How is molecular hydrogen formed in a galaxy halo through star formation wind interactions?

A 139 kpc H2 detection with unusually little atomic hydrogen

The study detects molecular hydrogen in absorption along the sight line to the quasar J0057+1446, which pierces the circumgalactic medium of the galaxy GAL J0057+1447 at a projected distance of 139 kpc — about 11 times the galaxy's r-band stellar radius and roughly half its virial radius. The sight line runs along the galaxy's minor axis. The detected H2 covers four excitation levels (J = 0, 1, 2, and 3) with a total column density of log10[N(H2)/cm^-2] = 16.52 ± 0.05.

That H2 column density is consistent with values measured in the interstellar medium. What is not ISM-like is the accompanying atomic gas: the Lyman-alpha measurement gives log10[N(H I)/cm^-2] = 18.52 with an asymmetric uncertainty of +0.13 and -3.58. The Ly-alpha feature is saturated and shows no damping wings, which firmly caps the upper bound on N(H I) but leaves the lower bound poorly constrained. The authors adopt the statistically preferred, larger N(H I) value and carry the smaller alternative inside the reported uncertainty.

The resulting molecular fraction, f(H2) = 2N(H2) / (2N(H2) + N(H I)), is 0.02 with an upper uncertainty of +0.94 and a lower one of -0.01. Placed against literature measurements of ISM systems across redshift, this system occupies a unique region of the f(H2) versus neutral hydrogen column density plane. The authors state it is the lowest neutral hydrogen column density cloud known to bear H2. One earlier source approaches it, but that target's N(H I) was inferred from N(O I) rather than measured directly.

One cold temperature instead of a core-envelope structure

The temperature of the H2 is constrained through the excitation diagram, which uses the relative populations of the measured J levels. The fit is consistent with a single, cold temperature of T = 245 K, with uncertainties of +25 and -21 K.

This is unusual. H2 clouds are typically fit by a multi-temperature solution in which the ground states (J = 0 and 1) are colder than the more excited states (J ≥ 2), a pattern interpreted as a cold core wrapped in a warmer shell. Forcing that two-component fit onto this system yields T = 762 K for the ground-state transitions and T = 246 K for the excited states — a temperature inversion the authors note has not been seen before and would be puzzling to explain.

The statistical comparison does not favor the more complex model: the Bayesian Information Criterion is roughly equal for both, at 6.26 for the single-temperature solution and 6.94 for the two-temperature one. The authors therefore conclude the H2 is likely at a single, cold temperature, and they read the absence of a core-envelope geometry as a possible hint about where the gas came from.

Super-solar metals in a cloud moving at about 100 km/s

The H2-bearing cloud is traveling at roughly 100 km/s relative to GAL J0057+1447. The detected absorption features sit near v ≈ -100 km/s, which is why the authors compute upper limits for undetected transitions over the velocity window from -150 to -50 km/s.

Metallicity is estimated in two ways. Because O I and H I are locked together by charge exchange and oxygen is not significantly depleted onto dust, O I / H I can be read as O / H, giving a metallicity of at least 1.9 solar; the O I absorption is saturated, so this is a lower limit. An upper limit from S II gives less than 5.2 solar. The system therefore has at least solar metallicity and likely super-solar, whereas the circumgalactic medium is typically found to be sub-solar. The authors note that outflowing winds have been observed across a wide range of metallicities, from under 20% to about 320% solar.

Twelve metal species are detected, spanning neutral gas (Na I, C I, O I, N I), low-ionization gas (Mg II, Si II, C II, N II, Si III), and high-ionization gas (C IV, N V, O VI). Most of the ionized metals are consistent with measurements from other circumgalactic medium surveys and are likely associated with the ambient halo rather than with the H2 cloud itself.

Neither photoionization equilibrium nor collisions reproduce the cloud

To test whether the absorbers sit in photoionization equilibrium, the authors compare their measurements with CLOUDY models at metallicities of 1, 2, 3, and 5 solar, varying the total hydrogen volume density from log10[n(H)/cm^-3] = -6.00 to 4.00 in 0.25 dex steps and N(H I) from 15.00 to 18.75 in 0.25 dex steps, all exposed to the extragalactic ultraviolet background.

Reproducing the observed molecular fraction requires log10[N(H I)/cm^-2] = 18.50 or 18.75, log10[n(H)/cm^-3] = 2.50 to 3.00, and Z = 1 or 2 solar. Those models imply a cloud length of roughly 10^-3 to 10^-2 pc, or about 200 to 2000 AU. But no single model can match all the metal absorbers at once: the best case matches 4 of 7 metal species, and the models consistently underproduce H2 by more than two orders of magnitude. The authors conclude the system is inconsistent with a single-phase photoionization equilibrium model.

