Black hole winds shake a galaxy cluster

Black holes are famous for swallowing everything nearby, so how does one stir hot gas 300,000 light-years away?

Welcome to FreeAstroScience. We opened this week’s big black hole paper expecting another violent jet story, and we found something stranger: a quiet, steady wind from quasar H1821+643 that out-muscles the jets, plus arithmetic we could check line by line.

Winds from the supermassive black hole powering quasar H1821+643 stir hot gas across roughly 300,000 light-years of the surrounding galaxy cluster, according to X-ray measurements by the XRISM satellite published on 28 July 2026. Measured turbulence carries about 100 times more energy than earlier estimates of black hole winds allowed, equal to several billion supernova explosions.

Between 4 and 10 September 2024, a Japanese X-ray satellite named XRISM held its gaze on one patch of the constellation Draco, stacking up 283.7 kiloseconds of clean exposure — a bit over three days of staring at a smudge of sky whose light left home 3.4 billion years ago.

Satoshi Yamada of Tohoku University, with colleagues at Kanazawa University and Tokyo Metropolitan University, was not there for the quasar itself. Instead, the prize was the gas around it: the 70-million-kelvin atmosphere of the galaxy cluster H1821+643, where every drifting iron atom broadcasts its motion in the width of an X-ray line. If black hole winds stir that atmosphere, the lines fatten. Their paper landed in Nature Astronomy on 28 July 2026, eight days before we wrote this.

What did XRISM actually see in H1821+643?

XRISM measured the width of a single emission line from Fe XXV (iron stripped of 24 of its 26 electrons) and found it broader than any nearby cluster core has ever shown. Its width corresponds to a velocity dispersion of 283 km/s (+26/−29), against 164 km/s in the Perseus cluster and no more than about 160 in every other nearby core that XRISM and its predecessor Hitomi have checked. In everyday units, the cluster’s gas is churning at roughly a million kilometers per hour.

H1821+643 sits 3.4 billion light-years away in Draco, and it is the nearest galaxy cluster anywhere to host a full-grown quasar at its center. A quasar is what you get when a supermassive black hole feeds hard enough to outshine its whole galaxy; we walked through what quasars are and why they blaze in June. Feeding it is a black hole of about 2.6 billion solar masses, and the system shines at up to 2 × 10⁴⁷ erg per second, brighter by 1 to 5 orders of magnitude than the other quasars in the study’s comparison set.

Behind the number sits Resolve, a spectrometer that separates X-ray energies to within 4.5 electron volts (a microcalorimeter senses each photon as a tiny pulse of heat, so the detector idles a twentieth of a degree above absolute zero). Precision like that turns a line profile into a speedometer.

One detail deep in the fitting shows somebody sweated this measurement. The strongest line of the group, the so-called w line, came in about 30% fainter than the model predicted, a possible sign of resonance scattering, so the team added an absorbing component, watched the dispersion slide from 295 to 283 km/s, and adopted the lower value as their conservative floor. Candor goes further: the scattering, the authors concede, is not strictly demanded by the statistics (the null-hypothesis probability is 22%), so both numbers appear in print. Honest brackets are worth more than tidy ones.

The lines were born far from the black hole

Nearly all of that broadened iron light comes from 20 to 100 kiloparsecs out — 65,000 to 326,000 light-years from the black hole, which is where the press release’s rounded 300,000 comes from. Convert it and the claim sharpens: by the scale bar in Tohoku’s own schematic, where a galaxy spans some 30,000 light-years, this wind’s reach is ten galaxy-widths.

Location was the hard part, because Resolve’s optics smear light, and a quasar this bright can bleed photons into neighboring pixels and counterfeit a cluster signal. Yamada’s team untangled the two with a joint fit that tracks where each photon landed and what energy it carried. Three separate checks then pin the address down.

