A 21-cm Forest With No Trees Still Constrains a Warmer Early Universe

The 21-cm forest is a strange kind of forest. It has no individual trees—no absorption lines that stand out clearly enough to count one by one. Yet a new analysis argues that the forest is still there in the statistical pattern of the data.

A forest with no individual trees

During the Epoch of Reionization, the intergalactic medium—the gas between galaxies—evolved from mostly neutral hydrogen into highly ionized plasma as the first stars, galaxies, and black holes formed and grew. Neutral hydrogen absorbs radio light at a wavelength of 21 centimeters. In the early universe, cosmic expansion stretches that wavelength into the meter band. If a bright radio quasar lies behind a patch of neutral intergalactic gas, its spectrum should carry a series of absorption features: the 21-cm forest.

The Lyα forest, the classic probe of this era, mainly samples gas that is already highly ionized. It saturates in neutral regions. The 21-cm forest, by contrast, is directly sensitive to neutral hydrogen. It is also measured as absorption against a compact background source, which helps it avoid the bright, spectrally smooth foregrounds that plague 21-cm emission experiments. The catch is that the signal is faint. No one has yet detected the 21-cm forest statistically.

What the uGMRT saw—and didn't see

A team led by Tomáš Šoltinský analyzed archival observations from the upgraded Giant Metrewave Radio Telescope (uGMRT) in India. The target was J352–15, a radio-loud quasar at redshift z=5.82 and the brightest known source of its kind from the era before reionization completed. Redshift is a measure of cosmic distance and time; higher values mean earlier in cosmic history.

The data came from observing runs in 2018 and 2022. After combining 17.5 hours on source and removing radio-frequency interference and poor-quality data, the team had 11.73 hours of effective integration. The resulting spectrum covered 209.0 to 222.5 MHz, corresponding to redshifts from about 5.4 to 5.82. The sensitivity reached 3.62 millijanskys per beam per 6.1 kHz channel.

Rather than hunting for individual absorption lines, the researchers measured the one-dimensional power spectrum of the quasar's spectrum. That statistic describes how much the radio flux fluctuates along the line of sight at different spectral scales. They found no statistically significant 21-cm forest signal. The measured power spectrum was consistent with the power spectrum of off-source sightlines, which contain only noise. A Kolmogorov–Smirnov test gave a p-value of 0.353, indicating no significant difference between the target and noise distributions. The result is the first observational estimate of the 21-cm forest's one-dimensional power spectrum during the Epoch of Reionization.

An independent analysis of archival Murchison Widefield Array data, from Western Australia, also measured a noise-consistent power spectrum, but its lower sensitivity and larger field of view prevented it from placing competitive constraints.

A null result that rules out the coldest models

The absence of a detection did not make the observation useless. The team compared the measured power spectrum with synthetic 21-cm forest spectra generated from cosmological simulations using the 21cmFAST framework. The simulations spanned a wide range of ionization states and X-ray pre-heating scenarios. Then they used Bayesian inference—a statistical method that weighs models against data and their uncertainties—to ask which combinations of mean neutral fraction and mean neutral gas temperature fit best.

The data did not pick out a single preferred model. Instead, they excluded part of the parameter space. At the 68% credible level, the analysis disfavours cold and substantially neutral intergalactic gas at z≈5.6. For a mean neutral fraction of 0.1, it disfavours neutral gas temperatures below about 27 K. At mean neutral fractions of 0.2, 0.5, and 0.8, it disfavours temperatures below roughly 40 K, 65 K, and 80 K, respectively. For very cold gas below about 18 K, neutral fractions above roughly 0.05 are disfavoured.

Those numbers matter because gas in an expanding universe cools unless something heats it. The lowest possible temperature from expansion alone is called the adiabatic cooling floor. The new limits suggest that the neutral intergalactic medium had already been substantially heated above that floor by z≈5.6. The constraints are derived within simulations that include X-ray pre-heating from early astrophysical sources, a mechanism that can warm gas even before reionization finishes.

Why that matters beyond one quasar

Existing probes constrain different parts of the reionization puzzle. The Lyα forest is powerful for measuring the ionization state of mostly ionized gas, but it has little direct leverage on the temperature of predominantly neutral gas. Tomographic 21-cm experiments are sensitive to that thermal state, but they must separate a faint cosmological signal from foreground emission that is orders of magnitude brighter.

The 21-cm forest offers a complementary route. It responds to both the abundance of neutral hydrogen and its temperature, so it can constrain the two together. The new limits cut into parameter space still allowed by current Lyα and 21-cm observations, meaning the forest is already adding information rather than merely confirming what was known.

What could still be missing

The result is a null detection, not a detection of the 21-cm forest. The constraints depend on the adopted simulations and modelling choices. The team did not include radiative feedback from the quasar itself, which could affect gas within tens of comoving megaparsecs. They also left out unresolved small-scale absorbers such as minihalos, and they assumed a single effective intergalactic state across a redshift interval where conditions may evolve. The masking of noisy channels could remove real absorption features, making the constraints conservative. The simulations also use an excursion-set treatment of reionization rather than full radiative-transfer hydrodynamics. The paper is an arXiv preprint, and the constraints are conditional on its modelling framework.

A statistical forest begins to bear fruit

The authors also forecast what deeper observations could do. A hypothetical 100-hour campaign, with 60% of data retained after flagging, would add 60 effective hours to the archival data, for a total of 71.73 effective hours. Extrapolating the empirical noise scaling, that would reach a sensitivity of 1.65 millijanskys per beam per channel. If the signal remains undetected, the lower limits on the neutral gas temperature would tighten to roughly 35 K, 51 K, 95 K, and 159 K at mean neutral fractions of 0.1, 0.2, 0.5, and 0.8, at the 68% credible level. If the gas is cold and substantially neutral, a statistical detection becomes more likely.

Either outcome would be scientifically valuable. A detection would open a new window on the neutral universe. A continued null result would keep shrinking the range of allowed thermal histories. The 21-cm forest still has no individual trees. But the statistical canopy is beginning to show, and it already suggests that the early universe's neutral gas was warmer than the coldest models allow.