Galaxies in the Epoch of Reionization Are All Bark and No Bite -- Plenty of Ionizing Photons, Low Escape Fractions

Abstract

Early results from JWST suggest that epoch-of-reionization (EoR) galaxies produce copious ionizing photons, which, if they escape efficiently, could cause reionization to occur too early. We study this problem using \jwst\ imaging and prism spectroscopy for 412 galaxies at 4.5 < z < 9.0. We fit these data simultaneously with stellar-population and nebular-emission models that include a parameter for the fraction of ionizing photons that escape the galaxy, $f_\mathrm{esc}$. We find that the ionization production efficiency, $ξ_\mathrm{ion}$ = Q(H) / L(UV), increases with redshift and decreasing UV luminosity, but shows significant scatter, $σ( \log ξ_\mathrm{ion})$ = 0.3 dex. The inferred escape fractions averaged over the population are low, ranging from $\langle f_\mathrm{esc} \rangle$ = $2.6\pm 1.4$\% at 6 < z < 9 to $6.5\pm 2.2$\% at 4.5 < z < 6 with weak or no indication of evolution with redshift. This implies that in our models most of the ionizing photons need to be absorbed to account for the nebular emission. We compute the impact of our results on reionization, including the distributions for $ξ_\mathrm{ion}$ and $f_\mathrm{esc}$, and the evolution and uncertainty of the UV luminosity function. Considering galaxies brighter than M(UV) < -16 mag, we would produce an IGM hydrogen-ionized fraction of $x_e = 0.5$ at 5.3 < z < 5.8, possibly too late compared to constraints from from QSO sightlines. Including fainter galaxies, M(UV) < -14 mag, we obtain $x_e = 0.5$ at 6.0 < z < 8.1, fully consistent with QSO and CMB data. This implies that EoR galaxies produce plenty of ionizing photons, but these do not efficiently escape. This may be a result of high gas column densities combined with burstier star-formation histories, which limit the time massive stars are able to clear channels through the gas for ionizing photons to escape.

Figures (17)

• Figure 1: Redshift and magnitude distributions of the spectroscopic sources at $4.5 < z < 9.0$ used in this work The bottom panel shows the redshift versus the NIRCam F277W magnitude distribution for galaxies in JADES and CEERS. A minority of the CEERS galaxies do not include NIRCam imaging, and we show the WFC3/F160W magnitude. The top panel shows the histograms show the distribution of sources from JADES, CEERS, and the total (the sum of the two), as labeled. JADES includes more galaxies overall, but weighted toward galaxies at the lower end of our redshift range. The CEERS sample is more evenly spread in redshift, and provides greater coverage at the higher redshift end of the redshift range.
• Figure 2: Examples of SED fits to galaxies in the CEERS sample. Each pair of plots shows the best-fit model for an individual galaxy, labeled by the MPT ID number. In each pair of plots, the top panel shows the model fit to the broad-band photometry. The black data points with error bars show the measured flux densities. The curves show a best-fit total model (green) and the contributions to that model from the stellar light (yellow) and nebular emission (blue). The green squares are the model photometry. The bottom panel of each galaxy shows the same model fit (green curve and shading) to the NIRSpec prism data (grey).
• Figure 3: Same as Figure \ref{['fig:sedfits_ceers']} but for example galaxies in the JADES sample.
• Figure 4: Comparison of $\xi_\mathrm{ion}$ derived from the H$\alpha$ to that from the H$\beta$ lines. The $\xi_\mathrm{ion}$ values are derived based on each Balmer emission line. In both cases we assume the dust attenuation of the stellar continua is equal to that of the nebular emission, $E(B-V)_\mathrm{nebular}=E(B-V)_\mathrm{stars}$. The solid line shows the unity relation. These results show H$\beta$ is a reasonable surrogate for H$\alpha$ when deriving $\xi_\mathrm{ion}$ for our datasets.
• Figure 5: UV absolute magnitude, $M_\mathrm{UV}$, versus the ionizing photon production efficiency, $\xi_\mathrm{ion}$. The data points show the results derived from the CEERS and JADES datasets, color-coded by redshift. The solid lines show linear fits to the relation in bins of redshift, as labeled. These are consistent with results derived based on the analysis of photometric broad-bands Simmonds_2024 and other spectroscopic analyses of $4 < z < 10$llerena_2024, with noticeable differences (e.g., Pahl_2024, see text). At the bright end, $\hbox{$M_\mathrm{UV}$} \lesssim -20$, our results are similar to the measurements at $z\sim 2$ from Shivaei_2018 and the canonical value for stellar populations, $\hbox{$\xi_\mathrm{ion}$} = 25.2$, argued by Robertson_2022.