic scatter. Thus we investigate the use of a machine learning framework trained on WISE-based infrared SFRs to predict SFR from radio continuum emission. Utilising MeerKAT Galaxy Cluster Legacy Survey (MGCLS) data, our analysis spans galaxy cluster cores ($R < R_{500}$), outer cluster regions ($R > R_{500}$), and the VLA-COSMOS field ($z \lesssim 6.5$). Across all environments we find no strong correlation, linear or otherwise, between predicted and observed radio flux densities. The lack of correlation persists even after removing AGN contamination. We find a consistent Gaussian Mixture Modelling breakpoint in our galaxy clusters near $\sim 0.2$\,mJy. Below this threshold, correlations improve modestly (from $\rho \simeq 0.21$ to $\rho \simeq 0.39$), particularly outside cluster environments, likely due to reduced AGN and intra-cluster contamination. Inverse modelling (predicting radio luminosity from SFR) yields a tighter ($\rho \simeq 0.98$), near-linear relation, underscoring an asymmetry in predictability. These results demonstrate that the radio–SFR connection is not only extremely weak, but also depends on a number of physical factors; highlighting the necessity of multi-wavelength and environmental context when interpreting radio observations. This is the first study to quantify the breakdown of the radio luminosity – SFR relation across both cluster and field environments to $z \sim 6.5$ using machine learning and multi-wavelength data.
| Stream | Science or Engineering |
|---|