The Collective Voice of Ly$α$ Emitters: Insights from JWST Stacked Spectroscopy

Abstract

We present a spatially resolved stacked analysis of 287 LAEs at $z>4$ observed with JWST/NIRSpec prism spectroscopy. By constructing a two-dimensional stack from public surveys (CAPERS, CEERS, JADES, and RUBIES), we probe the average internal structure of typical LAEs on sub-kiloparsec scales. We find a clear radial decoupling between resonant and non-resonant emission: while EW(H$β$) and other optical lines decline with radius, EW(Ly$α$) increases toward the outskirts, and the Ly$α$ escape fraction rises from $\sim16\%$ in the center to $\gtrsim24\%$ at larger radii. This behavior indicates that resonant scattering redistributes Ly$α$ photons into lower-density outer regions, where escape becomes more efficient. Optical diagnostics and $T_e$ measurements reveal low metallicities ($12+\log(\rm O/H)\simeq7.7\pm0.2$), high ionization parameters, negligible dust attenuation, and systematically elevated N/O ratios ($\log({\rm N/O})\sim-0.4$). The latter place typical LAEs among the growing population of nitrogen-enhanced high-redshift galaxies, pointing to rapid and possibly feedback-driven chemical enrichment. The inferred ionizing photon production efficiency, $\log(ξ_{\rm ion}/{\rm Hz\,erg^{-1}})\simeq25.2$, together with the high Ly$α$ escape fractions, suggests that these systems are efficient, though not extreme, contributors to the ionizing photon budget. Comparison with SPICE radiation-hydrodynamic simulations shows that bursty supernova feedback models naturally reproduce the observed radial trends in Ly$α$ escape, UV slope, and emission-line equivalent widths, linking the spatial redistribution of Ly$α$ to stochastic star formation and feedback-driven gas flows. Our results demonstrate that Ly$α$ emission, chemical enrichment, and feedback are tightly connected in typical $z>4$ LAEs. (Shortened version for arXiv; full abstract in the paper)

Figures (13)

• Figure 1: From top to bottom: EW(Ly$\alpha$), $M_{\rm UV}$ vs redshift in our sample of 287 LAEs. Histograms reporting the distributions of rest-frame EW(Ly$\alpha$), $M_{\rm UV}$, and redshift are also shown at the side of the main panels.
• Figure 2: 2D stacked spectrum of 287 LAEs using the NIRSpec/MSA prism spectra of LAEs at $z>4$. One pixel corresponds to 0.1 arcsec on the y-axis. Some of the brightest detected emission lines are marked with a vertical dashed white line.
• Figure 3: Spectrum extracted from pixel 0 of the LAEs stack. (Top) Best-fit models to the observed data (blue solid line) are shown as solid coloured lines. Different colors mark different wavelength groups for which the fit is performed independently, as described in Sect. \ref{['sec:method']}. (Bottom) Residuals of the best-fitting models shown in the top panel. 1$\sigma$ noise level is reported as shaded blue region.
• Figure 4: [N][ii][][6584]/H vs [Ne][iii]$_{\rm blend}$/[O][ii]. Our observed ratios in pixels $0$, $\pm1$ are compared with models from the NUVOLOSO project (Pérez-Dı́az et al. in prep.) at $\log(n)[\rm cm^{-3}]=2.0$ (gray scale) and $\log(n)[\rm cm^{-3}]=3.0$ (coloured scale). The grid is color-coded by discrete values of ionization parameter, $\log(U)$, and nitrogen abundance, $\log(\rm N/O)$, as reported in the respective color bars.
• Figure 5: [5007]/H--[Ne][iii][][3869]/[3727] line ratio ('OHNO') diagram for the LAE stack compared with observations and photo-ionization models. Spatially resolved ratios for our stack are plotted as solid circles in both panels and are colour coded based on the distance (in pixel) from the center of the galaxy (i.e., pixel 0). Error bars represent standard deviation. (Left) Our LAEs stack (coloured dots) is compared with $z\sim 0$ SDSS AGNs as blue colormap with contours, SMACS 06355, 10612 and 04590 red triangles; trussler2023; the left-most square of the three is 06355, a type-II AGN identified by brinchmann2023, high-z CIII] emitters arevalo2025, a sample of galaxies at $z>2$backhaus2024, a sample of star-forming galaxies in CEERS at $2.7<z<6.5$shapley2023, and at $7<z<9$tang2023. Contours represent the percentiles (from 15% to 95% for SDSS with a step of 10%) of the number of objects in a sample. (Right) overlaid to our results (as in the left panel) are the photoionization models of cameron2023katz2023 at hydrogen densities $\log n{\rm [cm^{-3}]}=2.0$ (gray scale grid) and $\log n{\rm [cm^{-3}]}=3.0$ (coloured scale grid). The grid shows the variation of the ionization parameter and metallicity (colour scales on the right-hand side of the figure, same limits are adopted for gray scale). To guide the eye, lighter colors represent lower metallicity and higher ionization. Y-axis is the same as the left panel.