What stirs a galaxy cluster: black-hole bubbles or a passing neighbor?

A heating problem with two possible drivers

Galaxy clusters contain hot, X-ray-emitting gas between their galaxies. In a “cool-core” cluster, the central gas loses energy through radiation relatively quickly. Without enough replacement heat, it could cool excessively. One possible heating source is the central supermassive black hole: its jets inflate bubbles that rise through the surrounding gas, stirring motions whose energy can eventually dissipate as heat. This interaction is called active galactic nucleus, or AGN, feedback.

Cluster mergers also stir gas, but generally on much larger scales. The question is how to separate these overlapping influences—and whether the balance changes as a cluster settles after an encounter.

Congyao Zhang and colleagues investigate this in A1795 and A2029, using Perseus as a reference. Their paper is an arXiv preprint, version 1 posted November 9, 2026; the supplied manuscript does not establish peer-reviewed acceptance.

Reading motion in X-ray light

XRISM measures shifts and broadening of X-ray spectral lines. Shifts reveal average motion toward or away from us. After spectral modeling, broadening constrains the spread of gas velocities along that direction, called velocity dispersion.

Importantly, a broad line does not uniquely identify turbulence from a black hole. Merger-driven sloshing and unresolved differences in coherent motion can also contribute. Moreover, different observations sample different depths of X-ray-bright gas, so similar dispersions need not represent motions on the same spatial scales.

The researchers model two drivers: large-scale merger motions and smaller-scale feedback motions concentrated near the center. A turbulent “cascade” describes energy passing from larger motions to smaller ones before dissipating. The proposed “double-cascade” picture therefore involves two sources feeding motions at distinct scales and radii—not simply two measured speeds.

A1795: strong evidence for central feedback

The team reanalyzed A1795’s central XRISM observation in four quadrants. They found no strong bulk motions: the measured line-of-sight speeds relative to the brightest cluster galaxy were about 50 kilometers per second or less. Velocity dispersions were larger, with little variation between quadrants, favoring motions on smaller scales rather than a strong, coherent flow.

To estimate the merger background, the authors used decaying turbulence from a Perseus-like simulation, selecting a stage that matched A1795’s outer measurement near 300 kiloparsecs. A kiloparsec is an astronomical distance unit.

Within this decomposition, feedback dominates the velocity dispersion inside roughly 100 kiloparsecs. Its modeled one-dimensional turbulent speed is about 120 kilometers per second. Even an alternative, stronger merger background contributes no more than about 20% of the modeled kinetic energy within roughly 70 kiloparsecs.

The authors connect this extended influence to rising bubbles and their wakes, supported by the cluster’s X-ray, radio and filament structures. They identify A1795 as another clear example of the double-driver picture after Perseus.

A2029: a merger can imitate feedback

For A2029, the researchers constructed a tailored simulation of an off-center encounter with a smaller group, fixing the mass ratio at 10:1. Their preferred snapshot, about 1.6 billion years after closest passage, broadly reproduces the observed sloshing spiral and the group’s configuration.

Merger motions alone produce substantial central velocity dispersion in this model. Adding a feedback component with a characteristic speed of about 100 kilometers per second can match the innermost measurement, but this is an approximate upper limit—not a secure detection. Alternative treatments of unresolved bulk motion can reduce the inferred feedback contribution to approximately zero.

In the simulation, sloshing displaces the central low-entropy gas—relatively cool, dense material—before it returns. The authors suggest that this temporarily interrupted black-hole fueling, with feedback now restarting. That interpretation is consistent with young radio lobes and little observed cold gas, but remains a proposed history rather than a directly observed sequence.

Heating remains conditional

For A1795 and Perseus, estimated feedback-driven turbulent heating is broadly comparable to radiative cooling. That makes turbulence a plausible contributor to preventing excessive cooling, not proof of a precisely balanced thermostat.

The estimates depend strongly on turbulent speed and eddy size. Those sizes are not directly measured by XRISM; the feedback model assumes uniformly distributed turbulence, and its normalization was adjusted visually. Merger geometry, viewing direction, spatial mixing and internal gravity waves introduce further uncertainty.

The broader proposal is evolutionary: mergers initially dominate gas motions and may suppress feedback; feedback-driven turbulence later re-emerges as gas resettles. Testing that picture requires more spatially resolved velocity maps and detailed modeling of individual clusters. Similar line widths alone cannot reveal which engine is doing the stirring.