Could Reionization's Long-Lived Relics Reveal Warm Dark Matter's Mass?

During reionization, hydrogen that had been neutral was reionized by early ionizing sources, including star-forming galaxies—not everywhere at once, but in expanding bubbles. That patchwork left a long-lived imprint on the intergalactic medium, the gas between galaxies. A new preprint argues that this cosmic memory can be used to probe one of the biggest unsolved problems in physics: what dark matter actually is.

Dark matter makes up roughly one-fourth of the universe, but its particle nature remains unknown. Warm dark matter, or WDM, is a broad class of candidates whose particles would retain non-negligible velocities. Those velocities would smooth out small cosmic structures, suppressing the formation of the tiniest clumps. That makes WDM hard to distinguish from cold dark matter, or CDM, on large scales. Current constraints therefore lean on small-scale probes: the Lyman-alpha forest, which is a set of absorption features produced when quasar light passes through hydrogen clouds, and observations of Milky Way satellite galaxies and stellar streams. The strongest existing lower bound, from a combination of strong gravitational lenses and Milky Way satellites, is about 9.7 keV.

The new idea uses a different scale entirely. Reionization was patchy: some regions were reionized early, others late. Gas that reionized earlier had more time to cool, recombine, and respond to pressure. Gas that reionized later had a different temperature and density history. As a result, the local reionization redshift leaves a 'memory' in two observable quantities: how transparent the intergalactic medium is to Lyman-alpha photons, and how much neutral hydrogen remains in small halos, which 21 cm intensity mapping can trace through the radio emission of hydrogen atoms.

Because reionization bubbles span several to tens of comoving megaparsecs, that memory creates fluctuations on large scales. WDM changes it in two ways. It alters the small-scale gas evolution after reionization, and it shifts when reionization happened by suppressing the low-mass halos that host early ionizing sources. The authors argue that these effects make the large-scale Lyman-alpha forest and 21 cm signals sensitive to WDM mass, even though WDM's primary signature is on small scales.

Reionization's memory in the intergalactic gas

To model the effect, the team combined two kinds of simulation. High-resolution hydrodynamical simulations followed small-scale gas structures in a box about 1,275 kiloparsecs across, resolving the response of mini-halos, filaments, and voids to reionization. Large semi-numerical simulations, using the code 21cmFAST, modeled the patchy reionization process across a 400 megaparsec box. The researchers then used Markov Chain Monte Carlo (a statistical sampling method) forecasts to estimate how well future surveys could measure the WDM particle mass.

A 19 percent WDM signal, and an opposite sign in 21 cm

The results are forecasts, not observations, but they show a clear pattern. In the Lyman-alpha forest, reionization relics enhance large-scale power. At redshift 4, a measure of cosmic time and distance, and a wavenumber of 0.05 per megaparsec, the difference between a 3 keV WDM model and CDM is about 19 percent. For a 9 keV model, the difference falls to about 2 percent. Without the reionization memory, the large-scale differences between WDM and CDM are largely negligible.

For 21 cm intensity mapping, the relic effect works in the opposite direction: it suppresses large-scale power. At redshift 5.5, the WDM-CDM difference is again about 19 percent for 3 keV and about 5 percent for 9 keV. The suppression is weaker for lighter WDM because fewer low-mass halos are available for reionization heating to affect.

From these signals, the team forecasts that a Lyman-alpha survey with capabilities mimicking the Dark Energy Spectroscopic Instrument, or DESI, could set a lower limit of 5.0 keV at 95 percent credible interval (the range the analysis assigns 95 percent probability). Adding a 5,000-hour 21 cm intensity-mapping survey like SKA1-Low could strengthen that to 7.1 keV. Looking further ahead, a Stage V Lyman-alpha survey could reach 30 keV, a Stage II intensity-mapping survey could reach 10 keV on its own, and the combination could reach 39 keV. That would be well beyond the current best bound of 9.7 keV.

Forecasts that reach 39 keV—for now

Several caveats matter. The constraints are projections based on assumed survey covariances and noise models; they are not measurements from real data. Instrument systematics, assumptions about cleaning the 21 cm foreground wedge, and the detailed quasar luminosity function could shift the forecasts. Reionization astrophysics is another uncertainty. The ionizing efficiency of early galaxies and the turnover mass of halos unable to host star-forming galaxies can partly mimic the effect of WDM in the global reionization history. When the authors allow these parameters to vary, the forecasted limits for DESI, DESI+SKA1-Low, Stage V Lyman-alpha, and Stage V+Stage II intensity mapping relax by 12%, 12%, 40%, and 36%, respectively. Helium reionization and possible variations in photoheating are not fully included. The hybrid simulation method may also underestimate coupling between large and small scales.

Even so, the work points to a broader principle: the smallest scales of dark-matter physics can leave signals on the largest cosmic scales when reionization intervenes. The team notes that current Lyman-alpha multipole measurements at redshift 2.3 are too uncertain to detect the predicted differences, which are below 1 percent there. Future measurements at redshift 3, and cross-correlations between the Lyman-alpha forest and 21 cm maps, could help.

The preprint has not yet been peer-reviewed. But if its forecasts hold, reionization relics could become a new kind of dark-matter probe. A cosmic transition that ended billions of years ago may still be whispering clues about what dark matter is made of.