Supernovae Driven Winds Impede Lyman Continuum Escape from Dwarf Galaxies in First 10 Myr

TL;DR

This paper investigates how Lyman continuum escape, $f_{\rm esc}^{\rm LyC}$, evolves in dwarf-like, compact star-forming galaxies by directly observing a time sequence of galactic winds in six $z\sim0.3$ local analogs. Using high-resolution HST/COS spectra and SALT radiative-transfer modeling, the authors measure wind properties across multiple ionization states (H I, N III, O VI) and connect them to $f_{\rm esc}^{\rm LyC}$. They find that LyC escape peaks early during radiation- and wind-dominated phases and then declines as SN-driven winds mass-load the ISM, increasing neutral/dust columns over about $10~\mathrm{Myr}$. This suggests early feedback is a critical regulator of reionization, a result that challenges simulations predicting LyC escape to peak after the onset of SNe, and calls for higher-resolution, radiation-hydrodynamic modeling and observations with upcoming facilities to refine our understanding of the reionization-era galaxies.

Abstract

Observations suggest that UV-bright, compact star-forming galaxies produce enough ionizing (Lyman continuum; LyC) photons to reionize the Universe. Yet, the efficiency of LyC escape and the roles of radiation, stellar winds, and supernovae remain uncertain. Using medium-resolution spectra of six nearly identical local star-forming galaxies, we directly trace, for the first time, the evolution of a multiphase wind through individual spectral lines alongside measurements of the LyC escape fraction. We find that LyC escape peaks early, during a period dominated by intense radiation and stellar winds but lacking a fast galactic wind. As the starbursts age, supernovae drive and accelerate the wind, progressively suppressing LyC escape. These results highlight the need for cosmological simulations to incorporate early feedback as a key driver of reionization.

Figures (7)

