Radio Spectral Energy Distribution of Low-$z$ Metal Poor Extreme Starburst Galaxies: Novel insights on the escape of ionizing photons

Abstract

Recent optical surveys have identified a rare population of low-$z$ ($z \sim 0.01 - 0.06$) extreme star-forming galaxies (xSFGs) characterized by very low metallicity, strong emission lines, extremely high specific star-formation rate, low stellar mass, and strong Ly~$α$ emission. Their global properties resemble recently discovered $z > 6$ reionization-era star-forming galaxies. We present new multi-frequency radio continuum (RC) observations of $8$ xSFGs using the upgraded Giant Metrewave Radio Telescope (uGMRT) at $1.25$ GHz, the Karl G. Jansky Very Large Array (VLA) at $1.5, 3.0, 6.0, 10.0$ and $15.0$ GHz, along with archival LOw Frequency ARray (LOFAR) data at $150$ MHz for several sources. These data allow construction of the radio spectral energy distribution (radio-SED) from $\sim 1$ GHz (down to $150$ MHz for some sources) to $15$ GHz, spanning nearly two orders of magnitude in frequency. We find that xSFGs exhibit a flat spectral index between $6$ and $15$ GHz, while a subset shows spectral turnovers at $0.3 - 3$ GHz. Our Bayesian radio-SED modeling indicates that these features are consistent with a high thermal fraction combined with free-free absorption, requiring high emission measures in some systems. By comparing modeled thermal radio emission with observed H$β$ line flux density, we find evidence for dust in several xSFGs. Finally, we confirm a previously reported correlation between Lyman continuum escape fraction, ionization state, and radio spectral index, particularly among strong leakers.

Figures (9)

• Figure 1: Bayesian radio-SED fitting results for J160810+352809 using the SFG-thin model. Top (bottom) row shows the fitting with flat (Gaussian) priors on _1GHz^th S$_{\mathrm{1GHz}}^{\mathrm{th}}$. The left panels shows RC detections with blue error bars (wth $1\sigma$ errors), and non-detections ($3\sigma$ upper limits) are shown in red arrows. The corresponding best-fit (median) SFG-thin model is shown in solid blue line. The thermal (non-thermal) components are shown in green (red) solid line. The shaded area shows the 16th and 84th percentile range derived from the MCMC samples. We show the fraction residuals ( (data - best-fit model)/best-fit model) in the lower subplot in the left column. The right column shows the 2D and 1D marginalized posteriors on _1GHz^th S$_{\mathrm{1GHz}}^{\mathrm{th}}$, _1GHz^nth S$_{\mathrm{1GHz}}^{\mathrm{nth}}$ and _nth $\alpha_\mathrm{nth}$ derived from the MCMC chains. The contours enclose $19.3\%, 68\%, 95\%$ and $99.7\%$ of the 2D marginalized posterior probability. The dashed lines on the 1D marginalized posteriors show the $16$th, $50$th (median) and $84$th percentiles. See Section \ref{['sec: radio-sed individual cases']} and Table \ref{['table: bayes fitting']} for details on the priors, choice of model and the bands used for the Bayesian fitting. Table \ref{['table: bayes parameters']} summarizes the constraints on the various parameters and constraints on f_1 GHz^th (SED) f$_{1 \mathrm{GHz}}^\mathrm{th (SED)}$.
• Figure 2: Bayesian radio-SED fitting results for J1032+4919 using the SFG-thick model. The various panels follow the same order as Figure \ref{['fig: BB13 bayes rado-sed fit']} in terms of the choice of priors on _1GHz^th S$_{\mathrm{1GHz}}^{\mathrm{th}}$. We constrain the _t $\nu_\mathrm{t}$ to $\sim 3$GHz across the different choice of priors (Table \ref{['table: bayes parameters']}). Overall we find that the radio-SED is well fit with a thermally dominant SFG-thick radio-SED with a turnover at $\sim 3$ GHz. See Section \ref{['sec: radio-sed individual cases']} for a detailed discussion on this source.
• Figure 3: Bayesian radio-SED fitting results for J150934+373146 using the SFG-thick model. The various panels follow the same order as Figure \ref{['fig: J1032 bayes rado-sed fit']} in terms of the choice of priors on _1GHz^th S$_{\mathrm{1GHz}}^{\mathrm{th}}$. We constrain the _t $\nu_\mathrm{t}$ to $\sim 0.34 - 0.39$ GHz across the different choice of priors (Table \ref{['table: bayes parameters']}). The radio-SED is well fit with a non-thermally dominant SFG-thick radio-SED typical of normal star-forming galaxies and unlike other xSFGs from our sample. See Section \ref{['sec: radio-sed individual cases']} for a detailed discussion on this source.
• Figure 4: Bayesian radio-SED fitting results for J0159+4919 using the SFG-thin model. The various panels follow the same order followed in Figure \ref{['fig: BB13 bayes rado-sed fit']}. See Section \ref{['sec: radio-sed individual cases']} and Table \ref{['table: bayes fitting']} for details on the priors, choice of model and the bands used for the Bayesian fitting. Table \ref{['table: bayes parameters']} summarizes the constraints on the various parameters and constraints on f_1 GHz^th (SED) f$_{1 \mathrm{GHz}}^\mathrm{th (SED)}$. We find that the radio-SED is well fit with a thermally dominant component.
• Figure 5: Bayesian radio-SED fitting results for J1355+4651 using the SFG-thin model. The various panels follow the same order as Figure \ref{['fig: BB13 bayes rado-sed fit']} in terms of the choice of priors on _1GHz^th S$_{\mathrm{1GHz}}^{\mathrm{th}}$. Overall we find that the radio-SED is well fit with a thermally dominant SFG-thin radio-SED. See Section \ref{['sec: radio-sed individual cases']} for a detailed discussion on this source. We notice weak systematic offset ($<2\sigma$) in the residuals at $10.0$ and $15.0$ GHz bands.