Constraints on the $z\sim6-13$ intergalactic medium from JWST spectroscopy of Lyman-alpha damping wings in galaxies

TL;DR

This study leverages JWST/NIRSpec prism spectra of 99 galaxies in the range $z\sim5.5-13$ to probe the intergalactic medium during the early stages of reionization. By forward-modeling Ly$\alpha$ damping wings with a joint treatment of diffuse IGM attenuation and local DLAs, and using sightlines through inhomogeneous reionization simulations, the authors map the distance to the first neutral patch, $D_b$, to the mean neutral fraction, $\overline{x}_{\mathrm{HI}}$, while marginalizing Ly$\alpha$ emission and local absorbers. They find a clear redshift evolution toward a more neutral IGM, with $\overline{x}_{\mathrm{HI}}$ around $0.33^{+0.18}_{-0.27}$ at $z\sim6.5$ and $0.64^{+0.17}_{-0.23}$ at $z\sim9.3$ (rising to $>0.70$ when excluding GNz11). The median HI column in local absorbers is $\log_{10}N_{\mathrm HI} \approx 20.8$ cm$^{-2}$ with little redshift evolution, and DLAs are more common in galaxies with close neighbors, suggesting absorption is enhanced in massive halos. The work demonstrates that high-S/N prism spectra can constrain IGM properties, but future deep prism and high-resolution grating data across larger samples will be essential to tighten these constraints and map the earliest stages of reionization. Overall, the paper advances a realistic, simulation-informed damping-wing analysis that accounts for IGM inhomogeneity and local absorbers, enabling robust inferences about the IGM state across $z>6$ and highlighting the importance of sample size and spectral resolution for disentangling IGM and ISM/CGM effects.

Abstract

JWST provides a unique dataset to study reionization's earliest stages, promising insights into the first galaxies. Many JWST/NIRSpec prism spectra reveal smooth Lyman-alpha breaks in z>5 galaxies, implying damping wing scattering by neutral hydrogen. We investigate what current prism spectra imply about the intergalactic medium (IGM), and how best to use NIRSpec spectra to recover IGM properties. We use a sample of 99 z~5.5-13 galaxies with high S/N prism spectra in the public archive, including 12 at z>10. We analyse these spectra using damping wing sightlines from inhomogeneous reionizing IGM simulations, mapping between the distance of a source from neutral IGM and the mean IGM neutral fraction. We marginalise over absorption by local HI around galaxies, and Lyman-alpha emission. We observe a decline in the median and variance of flux around the Lyman-alpha break with increasing redshift, consistent with an increasingly neutral IGM, as ionized regions become smaller and rarer. At $z\gtrsim9$ the spectra become consistent with an almost fully neutral IGM. We find S/N>15 per pixel is required to robustly estimate IGM properties from prism spectra. We fit a sub-sample of high S/N spectra and infer mean neutral fractions $\overline{x}_\mathrm{HI}=0.33^{+0.18}_{-0.27}, 0.64^{+0.17}_{-0.23}$ ($>0.70$ excluding GNz11) at $z \approx 6.5, 9.3$. We also investigate local HI absorption, finding median column density $\log_{10}N_\mathrm{HI}\approx10^{20.8}$ cm$^{-2}$, comparable to $z\sim3$ Lyman-break galaxies, with no significant redshift evolution. We find galaxies showing the strongest absorption are more likely to be in close associations (<500 pkpc), implying enhanced absorption in massive dark matter halos. Future deep prism and grating spectroscopy of z>9 sources will provide tighter constraints on the earliest stages of reionization, key for understanding the onset of star formation.

Figures (24)

