The universe’s “missing matter” was never thought to have vanished. Cosmologists could infer how much ordinary, atom-building matter should exist, but direct inventories of stars, cold gas and the hot gas in galaxy clusters did not recover all of it in the nearby universe. The discrepancy pointed instead to matter that is real, widespread and exceptionally difficult to see.
A census built from 69 precisely located fast radio bursts now places most of that elusive material in the thinnest part of the cosmic web. In a study published on 16 June 2025 in Nature Astronomy, Liam Connor and colleagues estimate that 76 per cent of ordinary matter lies in the intergalactic medium, with an uncertainty of plus 10 and minus 11 percentage points. About 15 per cent lies in extended galactic halos, while 9 per cent is in stars and cold galactic gas.
This is one study, not settled consensus. Its fractions are a statistical inference from a modest sample, foreground estimates and a model of where electrons reside. They are not a direct count of individual particles. Even so, the work makes a more discriminating partition of the late-universe baryon budget than earlier fast-burst measurements could provide.
The missing matter was ordinary
In cosmology, baryons are the family of ordinary particles that includes protons and neutrons, the ingredients of atomic nuclei. Measurements of the early universe, including the Planck cosmic microwave background analysis, tightly constrain how many baryons the cosmos contains. Big Bang nucleosynthesis reaches a compatible answer through the primordial abundances of light elements.
Yet stars are bright rather than abundant. Add stellar mass, cold galactic gas, hot cluster gas and other detectable reservoirs, and the local tally historically came up short of that early-universe prediction. The leading expectation was that much of the difference occupied the warm-hot intergalactic medium, spread through filaments so thinly that its emission and absorption signatures were hard to assemble into a complete census. Individual detections existed, but extrapolating them across an inhomogeneous cosmic web left substantial room between locating some of the reservoir and measuring its global share.
The 76/15/9 division is therefore not a breakdown of everything in the universe. It partitions ordinary matter alone. Dark matter is a separate, non-baryonic component, and dark energy belongs to the wider mass-energy budget. Ordinary matter makes up only about one-sixth of all matter, so 76 per cent of baryons is not 76 per cent of all cosmic matter.
How a millisecond burst weighs invisible gas
A fast radio burst, or FRB, is a brief flash of radio energy from a distant galaxy. Its physical source may differ from event to event, but that uncertainty does not prevent an FRB from working as a probe. What matters for the census is the signal’s journey through intervening plasma.
Ionized gas contains free electrons. When a burst passes through it, lower radio frequencies arrive slightly later than higher ones. The size of this frequency-dependent delay yields the dispersion measure, which is proportional to the column of free electrons integrated along the full sightline. Gas that is far too faint to photograph can still leave a measurable timing signature.
Distance turns that electron column into a cosmological tool. Once astronomers identify a burst’s host galaxy and measure its redshift, they can compare how far the signal travelled with how much plasma it encountered. They must also estimate the contributions from the Milky Way, the host galaxy and intervening galactic halos before attributing the balance to intergalactic space.
Why localization mattered
A burst without a securely identified host is much less useful because its distance remains uncertain. Thirty-nine of the 69 events in the study came from the Deep Synoptic Array-110 at California’s Owens Valley Radio Observatory. The array was designed not merely to detect FRBs, but to pinpoint them accurately enough for follow-up observations of their host galaxies.
The other 30 bursts came from telescopes around the world, primarily the Australian Square Kilometre Array Pathfinder. Optical observations from facilities including the Keck and Palomar observatories helped establish host redshifts. That combination, a radio dispersion measure tied to a known cosmic distance, is the essential unit in the analysis.
According to Caltech’s account of the project, the 69 localized bursts ranged from about 11.74 million to 9.1 billion light-years away. The most distant was FRB 20230521B. The long baseline supplied both nearby and remote probes of the accumulated electrons between galaxies rather than relying on a narrow slice of cosmic history.
The 76 per cent is an inference with error bars
The team’s intergalactic category means gas outside virialized galaxy halos, combining filaments and voids. The published estimate assigns 76 per cent of baryons to this reservoir, with a plus 10 and minus 11 percentage-point uncertainty. The central value is memorable, but the interval is part of the result and should travel with it.
The same analysis describes this gas as highly ionized and extremely diffuse. Its density is below roughly 0.001 particles per cubic centimetre, and its neutral fraction is below one part in 10,000. Those limits explain the observational problem: the gas can dominate the ordinary-matter inventory while producing little light of its own.
The remaining shares are about 15 per cent in the extended halos around galaxies and 9 per cent in stars and cold galactic gas. These are fitted population fractions, not three piles weighed independently. The open preprint record describes how simulations help calibrate the statistical fingerprints of the different components, including the uneven contributions from intervening halos.
What changed after the first FRB census
The crucial groundwork was the 2020 Macquart relation, named for the late astronomer J. Xavier Prochaska Macquart. Using a smaller set of localized bursts, that study showed that dispersion increases with redshift in a way consistent with the expected cosmic density of baryons. FRBs could recover the missing total.
Recovering the total was not the same as assigning matter to environments. The average dispersion-redshift trend mainly constrains how many baryons exist along the line of sight. Separating smooth intergalactic gas from the clumpier material around galaxies requires information in the distribution and scatter of individual measurements, plus knowledge of which foreground halos a sightline crosses.
The 69-burst analysis combines those clues. Diffuse gas produces a relatively smooth accumulation with distance, while halos add irregular contributions on selected sightlines. A previous SpaceDaily report on a record-distance localized FRB explains why reaching farther into the universe extends the lever arm for this kind of measurement.
A clue about galaxy feedback
If roughly three-quarters of baryons are outside halos, galaxies have not retained all the gas that gravity drew toward them. Supernovae, stellar winds and accreting supermassive black holes can heat gas and drive it outward. In simulations, the strength of that feedback determines how much material remains near galaxies and how much is expelled into the broader web.
The study argues that relatively efficient feedback is needed to reproduce the inferred split. That gives galaxy-formation models a new observational constraint: matching the number and appearance of galaxies is not sufficient if a simulation stores too much ordinary matter inside their halos. The baryon census tests the otherwise faint aftermath of billions of years of outflows.
Redistributed gas also changes how matter clusters on intermediate scales. The authors note that stronger baryonic feedback could reduce part of the S8 tension, a mismatch between some early- and late-universe estimates of clustering amplitude. That is a possible implication rather than a demonstrated resolution, and it depends on whether the inferred gas distribution survives larger and more varied FRB samples.
A census, not a map of every particle
The method is powerful precisely because electrons that are almost invisible in emission still delay radio waves. Its limitations follow the same line of reasoning. Each burst mixes contributions from the Milky Way, intergalactic space, foreground halos and its host. Errors in any of those components can shift the inferred partition even if the measured dispersion itself is precise.
Selection also matters. Only localized bursts with usable host information enter this kind of census, and that subset may not perfectly represent the full FRB population. The conversion from observed scatter to environmental fractions depends on halo prescriptions and on simulations such as IllustrisTNG. The quoted uncertainties reflect the statistical model, but no error bar can guarantee that every modelling assumption is complete.
Sixty-nine events are enough to sharpen the fractions, not enough to erase every degeneracy. Future arrays should supply thousands of localized sightlines, allowing the budget to be tested across redshift, galaxy type and foreground environment. The clearest conclusion for now is deliberately bounded: most ordinary matter long missing from visible inventories is consistent with very diffuse, ionized gas between galaxies.
About this article
This article is for general information and reflection. It is not professional advice. For your specific situation, consult a qualified professional.