For a few hundred million years after the Big Bang, the Universe was filled with a fog of neutral hydrogen gas that swallowed the ultraviolet light of the first galaxies. Then, roughly 13 billion years ago, the fog lifted. Starlight stripped the electrons from hydrogen atoms, a process called reionisation, and the cosmos became the transparent place telescopes look through today. Astronomers broadly agree that young galaxies did the work. The argument is over how their light got out, and a new Hubble study suggests that one popular answer, galactic winds blasted out by exploding stars, has the timing backwards.
The work, led by Cody Carr of the University of Michigan and published in Nature Astronomy on 15 October 2026, used the Hubble Space Telescope’s Cosmic Origins Spectrograph to study six compact, intensely star-forming galaxies in the relatively nearby Universe. All six come from the Low-redshift Lyman Continuum Survey, a Hubble programme built to find local galaxies that leak the kind of light that drove reionisation. They were picked to be near twins: each holds between one and six billion times the mass of the Sun, has a similar chemical make-up and forms 20 to 40 Suns’ worth of new stars a year. What separates them is the age of their starbursts. That turns the set into a natural experiment, as the University of Michigan’s announcement explains, in what happens to a galaxy’s light as its stars grow older.
Lyman continuum radiation is the light in question: ultraviolet photons with wavelengths shorter than 91.2 nanometres, energetic enough to knock the electron off a hydrogen atom. That appetite for hydrogen also makes it hard to see. Neutral gas inside a galaxy soaks it up long before it reaches open space, so the share that leaks out, known as the escape fraction, is the number that matters. Many explanations for how galaxies raise that number lean on supernovae. Thousands of massive stars exploding in quick succession drive a galactic wind, a large outflow of gas, and the thinking was that this wind would punch holes in the surrounding hydrogen and let the ultraviolet light through.
In practice, the six galaxies told a different story. The youngest starburst, with a mean stellar age of about 2.7 million years, leaked roughly 21 per cent of its Lyman continuum light, according to the freely available version of the paper. Its hydrogen was drifting outwards at only about 135 km/s. The older systems, whose light comes mostly from stars past three million years old, around the age at which the most massive stars begin to explode, had faster, stronger and dustier winds. In them, hydrogen raced out at 450 to nearly 800 km/s. The two weakest leakers let less than half a per cent of their ionising light escape.
Faster winds came with less light. The explanation the team favours is almost mechanical. Supernova energy does not simply clear a path; it sweeps up gas and dust around the starburst and drives it outwards as a thick, moving screen that sits squarely in the way. In the youngest galaxy, the light seems to escape during a brief window before any fast wind exists, when intense radiation and the winds of individual massive stars dominate. Over the first 10 million years or so, the supernova wind builds and progressively shuts that window. The light leaves first. The wind follows and closes the door behind it.
This matters because computer simulations of reionisation must decide how efficiently early galaxies leak, and many tie that leaking to supernova feedback, the energy exploding stars pump back into their surroundings. If escape is instead governed by the first few million years of a starburst, models will need to follow radiation and stellar winds in far finer detail, a point the authors make directly. Compact, bursty galaxies of this kind are also the closest local match the team could find to the galaxies seen in the reionisation era, which is why their behaviour carries weight beyond the six objects themselves.
There are real limits. Six galaxies is a small sample, and they are local stand-ins observed billions of years after reionisation ended, not the early galaxies themselves, whose escaping ultraviolet light is absorbed by intervening hydrogen long before it could reach a telescope. A single spectrum also blends a whole starburst into one average. The researchers acknowledge that small winds close to young star clusters could still help clear channels on scales Hubble cannot resolve in these galaxies, and they plan further Hubble observations to test it.
If the result holds, the galaxies that lit up the early Universe did their most important work in a cosmic instant: a few million years out of hundreds of millions. The supernovae that followed made the noise and the spectacle. By the time they arrived, the light had already gone.