• Rebuilding the line’s cumulative profile from Chandra’s temperature and density maps places over 90% of the Fe XXV flux between 20 and 100 kiloparsecs.
• The center is clean: Chandra’s sharpest look at the innermost 4 arcseconds finds only neutral iron, none of the hot ionized lines.
• Outer pixels barely register.

That last check matters: the outer field contributes less than 10% of the line emission, too little for photon leakage to manufacture the breadth.

How much energy is sloshing out there?

About 4 × 10⁶⁰ erg in random motion alone, a figure worth checking by hand, so we did.

Take the paper’s own ingredients: a gas mass of about 1.6 trillion solar masses in the 20-to-100-kiloparsec shell, and the dispersion of 283 km/s. Kinetic energy in random motions follows E = (3/2) × M × σ², where E is the energy, M is the gas mass and σ is the one-dimensional velocity spread. Multiply everything out and just under 4 × 10⁶⁰ erg falls out. It holds.

Tohoku’s press office translates that reservoir into the energy of several billion supernova explosions, and pegs it at roughly 100 times more than anyone had previously credited to black hole winds. Only the second claim has real teeth. Cavities carved into the cluster gas by this system’s own radio jets account for about 3 × 10⁵⁹ erg, an order of magnitude less than the wind-stirred motion, and jets were supposed to be the main event.

Now scale up. Add the thermal energy inside the presumed shock, about 4 × 10⁶¹ erg, compare it with the 4 × 10⁶² erg the black hole could plausibly have released over its feeding life, and the arithmetic says this quasar has coupled somewhere between 1% and 10% of its output into gas beyond 20 kiloparsecs. Earlier multiwavelength estimates put that coupling at 0.01% or less. Non-thermal motions hold 8.4% (+1.6/−1.8) of the core’s energy budget here, where clusters around weaker central engines manage 1 to 5%, and Yamada’s team permits itself one flourish about it: the gas motions in this system, they write, are “exceptionally vigorous”.

Table 1 — Core gas motion and non-thermal energy share across cluster centers (Yamada et al., Nature Astronomy, 28 July 2026)

Black hole winds that stir rather than blast

The word “blast” promises an explosion, and the spectra deliver nothing of the kind. Sound moves at about 1,300 km/s through gas this hot, near 6 keV. Fold the measured 283 km/s line-of-sight spread into a three-dimensional speed (multiply by the square root of 3) and the motions come out near 490 km/s, a Mach number of roughly 0.4, exactly the subsonic figure the paper quotes. A stir, not a bang.

Nothing about the engine is maxed out either. Its Eddington ratio, the fraction of the theoretical ceiling on how fast a black hole can feed, sits between 0.3 and 0.6, and the paper files the whole configuration under the “gentle” feedback mode expected in cool-core clusters. Deep Chandra images show no crisp shock fronts, which fits: the paper’s mock observations suggest any front beyond 100 kiloparsecs would have weakened to around Mach 1.1 and slipped below detectability.

So we will push back on the framing. Tohoku’s press release leads with “blast” and “explosive winds”, and we understand the pull of those words, but what the data supports is stranger and, we think, better. Persistence beats violence here: a subsonic wind that simply never lets up moves more energy than the jets ever carved.

What does this mean for galaxies and clusters?

It hands simulators the number they had been penciling in, and it removes an excuse. COLIBRE, one of the newest simulations, adopts thermal coupling efficiencies of 5 to 10% for quasar feedback; IllustrisTNG lets quasar-mode feedback dominate whenever black holes feed quickly. Both choices used to float free of any direct check beyond galactic scales. An observed 1 to 10% finally sits in the same range.

Yet the same models expect massive, settled clusters in the present-day universe to be ruled by jets rather than winds. H1821+643 answers with feeble jets and the most agitated core gas on record, around a quasar that is not even radio-loud. Either this system is a fluke of timing, or quasar-mode feedback keeps working in places the models have retired it, and we lean toward the second reading. So do the authors, whose closing frames H1821+643 as a nearby analog of the high-redshift quasars thought to regulate galaxy growth, preheat protoclusters, and enrich the space between galaxies.