Collisional ionization is also ruled out. Comparing metal ratios against equilibrium, non-equilibrium isobaric cooling, and non-equilibrium isochoric cooling models, none reproduce the measurements at any metallicity. The ionization state of the cloud is therefore not collisionally driven.

In situ formation during wind-halo interaction beats the alternatives

The paper weighs several origins for the gas: the quasar sight line passing through the interstellar medium of a dwarf galaxy hidden along the line of sight, debris from earlier tidal interactions, molecular outflows, and in situ formation during wind-circumgalactic medium interactions. The authors conclude the last is most likely.

The supporting logic rests on the combination of properties rather than any single measurement. The cloud sits 139 kpc out, far beyond the few to tens of kpc where molecular outflows are typically seen. It has a high molecular fraction for its very low neutral hydrogen column density, which matters because self-shielding from ionizing photons is expected only when N(H I) exceeds roughly 10^17 cm^-2 — a threshold this system falls below. It has super-solar metallicity, which fits material processed by star formation rather than pristine halo gas. And it excites at a single cold temperature with no core-envelope structure, which the authors treat as a clue to its origin.

The authors frame this as possibly the first detection of complex cooling enabled by wind-circumgalactic medium interactions far into a halo, and as a path toward modeling how star-formation-driven feedback affects an entire galaxy. That framing is the authors' interpretation of the measurements, not a directly measured causal chain.

A short-lived cloud that still constrains wind microphysics

The paper states that the properties of this cloud suggest the H2 is short-lived. The reasoning follows from the low neutral hydrogen column density: with N(H I) below the self-shielding threshold, the molecular gas would likely be destroyed by the extragalactic ultraviolet background, so the detection captures a transient state rather than a long-lived reservoir.

Even so, the authors argue the system carries crucial information about the microphysics of wind-circumgalactic medium interactions — the small-scale heating and cooling cascade that occurs when hot wind fluid sweeps through cooler halo clouds. That cascade is described in the paper as not well understood, and observations of wind-interstellar medium filaments beyond a few kpc from a starburst disk are not feasible with current technology.

The measurement approach itself carries a caveat the authors state plainly. Voigt profile fitting can turn multiple narrow clouds at similar velocities into a single wider component, and high-resolution Hectochelle optical spectra do show multiple Na I components. The authors attempted to refit the ultraviolet absorbers with velocity centroids fixed to the Na I components, but the signal-to-noise and resolution of the COS spectra prevented convergence. The reported measurements are therefore assumed to be dominated by the largest absorbing cloud, and higher-resolution observations could reveal a more complicated picture.

One sight line, one galaxy, and a lower limit on N(H I)

The formation conclusion applies to this specific system: the circumgalactic medium of GAL J0057+1447, probed along a single quasar sight line at a projected distance of 139 kpc along the galaxy's minor axis. GAL J0057+1447 is a star-forming galaxy with a star formation rate of 29 solar masses per year, a stellar mass of log10(M*/Msun) = 11.1, a specific star formation rate of 10^-9.6 per year, and a far-infrared luminosity about 100 times its near-ultraviolet luminosity, which makes it a luminous infrared galaxy.

The H I measurement is the weakest link in the quantitative picture. Because the Ly-alpha feature is saturated without damping wings, two Voigt profiles fit it: one with log10[N(H I)/cm^-2] = 18.52 ± 0.13 and b = 40.6 ± 1.1 km/s, and one with log10[N(H I)/cm^-2] = 15.44 ± 0.50 and b = 82.8 ± 1.2 km/s. The larger value is statistically preferred. Adopting the smaller one would push f(H2) to 96%, place the system as the highest molecular fraction by more than 20% while sitting nearly five orders of magnitude below the next system in neutral hydrogen, and imply a metallicity above 2300 solar from O I / H I — which the authors call not physical. They therefore treat these models as limits rather than a robust measurement of N(H I).

The practical consequence for readers is that the headline molecular fraction of 2% is tied to the adopted, statistically preferred N(H I), and the reported uncertainty already spans the alternative. The distinctive claim — H2 forming in situ where star-formation-driven winds meet the halo — rests on the combination of distance, molecular fraction, single cold temperature, super-solar metallicity, and the failure of photoionization and collisional models, not on any one of these alone.

Sources used in this answer