• Figure 1: Top panels: Example 300 cMpc $\times$ 300 cMpc slices of the ionization field (white patches show ionized gas, black neutral gas) in our (1.6 cGpc)$^3$ simulations at $\overline{x}_\textsc{hi} H\textsc{i}\xspace$x_ Hi$\xspace=[0.5,0.7,0.9]$ at $z=8$, as described in Section \ref{['sec:dw_sims']}. We show a 100 sq. arcmin field as a green square (similar to e.g., the CEERS and JADES survey coverage) for comparison. Bottom panels: Ly$\alpha$ transmission profiles, due to the optical depth from the IGM, to galaxies in the corresponding simulations. We truncate the transmission blueward of Ly$\alpha$ to account for residual neutral gas inside ionized regions Mason2020. We show the median (solid line), 68%, and 95% range (shaded regions) of transmission profiles from sightlines to $\sim2000$ galaxies in the simulations with $M_\textsc{uv}$M_uv$\xspace\sim-19$. When $\overline{x}_\textsc{hi} H\textsc{i}\xspace$x_ Hi$\xspace \,\hbox{$\mathchar \sim$} \hbox{$<$}\, 0.7$ the sightline variance in the IGM is significant, meaning a large number of sightlines are required to accurately estimate $\overline{x}_\textsc{hi} H\textsc{i}\xspace$(see Section \ref{['sec:disc']}).
• Figure 2: Median, 68%, 95% range of distances of $M_\textsc{uv}$M_uv$\xspace \sim -19.3$ galaxies (the median in our sample) to the neutral IGM, $D_b$ as a function of $\overline{x}_\textsc{hi} H\textsc{i}\xspace$ from the simulations used in this work Lu2024. The median $D_b$ closely tracks the characteristic (mean) bubble radius in the simulations (blue dashed line), and shows a very broad distribution as the size distribution of ionized bubbles is broad (see Figure \ref{['fig:T_igm_sightline']}) and galaxies can sit in a range of locations inside bubbles, not always in the centre.
• Figure 3: Mock spectra at $z=10$ demonstrating the impact of the distance from neutral IGM, $D_b$. Left (right) panels show the spectrum convolved to the resolution of the NIRSpec prism (G140M grating). The grey line shows the spectrum in an almost fully ionized IGM (i.e. Ly$\alpha$ is attenuated only blueward of resonance by the Gunn1965b optical depth). Coloured lines show the spectrum if the galaxy is a distance $D_b=1-100$ cMpc from the neutral IGM. The top panels show a case with no Ly$\alpha$ emission, the bottom panels show the same spectrum including Ly$\alpha$ emission with pre-IGM $EW=100$ Å, FWHM=200 km s$^{-1}$ and $\Delta v$Δ v$\xspace$=200 km s$^{-1}$, where weak Ly$\alpha$ due to the smallest $D_b$ can only be clearly identified in the G140M spectrum.
• Figure 4: Transmission ($e^{-\tau}$) as a function of wavelength around Ly$\alpha$ due to the neutral IGM and local absorbers for high resolution ($R\,\hbox{$\mathchar \sim$} \hbox{$>$}\,1000$, left panels) and convolved with the resolution of the prism (right panels). Top panels: For a source at $z=10$. The thick grey line shows the transmission expected in the fully neutral IGM (Section \ref{['sec:dw_sims']}). Dashed coloured lines show the absorption profiles expected for local absorbers in an ionized IGM (Section \ref{['sec:dw_DLA']}). Solid coloured lines show the profiles for the combinations of both local absorbers and neutral IGM. For $N_\textsc{hi} H\textsc{i}\xspace$N_ Hi$\xspace \,\hbox{$\mathchar \sim$} \hbox{$<$}\, 10^{20.5}$ cm$^{-2}$ the neutral IGM dominates the damping wing profile. For higher column densities $N_\textsc{hi} H\textsc{i}\xspace$N_ Hi$\xspace \,\hbox{$\mathchar \sim$} \hbox{$>$}\, 10^{22}$ cm$^{-2}$, the shape becomes dominated by the local absorption, though the neutral IGM causes more absorption at redder wavelengths than a local absorber alone. Bottom panels: The neutral IGM damping wing at $z=6,10,14$ (grey solid lines) compared to only local absorption (dashed coloured lines, same as top panel). By $z\sim14$ the IGM damping wing becomes similar in strength to a $N_\textsc{hi} H\textsc{i}\xspace$N_ Hi$\xspace \,\hbox{$\mathchar \sim$} \hbox{$>$}\, 10^{21.5}$ cm$^{-2}$ local absorber.
• Figure 5: UV magnitude versus spectroscopic redshift for our sample. We show sources from CEERS, JADES and UNCOVER in blue, orange and green respectively, and highlight sources with sufficient S/N($>15$) for robust IGM fitting (see Section \ref{['sec:fitting']}) with black outlines.