There is something humbling in that: the brightest engine we can find may be doing its most lasting work below the speed of sound.

Our own coverage bends back on itself at this point. In June we covered XRISM catching black hole winds silencing star formation in NGC 4151, a story about what a wind does inside its own galaxy, and the open worry was whether that influence stops at the galaxy’s edge. H1821+643 pushes the same mechanism out of the galaxy altogether: around this quasar, the wind heats the atmosphere of an entire neighborhood of galaxies. If that reading survives, stirring like this was routine in the young, crowded universe, the era of compact proto-groups like the one we toured in our look at SCGG-z5 and lambda-CDM.

One omission is deliberate. H1821+643 also hosts a giant radio halo more than a million light-years across, and the paper argues even minor mergers might build such structures, since this cool core has plainly never been smashed. Whether relaxed clusters can grow their own haloes is a separate argument, and it deserves its own article.

What the spectra still cannot untangle

Two honest limitations survive this paper, and the first is the shape of the motion. A velocity dispersion records spread along the line of sight, and the same 300 km/s could come from tangled turbulence or from smooth shear, layers of gas sliding past one another in order. Yamada’s team says plainly that the two cannot be separated yet, and Chandra images do show a cold front and some asymmetry in this core, classic fingerprints of gas sloshing after a gravitational nudge. The counterargument is quantitative: reproducing the width with bulk sloshing needs velocity components spanning roughly 600 km/s, far beyond the 180 or so seen in the visibly sloshing Centaurus cluster, while this cluster’s overall drift is consistent with zero (51 ± 46 km/s). We find the wind reading stronger, and we say so knowing the shear door stays open.

Neither source says how long the stirring has been running. A single September 2024 stare gives the speed of the gas today, not the age of the wind, and the paper claims no duration. That gap is real, and naming it beats papering over it.

Then comes the anchor of the energy budget. Everything in the 1-to-10% coupling assumes the quasar-driven shock now sits near 100 kiloparsecs, and the authors concede the estimate “strongly depends on the assumed shock radius”. Move the shock and the efficiency moves with it. On the measured motions alone, the floor still stands at about 1% coupling, a hundredfold above the old ceiling. None of it has to be taken on faith, because the raw spectra are public through NASA’s HEASARC archive.

The case for a patient wind

Strip the week’s coverage down to what the instruments delivered and you keep two measurements and a ratio. Iron lines emerge broadened to 283 km/s far outside the host galaxy. Stirred motion out there holds close to 4 × 10⁶⁰ erg, ten times what this cluster’s jets have carved into cavities. And the coupling lands at 1% or better, a hundred times the old ceiling, with the honest asterisk that shear and turbulence remain entangled in the data.

We built this walkthrough at FreeAstroScience because that hundredfold jump deserved to reach you with its working attached, in words that need no physics degree. Check our division, doubt the shock radius alongside us, and keep your reason lit after the tab closes, because a mind that stops questioning is exactly the dark a bad idea needs. Yamada’s team says a companion study of the quasar’s own structure is already in preparation. The moment it lands, come back and we will test the wind up close against the stir far out. FreeAstroScience, Rimini. Gerd Dani.

Sources

1. Yamada, S., Ueda, S., Noda, H., Fujita, Y., Mizumoto, M., Nagamine, K., Ricci, C., Ogawa, S., Kawamuro, T., Yamada, S., Terashima, Y. & Ueda, Y. (2026). Vigorous turbulence driven by quasar-mode feedback in a cluster core. Nature Astronomy, Springer Nature. Published 28 July 2026. https://doi.org/10.1038/s41550-026-02939-x
2. Tohoku University (2026). Black hole ‘blast’ reaches 300,000 light-years. Phys.org. Published 29 July 2026. https://phys.org/news/2026-07-black-hole-blast-years.html