First Statistical Detection of Cool Gas Outflows with JWST Towards Cosmic Dawn

TL;DR

This study tackles how cool gas outflows evolve over cosmic time by statistically detecting Mg II absorption in thousands of JWST/NIRSpec spectra across $1<z<10$ using SN-weighted stacking. The authors establish that the outflow strength, traced by $EW_{\rm out}$, scales with stellar mass $M_*$, while typical outflow velocities $v_{\rm out}$ hover around $\sim 350$ km s$^{-1}$ with no strong evolution for $z>3$, providing a robust baseline for feedback in the early universe. The combination of population-level stacking and selective analysis of high-S/N individual spectra reveals a persistent, unevolving feedback signature that challenges the notion of a dramatic shift in feedback physics at Cosmic Dawn, such as the feedback-free starburst scenario. These findings offer critical empirical constraints for modeling the baryon cycle and for calibrating galaxy formation simulations toward the Cosmic Dawn regime.

Abstract

Galactic-scale outflows are a crucial component of galaxy evolution, yet their properties in the early universe remain poorly constrained. We present the first statistical investigation of cool gas outflows in galaxies spanning a wide cosmic timeline from $z \approx 1$ to $z \approx 10$. Using thousands of public JWST/NIRSpec spectra, we employ a signal-to-noise weighted spectral stacking technique on the \ion{Mg}{2} $λ\lambda2796, 2803$ absorption doublet. We robustly detect blueshifted \ion{Mg}{2} absorption in all stellar mass and redshift bins. The outflow equivalent width exhibits a strong, positive correlation with stellar mass ($M_*$) at all epochs, increasing from $\sim 1$~Å at $M_* \approx 10^9~\mathrm{M}_\odot$ to over $3$~Å at $M_* > 10^{10.5}~\mathrm{M}_\odot$. Our work provides the first statistical constraints on cool outflows in the low-mass ($M_* \lesssim 10^{9.5}~\mathrm{M}_\odot$), high-redshift ($z > 3$) regime, vital for constraining feedback in the numerous progenitors of typical present-day galaxies. Crucially, the scaling relation between outflow properties and stellar mass shows no significant evolution at $z > 3$. This suggests a persistent, unevolving feedback mechanism governing the baryon cycle in the early universe, placing strong constraints on models that invoke a fundamental change in feedback physics at Cosmic Dawn, such as the feedback-free starburst model.

Figures (8)

• Figure 1: Sample selection and stellar mass distribution. The left panel displays the distribution of our galaxy samples in the stellar mass-redshift plane. The grey contours represent the parent sample of 24,927 galaxies, while the blue and red contours show the subsamples with valid spectral coverage of the Mg2 and Na1 D doublets, respectively. The three contour levels for each sample are derived from a Kernel Density Estimate (KDE) and correspond to the 16th, 50th, and 84th percentiles of the distribution. The middle and right panels present the Probability Density Functions (PDFs) of the stellar mass distributions for the Mg2- and Na1 D-available subsamples. The colored histograms correspond to the different redshift bins as labeled. To assess the representativeness of our sample, we overlay the evolving Galaxy Stellar Mass Function (GSMF) from Weaver2023. The brown shaded region represents the envelope of the GSMF across the redshift range $1.1 < z < 7.5$, derived from the Schechter function parameters (double for $z<3$ and single for $z>3$) presented in their work. The close agreement in shape demonstrates that our sample is broadly representative of the underlying galaxy population at these epochs.
• Figure 2: Stacked absorption feature in redshift bins. The left and right columns show the continuum-normalized stacked spectra centered on the Mg2 $\lambda\lambda2796, 2803$ and Na1 D $\lambda\lambda5890, 5896$ doublets, respectively. Each row corresponds to a different redshift bin, as indicated in the bottom-left corner of each panel. The blue solid line represents the weighted median spectrum, while the shaded region shows the 1$\sigma$ uncertainty derived from bootstrap resampling. The horizontal dashed line is the continuum level at unity, and the vertical red dashed lines mark the rest-frame wavelengths of the doublet components. The number of spectra in each stack ($N$) and the derived outflow velocity and EW are also annotated.
• Figure 3: Stacked Mg2 feature in stellar mass and redshift bins. Each colored box corresponds to a specific redshift bin, as indicated by the legend on the right. Within each box, the panels are arranged in order of increasing stellar mass from top to bottom. The stellar mass range and the number of spectra ($N$) in each stack are annotated in each panel. The blue solid line, shaded uncertainty region, and dashed lines are the same as in Figure \ref{['fig:stacked_z_bin']}.
• Figure 4: Scaling relations for the Mg2 absorption EW. Left panel: The outflow EW as a function of stellar mass. Our stacked measurements are shown as star symbols, color-coded by their weighted mean redshift. The grey dashed line shows a linear fit to the stacked data points. For comparison, small circles show the measurements from the individual galaxies presented in the Appendix \ref{['sec:inds']}, color-coded by their redshift. Right panel: The outflow EW as a function of redshift, with an additional top axis showing cosmic time. The star symbols again represent our stacked results, but all data points are now color-coded by their weighted mean stellar mass, as indicated by the colorbar. The grey dashed line is a linear fit to the stacked data.
• Figure 5: The outflow velocity versus stellar mass relation. Our measurements from the weighted stacking are shown as star symbols. The color of our data points corresponds to the mean redshift of the stack, as indicated by the color bar on the right. We compare our results with literature data from Weiner2009, Bordoloi2014, Davis2023, Yu2025, Liboni2025, and Valentino2025, as detailed in the legend. Our work provides new constraints, particularly at the low-mass, high-redshift ($z>3$) end of the parameter space.