Read Recompute the localized-burst baryon census first if you want to see the 2020 likelihood rebuilt step by step.
Which versions this guide reads
All four were read in full text. The journal references for the first two come from the reference list of the Connor preprint.
Macquart et al. 2020: what Figs. 2 and 3 do and do not show
Fig. 2 plots an estimated cosmic dispersion against redshift for every localised burst, over the mean relation for Planck 2015 and a shaded band that holds 90 percent of sightlines in a feedback model with F=0.2. Each point already has a host term of 50/(1+z)pccm−3 subtracted. The agreement in Fig. 2 therefore depends on that assumed host value and is a visual check, not the measurement. Macquart et al., 2020, Fig. 2
The measurement is Fig. 3. It fits four parameters, Ωbh70, F, μ and σhost, to the five gold-standard bursts and marginalises over the last three. The main text gives a fractional precision of 0.31 at 68 percent confidence, a ratio to the early-universe value of 1.1−0.6+0.5, and rules out Ωbh70<0.02 at 98.6 percent confidence. Adding two more bursts moves the best value from 0.051−0.015+0.014 to 0.042−0.012+0.011, shown in Extended Data Fig. 4. The Markov-chain version is Extended Data Fig. 5 and Extended Data Table 2.
The paper says the bursts are consistent with all the missing baryons sitting in the ionised intergalactic medium. Read that as a statement about the total diffuse electron column along the sightlines. A dispersion measure does not say where along the path the electrons are. Separating gas between halos from gas inside them needs the shape of the distribution, which is what the newer preprint uses.
Where the gas sits: halos and the space between
Prochaska and Zheng, 2019, Sections 3 and 7 supply the model behind most of this discussion. Halos above a minimum mass keep 75 percent of their cosmic share of baryons inside r200, in a modified NFW profile. Section 7.1 adds such halos along random sightlines and shows how the spread of the cosmic term depends on the minimum mass. These are model predictions: they say what the scatter would be if halos kept that much gas.
The Connor preprint turns this around and fits it. Its new step is to give the intergalactic medium and the halos separate parameters, fIGM and fX, referenced to redshift 0.1, with a likelihood built from a mock survey in the IllustrisTNG simulation. The fit gives fIGM=0.76−0.11+0.10 and fX=0.15−0.10+0.11, with a median rest-frame host of 130−23+25pccm−3. Connor et al., 2025, Fig. 2
The feature that drives the large intergalactic share is a floor: no burst beyond redshift 0.1 has an extragalactic dispersion per unit redshift below 800pccm−3. Fig. 3 of the preprint shows why that matters. Two universes with the same mean cosmic dispersion, one with fIGM=0.80 and fX=0.12 and one with 0.60 and 0.32, differ mainly below the median. Gas-rich halos would let more sightlines come in low.
Three things to keep apart when reading this section:
- The shares of baryons in clusters, 3.75±0.5 percent, and in groups of 1012.7 to 1014M⊙, 5.4±1.0 percent, come from X-ray and Sunyaev–Zeldovich measurements that the authors combine with the bursts. They are not measured by the bursts.
- The statement that the gas around individual galaxies holds only 0.35−0.25+0.30 of its cosmic share, for halos of 109 to 5×1012M⊙, is an inference from the budget once the other pieces are fixed.
- The fit uses one simulation, TNG300. The authors widen their Markov-chain errors by 30 percent to allow for model mismatch.
Joint cosmology
If the total baryon content is left free, the preprint finds Ωbh70=0.051±0.006 (Fig. 4), consistent with the early-universe value at the 10 percent level. An independent fit to the mean cosmic dispersion alone gives fd=0.94±0.05 (Extended Data Fig. 4). The paper also quotes H0=71±7kms−1Mpc−1, with the caveat that this requires fixing the baryon density at its early-universe value, so it is not an independent Hubble constant. Connor et al., 2025, Fig. 4 and main text
The same paper writes that its result may ease the lower clustering amplitude seen by weak-lensing surveys, if feedback pushes gas out of halos. That sentence is a suggestion, not a measurement. In the same way, the upper limit on the stellar baryon fraction, at most 9 percent, follows from fd and an independent count of cold gas.
Sort the claims
Open questions
How the gas fraction of a halo depends on its mass between 1011 and 1013M⊙ is still model-dependent. The host contribution is fitted as a log-normal because that shape is flexible, not because it is derived; bursts like FRB 20190520B show that a long tail exists. The Milky Way halo term is assumed at 50, within an expected range of 50 to 100. Surveys that miss high-dispersion bursts bias the inferred baryon density low, as the injection test in the method guide shows. And a census built on one simulation needs to be repeated with other feedback models before the split between halos and intergalactic gas is settled. Petroff et al., 2022, Section 4.1
Lab task
In the Lab, choose Feedback empties the halos and compare the band and the histogram with the default. The mean relation stays put, while the low side of the distribution tightens. That is the signal the 2025 preprint reads from its 69 bursts.