• Figure 1: Time evolution of stellar winds alongside the Lyman continuum escape fraction ($\boldsymbol{f_{\mathrm{esc}}^{\mathrm{LyC}}}$) in compact, highly star-forming galaxies observed in O VI 1032 Å, 1038 Å absorption lines. Galaxies are arranged in order of decreasing $f_{\rm esc}^{\rm LyC}$ from top to bottom, left to right and are colored according to their individual $f_{\rm esc}^{\rm LyC}$ values, as indicated by the color bar. Silver regions denote contamination by Milky Way absorption. The best fitting SED, revealing stellar components, is shown in magenta. Galaxies with higher $f_{\rm esc}^{\rm LyC}$ (e.g., J103344 + 635317, orange) exhibit deep, large-scale ($\sim$ 20 Å) absorption features, likely corresponding to the fast stellar winds of massive O-type stars. In contrast, galaxies with lower $f_{\rm esc}^{\rm LyC}$ (e.g., J154050 + 572442, purple), show shallow or absent, large-scale absorption features and enhanced small-scale ($\sim$ 3 Å) galactic wind absorption features, that fail to be captured by the stellar SED model. This trend likely reflects the deaths of the most massive O stars and the emergence of SN-driven winds as the starbursts age.
• Figure 2: Time evolution of a multiphase galactic wind and the LyC escape fraction ($\boldsymbol{f_{\rm esc}^{\rm LyC}}$) observed in the spectra of compact, highly star-forming galaxies. Each column corresponds to a different ionic line, H I (left), N III (middle), and O VI (right). The top row shows ionic abundances calculated at a metallicity of $0.4 Z_{\odot}$, redshift $z = 0.3$, hydrogen density $n_{\rm HI} = 1\ \mathrm{cm^{-3}}$, and ionization equilibrium in the presence of intergalactic and extragalactic ionizing radiation, as well as cosmic rays Ploeckinger2020. Together, the three lines trace the wind ionization structure from $\sim 10^2-10^6$ K. Galaxies are ordered from top to bottom with decreasing $f_{\rm esc}^{\rm LyC}$ and colored according to their $f_{\rm esc}^{\rm LyC}$ values, as indicated by the color bar on the right. Silver regions denote contamination from Milky Way absorption. All spectra are normalized by the best-fitting SED model. The left, middle, and right bottom panels show the H I 1026 Å, N III 990 Å, and O VI 1032 Å lines, respectively. $v_{90}$ values are shown in magenta with errors in cyan. In general, both wind speed (width) and mass (depth) increase as $f_{\rm esc}^{\rm LyC}$ declines in all lines. Except for J115855 + 312559 (N III 990Å, gold) and J154050 + 572442 (O VI 1032Å, purple), the ionized outflows closely follow their cooler counterparts, suggesting a multiphase outflow. These results imply that $f_{\rm esc}^{\rm LyC}$ peaks before the onset of SNe and declines following the subsequent mass loading of a SN-driven wind.
• Figure 3: An artist’s rendering of the LyC escape sequence alongside the evolution of a multiphase wind in compact, star-forming galaxies, as inferred from spectral lines tracing $\mathbf{10^2\text{–}10^6}$ K gas. Example observations are shown, with light blue regions marking Milky Way contamination (see Figures \ref{['fig:stellar_winds']} and \ref{['fig:cool_warm_phases']} for details). Red clouds represent the multiphase wind, UV-dominated regions in blue; lighter shades indicate lower densities. Radiation and fast stellar winds drive a slow wind early, without depositing significant dust or neutral gas. Gas clumping and intense ionizing feedback promote high $f_{\rm esc}^{\rm LyC}$ during this period. After the onset of SNe, mass and dust are expelled, lowering $f_{\rm esc}^{\rm LyC}$ on average despite the creation of low-density channels. At later times, SNe drive large-scale winds that block LyC escape, possibly enhanced by condensation from the warm gas. See also Jaskot2019Hayes2023_LyABait2024Carr2025_LyCFlury2022_diagnosticsFlury2025_ISM.
• Figure S1: LzLCS galaxies shown according to star formation surface density ($\Sigma_{\rm SFR}$) and the UV half-light radius ($r_{1/2}$) determined from NUV acquisition imaging in log space. Strong leakers ($>5\sigma$ LyC detection, $f_{\rm esc}^{\rm LyC}>5\%$), weak leakers ($>2\sigma$ LyC detection, not strong), and non-detections ($<2\sigma$ LyC detection) appear as red stars, black filled circles, and black open circles, respectively. The galaxies from our sample (HST-GO-17443) appear with a blue shadow. High $\Sigma_{\rm SFR}$ and compact morphology are associated with strong galactic winds, long proposed to facilitate LyC escape. However, $f_{\rm esc}^{\rm LyC}$ varies substantially within this region. Our galaxies were selected from within this parameter space, while controlling for the physical properties that govern the forces driving galactic winds (see Table \ref{['tab:galaxy_props']}), to investigate the observed variation in $f_{\rm esc}^{\rm LyC}$. As discussed in the main body of the paper, we attribute this variation to the time evolution of the starburst, with $f_{\rm esc}^{\rm LyC}$ peaking prior to the onset of the outflow. This sequence represents a significant departure from previous theoretical investigations of compact, star-forming galaxies, which found that $f_{\rm esc}^{\rm LyC}$ peaks at later stages—after winds have cleared the galaxy of neutral gas and dust Kimm2014. This findings will need to be addressed in the next generation of cosmological simulations of galaxy formation. Data are taken from Flury2022_dataFlury2022_data_erratum. See also Carr2025_LyC.
• Figure S2: Kendall Rank Correlations Upper Left: $v_{90}$ [km/s] versus $f_{\rm esc}^{\rm LyC}$ [%]; Upper Right: EW versus $f_{\rm esc}^{\rm LyC}$ [%]; Bottom Left: light fraction $f_*(4\leq t\leq 10~{\rm Myr})$ versus $f_{\rm esc}^{\rm LyC}$ [%]; Bottom Right: light fraction $f_*(4\leq t\leq 10~{\rm Myr})$ versus $v_{90}$ [km/s]. Upper limits are shown as gray arrowheads. Both $v_{90}$ and EW show strong and significant correlations with $f_{\rm esc}^{\rm LyC}$ ($\tau = -0.87$, $p = 0.02$). Monte Carlo sampling indicates that these correlations are moderately tight, with a one sigma dispersion of ($-0.87$--$-0.53$) and ($-0.93$--$-0.60$)in $\tau$. The light fraction $f_*(4\leq t\leq 10~{\rm Myr})$ shows a perfect correlation with $f_{\rm esc}^{\rm LyC}$ ($\tau = 1.0$, $p = 0.00$). However, the large age uncertainties allow for substantial dispersion, with a one sigma range of ($-0.73$--$-0.27$). A similar result is found for $v_{90}$ versus $f_*(4\leq t\leq 10~{\rm Myr})$ ($\tau = 0.87$, $p = 0.02$), with a one sigma dispersion of ($-0.73$--$-0.20$). For this reason, we conservatively interpret the correlations in the bottom row of panels as moderate.