2024A&A...683A.174W

Context. Radio wavelengths offer a unique possibility to trace the total star-formation rate (SFR) in galaxies, both obscured and unobscured. To probe the dust-unbiased star-formation history, an accurate measurement of the radio luminosity function (LF) for star-forming galaxies (SFGs) is crucial.
Aims. We make use of an SFG sample (5900 sources) from the Very Large Array (VLA) COSMOS 3 GHz data to perform a new modeling of the radio LF. By integrating the analytical LF, we aim to calculate the history of the cosmic SFR density (SFRD) from z ∼ 5 onwards.
Methods. For the first time, we use both models of the pure luminosity evolution (PLE) and joint luminosity+density evolution (LADE) to fit the LFs directly to the radio data using a full maximum-likelihood analysis, considering the sample completeness correction. We also incorporate updated observations of local radio LFs and radio source counts into the fitting process to obtain additional constraints.
Results. We find that the PLE model cannot be used to describe the evolution of the radio LF at high redshift (z > 2). By construct, our LADE models can successfully fit a large amount of data on radio LFs and source counts of SFGs from recent observations. The Akaike information criterion (AIC) also demonstrates that the LADE model is superior to the PLE model. We therefore conclude that density evolution is genuinely indispensable in modeling the evolution of SFG radio LFs. Our SFRD curve shows a good fit to the SFRD points derived by previous radio estimates. In view of the fact that our radio LFs are not biased, as opposed those of previous studies performed by fitting the 1/V_max LF points, our SFRD results should be an improvement on these previous estimates. Below z ∼ 1.5, our SFRD matches a published multiwavelength compilation, while our SFRD turns over at a slightly higher redshift (2 < z < 2.5) and falls more rapidly out to high redshift.