Extreme equivalent width-selected low-mass starbursts at $z=4-9$: insights into their role in cosmic reionization

TL;DR

This study characterizes extreme emission line galaxies at $z\sim4-9$ using JWST/NIRSpec spectra from CAPERS, CEERS, and RUBIES, assembling a sample of 160 EELGs selected photometrically for rest-frame EW$([O III]+H\beta)$. The objects are low-mass, compact systems with elevated sSFR and recent bursts, producing high ionizing photon budgets ($\xi_{ion}\sim10^{25.34}$ Hz erg$^{-1}$) but exhibiting heterogeneous LyC leakage (median $f_{esc}^{LyC}\sim4.4\%$; $\sim9\%$ with $f_{esc}^{LyC}>10\%$). Emission lines are predominantly powered by young stars, though a minority shows AGN-like ionization signatures or broad-line components. The analysis demonstrates that elevated sSFR and compact star formation are the primary drivers of extreme EWs and assesses the potential contribution of EELGs to cosmic reionization, highlighting the need to account for dust, burstiness, and size when predicting LyC escape in the early universe.

Abstract

We investigate the properties of extreme emission line galaxies (EELGs) at $z=4-9$ and their role in reionization. Compact, low-mass galaxies with intense optical emission lines are linked to elevated specific star formation rates (sSFRs) and recent bursts of star formation. Feedback in these systems may enable the leakage of ionizing radiation into the intergalactic medium. Using JWST/NIRSpec spectroscopy from the CAPERS, CEERS, and RUBIES surveys, we compile 160 NIRCam-selected EELGs in the EGS field. These galaxies show extreme rest-frame equivalent widths (EWs), with a median EW([O III]+H$β$)=1616Å and EW(H$α$)=763Å. They are low-mass (median log(M$_{\star}$/M$_{\odot}$)=8.26) with high sSFRs (median 43 Gyr$^{-1}$), above the $z\sim6$ main sequence. UV slopes are diverse, with a mean $β=-2.0$, and only 7% have extremely blue continua ($β<-2.6$). Emission-line diagnostics suggest stellar populations as the primary ionizing source, although an AGN fraction of 14% is found. These galaxies are efficient ionizing photon producers, with mean log($ξ_{\rm ion}$ [Hz erg$^{-1}$])=25.34, exceeding typical values at similar redshifts. Escape fractions, however, are heterogeneous: 9% of EELGs show escape fractions $>$10% for both Ly$α$ and LyC photons, while 82% lack detectable Ly$α$ emission. The median inferred LyC escape fraction is modest (4.4%) but enhanced in super-Eddington systems with sSFR >25 Gyr$^{-1}$. The galaxies are extremely compact, with a median effective radius of 0.49 kpc, and exhibit a recent star-formation burst. Our analysis indicates that sSFR and star-formation rate surface density are the primary drivers of their extreme emission line strengths.

Figures (21)

• Figure 1: Left panel: Comparison between $z_{phot}$ and $z_{spec}$. The dashed black line is the 1:1 relation. In red (blue), the galaxies with PRISM (MR) spectra. Inset: Distribution of $z_{phot}$ in the parent sample. Right panel: Distribution of $z_{spec}$, shown as superimposed histograms.
• Figure 2: Distribution of the sample in the stellar mass-SFR plane. The triangle symbols are upper limits due to low H${\beta}$ S/N. The dashed black and blue lines are the main-sequence of SF galaxies at $z=6$ from Iyer2018 and Speagle_2014, respectively, while the dashed red line represents the main-sequence at $4<z<10$ from Calabro2024sigmasfr. The cyan (magenta) circles are the sample of strong (weak) LyC leakers at $z\sim0.3$Flury2022IJaskot2024aJaskot2024b. The dashed-dotted lines represent constant sSFR lines.
• Figure 3: Comparison between EWs estimated from the photometry and from spectroscopy. Top panel: EW([O III]+H${\beta}$). The triangle symbols are upper limits due to low S/N of H${\beta}$. Bottom panel: EW(H${\alpha}$). The dashed black line is the 1:1 relation.
• Figure 4: Relation between H${\alpha}$/H${\beta}$ and EW([O III]). In the right y-axis, we show the corresponding E$(\mathrm{B-V})_\mathrm{neb}$. The gray shaded region corresponds to the mean error $\sim0.06$ dex in the H${\alpha}$/H${\beta}$ ratio.
• Figure 5: Relation between M$_{UV}$ and $\beta$. The black dashed line is the relation from Dottorini2024. The dashed green line represents the predictions from SAMs at $z=5$Yung2019. The red line traces the median values of $\beta$ as a function of $M_{UV}$, computed in bins of $M_{UV}$. The shaded region corresponds to the 1$\sigma$ observed scatter in the sample. Cyan and magenta symbols are as in Fig. \ref{['fig:MS']}.