A JWST Paschen-alpha Calibration of the Radio Luminosity-Star Formation Rate Relation at z~1.3

TL;DR

This study tests the reliability of radio luminosities as tracers of star formation at $z\sim1.3$ by calibrating the $L_{1.4\mathrm{GHz}}$–$SFR_{\mathrm{Pa}\alpha}$ relation using a JWST Paschen-$\alpha$ sample cross-matched with deep MeerKAT 1.23 GHz data. By removing AGN and blending effects and employing stacked non-detections alongside a Bayesian regression, the authors derive a power-law relation $\log(L_{1.4\mathrm{GHz}}) = (1.31\pm0.17)\log(\mathrm{SFR}_{\mathrm{Pa}\alpha}) + (21.36\pm0.17)$, consistent with local and $z\sim1$ calibrations. A toy model that delays and averages the synchrotron component by roughly $10-75$ Myr helps explain part of the observed scatter between radio luminosity and Paschen-$\alpha$ SFR, with implications for how radio emission traces evolving starbursts. The work supports the use of radio continuum as a dust-free SFR tracer at cosmic noon and provides physical insight into cosmic-ray electron lifetimes and magnetic-field-related processes relevant for upcoming SKA/ngVLA calibrations.

Abstract

As radio emission from normal galaxies is a dust-free tracer of star formation, tracing the star formation history of the Universe is a key goal of the SKA and ngVLA. In order to investigate how well radio luminosity traces star formation rate (SFR) in the early Universe, we have examined the radio properties of a JWST Paschen-alpha sample of galaxies at 1.0<=z<=1.8. In the GOODS-S field, we cross-matched a sample of 506 FRESCO Paschen-alpha emitters with the 1.23 GHz radio continuum data from the MeerKAT MIGHTEE survey finding 47 detections. After filtering for AGN (via X-ray detections, hot mid-infrared dust and extended radio emission), as well as blended sources, we obtained a sample of SFGs comprising: 11 cataloged radio detections, 18 non-cataloged detections (at ~3-5sigma) and 298 undetected sources. Stacking the 298 undetected sources we obtain a 3.3sigma detection in the radio. This sample, along with a local sample of Paschen-alpha emitters, lies along previous radio luminosity/SFR relations from local (z<0.2) to high redshift (z~1). Fitting the FRESCO data at 1.0<=z<=1.8 we find log(L_1.4GHz) = (1.31+/-0.17) x log(SFR_Pa-alpha) + (21.36+/-0.17) which is consistent with other literature relations. We can explain some of the observed scatter in the L_1.4GHz/SFR_Pa-alpha correlation by a toy model in which the synchrotron emission is a delayed/averaged tracer of the instantaneous Paschen-alpha SFR by ~10/75 Myr.

Figures (5)

• Figure 1: Attenuation corrected line luminosities with uncertainties plotted against redshift. (left) The 506 Paschen-$\alpha$ sources in the GOODS-South sample with red squares indicating the 47 sources which match the host galaxies of the cataloged radio sources from the E-CDFS MIGHTEE image at $1.23\,$GHz. (right) Radio cataloged detections of Paschen-$\alpha$ sources with AGN identified (as described in §\ref{['sec:agn']}). AGN are selected via X-ray detections from a deep Chandra image Luo:17, mid-IR warm dust from SED fitting using NIRCam/MIRI data Lyu:24 or from extended radio emission from MIGHTEE (§\ref{['sec:data:radio']}). One source is removed due to blended radio emission. Of 47 radio sources only 11 are SFGs free from AGN activity.
• Figure 2: Distribution in Paschen-$\alpha$ luminosity/distance parameter space of the local sample from tateuchi:15 as presented in §\ref{['sec:data:local']}. Sources with Seyfert or LINER indicators in their optical spectrum are marked. These are removed from further analysis to avoid contamination by AGN as discussed in §\ref{['analysis:local']}.
• Figure 3: Histogram of the SNR of MIGHTEE flux density pixels values at the positions of Paschen-$\alpha$ sources without counterparts in the radio catalog. The 18 sources with values above $3\sigma$ (the vertical dashed line) are considered non-cataloged detections.
• Figure 4: Radio luminosity plotted as function of Paschen-$\alpha$ luminosity for our various AGN removed sub-samples: the 11 cataloged radio-detections, the 18 non-cataloged detections and the stack of the 298 remaining non-detections after removing blended sources. The local sample of 14 Paschen-$\alpha$ emitters from tateuchi:15, minus AGN, is also plotted. Both axes have error bars plotted which are sometimes smaller than the symbol. Overlaid as lines are the expected 1.4 GHz free-free luminosity from a Paschen-$\alpha$ source as well as $\times 10$ and $\times 100$ this value which corresponds to 10% and 1% 1.4 GHz thermal fractions (TFs), respectively. Nearly all galaxies have TF in the range $1-10\%$ with the local sample having slightly higher TF.
• Figure 5: Radio luminosity plotted as function of Paschen-$\alpha$ SFR for our various sub-samples after the removal of AGN: the 11 cataloged radio detections, the 18 non-cataloged radio detections, the 14 local sources and the 298 remaining Paschen-$\alpha$ non-detections. We also present the stacked value of these non-detections although it is not used in the fitting. For comparison we present the H$\alpha$/radio luminosity results from duncan:20 at $z\sim 1.65$. We overlay various conversion factors from the literature including Bell:03molnar:21cook:24 and matthews:21 as well as indicating the redshift range they were evaluated over (the length of the line indicates the SFR range they were determined over). The best fit radio luminosity/SFR correlation is presented (see §\ref{['sec:mod:conv']}) with the pink tracks representing solutions within $1\sigma$ of the median fit of both parameters. We also overlay a starburst evolutionary track as described in §\ref{['sec:mod:evolution']} with the arrows indicating the direction of time. The starburst has a Cauchy-Lorentz (pseudo-exponential) rise and fall temporal form with the synchrotron component of the radio luminosity delayed by 20 Myr (compared to 10 Myr for the Paschen-$\alpha$ and free-fee components). The SFRs are averaged over previous 10 and 75 Myr for the Paschen-$\alpha$ SFR/free-free luminosity and synchrotron luminosity, respectively. This model can partly explain the scatter around the best fit line in this and other work.